ORIGINAL COMMUNICATIONS EIGHTH INTERNATIONAL CONGRESS OF APPLIED CHEMISTRY Washington and New York September 4 to 13, 1912 SECTION III6. EXPLOSIVES VOL. IV ORIGINAL COMMUNICATIONS EIGHTH INTERNATIONAL CONGRESS OF APPLIED CHEMISTRY Washington and New York September 4 to 13, 1912 SECTION III6. EXPLOSIVES VOL. IV The matter contained in this volume is printed in exact accordance with the manuscript submitted, as provided for in the rules governing papers and publications. La matiere de ce volume a 6t6 imprimfee strictement d'accord aveo le manuscrit fourni et lee regies gouvernant tous les documents et publications. Die in diesem Heft enthaltenen Beitrage sind genau in tJbereinstimmung mit den una unterbreiteten Manuskripten gedruckt, in Gemassheit der fur Beitrage und Verlagsartikel geltenden Bestimmungen. La materia di questo volume e stampata in accordo al manoscritto presentato ed in base alle regole que governano i documenti e le publicazioni. THE RUMFORD PRESS CONCORD- N'H-U-8. A- ORIGINAL COMMUNICATIONS TO THE EIGHTH INTERNATIONAL CONGRESS OF APPLIED CHEMISTRY APPROVED BY THE COMMITTEE ON PAPERS AND PUBLICATIONS IRVING W. FAY. CHAIRMAN T. LYNTON BRIGGS JOHN C. OLSEN P. W. FRERICHS JOSEPH W. RICHARDS A. C. LANGMUIR E. F. ROEBER A. F. SEEKER 246079 SECTION]in&. EXPLOSIVES EXECUTIVE COMMITTEE President: CHARLES E. MUNROE, Ph.D. V ice-President: THOMAS M. CHATARD, Ph.D., LL.D. Secretary: WALTER O. SNELLING, Ph.D. CHARLES L. REESE, Ph.D. CHARLES F. MCKENNA, Ph.D. SECTIONAL COMMITTEE HENRY L. ABBOTT, LL.D., Brig. General, U. S. A. A. S. BARKER, Rear Admiral, U. S. Navy. CHARLES P. BEISTLE, B.S. MARCUS BENJAMIN, Ph.D., Sc.D., LL.D. A. A. BRENEMAN, M.Sc. W. H. BUELL, Ph.B. PAUL BUTLER A. M. COMEY, Ph.D. B. W. DUNN, Colonel, U. S. A., D.S. F. H. GUNSOLUS, C.E. CLARENCE HALL and the Sectional A. L. KIBLER, M.S. W. I. KOLLER M. F. LlNDSLEY FREDERICK J. M. MASURY HUDSON MAXIM G. W. PATTERSON, B.S. R. S. PENNIMAN, B.S. C. G. STORM, M.S. JOSEPH STRAUSS, Commodore, U. S. N. ERASMUS M. WEAVER, Brig. General, U. S. A. JOHN P. WISSER, Col., U. S. A. EDWARD C. WORDEN, M.A. Executive Committee VOLUME 4 SECTION III6 . EXPLOSIVES CONTENTS BBISTLE, CHARLES P. The Determination of Exudation of Nitroglycerin from Dynamite 7 BENJAMIN, MARCUS. A Convenient Method for Testing the Color of Explosives 9 BROADBENT, ALFRED L., and SPARRE, FIN. Nitration of Anisol to Tri Nitro-Anisol 15 BRUNSWIG, DR. H. VON. Neue Initialzundung fur Sprengstoffe 10 CHATARD, THOMAS M. The Misuse of Explosives 23 DAUTRICHE, H. See TAFFANEL, J. DELEPINE, M. MARCEL. Sur L' Inflammability de L'Acetylene Melange de 30% d'Air Environ 25 FLURSCHEIM, DR. BERNHARD. Tetranitroaniline, a new High Explosive 31 GOPNER, VON C. Auszug aus dem Vortrage: Die Internationale Regelung der Vorschriften uber den Post-, Eisenbahn- und Seetransport explosiver, leicht brennbarer, aetzender, etc., Produkte 35 HIBBERT, HAROLD. The Preparation, Crystalline Structure and Physical Properties of the Two Forms of Solid Nitroglycerin 37 HYDE, A. L. Boiling Points of Solutions of Nitroglycerin 59 HYDE, A. L. Separation of Nitroglycerin from Nitrosubstitution Compounds ... 69 MASLAND, WALTER E., and SPARRE, FIN. Abstract of Paper on Hydrolysis of Tri-Nitro-Anisol by Alkalies and Water 77 MOIR, JAMES. A Plea for Improvement in the Methods of Chemical Testing of Mining Explosives ROBINSON, ARTHUR LEE. Detonator Troubles Experienced in the Construction of the Isthmian Canal 85 5 6 Original Communications: Eighth International SNELLINQ, WALTER O. Improved Densimeter 105 SPARRE, FIN. See BROADBENT, ALFRED L. STORM, C. G. The effect of the Nitrotoluenes on the Determination of Nitroglycerin by Means of the Nitrometer 117 TAFFANEL, J., and DAUTRICHE, H. Recherches de la Station d'essais de Lifvin sur les explosifs de sarete pour mines grisoutcuses et poussiereuses 127 WEBER, H. C. P. On a Modified Form of Stability Test 147 WlERMAN, S. A. A New Stability Test for Nitrocellulose Powders 157 Abstract THE DETERMINATION OF EXUDATION OF NITRO- GLYCERIN FROM DYNAMITE BY CHARLES P. BEISTLE South Amboy, N. J. The compression method of determining the exudation of dyna- mite was found to give results which depended upon the nature of the absorbent materials used, and did not agree with the 40 oven tests, which simulate the most severe conditions likely to be met with in transportation under normal conditions. By subjecting a sample of dynamite to centrifugal force, as suggested by Mr. T. J. Wraempelmeier, in a centrifugal machine, the cups of which describe a circle of 7 inches radius, rotated at a speed of 600 revolutions per minute for one minute, results were obtained which checked the 40 tests. 1 This method of testing affords a rapid and effective means of determining the exudation of nitroglycerin from dynamite, and has given highly satisfactory results, so that now practically no leaky dynamite is being manu- factured in this country or Canada. Report of the Chief Inspector, Bureau for the Safe Transportation of Explosives, Feb., 1909, p. 27. 7 A CONVENIENT METHOD FOR TESTING THE COLOR OF EXPLOSIVES BY MARCUS BENJAMIN Washington, D. C. Some years ago I was consulted as to whether there was any definite scientific method by which various ordinary colors or shades could be analyzed, that is, have their composition deter- mined in exact terms from which the original color or shade could again be synthesized or duplicated without any possibility of error. After some study I came to the conclusion that the practi- cal solution of this problem could be best worked out along the lines of the valuable researches 1 made by Ogden N. Rood, long professor of physics at Columbia University and perhaps, after Chevreul, the greatest authority on color. Rood determined with great exactness the wave lengths of the spectrum colors which he adopted as standards and then found corresponding pigments easily purchasable in the open market that might be used for comparison. According to determinations made in the physical laboratory of Columbia College, therefore, the wave lengths of the five standard colors chosen expressed in microns were as follows: red, 0.644; orange, 0.614; yellow, 0.585; green, 0.521; and blue, 0.425. After accepting these standards it became necessary to make them available for practical use. That is to say, if we wanted to prove that the color cinnabar, derived from the mineral of that name, consisted of exactly 78 parts of red and 22 parts of orange, we must have some convenient colors to compare it with for the reason that a spectroscope is not always accessible and, moreover, it is an instrument that requires a certain amount of skill for ma- nipulation. Accordingly for this part of the problem advantage was taken of the investigations made by J. Clark Maxwell. This eminent scientist used for his analyses of color a series of color discs which he rotated on a wheel. These color discs consisted of ! On a Color System. Amer. Jour. Sci., Vol. 44, 1892, pp. 263-270. 9 10 Original Communications: Eighth International [VOL. circular pieces of pasteboard coated with colored paper or painted with colored pastes. By overlapping these discs within a gradu- ated circle and rapidly rotating them on a wheel so as to produce an impression of a single mass of color, they could be made to correspond to any desired color, and especially so when a small piece of material of the color to be matched was placed in front of the discs, that is, near the center of the rotating instrument, which, though usually a wheel, was sometimes a top. Therefore it is easily possible by means of the Maxwell color wheel to determine the various colors in terms of five standard colors obtained from the spectrum, together with black and white. The five color discs selected were prepared by mixing the best (1) English vermilion; (2) mineral orange; (3) light chrome yellow; (4) emerald green; and (5) artificial ultramarine blue all pigments readily purchasable in any paint store with a thick solution of gum arabic in water until it had a consistency equal to that of 011 paint and applying it to the cardboard. Light cardboard or heavy drawing paper can be used. The white was cut from the purest white cardboard obtainable and the black one was made by painting a white disc with a mixture of the best lampblack in an alcoholic solution of shellac. The disc when finished should have an even, firm, and dull surface. The best size of which to make the discs is from three to five inches in diameter. Within a comparatively short time the desirability of some standard measurement for colors became necessary in the labora- tory of the U. S. Bureau of Mines in connection with the work of analyzing explosives, particularly dynamite and explosives of the class now so largely used in mines and designated "permissible explosives." 1 Prof. Charles E. Munroe, consulting expert to the Bureau, called attention to the method which has just been described, and on his recommendation it was adopted. He has since ad- vised me that it "has been regularly, I might properly say officially, *An explosive is called a permissible explosive when it is similar in all respects to the sample that passed certain tests by the national Bureau of Mines, and whten it is used in accordance with the conditions prescribed by that bureau. Investigations of Explosives used in Coal Mines. Bulletin 15, Bureau of Mines, 1912, p. 192. iv] Congress of Applied Chemistry 11 used in recording the characteristic color of the explosives sub- mitted for testing." Also that the adoption of this method "greatly improved the precision of their descriptions." As this method has never before been used in connection with explosives, and, as I understand, no description has ever been published, I have sought from Doctor Walter O. Snelling, the chemist in charge of this work in Pittsburgh, for information con- cerning it, and from facts kindly furnished by him I have pleasure in presenting the following brief description of the method used in his laboratory. As is generally known a part of the work of the Bureau of Mines consists in the study of explosives, including the elaboration of a series of tests by which explosives that can be safely fired in a mine in the presence of firedamp could be separated from those which cannot be so fired without causing explosion. The work of testing an explosive covers a long series of physical and chemical examinations and a large number of practical tests within a steel gallery, where the normal conditions of a mine, in regard to the presence of gas, etc., may be imitated. It soon became evident that even when an explosive had passed all of the required tests, it might, at some subsequent time, become so changed in composi- tion by the manufacturer (as the market price of the different constituents varied) as to be no longer as safe as was the original sample furnished to the Bureau. Accordingly, Doctor Snelling set about finding a method by which, through the aid of chemical analysis and a careful study of physical characteristics, it might be possible to detect any changes in the composition of an ex- plosive, and he began with comparing the report of the original explosive with the results obtained from samples which are col- lected from time to time by mining engineers in mines where the explosive was actually in use. Doctor Snelling devised an exact means of determining the absolute density of an explosive, etc., and arranged the work of the chemical laboratory in such a way as to make the examina- tion of the chemical constituents present in an explosive, and their percentages, a matter of ready determination. The color was a factor that did not admit of easy classification, and yet it was recognized that a change in color of the explosive would be 12 Original Communications: Eighth International [VOL. one of the changes that would result from a difference in composi- tion and one that would be most quickly detected. Therefore it seemed eminently desirable that this point should be carefully considered. From paint manufacturers a series of charts showing certain standard colors was obtained, and for a time an attempt was made to report the color of explosives by comparing them with the nearest color sample that could be obtained. It was at this point that Professer Munroe called Doctor Snell- ing's attention to the plan of using the color wheel as suggested in the earlier part of this article. Doctor Snelling readily recog- nized the value of this method and constructed a set of color discs, six in number, as follows: black, white, red, yellow, blue, and green. His discs were 20 centimeters in diameter and were so arranged as to be mounted upon an ordinary metallic centrifuge, such as is made by Williams, Brown, and Karle of Philadelphia, Pennsylvania. The chemists in the laboratory of the Bureau of Mines soon acquired the ability of very readily and quickly determining the color of any explosive in terms of the percentage of the color discs in use, starting in general from a known definite color composition of the color which appeared nearest to the one present in the explosive. For example, if the explosive were a light grayish- yellow, the color wheel would be set to the percentage of primary colors given in the standard previously determined, and then adjustment would be made until the color corresponded in every respect to the sample of explosive under examination. Subsequently Doctor Snelling devised various minor improve- ments in the method of examination. These included a movable celluloid cover over the top of the discs to protect them from dirt, and the use of a brass plate as the bed plate on which the discs should rest with graduations on its outer edge, so that the percent- age of any color disc which is present could be more easily read. He also found it to his advantage to use larger color discs and found those 25 centimeters in diameter the most satisfactory. The brass plate on which these rest is 29 centimeters in diameter, which allows an annular space two centimeters wide for gradua- tion, and a brass frame 29 centimeters in diameter, carrying a iv] Congress of Applied Chemistry 13 plate of celluloid which serves as the cover for the apparatus and will revolve with the color discs. By such means it is found possible to determine in a most simple and exact manner the color relations of the explosives which are under examination. NITRATION OF ANISOL TO TRI-NITRO-ANISOL BY ALFRED L. BROADBENT AND FIN SPARRE Wilmington, Delaware Tri-nitro-anisol, the tri-nitro derivative of anisol, may be looked upon as the methyl ester of picric acid. Theo- retically, four isomers of this ester are possible, but only one appears to be known, viz., 2, 4, 6-tri-nitro-anisol. The chemical literature treating of the preparation of this compound from anisol is very scant. Beilstein, Organische Chemie, II-2-691, states that it is formed by treating anisic acid or anisol with sulphuric- nitric acid, quoting an article by Cahours, Annalen der Chemie 1868, 69, 238. The same statement occurs in Watts' Dictionary of Chemistry, and V. Meyer and P. Jacobson's Lehrbuch der Organischen Chemie contains nothing more on this subject. The article by Cahours merely states that tri-nitro-anisol is formed by the action of a mixture of concentrated nitric and sulphuric acids on anisol. This investigation was undertaken owing to the aforesaid absence of information in the literature and to considerable difficulty experienced in preliminary work on the same subject. The anisol was made by a method described by Kolbe (Journal fur praktische Chemie, 11-27-424) through reaction between sodium phenylate and sodium methyl sulphate: C 6 H 5 ONa+CH 3 NaS0 4 = CeHsOCHs+NasSO^ Although there seemed to be no apparent reason why the nitration of anisol to tri-nitro-anisol should be accompanied by extraordinary difficulties, preliminary experiments indicated this to be the case. Nitric acid alone as well as nitric-sulphuric acid in a variety of concentrations and proportions were tried at lower and higher temperatures with but indifferent success. When the acids were too dilute the nitration was incomplete, while with acids of higher concentration the reaction was quite violent, abundant and suffocating nitrous fumes being evolved, and a 15 16 Original Communications: Eighth International [VOL. product resulted which was never quite free from the lower nitro- anisols and which was invariably contaminated with oxidation products. It was found, however, that as soon as one N0 2 group was introduced, further nitration could be readily accomplished in the usual manner. This fact led to an investigation of the possibilities of nitrating anisol to mono-nitro-anisol and the subsequent nitration of the latter. It was soon found that the difficulties which militated against the success of the one-step nitration were met with in this case, namely, low yields contami- nated with by-products of the reaction resulting from the oxidation of the side-chain. Experiments were next carried out to determine the possibilities of nitrating anisol sulphonic acid, the latter being prepared by dissolving anisol in concentrated sulphuric acid. Although a very pure product of tri-nitro-anisol was readily obtained the method was unsatisfactory owing to the low yields, which in no case exceeded 40% of the theoretical. In view of the foregoing indifferent results it was decided to investigate more fully the one-step nitration with mixed acids. The results were most gratifying, it being found that an excellent yield of tri-nitro-anisol of a high degree of purity could be quite readily obtained. The nitration is best carried out by placing 350 gms. of a mixed acid prepared by adding 130 gms. of nitric acid of S. G. 1.52 to 220 gms. of concentrated sulphuric acid, S. G. 1.84 in a 400 cc. beaker surrounded by an ice-salt freezing mixture. The liquid within the beaker is agitated by a propeller-shaped stirrer (driven by a belt connected with a water motor) revolving at the rate of about 250 revolutions per minute, and stirring downwards. The temperature of the liquid is recorded by a thermometer placed in the beaker in such a manner as not to interfere with the stirring. When the temperature of the acid reaches 5C. the anisol (30 gms.) is slowly introduced from a separatory funnel drop by drop at such a rate that the temperature of the liquid in the beaker does not exceed 0C. The stem of the separatory funnel should be drawn out to facilitate the introduction of the anisol in very small drops. With the above quantities from two to three hours are required for introducing the anisol. It is absolutely essential that the temperatures be kept below iv] Congress of Applied Chemistry 17 0C. at the beginning of the experiment. At higher temperatures oxidation as well as nitration occurs, and the reaction is most vio ent, combustion of the anisol having resulted in some instances. After adding the anisol the temperature is slowly raised to 65 -70C. and kept at this point for twenty minutes, stirring continu- ally. The mixture is then cooled (or poured into four or five volumes of water), whereupon the lemon-yellow colored tri- nitro-anisol separates in a solid crystalline cake. The latter is separated from the acid and purified by heating first with water at 70-80C. (at which temperature it is liquid), cooled, the water withdrawn and treatment repeated with a 2% solution of sodium carbonate, followed by two more treatments with water. After the final washing the water is removed and the product dried between filter paper and in a desiccator. The yield of tri-nitro- anisol is 55-60 gms., equal to about 85% of the theoretical from the anisol. The tri-nitro-anisol thus obtained melts between 64 and 65C. It is nearly insoluble in water, but is slowly hydrolized by the latter nto methyl alcohol and picric acid. It is quite soluble in alcoho and ether, and dissolves very readily in acetone. Its specific gravity is 1.408 at 20C. The above experiments were carried out at the Experimental Stat on of the E. I. duPont deNemours Powder Company. NEUE INITIALZUNDUNG FUR SPRENGSTOFFE VON DR. H. BRUNSWIG Berlin-Steglitz, Germany Seit einigen Jahren kommen Ziindschniire in den Handel, deren Sprengstoffseele bei geeigneter Initiierung detoniert. Be- festigt man eine solche Ziindschnur auf einer Bleiplatte und detoniert man die Ziindschnur mit Hiilfe einer Sprengkapsel, so entsteht auf der Bleiplatte, dort, wo die Ziindschnur auflag, eine kanalformige Vertiefung. Man beobachtet diese Explosions- wirkung regelmassig bei Bestimmung der Detonationsgeschwindig- keit von Sprengstoffen nach der Methode von Dautriche, bemerkt aber zugleich eine andere Erscheinung, welche in ver- schiedener Richtung Interesse beansprucht. Bei Bestimmung der Detonationsgeschwindigkeit von Spreng- stoffen nach Dautriche kommt es bekanntlich darauf an, den Treffpunkt von zwei in entgegengesetzter Richtung in der Deto- nations-ziindschnur verlaufenden Explosionswellen auf der unter- gelegten Bleiplatte festzuhalten. Es ist leicht, diesen Treffpunkt zu erkennen; denn der von den zwei aufeinanderstossenden Detonationswellen auf der Bleiplatte hervorgerufene Eindruck hat einen vollig anderen Charakter als der von der Detonations- ziindschnur selbst erzeugte. Der Eindruck, welchen die aufei- nanderstossenden Detonationswellen bewirken, gleicht einem Messerschnitt, der rechtwinklig zur Ziindschnur gefiihrt wurde; durch diese dagegen wird eine Vertiefung hervorgerufen, die unterhalb der Ziindschnur mit ihr parallel verlauft und einen halb zylindrischen Kanal bildet. Die schnittartige mechanische Wirkung der aufeinanderstossenden Detonationswellen ist be- sonders gut auf der Riickseite der Bleiplatte zu erkennen. Die Querstellung dieser Druckwirkung ist geradezu charakteristisch fur aufeinanderstossende Detonationswellen; sie ist eine Folge der ungeheuren Geschwindigkeit, mit welcher zwei Gasschichten hier zusammentreffen und sich seitlich ausbreiten. Wenn die Fortpflanzungsgeschwindigkeit der Detonation in Sprengstoffen, 19 20 Original Communications: Eighth International [VOL. soweit uns bekannt, hochstens 8 km pro Sekunde betragt, so kann sie in Treffpunkt der entgegengesetzt gerichteten Detonations- wellen, relativ zu einander gemessen, das Doppelte, also 16 km pro Sekunde erreichen. Eine merkwiirdige Eigenschaft des Treffpunktes aufeinander- stossender Detonationswellen, auf welche mein Vortrag die Aufmerksamkeit lenken mochte, ist seine kraftige Initialwirkung auf andere Sprengstoffe. Was die Ursache des iiberraschend gros- sen Initiiervermogens der auf einanderstossenden Detonationswellen ist, ob ihre eigentumliche, soeben gekennzeichnete seitliche Aus- breitung, oder ihre ungewohnlich grosse Verdichtung im Treff- punkt, lasst sich zurzeit nicht entscheiden, zumal man Spreng- stoffe mit Detonationsgeschwindigkeiten von 10 bis 20 km in der Sekunde nicht kennt. Bis auf weiteres kann man annehmen, dass eine spezifische Wirkung der aufeinanderstossenden Deto- nationswellen vorliegt und dass man sich iiberhaupt hier vor einem neuen Phanomen befindet. Es ist mit der Moglichkeit zu rechnen, dass die aufeinanderstossenden Detonationswellen noch andere technisch wertvolle Fahigkeiten besitzen, die bei Sprengstoffen nicht in dem Grade beobachtet werden. Was den Gegenstand dieses Vortrages bildet, betrifft also eine neue Art der Initiierung von Sprengstoffen, die von alien bis- herigen Ztindungen einschliesslich der Sprengkapselztindung wesentlich abweicht. Die neue Ztindungsweise stiitzt sich auf das bisher tibersehene Initiiervermogen des Treffpunktes zweier auf- einanderstossenden Detonationswellen. Bisher war bekannt, dass der Grad der Initiierung eines gegebenen Initialsprengstoffes bei gegebener Anordnung desselben abhangt von dessen Menge; aber man wusste nicht, dass mit der gleichen Menge Initialspreng- stoff, je nachdem seine Detonationswellen frei auslaufen oder sich entgegenkommen, verschieden starke Initialwirkungen her- vorgerufen werden konnen. Nach der bisherigen Anschauung musste man insbesondere annehmen, dass zwei in eine Spreng- stoffmasse eingefiihrte Detonationsztindschnure mit frei aus- laufenden Detonationswellen mindestens die gleiche initiierende Wirkung ausiiben wiirden, wie eine Zlindschnur allein, in der die Detonationswellen einander entgegen laufen. In Wahrheit ist die Wirkung im letzteren Falle erhablich starker; der Unterschied iv] Congress of Applied Chemistry 21 gegeniiber den bisherigen Initialwirkungen 1st ein unter Umstanden sehr bedeutender. Als 50 g Sprenggelatine in einer grossen Blei- kugel detoniert wurden, betrug der entstandene Hohlraum bei Ziindung mit Knallquecksilbersprengkapsel Nr. 8 2700 ccm, bei Ziindung mit aufeinanderstossenden Detonationswellen 3300 ccm, also um 22 Prozent mehr. Bei Sprengversuchen mit Gra- naten, die eine Sprengladung von 250 g gegossenem Trinitrotoluol besassen, wurden durch Initiierung mit zwei einfachen Deto- nationswellen insgesamt 50 wirksame Sprengstiicke gezahlt, bei Initiierung mit zwei aufeinanderstossenden Detonationswellen aber 206 wirksame Sprengstiicke. Bei Vergleichsversuchen, angestellt mit diinnen Detonationsziindschniiren unter sonst gleichen Bedingungen, tibertrug die einfache Detonationswelle die Detonation noch nicht durch einen Luftraum von 0,5 mm; die Kumulationswelle, wie man den Treffpunkt zweier aus ent- gegengesetzter Richtung zusammenstossenden Detonationswellen nennen konnte, aber auf mehr als 2 mm Entfernung. Es scheint moglich zu sein, die Wirkung der Kumulationswelle noch weiter zu verstarken, indem man das zugrunde liegende Prinzip wiederholt zur Anwendung bringt, z. B. indem man jeden Arm einer U-formig gestalteten Patrone durch Kumulations- zundung initiiert und die in dieser Ziindpatrone entstandene sekundare Kumulationswelle ihrerseits zur Initiierung der eigent- lichen Sprengladung verwendet. Ob auch eine Schwachung der Kumulationswelle in dem Sinne stattfinden konne, dass sich entgegenkommende Detonationswellen von ungleicher Phase gegenseitig aufheben, mag dahingestellt bleiben. Die zur praktischen Anwendung dieser neuen Initialziindung erforderliche Anordnung besteht also darin, die Sprengladung in unmittelbare Beriihrung mit dem Orte des Zusammenprallens der Detonationswellen zu bringen, entweder in der Art, dass die Sprengladung jene Stelle in sich aufnimmt, oder so, dass jene Stelle aussen an der Sprengladung anliegt. Es ist nach diesem Verfahren moglich, nicht mur die Wirkung der bekannten Sprengstoffe sehr zu verstarken, sondern auch solche Sprengstoffe und Sprengstoffmischungen zur vollkommenen Detonation zu bringen, welche mit den bisherigen Hulfsmitteln unvollstandi g detonieren und deshalb fur die Technik nicht 22 Original Communications: Eighth International zweckmassig ausgenutzt werden konnten. Sobald das neue Verfahren soweit ausgebildet sein wird, dass es eine unmittelbare Verwertung erlaubt, diirfte es von Bedeutung sein sowohl fur die Artillerie, fiir Granaten, Minen, Torpedos, als auch fiir den Bergbau, insbesondere durch Anwendung von unempfindlicheren Sprengstoffmischungen und fiir die Sprengtechnik im allgemeinen. Die neue Initialziindung zur Detonierung von Sprengstoffen ist der Centralstelle fiir wissenschaftlich-technische Untersuch- ungen in Neubabelsberg patentrechtlich geschiitzt; in Deutschland z. B. durch D. R. P. Nr. 245087. Abstract THE MISUSE OF EXPLOSIVES. ITS EXTENT AND PREVENTION BY THOMAS M. CHATARD Washington, D. C. Noting the statement of a Portuguese revolutionist that ''bombs are cheaper and, in the hands of untrained people, more effective than other weapons," this paper gives the results of a tabulation of 772 instances of the misuse of high explosives from 1903 to 1911, inclusive. In the United States alone, there were 213 explosions due to labor troubles and 238 Black Hand outrages, while 139 occurrences can only be ascribed to absolute lawlessness. This lawlessness is especially dwelt upon in the discussion of the results, which is confined to conditions existing in the United States. These conditions and the legal difficulties attending any legislation against and prosecution of such offenses, in this country, are particularly considered since any proposed remedy must be in conformity with them, if any practical benefit is to be expected, and the difficulties are greater here than abroad. The importance of the co-operation of the Sections of Explosives and of Law, of this International Congress, in securing adequate and satisfactory legislation and legal practice to this end is pointed out and the character and scope of this co-operation is indicated, while the need of such action is evidenced by the extent of this abuse and by the statements of prominent radical-socialistic leaders. The fundamental principle of the suggested legislation is that "the unlicensed possession of any high explosive shall be a punish- able offense, without reference to the intentions of the trans- gressor" and instances of such laws are given. A plan of a licens- ing system is outlined with examples of its practical working. "It is prevention, not punishment, that is to be sought for" and this can best be obtained by enacting reasonable laws, satis- factory to legitimate business, and then rigidly enforcing them. 23 SUR L'INFLAMMABILITE DE L' ACETYLENE MELANGE DE 30% D'AIR ENVIRON PAR M. MARCEL DELEPINE Agrege pres VEcole superieure de Pharmacie de Paris, Paris, France On salt que 1 'acetylene pur ne brule avec tout son eclat dans 1'air que s'il s'echappe des bees avec une pression assez consid- erable; si la pression est insuffisante, le brassage de 1'air avec le gaz est imparfait, la flamme est fuligineuse et I'orifice du bee s'encrasse. Aussi a-t-on du remedier a ces inconve"nients en utilisant des bruleurs-melangeurs d'air. On arrive egalement a de bons resultats en brulant non pas le gaz pur, mais le gaz prealablement melange d'une certaine quantite d'air. L 'experience montre qu'un melange contenant aux environs de 30% d'air et 70% d 'acetylene, brule avec un eclat plus grand encore que l'ace*tylene pur, tout en n'exigeant qu'une pression de quelques centimetres d'eau. Pour un meme eclairage, la depense est un peu moindre et les canalisations du gaz de houille ordinaire suffisent. On arrive tres simplement a re*aliser des melanges d'une composition determinee en divisant le tambour d'un compteur a gaz en deux compartiments soli- daires de capacites inegales: 1'un des compartiments debite de 1'acetylene qui le fait tourner par sa pression, 1'autre debite de 1'air; les deux gaz se melangent ensuite dans une chambre unique avant de se rendre aux canalisations. C'est d'un appareil sem- blable, le me*langeur-doseur OVING que je me suis servi. Etant donnee la grande puissance explosive de l'ace*tylne et de ses melanges avec 1'air, il etait bon de proce*der a quelques experiences sur 1'inflammabilit^ des melanges fournis par les appareils en question et je crois utile de les communiquer a titre de contribution a 1'etude des melanges combustibles. Get examen nitrite d'autant plus de nous arreter qu'il y a quelque indecision sur les limites d 'inflammabilite des melanges d'oxygene et d'air. LE CHATELIER (C. R. Ac. Sc., t. 121, p. 1144; 1895) donne les limites infe"rieure et supe*rieure, 2, 8 et 65% d 'acetylene; CLOWES (Report of the Britisch Association, 1896, 25 26 Original Communications: Eighth International [VOL. p. 746) donne 3 a 82% pour 1'inflammation par flamme; GERDES (Chem. und chem-techn. Vortrdge, t. 4, p. 249; 1899) indique 2, 5 a 80% pour un melange de 90 litres soumis a 1 'influence de Tetin- celle electrique. LE CHATELIER a, en outre, fait observer que le chiffre de 65% est un maximum constate* dans une masse de grand volume; les limites se resserrent notallement si Ton opere dans des tubes e"troits. Dans les experiences qui suivent, le melange fut analyse" chaque fois. Pour cela on remplissait les appareils destines aux essais par un balayage suffisant et re*coltait a leur extremite* meme une fraction destinee a 1'analyse. Celle-ci e*tait bien simple: il suffi- sait, en effet, de mesurer 1 'absorption produite dans le melange par le chlorure cuivreux ammoniacal qui s'empare a la fois de 1 'ac&tylene et de I'oxyg&ne, en ne laissant que 1 'azote et ses com- pagnons. J'ai d'ailleurs ve*rifie que Ton arrivait aux memes re*sultats que si 1'on absorbait successivement I'oxyg&ne par le pyrogallate et 1 'acetylene par le reactif cuivreux. Voici deux de ces analyses comparatives d'un meme e*chantillon : I. Gaz du me*langeur . . . ............. 29,2 c. cub. Res apres absorption de C 2 H 2 et O 2 6,1 c. cub. = azote D'ou, a ir=^i- = 7,7 c. cub. soit ........ 26,4% 0,79 II. Gaz depouille" de O 2 . V ............ 29,2 c. cub. Reste apres absorption de C 2 H 2 ...... 6,5 c. cub. = azote. f acetylene .......... 22,7 c. cub. 8,22, soit 26,6%. D ' ou > \ air = M- I 0,79 Total 30,92 I. Etincelk d' induction. Une premiere s^rie d 'experiences a montre que des melanges contenant de 25,4 a 31% d'air ne s 'enflammaient pas lorsqu'on les soumettait a des exces de pres- sion sur la pression atmospherique allant de 3,5 a 39 centimetres de mercure (1,5 atm. de pression totale) et qu'on y faisait passer des e*tincelles d 'induction de 2 millimetres de longueur environ. II se faisait tout au plus un petit de"p6t de charbon sur les poles. iv] Congress of Applied Chemistry 27 II. Fit de fer incandescent. Une deuxime se*rie a e*te* faite en portant au bon rouge, au moyen cTun courant electrique, un fil de fer de 2 centimetres de long place* au sein du melange gazeux renferme* dans une ampoule ovoide d'un litre; un bouchon de caoutchouc moyennement serre* dans le goulot laissait passer les conducteurs et servait en meme temps a clore 1'ampoule. Aucun des melanges contenant de 25 a 31% d'air, examines sous ses pressions allant de 0,3 a 15 centimetres de mercure m'a donne* d 'inflammation. On a observe* seulement un de*pot de charbon lanugineux sur le fil devenu cassant. Toutefois, une experience faite avec un courant suffisant pour fondre le fil et sous la surpression de 3 cm. 5, a couse* une inflam- mation qui a chasse* le bouchon de 1'appareil, avec production abondante de noir de fume*e. III. Fil de platine incandescent. On a remplace* le fil de fer par du fil de platine, en conservant le mme dispositif. Dans un premier groupe d 'experiences avec un fil de 0,1 mil- limetre de diam tre et 10 millimetres de longueur, sous des excs de pression variant de 0,5 a 11,3 centimetres de mercure et une teneur en air de 28%, on n'a rien observe*. On a m6me pu dans un cas fondre le fil de platine sans provoquer d 'inflammation. Par centre, dans un second groupe d 'experiences avec un fil de 0,2 millimetre de diametre et 20 a 30 millimetres de longueur, des melanges contenant de 26,5 a 32,6% d'air ont toujours ete enflammes, m&ne sous la pression tres reduite de quelques mil- limetres de mercure en sus de la pression ambiante. Ce m&me fil place*, non plus dans 1'ampoule d'un litre, mais dans 1'axe d'un tube de plomb de 20 millimetres de diam&tre et de 1,40 metre de long (avec ou sans garniture isolante de verre) n'a laisse* apparaitre que des traces de charbon dans le voisinage du fil de platine, malgre que 1 'incandescence eut ete* maintenue plus d'une minute et eut meme e*te pousse*e jusqu'a la fusion du fil; ce qui montrait que la combustion, fort restreinte, s'e*tait limite*e. IV. Fulminate de mercure. Une quatrieme se*rie d 'expedi- ences a enfin e*te* exe*cute*e en faisant detoner un peu de fulminate de mercure au sein de 1'ampoule d'un litre. La detonation du 28 Original Communications: Eighth International [VOL. fulminate etait provoquee par le contact d'un fil chauffe juste assez au moyen d'un courant electrique. Pour un poids de fulminate de 0,005 gr. detonant dans des melanges a 29% sous des exces de pression allant de 1 a 4 centi- metres de mercure, on n'a pas eu d 'inflammation. Pour un poids double, soit 0,01 gr., on a eu les re*sultats suivants: Air % du melange Pression en Resultat centim. de Hg 25,4 1 Rien. 25,4 . 4 Rien. 28,0 1 Rien. 28,0 1 Inflammation. 28,0 4 Inflammation. 23,0 9 Rien. 23,0 15 Explosion vive. 29,0 1 Inflammation. CONCLUSIONS II ressort nettement des experiences que de courtes etincelles sont sans effet, meme pour des compressions d'une atmosphere et demie. Le fil de fer est egalement peu efncace; cependant a sa temperature de fusion, il a provoque une inflammation. Les experiences avec les fils de platine montrent que ce metal est plus actif que le fer, sans doute parce qu'on peut le chauffer plus haut sans le fondre. Toutefois, les result ats negatifs obtenus avec le fil de 0,1 millimetre, meme a la temperature de fusion, montrent que la temperature en un point n'est pas le seul facteur a considerer : 1 'etendue de la surface chaude doit entrer en ligne de compte en permettant de decomposer dans le voisinage du fil une plus grande quantite* d 'acetylene dont 1'hydrogene a un moment donne* devient assez abondant pour former avec 1'air un melange explosif, meilleur transmetteur d 'inflammation (ou d 'explosion) que le fil chaud. C'est pourquoi un fil plus gros et plus long a egalite de temperature est plus efficace. L 'experience en tube de plomb qui ne permet sans doute pas le melange rapide de 1'hydrogene avec 1'oxygene conduit naturellement a un re- sultat negatif. Cette maniere d'etre du melange, suivant qu'il est dans un recipient tubulaire ou un recipient globoide, est a rapprocher de celle qui a ete constatee par MM. BERTHELOT iv] Congress of Applied Chemistry 29 et VIEILLE (Ann. chim. et phys. [7], t. 16, p. 24; 1899) pour la decomposition de 1 'acetylene pur. Alors que le fil rougi ou Pamorce de*truisent 1 'acetylene dans des vases de grande capacite*, ils ne produisent qu'une faible decomposition locale dans des tubes e*troits, les conditions de pression etant elgales d'ailleurs. Les experiences faites avec le fulminate montrent que cet agent est tres actif. Son influence depend de la dose puisque les explo- sions de 0,005 gr. n'ont produit aucun effet, sous une surpression de 1 a 4 centimetres de mercure et une teneur de 29% d'air, alors que ce meme melange et meme celui a 28% s'enflamme sous une pression de 1 centimetre, s'il y a 0,01 gr. de fulminate. Pour une meme dose, la composition a une influence notable; avec 23% d'air, il faut depasser 9 centimetres de mercure pour avoir rinflammation, tandis qu'avec 28%, elle a lieu pour 4 centimetres et meme 1 centimetre; il en est de meme pour le melange a 29%. L 'experience a 23% d'air montre aussi que la pression est un fac- teur efficace qui peut compenser le deficit en air. Les inflammation generalise*es qui ont produit la chasse du bouchon seraient, a mon avis, dues a trois phenomenes successifs: decomposition au contact des fils ou des amorces d'un volume determine d 'acetylene dans un premier temps et remplacement du gaz detruit par son volume d'hydrogene; combinaison ex- plosive du melange d'hydrogene et d'air qui s'est ainsi substitue au melange primitif, dans un deuxieme temps; le melange nouveau aurait en effet la composition : 18% (2H 2 + O 2 ) + 24% azote + 54% de C 2 H 2 ou de H 2 , en moyenne; si ce melange est forme en quantite suffisante dans un temps tres court, il est alors capable dans un troisieme temps, en raison de la grande vitesse de son onde ex- plosive de propager 1 'inflammation dans le reste du melange non decompose ou d'y produire une decomposition notable de 1 'acety- lene. La seule veritable explosion qui ait e"te remarquee est celle qu avait provoquee la charge de 0,01 gr. de fulminate sous la sur pression de 15 centimetres de mercure. Toutes ces hypotheses se coordonnent avec la constatation d'une plus facile inflammation si 1'on augmente Pe*nergie des processus primordiaux de la decomposition, ce qui peut se re*ali- ser, soit en augmentant le volume de la zone atteninte, par 1'em- ploi d'un gros fil ou d'une plus grosse amorce, soit en augmentant la pression. TETRANITROANILINE, A NEW HIGH EXPLOSIVE BY DR. BERNHARD FLURSCHEIM Rushmoor Fleet, Hampshire, England A compound which would combine the stability of an aromatic nitro-derivative with the explosive power of an aliphatic nitro- ether has long been an object of research work on explosives. The nearest approach to this combination of properties hither- to realised seems to have been tetryl (tetranitromethylaniline) . Partly owing, however, to its expensive manufacture, and partly to its small resistance to heat and mechanical influences, due undoubtedly to the non-aromatic linking of one of the nitro- groups, the use of tetryl seems practically to have been confined to detonators. It may therefore be of interest to record some of the prop- erties of an entirely aromatic compound of quite recent discovery, "tetranitroaniline," as this has been found to combine perfect aromatic stability with quite exceptional "brisance," while also complying with other industrial postulates, such as high specific gravity and economic manufacture. Tetranitroaniline (abbreviated T. N. A.) is obtained by various methods which will be published shortly. The commercial method consists in treating meta-nitroani ine with a mixture of sulphuric and nitric acids, whereupon the tetranitroaniline sep- arates from the undiluted acid in pure crystals which are filtered off and washed with water, the waste acid serving for the manu- facture of nitric acid. The meta-nitroaniline in its turn is obtained by another new process whereby commercial dinitrobenzol is reduced by means of sodiumbisulphide and water, the meta- nitroaniline thus produced being suitable for nitration without previous purification, and the sodiumbisulphide used being re- covered in the shape of sodiumhyposdphite. In this way, com- mercial dini robenzol yields almost its own weight of pure tet- ranitroaniline. 31 32 Original Communications: Eighth International [VOL. T. N. A. has the formula ^,~ NH 2 and conta'ns 25.6% of nitrogen and 46.9% of oxygen. It is a yellow crystalline compound. Its absolute specific gravity is 1.85. It melts with decomposition, but without explosion, at between 208 and 215 C. according to the rate of heating. One part is soluble in about 6 parts of acetone. It can be re-crystal- lised from a number of solvents, notably from xylene and from liquid aromatic nitro-compounds. It is not attacked by cold water, in which it is insoluble. By heating it with water it is slowly transformed into trinitroaminophenol, a derivative of picric acid T. N. A. has neither acid, nor basic properties. It can be melted with trinitrotoluol, dinitrobenzol and other nitro- compounds, no decomposition occurring even at 150 C. It gives Abel tests of over an hour. On a lighted piece of paper it burns gradually, without explosion. When spread on wood or stone and lighted, the flame does not spread. Under the falling hammer it detonates with 5 kilograms at 35 cm. (tetryl at 22 cm.). When detonated by a primer, it gives no residue or smoke. Its explosive power, as measured in a lead block, is greater than that of any other solid compound, but is inferior to that of nitro- glycerine. The enlargement of the cavity by T. N. A. is, for instance, 40 to 50% greater than for trinitrotoluol and picric acid and 20% greater than for tetryl and gun-cotton. T. N. A. can be compressed at high pressures, either alone with nitrates in suitable proportions, without losing its capacity of being easily detonated, _as is the case with other aromatic nitro- compounds. This was, or instance, ascertained for a mixture of 35% T. N. A. with 65% Ammoniumnitrate, compressed to a specific gravity of 1.6; also for a mixture of 50% T. N. A., 40% Bariumnitrate and 10% Potassiumnitrate, compressed at 145 atmospheres to a density of 1.8 (without allowing for the central cavity of the cartridge). The latter mixture was compared with an analogous composition made from gun-cotton and ni- iv] Congress of Applied Chemistry 33 trates (with which a density of 1.275 was obtained on compression), cartridges of both being fired by means of a Nr. 6 detonator on hardened iron bars, when a considerably deeper groove was produced by the T. N. A. composition. In detonators T. N. A. produces strong effects with a small amount of fulminate as primer. With T. N. A. in propelling charges, in charges for projectiles and in detonating fuses, some interesting results are being obtained. The following calorimetric tests were carried out by Mr. W. Macnab, F. I. C.: Results of firing Tetranitroaniline and some mixtures in large calorimetric bomb from which the air had been pumped out. TETRANITROANILINE cc. perm. Composition of perm, gas cals per gr. gas per Charge Case (water liq.) gram CO2 CO CH4 H N 55 grm. glass 1017 827 21.5 38.3 1.2 15.1 23.9 100 grm. glass 1018 25.0 33.1 0.6 18.5 22.8 Mixtures 12 3 Ammonium nitrate 86 81 87 Trinitrotoluene 8 Tetranitroaniline 13 "Tetra" & Dinitrobenzene 7 Black charcoal 66 6 100 100 100 cals. per grm. (water liq.) 1131 1160 1154 cals. per grm. (gaseous) 903 933 913 cc. perm, gas per gram 411 443 416 COMPOSITION OP GASES C02 26.3 37.7 26.5 CO 3.3 3.5 4.1 H 8.5 4.1 8.5 N 61.9 54.7 60.9 100.0 100.0 100.0 No. 8 Fulminate detonators used. AUSZUG AUS DEM VORTRAGE: DIE INTERNATIONALE REGELUNG DER VORSCHRIFTEN UBER DEN POST-, EISENBAHN- UND SEETRANSPORT EXPLOSIVER, LEICHT BRENNBARER, AETZENDER, ETC., PRODUKTE VON C. GOPNER Hamburg, Germany I. Beziiglich des internationalen Postversands wird dem Wun- sche Ausdruck gegeben, den schon von Herrn Dr. C. A. von Martius auf dem 6. internationalen Kongresse zu Rom vorgeschlagenen Wortlaut: " dass eine genauere Definition oder speziellere Auf- " zeichnung der als explosiv, leicht entziindlich, von der Befor- ' ' derung auszuschliessenden Gegenstande durch Internationale " Uebereinkunft gegeben werde." zu wiederholen und diessen Wunsch in Form einer Resolution zum Ausdruck zu bringen. II. Eisenbahnversand gefahrlicher giiter. Es wird die lange Frist, welche die von den beauftragten Del- egierten der Staaten, welche das internationale Uebereinkommen iiber den Eisenbahnfrachtverkehr abgeschlossen haben, gefassten Beschliisse bis zu ihrer Ratifizierung und in Kraftsetzung ge- brauchen, bemangelt. Wenn diese bis jetzt 2-3 Jahre beansprucht haben, so wird es nach dem neuen Wortlaute des Uebereinkommens 7-8 Jahre dauern, bis Aenderungen und Erganzungen, namentlich der Vorschriften tiber die bedingungsweise zum Transport auf Eisenbahnen zugelassenen Gegenstande, Giiltigkeit erlangen. Es wird deshalb vorgeschlagen, derartige Sachen sofort nach ihrem Eingange zu beraten, dariiber Beschluss zu fassen und schnellstens zur Giiltigkeit zu bringen. Es wird als wtinschenswert bezeichnet, dass nicht nur alle europaischen Staaten, sondern auch die der tibrigen Continente dem internationalen Uebereinkommen fur den Eisenbahnfracht- verkehr beitreten. 35 36 Original Communications: Eighth International Es wird vorgeschlagen, die Vorschriften aller Staaten iiber den Frachtverkehr, resp. liber die bedingungsweise zum Eisen- bahnversand zugelassenen Gegenstande, zu sammeln, das Ver- zeichnis derselben zu erganzen, genaue Vorschriften iiber die in einem Versandsttick zusammen zu packenden Gegenstande zu schaffen, und ein Verzeichnis der Giiter, welche zum Eilgutverkehr zuzulassen sind, anzufertigen. III. Beforderung gefahrlicher Gtiter mit Kauffahrteischiffen. Am 1. Juni dieses Jahres ist von am Seeverkehr beteiligten deutschen Staaten eine Verordnung in Kraft getreten, welche diese Sache in systematischer Weise regelt und sich im Grossen und Ganzen an die fur den Eisenbahnverkehr in Deutschland geltenden Bestimmungen anschliesst. Es wird ein Antrag gestellt, die Regierung der Vereinigten Staaten zu veranlassen: 1. zu untersuchen, ob es im Interesse von Handel, Industrie und Schifffahrt liegt, einheitliche Vorschriften fiir den Trans- port von gefahrlichen Gtitern mit Kauffahrteischiffen zu bes- itzen, und 2. im Falle die Regierung der Vereinigten Staaten diese Meinung teilt, eine Konferenz der an der Frage interessierten Seestaaten zu berufen, mit dem Auftrage, solche Vorschriften zu entwerfen und zu ratifizieren. THE PREPARATION, CRYSTALLINE STRUCTURE, AND PHYSICAL PROPERTIES OF THE TWO FORMS OF SOLID NITROGLYCERINE BY HAROLD HIBBERT Wilmington, Delaware INTRODUCTION On account of the extended use to which nitroglycerine is put in the explosives industry, a knowledge of its physical properties is of great importance, and this is especially true regarding the " solid product," viz., frozen nitroglycerine explosives, due largely to the fact that nitroglycerine may often be supercooled to an extraordinary degree ( 70 C.) without showing any tendency to solidify. An exact determination of the freezing- and melting- points occasions some difficulty so that the value of these constants as determined by various investigators varies considerably. In the patent specification of Nobel (U. S. patent No. 57, 175, issued August 14, 1866) the patentee draws attention to the fact that his nitroglycerine product differs from that previously ob- tained by Sobrero in that it can be readily induced to crystallize at a comparatively low temperature, viz., around 55 F. (equiva- lent to 12.8 C.) and it is an interesting fact that though this value is not quoted in the literature on the subject it is in very close agreement with that obtained in recent investigations, the more reliable of which are those carried out by Nauckhoff and Kast. Nauckhoff (Zeit. angew. Chem. 1905, 18, pp. 11 and 53) in a very interesting paper on the physical properties of nitroglycerine found the melting-point to be 12-12.4 C. The subject was taken up later by Kast (Zeit. Schiess. u. Spreng., 1906, 1, p. 225) who introduced another complication by his claim to have discovered that solid nitroglycerine is capable of existing in two isomeric forms, a labile isomeride melting at 2.8 C. and a stable one melt- ing at 13.5 C. In a later paper on this subject by Nauckhoff 37 38 Original Communications: Eighth International [VOL. (Zeit. Schiess. u. Spreng., 1911, 6, p. 124) in which a description of the crystalline structure of the higher-melting isomeride is given, this author confirms the above value for the melting-point of the stable form obtained by Kast, viz., 13.5 C., but differs from him in finding this to be identical with the value for the freezing point, coming to the conclusion that the value (12-12.4 C.) obtained previously was due to the nitroglycerine used by him not having been sufficiently pure. He was not successful, however, in pre- paring the lower-melting labile isomeride. In connection with another investigation in which frozen nitro- glycerine was desired, the author prepared a sample by freezing a mixture of nitroglycerine and wood-pulp and inoculating from this, whereby a product was obtained which melted sharply at 1.0 C. and was apparently identical with the labile isomeride of Kast. Since Kast's work has not up to the present apparently been generally accepted, and especially in view of Nauckhoff's failure to duplicate the results, it was deemed advisable to repeat it and to ascertain whether or not there is, in reality, a second iso- meric labile form of solid nitroglycerine. This was the more necessary since apparently the labile isomeride had been obtained in a purely accidental manner and the conditions for its prepara- tion had not been defined. This investigation has now been carried out and the existence of the second solid isomeride defi- nitely established. Furthermore, its physical properties (melting- and freezing-point, crystalline structure, etc.) have been accu- rately determined and the conditions under which it may be ob- tained settled with considerable exactitude. According to our measurements the labile isomeride freezes at 1.9 C. and melts at 2.0 C., while the stable form freezes at 13.0 C. and melts at 13.2 C. Photomicrographs of both forms have also been obtained. EXPERIMENTAL PART Conditions under which the Two Forms are Obtained. It is a well- known fact that various dynamite mixtures, e.g., nitroglycerine, with various " dopes," such as wood-pulp with sodium, potassium or ammonium nitrate, freeze much more readily than does nitro- glycerine alone, and as it is customary to make use of this fact in rv] Congress of Applied Chemistry 39 preparing solid nitroglycerine, a mixture of nitroglycerine with wood-pulp was taken, cooled to 40 C., stirred with a glass rod and then the frozen product used for inoculating another sample of the same nitroglycerine also cooled to the same temperature. To the author's surprise, and as already stated, the crystalline product so obtained was found to melt around 2 C., and thus represented Kast's labile form. This experiment was repeated many times with different samples of nitroglycerine recently prepared and not previously frozen, and always with the same result. When, how- ever, a mixture of the same nitroglycerine with wood-pulp and powdered sodium nitrate was taken, then on carrying out the same experiment under the above conditions, (i.e., at a temperature of 40 C.) the product crystallized out more readily but was found to possess a melting-point of around 13 C., i.e., it was the stable modification. This experiment has also been repeated very many times and always with the same result, viz., the isolation of the higher-melting isomeride, although its formation appeared in a large number of cases to proceed through the labile form. If in the above experiments a sample of previously frozen nitroglycerine is used instead of one which has not been previously frozen, the effect of mixing with wood-pulp and cooling as above sometimes gives the labile and sometimes the stable modification. Even in those cases where the labile form was obtained it would almost invariably go over very readily (e.g., by rubbing vigorously with a glass rod) into the stable isomeride. The addition of sodium nitrate to the wood-pulp, when mixed with nitroglycerine which has been previously frozen, invariably gives under these condi- tions the stable form. Other substances may be substituted for the wood-pulp in the preparation of the labile derivative, e.g., powdered glass-wool, cotton-wool, powdered unglazed porcelain, etc., the most efficient of all being powdered glass-wool. The following methods are to be recommended for the preparation of the two isomerides: Preparation of the Labile Isomeride. Unless nitroglycerine which has not been previously frozen, and preferably recently made, is employed, no guarantee can be giventh at the labile form will separate out in the following experiments, but with this proviso the following procedure (as shown in the numerous ex- 4:0 Original Communications: Eighth International [VOL. periments carried out by us) will invariably give the labile isomer- ide. As thus obtained it is not convertible into the stable form by friction, i.e., rubbing with a glass rod, and is in reality a com- paratively stable product below C. provided it be kept free from inoculation with the higher-melting isomeride. The nitro- glycerine (2 to 3 drops) is "absorbed" in powdered glass-wool, cooled to 40 C., and stirred with a glass rod at this temperature. A trace of the frozen product is then transferred to another sample of the nitroglycerine also cooled to 40 C. and the sides of the tube just below the surface of the liquid rubbed with a glass rod; after a few seconds crystallization to the labile form commences. Preparation of the Stable Isomeride. The procedure corresponds exactly with that indicated above for the labile form except that a mixture of wood-pulp and sodium nitrate is employed instead of the glass-wool, and preferably, although not necessarily, pre- viously frozen nitroglycerine. The fact that Kast succeeded in obtaining the labile isomeride while Nauckhoff was unsuccessful is in all probability due to the accidental circumstance that Kast used a sample of nitroglycerine which had been recently made and not previously frozen, while Nauckhoff in all his experiments may have worked with samples which had been previously frozen or the manner of preparation of the crystals used for inoculation may have been unsuitable. Method of Determining the Freezing-Point of the Labile Isomeride. The apparatus adopted was that proposed by Kast (Zeit. Schiess. u. Spreng., 1906,7, p. 225) consisting, as shown on the accompany- ing blueprint, of a hard glass tube (6" by f ") surrounded by an outer tube to serve as an air-mantle which in turn is sur- rounded by a freezing mixture of ice and salt. A normal ther- mometer, graduated to 0.01 C. and calibrated at the Reichsen- stalt, Berlin, was employed for taking the temperature, the stirring being effected by means of a platinum stirrer operated mechanically in the manner shown. This device, due to the flexibility of the stirrer, obviates the danger of the latter becoming imbedded in the frozen material, and a true vertical motion is maintained by the " sleeve attachment" in which a brass bar slides, to which the platinum stirrer is attached. In our experiments the tem- perature of the outer bath was 10 to - 14 C., the nitroglycerine PHOTOMICROGRAPHS OF THE LABILE ISOMERIDE OF NITROGLYCERINE. IV] Congress of Applied Chemistry 41 being cooled to 5 to 6C. before being "seeded" with a small crystal of either isomeride introduced on the bulb of the thermometer. The stirrer was then agitated vigorously and the maximum temperature indicated by the thermometer was taken as the freezing-point of the sample. The time elapsing before this temperature was reached varied only slightly (between 9 and 12 minutes for both forms) with the different samples. Also it made no appreciable difference in the value obtained for the freezing-point whether the product was supercooled in one case to 4 C. and in another case to 8 C. In all experiments, except in the one or two special cases noted, a quantity of 8 cc. of the sample was employed. This is only about half the quantity em- ployed by Kast but a comparative experiment using double this amount, viz., 16 cc., was carried out and gave an identical value for the freezing-point, as indicated below: EFFECT OF USING DIFFERENT QUANTITIES OF NITROGLYCERINE (Comparative Experiments with Sample No. 3) Quantity employed. Temperature to which product was cooled. Time taken to obtain maximum temperature. Labile Isomeride. Freezing Point. Melting Point. 8 CC. 16 cc. c. -5 -5 mins. 10 13 c. 1.3 1.3 c. 1.7 1.6 Duplicate tests gave identical results. When the rate of stirring was decreased so that the length of time to attain the maximum temperature increased from 8 to 20 minutes, practically no difference in the value for the freezing- point was found. Thus in one case (No. 3 sample) when rapid stirring was employed the freezing-point was found to be 1.3 C. and with slow stirring, 1.22 C. Determination of the Freezing-Point of the Stable Isomeride. Exactly the same procedure was adopted, except that the nitro- glycerine was first cooled to 5 C. to 6 C., the temperature of the outside bath being maintained at C. 42 Original Communications: Eighth International [VOL. Method of Determining the Melting-Point of the Two Isomerides. This was determined by replacing the outer cooling bath used in the determination of the freezing-point by water at a tempera- ture of 15 C. and then stirring until the bulb of the thermometer just became visable. This point corresponds, as pointed out by Kast (loc. cit. p. 225) , to the point above which the thermometer commences to rise very rapidly. The melting-point of the stable isomeride was determined in a similar manner except that the temperature of the outer bath was raised to 25 to 30 C. The results obtained, together with the description of the samples of nitroglycerine employed, including the analytical data, are given in Table I. TABLE I No. Description of Product % Ni- trogen Labile Isomeride Stable Isomeride F. P. M. P. F. P . M. P. 1 Product was made by nitrating dynamite glyc- C. C. oC, C. erine with the usual nitrating acid. 1 The nitro- 0.9 glycerine obtained was thoroughly washed with dilute carbonate solution and then with water, 18.47 1.0 1.4 12.2 12.6 finally dried over solid potash in a vacuum des- 18.47 12.3 12.7 iccator; it possessed a light straw color. 1.0 2 This was prepared in a similar manner to No. 1 1.3 1.7 12.5 12.9 from the same dynamite glycerine. Product 18.50 possessed a somewhat darker color than No. 1. 1.3 1.7 12.5 12.8 3 This was a commercial sample of nitroglycerine shipped to the Experimental Station from the 1.1 1.5 not 12.6 Repauno Works, Gibbstown, N. J., in acetone 18.45 deter- solution about 1905. It was purified as above. 1.0 1.5 mined 12.6 Product possessed a bright yellow color. 4 This sample of nitroglycerine was prepared from C.P. glycerine, obtained by redistilling the C.P. trade product under reduced pressure and then nitrating this colorless product with the same 18.50 18.50 1.0 0.9 1.4 1.4 12.2 12.2 12.5 12.6 acid as used for preparation of samples No. 1 and No. 2. It was purified in the same manner and was a bright, practically colorless liquid. 25 This was prepared by extracting a sample of 40% "Atlas Powder" (a straight dynamite) with 18.32 0.4 1.2 11.3 12.4 ether and evaporating off the ether under reduced pressure at ordinary temperature. 18.36 0.4 1.2 11.4 12.5 Product was a dark yellow color. 1 This nitrating acid had the following composition: Actual H 2 SO 4 61.35< Actual HNOs 32.K H(NO)SO< HsO 2 This sample was received in a frozen condition and was thawed out prior to extracting with ether. IV Congress of Applied Chemistry 43 In some experiments on the lowering of the freezing-point of nitroglycerine, induced by various additions, carried out in 1904 at the Eastern Laboratory of the E. I. duPont deNemours Powder Company, at Gibbstown, N. J., a value for the freezing- point of the technical product equal to 12.8 C. was obtained, identical with the value given by Nobel in 1866 (loc. cit.). A review of the above figures shows that the values obtained in the present investigation are approximately 1 C. lower than those obtained by Kast and Nauckhoff, which indicated the pos- sibility of the various samples either containing a trace of moisture or a small amount of some impurity, e.g., nitrated polyglycerines. Accordingly, samples No. 2, No. 3 and No. 4 were submitted to a process of fractional recrystallization by cooling, seeding with the stable isomeride and after partially freezing, pouring off the mother liquor. The crystals thus obtained were then melted, and the liquid product dried for a week over anhydrous barium oxide and phosphorus pentoxide, under a pressure of 15 mm. The freezing- point and melting-point of each sample was then redetermined, when the following values were obtained: TABLE II Labile Isomeride Stable Isomeride Freezing Point Melting Point Freezing Point Melting Point Sample No. 2 Sample No. 3 Sample No. 4 1.7 1.9 not det 1.9 2.0 ermined 12.8 13.0 12.4 13.0 13.2 12.7 Sample No. 6 1.8 2.0 12.9 13.1 A second fractional recrystallization did not result in any altera- tion in the above values. It will be observed that in each case there is a marked rise in the freezing- and melting-points, and that those obtained with samples No. 2 and No. 3 are in sub- stantial agreement with Kast's figures. Curiously enough the sample of nitroglycerine from supposedly C.P. glycerine gave the lowest values, pointing to some impurity in the glycerine itself. In consequence another specimen of refined glycerine was taken 44 Original Communications: Eighth International [VOL. and without submitting it to a distillation under reduced pressures as in the previous case, it was nitrated directly at a temperature of + 15 C. with the same nitrating acid as used previously, and in this way after carefully washing, drying and recrystallizing twice by a process of partial freezing as above, was obtained as a per- fectly colorless glistening product. It was then dried for 4 to 5 days in a vacuum over anhydrous barium oxide and phosphorus pentoxide. This specimen (sample No. 6) gave much higher values as indicated in Table II, and apparently represented a much purer specimen than sample No. 4. It is of interest as pointing to the greater purity of these recrystallized products, that the time taken to obtain the maximum reading for the freezing-point was around 18 to 20 minutes, compared with 9 to 12 minutes in the earlier experiments, showing that crystallization took place more slowly as was to be expected with a purer product. A comparison of the above values with those obtained by Nauck- hoff and Kast is given below in Table III. TABLE III Description Labile Stable of sample of Appearance % Nitro- Isomeride Isomeride ine used F. P. M. P. F. P. M P C. C. C. 12.6 Nauckhoff (1905) C.P. Pale Yellow 18.48 not dete rmined 12.3 to 12.4 Kast (1906) C.P. Almost colorless 18.50 2.1 2.8 13.2 13.5 Nauckhoff (1911) C.P. Almost colorless not dete rmined 13.5 13.5 Hibbert (1912) C.P. Colorless 18.50 1.8 2.0 12.9 13.1 No appreciable difference was found in the individual values for the freezing- and melting-point, an observation not altogether in agreement with Kast's figures but in harmony with the later work of Nauckhoff (loc. tit.). Solubility of the Isomerides in Various Solvents. An attempt was made to ascertain whether some difference in behavior of the two isomerides towards various solvents could not be detected and for this purpose solubility experiments were carried out at 10 C. At this temperature both dissolve readily in acetone, ethyl acetate, ether, methyl and ethyl alcohol, but more sparingly iv] Congress of Applied Chemistry 45 in butylene nitrate. They are insoluble in carbon tetrachloride and chloroform at the same temperature of 10 C. The experiment with butylene nitrate was carried out with a view to making molecular weight determinations with this solvent, which freezes around 16 C., and our experiments served to show that it could probably be made to serve this purpose. Sensitiveness to Shock of Liquid Nitroglycerine and of the Two Solid Isomerides. Experiments were carried out by allowing a weight of two pounds to drop onto the substance placed in the cup of a steel die 1 mm. deep, the test being conducted out of doors with the atmospheric temperature around 1 to 3 C. The liquid product was supercooled to this temperature before being used, a small drop being placed in the center of the die, while with the crystalline isomerides the powdered substance was placed on the bottom of the die in a thin layer. The following figures represent the maximum height at which no explosion occurred: Liquid nitroglycerine (substance supercooled to -2 to -3 C.) 6 inches Stable form, M.P. 12.6 C 10 to 12 inches Labile form, M.P. 1.0 C 14 to 16 inches These experiments would seem to establish the fact that the liquid nitroglycerine under the above conditions is much more sensitive to shock than either of the two solid isomerides, but with regard to the latter a point which should be mentioned is that the stable form was somewhat more finely powdered than the labile. Crystalline Structure of the Two Isomerides. Nauckhoff in his recent paper (loc. cit.) quotes the measurements carried out for him by Flink to show that the stable form of nitroglycerine crystallizes in the bipyramidal class of the rhombic system, and full crystallographic data are given regarding the same. A crystal- lographic investigation conducted by us with the kind assistance of Prof. Amos C. Brown of the University of Pennsylvania, using the polarizing microscope, convinced us of the accuracy of the above classification, and, as Nauckhoff was unable to obtain the labile isomeride, we have also undertaken an investigation of its crystalline structure, and there appears to be little doubt (judging 46 Original Communications: Eighth International [VOL. from measurements of the angle of extinction, etc., made by means of the polarizing microscope) that the body crystallizes in the triclinic system. We were fortunate in securing excellent photo- micrographs of both forms, prints of which are shown on the follow- ing pages. In this connection it should be pointed out that both for crystallographic and microscopic work on the labile isomeride it is strongly advisable to use nitroglycerine which has been re- cently made and not previously frozen, since the product thus obtained apparently does not change into the stable form so readily. Conditions Influencing the Transformation of the Labile into the Stable Isomeride. It must be stated at once that while in every case when cooled nitroglycerine was inoculated with either form the same isomeride crystallized out, yet in no case was it found possible to convert the higher melting isomeride (stable form) di- rectly into the lower melting (labile form) without proceeding through the liquid phase. On the other hand the introduction of a trace of the stable isomeride into the solid or partially melted labile form caused a practically instantaneous and complete transformation of the labile into the higher melting product, accompanied by a considerable rise in temperature as indicated in the following experiments: A quantity of 10 grams of the frozen labile form was cooled to 2 to 4 C. and then the product inoculated with a quantity of the stable isomeride, introduced on the end of the thermometer bulb. The mixture was stirred vigorously causing the labile isom- eride to change over almost instantly into the stable modifica- tion, accompanied by a rise in temperature of some 10 C., such rise taking place within 30 to 40 seconds. When a similar experi- ment was tried with liquid nitroglycerine, supercooled to the same temperature, the rise in temperature on inoculating with the stable form was approximately of the same order of magnitude, so that apparently the heat evolved in the change from the liquid to the labile form would seem to be much less pronounced in character, the principal " energy-change " apparently being found in the conversion of the lower melting solid into the higher melting solid. At temperatures below 1 C. the change (lower-melting isomeride higher-melting isomeride) is apparently an irreversible one. The crystalline labile isomeride may be kept for a considerable rv] Congress of Applied Chemistry 47 time (one to two weeks) without change but the crystals then slowly lose their transparency, becoming opaque and the body is slowly converted into the higher melting derivative. The ease of transformation of the lower melting into the higher melting product apparently depends on a number of various subtle conditions. Thus, for example, the labile form prepared from nitroglycerine which has been previously frozen as the stable modification would appear to be more easily converted into the stable isomeride than when the labile product is obtained from recently prepared nitro- glycerine which either has not been frozen or at least only as the labile isomeride. In the former case the application of friction, such as rubbing the product, prepared either at 20 C., 40 or 60 C. with a glass rod was in general sufficient to bring about the transformation; in fact it took place at times so readily that it was very difficult to know just exactly to what the sudden change over into the higher melting product was due. These conditions were not found to arise, however, when the labile form was pre- pared from a sample of nitroglycerine freshly prepared and not previously frozen. The following experiments were all carried out with a sample of the labile form prepared from recently made and not previously frozen nitroglycerine. (A) Effect of Friction. The labile isomeride could not be con- verted into the stable form either at -5, -20, -40 or -60 C. by rubbing vigorously with a glass rod. (B) Effect of "Seeding" the Labile with the Stable Isomeride. When a quantity of the labile derivative was powdered out of doors in a cold mortar, the temperature being between and 2 C., and a trace of the stable isomeride incorporated, then although there was no suspicion of melting visible, the whole mass went over rapidly into the stable form, thus proving that the presence of the " solid " stable derivative was sufficient to induce the change into the labile product. This experiment is of con- siderable importance as indicated later in the theoretical discus- sion, in enabling us to reach a decision as to whether the deriva- tives in question are "purely physical" or "chemical" isomerides. 48 Original Communications: Eighth International [VOL. (C) Effect of Various Metallic Salts in Inducing the Change, Labile Stable. In attempting to find an explanation of the fact that while with wood-pulp and not previously frozen nitro- glycerine at 40 C. the labile form separates out, while on the other hand, with a mixture of the same nitroglycerine, wood-pulp and sodium nitrate, we get the stable isomeride under these con- ditions, several interesting facts came to light, without, however, enabling us to solve this peculiar problem. It was thought at first that the phenomenon was possibly connected in some way with the isomorphism of sodium nitrate with the stable form of nitroglycerine, but this was soon shown to be without foundation. The further possibility that the sodium nitrate might form some kind of double compound with nitroglycerine which would be isomorphous with the stable form was disproved by saturating nitroglycerine with sodium nitrate at the ordinary temperature, cooling this solution to 40 C. and "seeding" with the labile isomeride. The product crystallized out completely as the labile modification and this independent of whether the nitroglycerine employed had been previously frozen or not. A saturated solu- tion of either potassium or ammonium nitrate behaved similarly. Further experiments by taking respective samples of nitroglycer- ine which had and had not been previously frozen, either perfectly dry or containing a little moisture, "seeding" with the labile form, melting this completely by keeping at 25 C. for one or two minutes, and then adding powdered sodium, potassium or ammo- nium nitrate, and cooling to 40 C., gave rise to no separation of crystals, but on scratching the walls of the test tube vigorously with a glass rod under the surface of the liquid, the "labile" derivative separated out in every case. The experiments seem to prove that the presence of the metallic salts alone is not sufficient to cause the change from labile into stable. When, however, the liquid product obtained by melting the labile isomeride was kept in the liquid condition for a longer period, e.g., about one hour at 20 C., the product could not be induced to crystallize by cooling to 40 C., even in the presence of the above salts. Various other experiments were tried, e.g., by inoculating the solid labile form obtained from not previously frozen nitroglycerine with pow- dered sodium, potassium and ammonium nitrates, and rubbing * PHOTOMICROGRAPHS OF THE STABLE ISOMERIDE OF NITROGLYCERINE. iv] Congress of Applied Chemistry 49 the product vigorously with a glass rod. The presence of sodium and ammonium nitrate was effective in bringing about the con- version of the labile into the stable form in a number of experi- ments tried, but under the same conditions no such conversion could be induced when potassium nitrate was used. This marked difference in the behavior of potassium as contrasted with that of sodium and ammonium nitrate proved to be an outstanding feature throughout this work, since neither when used with nitro- glycerine alone nor in conjunction with wood-pulp, glass-wool, etc., was it capable of inducing the conversion of the labile into the stable modification. The assumption that this difference in behavior may be occasioned by some impurities in the sample of sodium nitrate employed would seem to be unwarranted, since the liquid nitroglycerine could be supercooled in the presence of the same sample of powdered nitrate and no crystallization ensued. Also as stated above even when the solid labile form was incor- porated with the powdered sodium nitrate this treatment was not always successful in causing the conversion of the labile into the stable modification. A further peculiarity is the fact that when the wood-pulp-sodium-nitrate mixture is replaced by a mixture of ground glass-wool with sodium nitrate the behavior of these two mixtures with not previously frozen nitroglycerine is quite different, since in the latter case the sodium nitrate apparently exerts no effect on the labile form, this latter invariably crystalliz- ing out in the limited number of experiments carried out by us. It was noticeable, however, that when the sample of nitroglycerine employed contained moisture the labile form separating out showed a much more marked tendency to go over into the stable isomeride, especially on rubbing with a glass rod. Previous experiments would seem to indicate that the change of the labile into the stable isomeride is not generally induced by friction or by the presence of sodium nitrate or wood-pulp acting alone, but arises rather from the joint action of all three factors. In view of the fact just mentioned that the presence of moisture seems to facilitate the conversion of the labile form when a metallic salt is present, the action of wood-pulp may possibly be connected in some way with its moisture content. It would seem next to impossible to give a satisfactory explanation of such abnormal behavior, but for the 50 Original Communications: Eighth International [VOL. purpose of obtaining a general oversight, the results obtained have been tabulated below. The procedure in each case was to cool the product first to 40 C. and then stir it vigorously with a glass rod, afterwards to transfer a trace of this frozen product to another sample of the same nitroglycerine cooled to the same temperature, and then rub the inner surface of the test tube below the liquid vigorously with a glass rod for some seconds. TABLE IV No. of Exp. Description of sample of nitroglycerine used Mixture incorpor- ated with the nitroglycerine Temperature to which prod- uct was cooled Nature of product separating out 1 2 3 4 5 The nitroglycerine was em- ployed both as the pre- viously and not previous- ly frozen product, also an- hydrous and containing a little moisture. Same Sodium nitrate and wood-pulp Potassium nitrate and wood-pulp or glass-wool Ammonium ni- trate and wood- pulp or glass- wool Sodium nitrate and glass-wool Same 40 C. 40 C. 40 C 40 C. 40 C. Stable isomeride almost in- variably, though labile form appeared tran- siently in many cases. Labile isomeride invari- ably obtained. Labile isomeride sepa- rated out almost invari- ably. The labile isomeride sepa- rated but presence of moisture seemed to give a product more easily converted into the stable form. Sometimes the labile and sometimes the stable isomeride. Same Nitroglycerine not pre- viously frozen. Nitroglycerine previously frozen. As far as our experiments go there appears to be no doubt but that previously frozen nitroglycerine freezes much more readily than not previously frozen, and this is especially true with recently frozen samples. When nitroglycerine is frozen as the labile form, then melted and kept in the liquid state for not more than about one minute, the labile isomeride was found to separate out again on cooling to 40 C., while when frozen as the stable derivative, then melted and again cooled, the stable modification crystallizes out when "seeding" is not resorted to. If nitroglycerine is first frozen as the labile isomeride, then melted completely, this " seeded" with the stable form, the stable isomeride so obtained, melted for two or three minutes, cooled and then " seeded" with iv] Congress of Applied Chemistry 51 the labile form, the labile isomeride thus obtained melted for about one minute, and then cooled without further inoculation, we get the labile modification crystallizing out, which, however, under these conditions changes over almost immediately into the stable isomeride on rubbing with a glass rod. It is an interesting fact that the labile isomeride can be melted, the liquid product heated to a temperature of 20 to 30 C., i.e., some 8 to 18 C. above the melting point of the stable form, maintained there for one minute, and then on cooling to 40 C. and scratching the inner surface of the tube can be caused to separate again as the labile modification. The above change occurs when the labile form is kept only for a short period as indicated at a temperature of 20 to 30 C. With a longer period than this it was not found possible to induce spontaneous crystallization of either form, though as far as could be judged no trace of solid was left in the first experiment, this being shown by the fact that cooling alone to 40 C. produced no crystallization, this being only induced by scratching with a glass rod. These experiments would seem to indicate that the two isomerides may exist in the liquid state, but if we conclude, as is assumed later, that we are dealing only with ''physical" isomerides, then the work of Schaum (Ann. 300, 209 [1898]) and others does not bear out such a contention. We were not successful in effecting the crystallization of nitro- glycerine at a temperature of 17 to 20 C. (even on standing for 24 hours) with recently made and not previously frozen samples, even when vigorous scratching of the walls of the test tube by a glass rod was resorted to. Furthermore, at this temperature the addition of wood-pulp alone or wood-pulp with sodium, potassium or ammonium nitrate used individually or collectively was in- effective in bringing about the crystallization of the nitroglycerine at this temperature, while with a previously and especially a re- cently frozen product the freezing could be induced without very much difficulty without any addition or inoculation. THEORETICAL DISCUSSION It is necessary at the outset to consider whether in the case of the two isomeric forms of nitroglycerine described above we are dealing with "purely physical" as distinct from " chemical" 52 Original Communications: Eighth International [VOL. isomerides. In the first place it cannot be stated with certainty that we are not simply dealing with two polymers, and this ques- tion cannot be decided until molecular weight determinations of the two forms have been made, for which purpose butylene glycol dinitrate, which freezes around 16 C., would probably serve as a good solvent. There is also the point that " structural" isomerism is possible as indicated by the two formulae (" A" and "B") shown below: .0 CH 2 -0-N / '% CH 2 -0-N^_ / \ CH N CH 2 -0-N I \ / V) (W-O-N \ "A" "B" However, a review of the experimental evidence obtained, and especially in view of the extraordinary far-reaching analogy be- tween these derivatives and the two isomeric forms of benzophe- none convinces us that we are dealing here with a case of " physical" rather than with one of "chemical" isomerism. In a very inter- teresting paper on isomerism published some years ago by Schaum (Ann. 1898, 800, p. 209), the question of physical and chemical isomerism is discussed very thoroughly from a physico-chemical standpoint, with especial reference to the case of the two isomeric forms of benzophenone. According to this author the criterion between chemical and physical isomerism is to be found in the fact that with purely physical isomerides, inoculation of the solid labile with the solid stable form brings about the complete con- version of the former in the absence of any solvent. If it is not possible to convert the solid labile form into the solid stable iso- meride directly, i.e., by simple inoculation, but only by passing through the liquid phase or by the use of a solvent, then we are iv] Congress of Applied Chemistry 53 dealing, according to Schaum, with " chemical" isomerides. From his own experimental work and that of others, this author arrives at the conclusion that physical isomerism does not exist in the liquid but only in the soild state; in other words, the liquids obtained by melting both forms are identical in character. Viewed from this standpoint, the two isomeric forms of benzophenone are physical, not chemical isomerides. Since the solid labile form of nitroglycerine when incorporated with the stable form goes over practically instantaneously into the higher melting isomeride (page 17) no solvent being present, it would seem that we are also dealing here with physical rather than chemical isomerides. In order the better to grasp the analogy between the isomeric forms of nitroglycerine and those of benzophenone, the properties of the latter are quoted below in brief form. The stable form of benzophenone melts afc 48 to 49 C. and the second isomeride (labile form) at 26 to 27 C. This latter deriva- tive was discovered by Zincke in 1871 (Ann. 159, p. 377) who prepared it by the distillation of benzophenone obtained as an oxidation product from diphenylmethane. He showed that its behavior with solvents was different from that of the stable form, e.g., it could not be recrystallized from alcohol, etc., and apparently crystallizes in the monoclinic system, the stable on the other hand belonging to the rhombic class. The transparent crystals of the labile isomeride after standing for several weeks become opaque, changing over slowly into the stable modification. When a quan- tity of the labile isomeride was inoculated with a trace of the stable form there was an instantaneous conversion into the higher melting derivative accompanied by evolution of considerable heat. If a sample of supercooled liquid benzophenone is inoculated simultaneously at different points with a trace of both isomerides, both forms develop, though the stable form separates out much more rapidly and in a short time the whole product is converted into the stable modification. It was not found possible to re- convert the solid stable isomeride directly into the solid labile form. Later work by K. Auwers and V. Meyer, Ber. 1889, 22, 550, developed the following facts: The most reliable method for obtaining the labile form consists in preparing freshly distilled 54 Original Communications: Eighth International [VOL. benzophenone and cooling this to C. The liquid distillate may be kept at room temperature for days without solidifying but on cooling to C. it crystallizes out more or less slowly as the lower melting labile form. It can then be melted, solidified, and re- melted repeatedly without change, but if the solid product is rubbed with a glass rod or pestle, or if a trace of the solid stable form is introduced, we have an immediate conversion of the labile into the stable isomeride accompanied by a considerable rise in tem- perature. In the article referred to above, Schaum shows that it is not necessary to heat benzophenone to the boiling point in order to obtain the labile modification, that although this is a reliable method, yet it is possible by heating the stable form to only a few iv] Congress of Applied Chemistry 55 degrees above its melting point and then cooling, to occasionally obtain the labile modification. He points out further that "Wenn wir beispielweise fluessiges Benzophenon unterkuehlen so kommen wir in das metastabile Gebiet. Beim Ueberschreiten der Metastabilitaetsgrenzen bei fallender Temperatur gelangt dann der Koerper in ein labiles Gebiet in welchem er sich auch bei Keimfreiheit in eine stabilere Form unwandeln muss," and in fact was able to show that with the product in the metastable condition indicated in the quotation above, the vibration or shock caused by scratching the sealed glass tube containing the benzo- phenone with a file, provided a sufficient disturbance to cause the change over into the stable modification. However, to properly characterize the "metastable condition" he considers it necessary to add (p. 217): "Ausser der Anwesenheit der stabilen Phase auch noch allerhand aeussere Einwirkungen die Umwandlung hervorzubringen im Stande sind," a statement which we can heartily support in view of the anomalous results met with by us in the examination of the nitroglycerine isomerides. A careful review of the experimental work performed on the isomeric forms of nitroglycerine shows that the above properties of benzophenone are almost identical in every respect with those of the nitroglycerine isomerides. Thus, for example, the labile form of nitroglycerine can only be obtained with certainty by one method, viz., the use of a freshly prepared, not previously frozen product. Further, it is converted into the stable form instantly by direct inoculation and slowly on standing at ordinary tempera- ture. The stable cannot be reconverted directly into the labile form, and when used to inoculate supercooled liquid nitroglycerine simultaneously with the labile form it crystallizes out more rapidly than the latter, finally causing the solidification to take place entirely as the stable derivative. On rubbing with a glass rod, the labile form was in a number of cases (dependent on the mode of preparation) transformed into the stable isomeride, and in harmony with the behavior of the labile form of benzophenone the labile isomeride of nitroglycerine could be melted and solidified re- peatedly without change. It is also a curious fact that both the stable form of nitroglycerine and that of benzophenone crystallize in the rhombic system, the labile form of nitroglycerine crystalliz- 56 Original Communications: Eighth International [VOL. ing in the triclinic and possibly also that of benzophenone, since Zincke (loc. tit.) gives the monoclinic system for this body but admits the uncertainty of his identification. Finally, as indicated on page 17, there is considerable heat evolved in the change from the labile into the stable isomeride. It would be difficult to im- agine a closer analogy than between the isomeric forms of these two substances. The assumption would appear to be justified that we are also dealing in the case of nitroglycerine with a case of "physical" and not "chemical" isomerism, i.e., that the two derivatives differ only in the "energy-value" represented by their different crystalline structures. Sensitiveness of Frozen Dynamite to Shock. The view previously held by many technical chemists that frozen dynamite is more sensitive to shock than the liquid product is gradually being abandoned, and as indicated in our own experiments on the " drop test " is incorrect. It must, however, be pointed out that such remarks refer to the products themselves and not to the conditions arising during the change of physical state when the liquid passes over into either of the two solid forms. It has been shown that the labile isomeride may be induced to go over, under certain conditions, practically spontaneously into the stable modification, considerable heat being evolved in the process. It is difficult to imagine that the heat thus evolved could provide a sufficient " energy-stimulus " to induce a premature explosion; rather are the conditions for such to be sought for, in all probability in the peculiar synchronistic state possessed by the nitroglycerine mole- cule in the actual change of the liquid into the solid state or espe- cially during that of the lower into the higher melting isomeride, the latter being possibly analogous to the change of state of the yellow modification of mercuric iodide into the red variety brought about by rubbing with a metallic substance. CONCLUSIONS 1. It has been shown that solid nitroglycerine exists in two forms, a labile and a stable isomeride possessing the following freezing- and melting-points: Labile Form Stable Form F.P. 1.9 C. F.P. 13.0 C. M.P. 2.0 C. M.P. 13.2 C. rv] Congress of Applied Chemistry 57 which values are in substantial agreement with those obtained by Kast and Nauckhoff. 2. Methods for the preparation of both isomeric forms have been developed and their various physical properties, solubility in sol- vents, sensitiveness to shock, crystallographic structure, the con- ditions under which one is converted into the other, and the influence of metallic salts in effecting such conversion have been investigated. Photomicrographs of both isomerides have also been prepared. 3. The substitution of potassium for sodium nitrate in the " dope " employed in making dynamite would presumably give a product freezing at a lower temperature, since with the former salt there is no tendency towards the formation of the higher-melting stable isomeride, although we are not aware of any observations indicating that dynamite made with commercial saltpetre is more difficult to freeze than that made with sodium nitrate. 4. The extraordinarily far-reaching analogy between the iso- meric forms of nitroglycerine and benzophenone has been com- mented upon and the evidence pointed out in favor of regarding the isomerism as " physical" rather than " chemical." The author wishes to express his thanks to the Director of the Experimen tall Station and the officials of the duPont Powder Company for their kind permission to publish the preceding investigation. BOILING POINTS OF SOLUTIONS OF NITROGLYCERIN BY A. L. HYDE Bureau of Mines, Pittsburgh, Pa. The following work is the result of an attempt to obtain a method for determin'ng approximately the mean molecular weight of mixtures of nitroglycerin with allied bodies. The determination of rise in boiling point seemed to be the most promising method since most of the solvents for nitroglycerin freeze at so low a temperature as to make the freezing-point method not feasible. From the nature of the case no very exact determinations could be expected from the boiling-point method, but it was thought that by using fairly concentrated solutions so that the rise in boiling point would be considerable, information of some value might be obtained. Accordingly all the solutions used were of considerable concentration. The apparatus used was a slight modification of the ordinary molecular-weight apparatus of Beckmann. Since it was important that the concentration of the solutions should change as little as possible during the observation, it was necessary to condense the vaporized solvent as near to the boiling solution as possible. This was done by means of a coil of copper tubing wound around the outside of the tube containing the solution. This coil was just above the surface of. the solution and cold water was kept circulating through it. In this way only a small fraction of the solvent was at any time away from the main body of the solution. Heating of the solution was accomplished by means of a coil of fine platinum wire sealed into the bottom of the tube and heated by an electric current. In this way the heating was done within the body of the solution instead of from the outside and over- heating was avoided. The thermometer used was an ordinary Beckmann reading to hundredths of a degree. The procedure was as follows : A suitable quantity of the pure solvent was poured into the tube and the thermometer inserted. The solvent was then boiled and when equilibrium was established the boiling point was noted. The tube was then emptied and 59 60 Original Communications: Eighth International [VOL. dried. A quantity of nitroglycerin was then placed in a small tared beaker and accurately weighed. A suitable quantity of solvent was poured upon the nitroglycerin, the beaker rotated until a homogeneous solution was obtained and the total solution weighed. This solution was quickly transferred to the apparatus and its boiling point determined under the same conditions as that of the pure solvent. The difference in the boiling points after proper correction represents the influence of the quantity of nitroglycerin present. After several determinations the boiling point of the pure solvent was again determined to find if there had been any change due to variation in atmospheric pressure. In case of change, usually only a few hundredths of a degree, it was assumed that the rate of change had been constant and cal- culations made accordingly. Assuming that the law for dilute solutions holds also in the case of concentrated solutions, the molecular weight of nitroglycerin should be given by the formula, m= , where g is grams of nitro- a glycerin per 100 grams of solvent a is the rise in boiling point, and r is the boiling point constant of the solvent used. As a matter of fact, it will be shown that this law does not hold true in most concentrated solutions of nitroglycerin. Only one of the solvents tried shows even approximate conformity to this law. The solvents tried were ether, acetone, methyl alcohol, chloroform, and ethyl acetate. The results with each of these are given below. Ether. Several precautions are necessary when using ether as a solvent. In the first place, the low boiling point of ether makes its solutions more subject to changes in concentration than is the case when higher boiling solvents are used. Hence the weighing should be made in a stoppered weighing bottle and both weighing and transfer to the boiling-point apparatus should be as rapid as possible. The low boiling point also makes necessary a consider- able condensing surface hi the apparatus to prevent escape of part of the solvent. Again, the electric current used for heating must be carefully regulated, since excessive heating of the platinum wire in contact with ether causes formation of formaldehyde. iv] Congress of Applied Chemistry 61 This of course changes the boiling point. Below are results ob- tained with ether as solvent. In this table, (g) represents grams of nitoglycerin in 100 grams of solvent, (a) represents rise in boiling point in degrees Centigrade, and (m) is the apparent mole- or j 1 cular weight of nitroglycerin calculated by the formula, m = . a The value of (r) used is 21.1, which is the value for ether given in the Chemiker Kalender, 1910. gam 10.81 .86 264 17.70 1.35 277 34.51 2.29 318 72.63 3.79 404 Since the theoretical molecular weight of nitroglycerin is 227, these results seem to indicate that in concentrated ether solutions of nitroglycerin there is association of the molecules of nitrogly- cerin and that this association increases with increase of concen- tration. Acetone. Only two determinations were made with acetone as a solvent. These results are as follows, the value of (r) used in calculating (m being 16.7). gam 21.00 1.68 209 48.19 4.26 189 These results as far as they indicate anything indicate a dissocia- tion of the nitroglycerin molecule instead of an association as was the case when ether was used as solvent. Methyl alcohol. The boiling-point constant (r) for methyl alcohol is low 8.8. This makes necessary the use of very concen- trated solutions of nitroglycerin if the rise in boiling point is to be considerable. Accordingly the solutions with methyl alcohol as solvent were considerably more concentrated than the ether 62 Original Communications: Eighth International [VOL. solutions already mentioned, as follows: g 17.54 39.03 75.10 111.38 The results with this solvent were a m .55 281 1.21 284 2.01 328 2.56 383 Like the results with ether, these results seem to indicate an association of the nitroglycerin molecules. Also the association increases with increase of concentration, though the degree of association is not as great as with ether solutions. Chloroform. The boiling-point constant for chloroform as a solvent is 36.6. Hence the solutions used were not as concentrated as the methyl alcohol solutions. Below are results with this solvent. gam 14.69 1.66 324 29.26 2.69 398 57.67 3.97 532 74.93 4.59 598 An inspection of these results shows a very great degree of associa- tion of the nitroglycerin molecules. The highest concentration tried shows an apparent molecular weight of considerably more than twice the theoretical. Even the lowest concentration used gives a molecular weight far above theoretical. Here again as was the case with ether and methyl alcohol, increased association is shown with increased concentration. Ethyl acetate. Results using ethyl acetate as solvent are given below. The value of (r) used for calculating (m) is 26.1. gam 10.60 1.18 234 22.30 2.47 235 29.00 3.65 235 34.66 3.95 229 35.15 3.97 231 rv] Congress of Applied Chemistry 63 Of the solvents used, this is the only one whose rise in boiling point is approximately proportional to the concentration even at the rather high concentrations used. The uncertainty of the deter- mination of the value (a) with the apparatus used is probably several hundredths of a degree so that the calculated value of (m) when using this solvent at least approximates the theoretical value. Inspection of the results obtained with these five common sol- vents for nitroglycerin shows that of the five, ethyl acetate alone gives results nearly independent of concentration. Hence if it is desired to obtain approximately the mean molecular weight of a mixture containing nitroglycerin, this is the only one of the solvents tried which is available. Ethyl acetate is a good solvent for most of the substance's which are likely to occur in the ether extract of explosives. It is therefore well suited for the purpose in view. An example of the use of the method as a check upon other determinations may be given. The ether extract from a certain explosive is supposed to consist of a mixture of nitroglycerin and tetranitrodiglycerin. Since the molecular weight of the latter is 346, a mixture of the two should show a mean molecular weight considerably higher than pure nitroglycerin. Two determinations were made upon this mixture with following results : gam 30.50 3.06 260 35.10 3.50 262 This molecular weight corresponds to a percentage of about 29 per cent of tetranitrodiglycerin in the mixture. A nitrogen determination of the mixture in the nitrometer gave a nitrogen content of 17.74 per cent. Since tetranitrodiglycerin contains 16.19 per cent nitrogen and nitroglycerin contains 18.5 per cent nitrogen, the nitrogen content found corresponds to about 33 per cent of tetranitrodiglycerin in the mixture. Con- sidering the fact that a variation of only .10 per cent in the nitro- gen content found will make a difference of over 4 per cent in the apparent amount of tetranitrodiglycerin present, the results obtained by the two methods agree fairly well. 64 Original Communications: Eighth International [VOL. These facts lead to the conclusion that a fairly satisfactory determination of mean molecular weight of nitroglycerin mixtures may be made by noting rise in boiling point of ethyl acetate solu- tions of these mixtures. A few observations upon the above results, though not bearing upon the problem in hand, may be given for what they are worth. In the results with the solvents ether, methyl alcohol, and chloro- form, though the rise in boiling point does not follow the law for dilute solutions, yet there is a certain regularity. Consider first the results with chloroform. If the values of (g) and (a) be plotted on the x and y axes respectively of a system of co-ordinates, the points obtained lie along an apparently regular curve. If the rise in boiling point (a) were directly proportioned to the con- centration (g) as in dilute solutions, this curve would of course be a straight line. But the values actually found give a curve which bends towards the x axis as concentration increases. Though this curve appears to be a regular one, attempts to find a simple algebraic expression for it were unsuccessful. If, however, the concentration (g), instead of representing grams of nitroglycerin per 100 grams of chloroform, be expressed in grams of nitrogly- cerin per 100 grams of total solution, the curve obtained seems to follow a simple law. If the values of (g) be recalculated thus and these values be called (h), will then be the percentage concentration of the solu- tions. The values of (h) corresponding to the values of (a) with chloroform as solvent are as follows: a = 1.66 h = 12.8 a = 2.69 h = 22.7 a = 3.97 h = 36.6 a = 4.59 h = 42.3 Now, if the variations from the law of dilute solutions are really due to association of the molecules, it seems reasonable to suppose that this association should be a function of the concentration (h). Hence we might expect that the curve would be represented by some such expression as a ch n where c and n are constants. This assumption may be tested by taking any two corresponding sets of values of a and h and determining the values of c and n. rv] Congress of Applied Chemistry 65 Using these values of c and n, the values of (a) corresponding to the remaining values of (h) may be calculated and compared with the values of (a) actually determined. Taking the first two sets of values of (a) and (h) for determination of (c) and (n), the calculation is as follows: 1.66 = cxl2.8 n 2.69 = cx22.7 n 2.69cxl2.8 n = 1.66cx22.7 n 1.62xl2.8n = 22.7 n Let 22.7 = d, 12.8 = k, and d n = m Then: d, 10 log m n= logm = 1.3560 m and n = k log / m \ 10 log 1.62 10 log m .2095 1.62 /- ioiogk 1.1072 1.1072 Hence : 10 logm_ 10 logm 1.356 1.1072 M -t ~< A f~\ log m = 1.1418 1.356 c 8.556c = l.e6 c= .194 Thus the expression a = .194 h' 842 satisfies the first two sets of values of a and h. 66 Original Communications: Eighth International [VOL. Calculation of the remaining values of (a), using these values for c and n, gives 4.02 and 4.55, while the values actually deter- mined were 3.97 and 4.59. || Ether solutions of nitroglycerin as already noted also show association of the molecules. The boiling-point curve is of the same character as that for chloroform, but does not bend toward the x axis as much. Recalculation of the values for (g) as before gives the following values of h: a= .86 h = 9.75 a = 1.35 h = 15.04 a = 2.29 h = 25.65 a = 3.79 h = 42.06 Taking the extreme values of a and h and calculating as before gives the values c = .085 and n = 1.015. Using these values and calculating from the expression a = .085 h 1 ' 015 the remaining values of (a), give the values 1.33 and 2.29 as against values actu- ally determined of 1.35 and 2.29. Fina ly taking the results with methyl alcohol and recalculating g gives as values for h: a = .55 h = 14.9 a = 1.21 h = 28.0 a = 2.01 h=42.9 a = 2.56 h = 52.7 Calculation from the extreme values of a and h gives c = .0205 and n = 1.218. From these values are obtained values for (a) of 1.19 and 2.00. The values actually found were 1.21 and 2.01. Thus the expression a = .0205 h 1<218 approximates very closely to the boiling-point curve of this solvent. The above calculations seem to show that with the three sol- vents dealt with, the rise in boiling point follows a law whose expression is of the form a = ch n where a is the rise in boiling point, h the percentage composition of the solution, and c and n are constants. The agreement between calculated and determined values of a for all three solvents is again shown below: rv] Congress of Applied Chemistry 67 a calculated a determined 4.02 3.97 4.55 4.59 1.33 1.35 2.29 2.29 1.19 1.21 2.00 2.01 The writer has been unable to find any rational explanation for this expression, nor to find any relation between the values of c and n for the different solvents, and it is of course possible that the formula is not the expression of any law at all. But the fact that the rise in boiling point of three different solvents should follow so closely the course demanded by the formula makes it seem improbable that the agreement is wholly fanciful. SEPARATION OF NITROGLYCERIN FROM NITRO- SUBSTITUTION COMPOUNDS BY A. L. HYDE Bureau of Mines, Pittsburgh, Pa. The separation of nitroglycerin from nitro bodies is difficult because practically all solvents for nitroglycerin are also solvents for the different nitro bodies and vice versa. There are, however, considerable differences in the degree of solubility in different solvents and the following method of separation depends upon this fact. There are several solvents in which nitro compounds are more soluble than is nitroglycerin. Among these are acetone and acetone water mixtures, carbon tetrachloride, and carbon bisul- phide. There are also several solvents which dissolve nitroglycerin more readily than they do nitro compounds. Among these are ether, formic acid, acetic acid, and mixtures of these acids with water. It is evident that if two solvents which do not themselves mix are shaken together with a mixture of nitroglycerin and a nitro- compound there will be a partial separation of the mixture provided that in one of these solvents nitroglycerin is more soluble than the nitro compound and that in the other solvent the nitro com- pound is more soluble than nitroglycerin. Hence the conditions that two solvents shall be available for this method of separation are that they shall be nearly insoluble in each other and that one shall dissolve nitroglycerin more readily than it dissolves nitro compounds, while the other dissolves nitro compounds more readily than it dissolves nitroglycerin. The solvents mentioned above may be examined with a view of determining how far they satisfy these conditions. Acetone mixes with almost all organic solvents including ether, formic acid and acetic acid. This solvent therefore is not available. Carbon tetrachloride mixes with ether and acetic acid but only slightly with formic acid. If other conditions were favorable, carbon tetrachloride and formic acid water mixtures might therefore be used for the separation. Carbon bisulphide mixes with ether, 69 70 Original Communications: Eighth International [VOL. but only slightly with acetic or formic acids. It therefore appears that carbon bisulphide and either acetic or formic acid water mixtures would be available. Preliminary experiments with all these solvents showed that a single shaking with any pair of solvents gives only a very in- complete separation of nitroglycerin from nitro compounds. A sort of fractional separation was then attempted. The first two solvents tried were carbon tetrachloride and formic acid. A number of experiments showed that a mixture of 65 parts of formic acid to 35 parts of water by volume was the most promising. The method used was as follows: A mixture containing known quantities of nitroglycerin and liquid trinitrotoluene was poured into a separating flask and the small remainder in the beaker washed in with a definite volume of carbon tetrachloride. A definite volume of formic acid water mixture was then added and the whole shaken in the separating flask. After separation of the liquids, the carbon tetrachloride was run into a beaker. A second portion of carbon tetrachloride was added to the formic acid mixture and shaken with it and run into a second beaker. A third portion of carbon tetrachloride was added and run into a third beaker. The formic acid mixture was then removed from the flask and placed in a beaker. A second portion of formic acid mixture was then shaken with the three portions of carbon tetrachloride in turn and finally placed in another beaker. A third portion of formic acid mixture was treated in the same way. The three portions of carbon tetrachloride were then poured to- gether and the solvent evaporated off on the steam bath and the residue consisting of trinitrotoluene and some nitroglycerin weighed. Sulphuric acid was then added to this residue and the amount of nitroglycerin present determined on the nitrometer. A number of mixtures of trinitrotoluene and nitroglycerin were treated in this manner using different proportions of the two solvents, but the results were not very promising. One single result may be given. A mixture consisting of 1.790 grams of liquid trinitrotoluene and 4.713 grams of nitroglycerin was treated as above, each portion of formic acid mixture used being 30 cc. and each portion of carbon tetrachloride 45 cc. The residue after evaporation of the carbon tetrachloride weighed 1.673 grams or iv] Congress of Applied Chemistry 71 about 94 per cent of the weight of nitrotoluene originally present. A determination of the NO given off in the nitrometer, however, showed that this still contained about .4 gram of nitroglycerin. This was of course far too much for hope of quantitative results. Also there seemed to have been some action between the carbon tetrachlorid and the nitrotoluene, as upon adding sulphuric acid to the residue an odor of chlorine was noticeable. Still another difficulty is the high boiling point of carbon tetrachloride and the danger of losing nitrotoluene during the evaporation. These two solvents were therefore abandoned. The method of fractional separation as above described was next tried with the solvents carbon bisulphide and acetic acid water mixtures. Several different mixtures of acetic acid and water were tried and the one finally chosen as promising best results was a mixture containing 65 parts acetic acid (99.5 per cent pure) and 35 parts water by volume. This mixture was used with freshly distilled carbon bisulphide in the proportions of 22.5 cc. acetic acid mixture to 50 cc. of carbon bisulphide and the fractionation carried on as above described using three por- tions of each solvent. A mixture consisting of 1.007 grams of liquid trinitrotoluene and 1.743 grams of nitroglycerin treated in this way gave a residue after evaporation of the carbon bisul- phide which weighed .973 gram. Determination in the nitro- meter showed that this residue contained about .2 gram of nitro- glycerin. This separation was still too incomplete for quantitative pur- poses and fractionation of another sample was carried one step further, using four portions each of acetic acid mixture and carbon bisulphide in the same proportions as before. In this way a mixture containing 1.515 grams of liquid trinitrotoluol and 1.915 grams of nitroglycerin gave a residue weighing 1.439 grams containing about .15 grams of nitroglycerin. Another mixture containing .487 gram of liquid trinitrotoluene and 3.076 grams of nitroglycerin treated in the same way gave a residue weighing .599 gram. It was thought that results more nearly quantitative might be obtained by correcting for the amount of nitroglycerin remaining in the carbon bisulphide and the nitrotoluene remaining in the acetic acid mixture. With this end in view several separa- 72 Original Communications: Eighth International [VOL. tions were made using three portions each of carbon bisulphide and acetic acid water mixture. The results, however, showed rather large variations. They were as follows: A mixture con- taining 1.527 grams of liquid trinitrotoluene and 2.168 grams of nitroglycerin gave a residue after evaporation of the carbon disul- phide weighing 1.475 grams. A mixture containing .509 gram of liquid trinitrotoluene and 3.484 grams of nitroglycerin gave a residue weighing .695 gram. A mixture containing 1.083 grams of crystalline dinitrotoluene and 2.094 grams of nitroglycerin gave a residue weighing 1.223 grams. A mixt^aie containing .308 gram of crystalline dinitrotoluene and 1.930 1 grams of nitroglycerin gave a residue weighing .451 gram. A mixture containing .751 gram of paranitrotoluene and 2.111 grams of nitroglycerin gave a residue weighing .753 gram. A mixture containing 1.142 grams of paranitrotoluene and 2.416 grams of nitroglycerin gave a residue weighing 1.398 grams. If it is assumed that under the conditions of the separation one-tenth of the nitroglycerin originally present remains in the carbon bisulphide while from 80 to 90 per cent of the nitrotoluene is found there, these results are much more concordant than they appear at first glance. Even with this correction, however, the results show that this method is available only in a rough way. Carbon bisulphide and acetic acid are not absolutely insoluble in each other and most of the variations are probably due to this fact together with the difficulty of carrying on so many manipulations with the different portions of solvents in a uniform manner. The slight volatility of paranitrotoluene also plays some part in the last two results. This method of procedure was abandoned for the one which follows. It was thought that it might be possible to remove practically all the nitroglycerin from the nitro body and still leave a fairly definite fraction of the latter in the carbon bisulphide. This proved to be the case. A preliminary separation of a mixture containing 1.631 grams of liquid trinitrotoluene and 2.078 grams of nitroglycerin with 75 cc. of carbon bisulphide shaken with 4 portions of 30 cc. each of acetic acid water mixture gave in the carbon bisulphide about one-third of the nitrotoluene containing less than 10 mg. of nitroglycerin. A uniform method of procedure was then adopted and the fraction of different nitro compounds iv] Congress of Applied Chemistry 73 left in the carbon bisulphide determined. The results show a uniformity sufficient for a fairly accurate determination of the different nitro compounds in the presence of nitroglycerin. The following method of procedure was the one adopted. A mixture of 65 parts acetic acid (99.5 per cent pure) and 35 parts water by volume was made up. Carbon bisulphide was freshly distilled. The mixture of nitroglycerin and nitro compound in a small beaker was poured into a medium-sized separating flask and the remaining contents of the beaker washed into the flask with 30 cc. of the acetic acid water mixture. 75 cc. of carbon bisulphide was added to the flask and the whole shaken and then allowed to separate. The carbon bisulphide was then drawn off into a small Erlenmeyer flask and the acetic acid solution poured into a beaker. The carbon bisulphide was poured back into the separating flask and shaken with a second 30-cc. portion of acetic acid mixture. This was repeated with two more portions of acid mixture so that the 75 cc. of carbon bisulphide had finally been treated with 120 cc. of acetic acid mixture in 30 cc. portions. Finally the carbon bisulphide was washed with one 75-cc. portion of water to remove the acetic acid taken up and afterwards run into a small beaker. It was then ready for the evaporation. Since some of the nitro compounds are slightly volatile at the boiling point of carbon bisulphide, a special method of evaporation of the solvent was used. A bell jar having a hole in the top and one in the side was placed over the beaker containing the carbon bisulphide solution. A rubber stopper through which was passed a glass tube was placed in the top hole. The glass tube reached down to within about an inch of the top of the liquid in the beaker. A rubber tube was connected to the side opening of the bell jar to carry off the vapors of carbon bisulphide. Air, dried by passing through two calcium chloride tubes, was blown down through the glass tube upon the surface of the liquid. The liquid immediately became so cold that the loss of nitrotoluene by evaporation was very slight. Immediately after the carbon bisulphide was evapo- rated which could be determined by the melting of the few very fine particles of ice present, the beaker was placed in a desiccator over sulphuric acid and weighed after 12 hours. A test to deter- mine the evaporation of the nitro compound when using this 74 Original Communications: Eighth International [VOL. apparatus was made. The loss from 2.382 grams of orthonitro- toluene (the most volatile of the nitrotoluenes) was 7 mg. in evaporating 50 cc. of carbon bisulphide from it. The following tables show the results when using the above method. Per cent of Mixture treated Recovered from original nitro com- Orthonitrotoluene Nitroglycerin CS solution pound recovered 2.475 grams 1.841 grams 1.889 grams 76.4 1.069 grams 4.099 grams .790 gram 73.9 1.281 grams 2.705 grams .955 gram 74.6 Average percentage of nitro compound recovered, 75 per cent. Per cent of Liquid Recovered from original nitro com- Dinitrotoluene Nitroglycerin CSi solution pound recovered 2.144 grams 1.566 grams .782 gram 36.5 1.014 grams 3.902 grams .357 gram 35.2 1.169 grams 3.741 grams .415 gram 35.5 Average percentage of nitro compound recovered, 35.7 per cent. Mixture treated Per cent of Liquid Recovered from original nitro corn- Trinitrotoluene Nitroglycerin CS solution pound recovered 2.033 grams 1.886 grams .574 gram 28.2 1.236 grams 2.792 grams .340 gram 27.5 1.118 grams 4.625 grams .296 gram 26.5 Average percentage of nitro compound recovered, 27.4 per cent. Per cent of Dinitrotoluene Recovered from original nitro com- M.P.66-68 C. Nitroglycerin CSs solution pound recovered 1.077 grams 3.196 grams .425 gram 39.5 1.807 grams 1.472 grams .758 gram 41.9 1.606 grams 3.035 grams .647 gram 40.3 Average percentage of nitro compound recovered, 40.6 per cent. Per cent of Recovered from original nitro com- Paranitro toluene Nitroglycerin CS Z solution pound recovered 2.282 grams 1.444 grams 1.756 grams 77.0 1.008 grams 3.425 grams .788 gram 78.1 1.615 grams 3.074 grams 1.227 grams 76.0 Average percentage of nitro compound recovered, 77.0 per cent. rv] Congress of Applied Chemistry 75 Per cent of Dinitrotoluene Recovered from original nitro com- M.P. 48 C. Nitroglycerin CS solution pound recovered 2.192 grams 1.754 grams .893 gram 40.7 1.073 grams 3.752 grams .421 gram 39.2 1.504 grams 3.476 grams .584 gram 38.8 Average percentage of nitro compound recovered, 39.6 per cent. These results show that while no very great exactness can be claimed for the method as a quantitative one, still it is sufficiently accurate for practical purposes in many cases. The mixtures tried represent a rather wide range in the proportion of nitrogly- cerin to nitro compound, yet the error involved in the determination of a nitro compound by assuming the average percentage in each case would not be excessive for practical purposes. The average amount of each nitro compound recovered by the above treatment is given below. Orthonitrotoluene 75.0 per cent Liquid dinitrotoluene 35.7 per cent Liquid trinitrotoluene 27.4 per cent Dinitrotoluene M. P. 66-68C. 40.6 per cent Paranitrotoluene 77.0 per cent Dinitrotoluene M. P. 48 C. 39.6 per cent Assuming the above averages as the basis of calculation, a determination of the nitro compound in the 18 mixtures shown above would give percentages of the amount actually present as shown in the following table. The mixtures are numbered in the same order as given above. Mixture 1-101.8 per cent Mixture 10- 97.3 per cent Mixture 2- 98.5 per cent Mixture 11-103.2 per cent Mixture 3- 99.4 per cent Mixture 12- 99.2 per cent Mixture 4-102.2 per cent Mixture 13-100.0 per cent Mixture 5- 98.6 per cent Mixture 14-101.4 per cent Mixture 6- 99.4 per cent Mixture 15- 98.7 per cent Mixture 7-102.9 per cent Mixture 16-102.7 per cent Mixture 8-100.4 per cent Mixture 17- 98.9 per cent Mixture 9- 96.7 per cent Mixture 18- 97.9 per cent 76 Original Communications: Eighth International There are several errors involved in this method which prevent its being exactly quantitative. In the first place, the final product weighed is not the pure nitro-compound but contains in every case a few milligrams of nitroglycerin, the exact amount depending upon the quantity originally present in the mixture. The separa- tion could of course be carried further, but in that case the frac- tion of nitro-compound recovered would be decreased and all errors of manipulation would consequently be multiplied by a larger factor. Again, the solubilities of nitroglycerin and the nitro-compound in each other play a part which varies somewhat with different proportions. With the more volatile nitro-com- pounds there is also some error involved in evaporating off the solvent and drying in the desiccator. Finally there are the errors which are unavoidable in the separation of two liquids by means of a separating flask, such as solubility of one in the other, evapora- tion of the liquids during manipulation, etc. Keeping in mind these limitations, it seems probable that the method may have a fairly wide application. Qualitative tests upon the nitronaphthalenes show that it may be applied to mix- tures of these substances with nitroglycerin and probably to most other nitro-compounds. The great difference between the fractions of mononitro-compounds and dinitro-compounds recovered from the CS2 solution in the above mixtures also indicates that a separa- tion of these compounds might be possible by varying the relative proportions of the solvents. It must be remembered that the method is strictly empirical and the manipulations must be carried out in a definite manner in order to obtain results even approximately quantitative. Abstract HYDROLYSIS OF TRI-NITRO-ANISOL BY ALKALIES AND WATER BT WALTER E. MASLAND AND FIN SPARRB Wilmington, Delaware Very little information is to be found in chemical literature in regard to the stability of tri-nitro-anisol, the methyl ester of tri- nitro-phenol or picric acid, in the presence of alkalies and water. This question being of considerable importance in connection with the possible application of this material in the explosives industry, an investigation of the stability was undertaken. Pure tri-nitro-anisol was prepared by the nitration of mono- nitro-anisol with a strong mixture of sulphuric and nitric acids. The purity of the mono-and tri-nitro derivatives was proved by determinations of molecular weight, percentage of methoxyl radicals, etc. The following information was obtained from an investigation of the hydrolyzing action of water and alkalies: 1. Alkaline carbonate solutions, slowly in the cold, more rapidly when hot, hydrolyze the ester with formation of alkali picrates. 2. Pure hot water slowly hydrolyzes the ester with the forma- tion of picric acid. 3. Pure cold water appears to have the same but much weaker hydrolyzing action. The identity of the products of hydrolysis was shown by ana- lytical tests, molecular weight determinations, etc. 77 A PLEA FOR IMPROVEMENT IN THE METHODS OF CHEMICAL TESTING OF MINING EXPLOSIVES BY JAMES Mom, D.Sc., M.A. Chemical Laboratory of the Department of Mines, Union of South Africa (Johannesburg) The subject of the testing of explosives divides itself, on the chemical side, into (1) the stability or heat-test, and (2) proximate analysis. Of the known stability tests, I shall deal only with the official heat-test of the British Home Office, as being the only one capable of giving a result in a reasonable time which is a fundamental requirement of an analytical laboratory. This heat-test consists, as is well known, in observing the time required to produce a standard iodine-stain on iodised starch-paper (moistened in a thin line by 50% glycerol solution) by means of the traces of nitric peroxide evolved when explosives are heated to such temperatures as 70 to 80C. The test was invented (about the year 1875, I believe) to deal with nitro- glycerine and cordite, and acts well with these and similar explo- sives (i. e., with an experimental error of 5% or so); but my ex- perience of it with mining explosives, for which the test is prescribed by most of the Governments of the world, is quite unsatisfactory. The defects are of two kinds: (1) those due to the nature of the reagent, and (2) those caused by ingredients of the explosives tested or by impurities in the French-chalk used to reduce the sample to a powder. In regard to the first kind of defect, the main difficulty is the volatility of the iodine or the sensitiveness to heat of the coloured compound with starch, in consequence of which the results are liable to variation whenever slight changes of the depth of immersion in the water-bath, or of the precise position of the test-paper in the tube, happen to occur. The test itself is of course nowadays looked on as probably the least sensitive of all the known tests for nitrous fumes, and would deserve, were it not for an advantage to be discussed later on, to 79 80 Original Communications: Eighth International [VOL. be relegated to the oblivion of University lectures, and replaced by any of the good organic reagents for nitrous acid, such as indole, ra-phenylenediamine, or Ilosvay's reagent. Of the defects due to the nature of the explosives tested, it may be said that all are due to the presence of some volatile substance wh ch is capable of combining either with iodine or with nitric peroxide, and so spoiling the test. The best-known case is that of mercuric chloride, the presence of which, even in quantity so little as 1 in 50,000, almost entirely masks the action of the test-paper. Even water, however, when present to the extent of over 1% in gelignite for example, makes the test insensitive by 50% or so. This is due to the reaction between steam and NO 2 (H^O-f- 3N0 2 = 2HN03-{-NO) whereby the greater part of the nitrous fumes are rendered inert by oxidation to (dilute) nitric acid : (this occurs also in the cold, but to a smaller extent) . It may of course be said that it is easy to dry the explosive before heat-testing. This can be done in a sulphuric acid vacuum in 24 hours, if the sample is cut into thin slices. The desiccator must be kept in the dark, and not used too often, else fumes from ab- sorbed nitroglycerine vapour will vitiate the test. As a rule, how- ever, heat-tests are urgent work and cannot be postponed for a day or two. A more serious factor is the use of pinewood sawdust as absorbent in dynamites, etc. This on heating evolves terpenes which absorb any N0 2 that is being formed. Thus, a blasting-gelatine giving a heat-test of 15 minutes, will, on incorporation with some of this sawdust, give a test of 30 minutes as a rule. To my mind, there is no moral or legal distinction between this use of terpenes to mask the official test and the use of mercury which has already brought many dynamite-factories into conflict with their Govern- ments. While I am on this branch of the subject, I should like to say that I think that the official procedure for testing 'Dynamite F (Kieselguhr dynamite) by extraction with water, is perfectly useless. Nitroglycerine when properly washed, is a perfectly stable substance, and will give a very long heat-test; and so it does when extracted from these dynamite samples, even when iv] Congress of Applied Chemistry 81 the samples, tested directly as if they were ' 'Dynamite II," give very poor tests. I shall now mention a few factors which act in the opposite direction, and shorten the test-period. The first of these is ex- posure to bright light. This is very marked in such a high-lying and sunny country as South Africa; thus, one minute's exposure of an explosive to the sun will ruin its heat-test, and has even been known to lead to explosion. The second is the presence of very finely-divided carbon in grey specimens of French chalk. This acts catalytically on the explosive and lowers the heat-test. It is liable to be produced when commercial French chalk (which contains an objectionable scented impurity) is dried at too high a temperature. A third and more doubtful factor is low atmos- pheric pressure: certainly tests done in Johannesburg at 24J inches barometer seem to give shorter periods than tests on the same samples done at the sea-coast. This is hard to explain, since the reaction-velocity ought to depend on temperature alone. I have suggested that the official iodide test should be replaced by ore of the organic tests, but a strong factor in favour of the old test is the fact that its time of appearance is much more definite than that of the organic tests. This is due to a cyclic process involving the constant liberation of nitric oxide (NO) and its re-oxidation to N0 2 , so that a small quantity of nitrous fumes liberates a considerable quantity of iodine. The reactions are: (1) 2 N0 2 +H 2 (on wet part of paper) -HN0 3 +HNO 2 . (2) HNCM HN0 2 +KI = KN0 3 +H 2 0+NO+I. (3) NO+i0 2 = N02 (assuming excess of air). Thus half the original N0 2 is regenerated, and the whole process can therefore be represented by the single equation: (4) 2 NO.+2KI+0 2 = 2KN0 3 +I 2 . In consequence of this, the iodine colour deepens quite rapidly very soon after the first trace makes its appearance: whereas in the organic tests, the deepening of colour is progressive all through, and constant comparison with a standard would be necessary. I have however got good results with a test liquid made of equal parts of glycerol and Ilosvay's reagent (0.6 gram of sul- 82 Original Communications: Eighth International [VOL. phanilic acid and 0.5 gram of alphanaphthylamine dissolved together in 100 cc. of water with a few drops of acetic acid). This is approximately seven times more sensitive than KI towards nitrous fumes, and is not affected by mercury or terpenes or mois- ture. To bring it down to about the same sensitiveness as the official test, a lower temperature of the heat-test water-bath may be used (approximately 45C. instead of 71), and an arbitrary pink standard, similar to pink blotting-paper for example, may be chosen to work to. A solution of pure indole, say 1 in 1,000 of 50% glycerine, gives similar results. Guttmann's test is useless for moist explosives, because the green colour is not produced unless the sulphuric acid remains undiluted, which is not the case in practice. To summarise this branch of the subject, the official iodide test does not work well with mining explosives, except blasting- gelatine and Kieselguhr dynamite. The second division of my remarks deals with the question of proximate analysis of complex nitroglycerine explosives, wherein I have met with sources of serious error not mentioned in the current works on the subject. Take the case of that familiar explosive gelignite. This is a mixture of a 'thin' blasting-gelatine with a dope composed of fine sawdust and saltpetre. When the analysis is commenced (as usual) by drying in a vacuum over sulphuric acid, a trace of the nitroglyce ine is lost by volatilisa- tion unless one stops short of the stage of complete drying, namely by giving 24 hours for the process. In the second stage, viz., ether extraction by the Soxhlet or other method, I find that serious error is caused by extraction of fatty and resinous matter from the sawdust. This may be only say J% in the case of gelignite, but may be several parts per cent, in the case of 'Dynamite II.' The residue is a mixture of nitrocotton, sawdust and saltpetre, and may be worked up in either of two ways, both of which are subject to unavoidable error. The usual method is that o! boiling with water (after moistening with alcohol to cause the sample to sink). This is intended to remove saltpetre and soluble alkali, but also generally removes some soluble constituent (sugar, saponin, etc.) from the wood-pulp, so that a determination by difference is too high, even if correct allowance is made for any iv] Congress of Applied Chemistry 83 alkali after titration. My attempts to improve on this by pre- cipitation w'th 'nitron' (phenylimino-diphenyltriazole) gave results generally % too low; and in South Africa, where sodium nitrate is permitted in place of saltpetre, the evaporation methods are practically impossible. This method is finished off by extracting the nitrocotton with ether-alcohol or ether-acetone, and finally, the residue of the wood is examined for insoluble alkali (such as calcium-carbonate) by t'.tration. The alternative procedure is to remove the nitrocotton first by the above solvent-mixtures, and, as everyone knows, great manip- ulative difficulties are met with in filtering this viscous and volatile mixture. For special work where accurate values of the nitro- cotton content alone are required, I prefer to measure the bulk of the mixture (made by several hours' shaking in a closed and graduated vessel^, allow to settle, and decant and evaporate the upper half of the extract, allowing for the bulky insoluble at about ha ] f its apparent volume. The residue of sawdust and saltpetre can now be separated by cold water quite satisfactorily, whereas in presence of nitrocotton it is necessary to boil with water in order to get at the saltpetre. The use of sodium monosulphide for dissolving nitrocotton is only possible where wood-pulp is absent, since it apparently dissolves every constituent of wood except the cellulose, and even the latter absorbs sodium quite strongly. Another 'thorn in the flesh' of the explosives-analyst is the recent introduction of substitutes for part of the nitroglycerine in explosives, such as trinitrotoluene. Towards solvents these substances behave e: actly like nitroglycerine, and the only possible method of analysis appears to consist in evaporating the ether- extract and determining the amount of nitroglycerine by the nitrometer. (Trinitrotoluene does not react with sulphuric acid and mercury). Owing to volatilisation of nitroglycerine and other causes, this gives very poor results. Other methods of determining nitrogen are useless because both trinitrotoluene and nitroglycerine contain the same percentage of nitrogen. DETONATOR TROUBLES EXPERIENCED IN THE CONSTRUCTION OF THE ISTHMIAN CANAL BY ARTHUR LEE ROBINSON Gorgona, Panama Canal Zone A preliminary survey of the work to be performed in the con- struction of the Isthmian Canal would have led the majority of engineers to presuppose that the ordinary commercial products of the United States and Europe would meet all the requirements and necessities of such a service. Yet the actual construction of the Isthmian Canal has called for improvements in many classes of machinery and the special manufacture of many other products in order to meet the exigencies of this service. After the years of mining operations in all countries of the world it would be no more than natural to presume that the usual methods and materials employed at home and abroad for blasting purposes would fully answer all requirements of the apparently simple work of breaking ground for excavation by steam shovels on this Isthmian Canal. However, troubles were encountered with detonators of such a serious nature as to require changes in their structu e by manufacturers and also changes in the usual method of firing. These troubles originated from two distinct sources and will be discussed in this paper under the following heads: First, The Absorption of Moisture by Detonators; Second, The Nonuniformity of Detonator Current Explosive Values. THE ABSORPTION OF MOISTURE BY DETONATORS While the United States Government formally entered upon the work of Canal construction in May, 1904, comparatively little excavation was begun or accomplished during that year. Among the early requisitions for material and supplies was an order for 500,000 pounds of dynamite, which material after receipt was found sufficient to last through the years 1904 and 1905. In 1906, 1,400,000 pounds was purchased and used. Throughout the two and one-half years of operation in 1904, 85 86 Original Communications: Eighth International [VOL. 1905 and 1906 the construction work required comparatively light surface mining and very little heavy material was encoun- tered. Holes were drilled and blasted in small lots and no ex- plosive troubles of any serious moment were encountered. During the year 1907, mining operations assumed much larger proportions, about one hundred four-inch Star Well Drills and one hundred and fifty Tripod Drills being in operation. Drill holes were spaced on centers varying from twelve to eighteen feet and with depths of from fifteen to as high as sixty feet, blasting charges ranging from two hundred to as high as one thousand pounds of dynamite per hole, depending upon the class of material being mined. Five million eighty-seven thousand pounds of dynamite were ordered and used during this year. Serious troubles then began to develop from the encountering by steam shovels of unexploded dynamite. A number of shovels were damaged and a number of operators injured. It was im- mediately ascertained that this unexploded dynamite resulted from detonator failures and not from the quality of dynamite being used. Investigations were therefore started under the direction of Lieutenant-Colonel D. D. Gaillard, U. S. A., Corps of Engineers, and Division Engineer of the Central Division, along the line of determining the cause of detonator failures. The heavy charging of holes and the condition of work along the Canal often required detonators to be left in charged holes for twenty-four hours or more before blast was set off. A number of tests were immediately made upon the imperviousness of detonators to the waters encountered in these drill holes. De- tonators were left immersed under varying heads of waters taken from drill holes for from twenty-four to thirty-six hours and were then dissected to ascertain to what extent water had penetrated detonator shell. It was discovered that some of the brands of detonators then in use became thoroughly saturated and the ful- minate ruined after a twenty-four hour immersion. While other classes did not absorb as much water, nearly all were found to have become affected. Three different brands of detonators manufactured in the United States and one brand of detonator manufactured in Germany were tested under heads of from twenty-five to forty feet of water and none were found impervious iv] Congress of Applied Chemistry 87 after immersion from twenty-four to thirty-six hours. This discovery was most surprising in view of the fact that these detonators were the best products then obtainable in the com- mercial market. The only difference between the requirements of the work on the Isthmian Canal and that of ordinary mining work was the length of time which detonators were left in charged holes before being fired, and it is therefore presumed that the manufacturers had not heretofore been called upon to furnish a detonator which would stand immersion under a head of water for more than from two to three hours. As an example of the deterioration of one brand of detonator after immersion, the results of a series of tests are given in accompanying table, marked Exhibit "A." When it was discovered that detonators were not impervious to drill hole waters, it was assumed that these waters contained some free acid which attacked the detonator shells. This assump- tion seemed to have good foundation from the fact that a micro- scopic examination of detonator shells after immersion showed minute holes just below the sulphur plug. A number of these detonators and samples of the water from drill holes was then sent to the Chief Sanitary Officer at Ancon for a report as to what acids, if any, were causing the deterioration of detonator shells. Tests and examinations made in Ancon Laboratory showed that this shell deterioration was not due to free acids, but was due primarily to the composition and construction of the exploder it- self. A summation of the conclusion of these tests is as follows: 1. That the different kinds of water used for detonator im- mersion made little difference in the net results when the same conditions of time of immersion and pressure were maintained. 2. That the number of failures to get a normal explosion of cap increased almost geometrically with the length of time of immersion and the hydrostatic pressure to which submitted. 3. That corrosion began on the internal portion of shell at or near the point of contact between sulphur plug and the loosely packed portion of the charge usually opposite or a little above the level of the platinum bridge. A full copy of the report of Ancon Laboratory is herewith attached, marked Exhibit "B." 88 Original Communications: Eighth International [VOL. Detonator manufacturers were notified of the troubles encoun- tered, and two of these manufacturers later developed and fur- nished to the Commission detonators which, after test, proved practically impervious to water under heads of forty to sixty feet. In one brand of these detonators a composition or cement plug is inserted in shell just above the loose fulminate and at the point !v > where shell deterioration had occurred. Above this composition plug a black viscous compound is inserted and above this compound the usual sulphur plug is put in as a cap. If this sulphur plug causes any deterioration of shell no leakage through deterioration at this point can reach the fulminate. The other detonator is composed of a double jacket or shell. The inner shell contains the usual packed fulminate at the bottom, a small quantity of gun cotton, the bridge, and the usual sulphur plug at the top. This inner shell is set into an outer shell, filled with a black compound resembling fluid tar which completely envelops the inner or explosive shell. If inner shell is deteriorated or perforated by sulphur action, the outer shell and compound will prevent any moisture reaching fulminate. The detailed con- struction of these two classes of shells is shown in attached sketch, marked Exhibit "C." With impervious detonators it was assumed that accidents from unexploded dynamite would become matters of history. However, in the spring of 1908 the number of missed holes and the consequent amount of unexploded dynamite dug up by steam shovels reached such proportions as to call for a careful study of all conditions which might in any way cause misfires. THE NON-UNIFORMITY OF DETONATOR CURRENT EXPLOSIVE VALUES With impervious detonators in service, and the knowledge that the dynamite being used was of the best grade obtainable, it was then decided to investigate the power for exploding de- tonators. While these investigations were being carried on, the necessity for reducing the number of missed holes caused the issuance of instructions to reduce the spacing of drill holes to thirteen feet as a maximum so as to thus utilize more largely the explosion iv] Congress of Applied Chemistry 89 of one hole to detonate surrounding holes (in which detonator had misfired) by either sympathetic explosion or by means of the hot gases from exploded holes penetrating the fissures to ad- joining holes. While the issuance of the above orders largely assisted in reducing the amounts of unexploded dynamite, it must be understood that results obtained from explosives de- tonated by surrounding holes, instead of by their own detonator, are very poor, in that the holes detonating first open up fissures and cracks which prevent the full force of the second detonation being obtained. In this connection, the attention of this Congress is invited to the fact that many detonator failures in ordinary mining operations are never discovered, because adjoining; holes explode the charge of the defective detonator. Until the spring of 1908 all blasts in the Canal excavation were discharged by means of hand batteries operating from ten to as high as seventy detonators wired in series. These batteries had been furnished the Isthmian Canal Commission in various types and sizes, marked and guaranteed as being good to explode from ten to one hundred detonators. The manufacturers furnish- ing these batteries also furnished small rheostats which contained resistances marked for field purposes as representing specified numbers of caps. For example: If a battery was marked good for seventy detonators, one of these rheostats, set at a resistance equal to seventy detonators with thirty foot leads, could be placed in circuit with the battery and one detonator. The battery would then be operated and if detonator exploded, battery was con- sidered to be good for the explosion of seventy detonators, in that it had fired one detonator with current passing through a resistance equal to the other sixty-nine or seventy detonators. It was the practice to test all of these batteries with these rheostats before putting them into service, and while a few were found defective, the majority were found fully capable of exploding the one de- tonator through the resistance of the full number of detonators for which battery was marked good. A short description of these batteries (commonly used in all blasting operations) is herewith considered apropos for the benefit of those not familiar with their construction. The common type of these batteries is a small series of com- 90 Original Communications: Eighth International [VOL. pound series wound generator, located on a horizontal plane as a base, in a rectangular wooden box. A vertical rod projects through the top of box, one side of which rod has been cut into teeth forming a rack. A train of gears operating between this rack and generator revolves the armature of generator as this rack rod is pushed down into box. The rate of armature revolu- tion depends entirely upon the speed with which the rack rod is moved. The rack rod upon reaching the bottom of box breaks one contact and connects another which throws the generator in series with the external circuit (firing wires and detonators). From this construction it is readily understood that with the beginning of the push down movement of rack rod, the generator voltage builds up in the form of a curve, reaching its maximum when rack rod reaches the bottom of box and throws external circuit into series with generator. It is also readily seen that the power applied to the generator by the operator ceases at the instant this rack rod reaches the bottom of box, and thus at the same in- stant that firing circuit is connected, and from that time on, the gen- erator's only source of power is that derived from the momentum of the revolving armature. The voltage characteristic of blasting batteries is, therefore, a curve, which reaches a maximum at the instant that blasting circuit is connected to generator, the peak of which curve could be nothing more than a point, in that voltage immediately begins to drop with the connection of the blasting circuit. We, therefore, obtain from one of these blasting batteries only a pulsation of current lasting through a fractional part of a second. It is important that the brevity of this current pulsa- tion be thoroughly appreciated in order to understand the weak- ness and uncertainty of blasting batteries in exploding detonators. We have on our construction work no instruments delicate enough to accurately measure and record the voltage curve of one of these batteries, but a study of such a curve would be extremely interest- ing, and would, I am sure, prove the conclusions reached through practical experiments and from crude experimental apparatus. After ascertaining that the field batteries were sufficiently power- ful to explode one detonator through a resistance equal to the num- ber of detonators for which battery was guaranteed, it occurred to the writer that our detonator troubles were probably due to the series iv] Congress of Applied Chemistry 91 method of connecting these detonators in circuit. With only an instantaneous pulsation of current the explosion of a large number of detonators must be impossible unless said detonators are so uniform in construction as to allow of their bridge being heated to the explosive point by the same amount of current with the same time element. A study was then determined upon and made of the two vari- ables entering into the heating limits of these detonator bridges. It was quickly ascertained: 1. That with the time element of current flow constant, detonators exploded with variable amounts of current. For example: In testing twenty-five detonators, picked at random from stock, it was found that with the time element constant the current value for exploding these detonators varied from forty-four one-hundredths to seventy-six one-hundredths of an ampere. 2. That with current flow constant, detonators exploded under variable time elements. For example: It was found from tests that with a constant current value of five-tenths of an ampere and with variable time element detonators exploded with above amount of current flow through them in from two one-hundreths to seven one-hundredths of a second. With this knowledge it was readily understood why a large number of detonators wired in series could not be counted upon to explode from a momentary current from a blasting battery, even though that battery had fired one detonator through a resistance equal to the number of detonators for which battery was guaranteed. With fifty detonators wired in series and requiring an exploding current value varying from forty-four one-hundredths to seventy- six one-hundredths of an ampere, that detonator requiring the lowest current value to explode would naturally be the first to break down, thus breaking the circuit and leaving detonators requiring high current value unexploded. The apparatus used in obtaining the above current and time values is fully described in an attached appendix, marked Ex- hibit "D." This apparatus was the best that could be obtained on the Isthmus, and while it may appear crude from a scientific 92 Original Communications: Eighth International [VOL. standpoint, it was considered and proved to be accurate enough to demonstrate by comparative results the non-uniformity of the detonators tested. The results also unquestionably demonstrated the impracti- cability of avoiding missed holes when using a large number of detonators wired in series, no matter what or how powerful the source of current supply. In order, however, to verify the conclusions that missed holes were sure to be encountered where detonators were wired in series, a number of tests were made with one hundred detonators con- nected in series with a fifteen kilowatt transformer with voltage of 110, 220 and 440. In no instance were we able to explode the entire one hundred detonators with the first closing of switch, although in every instance it was found that there were no de- fective detonators in circuit. For example : In testing one hundred detonators we would explode from ninety to ninety-five upon the first closing of switch. Those which did not explode would be re-connected in circuit with transforner and would all explode but one or two upon the second closing of switch. The opening of unexploded detonator showed same to be double bridged. We thus obtained conclusive evidence that double-bridged de- tonators or those requiring high current values for exploding could not be counted upon to detonate when wired in series with detonators requiring low current values for exploding. A number of experiments were then made by wiring detonators in parallel with source of current from the same transformer. In every case where detonators were wired in parallel, all exploded at the first closing of switch. While without question those de- tonators requiring the lowest current value exploded first, the time element between the explosion of detonators was inappreciable to either sight or hearing of investigator standing within two hundred feet of the detonators, which were laid upon the open ground. The results of experiments caused orders for the immediate construction of some eight miles of primary pole line along the banks of the Canal between Bas Obispo and Pedro Miguel, the installation of transformers at every thousand feet along this pole line, and the stringing of secondary wire from transformers iv] Congress of Applied Chemistry 93 throughout the blasting districts in this eight miles of territory. The use of blasting batteries was immediately prohibited throughout this section, except for the springing of holes and for such other light work as would not result in accidents in case detonators failed to explode. In all heavy mining operations detonators were connected in parallel to feed wires of number Twelve B. and S. gauge, strung on stakes over the holes to be exploded. These feed wires were then connected by blasting wiremen to the secondaries of number Two Brown and Sharpe gauge wire which had been strung from transformers along the ground on portable trestles. Some little time was required to initiate and educate the mining forces into the practice of parallel wiring and firing of detonators from power line. Some prejudice was encountered on the part of old miners who had been accustomed throughout years of experience to the use of blasting batteries and series wiring of detonators. A number of these miners insisted that they had never had any missed holes when using blasting batteries and offered to wager the writer that they could fire with their own particular blasting batteries any number of detonators (up to the battery capacity) wired in series. In order to convince these men of their error in judgment, the writer allowed a number to demonstrate to him personally the handling of blasting battery with detonators wired in series. In only one instance did the battery (marked good for fifty holes) explode ten detonators connected in series. The surprise and consternation of the operators was remarkable and can only be accounted for by the fact that throughout their mining experience they had been using blasting batteries for the explosion of detonators in holes close enough together to cause the explosion of one hole by another, thus giving the miner the impression that all detonators had fired. The writer is of the opinion that not more than ninety-five per cent of detonators connected in series have ever exploded, no matter what source of current had been utilized. The installation of power lines and the adoption of the parallel wiring and firing of detonators produced such an immediate reduction of missed holes and increased efficiency of explosives that this method was adopted throughout practically the entire Canal construction. 94 Original Communications: Eighth International [VOL. The author believes, and our experience seems to verify the conclusion, that it is impracticable to manufacture for commercial purposes detonators which would be so uniform in the required current value and time element for explosion as to make the use of blasting batteries and series wiring (regardless of the magnitude of power source) as safe and sure as the parallel wiring in which non-uniform and double-bridged detonators are sure to explode. The tests made on the Isthmus and the practical results obtained from the parallel wiring and firing of detonators confirm most positively the conclusions that blasting batteries and series wiring of detonators is an unsafe and questionable practice. It would be extremely hard to estimate the amount of money which the Isthmian Canal Commission has saved by the change from series to parallel wiring and firing of detonators. The unexploded holes due to detonators misfiring caused large losses by ground being insufficiently broken and the consequent result of delays incident to "dobeying," to say nothing of the extra amount of dynamite used. In addition, a number of steam shovel cranemen were hurt by digging into unexploded dynamite, which resulted in a timidity on the part of steam shovel operators, which largely reduced their output. The firing of detonators wired in parallel naturally requires a large current output and consequently a much more expensive current source than the ordinary hand battery. The magnitude of the work in hand must determine the advisability of making the necessary installation for parallel wiring and firing of detonators. If the work is of sufficient magnitude to warrant such an ex- penditure, the danger to operators is greatly reduced, the efficiency of the explosives is largely increased, and the general results obtained are much more economical and satisfactory. EXHIBIT "A" Test No. 1 of Fifteen 35-foot Electric Fuses Fuses immersed together in 25-foot well drill hole having 19 feet water in hole. Fuses remained in hole June 2, 5 P.M. to June 4, 7 A.M. 38 hours in water. Connected 15 and pumped battery. Trial No. 114 exploded; 1 failed. Trial No. 21 exploded; none failed. iv] Congress of Applied Chemistry 95 Test No. 2 of Sixteen 16-foot Electric Fuses Fuses made up into primers to spring 25-foot holes having 10 to 20 feet of water in each. Fuses remained in hole June 3, 3 P.M. to June 4, 8 A.M. 17 hours. Connected sixteen and pumped battery failed. Tried singly, second failed ; fourth and fifth exploded together from proximity; tenth, eleventh and twelfth exploded together from proximity and burned dynamite in hole making smoke; fourteenth failed. Judging from appear- ances, exploders did not properly detonate dynamite in many instances. Test No. 3 of Sixteen 20-foot Electric Fuses Fuses immersed as in Test No. 2 and tried in same manner singly after all failed connected. Third failed completely, when opened, found wet fulminate inside. Test No. 4 of Eighteen 2 4-foot Electric Fuses Fuses immersed as in tests Nos. 2 and 3 and tried connected and failed, then singly to get results. Fifth, sixth and seventh, eighth and ninth holes exploded by proximity. One failed com- pletely. Test No. 6 of Fifteen 30-foot Electric Fuses Fuses immersed together in 25-foot well drill holes having 20 feet water in hole. Fuses remained in hole June 2, 5 P.M. to June 5, 7 A.M. 62 hours in water. Connected 15 and pumped battery, failed. Tried 5, exploded; tried 5, failed; tried 5, 3 ex- ploded. Opened two, found wet fulminate. Single No. 1 exploded ; No. 2 exploded and burned; No. 3 like "nigger chaser"; No. 4 failed completely; No. 5 burned like a match. N.B. After as many as 15 times having battery pumped, fulminate could be made to burn ineffectively. Test No. 5 of Ten 35-foot Electric Fuses Fuses made into primers of one stick dynamite each immersed in 25-foot well drill holes having depth 18 feet water. Fuses remained in hole June 2, 5 P.M. to June 5, 7 A.M. 62 hours. Con- nected 7, 2 on surface of ground, 5 in holes. Result, all had no good effect to detonate, the two on surface having only 96 Original Communications: Eighth International [VOL. blown dynamite into pieces. Connected three in same hole and pulled battery which exploded with poor effect, one on being pulled out showed low explosive effect, blowing hole in side of copper shell only. EXHIBIT "B" BOAKD OF HEALTH LABORATORY, ANCON HOSPITAL, May 21, 1908. To the Superintendent, Ancon Hospital. Sir: In compliance with the request of the Chief Sanitary Officer I have the honor to render the following report: Chemical No. 174. Report on an investigation into the cause of corrosion of copper shells of electric exploders and failure to detonate the charge of Powder. I. Waters from Test Blast Holes. Both samples of water submitted showed practically the same amount of alkalinity and sulphur trioxide (S0 3 ) as follows: Alkalinity 204-212. Sulphur Trioxide 0.00943 % (per cc.). The alkalinity is nearly twice as much as that of the highest of the reservoir waters during the height of the dry season, namely, Gorgona, which was 114. The sulphur is not present as free sulphuric acid, but in combina- tion with the bases present which is clearly shown by the high amount of alkalinity. From this fact it may be inferred that the corrosive action of the sulphuric acid is largely neutralized by its presence in a state of combination with the bases. This low direct corrosive effect of the water from blast holes is also clearly shown by comparative experiments which I have made with distilled water and Ancon Tap water. II. Cause of Corrosion of Copper Shells of Electric Exploders. The results of a series of experiments under a wide variety of conditions prove fairly conclusively that the cause of the corrosion is internal and due primarily to the composition and construction of the exploder itself. The experiments (briefly stated) were carried out as follows: The several types of exploders submitted were soaked in distilled water and water from test blast holes under a head of about one foot, some of which were bared of the paraffine or asphaltum like iv] Congress of Applied Chemistry 97 covering where the wires enter the copper shell, while the remainder were left in the normal condition. Following this series another set of the exploders with the ends treated in the same manner were immersed in a section of three inch gas-pipe filled with Ancon Tap water at depths of about twelve and twenty feet, depending on the length of the fuse wires. Tests of 85 fuses of the different types (mostly of the older forms of exploder) by the Magneto exploding battery, showed: 1. That the different kinds of water used made little difference in the net result when the same conditions of time of immersion and pressure were maintained. 2. That the number of failures to get a normal explosion of the cap increased, I might say, almost geometrically with the length of time of immersion and the hydrostatic pressure to which submitted (particularly the latter). A detailed examination of those exploders which failed to give a normal explosion was made and the following results were deduced and verified in most cases by direct experiment. 1. The corrosion occurs at or near the point of contact between the sulphur plug and the loosely packed portion of the charge, usually opposite or a little above the level of the platinum bridge. 2. An examination of the interior surface of the copper shell, both above and below the point of corrosion, shows that the copper is more or less corroded wherever the sulphur comes in contact with it and an amalgam of metallic mercury occurs in the neighbor- hood where the break in the shell occurs, often on the outside as well as the inside of the shell. Small globules of mercury may often be seen scattered all over that portion of the shell which comes in contact with the loosely packed upper and darker colored crystalline-like portion of the charge, which I take to be the fulminate. Finally there seems to be little or no corrosion in the lower part of the shell where the lighter colored compressed powder form of the charge is located. This latter portion of the charge occasionally failed to be detonated, although the upper portion of the charge had exploded. (This lower compressed portion of the charge is said to be gun- cotton, and its general appearance and character agree with this statement.) 98 Original Communications: Eighth International [VOL. III. Cause of Failure to Detonate the Charge of Powder. Taking into consideration all of the foregoing experimental facts and data, I believe that the prime cause of failure to detonate the powder is due to the entrance of moisture (principally when subjected to pressureomder water) into that portion of the charge in the vicinity of the fuse. In the majority of cases this moisture gets in directly through the wall of the shell in the immediate vicinity of the fuse. However, it may occasionally get in from the end where the wires enter, since defects in the copper shell and sulphur plug have been found sufficient to allow of the entrance of water, especially when under pressure. The entrance of water into one make of exploder, in which there is a considerable thickness of rather porous paper between the sulphur plug and interior surface of the copper shell, probably takes place by means of this bibulous layer of paper from the end where the wires enter. The three of this variety which were submitted, all of which had failed to explode (one in a blast hole and the other two under experimental conditions), bear out this statement. The copper shells also show little corrosion in the neighborhood of the fuse and upper portion of charge and practically none above the fuse since the sulphur does not come into direct contact with the copper. The fact that the sulphur is put in the shell in the melted con- dition, accounts for its marked corrosive action everywhere it comes in direct contact with the copper, and also the hot sulphur coming in contact with the fulminate in the vicinity of the fuse probably accounts for the setting free of mercury which, together with the nitrogen compounds formed at the same time, sulphide of copper previously formed, and also a possible sulphide of mer- cury, furnish a sufficient supply of corrosives to further eat away the copper shell. Keeping these facts and conditions in mind, one can readily see .how length of time of storage, length of time of immersion and hydrostatic pressure would effect the deterioration and failure of the exploder to do the work for which it was intended. I have also discovered marked corrosion of the bare copper wires just above the bridge and even loosening of the platinum bridge which might account for the failure of same, but this I think is of minor importance. iv] Congress of Applied Chemistry 99 The recent modification of another make of exploder also shows the same process at work on the interior, but modified by the plug (also containing sulphur) existing between the sulphur fillings and charge, and the thick asphaltum-like exterior coating. Only four of the new form were submitted; one of the first two, bared of coating where wires enter and soaked in one foot of dis- tilled water for forty-eight hours, exploded only with a very feeble report. One of the second two bared at the end and the other not were soaked in twenty feet of tap water for forty-eight hours, and, when taken out, the ends of both were broken off at the usual place, and the same conditions found to exist as in the case of the older variety. It was also noted that in handling this variety the wires at point of entrance are very apt to be loosened from the asphaltum-like protection and thus afford another point for water to gain entrance along the insulation of the wires, and the sulphur filling. I have also found that if sulphur is simply melted in an empty copper shell, upon cooling the shell is often found to contain very fine cracks, usually running lengthwise, due presumably to the sudden expansion of the sulphur upon solidifying. Respectfully, (Signed) R. W. NAUSS, M.D., Chemist. (Signed) G. H. WHIPPLB, Acting Chief of Laboratory. W.... EXHIBIT " D " Description of Apparatus employed in obtaining comparative values of current and time element required for exploding Detonators. While the apparatus described below may appear crude and unreliable in obtaining accurate data, it must be borne in mind that the only object of tests was to ascertain values, which, by comparison, would show whether detonators required a uniform current and time element for their explosion. Comparing current explosive values with time element constant. For this purpose current was taken from a 15 K.W.-110 Volt, B.C. Generator, which generator was kept operating at a constant n Congress of Applied Chemistry 103 speed and constant voltage on a uniform load. Current for test purposes was passed through a bank of lamps, then through ammeter and rheostat to a double pole double throw switch. With this double pole switch in one position, rheostat was set so as to give the desired amount of current, usually beginning with 30/100 ampere. A piece of hard fiber with a copper insert as shown in diagram was mounted upon a table. A piece of spring copper connected with coil spring was pivoted on this fiber. One of the terminals of double throw switch as shown in diagram was connected to the dethc&jitor to be tested. Other side of detona- tor was connected ^ the strip of spring copper. The copper insert in fiber was connected with the other terminal of double throw switch. After rheostat had been set for 30/100 ampere, the spring copper or switch blade was pulled back across copper insert and holding plug inserted. Double throw switch was then placed in position to connect detonator in circuit. Holding plug was then removed, which allowed the blade of spring copper to make a wiping contact across the copper insert for only such length of time as was required for this blade to pass over the copper insert. If detonator under test did not explode, double throw switch was then placed in such position as to throw detonator out of circuit and rheostat was adjusted so as to give 32/100 ampere. Blade of detonator switch was again pulled back and held by plug. Double throw switch was then placed in position to connect detonator in circuit and holding plug removed, allowing blade to again make a wiping contact across copper insert. Each detonator was continued under test by the above method with current values being in- creased by increments of 2/100 ampere until detonator finally exploded. With the same coil spring in service and the spring blade drawn back to the point of holding plug at each test, it was assumed that the time element for current flow would be nearly enough constant for all practical and comparative purposes. As stated in the body of this paper, detonator current explosive values varied from 44/100 to 76/100 ampere. Comparing time element explosive values with current constant. The devising of some method by which detonators could be connected in circuit for 1, 2, or 3/100 of a second, required som 104 Original Communications: Eighth International expediency in that we are working on a construction job with no refined instruments or testing equipment on hand. A 150 H.P. Ball Engine operating a 100 K.W. 3-Phase 60-Cycle Generator was brought into service. This engine was put under constant load and the speed held at 120 revolutions per minute. Tests of speed by tachometer showed variation of not more than one revolution per minute. The fly-wheel of this engine was covered with a shellacked canvas, which canvas was then cut out in a form as shown in attached diagram. A stand carrying a flat copper brush was then mounted in front of fly-w^dd so that brush could be given a wiping contact on face of fly-wheel over any of the seven spaces cut in canvas. With brush bearing on face of fly- wheel at right side, the amount of bare iron over which brush would pass (four and one-half inches) represented a time element of 1/100 of a second. With brush removed to second space and wiping over nine inches of the bare face of fly-wheel, we obtained a time element of 2/100 of a second. With brush bearing on third space and wiping over 13J inches of bare face of fly-wheel, we obtained a time element of 3/100 of a second. With the seven spaces on fly-wheel we were able to obtain time elements of 1, 2, 3, 4, 5, 6, and 7/100 second. Current for testing detonators was obtained from the same source as in other tests, passing through bank of lamps, ammeter and rheostat to double pole switch. The lower set of terminals of this double pole switch were con- nected through detonator, brush and engine shaft. The current value was set at 5/10 ampere, and the time element then varied from 1/100 second up until detonator exploded The brush would be released and allowed to press against fly-wheel just long enough to insure having passed over the bare face of wheel two or three times or during two or three revolutions of fly-wheel. Under these tests detonators were found to explode with 5/10 ampere and with time elements varying from 2 to 14/100 second. These detonators which did not explode under time elements of 7/100 second were set to one side and later tested with engine running at 60 revolutions, under which conditions each space on face of fly-wheel represented two instead of 1/100 second, giving us a time element up to 14/100 second. IMPROVED DENSIMETER BY WALTER O. SNELLING Pittsburgh, Pa. The specific gravity of gunpowder, black blasting powder, and similar explosives, is usually expressed as " gravimetric density," and represents the apparent specific gravity of a selected volume of explosive, the space between the grains of powder not being allowed for. In the study of ballistics, and in comparison of the work done by varying grades of black blasting powder, even the rather meager information which is furnished by the " gravi- metric density" of a sample is of considerable value in compari- son between powders, and several types of apparatus have been designed to enable the gravimetric density of a powder to be readily determined. The determination is in any case a simple matter, and practically consists of nothing more than the deter- mination of the weight of powder required to fill even full a se- lected vessel of known volume. The true or absolute density of black powder or other similar type of explosive has long been recognized to be a decidedly more important factor than the apparent specific gravity or "gravi- metric density." It is well known that exact comparison be- tween two explosives can only be made on the basis of their true density, since two powders of widely different density can have the same apparent specific gravity where the denser powder is in large, angular grains and the lighter powder is in small and well- rounded grains. In such a case the small and rounded grains pack so much closer than the larger angular grains that even a considerable divergence in true density may be entirely without effect in producing a difference in the apparent specific gravity. Owing to the solubility in water of the nitrates present, the usual picnometer method of determining density has to be modi- fied for use with black powder and similar explosives, and several types of densimeter have been devised for such use. Of these, densimeters using mercury to replace the air filling the spaces 105 106 Original Communications: Eighth International in and around the grains of powder have been found to be the most satisfactory (densimeter of Bianchi, 1 Ricq, 2 Bode, 3 Michel- son, 4 Hoffmann, 5 Marchand 6 ), but other types of densimeter are also well known in which the volume occupied by the powder is measured by a change in volume or pressure of air or other gas present in a vessel with the powder, application being made of Boyle's law to estimate the portion of the known volume of the reservoir which is filled with powder (Holecek volumeter, 7 Say stereometer, 8 Kopp volumeter 8 ). The density of powder is also sometimes determined in the ordinary picnometer bottle, making use of absolute alcohol in place of water. Absolute alcohol is not without effect upon the constituents of gunpowder, and the re- sults obtained by this method are not as accurate as those reached in an apparatus using mercury. The method has, however, the advantage of not requiring any special apparatus. In the study of black powder and black blasting powder by the Bureau of Mines, it was found desirable to distinguish powders on the basis of their true or absolute density, and a comparison of the many types of apparatus which have been devised for this purpose indicated the possibility of arranging apparatus of simpler construction, which would equally well fulfill the purpose of determining real or absolute density of black powder. The im- proved densimeter, devised by the writer, is to be considered as only a variant of the many devices which have been previously used in the determination of absolute density. In the new densimeter, advantage is taken of the production of a Torricellian vacuum within the apparatus itself, thus entirely eliminating the use of a separate air pump, and the device as x Kast, Anleitung zur Chem. und Phys. Untersuchung der Spreng-und Zund-stoffe, page 1001. 2 Guttmann, The manufacture of explosives, page 306. 8 Guttmann, The manufacture of explosives, page 304. 4 Ann. Kept. Secy. U. S. Navy, 1879, page 70. 6 Guttmann, The manufacture of explosives, page 299. "Guttmann, The manufacture of explosives, page 297. 7 Kast, Anleitung zur Chem. und Phys. Untersuchung der Spreng-und Zund- stoffe, page 1002. 8 Guttmann, The manufacture of explosives, page 299. 'M- 110 Original Communications: Eighth International [VOL. constructed has been found to fully answer the many require- ments which have been met with in the practical examination of blasting powder, and has proved to be an apparatus possessing not only a high degree of accuracy, but also of extremely simple operation. In Figure 1 a vertical section of the apparatus is shown, (a) representing the vessel within which the Torricellian vacuum is produced, (b) the cover of this vessel, (c) a valve placed in the top of the cover, to allow exit of air entrapped within the reservoir. The flange around the reservoir (a), as well as the flange upon the cover (6), is to provide a mercury seal covering such parts of the apparatus as are open to the outside air. (d) is an iron pipe, which must be about 1 meter in length, leading from the vessel (a) to the mercury reservoir (e). The reservoir (e) is of iron, completely closed except for the two pipes (d) and (/). The pipe (f) is connected to a source of water under pressure at valve (w), and an outlet pipe at valve (x). (i) is a framework designed to hold the picnometer bottle or other suitable vessel (h). To determine the absolute density of a sample of powder, the picnometer bottle (h) is partly or wholly filled with the material whose density is to be determined, and is then placed firmly in the clamp or framework (i). The framework and picnometer bottle are not placed in the reservoir (a), taking the position shown in the drawing. The frame and bottle are lighter than mercury, but are held in proper position in the mercury in the reservoir by means of the small projection shown, set in the wall of the vessel (a). The cover (6) is placed upon the reservoir, and is screwed down. The valve (x) being first opened quite wide, valve (w) is opened until there is a steady flow of water from the supply pipe through the outlet pipe (x). Valve (x) is now closed, and as the outlet pipe is thus cut off, the water is forced into the res- ervoir (e) and raises the mercury through the pipe (d). The valve (c) being open, mercury is allowed to flow in until the res- ervoir (a) is completely full, and until the mercury has also over- flowed so as to fill the mercury seals around the cover (6) and around the valve (c). When this condition has been reached valve (z) is opened, and valve (c) is closed, and as the pipe (d) is of greater vertical height than 760 mm., the mercury in the IV] Congress of Applied Chemistry 111 FIGURE 4 upper part of the reservoir (a) flows out through the pipe (d) and a partial vacuum is produced in the upper portion of the res- ervoir (a). The air surrounding the grains of powder in the pic- nometer (h) rises into the evacuated space, and at the end of a few minutes has been largely entrapped in the conical space in the cover (b) . The valve (x) is now closed, the valve (c) is opened, and because of the pressure within the apparatus being greater than the atmospheric pressure, owing to the pressure of the water, the bubbles of air entrapped rise through the mercury seal to 112 Original Communications: Eighth International [VOL. the upper part of the cover (6) . Valve (c) is closed as soon as the air has been completely driven out, and valve (x) is opened, and this operation is repeated until no further bubbles of air rise through the mercury seal when the valve (x) is closed and the valve (c) is opened. When this condition has been reached it will be found that all the air in the picnometer bottle (h) has been replaced by mercury, and upon now taking off the cover of the apparatus, removing the picnometer bottle (h) from the framework and weighing it, a weight is obtained which represents: Wt. of bottle (h) + wt. of powder taken + wt. of mercury filling all portions of the bottle not filled with powder. Since the weight of the empty bottle, and the weight of the bottle completely filled with mercury, form the known constants of the apparatus, being determined with accuracy when the appara- tus is first used, and as the specific gravity of mercury is also known, it is evident that there is now available all the information required to calculate the density of the powder under examination. In order that the method of calculation may be fully understood, an illustrative example is here given: Weight of picnometer 18.4 grams Weight of picnometer full of mercury 1375.8 grams Weight of mercury required to completely fill the picnometer 1357.4 grams Weight of picnometer plus sample of powder taken 97.6 grams Weight of picnometer 18.4 grams Weight of sample of powder taken 79.2 grams Weight of picnometer plus powder plus mercury re- placing air spaces 886.1 grams Weight of picnometer plus powder 97.6 grams Weight of mercury replacing all spaces in picnometer 788.5 grams The mercury displaced by powder equals 1357.4 grams minus 788.5 grams, or 568.9 grams mercury which would be required iv] Congress of Applied Chemistry 113 to occupy the same volume as the powder contained within the picnometer. This amount, 568.9 grams, divided by 13.55, the specific gravity of mercury, gives 41.98 as the volume in cc. of the sample of powder taken. Since the weight of this amount of pow- der is 79.2 grams, the density of the powder must be 79.2 divided by 41.98, or 1.886, absolute density. It is of course evident that when determinations are made at other than standard conditions of temperature correction for the decreased density of mercury due to expansion must be made, by applying the customary formula. The powder in this particular experiment was of the granu- lation known as FFF. From this example it will be understood that the operation of determining the absolute density of powder by the densimeter consists essentially in determining the weight of mercury required to completely fill a picnometer bottle, and then determining the amount of mercury which corresponds to the volume occupied by the sample of powder taken. From this data, with the known specific gravity of mercury, the absolute density of the powder is at once obtained. The gravimetric density of the powder in the above sample, as determined by weighing a given volume of the powder, poured into a vessel of the standard size and shape, was 1.176. The absolute density, 1.866, is thus seen to differ greatly from the ordinary or gravimetric density. The per cent of open spaces between the grains of powder in this particular sample is found from the above data to be 35.16 per cent of the entire volume taken up by the powder. The following results, representing the study of 29 samples of black blasting powder, collected from one of the coal-mining fields in which black blasting powder is still largely used, give a good idea of the usefulness of a comparison of the apparent and true density of black powder. These 29 samples were all differ- ent, and represented the product of a number of manufacturers. The table shows the gravimetric density and the absolute density of each of the samples, and the per cent of voids. 114 Original Communications : Eighth International [VOL. Improved densimeter Sample Absolute Gravimetric Per cent of number density density voids 1 1.791 1.15 35.79 2 1.775 1.18 33.52 3 1.859 1.17 37.06 4 1.779 1.13 36.48 5 1.749 1.13 35.39 6 1.764 1.15 34.81 7 1.783 1.18 33.82 8 1.764 1.19 32.54 9 1.744 1.17 32.91 10 1.859 1.17 37.06 11 1.775 1.17 34.08 12 1.789 1.16 35.16 13 1.857 1.20 35.38 14 1.766 1.18 33.18 15 1.744 1.14 34.63 16 1.830 1.15 37.16 17 1.865 1.20 35.66 18 1.759 1.18 32.92 19 1.761 1.17 33.56 20 1.783 1.17 34.38 21 1.739 1.16 33.29 22 1.731 1.09 37.03 23 1.879 1.18 37.20 24 1.851 1.15 37.87 25 1.772 1.16 34.54 26 1.775 1.15 35.21 27 1.745 1.16 33.52 28 1.798 1.19 33.82 29 1.855 1.18 36.39 This table shows clearly how unreliable the gravimetric den- sity of an explosive is, as a measure of its true density. Explo- sive No. 3, one of the explosives whose true density was very high (1.859), shows exactly the same gravimetric density (1.17) as does sample 9, whose true density is one of the lowest of the rv] Congress of Applied Chemistry 115 29 examined. A still more marked variation is that shown between samples 21 and 23. Although the gravimetric density of these two samples differs by but .02, the absolute density of the two samples is represented by the difference between 1.879 in the case of sample 23 and 1.739 in the case of sample 21. The importance of determinations .of absolute density has long been recognized in connection with the study of military powders, and the usefulness of determinations of absolute density of black powder for mining purposes is now receiving greater recognition than it has at any previous time. Being unaffected by the size of the grains of powder or their angularity, absolute density presents at once a measure of the physical uniformity of black blasting powders in regard to specific gravity. Since dynamite is usually packed hard, it is commonly believed that the true density and the apparent density of the cartridge differ but little, and that the well packed stick of dynamite pre- sents few, if any, interstices or spaces filled with air. To test this point a sample of 40 per cent dynamite was taken, and its absolute density was determined upon the densimeter. The apparent specific gravity of this explosive had been found by careful measurement to be 1.24; showing that even in this well packed sample of dynamite there was present 25.53 per cent open or pore space. A number of other samples of the dynamite class were similarly examined, and the per cent of pore space was found to vary in different samples from 18.62 per cent to a maximum of over 30 per cent. In conclusion, it may be stated that the densimeter of the con- struction herein described has for more than a year been in prac- tical use in the laboratory of the Bureau of Mines, and in the course of that time has been used in a wide range of work. In this time it has given perfect satisfaction, and has proved to be an instrument of decided accuracy. Check and duplicate de- terminations agree to the third decimal place a degree of accu- racy which is seldom attained in work of this sort, but which be- comes possible in this instrument because of the fact that all weights made represent the mercury occupying the volume of the powder, and accordingly errors hi weighing produce very 116 Original Communications: Eighth International small differences in the absolute density, owing to the high spe- cific gravity of the mercury upon which the weights are based. In the determination of the density of black blasting powder this instrument furnishes a means of conveniently, quickly and accurately determining the absolute density of a sample of powder, and thus enables comparisons to be readily made. In construc- tion and operation this instrument seems to be simpler than any of the types of apparatus now commercially available for the purpose of determining the absolute density of black powder. THE EFFECT OF THE NITROTOLUENES ON THE DE- TERMINATION OF NITROGLYCERIN BY MEANS OF THE NITROMETER BT C. G. STORM Bureau of Mines, Pittsburgh, Pa. Among the coal mining explosives submitted to the Bureau of Mines for tests, a considerable percentage contain nitroglycer- in together with nitrotoluenes, the principal object of the latter ingredient being to reduce the freezing point of the mixture. The nitrotoluenes employed for this purpose are generally the commercial liquid di- or tri-nitrotoluenes, which are mixtures of the various nitrosubstitu tion products of toluene. Frequently the more or less pure crystalline di-or tri-nitro compounds are used, and less often liquid ortho-nitro toluene. In the analysis of such explosives, the nitroglycerin together with the nitrotoluene is obtained in the ether extract, the ether removed by evaporation in a small weighed beaker and the ex- tracted material weighed. The evaporation is allowed to proceed spontaneously at room temperature, usually over night, and the beaker is then placed in a vacuum desiccator over sulphuric acid until no odor of ether remains and any moisture condensed by the evaporation of the ether has been removed. The amount of the extracted material may be determined by a direct weighing of the beaker and contents after desiccating over night, or the desiccation may be continued only long enough to remove the ether and all but traces of moisture, for which about 3 or 4 hours is usually sufficient, and the weight of extracted material found from the loss of weight on extraction, allowance being made for the moisture originally present in the explosive. The desiccated mixture of nitroglycerin and nitrosubstitution product is next transferred to the decomposition bulb of a nitro- meter by dissolving from the beaker with sulphuric acid (sp. gr. 1.84) and nitrogen determined in the usual manner. 1 If Bulletin No. Bureau of Mines, " Analysis of Explosives." 117 118 Original Communications: Eighth International [VOL. crystalline nitro compounds insoluble in sulphuric acid are present, the entire mass is washed into the nitrometer with the acid. By the reaction with mercury and sulphuric acid in the nitro- meter, nitrates or nitrites, both inorganic or organic, are decom- posed, liberating all of their nitrogen as nitric oxide (NO). From the volume of NO obtained the per cent of such nitrogen is de- termined and from this the nitrate content of the sample. Nitro- substitution compounds are, however, not decomposed by the action of sulphuric acid and mercury, and hence yield no nitric oxide in the nitrometer. It would therefore appear that the amount of nitroglycerin present in a mixture with a nitrosubstitution compound could be accurately ascertained by a determination of the ester nitrogen by means of the nitrometer, and this method has up to the present time been the one most commonly used for the analysis of such mixtures. Mono- and di-nitrotoluenes are, however, capable of combining with more nitric acid, and it seemed reasonable to believe that such compounds might, in the presence of sulphuric acid, react with a portion of the nitric acid liberated from the nitroglycerin by the sulphuric acid, and thereby become more highly nitrated. Such reaction would naturally diminish the volume of nitric oxide liberated, and the amount of nitroglycerin calculated from this gas volume would be correspondingly low. A series of experiments were made to determine the effects of known amounts of the various nitrotoluenes on the determina- tion of nitroglycerin in the nitrometer. Various mixtures of each of the nitrotoluenes with nitroglycerin of known purity were weighed in a small beaker, about 10 cc. of sulphuric acid (95- 96%) added quickly, the mixture stirred and transferred to the generating bulb of the nitrometer, the beaker and nitrometer cup being completely washed with several portions of acid, a total of about 25 cc. of the sulphuric acid being used. The gen- erator was then shaken in the usual manner until the decompo- sition was complete, the NO transferred to the reading tube, and its volume noted. By dividing the reading of the gas vol ume found by the theoretical volume representing 1 gram of nitro- glycerin (18.50%), the weight of nitroglycerin found was deter- mined. Substr acting this weight from the weight of pure nitro- IT] Congress of Applied Chemistry 119 glycerin used gives the error of the determination due to the pres- ence of the nitrosubstitution compound. The results of the determinations are given in the following table. All of the nitrotoluenes used were commercial products manu- factured by Leitch & Co., Huddersfield, Eng., a grade commonly used by powder manufacturers in this country. TABLE I DETERMINATION OF NITROGLYCERIN IN THE PRESENCE OP NITROTOLUENES Test. II III IV V VI Nitro- Nitro- Nitro- Nitro- Nitrometer glycerin glycerin glycerin toluene reading found lost (grams) 1 (grams) (per cent) (grams) (grams) NITROGLYCERIN AND ORTHONITROTOLUENE VII Nitroglycerin lost per gram of nitrotoluene used 1. .9904 .7236 10.88 .5881 .4023 .5560 2. .8510 .4912 10.70 .5784 .2726 .5549 3. .9016 .2766 13.86 .7492 .1524 .5510 4. .8059 .2526 12.32 .6660 .1399 .5539 5. 1.1083 .7618 12.71 .6870 .4213 .5530 6. 1.3081 1.0273 13.74 .7481 .5600 .5452 7. .8515 .3084 12.60 .6811 .1704 .5525 8. .5101 1.0005 (No evolution of gas obtained.) 9. .5530 1.0000 (A very small bubble of gas gen- erated, volume too small to read.) NITROGLYCERIN AND PARANITROTOLUENE 10. 1.1057 .7505 12.78 .6908 .4149 .5515 11. 1.0050 .5000 13.50 .7297 .2753 .5506 NITROGLYCERIN AND LIQUID DINITROTOLUENE 12. 13. 14. 15. 16. .8134 1.1000 13.70 .7405 .0729 .8954 2.4364 13.63 .7367 .1587 .7408 .4982 13.17 .7119 .0289 .8500 2.0102 13.35 .7216 .1284 .7569 .5104 13.43 .7259 .0310 .0663 .0651 .0580 .0639 .0607 1 The weights in column I represent pure nitroglycerin in the samples taken, calculated from its purity (99.91%) as determined by the nitrometer. 120 Original Communications: Eighth International [VOL. NITROGLYCERIN AND LIQUID TRINITROTOLUENE 17. .7004 .9469 12.97 .7011 .0007 (gain) 18. .7118 2.5245 13.13 .7097 .0008 19. .7187 1.0236 13.28 .7179 .0008 20. .7049 2.0186 13.00 .7027 .0011 NITROGLYCERIN AND DINITROTOLUENE, (M. P. 66-68) 21. .7362 .6963 13.61 .7357 .0005 22. .7364 .7347 13.63 .7366 .0002 (gain)- NITROGLYCERIN AND TRINITROTOLUENE, (M. P. 80-81) 23. .7300 .7023 13.51 .7303 .0003 (gain)- 24. .7285 .7516 13.47 .7285 .0000 It will be noted from the above table that neither pure di- nitrotoluene nor trinitrotoluene have any effect on the determi- nation of nitroglycerin. The liquid trinitrotoluene has only a slight influence. On the other hand, the mononitrotoluenes (both ortho and para) and liquid dinitrotoluene each take up a definite amount of the nitric acid resulting from the decompo- sition of the nitroglycerin, causing an error in the determination depending on the amount of such nitro compound present. Tak- ing an average of the values in column VII for tests 1-7 inclusive, it is noted that one gram of mononitrotoluene (ortho) causes a loss of .5530 grams of nitroglycerin, equal to .1023 grams of nitrogen. Since, theoretically, .1022 gram of nitrogen would be taken up in the conversion of one gram of mononitrotoluene to dinitrotoluene (C 7 H 7 N0 2 = 10.22%N), the results appear to indicate that the mononitrotoluene is completely converted to dinitrotoluene. In test 8 the nitric acid evolved from the nitro- glycerin was entirely taken up by the large excess of nitrotoluene present, so that no generation of gas resulted. In test 9 the mix- ture was so proportioned that an approximate condition of equilibrium existed, that is, the amount of nitroglycerin was just about sufficient to provide for the conversion of the mono- nitrotoluene to dinitrotoluene, and leave no nitric oxide to be set free. The small bubble of gas obtained may readily be accounted for by a slight error in weighing, since a volume of 0.10 cc. NO would be obtained from .0003 gram cf nitroglycerin in excess of the theoretical amount necessary for equilibrium. iv] Congress of Applied Chemistry 121 Tests 10 and 11 show an average loss of .5510 gram nitrogly- cerin, equivalent to .1019 gram of nitrogen, per gram of para- nitrotolueme, or practically the same figure as in the case of or- thonitrotoluene. Liquid dinitrotoluene (tests 12-16 inc.) causes only a rela- tively slight loss, and as it is evident from tests 21-22 that dinitro- toluene itself has no effect, it would seem reasonable to assume that the loss is due to the presence of mononitrotoluene in the liquid dinitro compound. The average loss of nitroglycerin in tests 12-16 (.0628 gram for each gram of liquid dinitrotoluene) is equivalent to the loss which would be found if each gram of the latter contained - = .1136 gram, or 11.36%, mononitro- .5530 toluene. In order to prove the presence of the mono compound in the liquid dinitrotoluene, 50 cc. of the latter were distilled with steam, the distillate shaken with ether, the ether solution separated and evaporated by warming gently. A yellow liquid residue was obtained which had a strong odor of orthonitrotol- uene. On adding a weighed portion of this residue to a weighed amount of nitroglycerin and determining the latter in the nitro- meter, a definite loss of nitroglycerin was noted, as shown below. Nitro- Nitro- Nitroglycerin Nitro- Nitro- Nitrometer glycerin glycerin lost per gram glycerin toluene reading found lost of nitrotoluene Test. (grams) (grams) (per cent) (grams) (grams) (grams) 25. 1.0045 .8004 11.90 .6432 .3613 .4514 26. 1.0162 .8350 11.84 .6400 .3762 .4505 Since one gram of pure mononitrotoluene caused a loss of .5530 gram nitroglycerin (Table I), the above loss of .4510 gram would indicate that the material distilled from the liquid dinitrotoluene contained 81.55% mononitrotoluene I- - = 81.55%). V .5530 J No attempt was made to determine the total content of mononi- trotoluene in the liquid dinitrotoluene, owing to the fact that distillation with steam affords only slow and approximate separa- tion, small amounts of both di- and tri-compounds passing over in the distillate. Addition of acetone and potassium hydroxide gave the characteristic red coloration produced by trinitrotoluene. This color completely masks the blue and yellow colors characteris- 122 Original Communications: Eighth International [VOL. tic of the di- and mono-compounds respectively. However, the experiment appears to confirm the presence of mononitrotoluene in the liquid dinitrotoluene, and verifies the conclusions drawn from the results in Table I. By a similar procedure a small amount of a yellow distillate was separated from liquid trinitrotoluene. A determination of a weighed amount of nitroglycerin mixed with this distillate showed a loss of only .0464 gram of nitroglycerin per gram of the distillate, indicating about 8.4% mononitrotoluene in the latter. As the amount of distillate (.68 gram) was only about 1% of the weight of liquid trinitrotoluene used, it can be seen that the latter contains only about .08% mononitrotoluene, which amount would have a negligible effect on the determination of nitroglycerin in a mixture containing nitroglycerin and iquid trinitrotoluene. (See tests 17-20, Table I.) The following additional experiments were made to show that the effect of mononitrotoluene on the determination of nitro- glycerin is not influenced by the presence of the higher nitrotol- uenes. Mixtures were prepared similar to those in Table I, in which orthonitrotoluene and a higher nitro compound were both added to nitroglycerin. Determinations of the latter in- gredient in the nitrometer gave the following results: TABLE II DETERMINATION OF NITROGLYCERIN IN THE PRESENCE OF MIX- TURES OF NITROTOLUENES Nitro- Nitro- Nitroglycerin Nitro- Nitro- Nitrometer glycerin gtycerin lost per gram of Test. glycerin toluenes reading found lost mononitrotoluene (grams) (grams) (per cent) (grams) (grams) (grams) 27. 1.3118 1.0273 a 13.74 .7481 .5637 .5487 .4985 b 28. .8523 .3084 a 12.60 .6811 .1712 .5551 1.7654 b 29. .9071 .2638 a 14.05 .7595 .1476 30. 1.1772 1.0173 b .8073 a 13.50 1.0048 d .7297 .4475 .5595 .5543 Note: In Table II, a = orthonitrotoluene, 6= dinitrotoluene (m. p. 66-68), C=liquid trinitrotoluene, d= trinitrotoluene (m. p. 80-81), iv] Congress of Applied Chemistry 123 These values for the loss of nitroglycerin per gram of mono- nitrotoluene agree closely with those found in Table I. It now remained to prove that in the tests described, there is a complete conversion of mononitrotoluene to dinitrotoluene when the amount of nitroglycerin present is sufficient to furnish the necessary quantity of nitric acid. The clear acid solution from the nitrometer in test 9 (Table I) was drawn off as completely as possible and poured into about 250 cc. of water, with stirring. After cooling, the emulsion was shaken with ether in a separatory funnel, the lower acid layer drawn off and the clear ether solution evaporated in a weighed beaker. The residue obtained was a light-yellow crystalline mass weighing 1.2650 grams. The 1 gram of orthonitrotoluene used would theoretically be equivalent to 1.3284 grams dinitrotoluene. A second trial gave 1.2520 grams of the crystalline substance. This material gave the characteristic color test for dinitrotolu- ene, a blue coloration on adding potassium hydroxide to an ace- tone solution of the crystals, showing that not even small amounts of trinitrotoluene were produced as the red color from the latter would have masked the blue color produced by the dinitrotoluene. The fact that mixtures of the mononitrotoluenes with nitro- glycerin became warm on the addition of sulphuric acid, while the separate ingredients did not, indicated that the nitration of the mononitrotoluene occurred immediately on adding the acid to the mixture, and before transferring it to the nitrometer. Experiments were therefore made with both ortho and para nitrotoluenes, one gram of each in separate beakers, being dissolved in 10 cc. sulphuric acid, and .5530 gram nitroglycerin dissolved in 10 cc. sulphuric acid added to each. The mixtures were allowed to cool, poured into water, shaken with ether and the ether ex- tract evaporated. The 1 gram orthonitrotoluene yielded 1.3143 grams, the 1 gram paranitrotoluene 1.2996 grams of crystalline product (theory 1.3284 grams dinitrotoluene). The two sub- stances obtained were different in appearance, that from the para compound crystallizing readily from the ether in large needles while that from the ortho compound became a solid maaauafter all the ether had volatilized. The former, recrystallizedjfrom absolute alcohol showed a melting point of 70, corresponding 124 Original Communications: Eighth International [VOL. with 2.4 dinitrotoluene (m. p. 70.5), while the latter recryst alii zed from absolute alcohol in the form of fine needles, had a melting point of about 50, corresponding to 2.5. dinitrotoluene (m. p. 48). In a further experiment, one gram of orthonitrotoluene dissolved in 10 cc. sulphuric acid was treated with a mixture of 10 cc. of sulphuric acid and .657 gram of 70% nitric acid (equivalent to .5530 gram nitroglycerin). From this mixture 1.25 grams di- nitrotoluene with a melting point of 50 was separated by the method described above. The fact that the yield of dinitrotoluene was a trifle low is in a large measure due to inaccuracies of the method of separation and to the volatility of the nitro compounds with ether. CONCLUSIONS In the determination of nitroglycerin in the presence of one or more of the nitrosubstitution products of toluene by means of the nitrometer, no errors are introduced by the action of pure di- or tri-nitrotoluenes. The mononitrotoluenes, both ortho and para, are, however, quantitatively converted to dinitrotoluene by the nitrating action of the nitric acid liberated from the nitro- glycerin by the sulphuric acid. The large excess of the latter takes up the water liberated by the reaction and therefore permits the reaction to proceed until either all of the nitric acid has been used or all of the mononitrotoluene nitrated to the dinitro-compound. Thus, if in a mixture of nitroglycerin and mononitrotoluene, the nitroglycerin present comprises not more than 35.6% of the total, no nitric oxide will be generated in the nitrometer. If the pro- portion of nitroglycerin is greater than this amount the error of the determination will be equal to .55 gram of nitroglycerin for every gram of mononitrotoluene present. Orthonitrotoluene was found to be converted to the 2.5. dinitrotoluene, while the para- nitrotoluene produced the 2.4. dinitrotoluene. These relations are not influenced by the presence of the higher nitrotoluenes, since trinitrotoluene cannot be produced from either the mono- or di-compounds except by the aid of higher temperatures and greater concentration of acids. It is therefore apparent that in the analysis of unknown mixtures of nitroglycerin with nitrotoluenes, the results obtained with rv] Congress of Applied Chemistry 125 the nitrometer are of no value if mononitrotoluene is present in appreciable amounts. The liquid mixtures of somewhat indefinite composition, known as liquid di- or tri-nitrotoluenes, affect the determination according to the amount of mononitrotoluene present. In the case of the liquid dinitrotoluene used in the above experiments, the results indicated a content of 11.36 per cent of mononitrotoluene, 1 gram of the liquid dinitrotoluene causing an error of .0628 gram of nitroglycerin. In an explosive containing 25% nitroglycerin and 10% liquid dinitrotoluene, the error in the determination of the nitroglycerin by the nitrometer would then be .628%, i.e., the per cent of nitroglycerin found would be 24.37 instead of 25.00. Most of the usual types of low freezing dynamites containing the liquid di- or tri-nitrotoluenes can therefore be determined by means of the nitrometer without serious error. The writer desires to acknowledge the valuable assistance of Mr. J. H. Hunter and Mr. A. L. Hyde, both of the Explosives Chemical Laboratory, Bureau of Mines, in the analytical work of this paper. RECHERCHES DE LA STATION D'ESSAIS DE LIEVIN SUR LES EXPLOSIFS DE SURETE POUR MINES GRISOUTEUSES ET POUSSIEREUSES PAR MM. J. TAFPANEL, ET H. DAUTRICHE Station d'essais de Lievin, France I. Un explosif de surete ideal serait celui qui, en aucune cir- constance, ne pourrait allumer le grisou ou les poussires de houille. Un tel explosif n'existe pas, et il ne peut s'agir que de securite relative. La difficult^ est de fixer les regies suivant lesquelles les explosifs seront compare's, afin d'eliminer les moins stirs et de reserver pour les mines grisouteuses et poussiereuses ceux pour lesquels la probability d'innammation est le plus re*duite. En France, jusqu'il y a quelques mois, la regie etait de n'ad- mettre dans ces mines que des explosifs detonants, a 1'exclusion de la poudre noire, et avec la double condition que leur temperature theorique de detonation soit infe*rieure a 1500 ou 1900 degres centigrades, suivant le genre des travaux, et qu'il ne se trouve aucun gaz combustible parmi les produits de la detonation. Par cette regie on reduisait trois des plus importantes causes d'in- nammation: dure*e de la flamme, temperature de la flamme, risque de flamme secondaire apres melange avec Fair de la galerie. En d'autres pays, la rgle est de classer Is explosifs d'apr&s les resultats bruts d'essais pratiques d'inflammation, executes dans des conditions invariables, en tirant Pexplosif dans un canon d'acier, le plus sou vent sans bourrage. Par cette regie, on admet implicitement que les conditions de tir realisees dans 1'essai presentent assez d'analogie avec celles de la pratique pour que le classement resultant du tir artificiel soit applicable aux diverses variantes du tir reel. Aucune de ces regies ne donne complete satisfaction. L'an- cienne regie fran9aise ne tenait pas compte de tous les facteurs du probleme en sorte qu'il pouvait exister des explosifs relative- ment bons parmi ceux qu'elle excluait. Quant au classement par les charges limites obtenues dans les essais pratiques, il est 127 128 Original Communications: Eighth International [VOL. variable suivant les conditions des essais, et comme ces condi- tions different toujours quelque peu de celles de la pratique, il y a incertitude sur la valeur pratique du classement obtenu; d'autre part les conditions les plus favorables a Pinflammation ne sont pas ne*cessairement les memes pour les divers types d'ex- plosifs, en sorte qu'en adoptant une formule invariable d'essais, on peut avantager certains types au detriment des autres. Pour aboutir a de meilleures regies, il est necessaire d'e*tudier de plus pres le mode de fanctionnement des explosifs dans les diverses circonstances de P experience et de la pratique, de recher- cher quels sont les divers mecanismes d'inflammation du grisou ou des poussieres et d'approprier les me'triodes d'essai a la nature de Pexplosif essaye*, selon le mode d'inflammation envisage. Cette ample etude a ete* entreprise par la Station d'essais de Lie*vin; elle est loin d'etre terminee; nous pouvons cependant indiquer des maintenant quelques uns des resultats obtenus. II. Trois principes importants ont 6te* etablis ou confirmes par les travaux de la Station d'essais de Lievin. Le premier principe est le suivant: La decomposition complete, selon la formule theorique, est presque toujours realisee, dans la combustion en vase clos, lorsque la densite de chargement n'est pas trop faible; elle paraU devoir etre egalement realisee dans les conditions de la pratique, en trous de mine bien bourres; mais, pour la plupart des explosifs, la decomposition est ires incomplete dans le tir au canon d'acier, tel qu'on le pratique dans les essais, meme avec les densites de chargement qui donnent la decomposition complete en vase clos; elle est encore plus incomplete dans le tir a I'air libre. Cette proposition a ete* etablie par des mesures calorimetriques et des analyses de gaz. Des combustions en vase clos ont ete realisees dans la bombe de Sarrau et Vieille, placee dans un calorimetre; les quantites de chaleur degagees (grandes calories par kilogramme d'explosif), mesurees apres condensation de la vapeur d'eau, sont portees au tableau suivant, en regard des quantity's theoriques calcule*es dans Phypothese d'une decomposition complete 1 . ^Compositions centesimales des explosifs. IV] Congress of Applied Chemistry 129 Designation de 1'explosif Chaleur theorique (eau condensee) Chaleur mesuree pour lea densit6s de chargement ci-dessous 0,05 0,1 0,2 0,3 0,4 Kohlencarbonit . ... 594 767 776 957 941 895 1205 861 030 ssi 861 642 783 816 951 978 888 1222 939 722 Grisou-dynamite couche .... Grisou-naphtalite couche sal- pe'tre'e Sabulite Grisou-dynamite roche Grisou-dynamite roche sal- pe"tr4e Dynamite nl a a a Designation de 1'explosif trate d'ammoniu N'O'H* 1 Sfc | trate de baryum N?OBa 1 ton azotique C21H2O8N8 |g J3O |2> "3 initronaphtaline CioH5Q8N 1 a j lorure d'ammoni NH840 >670 220 Grisou-dynamite roche En presence des poussieres, les sels alcalins n'ont pas d'in- fluence nette; c'est une preuve que le me*canisme d'inflammation est different. 2 Influence de la densite* de 1'explosif. On sait que la densite* d'encartouchage influe sur le mode de detonation. Pour un diametre de cartouche donne*, la vitesse de detonation varie avec la densite* de 1'explosif; elle croit d'abord regulierement avec cette densite* jusqu'a une certaine densit^ 136 Original Communications: Eighth International [VOL. limite, & partir de laquelle la vitesse de detonation decroit rapide- ment, jusqu'a ce que Pexplosif ne detone plus. Nous avons constate que la densite d'encartouchage avait egeiement une grande influence sur le risque d'inflammation du grisou. Ainsi la grisou-dynamite roche, tir^e en cartouches de 30 m/m, dans un canon de 40 m/m., a donne* une charge limite de 150 grammes pour les densites comprises entre 1, 4 et 1, 5 et une charge limite supeYieure a 725 grammes pour les densites in- ferieures a 1, 4 et notamment pour celles comprises entre 1, 3 et 1, 4. II est remarquable qu'une aussi faible variation de densite entraine une aussi grande difference dans la charge limite; cette difference ne peut d'ailleurs aucunement s'expliquer par la ^valeur le*gerement plus grande de la densite* de chargement, dans le cas de la plus forte densite d'encartouchage; car une charge limite presque aussi basse (200 grammes) a e"te obtenue avec des cartouches de 20 m/m. de diametre qui, a la densite* de 1, 4 a 1, 5, re*alisaient, en raison de leur diametre re*duit, une densite de chargement beaucoup plus faible que les echantillons com- paratifs. Les essais calorimetriques montrent d'ailleurs que les variations observers dans les charges limites ne correspondent pas a des differences notables dans les chaleurs d'e*gagees. Des resultats analogues ont e*te* obtenus avec les grisou-naph- talites couche et roche. On a trouve* chaque fois une certaine densite limite se*parant des densit^s fortes ou la charge limite est relativement basse, et des densites plus faibles ou la charge limite est relativement e*leve"e, pour des conditions de tir determinees. La valeur de cette densite limite varie suivant la nature de Pexplosif et son degre d'humidite; elle s'est trouvee parfois, mais non toujours, coincider avec la densite limite au point de vue de la vitesse de detonation; comme elle n'est pas eioignee des densites usuelles d'encartouchage, de legeres variations dans la fabrication des cartouches ou dans le degre d'humidite sont capables de modifier profondement les charges limites de securite, soit dans les condi- tions des essais, soit peut-etre aussi dans les conditions du tir reel. 3 Influence du degre de trituration. rv] Congress of Applied Chemistry 137 Les explosifs usuels sont formes de plusieurs composes qui sont supposes reagir les uns sur les autres au moment de la detonation. Si le melange n'est pas tout a fait intime et parfait, diverses anomalies peuvent se produire; si la detente est tre"s rapide, il pent arriver que certaines reactions ou decompositions n'aient pas le temps de se produire; dans d'autres cas, ou les gaz sont mieux maintenus en contact, on pourra avoir des reactions plus completes, mais d'une dure*e prolongee. Ces causes peuvent agir differemment suivant les conditions du tir et le me*canisme de 1'inflammation; une mauvaise tritura- tion a recuie la charge limite d'une grisou-naphtalite couche tiree en cartouches suspendues, tandis qu'elle a fortement abaisse la charge limite d'une kohlencarbonit tir^e dans le mortier d'acier en presence des poussieTes. 4 Influences diverses. D 'autres causes concernant la constitution chimique et physique de Pexplosif, notamment le grenage, le degre* d'humidite*, influent sur la securite; nous ne nous attarderons pas a discuter ces effets qui sont d'ailleurs generalement connus. V. Influences relatives aux conditions de tir. 1 Influence du diamtre du canon et de la densite de charge- ment. Cette double influence a e*te* constatee bien des fois dans les galeries d'essais; elle a donne lieu, a la Station d'essais de LieVin, a de nombreuses recherches, dans le detail desquelles il serait trop long d'entrer. II suffira de rappeler que d'une maniere generale les charges limites sont d'autant plus basses que le diametre de P dme du canon est plus faible, et que la densite de chargement est plus forte. Toutefois, en presence du grisou, les conditions les plus dures sont realise*es quand la cartouche est librement suspendue au sein du melange explosif; la grisou-dynamite couche dont la detonation dans ces conditions est particulierement incomplete, n'enflamme generalement le grisou qu'a partir de 250 grammes; les autres explosifs essayes renflamment a partir de 100 ou 50 grammes; c'est notamment le cas pour la kohlencarbonit. 2 Influence de la nature des enveloppes. 138 Original Communications: Eighth International [VOL. L'enveloppe des cartouches, si elle est composee de corps suscep- tibles de re*agir chimiquement sur les produits gazeux de la deto- nation, participe aux reactions explosives. Conditions de 1'essai 02 CO2 CO H2 CH< Observations Decomposition theorique. . . . Cartouche place dans le canon d'acier, sans aucune enveloppe % 11,4 I' 1 Q % 21,9 19,0 a oo o % 0,0 a 4 1 % 0,0 a 1 A % 0,0 4 essais 8 prises de Memes dispositions, mais avec leger bourrage 19,1 ^fc,- 1 1,4 A, 1,0 A gaz. 2 essais Cartouche entouree d'une enveloppe de papier d'ami- ante (sans bourrage) a 3,0 0,1 a Q Q a 22/2 18,0 a 20 *> 4,8 1,0 a 0,5 0,6 1,3 gaz. 2 essais 3 'prises de Cartouche entouree d'un papier paraffin^ O,O -2,9 a 13,3 5,3 a 1,7 a 5,0 a 3 essais M6mes dispositions, mais avec leger bourrage Cartouche entouree d'une feuille d' aluminium (sans bourrage) -0,8 T -0,9 -2,9 a -0 8 19,5 18,3 a 23,1 19,8 a 21 13,1 9,6 / 1 1 6,2 0,6 a 1,2 0,7 a 1 *> 8,6 3,6 a 4,5 gaz. 2 essais 3 p-. de gaz. 1 essai 2 prise ; de Cartouche entouree d'une feuille d'etain u,o -2,0 a V ^J- 1 - 1,8 a X,J g- :A ::::::- o -o SO -1C \/ ; V -010 -oo - -iO(N -Oi-H to o *o o CO -b- O O5 00 A A H 10 -o t^ 'lO O -iH -o -oo -ico .b * -TH -COO5 -I>00 !> -00 CO iter T^cueil des flammes secondaires en se plagant du c6te* des limites superieures d'inflammabilite; mais il est sans doute preferable, du moment que, dans le tir au charbon, on ne peut garantir Pabsence de gaz combustibles, d'agir sur la composition de 1'explosif pour diminuer le risque que ces gaz ne s'enflamment, en utilisant, par exemple, les propriete*s des sels alcalins. 3 Influence de la longueur du canon. La charge limite depend non seulement du diamtre du canon, comme il a ete souvent ve*rifie, mais aussi de sa longueur. Dans un canon de 55 millimetres de diametre inte*rieur et 1 m 20 de longueur la charge limite de la grisou-dynamite roche salpetre"e entoure*e de papier ordinaire, est de 275 grammes; mais si Ton reduit la longueur du canon en bourrant le fond avec de Pargile, la charge limite croit progressivement comme le montre le tableau suivant : Longueur du canon. . 1 m 20 1,15 1,00 0,60 metre Charge limite 275 375 450 525 grammes. Cette importante variation de charge limite, qui va presque du simple au double, n'est pas imputable a la variation de position des cartouches par rapport a 1'orifice; car si, laissant celles-ci sur les m 60 voisins de Porifice, comme dans les essais qui donnerent la charge limite de 525 grammes on rend au canon sa longueur entiere de 1 m 20, on retrouve la charge limite initiale de 275 grammes. Les cir Constances qui elevent la charge limite sont done celles pour lesquelles une plus grande surface de me*tal est au contact des gaz de la detonation et Ton est d'autant plus con- duit a songer a la participation probable des parois a la reaction, que Ton constate dans la chaudiere calorimetrique, un certain accroissement du nombre de calories degagees (20%, dans le cas etudie, pour la grisou-dynamite roche tire*e dans un meme canon de 55 millimetres ayant 1 m ou m 50 de profondeur). Mais on ne peut pas faire abstraction d'autres causes de difference 142 Original Communications: Eighth International [VOL. dans le mode de detente des gaz, les pressions exercees sur le milieu ambiant et la duree de la flamme. 4 Influence du bourrage. II est bien connu que le bourrage augmente la charge limite. Des Etudes comparatives ont ete faites sur Pinfluence relative des divers modes de bourrage dans Pessai au canon, bourrage interieur selon la pratique courante, bourrage exterieur selon la formule proposed par les experimentateurs beiges a titre de sur- croit de garantie. Les essais execute's en presence des poussieres montrent que les divers modes de bourrage, compares a poids egal, se classent comme suit dans Pordre d'efficacite decroissante : bourrage interieur au contact de la charge, bourrage interieur distant de la charge, bourrage exte*rieur obstruant Porifice, bour- rage exterieur en tas devant Porifice, mais ne Pobstruant pas. II faut deux a quatre fois plus de bourre sous forme de bourrage exterieur que sous forme de bourrage interieur, pour determiner un mme recul de la charge limite; mais la quantity de bourre que Pon peut accumuler devant un trou de mine n'est pas limitee comme celle que Pon peut introduire dans le trou, et si Pon com- bine les deux systemes, ainsi que Pentendent expressement les promoteurs du bourrage exterieur, il n'en peut re*sulter, dans le cas ge*ne*ral, qu'un se*rieux appoint de se*curite. 5 Deflagrations fusantes. Les explosifs a Pazotate d'ammoniaque etudie"s ci-dessus, grisou-dynamites et grisou-naphtalites, ne paraissent a priori susceptibles que de deux modes de decomposition: la detonation amorce*e au fulminate et Pexplosion ou combustion sous forte pression (essai a la bombe); a la pression atmospherique ces explosifs ne propagent pas la flamme. Or, on a signale que, dans la mine, certains trous de mine au charbon, ayant mal travaille, donnaient lieu a une flamme rouge persistant au fond du trou, avec un bruit particulier assimile a celui de la graisse bouillante, et un de"gagement abondant de fumees acres. La "deflagration fusante" ainsi caracterisee peut durer une ou plusieurs minutes; une duree superieure a une demi-heure aurait meme ete const atee. Ce ph^nomene risque de passer inapergu s'il est d'usage d'attendre quelque temps avant de revenir au chantier; aussi est-il peut-etre plus frequent qu'on ne le suppose parfois; il constitue un danger iv] Congress of Applied Chemistry 143 se*rieux dans les mines grisouteuses. II n'a jamais e*te* constate* a 1'occasion de trous fore's dans le rocher. Les essais ont montre* que la deflagration fusante e*tait due au voisinage de la paroi de charbon; on a reproduit le phenomene en allumant avec une flamme un melange d'explosif, ou simple- ment d'azotate d'ammoniaque et de charbon; il n'est meme pas necessaire, pour entretenir la combustion fusante, que le melange soit preexist ant; il suffit de faire tomber de temps en temps des fragments de charbon sur la cartouche en train de fuser; on a meme pu, au moyen d'un amorgage approprie, realiser des de*- flagrations fusantes dans le canon d'acier, ou la cartouche d'ex- plosif avait etc* entoure*e de charbon pulverulent. Ce dangereux phenomene, qu'il faut craindre avec tous les explosifs a 1'azotate d'ammoniaque, ne peut survenir qu'a la suite d'un rate de transmission de la detonation. II convient, pour 1'eviter, de curer soigneusement le trou de mine et de mettre les cartouches bien en contact. II faut en outre que 1'explosif ait une aptitude suffisante a la transmission de la detonation. On est ainsi conduit a etudier la question de 1'amorc.age des explosifs dans ses rapports avec la question de la transmission de la detonation de la cartouche amorce aux cartouches suivantes. Les distances de transmission sont mesure'es, pour plus de com- modite*, sur plaques de plomb; on a d'ailleurs verifie que pour le grisou-naphtalite couche les distances de transmission &taient les memes dans le canon. On a constate que la distance maximum de transmission e*tait notablement plus faible a partir de I'extremit4 de la cartouche amorce oil Ton a introduit la capsule de fulminate, qu'a partir de Textremite opposee; on trouve, par exemple, pour les naphtalites couche, 2 centimetres d'un cote*, 5 centimetres de 1'autre. II y a done inte*ret, pour la bonne transmission de la detonation, a pratiquer 1'amorc.age direct (amorce cote du bourrage) plutot que 1'amorc.age inverse (amorce cote de la derniere cartouche). C'est la distance a 1'amorce fulminante qui importe plutot que le sens de transmission de la detonation; en amor gage direct, une cartouche longue transmet mieux qu'une courte; si la de*tonateur e*tait au milieu de la cartouche, la transmission se ferait aussi bien par les deux extremity's. 144 Original Communications: Eighth International [VOL. 6 Le cordeau detonant. M. Lheure, Ingenieur des Poudres et Salpetres, a imaging de plager, sur toute la longueur de la charge, un cordeau detonant; ce dispositif supprime les rate's de transmission aux intervalles de cartouches et accroit dans une certaine mesure la densite limite a partir de laquelle commencent les detonations incom- pletes. Ce sont d'incontestables avantages. Mais ce qui est vrai des enveloppes Pest e*galement du cordeau detonant: les produits de la detonation de 1'explosif contenu dans le cordeau, et meme le plomb qui en constitue 1'enveloppe peuvent participer aux reactions explosives de telle sorte qu'au point de vue des chaleurs degagees le resultat est alors le meme que si les constituants du cordeau avaient e*te* intimement meles aux car- touches. Ces combustions peuvent tre parfois suffisamment rapides pour accroitre notablement la puissance des explosifs amorce's au cordeau. L'influence de cet amorgage sur les charges limites au canon a ete etudiee sur des cartouches de grisou-naphtalite couche com- prime'e. II a fallu faire des essais avec bourrage afin que le risque d'inflammation resultant de la detonation du cordeau en atmo- sphere grisouteuse ne risquat pas de marquer 1'effet du cordeau sur 1'explosif et sur sa plus ou moins grande aptitude a enflammer le grisou. II est apparu une difference systematique indiquant que le cordeau augmente le risque d'inflammation; mais la diffe- rence est faible et 1'on retablit 1'egalite de risque en ajoutant seulement deux centimetres de bourrage. II semble bien que, dans les conditions de ces essais, les combustions secondaires ont ete peu importantes. CONCLUSION Ce rapide apergu de quelques uns des aspects de la question des explosifs de surete, telle qu'elle est etudiee a la Station d'essais de Lievin, montre toute la complexite du probleme. Les formules simples par lesquelles on a coutume de definir les explosifs de surete, aussi bien Fancienne formule theorique du reglement frangais que les formules pratiques usitees a Tetranger, apres avoir rendu de trs grands services, auront sans doute bientot fait leur temps; deja, en France, la nouvelle reglementation cesse rv] Congress of Applied Chemistry 145 de fixer une regie uniforme pour caracte*riser les explosifs de surete* ; il est reserve* a la Commission du grisou de juger, dans chaque cas, des epreuves a faire subir aux nouveaux explosifs proposes, e*preuves qui peuvent dependre du type d'explosif et qui eVolueront avec les progress de nos connaissances. La Station d'essais de LieVin, a qui incombe le soin de l'expe*rimentation, s'attache a faire, de chaque explosif, une etude aussi complete que possible, en variant les conditions d'essais et examinant les divers risques. II est avant tout desirable que Ton poursuive, dans les diverses Stations d'essais, des etudes rationnelles et systematiques, dans le but de mieux connaitre les divers mecanismes d'inflammation et le role des diverses proprie"tes de Pexplosif dans chacun de ces me*canismes. La Station d'essais de LieVin s'est oriente*e resolu- ment dans cette voie des recherches generates et s'est outill^e en consequence; elle acquiert de jour en jour la conviction plus intime que les classements actuels d'explosifs de surete sont fortement sujets a critique, et que les critiques ne pourront etre levies, que de serieux progres ne pourront etre realises que grace a une ^tude scientifique approfondie de la question. Published by permission of the Director of the Bureau of Standards ON A MODIFIED FORM OF STABILITY TEST BY H. C. P. WEBER Bureau of Standards, Washington, D. C. Some time ago an investigation on the stability of nitrocellulose plastics was undertaken at the Bureau of Standards and the question of the stability of these materials at normal and elevated temperatures was one of the questions studied. One of the stability tests, employed in that investigation, seems of sufficient interest to warrant calling attention to it, especially since it does not seem possible to carry the investigation further at present. The papers to which reference has been made in connection with this subject are tabulated at the end of this paper and will not be cited again in detail. There is perhaps no need for going into details regarding all the various tests proposed. While there are many of them, each having its own particular advantage, only a few are at all generally applied. The reason for adding to their number is that while this test is an explosion test, and therefore simple and rapid, it is in reality a determination of the change of decomposition veloc- ity with rise in temperature and, as such, a measure of stability. Various investigators have touched upon the influence of the rate of heating on the result, whether it be in the explosion tests or in methods depending on the amount or rate of gas evolution. For this reason the rate of heating in the explosion test is defined within certain limits. The decomposition of nitrocellulose is autocatalytic and when a certain surrounding temperature is attained, say 135 C ., nearly all samples of nitrocellulose will explode if kept in surroundings of that temperature long enough. The temperature of the decomposing material may be a few or many degrees above 135. In the investigation of nitrocellulose plastics we have repeatedly seen differences of 30 and 40 between 147 148 Original Communications: Eighth International [VOL. the temperature of the surroundings and of the sample when the substance went off. The amount of this difference depends on the mass of the material and its heat conductivity, and on the heat conductivity of the system used for test. These factors enter into the German 135 test as well as into the ordinary high temperature explosion test. In the former the time will vary with the heat insulation, in the latter the explosion temperature will vary with the rate of heating. The apparatus used for the test is shown in Fig. 1. The heating bath consists of a crucible of iron or nickel, about 10 cm. in dia- meter and of approximately the same depth. The cover is of sheet metal about 3 mm. thick, with a flange that fits snugly into the crucible and projects slightly beyond the rim. One hole through the centre of the cover is just large enough to permit the thermometer to pass. Symmetrically distributed around the centre of the plate are 8 openings 15 mm. in diameter. The heavy metal supporting tubes are about 4 cm. in length and about 12 mm. internal diameter. The lower end of the tube is flanged so that the tube rests securely on the cover. The test tubes are about 9 cm. long and must be of such diameter that they will just slip freely into the supporting tubes. A number of extra caps are provided to cover openings not intended to be used during the test. The heating liquid may be either paraffine, glycerine, or similar inert liquid which may be heated to 200 without boiling or fuming strongly. 1 The test tubes should dip about 4 or 5 cm. into the heated liquid so that their ends will be at about the centre of the heated mass and may readily be removed one at a time and replaced by fresh ones. For each explosion a clean tube should be taken. The thermometer is supported by a metal clip which rests on the cover, the bulb being on a level with the lower ends of the test tubes. The thermometer used was standardized. Since the mercury thread projected but little above the highly heated zone the stem correction was found to be negligible. This should be checked with each apparatus and thermometer for the various temperatures, when the apparatus is put together. 1( rhis form of apparatus has been devised by C. E. Waters in connection with work on lubricating oils. iv] Congress of Applied Chemistry 149 FIGURE 1 The heating bath is suspended in a conical piece of sheet metal wrapped with asbestos. The metal shield (Met. Sh.) is cut so that the crucible will hang securely in the upper smaller opening, while its larger end rests in the flanged tripod rim. With this apparatus and a small gas flame it has been easily possible to maintain the temperature constant for 15 or 20 minutes within half a degree. It is most convenient to have the burner set so 150 Original Communications: Eighth International [VOL. that there is a tendency for the temperature to fall and to use a small accessory flame momentarily whenever necessary. With a temperature regulator or with electrical heating the ease of manipulation might no doubt be increased but this is a matter of detail. One or two stop watches 1 complete the equipment. When the apparatus has attained equilibrium at the desired temperature one or more of the test tubes is loaded by dropping in the sample of powder, the stopwatch is started and a cork is dropped into the mouth of the test tube. The time until the explosion takes place is then noted. The grains of the six pounder smokeless powder are of conven- ient size to use directly. Powders of larger caliber should be cut into pieces weighing about 0.2g each. Each sample of powder was tested at four temperatures, 160, 170, 180, and 200. For the present purpose at least three tests were made at each temperature interval and curve was drawn through the average values. The following series on powder A shows how closely duplicates may be expected to agree: Max. Av. Variations 200 1'44" 1'48" 1'44" l'48" l'45" 1'46" 2% 180 2'55" 3'06" 3' 10" 3 / 04 // 5% 170 4 / 17 // 4'30" 4'28" 4'25" 3% 160 14'20" 17'40" 17'34" 16'36" 14% In general the discrepancies appear to be greater at the lower temperatures. This is to be expected since the curves given further on show to what extent the influence of small temperature varia- tions is magnified in the region of 160. Furthermore the irregu- larities are more pronounced in the "poor" powders. The following set shows what can be expected as to reproduci- bility of the complete curve. The sample used was L and the second test was made one month later than the first. The averages only are given. L I have found the type of stop watch with two second hands, which may be stopped independently, the most convenient form. With two of these at least four samples may be observed simultaneously. iv] Congress of Applied Chemistry 151 200 180 170 160 I. 1'31" 2'18" 3'39" 7'n" II. 1'27" 2'31" *'&" 7'30" The following table and Fig. 2, give the results obtained with ten samples of smokeless and two samples of nitrocellulose. The samples were obtained through the Navy Department and I am indebted to the courtesy of G. W. Patterson, Powder Expert at Indian Head, for the selection of three classes, good, fair, and poor; and for the description of these samples, which I quote for com- parison with the explosion periods. Nitrocellulose A. "Specially prepared. Heat test, potassium-iodide starch, at 65.5C., 4 min.; German test at 135C., 9 min. for litmus red." Nitrocellulose B. "Heat test, potassium-iodide starch at 65.5C., 42 min., German test at 13. 5C., 38 min., for litmus red." Powder Sample A. "Six-pounder Diphenylamine as stabilizer. German test at 135C., litmus red, 2 hrs. 35 min., explosion 5 hrs. plus. Surveillance test at 80C. ; 87 days; at 65.5C., 307 days." Sample B. "Medium caliber. Diphenylamine as stabilizer. German test at 135C., litmus red, 2 hrs. 17 min., explosion 5 hrs. plus. Surveillance test at 65.5C., 271 days." Sample C. "Large caliber. Diphenylamine as stabilizer. German test at 135C., litmus red 1 hr. 25 min., explosion, 5 hrs. plus. Surveillance test at 65.5C., 245 days plus." Sample D. "Large caliber. No stabilizer. Rosaniline as indicator. Ger- man test at 135C., litmus red, 2 hrs., explosion 5 hrs. plus. Surveillance test at 65.5C., 60 days." Sample E. "Medium caliber. No stabilizer. Rosaniline as indicator. German test at 135C., litmus red, 1 hr. 35 min., explosion 5 hrs. plus. Sur- veillance test at 65.5C., 74 days." Sample F. "Six-pounder. Contains rosaniline and diphenylamine. Ger- man test at 135C., litmus red, 2 hrs. 25 min., explosion 5 hrs. plus.|;Sur- veiUance test at 65.5C., 375 days, at 80C., 64 days." Sample H. "Manufactured in 1901. When last tested it gave German test at 135C., explosion in 41 min. Surveillance test at 65.5C. (1907) 79 days." Sample I. ' 'Manufactured about 1901. When last tested it gave German test at 135C., explosion in 33 min., Surveillance test at 65.5C, 82 days." Samples K & L. "These are both in very poor condition, giving only one day Surveillance test." 152 Original Communications: Eighth International [VOL. FIGURE 2 IV] Congress of Applied Chemistry 153 The explosion periods obtained on these samples are as follows. The value underscored is the average value : EXPLOSION PERIODS 200C. 180C. 170C. 160C. Nitrocellu- 33" 5'20" over 30' over 30' lose A. 58" 611" 37" 6'56" 43" 619" B 33" 4'06" 16' over 30' 42" 5'55" 17'30" 34" 6'06" 16'45" 36" 5'22" Powder 1'44" 1'45" 2'58" 4'30" 14'20" A 1'48" 1'46" 3'06" 417" 17'40" 1'44" 310" 4'28" 17'34" 1'48" 3'04" 4'25" 16'36" B 1'54" 1'41" 3'37" 3'27" 5'23" 5'44" 14'20" 1'15" 1'37" 3'54" 3'57" 5'26" 6'02" 13'45" 1'65" 1'27" 3'44" 3'44" 5'36" 5'53" 17' 7" 1'32" 1'25" 5'28" 5'57" 14'57" I /qq// 1 'W J. OO J- OU 168 C 2'04" 4'05" 617" 17'45" 2'04" 315" 6'28" 18'25" 210" 4'50" 6'55" 1818" 2'06" 4 / 05" 6'34" 18'09" 4 / 20" 615" 7'00" D 1'45" 2'10" 4'03" 5'53" 6'50" 19' 1'47" 2'30" 4'08" 5'57" 6'56" 26'49" 2' 2' 410" 6' 6'56" 28'30" 1'58" 24'46" 172 171 E 1'35" 210" 3'55" 610" 1015" 2'10" 210" 4'06" 5'58" 10'24" 210" 418" 6'08" 10'29" 2'05" 4'06" 6'05" 10'23" F 1'40" 1'45" 2'47" 3'33" 1516" 1'41" i'38" 2'45" 3'23 r/ 15'47" 1'48" 1'43" 2'49" 3'31" 17'40" 1'45" 2'47" 3'29" 1614" 154 Original Communications: Eighth International [VOL. EXPLOSION PERIODS H 115" 1'38" 3'32" 4'45" 6'47" 1'29" 1'22" 3'35" 417" 6'48" 110" 1'28" 3'28" 4'44" 6'50" i 'i e;" o/oo// AfOKff R'AQ" I 1 1O 2'04" 2'05" O d O 1o 3'05" 3'50" 613" 210" 2'05" 3'05" 3'49" 6'00" 2'02" 2'04" 3'20" 3'56" 617" 2'00" 310" 3'50" 610" K 2'05" 1'50" 3'29" 5'55" 1112" 12'28" 2'05" 1'52" 3'30" 5'22" 11'8" 13' 214" 2'04" 3'50" 615" 12'38" 11'54" 217" 3'36" 5'51" ll'OO" L 1'31" 1'30" 213" 3'38" 1'22" 1'40" 218" 3'33" 1'27" 1'31" 2'22" 3'47" 1'35" 218" 3'39" K at 183 3'20", 3'29"; at 160 12'43", 13'20" F at 178 3'00", 313" A at 199 1'47"; at 198, 1'53"; at 197 1'52"; at 195 I r 56" at 193.5 212. The curves embodying these results are given in Fig. 2. The curve marked "theoretical curve" is obtained on the assumption that the reaction velocity doubles for every 10C. While the curves obtained from the explosion periods are of the same general type, it is apparent that the relation is not so simple as that shown by the "theoretical curve." What influence the stabilizer has upon the direction of the curves it is difficult to say with the data at hand. It does not follow that the stabilizer will affect the explosion test in the same manner as it does the heat test or the surveillance test. The stabilizer, while it removes the products of decomposition, may of itself act as a positive or negative cata- lyzer. If the decomposition be considered as the dissociation of an ester, the presence of a substance removing the products of de- composition will increase the rate. This has been noticed by Mit- tasch 1 for nitrocellulose with various basic additions and has been confirmed by myself in the case of pyroxyline plastics containing zinc oxide, cit. iv] Congress of Applied Chemistry 155 That the curves do actually represent the stability of the powder with changing temperature and are not accidental is shown by the reproducibility of the various points along the curve, as shown in the table; by the fact that the curve having been deter- mined by four points, determinations made at varying tempera- tures are found to fall on the curve; and by the fact that the com- plete curve can be reproduced at widely varying intervals. See curves I/ and L". The curves are undoubtedly characteristic for the sample. The deviations for individual points are not large enough to affect the general trend of the curve. What temperatures are chosen must probably be left to individual requirements. 190 would perhaps be preferable to 200. At 150 we should have the apparent advantage that the differences between various samples become greater and the disadvantage that the test becomes slower and the results are more subject to accidental influences. The bend of the curve between 180C. and 160C. is, perhaps, its most characteristic portion. The curves fall into three distinct groups for the samples tested, corresponding to their general classification of good, fair, and bad. The stable powders have a pronounced bend, while the ratio of explosion periods at 200 and 160C. is at least 2:9. In the unstable samples this ratio falls as low as 2:3, and the points do not fit a smooth curve so well. The curves for the two samples of raw nitrocellulose are somewhat peculiar, being much flatter and corresponding more nearly to the theoretical curve. The powders do not always fall in exactly the same order by this explosion test as they do by the surveillance or the heat test, but I think this is true to the same extent for the 135 German explosion test and the ones mentioned. This appears by com- parison of the customary tests on samples D and H. D. German test 135 Litmus red 2 hrs., explosion 5 hrs.; sur- veillance 60. H. German test; explosion 41 min.; surveillance 79. From the results given it is evident that one explosion tempera- ture, even if time is considered, does not give much information while the determination of the characteristic curve does yield definite and specific information. On account of the complexity of the conditions the test can hardly be expected to tell all that is to be known, but I believe that with sufficient data it may even be made to throw some 156 Original Communications: Eighth International light on the actual effect of the stabilizer on the natural decompo- sition velocity of the powder, as distinguished from the length of time before the decomposition products become noticeable. The proposed method gives more accurate determinations of the explosion temperature than the method of heating with rising temperature. It gives a better comparison of the relative stability of explosive substances. The test is in effect a determination of the rate of change of decomposition velocity with change of tem- perature and is as such characteristic for each sample. LITERATURE REFERENCES Aspinwall. Stability Tests for Smokeless Powder and Nitro-explosives. Jour. Soc. Chem. Ind., 21, 687. Bergmann & Junk. Stability of Nitrocellulose. Z. Ang. Chem. 17, 17, 982, 1018, 1074. Cullen. Note on the so-called "Heat Test" for Explosives. Jour. Soc. Chem. Ind., 20, 8. * Escales.-^Stability of Nitrocellulose. Zeit. Angew. Chem. 18, 940. Meth- ods for Testing Stability of Explosives in Various Countries. Z. ges. Schiess- und Sprengstoffwesen, 5, 21, 72, 210. Finzi. Ignition Points of Nitrocellulose and Smokeless Powders. Gazz. Chim. Ital., 39, 1, 549. Jacque". German Railway Administration: Tests and Regulations of the. Z. ges. Schiess-u. Sprengstoffwesen, 4, 175. Causes and Methods of Determining Decomposition of Nitrocellulose. Z. ges. Schiess-u. Sprengstoffwesen, 1, 395. Lunge & Beibie. Contributions to the Knowledge of Nitrocellulose. Z. Angew. Chem., 14, 543, 561. Mittasch. Stability of Nitrocellulose. Z. Angew. Chem., 16, 16, 929. Obermiiller. Mitteil. aus dem Berliner Bezirks-verein, etc., Oct., 1904. See Wilcox, J. Am. Chem. Soc., 80, 271. Patterson. Stability Tests of Smokeless Powder. 7th International Cong. Applied Chem. Explosives, p. 99. Pleus. Some Improvements in the Apparatus for the Obermiiller Manometer Test. Z. ges. Schiess-und Sprengstoffwesen, 5, 121. Robertson. On the Will Test for Nitrocellulose, J. Soc. Chem. Ind., 21, 819. Rubin. Testing Regulations and Black Powder Safety Explosive in England. Z. ges. Schiess-u. Sprengstoffwesen, 4, 21. Saposhnikov. Rate of Decomposition of Nitrocellulose and Temperature. Russ. Phys. Chem. Soc., 38, 1186. Snelling & Storm. Behavior of Nitroglycerine when Heated. Bureau of Mines, Technical Paper 12, 1912. Sy. Stability of Nitrocelulose, J. Am. Chem. Soc., 25, 549; Z. Angew. Chem., 18, 1824. Wilcox. Decomposition Curves of Nitrocellulose. J. Am. Chem. Soc., SO, 271. Will. Stability of NitroceUulose. J. Soc. Chem. Ind., 20, 602; Stability of CeUuloid. Z. Angew. Chem., 19, 1386. Zschokke. Testing of Explosives. Z. ges. Schiess-u. Sprengstoffwesen, 6, 241. (Published by permission of the Navy Department) A NEW STABILITY TEST FOR NITROCELLULOSE POWDERS BY S. A. WEIRMAN U. S. Navy Ordnance Laboratory, Olongapo, P. I. The various stability tests for nitrocellulose powders in general use at the present time, depend, with various differences of details, upon one process: the subjection of the powder to temperatures sufficiently high to decompose the powder within a reasonable period of time, and from the result thus obtained to pass judg- ment upon the keeping qualities of the powder at the lower tem- perature of actual storage and use, i. e. the "working temperature" of the powder. While experience has proven the value of these tests as criteria of the actual stability and possible life of powder, yet so many other and complicated factors enter into the problem that it is by no means possible to consider increased temperature as the sole cause of lowered stability and therefore apply a mathematical proportion between the duration of test of a powder, even at a temperature but slightly higher than the working temperature, and the life of a powder at the working temperature. Obviously, the closer the test approaches the working temperature of the powder, the more accurate will be the test, but, for the same reason, the time consumed in concluding the test increases so rapidly as to soon place a limit on the minimum temperature which can be used to complete a test within a reasonable period. In the opposite direction it is evident that a test at any tem- perature sufficiently high to volatilize or decompose any of the constituents of the powder cannot be compared with any great degree of accuracy to the resistance of the powder at a working temperature where such decomposition will not occur. In one instance there is a rapid decomposition caused by the breaking down of a constituent of the powder and the accompanying acceleration of decomposition caused thereby, while in the other 157 158 Original Communications: Eighth International [VOL. instance there is but the slow gradual decomposition of the powder in its entirety. It must be acknowledged that all heat tests approach the problem of stability from but one direction; that of decomposition by heat. The 65.5C. Surveillance Test; the 65.5C. and 80C. Kl-Starch Test; the 110C. Vieille Test; the 115C. Ordnance Test and the 134.5C. German Test differ in the main only by difference in temperature employed. With these things in mind and realizing the value of a corrob- oratory test approaching the problem of stability from a different direction than that of decomposition by heat, the attention of the writer was directed to the question of moisture effect on nitrocellulose powders during some years' experience with these powders in the warm, moisture-laden climate of the tropics, where the pronounced effect which moisture had upon the keeping qualities of powder and the tests of the same was very evident. Various experiments and experience with powder stored in warm, damp magazines both ashore and aboard ship proved conclusively that while moisture in powder is most deleterious, yet the presence of the moisture may be regarded as an effect and proof of low stability rather than a cause, and therefore the amount of moisture present, or rather the ability of the powder to take up and retain moisture, serves as a fair criterion of the keeping qualities of the powder. This may be explained as follows: A newly made nitrocellulose powder consists of nitrocellulose of certain limits of nitration dissolved into a plastic mass by a solvent, generally of ether- alcohol. After the pressing and drying processes there remains a hard colloid containing a slight amount of residual alcohol from the ether-alcohol solvent, the amount depending principally upon the size and shape of the powder grains. Of moisture there is but the very small amount originally in the solvent and remain- ing with the residual alcohol incorporated with the powder grain in addition to the surface moisture deposited on the powder during the process of packing. In the course of years of storage and use with more or less exposure to changes of temperature and humidity, the residual solvent slowly evaporates and is replaced by a deposit of moisture. rv] Congress of Applied Chemistry 159 This moisture probably produces a hydrolytic decomposition and serves to accelerate the progress of decomposition within the powder grain. As the organic ingredients of the powder gradually break down additional H 2 O is formed, an accurate measure of which might serve as a true index of the actual condition of the powder. But at any time the amount of moisture present in a powder is a certain criterion of its stability as showing the porosity and vulnerability of the powder grain, and while density, residual solvent and size and shape of grain may all have their influence on the moisture content, yet for the purpose of a stability test these factors may be disregarded. The following test may then be developed as correctly indicating the stability of a powder independent of any heat test and approaching the problem from an entirely different direction. A suitable powder sample, ten grams are sufficient, is placed in a weighing-bottle and weighed. Then placed in a sulphuric or vacuum desiccator for a suitable period. Seven days have been found to be fully sufficient for complete drying. Sample then weighed and loss of weight obtained. Expose sample in saturated atmosphere, preferably each test at the same temperature, until equilibrium is established. Forty-eight hours in a Hempel des- iccator containing water has been found sufficient. Obtain gain in weight. Loss in weight plus gain in weight, expressed in per cent of weight after desiccation equals moisture range. This range corrected for surface area moisture converts all size powder grains to same standard for comparison. This does not necessitate a measurement of the powder grains each time as the original dimensions of the grain when manufactured will serve. The observed moisture range "r" divided by surface area "a" equals actual moisture range "R". = R. This result will be found to a parallel the results obtained from the Kl-Starch Test and help to explain the sudden unaccountable drop in test experienced with some powders in the 65.5C. Surveillance Test after exposure to unfavorable hygrometric conditions. ORIGINAL COMMUNICATIONS EIGHTH INTERNATIONAL CONGRESS OF APPLIED CHEMISTRY Washington and New York September 4 to 13, 1912 SECTION IIIc: SILICATE INDUSTRIES VOL. V ORIGINAL COMMUNICATIONS EIGHTH INTERNATIONAL CONGRESS OF APPLIED CHEMISTRY Washington and New York September 4 to 13, 1912 SECTION IIIc: SILICATE INDUSTRIES VOL. V The matter contained in this volume is printed in exact accordance with the manuscript submitted, as provided for in the rules governing papers and publications. La matiere de ce volume a 6t6 imprim6e strictement d'accord avec le manuscrit fourni et les regies gouvernant tous les documents et publications. Die in diesem Heft enthaltenen Beitrage sind genau in tJbereinstimmung mit den uns unterbreiteten Manuskripten gedruckt, in Gemassheit der fur Beitrage und Verlagsartikel geltenden Bestimmungen. La materia di questo volume e stampata in accordo al manoscritto presentato ed in base alle regole que governano i document! e le publicazioni. THE RUMFORD PRESS CONCORDN-HU'SA ORIGINAL COMMUNICATIONS TO THE EIGHTH INTERNATIONAL CONGRESS OF APPLIED CHEMISTRY APPROVED BY THE COMMITTEE ON PAPERS AND PUBLICATIONS IRVING W. FAY, CHAIRMAN T. LYNTON BRIGGS JOHN C. OLSEN F. W. FRERICHS JOSEPH W. RICHARDS A. C. LANGMUIR E. F. ROBBER A. F. SEEKER SECTION IIIc. SILICATE INDUSTRIES. EXECUTIVE COMMITTEE. President: ALLERTON S. CUSHMAN, PH.D. Vice- President: KARL LANGENBBCK Secretary: ZOLTAN DE HORTATH EDWARD ORTON, JR., E.M. HEINRICH RIBS, PH.D. SECTIONAL COMMITTEE. L. E. BARRINGER, E.M. S. V. PEPPBL CHAS. F. BINNS, M.S. R. C. PURDY A. V. BLEININGER, B.S. CLIFFORD RICHARDSON, F.C.S S. G. BURT, A.B. KENNETH SEAVER, S.B. ELLIS LOVEJOY, E.M. ALEXANDER SILVERMAN, M.S. RICHARD K. MEADE, M.S. F. W. WALKER S. B. NEWBERRY, Pn.D. ARTHUR S. WATTS L. W. PAGE H. A. WHEELER, E.M. and the Sectional Executive Committee. VOLUME V. SECTION IIIc: SILICATE INDUSTRIES. CONTENTS. BINNS, CHARLES F. AND MAKELEY, C. H. The Coloring Power of Iron Compounds in Burned Clay 7 BLEININGER, A. V. The Effect of Electrolytes upon Clay in the Plastic State 17 COGGESHALL, GEORGE W. AND CtJSHMAN, ALLERTON S. The Production of Available Potash from the Natural Silicates . . 33 CUSHMAN, ALLERTON S. AND COGGESHALL, GEORGE W. The Production of Available Potash from the Natural Silicates . . 33 CUSHMAN, ALLERTON S. Notes on a Study of the Temperature Gradients of Setting Portland Cement 51 FRZNK, ROBERT L. Causes of Breakage in Glass Manufacture and Method of Differ- entiating Chemico-Heterogeneic Strains from Cooling Strains .... 57 HADLEY, HARRY F. AND MCFARLAND, DAVID F. The Use of the Higher Phenols in Testing for Free Lime in Port- land Cement 83 KLEIN, A. A. AND PHILLIPS, A. J. Magnesia in Portland Cement 73 MAKELEY, C. H. AND BINNS, CHARLES F. The Coloring Power of Iron Compounds in Burned Clay 7 MCFARLAND, DAVID F. AND HADLEY, HARRY F. The Use of the Higher Phenols in Testing for Free Lime in Port- land Cement 83 PHILLIPS, A. J. AND KLEIN, A. A. Magnesia in Portland Cement 73 REIBLING, W. C. AND REYES, F. D. The Physical and Chemical Properties of Portland Cement 91 REYES, F. D. AND REIBLING, W. C. The Physical and Chemical Properties of Portland Cement 91 SCHMIDT, WALTER A. The Control of Dust in Portland Cement Manufacture by the Cot- trell Electrical Precipitation Processes 117 6 Contents [VOL. SILVERMAN, ALEXANDER Glass Formulas A Criticism 125 STALET, HOMER F. The Viscosity ofBorate Glasses 127 THE COLORING POWER OF IRON COMPOUNDS IN BURNED CLAY CHARLES F. BINNS AND C. H. MAKELEY Alfred, N. Y. The phenomena which attend the production of color in un- glazed clay wares have attracted the attention of many investi- gators. It has been shown that iron is the most prolific colorant and that a great variety of hues are produced by it. These hues are influenced by the amount of iron present, by the form in which the iron occurs, by other ingredients in the clay and by the nature and temperature of the burning. The subject was first investigated by Hermann A. Seger whose collected works have been translated by the American Ceramic Society. He says 1 " It is known that the color goes through all the variations from white through yellow, orange, red, blue, brown up to black; all these colors are produced by the combi- nations of iron "; and again, 2 " It is known that the oxide of iron, which in the majority of cases must be considered the only color- ing constituent, may produce a great variety of shades from yellowish red to violet black, according to its state of division, and it will always assume a darker color when the clay is exposed to a high temperature." The normal burning of a red clay takes place under oxydizing conditions, hence the iron will always ultimately become ferric oxide unless it is partly combined as a silicate. Nevertheless the original source of the iron seems to be of profound significance in its influence upon color. Prof. Orton says " The ferruginous minerals which are most commonly found in clays are: (a) Ferric oxide, anhydrous or in various stages of hydration. (b) Ferrous carbonate. (c) Ferric sulphide or Pyrite. Collected Writings, Vol. I, p. 343. 'Ibid, p. 349. 8 Original Communications: Eighth International [VOL. (d) Ferrous silicate minerals, like Biotite, Hornblend, and many others. (e) Ferric sand minerals, like Magnetite, Menaccanite, Chromite, etc. It is from the first three minerals 1 that we derive the principal beneficial or detrimental effects of iron." And again, " The effect of these various minerals on the color of clays is not in itself strongly marked or characteristic. That is, as good a red color may be developed from a clay containing its iron as ferrous carbonate as from ferric hydroxide." " Ferric sulphide 2 is invariably found in granular form. These gran- ules may be large or small but they are never small enough to give a red color." Certain opinions have been advanced also by the above quoted writers upon the influence of the other ingredients of the clay upon the color of the iron. It is admitted on all sides that lime exerts a powerful effect in producing a buff color in a clay which otherwise would burn red. Magnesia has a similar influence but somewhat less marked. The substance upon which there is a difference of opinion is alumina. Seger says, "If we wish to classify the clays according to the colors which the mass assumes on burning 3 we can divide them into four groups: 1. Clays high in alumina and low in iron. These burn white or to a scarcely noticeable color. 2. Clays high in alumina containing moderate amounts of iron ; their color ranges from pale yellow to buff. 3. Clays low in alumina and high in iron, the brick clays, burn- ing red. 4. Clays low in alumina, high in iron and lime, the brick clays burning yellow, or clay marl/' Prof. Orton (p. 389) " cannot give unconditional assent. The evidence Seger marshalls is strong and undoubtedly points to his conclusion. But if it were true, then synthetic mixtures should 1U On the Role played by Iron in the Burning of Clays." Transactions of the American Ceramic Society, Vol. V, p. 384. 2 Ibid, p. 387. 'Collected Writings, p. 109. v] Congress of Applied Chemistry 9 easily give the buff color, which has not been the writer's ex- perience. We certainly find an amazing uniformity in the color of buff burning clays, while their iron-alumina ratios fluctuate very greatly. Some fire clays contain 40 per cent, of alumina and 0.5 per cent, of iron, and burn to a good buff. Others contain 15 or 20 per cent, of alumina and 2.5 to 4 per cent, of iron and burn to almost exactly the same tint. Other clays of different geolog- ical history, containing about the same alumina and iron burn to a fine red color." The present investigation was undertaken with the view of throwing some light upon the case and is an attempt to ascertain the effect of the most common sources of iron under the influence of silica and alumina respectively. The influence of temperature was also considered so that the tests were prepared in duplicate, one series being burned at 1200 degrees C. and the other at 1270 degrees C. These are high temperatures for red burning clays but the mixtures made were very much more refractory than would be the case with natural clays. The foundation of the mixtures is an English plastic ball clay which, together with the iron, forms 40 per cent, of the mass. The remaining 60 per cent, is ground quartz in the first member and pure alumina in the last. The intermediate members contain 40 and 20 per cent, of the quartz and 20 and 40 per cent, of the alumina respectively; the iron content, calculated as ferric oxide, is 2.5 per cent, in the first line, 5 per cent, in the second, and 10 per cent, in the third, these amounts being subtracted from the ball clay so that the iron content is constant in a hori- zontal direction and the other ingredients are constant in a ver- tical direction. The first series was made up with commercial ferric oxide, the whole mixture being ground together in water in a porcelain ball mill. An inspection of the results shows that no red color can be expected from this source. The prevailing tone is a pinkish gray, the color being somewhat lightened as the content of alumina increases. At the lower fire the alumina produces no change in hue but simply a lighter tint. This is probably due to the fact that the alumina is more bulky than the quartz and, consequent- ly, the whitening effect is greater. At the greater heat, however, 10 Original Communications: Eighth International [VOL. the tone of the color is changed, as the alumina increases, to buffs of varying strength. This is especially the case at the 5 per cent, iron content, though it is apparent in every instance. In the second series the iron was introduced as a precipitate, thus stimulating the effect of limonite. A solution of ferrous sulphate was prepared and the volume necessary to contain the proper amount of iron was added to the clay mixture after grind- ing. Sufficient ammonia to precipitate the hydroxide was then mixed with the fluid which was then washed free from ammonium sulphate. It was not found necessary to oxydize the solution in advance as the clay speedily changed from green to orange from atmospheric influence. The results are interesting. At the lower fire a number of reds appear. The alumina shows the lightening effect already mentioned but no actual change of hue. At the higher temperature, however, there are some marked changes especially in the 5 per cent. iron. Here are two excellent buff colors showing that alumina has a marked effect when the heat is sufficient to cause it t react. It appears that alumina is powerless to overcome the red tones when the iron stands as high as 10 per cent. The third series contains the iron in the form of siderite, FeCOs. A supply of the mineral was obtained and the iron content ascer- tained. The proper quantity was then ground with the clay. At the low fire the mineral has no tendency to produce a red or even a brown color. Nothing appears but various shades of gray. At the higher temperature the prevailing color is brown, in- fluenced towards gray on the increase of alumina content. In the fourth and last series pyrite was the source of the iron. The procedure was the same as in the third series. Some interest- ing developments have arisen from the fact that the pyrite has oxydized in part as the clay samples were slowly dried. In this the result may be regarded as a combination of series 1 and series 2. Where the clay has dried rapidly as on the edges the iron appears as the anhydrous oxide but where a previous oxidation has taken place the effect appears as though the hydrous oxide had been formed. At the higher temperature there is an inter- esting array of buff colors, especially at the 5 per cent, iron content. y] Congress of Applied Chemistry 11 There are certain points which appear evident from these experiments. 1. It is not possible to produce red colors in burned clay by the use of pulverized iron bearing minerals, however finely they may be ground, but buff tones are produced under the influence of alumina and at a temperature at which the clay approaches vitrification. These buff colors are apparently due to the blend- ing of a multitude of minute brown specks. 2. Red colors are the result of a precipitation of a colloidal iron compound in the clay mass. This precipitation apparently results from a solution of ferrous sulphate, which is itself the result of the oxidation of pyrite, either becoming oxydized with the separation of limonite or meeting with carbon dioxide in some form with the resulting precipitation of ferrous carbonate. This is the only way of explaining the statement of Prof. Orton, quoted above, that " as good a red color may be developed from a clay containing its iron as ferrous carbonate as from ferric hydroxide." Siderite does not decompose under ordinary conditions and in the finely ground form no red is produced. 3. Pyrite is responsible for several phenomena. As already stated it is the parent of other forms of iron and while it is true as stated by Orton that the granules of pyrite " are never small enough to produce a red color ," it is also true that pyrite is extremely susceptible of oxidation. Unless the clay containing this mineral is dried very rapidly ferrous sulphate and ultimately ferric hydroxide will be found. There are examples of this in the specimens shown; in fact, in these there is the actual birth of a red clay. 4. Alumina is undoubtedly responsible for the production of buff colors and in this the opinion of Seger is confirmed. At the same time it must be admitted that the effect of alumina, added, as in these experiments, in the pure form, may be different from that of combined alumina. Possibly the different behavior of clays with a similar content of alumina and iron may be ac- counted for in this way. I. Fe,( 5.< FPaO. 10!$ PLATB FT Pe,< ?.' 10. FERRIC OXIDE Cone 3 1200* C. Si0 8 60 Si0 2 40 SiO a 20 SiO, Ai^O. Al0, 20 AlaO, 40 Al,0, -60 Gray Gray Light Gray Light Gray Pinlcleh Gray Gray Light Gray- Light Gray tight Pink Chocolate pinkish Gray Gray Light Gray FERRIC OXIDK Cone 5 IS^ S10 2 50 Si0 2 40 Si0 8 20 SIO, A1 2 3 Al a O a ?0 Al 2 a 40 Al,0, 50 arayleh Buff Buff Light Bftff Cream Dark Pinkish Gray Dark Buff Light Buff Cream Pink Chocolate Brown Grayish Buff Dark Buff PLATS III. 10. ( PLATS TV 10. 0# B R R I C HYDROXIDE Cone 3. ' 1200* C. S10, 60 Si0 2 40 810, 20 810 t Al,0 a A1 8 0, 20 A1 8 0, 40 Al0, 60 Pinkish Red Light Red Pinkish Buff Light Pink Buff Bright Bed Bright Red Pale Red Light Pink Strong Red I Bright Red Bright Red Pale Bed B R R T C HYDBOXTDB Cone 6. 12?0"C. SiO, 60- S10, 40 810, 20 Sl0 8 - Al,0 8 Al,0 3 20 Al,0, 40 Al a 9 50 Pinkish Red Strong Buff Buff Light Buff Medium Bed Light Red Strong Buff Buff Dejrk Red lledlun Red Red Red PLA22S..V. BIO, 60 Ai 8 0, - SIDBRITB Cone 3. 1200" C. S10 8 40 Al a 8 20 SiO a 20 40 A1 8 8 60 Cray Gray Medium Gray Light Gray Dark Gray Dark Gray Gray Gray Strong Gray Dark Gray Dark Gray Gray Pe a < 2.', Pe 2 0j io!o;l SlOa 6Q A1 4 3 -* SIDBRITB Cone 6. 1270*0. Si0 2 40 Al,0, 20 20 A1 2 3 40 SiO* A1 3 3 60 STOWQ Brown Buff Light Brown Light Brown Dark Brown Brown Brown Gray i Black Dark Brown Brown Dark Gray PL&EX VIZ. 810, 41,0, Gray Fed Pinkish Gray Purple Dark Brown PLATE VIII. . a . 10.0% t^t R t t Z Co no. 3. 1200* C. 40 20 810, 30 Al a O s 40 Pinkish Gray Purple Perk Red Orar Gray Red Gray Red \ > ^ B I T E Cone 6. 1270* C SiO- 60 Gray Gray Bed Gray Red SiO a 60 Si0 2 40 Si0 2 30 6J0 3 Al n O^ Al.O. 20 A1^O_ 4O A1_O- fin Dark Buff Buff Buff Light Buff Purpla Brown BuTf Buff Buff Dark Brown Red Purple Brown Purple Brown Purple Brown THE EFFECT OF ELECTROLYTES UPON CLAY IN THE PLASTIC STATE BY A. V. BLEININGER Ceramics Department, University of Illinois, Urbana, Illinois It is a well known fact that clays, suspended in water, are affected in a most decided manner by the presence of electrolytes. This susceptibility of clay slips is illustrated in a striking manner by the use of sodium carbonate and sodium silicate in the casting process where small amounts of these reagents bring about a considerable reduction in the amount of water required to cause fluidity and consequently in the drying shrinkage. According to the opinion of many writers, this and similar phenomena are con- nected with the colloidal nature of certain constituents of clays. Thus Schloesing 1 separated from kaolin, 1.47 per cent, of a volum- inous material which he called colloidal. Van Bemmelen, 2 Rohland, 3 Cushman, 4 Ashley 5 and others ascribe such properties as the plasticity of clays, the cementing power of ground rock, etc., to the content of colloid matter. That clays contain material of a colloid character seems to have been proven conclusively. The amounts of such substance, however, are not as great as has been supposed and the general properties of the clays are governed to a very considerable extent by the large quantities of granular matter usually present. The latter occurs in all sizes from coarse grains down to particles so small that they show Brownian movement. It is obvious that only the finest material remaining in suspension is affected by the J The Constitution of Clays: Compt. Rend., Vol. 79, 1874, pp. 376-380, 473-477. 2 Chem. Zentralblatt (1882) p. 1255; z.f.anorg. Chem. 18, p. 14 (1898); 23, p. 321 (1900). 3 Zeitschrift f. anorg. Chem., 1902, 158; 1904, 325. 4 J. Am. Chem. Soc. 1903, 451; Trans. Am. Ceram. Soc. 6, 65. 'Bull. 388, U. S. Geol. Survey. 2 17 18 Original Communications: Eighth International [VOL. presence of electrolytes and other conditions. As is well known, the behavior of suspended particles is governed by the slight quan- tities of electrolytes present in solution. The direction in which suspension colloids travel under the influence of an electric cur- rent is determined principally by the small amount of dissolved substance and is not so much a function of the kind of colloid itself. Thus a slight increase in the hydrogen ions in water impart to a hydrosol a positive charge so that its particles move towards the negative pole. On the other hand, an addition of hydroxyl ions brings about a negative charge causing the particles to move towards the positive pole. The migration velocity of all hydrosols is practically the same and according to Whitney and Blake, they possess about the same velocity as the mono valent ion of an inorganic salt. The addition of ions possessing a charge opposite to that of the colloid particles causes them to precipitate or to coagulate. Hence a negatively charged colloid is coagulated by a positive ion. Ions of the same sign either do not effect coagula- tion or tend to keep the sol in its original condition. The coagulation capacity of an ion is proportional to its valency. A bi valent ion has a greater coagulation capacity than a mono valent one in the ratio of y/ 2 =1:6 :1. A smaller amount of electrolyte is required if the reagent is added at once than when it is introduced in smaller portions. Oppositely charged hydrosols coagulate each other. The flocculating colloid invariably carries part of the electrolyte down with it and it seems to prefer the ion which has the same charge. In the case of a clay suspension we may then assume that each particle is surrounded by an adsorption envelope which is formed by an interchange with the dispersion medium. Flocculation consists in the consolidation of the particles after the addition of the electrolyte. In order that the particles may unite, the thin envelope of liquid must be broken. This does not take place simultaneously but slowly. As soon as the particles have become large enough for the Brownian movement to cease, coagulation becomes more rapid. If the particles are uncharged they adhere most readily and hence the soil is least constant. A kaolin suspension is very sensitive to impurities. The charges of the kaolin particles are very slight and according to v] Congress of Applied Chemistry 19 Bodlaender negative in character. 1 They are coagulated by di and tri valent cations more strongly than by mono valent ones. Bases have a high coagulating value referred to the clay particles while non electrolytes do not precipitate. Acids coagulate more strongly than di valent cations. On the other hand, OH'ions possess a strongly de-flocculating character. 2 Ashley has sum- marized the principles underlying the effect of various electrolytes and non electrolytes upon clay suspensions quite satisfactorily. 3 Giving our attention to some of the practical applications of these conceptions, it was found by Simonis 4 and Boettcher 5 that the best condition of a slip for casting is that corresponding to maximum deflocculation which at the same time, shows mini- mum viscosity. Ashley was inclined to think that a point just short of this state might give best results. The plastic state might be considered as a special case of clay in suspension, in which the particles approach each other so closely that cohesive forces come into play. During the process of drying, the particles approach each other more and more closely until they come in contact. The water which has been evaporated up to this point is termed shrinkage water. The remaining water content filling the interstices between the clay particles is called pore water. This term includes also part of that portion which upon heating the clay to 110 still persists in adhering to the mass and which is expelled only at still higher temperatures, the last traces being driven off only at red heat, and known as hygroscopic moisture. It has been realized for some time that the properties of the clays are influenced by the presence of electrolytes. Seger 6 explains the increase in the plasticity of clay upon storing by the assumption that the fermentation of organic substances results in acids which neutralize the alkalinity due to decomposed ^ottinger Nachrichten, 1893, p. 267. 2 Freundlich, Kapillar Chemie, chapter on suspension colloids. 3 Trans. Am. Ceram. Soc. 12, p. 768. 4 Sprechsaal 1905, No. 31. 'Sprechsaal 1909, Nos. 9-17. Tonindustrie Zeitimg 1891, p. 813. 20 Original Communications: Eighth International [VOL. feldspar, and in addition bring about the desired " sour " condi- tion which accompanies the improvement in the working qualities. Rohland 1 discusses this subject at length and comes to the con- clusion that the plasticity of clays is increased by the presence of H ions, the OH' ions on the other hand being active in the opposite sense. Other means of accomplishing the same result consist in the addition of colloids like tannin, dextrine etc., as has been shown by the work of Acheson, and the storing of the clay in cool and moist places. Coagulation is coincident with increase in plasticity and is primarily due to the presence of hydrogen ions; it is retarded by the hydroxyl ions. The salts of strong bases and weak acids which dissociate OH' ions hydrolyt- ically produce an effect similar to that of the hydroxyl ions. Neutral salts, Rohland goes on to say, with but few exceptions, are indifferent in their effect though some appear to show a contradictory behavior which has not been explained. " The effect of the hydroxyl ions may be weakened, compensated or strengthened by the cation of the salt in question. Thus borax is an example of the first class and Na2CO 3 of the second." The same author further says that with some clays the addition of Na2COs brings about an improvement in plasticity while ordinarily the same reagent behaves in the opposite way due to the hydrolytic dissociation of OH' ions. It is possible that the effect of the OH 'ions might be compensated by the presence of the CO 3 " ions. The perusal of literature dealing with this subject shows a decided lack of determinations resulting in numerical values, in fact, practically no results are available as far as the effect of various reagents upon clays in the plastic state is concerned. For this reason, it was thought desirable to begin the study of certain electrolytes as regards their action upon some well known plastic clays without reference to theoretical considerations. The most obvious criterion to be used in this connection is the shrink- age in volume which clays undergo in being dried under constant conditions. It is evident that any effect caused by the addition of reagents to clay will be at once indicated by the shrinkage. Tone, pp. 35-49. v] Congress of Applied Chemistry 21 From what we know of the practical properties of clay, we are justified in assuming that shrinkage is a function of plasticity. In these series of experiments, two clays were employed, plastic Georgia kaolin and Tennessee ball clay. The method of procedure consisted in preparing a thoroughly mixed sample of the clay in question, weighing out sufficiently large portions and adding the desired amount of the salt in solution. After preparing a thor- oughly worked mass of the desired plastic consistency, it was stored in a moist chamber for 24 hours in order to make reasonable allowance for any time effect. The lump of clay was then molded in a brass form into bars 10 x 2.5 x 0.625 cm, which were at once weighed. The volume of these specimens was determined in a voluminometer connected to a burette which permitted of reading easily to 0.05 cc. The measuring liquid used was petroleum from which the lighter constituents had been driven off by long con- tinued heating. For each concentration of salt, three bars were made and measured. The specimens were then allowed to dry at the laboratory temperature for three days after which time they were heated in an oven regulated by a thermostat to 110, to constant weight. The dry bars were at once weighed and immersed in petroleum until completely saturated when they were placed in the voluminometer for the determination of the dry volumes. In order to establish the limits of the working consistencies of such clays, samples of three clays were carefully made up into bars representing the extremes of the plastic state, i.e., so dry at one end of the series that the mass could barely be worked and as wet at the other end as the clay would permit. The drying shrinkages were then determined as usual. The curves of Fig. 1 represent the results of this work. It is at once observed that the kaolins possess but a short range of working consistency, as is to be expected, and the middle point of each curve represents the best molding condition. In fact, for clays of this type the proper consistency is determined by the " feel " with considerable accm- racy. In the case of the ball clay, however, the range is far more extensive and the best consistency is not so readily made evident by the working of the plastic mass. But here likewise the middle point of the curve stands for the most satisfactory molding con- dition. The far greater amount of water required by the ball 22 Original Communications: Eighth International [VOL. clay and the accompanying large shrinkage as compared with the kaolins is characteristic of this type of material and differentiates it sharply. The amount of water required by each clay for the best working consistency was thus determined and maintained r *. i v^ 54 'Showing And Wati Curves Re/afto rCcmtef for Thr i &etwe< t When ee- Cfays n Volum* tfater /s ' ShrinHa Varied ie From / ^or Wot 'inimun king Cl 7 To Ma ly In 7 Ktmom , ha Plot p %rCent t/c 3tatt 5 2 s J?/ V s to ** in tfy / I. *& / PER 'CENT VOLUME < 5 * 5 K j cf / r$* A i? y V 3O 34 38 42 46 SO PERCENT HATER BY WEIGHT 0# DR Y WEIGHT Figure 1. constant in most of the subsequent series employing solutions of the various salts. In several series that consistency was main- tained which gave the best working condition, a procedure which introduces slight variations. The electrolytes used were NaCl, CaCl 2 , A1C1 3 , Na 2 SO 4 and BaCl2. The first three reagents were employed for the purpose of Congress of Applied Chemistry 23 determining the effect, if any, of valency. In Fig. 2, the curves showing the effect of the chlorides of sodium, calcium and aluminum upon the shrinkage of Georgia kaolin are given and it is noted that both the CaCla and A1C1 3 are more effective than the NaCl in MS W A50 .0625 .075 JO GRAMS ELECTROLYTE PER IOO GRAMS DRY CLAY Figure 2. amounts less than 0.04 per cent. With a content of 0.05 per cent, the sodium salt brings about the same decrease in shrinkage as smaller amounts of the other chlorides. Based on molecular equivalents, these reagents are effective in the order of the valency of the respective metals, the A1C1 3 being most active and the NaCl least in decreasing shrinkage. The interesting fact is 24 Original Communications: Eighth International [VOL. brought out that very small additions of any of these reagents, maintaining the same amount of liquid, cause a marked increase in the shrinkage above that obtained with the same volume of water. In each case however, a sudden drop takes place down to approximately the same shrinkage. The cause of the dual be- havior of the electrolytes is to be sought in dissociation phe- nomena. It was found that the clay in question possessed a dis- tinct acid reaction. In the case of Aids, for instance hydrolytic dissociation would tend to produce additional H'ions and conse- quently an increase in plasticity and in shrinkage. According to Rohland such chlorides as NaCl and CaC^ should be entirely neutral. This is not the case since all of the three salts in greater concentrations produced a measurable decrease in shrinkage. The use of NaCl is of special interest in this connection since experiments carried on in this laboratory showed that in the case of exceedingly plastic clays of tertiary origin the plasticity was greatly decreased by the use of salts solution which was strikingly demonstrated by their drying behavior. Brickettes made from the untreated clay cracked and checked very badly while speci- mens made up with a NaCl solution dried normally without the slightest evidence of cracking and at the same time possessed a greatly reduced drying shrinkage. In order to show to what extent the relation between shrinkage and pore water is affected by the use of these reagents the total and shrinkage water were calculated in terms of the true clay volume. For this purpose the density of the powdered clay was determined by means of the pycnometer under the usual precau- tions. The volume of the shrinkage water was then calculated from the evident relation: 100 (vi v 2 ) = per cent, shrinkage water (by volume) w d where Vi = volume of wet brickette 2 = volume of dried brickette w = weight of brickette dried at 110 d = density of clay. Similarly, the volume of the total water referred to that of the clay was computed. v] Congress of Applied Chemistry 25 In Fig. 3, this relation is shown graphically for the addition of NaCl. It is seen that the shrinkage water volume was first increased and then slightly decreased. The decrease is due to a drop|in the total water and a rise in the pore water curves. The A/ACL WITH GEORGIA KAOLIN 00 ^00 ~ V- S" 1 X * * d ^ Shr nkaqe \f ater Vo '(/me "^S, <^-~- ^ PERCENT VOLUMES IN TEKM3 OF TRL 8 $ & t & Pot e Water Volume. Tru z Ctay V olume. L /ne o$- e '.$ *5 040 .50 ofczs ^;S JO GRAMS N*Ct t>ft IOO GRAMS DRY CLAY Figure 3. effect of sodium chloride upon Tennessee ball clay is shown in Fig. 4, where again the shrinkage is first increased and later diminished by greater concentrations of the salt. In Fig. 5, the volume rela- tions are again indicated and it is observed that the small reduc- tion in shrinkage is due principally to the decrease in the amount 26 Original Communications: Eighth International [VOL. PERCENT VOL -^ ^^ -^ PRCLHT,H*Ct IN TERMS OF DRY TflHd$ BALL Figure 4. AY A SHRINKAGE. POKE. WATE: 100 GKAM3 Figure 5. Congress of Applied Chemistry 27 of total water. The relation of total water content and shrinkage water established for the clay when made up with pure water also was disturbed by the use of the reagents as is seen from the AM, WITH G5ORG/A KAOLIM PERCEMT VOLUMES- IN TERMS Of TRUE CLAY VOLUME J&.Stf* &4, considered by Rohland to be neutral, was to increase the shrinkage decidedly as is indicated in the curve PERCENT Nfi t SO, IN TERMS or SKY TNAl53eE BALL CLAY Figure 7. of Fig. 7, referring to Tennessee ball clay. The maximum shown in the first part of the curve cannot be explained satisfactorily at this time. From Fig. 8, it is seen that the greater shrinkage is due to the increase in shrinkage water at the expense of the pore water. The clay grains are thus separated more widely from each other in the plastic state but arrange themselves more compactly upon drying. The effect of BaCl 2 , upon Tennessee ball clay, Fig. 9, seems to be more complicated than the preceding cases. One well marked and a slighter maximum point are noted. Larger quanti- ties of the reagents bring about a second reduction in shrinkage v] Congress of Applied Chemistry 29 go B >o I 70 Water V 107 / too GRAV& TNN33E BALL CLA Y A78 Figure 8. AY VOLUME. c, Cirve PERCENT &aCl,.-2h t O IN TERMS Or DRY TENNESSEE BALL CLAY Figure 9. 30 Original Communications: Eighth International [VOL. followed -by a third increase after which the drying contraction appears to remain constant. From the volume relations, Fig. 9, it is apparent that the reduction in shrinkage is due to the diminu- tion of the amount of total water which, at the end of the series is counteracted, in part, by the conversion of pore water into shrinkage water. Shrink-aye Water Pore. Water ,. SI *| BACI t PE*R /OO GRAM3 OF TENNESSEE BALI. CLAY Figure 10. Summary: The relation between the water content and the drying shrinkage (by volume) of two kaolins and one ball clay was determined for a series of conditions ranging from an ex- tremely dry to a very soft consistency. Characteristic short curve were obtained for the kaolins and a long range for the ball clay. The effect of varying concentrations of the electrolytes NaCl, CaCl 2 , A1C1 3 , Na 2 SO 4 , and BaCl 2 upon the drying shrinkage of plastic Georgia kaolin and Tennessee ball clay was studied. It seems unlikely that the effects of electrolytes upon clays in the plastic state can be classified according to such simple theoretical assumptions as have been made by Rohland, owing to the com- plex conditions which involve the presence of salts in the natural v] Congress of Applied Chemistry 31 clay, the phenomena of adsorption, ionic and hydrolytic dissocia- tion and the constantly changing concentration during drying. The electrolytes studied in this work showed a tendency to increase the shrinkage and presumably the plasticity when pres- ent in small concentrations. Maxima and minima seem to occur in this connection showing that extremely small additions of reagents suffice to disturb the equilibrium. As the concentration of the electrolyte increases there is a distinct tendency to diminish the shrinkage with an absence of maxima and minima, excepting in the case of Na2SO4 which seems to increase contraction. The positive or negative effect of the electrolyte is hence a function of the concentration up to a limiting point above which any addi- tions produce no further marked change. Expressing the concentration of the electrolytes in moles it is found that the activity of the reagents in reducing shrinkage is of the order of their valences. The influence of these salts extends not only to the spacing of the clay grains in the plastic state but als to their structure in the dry condition, resulting in a looser or more compact arrangement. THE PRODUCTION OF AVAILABLE POTASH FROM THE NATURAL SILICITES BY ALLERTON S. CUSHMAN AND GEORGE W. COGGESHALL Washington, D. C. The great demand which has recently arisen for an American supply of potash in available form for agriculture, has stimulated not only the search for new sources of this material but also experiments on a large and practical scale of operation, in the attempt to develop a method of making the vast supply of potash locked up in feldspars and feldspathic rocks either directly water soluble or sufficiently soluble in dilute acids to insure a product which shall be useful as a fertilizer. The natural silicites com- mercially available as sources of potash are chiefly orthoclase and leucite. Both of these minerals are potassium-aluminum silicates. The theoretical formula for orthoclase is written K 2 O.Al 2 O 3 .6SiO 2 , and for leucite K 2 O.Al 2 3 .4Si02. The principal sodium feldspar albite, has the theoretical formula: Na 2 O.Al 2 Oa.6Si0 2 . It is well known that these feldspars run into and substitute each other in various proportions so that the products from different quarries will vary widely in respect to their soda and potash contents. There is an enormous supply of feldspar in the United States, both east and west, which could be made economically possible as a source of potash supply, provided the cost of production can be made low enough to compete with the potash-holding manure salts which are at present so largely imported from Germany. Although it must be admitted that the imported potash salts are richer in potash than any product that can ever be made from American feldspars, it should also be remembered that the crude German manure salts contain large quantities of chloride and sulphates of elements which are not only undesirable in the fertilizer but which may do actual harm under certain conditions. It is this fact which gives encouragement to the attempt to pro- duce from American feldspars a straight potash fertilizer which 3 33 34 Original Communications: Eighth International [VOL. could be used in exactly the same way that hardwood ashes have been found useful. Six general methods have been proposed for decomposing the natural silicates, in the effort to obtain water-soluble potash salts. I. Adaptation of Natural Agencies. In the processes of Nature, the slow action of moisture and atmospheric agencies, including the action of carbonic acid gas, is known to have a decomposing or kaolinizing action upon the feldspars. Immense deposits of feldspar and granitic rocks have thus been decomposed, with the formation of large beds of kaolin and clays from which the potash has been leached into the surrounding valley. For this reason, the valleys between feldspathic and granitic hills are usually highly productive of the crops which require large amounts of potash, such as tobacco, potatoes, large fruits, berries, etc. There have been a few processes proposed, which depend principally upon the natural reactions hastened by pressure and other agencies. In 1904 Blackmore (U. S. Patent 772,206) proposed the action of carbon dioxide gas under five hundred pounds pres- sure upon a cream of the ground mineral, repeated intermittently for several hours, in the attempt to produce a yield of carbonate of potash. Ten years earlier the same experimenter (U. S. Patent 513,001) had proposed using lime, calcium chloride and steam pressure in an autoclave to produce chloride. In 1910 Coates (U. S. Patent 947,795) proposed the addition of bacteria for the decomposition of feldspar. In 1910 Carpenter (U. S. Patent 959,841) proposed to heat the mineral intensely and cool suddenly by plunging in water, in the effort to render the feldspar amor- phous, in the hope of making it more available for plant growth. None of the above processes have as yet been shown to possess industrial possibilities. II. Wet Processes of a Chemical Nature. Levi in 1904 (French patent 344,246 and English patent 13,875) and Piva in 1905 (French patent 351,338) proposed methods for treating leucite by means of solutions of alkali or alkali earth hydrates, generally under increased pressure. The same general method for treating feldspar was claimed by Swayze in 1907 (U. S. patent 862,676) and by Gibbs in 1909 (U. S. patent 910,662). v] Congress of Applied Chemistry 35 Also Gibbs in 1904 (U. S. patents 772,612 and 772,657) pro- posed a process of treatment with hydrofluosilicic acid, and sub- sequently with sulphuric acid, in order to produce potassium sulphate. In 1907 Cushman was granted U. S. patent 851,922, a public patent which proposed the sludging of finely ground feldspar with water, the addition of a small amount of hydro- fluoric acid and electrolyzing the mixture in wooden cells pro- vided with wooden diaphragms. Under this process both potas- sium and aluminum hydrate passed through the cell diaphragm into the cathode compartment. This process, although perfectly practical, has not yet been made commercially possible owing to the high cost of hydrofluoric acid and the large amount of by- products formed in the process. None of the above processes have as yet been made commercial possibilities. III. Dry Processes of a Chemical Nature, in which the Potash Salts are Volatilized. In processes of this nature, fluxes, and in some cases fuel, for reducing purposes is ground and mixed with the feldspar, the mixture being subsequently heated until the potash salts are volatilized and collected either in the stack dust or partially collected from the gases by passing them through or over water. Swayze in 1905 (U. S. patent 789,074) heated ground feldspar with gypsum and carbon, and proposed to collect the volatilized sulphate. Spencer and Eckel in 1909 (U. S. patent 912,266) made a cement mixed with calcareous and silicious fluxes and green sand, a natural potash-bearing iron silicate, clinkered the mixture in a rotary cement furnace, and obtained a Portland cement, at the same time collecting the potash in the stack dust and the flue gases. In 1911 Eckel (U. S. patent 1,011,172) proposed a somewhat similar method but heated only high enough to drive off the potash salts and not high enough to clinker the mixture. Again in 1911 Eckel (U. S. patent 1,011,- 173) melted a mixture of green sand, limestone and fuel, tapped off the melted iron and slag, and recovered the potash salts from the flue gases. Some of the processes under this heading have been tried on a large scale. No great difficulty is recorded in driving off the potash in the furnaces ; but obstacles were encountered in the attempt to collect the potash from the gases. As a by-product 36 Original Communications: Eighth International [VOL. operation in the manufacture of cements, these processes may yet come to be of some industrial importance. IV. Dry Processes which Propose to Separate Potash as Hydrox- ide or Carbonate. The old method of Bickell, proposed in 1856, (U. S. patent 16,111), which depended upon heating a mixture of felds- par, lime, and natural phosphate rock or guano to a bright red heat, has not as yet been proved practical or successful. The process of the Soc. Romana Solfati in 1905 (French patent 352,275), which proposes the roasting of leucite with carbonate, hydrate or nitrate of soda and lime and subsequently the passage of steam through the roasted product to produce sodium aluminate and potassium carbonate, is possible from a chemical standpoint but the high cost of operation has not permitted the process to come into commercial use. V. Dry Processes Producing the Chloride. These processes have been most experimented with upon the mill scale of opera- tion. In 1900 Rhodin (U. S. patent 641,406) and in 1901 (J. Soc. Chem. Indus. XX, 439) proposed fritting feldspar with lime and salt. According to the published results ; this experimenter ob- tained good yields although the process has not become a com- mercial success. In 1907 McKee (U. S. patent 869,011) sug- gested heating a potash-bearing material containing mica with lime, salt and carbon in order to obtain a yield of potassium chloride. Cushman in 1911 (U. S. patent 987,436) proposed mixing feldspar with lime and salts of a mineral acid capable of decomposing the silicate, giving the mixture special treatment previous to heating in a rotary furnace in order to produce the chloride. This method has been tried out on a large mill scale of operation, and the results obtained will be discussed later on in this paper. VI. Dry Processes Producing Sulphates. In 1911 Thompson (U. S. patent 995,105) proposed heating to a bright red heat a mixture of feldspar, sodium acid sulphate and sodium chloride, and subsequently leaching out the potassium sulphate produced. This experimenter claims that potassium chloride is first formed, which is subsequently changed to the sulphate by the action of the acid sulphate. It is stated that this process has recently v] Congress of Applied Chemistry 37 been tried on a commercial scale of operation. Sodium acid sul- phate is a by-product that is reasonably cheap although a large quantity is not available. The lack of an abundant supply of acid sulphite is perhaps the greatest drawback to the commer- cializing on a large scale of this process, although it is possible that it may still become of some industrial importance. Hart in 1911 (U. S. patent 997,671) proposed to fuse feldspar with some barium compound, such as the sulphate, together with carbon, to pulverize the cool melt and subsequently to digest the product with sulphuric acid and thus produce in solution potash alum and a residue of barium sulphate and silica which is claimed to be useful as a paint pigment. Hart claims that some of the potash is volatilized during fusion. Since the fusion temperature is 1500 C., it is probable that a considerable portion of the potash does volatilize, and it is possible that this difficulty may interfere with the commercial success of the process. Wadman in 1907 (U. S. patent 847,856) proposed heating lepidolite with potassium sulphate and leaching the product with sulphuric acid in order to obtain sulphates of lithium and potash. A chronological list of the patents which have been granted for the treatment of the silicates for the production of available potash is given in Table I. 38 Original Communications: Eighth International [VOL. TABLE I. Proposed Extraction Processes Chronologically Arranged Patentee Year Process Product IV Bickell 1856 Lime, Ca 3 (PO4)2 red heat Caustic I Blackmore 1894 Lime, powdered CaCl 2 , H 2 O, steam KC1 V Rhodin 1900 Lime, salt, heat under melt- ing KC1 II Levi (leucite) 1904 Ca(OH) 2 or NaOH, pres- sure 16 atmospheres K silicate II Gibbs 1904 H 2 SiF 6 and H 2 SO 4 K 2 S0 4 I Blackmore 1904 CO 2 500 Ib. pressure repeat- ing K 2 CO 3 II Piva 1905 (Leucite) KOH, NaOH, K silicate steam 25 atmospheres K aluminate IV Soc. Romana Sol- fati 1905 (Leucite) alkali, carbon, CaO red heat K 2 C0 3 III Swayze 1905 Gypsum and carbon, fuse, volatilize K 2 SO 4 VI Wadman 1907 Lepidolite, K 2 SO 4 H 2 SO 4 K 2 S0 4 II Cushman 1907 Water and HF1 electrolysis KOH II Swayze 1907 Heat alone then KOH solu- K silicate tion K aluminate V McKee 1907 " containing mica " with CaO, NaCl and C KC1 II Gibbs 1909 Ca(OH) 2 steam 150 Ibs. KOH III Spencer and Eckel 1909 Green sand cement mix vol- atilize K salts I Coates 1910 Bacterial action I Carpenter 1910 Intense heat, sudden cool- ing alone V Cushman 1911 CaO, CaCl 2 etc., clumps, red heat KC1 VI Thompson 1911 NaHSO 4 , NaCl bright red K 2 SO 4 VI Hart 1911 Ba compound as BaSO 4 and C, fuse, H 2 SO 4 Alum III Eckel 1911 Cement mix but not over 900 C. with green sand K 2 O volatilize K 2 S0 4 III Eckel 1911 Green sand, CaCO and C. melt iron volatilize K 2 S0 4 v] Congress of Applied Chemistry 39 It would appear that the most promising processes for making potash available from the natural silicates on a commercial scale of operation are those which are conducted in the dry way but without actual fusion of the reacting mixture. Potash salts volat- ilize readily at the high temperatures necessary for the fusion of the silicates, and the collection of the volatilized potash from the stack gases has not yet been carried out economically. A con- siderable portion of the potash does not settle in the dust chamber and if water sprays are used for washing the gases, the potash solutions are very dilute and the cost of evaporation becomes prohibitive. Furthermore, water sprays are found to interfere with the draft regulation, even when the use of fans is resorted to. The maintenance of artificial draft is an expensive and difficult matter, and is very likely to interfere with the proper control of the furnace temperatures. For work on the large scale of mill operation, a continuous process must be used, avoiding fusion and with the regulation of temperature to the exact point at which appreciable quantities of potash do not volatilize. The fluxes and reacting substances must be cheap, available in large quantity, and the yields of water soluble potash salts must be high. The process which has seemed to us to give the most prom- ise of successful adaptation to commercial ends is that of Gush- man (U. S. patent 987,436) coupled with the method of prepara- tion of the materials before furnacing, proposed and developed by Coggeshall (U. S. patent 987,554). This process has recently been given extensive trials on a large scale and interesting results have been obtained. The process consists essentially in powdering 100 parts of potash feldspar rock together with about 20 parts of lime and with or without 10 to 20 parts of rock salt. This powdered mixture is fed to the top of a moving drum about three feet in diameter, in a layer about half an inch deep. As soon as the layer is formed a strong solution of calcium chloride is applied from a series of small tubes. The CaCl 2 at once unites with the lime forming a so-called oxy- chloride cement and a large portion of the mixed powder is there- by at once formed into " clumps " or aggregates lying in a bed of surplus powder. As the drum revolves the bed is removed by a scraper to a belt which delivers the mixture to a screen which 40 Original Communications: Eighth International [VOL. separates the clumps from the residual powder. The powder is returned by a screw conveyor and elevator to the hopper above the drum again. The clumps are about the size of peas and pass from the screen directly to a rotary kiln similar to those used in burning Portland cement. The kiln is heated by a blast of air and powdered coal in the usual manner. The clumps pass regularly down through the increasingly heated portions of the rotating kiln and roll out at the end, prac- tically without alteration in size and shape. A large percentage of the total potash present in the feldspar is converted into potassium chloride during the heat treatment, and very little is volatilized. The dry clumps are of a pale yellow color outside, due to the iron in the ash of the bituminous coal used, but they are snow white inside. The clumps are finally ground, producing a pale yellow material containing as much water-soluble K 2 as hardwood ashes, although the potash is in the form of chloride, and the product also contains considerable free lime. Up to the present time no attempt has been made on a large scale to leach out the soluble potash. The ground material is being given field tests as a straight potash fertilizer containing lime. A Resume of the Large Scale Experiments. Potash Feldspars were obtained from five different localities. Eleven carloads were used in the trials, amounting to a total of 385 tons. Each carload was ground and analyzed separately. The lowest in potash ran 6 per cent. K 2 and 3 per cent. Na^O, the highest 11.3 per cent. K 2 and 3.1 per cent Na2O. The bulk of the spar ran 10 per cent, potash and 2 per cent, soda, and the results given in this paper were obtained on the 10 per cent. spar. The lime was a high calcium quick-lime, running about 90 per cent. CaO and 5.6 per cent. MgO. The salt was rock salt from New York State and ran about 98 per cent. NaCl. The calcium chloride was obtained from the Solvay Process Company. It was in the solid form and contained about 75 per cent. CaCl 2 and 25 per cent, water. All of the above materials are available in very large quantities and at low cost. The calcium chloride is a by-product in the form v] Congress of Applied Chemistry 41 of a moderately strong solution, and but a small proportion is concentrated at the present time, as the chief use is for refrigerat- ing purposes. Vast quantities are now run to waste. The solid form was used in these trials merely for convenience. Many heats were made with mixtures of varying proportions, but the two mixtures used in the work here described were : Feldspar 100 Feldspar 100 Lime 20 Lime 20 Salt 10 Salt 20 The feldspar, lime and salt were separately crushed in gyratory crushers and rolls, and dried in a rotary drier. In continuous work the proper mixture would be made at this point by con- tinuous weighing machines, but as a number of different mix- tures were to be tried, each of the three raw materials was ground separately in Huntington mills and put into bins. This pre- liminary grinding of the feldspar and salt was to about 65 per cent, through a 100-mesh sieve, of the lime about 83 per cent, through the 100-mesh. The weight per cubic foot of each powder of the above fineness was then ascertained and measur- ing boxes were built so that the materials could be separately measured out and run together into a large mixing machine. Almost a ton was thus mixed each time. The mixture was then conveyed to a tube-mill and further ground to a fineness of from 97 per cent, to 99.5 per cent, through a 100-mesh sieve, and then conveyed to the bin over the clumper and kiln. The calcium chloride masses were broken up and thrown on a perforated grid in a large tank holding about 48 tons. Water was run in and the chloride dissolved most readily. The solution was run out when about 42 degrees Beaume into two large sump tanks, and brought to a constant strength of about 42 per cent. CaCl2. This was then pumped up to an elevated tank and piped from there, through a constant-level tank, to the dropper tubes of the clumper placed in a row above the drum. This drum is 15.5 feet long and 3 feet in diameter, and is horizontal. There are 15 valved pipes, each one feeding an adjustable pipe holding 38 short dropping tubes of brass Y& i ncn internal diameter, and set -$ inch apart. 42 Original Communications: Eighth International [VOL. The finely ground mixed powder is taken from the bin by a chute, elevator and screw conveyor and distributed in a long hopper trough over the drum. It is taken from the trough by a roll device and spread evenly on the moving drum at its topmost point. The drum has a surface velocity of about 1.6 inches per second, the layer of powder advancing at this rate. It was found that by dropping the liquid very rapidly upon the powder, the clumps could be made rapidly enough to give a full feed to the short rotary kiln when only one-third of the trough and droppers and drum is used. A clumper drum 5 feet long produces every hour almost two tons of fresh clumps and con- siderably over a ton and a half of burned product with the kiln used in these trials. The excess of powder passes through a screen and goes to the same elevator which lifts the original material from the bin. The amount of actual CaCl2 in the fresh lime is regulated to about 20 parts to each 100 parts of feldspar in the mixture. The clumps leave the screen in rounded form and flow directly into the kiln. The reason for the above procedure will now be explained. In the first place, calcium chloride reacts very efficiently under these conditions with the feldspar by replacing the potassium with calcium, thus forming calcium silicate and potassium chloride. Anhydrous calcium chloride is expensive to produce and it is im- practicable to grind it into a mixture on a large scale on account of the rapid absorption of moisture. Even if such a dry mixture could easily be made, its use would present certain disadvantages. When a reaction between an ore and solid fluxes is produced by heating up to the fusing temperature, the reaction takes place on the surface of the particles alone and only at the points where the ore is in actual contact with the flux particles. Finer grinding will produce a larger surface area and thus a greater number of actual contact points, leading to a larger yield. There is, how- ever, a degree of fineness beyond which it is not wise to go, on account of the cost of extremely fine grinding. Another factor in the problem is brought out by the following experiments. A batch of ore and the theoretical amount of solid flux were ground together to just pass a 50-mesh sieve. This powder, when subjected to a certain heat treatment, gave a v] Congress of Applied Chemistry 43 reaction yield of about 35 per cent, of the theoretical. The mix- ture was then ground to just pass a 100-mesh sieve and given the same heat treatment. A reaction yield was obtained of about 65 per cent, of the theoretical. The mixture was then ground to pass a 200-mesh sieve and again reheated as before. A smaller yield was obtained than when the material just passed the 100- mesh, although the particles were undoubtedly only half the average diameter with about four times the surface area and should therefore have had far more points of contact. Upon weighing equal volumes of the 50-mesh, 100-mesh and 200-mesh powders, it was found that the latter contained far less material and it became apparent that the 200-mesh powder consisted for over 54 per cent, of its Volume simply of voids. Such finely ground powders are well known to "surge", that is, to show the tendency to flow like water through orifices in a manner resembl- ing fountains. Material ground as fine as this is the cause of much trouble at spout slides and conveyors. Each particle of a material of this extreme fineness is undoubtedly surrounded by a film of air, the actual contact with the surfaces is lessened and friction almost eliminated. When allowed to flow into a bin, such a powder assumes an almost horizontal surface, there is practically no angle of repose. Unquestionably the lessened contact caused the low yields in the finely ground mixtures. Some of the finer material was briquetted and the subsequent heat yield about 85 per cent, of the theoretical. Briquetting is, however, expensive and usually necessitates the addition of a binding agent foreign to the reaction. As a result of these investigations, the method was developed for aggregating fine powders by dropping a suitable liquid upon an excess of the powder in such a way as to cause a temporary bond to form, thus practically eliminating the air films or voids around the individual particles and permitting actual surface contact. Under these conditions, with the same ore and flux used in the experiments described above, the same heat treatment yielded within 3 per cent, of the theoretical quantity present. This method of aggregating finely powdered materials previous to furnacing has already been used in several different ways. For example, in an ore mixture in which the fluxing material is an 44 Original Communications: Eighth International [VOL. alkaline carbonate, such as sodium or potassium, which forms crystalline salts containing water of crystallization, if the car- bonate is used in the partially anhydrous condition and ground with the ore water alone dropped upon the mix in the manner described formed at once a crystalline carbonate which binds the particles of ore and flux into separate clumps, which are hard enough to withstand screening, while the air films are practically eliminated. Using such a mixture and process as this, a practi- cally theoretical yield was obtained, although the flux was used only in the exact molecular proportion called for by the reaction. By this clumping process a very intimate contact of reaction of surfaces is readily obtained at a low cost. The quantity of flux necessary to complete the reaction is greatly reduced, the duration and temperature of the heat treatment is lessened and working with rotary kilns dusting and stack losses are almost entirely eliminated. The clumps are beautifully adapted to the feed mechanism of rotary kilns, as they flow easily, do not dust and take the heat more evenly than fine powders. Now that the temperature conditions in rotary kilns can be accurately con- trolled, it would seem that many chemical and metallurgical reactions which are now performed by intermittent processes and with low yields could be much more economically carried out in continuous rotary kilns, taking advantage of this new method of forming aggregates previous to furnacing. In the application of this method to the treatment of felds- pathic rock, advantage was taken of the fact that a solution of calcium chloride acts upon dehydrated lime to form the oxy- chloride which is a strong cementing compound. It was found that the formation of calcium oxychloride gave a sufficiently strong bond to enable the aggregates to withstand the operation of screening and the burden in the kiln. The theoretical quantity of calcium chloride flux required depends upon the total quantity of K^O and Na20 present in the mix, as it is evident that the soda must also be liberated in pro- portion to its content. The feldspar ore used ran 10 per cent. K 2 and 2 per cent. Na2O which required theoretically 15.5 parts of calcic chloride. In all our trials some slight excess of calcium chloride has been used. The strength of the solution and the v] Congress of Applied Chemistry 45 method of treatment has been such that about 20 parts of actual calcium chloride are present in the fresh clumps to every 100 parts of feldspar. The 20 parts of lime used is for the purpose of form- ing the aggregates, and this lime remains practically unchanged in the finished product. The presence of lime in a potash fer- tilizer will be found advantageous to most soils, and it is generally admitted that lime increases the manurial value of a fertilizer. If the object was to leach out the soluble potash salts from the product, a much smaller amount of lime could be used without interfering with the formation of hard clumps. The salt is added because it has been found to aid the heat reaction, probably mechanically as will be explained later on. The fresh clumps contain from 16 to 20 per cent, of moisture, which is, of course, evaporated in the upper part of the kiln. The rotary kiln used in these trials was one of the old bottle shape cement kilns with a total length of slightly over fifty-five feet, the upper twenty feet having a diameter of 4 feet clear inside the firebrick lining, the lower portion widening out to nearly 6 feet inside diameter. The pitch was f inch per foot and the most suitable speed was found to be one revolution in about 2 J minutes. The conditions of the heat treatment are very important. The kiln used was too short to yield the best results, and after the preliminary experiments changes were made which caused the material to take about If hours to pass through the length of the kiln. The temperature of the gases issuing from the upper end of the kiln were read continually with a thermo-couple pyrometer fitted with a 15-foot fire end and temperatures were also taken from time to time at the firing platform. A furnace wall tem- perature of about 1370 C. is required for efficient burning of powdered bituminous coal. This is, however, much too high a temperature for potash work in a rotary kiln. This difficulty called for careful experimental investigations and adjustments of the heat treatment before the proper yields could be obtained. If a longer kiln had been available, there is every reason to be- lieve that a more efficient use of the heat could have been ob- tained. The coal used was a fairly high volatile bituminous coal. It was ground to about 94 per cent, through a 100-mesh sieve and blown into the furnace under an air pressure of about ten pounds per square inch. 46 Original Communications: Eighth International [VOL. During the progress of the clumps down the kiln the following reactions probably take place. At the entrance to the kiln the water begins to evaporate. As the hotter zone is approached, the temperature rises high enough to melt calcium chloride and salt. Whether the calcium chloride is free to melt is not known to us, as the exact composition of the oxychloride compound formed has not yet been determined. The results of our work seem to prove that the reacting chlorine is more readily evolved from the oxychloride compound than it is from calcium chloride alone. The melting of the salt, however, continues the bond of the re- acting particles, causing them to thoroughly " wet " each other, and from this point on the attact on the silicate proc.eeds rapidly. During the heating usually from 1 to 2 per cent, of Na2O is volat- ilized. When operating with no salt present, the yield of soluble potas- sium chloride was 47.5 per cent, of that originally present in the feldspar. On adding to the mixture 10 parts of salt to each 100 of spar, a test heat yielded 64 per cent, but of this 9 per cent, was lost by volatilization, giving a yield of 55 per cent, net in the final product. On adding 20 parts of salt to the mixture the yield grows to 69.2 per cent, with no volatilization and to 75 per cent, under heat conditions which caused a volatilization of 7 per cent., leaving a net yield of 68 per cent, of that originally present. In the case of clumps made from a mixture of 100 parts of feldspar containing 10 per cent. K 2 O and 2 per cent. Na 2 0, 20 parts of lime, 20 parts of salt and 20 parts of calcium chloride, the theo- retical composition if no volatilization loss takes place is shown compared with the actual results obtained in the following table. Theory Analysis Total K 2 O 6.25% 5.8% Water soluble K 2 .... 4.2% Equals 6.65% KC1 Loss of K 2 O 5% As KC1 already formed Total Na 2 O 7.62% 7.1% 52% made into NaCl Water soluble Na 2 6.37% 5.1% Showing 1.79% vapor- ized as NaCl or 26% of that present Congress of Applied Chemistry 47 This particular product contained 11.2 per cent, of free lime and total lime by analysis 15.5 per cent. There was also in this sample about 5 per cent, of free unchanged calcic chloride. The amount of calcic chloride in the various runs made up to the present time have been reduced gradually to about 1 per cent., and it is felt that in the future better conditions of heat treat- ment will make complete use of the calcic chloride and at the same time raise the yields of soluble potash. In later runs in which only 10 parts of salt were present in the mix, the theoretical and actual analysis of the product was as follows : Total K 2 O Water soluble K 2 Vaporization loss of soluble K 2 K 2 O insoluble in water Total Na 2 O Water soluble Na 2 O Theory Analysis 6.66% 4.15% 5.62% 4.5% 1.04% 1.12% 3.7% Equals 7.12% KC1 As KC1 already formed Showing 0.45% vapor- ized as NaCl or 11% of that present This product contained 12.25 per cent, of free lime, the total potash rendered soluble was 5.54 per cent, of the product or 83.2 per cent, of the total quantity present, but as 15.6 per cent, had been volatilized the net yield in the product amounted to 57.6 per cent. The material which was later made continuously according to the process described above carries 4.5 per cent, of water soluble K 2 O in the form of 7.12 per cent, potassium chloride and in addi- tion to this material carries only 1.12 per cent. K 2 O insoluble in water. It is well known that a 2 per cent, citric acid solution will extract, when used according to the Wagner method somewhat more K 2 than can be made directly water soluble. This fact is of considerable interest when the product is to be used directly as a potash fertilizer. Conclusion. It is believed that under better conditions of heat treatment which can be obtained with longer kilns and with a somewhat different arrangement of the combustion chamber, 48 Original Communications: Eighth International [VOL. that slightly better yields than those reported can be obtained. It should be remembered that the kiln used in these experimental trials was originally designed for burning cement, but this type of kiln has long been superseded by improved forms. In order to get the proper heat treatment in the middle of the kiln to com- plete the reaction, it was necessary to have the upper part too hot. This condition will not maintain in a properly designed kiln. It is also believed that the use of oil as fuel would have allowed an easier regulation of the heat treatment but the trials so far undertaken have been made under conditions which were found available at the time. The subject of the costs of this process and of the product cannot be gone into in detail at this time but a few general state- ments may be made. The production of water soluble potash in feldspathic rock is essentially a low grade proposition, and the commercial success of such a process depends upon the low cost of the various operations. The manufacture of a straight potash fertilizer containing as valuable ingredients only potash and lime must be carried out on a very large scale and by the most modern methods of continuous operation. With regard to the clumping process, the trials have shown that this operation can be practical- ly carried out as a continuous process and at an exceedingly low charge per ton of product. The process may be directly compared with that of the manu- facture of Portland cement. It is a little easier to grind feldspar and lime than the shales and limestones used in cement manu- facture. Drying will cost no more. Chemical control of the raw mixes will not be more expensive and perhaps much less. Clump- ing, as has been shown, adds a very small charge to the expense of treatment. The cost of furnacing the feldspar mix will be less than similar charges in the cement industry, as the temperatures required are much lower and less coal is consumed. The product from the potash kiln is comparatively soft and pulverizes easily in hammer mills, while the charges on the cement industry for grinding clinker is an important item. Again the softer potash product merely requires to be ground fine enough for use as a fertilizer, whereas cement clinker must be ground very fine and costs rise rapidly with increasing fineness. Repair bills in the* v] Congress of Applied Chemistry 49 case of feldspar treatment should be much smaller than in cement manufacture. The charge for raw materials is somewhat larger than in the case of cement but this is more than met by the smaller costs of operation. The potash fertilizer as now produced should be the equal in fertilizing value of the ordinary grades of hardwood ashes. The product carries practically the same content of water-soluble potash and somewhat more lime than wood ashes. There is every reason to believe that if the process becomes an industry that the yields of water soluble potash can be considerably im- proved. The material yielded is not a fused product, it is friable as an ash and it has the physical texture to make it a valuable aid to soil structure. The success of the product, must, of course, depend upon the results obtained under test conditions in its experimental use as a fertilizer. If results are obtained which are as good or better than those which usually attend the proper use of high grade wood ashes, it is believed that there should be no reason why this product cannot be successfully produced and introduced, especially in those parts of the country where potash feldspars, fuel and shipping facilities are available. Summary. In this paper a summary is given of the various processes which have been proposed for making the potash in the natural silicates available as a fertilizer. Experimental trials of a new rotary kiln process for treating feldspar are described, which depends upon a previous treatment before f urnacing, consisting of a method of aggregating or clump- ing the mix so that chemical contact of the reacting substances is brought about during the subsequent processing. The quali- tative and quantitative results obtained on a number of experi- mental trials on a mill scale of operation are presented and dis- cussed. It is shown that it is possible to economically manu- facture a potash fertilizer containing free lime from feldspar and for a sufficiently low cost to make an industry based upon the method, worthy of consideration. 4 NOTES ON A STUDY OF THE TEMPERATURE GRAD- IENTS OF SETTING PORTLAND CEMENT BY ALLEETON S. CUSHMAN The Institute of Industrial Research, Washington, D. C. The reactions that take place when hydraulic cements are tempered with water and while the mixture is hardening are as yet not understood. It is true that many theories have been advanced in regard to the hardening process or processes but more data is required before much that now seems inexplicable can be understood. Since all chemical reactions are accompanied by definite and measurable thermal changes, complete temperature records of hardening cements should yield interesting and valuable data. It is well known that if a portland cement clinker is ground without the addition of from 2 to 3 per cent, of calcic sulphate or gypsum to act as a restrainer it will be " flashy." By " flashy " is meant the tendency to harden very quickly, so quickly in fact that in many cases it is impossible to mold the wetted cement into a plastic mass. While this sudden hardening is going on, a considerable amount of heat is generated so that the mass feels hot to the hand. The temperature rises about 10 to 15 C. but the heat reaction lasts only a short time and after cooling no further heat reaction takes place. When however, a portland cement has been properly restrained by grinding with it 2 or 3 per cent, of gypsum (plaster) the conditions of thermal activity are changed in a quite extraordinary manner. On mixing a nor- mal Portland Cement with sufficient water to form a normally plastic mass, a certain amount of heat is immediately disengaged although not so much as in the case of an unplastered cement. The plastic mass soon cools down to the air temperature and generally falls somewhat below the surrounding temperature showing that a decided cooling effect is taking place. If now the plastic mass is allowed to stand quiescent in a constant tempera- ture chamber, nothing of moment happens if the cement be a 51 52 Original Communications: Eighth International VOL. normal standard brand, for a period of from four to eight hours. At a given time however, for every mixture a secondary heat rise begins, and increases more or less rapidly to a definite maximum. After this rise is completed the cement has attained its final set and a gradual cooling takes place to the temperature of the surrounding air and nothing further happens. If an imperfect, damaged or lumpy cement is under observation the temperature gradient for the rise may show aberrations. That is to say, a sudden rise may be followed by a temporary cooling only to be followed by another rise. The wonderful effect of a small percentage of gypsum plaster in thus controlling and regulating the temperature gradients or reaction of setting cement is little understood and indeed presents certain anomalous occurrences for our consideration as will be shown later on. The first successful attempt to record the temperature gradient of setting cement as far as the writer has been able to ascertain was made by Gary who used a photographic recording device which has been fully described by Burcartz. 1 The method consisted of placing the bulb of an ordinary glass thermometer in the cement paste. The whole arrangement was enclosed in a box through which a beam of light was made to impinge through a slot, upon the graduated stem of the thermometer and then upon a travelling photographic film. As the mercury rose or fell the beam of light was cut by the shadow of the mercury column and thus a con- tinuous temperature gradient record was obtained. The only criticism of this method that can be made is that it calls for an expensive and delicately adjusted piece of apparatus which few laboratories would care to install, and in which the temperature changes cannot be watched while they are taking place. The apparatus used by the writer is simple, comparatively inexpensive and can be installed and used in any laboratory for making daily records. The apparatus is shown in Fig. 1. A double walled wooden box as shown in Fig. 7, used simply to avoid any sudden changes which may take place in the laboratory temperature during a test run. An ordinary so-called " fireless . Record Dec. llth, 1909. Congress of Applied Chemistry 53 cooker " such as can be bought at any kitchen supply store answers very well for this purpose. The recording thermometer is of the Tycos type and consists of a copper plated steel tapered mercury filled bulb 9 c. m. long by about 2 c. m. in its maximum diameter. The bulb is connected to the recording dial by a Fiq.1- flexible steel capillary tube. The recording dial has a range from 10% to 120 F. The recorder is fairly accurate for the middle range and is easily calibrated and adjustable. ' In ordinary tests as carried out in the writer's laboratory 1 kilogram of the neat cement is tempered with 250 c. c. of water to make a homogeneous plastic paste which is packed into a No. 2 open top tin can. The thermometer bulb is not inserted until the primary heat effect which always develops when cement is kneaded with water, is over, and the paste reaches approximately the same temperature as the calorimeter box. In the meantime the copper plated thermometer bulb is smeared with vaseline and wrapped with several folds of fairly heavy tin foil. The object of these precautions is two fold; in part to guard against the " freezing " in of the bulb when the cement hardens and in part to overcome any possible pressure on the walls of the bulb if the cement shrinks or expands during the hardening process. With 54 Original Communications: Eighth International [VOL. 25 per cent, of water the consistency is somewhat softer than the normal, but experience has shown that the wetter mixture gives better results under the conditions of the test. When all is ready the covered bulb is pushed into the cement paste, care being taken that it is not pushed below the surface so that the cement can close over the shoulder of the bulb and so imprison it when hardening takes place. When these precautions are taken the apparatus gives no trouble and the bulb is easily withdrawn from the hard cement at the end of the test. The temperature gradients are usually taken for a twenty-four hour period, al- though this is not necessary unless the full cooling curve is desired. In presenting these notes on the temperature gradients of set- ting cements, it is not the intention of the writer to draw any conclusions at this time in regard to the mechanism of the harden- ing reactions. The curves obtained on the revolving scale are transferred to centigrade degrees and plotted in rectangular coordinates as is shown in the illustrations, plates I and II, curves 1 to 32. An inspection of the curves will show that in some cases the tempera- ture gradients are much steeper and more sudden than in others. Curves 6, 7, 8 and 21 represent cases in which the water was simply poured on to the dry cement without previously kneading the mass to a paste. In no case of this kind is a rise in tem- perature noted following the final set or hardening, at about seven to eight hours. In the cases of some brands of cement as is shown in curves 9, 10, 11 and 12 the rise in temperature is constant and gradual to a maximum which usually occurs at about ten to eleven hours. In other cases notably in curves 17 and 29, the rise is sharp and sudden. As both types of cement pass muster in the standard tests, it is not possible at the present time to state what the ideal temperature gradient curve for a cement should be. That the maxima and shape of the curves is modified by the addition of various salts to the tempering water is shown in curves 13, 14, 15, 16, 17, 18, 19, 20 and 22. Perhaps the most extraordinary curve is number 19 which shows the heating effect produced by saturating the water with . H * h ? 3 j*~-*t "" am '** '. **" "*!*"*>'- j-^j i*^v 1 4 i-T^ J ~ Plntel. 5*6 v] Congress of Applied Chemistry 55 calcium sulphate. In this case the temperature rose above the scale of the recording device and the test piece became uncomfort- ably hot to the hand. Since calcic sulphate is used as a re- strainer when ground with a cement, this extraordinary effect of calcic sulphate solution is difficult to explain. Curves 25, 26, 27 and 28 were from cements which did not stand test and had been rejected. The abnormality of these curves is at once apparent to the eye and furnishes the best argument as to the value of a study of the temperature gradient as an additional method of control in cement testing. In conclusion the author wishes to point out that these notes on the study of temperature gradients are offered not as data on which to establish theories but to stimulate other workers to include similar investigations in their studies of the hardening of hydraulic cements. CAUSES OF BREAKAGE IN GLASS MANUFACTURE AND METHOD OF DIFFERENTIATING CHEMICO- HETEROGENEIC STRAINS FROM COOLING STRAINS BY ROBERT L. FRINK Columbus, Ohio The manufacture of Glass, as carried on in the majority of factories wherein pressed ware, table ware, plate and window glass is made, is conducted in substantially the same manner, with perhaps a few minor improvements, as has been the case for ages, and up until a few years ago there had been no inventions presented or improvements made in the art for hundreds of years, and to-day the actual making of the glass, consisting of the pro- portioning of the raw materials, charging them into and firing the furnace for melting them, shows very little, if any, improve- ment, and substantially no acquisition of knowledge as to the real properties of the glasses, and the cause and effect of modifica- tions, or variable conditions, surrounding and entering into the production. I have many times been asked by progressive Glassmakers, where they could obtain literature on the subject of control of melting furnaces and compositions, to which I have been com- pelled to reply that I knew of none. Hovestadt, Winkelman, Tillotson, Kennedy, Schott, Lorentz and Lorenz, and in an indirect manner, Wedgewood, Pirsson, Day, Wright, Rosenbusch and Winchell, have each furnished us with more or less data on which might be based certain deductions, and which might be assumed or be attributed as the effect of certain causes, but these, with the exception of perhaps a few experiments of Day, Lorenz and Hovestadt, are based on laboratory operations and condi- tions, and do not represent and, as a consequence, do not reveal the problems entering into the manufacture of Glass from a commercial standpoint, and which confront the glass maker each hour and day. I have many times discussed the subject of Glass- 57 58 Original Communications: Eighth International [VOL. making with our various College Professors and Deans of depart- ments, and to my question, " Why do you not make more of a study of glass?" the invariable reply is made: " Because of the lack of problems in it " a conclusion which they have arrived at, no doubt, by reason of the non-progressiveness of the majority of our glassmakers and manufacturers. However, to-day the Glass industry is assuming a far different attitude and position from that of a few years ago, brought about by the advent of machines and mechanical manipulation of glass, and the release of the industry from the bondage of Unionism, which so com- pletely held it in its grip and throttled its progressiveness for centuries. Competition has become so keen that at the present time the manufacturer must look to economies in order that he may be able to " make ends meet," to say nothing of declaring dividends, and whilst some are still content to conduct their business accord- ing to the old rules and custom, nevertheless there are a few who are beginning to realize that the chemist and industrial engineer have got a place in their business, and that the successful filling of that place means the success or failure of their enterprise. I regret very much that I am unable to present to you in this paper that which I had intended when I assigned its title, but, owing to illness, I have been unable to complete many experi- ments and investigations anticipated and started at the time I undertook the preparation of this article, and, therefore, I shall have to be content in here presenting such data as I have collected as bearing upon several of the very vital problems, the solution of which would place in the Glass manufacturer's hands means and methods, the application of which would make for him a handsome dividend each year. Much time, money and energy have been expended in designing, experimenting with and perfecting various mechanical devices for the cheapening of production, by eliminating labor, or reduc- ing the degree of skill necessary to produce the finished product. However, still there followed, and there was always present, the uncertain factor of quantity of production controlled by the amount of good ware produced and that quantity of glass which it was necessary to melt to obtain such quantity of saleable prod- v] Congress of Applied Chemistry 59 uct, the difference usually being accounted for, in part, by breakage and imperfections, the one, in many instances, being identical with the other, that is to say, the cause of certain imper- fections is likewise the cause of breakage. The manufacture of Window Glass by the various mechanical processes, suffers particularly from breakage losses, and it was in the investigation of these losses, and their causes, that revealed some quite remarkable conditions, or at least they appeared quite remarkable at the time, being directly contrary to the accepted theories and beliefs held among tradesmen and operators. The experiments and results hereafter cited, were carried on, with the exception of a few instances, with glass that was made by the mechanical process of drawing glass cylinders from a bath of molten glass, the glass being a soda-lime-silicate, the batch form- ula of which consisted of: Sand 1,000 Ibs. Crushed Lime Stone 330 Ibs. Salt Cake 380 Ibs. Ground Coal 18 Ibs. and produced what was known to the trade as a salt-cake " nor- mal " glass. The glass was melted in a tank furnace, the dimen- sions of which were approximately 18 ft. inside x 80 ft. long, and a depth of glass of 5 ft. The fuel used was natural gas, which was applied in a method similar to the Siemens' regenerative proc- ess. On each side of the furnace there were five ports arising from the regenerative chamber through which the air passed, being heated by the retained heat of the checker work in this chamber, and as the air entered the furnace the gas was applied and mixed just inside the furnace wall, passing across the furnace and out of the ports of the same construction on the other side, the direction of flow of gas and air being reversed every 20 minutes, in the usual well known manner. The furnace temperature at the melting end was maintained, as uniformly as possible, at about 1500 deg. C. while at the ladling, or working end, a temperature was main- tained at from 1100 to 1150 deg. C. At the inauguration of the mechanical process, the glass was melted, and the furnace operation was conducted in the same 60 Original Communications: Eighth International [VOL. manner as when cylinders were made by the hand method. It was found, however, that considerable breakage occurred, which was characterized as " breaking off the pipe " " breaking on the horse " and " breaking in the flattening oven " meaning in the first instance that the glass would break where it was attached to the blow-pipe, or bait, to which the glass adhered, and by means of which it was withdrawn in cylinder form, from the bath of molten glass. In the second instance there were two forms of breakage, one being apparently from no cause, and which occurred shortly after the cylinder was taken down and laid upon horizont- al supports called the " horse," these supports consisting of blocks of wood held in metal pieces and supported by springs. As stated, shortly after the roller was placed upon these sup- ports, without any warning, and apparently from no cause, it would suddenly seem to explode and would break in thousands of pieces from one end to the other of the 20-ft. length. Another form of " breakage on the horse " was that which occurred when the operator had separated the roller into its respective flattening lengths, it being necessary to make from 6 to 8 " cuts " this being done by passing a wire around the cylinder, electrically heating the wire, and heating the cylinder along a line of narrow dimensions, applying to this line a cold or moistened iron, which caused a sudden contraction and fracture of the cylinder, usually along the line. However, many times this would not be the case, and instead of the fracture following the narrow heated area, it would depart from this path, and perhaps run in " cork-screw " fashion throughout the whole 15 or 20-ft. length. Again at times the fracture would perhaps branch off at a tangent, and produce a rupture of all manner of fantastic designs, at times going forward for a distance of a few feet or more, perhaps returning in a circle of 6 to 12 inch radius, reverse its direction and, in a smaller circle, start out and encircle the roller, or progress along its length. At other times the frac- ture, under these conditions, would assume a straight line, or perhaps a spiral line along the length of the cylinder. The breakage during the first six or eight months of operation by mechanical process would average 70%, and at short intervals go as high as 95%. After a time of experimenting it was decided v] Congress of Applied Chemistry 61 to change the batch mixture, supplanting the major portion of the salt cake with soda ash, and finally it was found that with a batch consisting of: Sand 1,000 Ibs. Crushed Lime Stone 328 Ibs. Soda Ash (58%) 240 Ibs. Salt Cake 80 Ibs. the breakage was very materially reduced, also that many of the imperfections, more particularly those known as " cords " and " strings/' were reduced, these being narrow areas where the glass was somewhat thicker in section, usually being caused by there being a local area, in the mass of glass from which the cylinder was drawn, which had suffered a reduction in temperature, or being due to a lack of chemical homogenity in the metal. How- ever, a great deal of breakage occurred, and it was found that it was very ununiform, being perhaps 40 per cent, one day and the next day jumping to 50 or 60 per cent. It was also found that the breakage in the flattening oven would increase at times when the breakage on the machine was at its lowest. These conditions were all very puzzling, and up to that time there seemed to be no literature on the subject of " Glass " which would give any definite accounting, even in a general way, or any specific cause for the breakage occurring. A careful study had been made of all the furnace conditions, taking into account the amount and composition of material charged into the furnace, the gas and air consumed, draft conditions, waste gas analyses, temperature, etc. To a degree, variations in these conditions coincided with the breakage, yet no one variable condition did so with sufficient definiteness whereby it could be accredited with having been the cause of any particular form of breakage. There would also occur at times, a breakage of from 15 to 20 per cent, in the flattening oven, the primary cause being the checking or shrending of the surface of the cylinder by the elec- trically heated wire causing a too high local temperature, which would shrend or check the outside surface to a depth of perhaps .001 inch, or in some instances .013 or .014 of an inch. However, this same checking might occur at intervals and produce no 62 Original Communications: Eighth International [VOL. breakage in the oven, and again the same temperature of wire, under the same manipulation, would not produce this checking, all of which appeared to be problems governed by the properties of the glass. The explosive nature noted in " breaking on the horse " would also be noted in the flattening oven, where, in per- haps 40 to 60 seconds after the split cylinder was " shoved in," and was resting under the shade stone, it would suddenly become disrupted and break in thousands of pieces. These same conditions are not only noted in the production of window glass, but may also be seen, in modified forms, in the manufacture of plate glass, tumblers, lamp chimneys, pressed ware of all description, and even in cut glass. A particular instance which has come to my notice, and which has revealed some quite unusual and aggravating conditions, is that of the manufacture of sidewalk glass. I have more specific data on window glass, and will confine myself to presenting this data, which may be used and applied where conditions are analogous in other branches of the art. Examinations of specimens of glass cut from cylinders which exploded on the horse, or from sample pieces taken from such cylinders, immediately after having come from the machine which formed same, or from the workman's hands who made the article, revealed excessive strains existing on the exterior, as also consid- erable strains on the interior surfaces of the cylinders. It was determined that they were compression strains which penetrated the glass to about 12 per cent, of its thickness on the outside, and 8 per cent, on the inside, but which in some instances would be as great as 25 per cent, on the outside and not over 3 or 4 per cent, on the inside. This determination was made by ascertain- ing the double refraction. This consisted of cutting from a cylin- der (by means of a diamond or wheel) specimens ^ inch wide, placing them on edge in a holder, or substage of a Petrographic Microscope, at 45 deg. to the plane of the Nicols. By introducing a quartz wedge to give a phase difference as to produce red of the first order, or a selenite plate, so that their c direction was at right angle to the plane of the specimen, and on then viewing the specimen with the 1-inch ocular and 1 J-inch objective, and using crossed nicols, it was found that the orientation of the strained v] Congress of Applied Chemistry 63 portion of the specimen was distinct and well denned, and pro- duced a bright yellow color, denoting a crushing strain. This was contrary to the belief of practical men, who held that the outside and inside surfaces were under tension, the fact being, however, that when the cylinder walls are formed and withdrawn from the surface of the bath of semi-molten metal, its dimensions are formed and fixed by the solidifying of the exterior and interior surfaces, while the interior section of the sheet, or cylinder, is yet in a mobile, or plastic state, which later cools and contracts upon the already solid and fixed exterior and interior surfaces, thereby producing a compression upon these surfaces, and a tension upon the interior of the section, whose orientation produces wave lengths the path difference of which ranges from .001 to .004, and in extreme cases .OOGft/A, this being more clearly differentiated by the introduction of the selenite plate, or a quartz wedge, in order to get a phase difference, or retardation of the light rays of about 525 to 550 /x/x, and which, as stated above, determines the orientation, that is, determines whether the specimen under examination be under compression or ten- sion, and to some extent the degree of such strain if the thickness of the specimen be known. This has been positively proven, and in fact by the use of a special Fedorow wedge, designed and com- puted by Fred Eugene Wright of the Carnegie Geophysical Laboratory, it is possible to calibrate the exact amount of strain introduced in pounds per square inch, or grams per square centimeter. By use of the Wright wedge, or compensator, it has been very easy to demonstrate the enormous strains which have been intro- duced by reason of the chemico-heterogeneity, as also those strains caused by cooling, or unequal solidification and contrac- tion during the formation of the article. It is found that the com- pression strains on the exterior of the cylinder, may, at times, reach the enormous pressure of 90,000 Ibs. per square inch. However, it seems that these pressures are not so directly re- sponsible for breakage as are the strains introduced by, what I have termed, a physical heterogeneity, meaning by the term an introduction, into a sheet or article of glass, of a local or narrow area of metal which has been cooled and solidified previous to 64 Original Communications: Eighth International [VOL. the solidification of the balance, or major portions of the article. Further, when the compression strains occur by reason of the formation of the article, as above explained, and the mass is not homogeneous in its composition, small portions or areas being of a greater or less density, this immediately creates a path or line of least resistance, through which fracture will occur should there be any occasion whatsoever for these conflicting strains to exert themselves, as in separating the cylinder with the hot wire, or sudden cooling of the surface of the glass, introducing additional crushing strain at or near the junction of this strain, which then sets up either chemical or physical heterogeneity, in which event a complete disruption of the article will usually occur, or should the chemical differences in composition lie in definite lines of cork- screw fashion throughout the length of the cylinder, this would be the path of the fracture, as cited. Should it be so disseminated as to form a considerable portion of the mass comprising the cylinder or article, it will then have the explosive effect; or should the exterior have suffered sufficiently rapid reduction in temperature over that of the interior, there is a tendency to produce this explosive effect, although in the latter conditions the edges of the fractures will have a shivered appearance, denoting a tendency to separate the article through its longitudinal section, producing some sharp edges. It has been stated that this breakage has been produced pri- marily by conflicting strains set up by tension and compression due to physical and chemical heterogeneity. This was determined by two methods, one being by etching the glass with HF, and making an analysis of that portion dissolved, in the following manner: A specimen of the glass was cut to a size sufficient to cover a platinum dish about ij inches in diameter, the specimen being about 2 inches in diameter, if possible. They were washed and made perfectly clean, rinsed with alcohol and ether, dried and weighed. There was placed in the dish 10 cc. of HF, which was then covered with the specimen, subjected to a temperature of water bath (about 90 deg. C.) until all of the acid had been volat- ilized and had acted upon this specimen, after which it was boiled in HC1 and washed perfectly free from any adhering resi- Fig. I. Fig. II. Fig. III. Fig. IV. Fig. V. Fig.VI. Fig. VII. v] Congress of Applied Chemistry 65 due. H 2 S0 4 was then added, evaporated to dryness, and sub- jected to low red heat to expell all the silicon fluoride, analysis then being made of that portion dissolved from the glass. The specimen was then again washed with alcohol and ether, and weighed so as to determine the amount dissolved by the HF. It was found that the portions affected by HF, when examined by the microscope and polarized light, showed striae, and which, as usually was the case, showed that the heavy cord, or ream, would suffer more or less from the effect of the HF, depending upon whether same contained a greater or less amount of lime silicates than was present in the surrounding clear glass. Fur- ther, if this etched portion be closely examined by higher magnifi- cation, there would be distinguished areas of ununiform solubility, and many times well defined crystals, or the outlines of such crystals in the matrix of higher or lower silicates in which they had resided. There also appeared to be a considerable difference in the appearance of the etched portion of various samples, some having a very close granular structure, as is shown in the accom- panying Fig. 1, and others having a nodular appearance, of 1 or 2 mm in diameter, as is shown in Figs. 2 and 3, although such structures were not perfectly round. Fig. 4 illustrates what seems as an impossible narrow area of chemical heterogeneity. As is seen in the center of the photo- graph, the acid etched the glass to a considerable depth, and a similar area near to the margin. In this specimen it was found that the Soda ran abnormally high in the analysis of the etching residue. Fig. 5 is an illustration of an etching wherein the acid has dis- solved the crystals from the matrix in which they lay, and in this analysis the Silica was high. Under conditions as shown in Fig. 5, where there appeared to be crystalline compounds, which, although perfectly transparent to the eye, with the aid of the microscope showed an unusual thickness or layer of compression strain on the exterior of the cylinder, and the breakage in this instance was very great, while under conditions as shown in Fig. 3, the breakage would amount to from 7 to 12 per cent, it being exceptionally low. 5 66 Original Communications: Eighth International [VOL. In viewing a specimen cut in transverse section, or in looking at the edge of a cylinder lengthwise, in cases of poor melting, or chemical heterogeneity, a striated condition would be very pro- * Fi. 6 showinc- Fig. 7 is a good example of a well made glass, and one in which the breakage was very low. As will be seen, the outside shows a strain somewhat heavier than the inside, but the specimen is perfectly free from striae. The photo-micrographs, Figs. 6 and 7, were taken with crossed nicols, but with no selenite plate intervening. Fig. 7 is under lower magnification than are the other illustrations, it being about 20 diameters while the others are magnified to 80 diameters. While the method above referred to gives us specific and defi- nite information as to the lack of chemical homogeneity of the glass, nevertheless same is subject to more or less inaccuracies, and there arise many conditions in composition which are more or less confusing, particularly as to just what element the cord or disrupting cause is due to. Therefore I have sought for other methods and means, which do not require skill and time to deter- mine these chemical changes and variables, that may be applied with definiteness and at the same time enable one to apply the results in operating the furnace and in controlling the batch mix- ture, and I believe that the determination of the refractive indices of the glasses, offers a solution of the problem. As has recently been stated by Tillotson, it is quite possible to, within certain limits, affix a specific refractive index for each of the ingredients used in the making of commercial glass, and such being the case it seems that it would be quite possible, by a sys- tem or a series of analyses and refractive indices determination of the various combinations of silicate of lime and soda, to acquire information of sufficient accuracy as to denote the exact cause of this heterogeneity, and while it is of course true that fire condi- tions may bring about sufficient disturbing influence upon the melting glass as to be the cause of breakage, yet I believe that this reveals itself in disturbing the refractive index to a much less degree, if at all, and that one will be enabled to differentiate this condition from improper chemical combinations. v] Congress of Applied Chemistry 67 It was this part of my investigations that I was unable to com- plete. However, I have made a number of determinations of the refractive indices of glasses as they came from the furnace, and have compared such determinations with the analyses, etchings and microscopic examinations, as above referred to, and find that they are much more accurate and definite in their definition as to the cause of breakage, although perhaps not quite so specific as to the elementary cause, that is to say, one can very easily deter- mine whether the breakage is due to chemical heterogeneity, physical heterogeneity, or to excessive strain produced by too rapid cooling, as is shown by the following instances: The method used was as follows samples were taken from the cylinders, or articles, as they came from the machines or work- men, crushed by tapping in an agate mortar so that the major portion would pass a 60 sieve. The refractive index of the speci- mens was determined by the Becke line method, using a Bausch & Lomb Petrographic Microscope, applying the 1 inch ocular and 8 mm and 4 mm objectives. A series of oils of different composi- tions were compounded, the refractive index of same being varied by 2 points in the third decimal place, I using for the major por- tion Oil of Juniper (n = 1.4761), Oil of Sassafras (n = 1.5233) and Oil of Cinnamon (n = 1.5773), and making mixtures of these compound oils as requirements demanded to get the proper index to the fourth or fifth decimal place. The instrument used was a Fuess, Model II, Refractometer. It was found that one could easily determine the indices to the fourth or fifth decimal, although no attempt was made, with the exception of peculiar cases, to reach this degree of accuracy. To illustrate the possi- bilities of this method, one or two instances will suffice. A sample of glass which was taken at the time when breakage was very heavy, and in which several heavy cords were seen in each cylinder, showed a refractive index of the clear glass (Speci- men C-251) of 1.5253, while the cord (Specimen C-252) indicated 1.5088. In another instance, where the tank had been running very badly, the gas being low and the fire being very poor, much diffi- culty had been experienced in melting sufficient batch to keep the glass at the proper level, and much trouble experienced from 68 Original Communications: Eighth International [VOL. cords being produced by reason of the cold air blowing into the gathering holes, and which produced what is known as the " cold cord." In this sample it was found that the thickness of the clear glass was .145 inch, while the chemical cord was .157 inch thick, and the cold cord about .15 inch thick. The natural glass (Specimen C-253) had a refractive index of 1.522. The chemical cord (Specimen C-254), in which there were seen small nodes, or thickened portions of about .1 inch or 2.5 mm in diameter, had an index of 1.519. Specimens of the physical cord (C-255), which were trimmed as free from the surrounding glass as possible, gave an index of 1.521, and there also appeared several crystals, more or less anisotropic, having an index of 1.517. Many times in the crushing of glass it is noted that there appear to be portions that are more friable, or brittle, than are others, and in one sample where this appeared to be quite marked, it was found that the main portion of the part investigated, same being a chemical cord (Specimen C-257) gave an index of 1.518, while that for the hard, or unfriable portion, was 1.522, and which corresponds with the index of the clear glass surround- ing the cord. In making determinations of several other speci- mens of Chemical Cord C-257, concordant results were obtained with those denoted, the indices remaining at 1.518 and 1.519. In specimen C-258, being glass from the same tank, the batch composition being identical, and the conditions throughout being as near the same as could be expected, in commercial opera- tion, as when C-251 was taken, but being from a several months' later production, I obtained an index for the clear glass of 1.52, while the cord in this case (Specimen C-259) gave an index of 1.5237, this, as will be noted, being directly opposite to the con- ditions as found in C-253, C-254 and C-255. Several months after the above mentioned specimens were taken, the glass from this furnace showed an index of 1.524, and which was no doubt due to a reduction in Soda, the Salt Cake in the batch analysis showing that the Soda content was nearly 1 per cent, lower, and the Silica content nearly .6 per cent, higher than heretofore. In the making of lenses there is often considerable difficulty experienced in producing the blanks free from reams or striae, v] Congress of Applied Chemistry 69 and it has been quite a problem, and a matter of much conjecture among the manufacturers, as to what is the due cause whether it be from erosion of the pots, or due to the chilling of the glass during the process of melting and it would seem that the results found in my experiments would answer this question quite deci- sively, inasmuch as that portion of the lens which was clear gave an index of 1.512, while the portion which showed striae gave 1.5136. In the manufacture of bottles, lamp chimneys, globes, table ware, and particularly lantern lenses, automobile reflectors, semaphores, tumblers, or other articles which are required to stand variables temperatures, the question of breakage is a very vital one, and should it be possible to work out a complete and definite method as simple as is the determination of the refractive indices, as above outlined, it would be something which would be accepted and hailed with delight by every progressive manufac- turer in the country. He knows, of course, that his ware is break- ing either because of injudicious handling or because of improper making, but rarely, if ever, does he know, and in fact it is quite difficult to convince him that this breakage is due to the mix- ing, melting, or control of the conditions surrounding the han- dling of his glass prior to the time that it reaches the solid state, and it would be of inestimable value and benefit if our industrial scientists could place in his hands some simple, non-scientific but practical method and means whereby he could with certainty and definiteness control his operations and obtain uniform results. The Glass industry is undergoing radical and revolutionary improvements, much money and time being spent in experi- menting with and developing these improvements, relying upon protection of Patents so as to give sufficient monopoly to warrant the vast expenditures necessary to demonstrate the practicability of such inventions and improvements. However, a little investi- gation of the Patents issued in the United States, and the basis for the claims in such Patents, will soon reveal the fact that there is substantially no understanding of the laws and principles gov- erning the properties of glass, nor of the chemical actions and reactions controlling its composition, and among some of the more educated capitalists controlling certain branches of the industry, 70 Original Communications: Eighth International [VOL. this has come to be a serious consideration in the point of further investment towards getting improvements, for, unfortunately, we Americans are prone to shun investment unless large returns are in sight, and, unlike our foreign relations and associates, we do not depend upon our research laboratories and competent and loyal help, to carry out the more salient and less obvious factors in our manufacturing processes rather than relying upon published facts and theories, protected by Patents, law and juris- diction. I have been connected with the Glass industry in its various branches, from the point of employee, operator and inventor, and in the research laboratory, and have also been more or less in intimate touch with other industries, and in my opinion, without doubt, the Glass industry is the most neglected, but has more intricate problems than any other that has come to my attention, and its problems are difficult ones by reason of the fact that but few of them manifest themselves by the production of a direct result, but in the majority of cases are hidden, or conceal them- selves, by reason of the very transparency of the material. I earnestly hope that our scientists, industrial engineers, pro- fessors, and students, will take a greater and a deeper interest in the problems confronting the manufacturer, and I shall cer- tainly be very pleased to devote my time and laboratory, and give personal escort to those sufficiently interested, to the plants with which I am connected, for I feel that it is only in the prac- tical operating plant that these problems manifest themselves, and where real and beneficial results can be accomplished which can be made to produce profits and dividends, and without such a final conclusion to the work, very little, if any, progress will be made in getting at the more concrete factors of this industry, for it is doubtful if there is any University, educational institu- tion, or private individual who would feel like spending a sum sufficient to erect and operate furnaces for experimental research purposes, of dimensions that would produce the conditions found in practical operating plants. With the assistance of the manufacturers and capitalists who are operating the plants for a profit, and with the aid of judicious scientific investigation, which they can be shown will produce v] Congress of Applied Chemistry 71 added profit, it seems that it is possible, by proper investigation and research, to produce glasses, and the apparatus for manip- ulating the same, as to greatly broaden the field of its use, either by improving its properties in the way of resistance to fracture and heat, or reducing the losses, for it is true that it is the cheapest commercial commodity in the crude state of any of our building materials. Is it not possible that we may step from fancy and fiction into fact and reality, and build our houses of glass, conduct our water supplies through glass, and sanitate and insulate in a manner that would be of great universal benefit? MAGNESIA IN PORTLAND CEMENT BY A. A. KLEIN AND A. J. PHILLIPS Bureau of Standards, Pittsburg, Pa. In the testing of cements by this laboratory a number were found which were rejected on account of their high MgO content, the limit for acceptance having been placed at 3%. While in some cases the MgO content exceeded 7%, a study of the physi- cal tests revealed little cause for their rejection, a few having failed in the boiling test but on ageing and retesting the pats were satisfactory. On this account it was decided to study more completely a number of picked samples with increasing MgO content in order to determine how the MgO was combined, and if any reason could be found for making a sharp rejection at 3%. The analysis of these samples are given in Table I. A petro- TABLE I. 1 2 3 4 5 6 7 8 9 10 11 SiO 2 , 21.38 21.57 21.06 21.63 22.58 20.65 20.94 21.87 21.92 21.68 21.47 A1 2 3 , 8.22 9.07 8.10 7.55 8.30 6.52 6.36 7.09 7.07 6.97 7.19 Fe 2 3 , 2.22 2.36 2.51 2.65 2.09 2.36 2.43 2.34 2.31 2.29 2.34 CaO, 62.29 61.14 62.44 62.12 60.92 61.49 61.92 57.47 57.46 58.32 57.94 MgO, 2.87 3.09 3.32 3.53 3.70 4.55 4.79 6.02 6.28 6.55 7.07 S0 3 , 1.94 1.38 1.70 1.65 1.58 1.58 1.58 2.05 2.12 1.78 1.70 Ig, .53 .68 .56 .45 .40 1.85 1.19 2.24 2.05 1.68 1.55 graphic examination, using the methods employed by the Geo- physical Laboratory in their investigation of the ternary system CaO-SiO^A^Os, 1 showed that the most abundant constituents present were the B 2CaO.SiO 2 and 3CaO.Al 2 O 3 while 3CaO.Si0 2 and 5Ca0.3Al 2 O 3 were present only in small quantities. Con- trary to the theories of some former investigators, neither MgO 'J. Ind. Eng. Chem. 3, 221-27. 73 74 Original Communications: Eighth International [VOL. nor MgO.Al 2 3 were found present. Monochramatic Na light was used in their investigation, since 3CaO.Al 2 O3, MgO and MgO. A1 2 O 3 are all isotropic with indices of refraction very close to one another. There was observed condiserable CaC0 3 which origin- ally must have been present as free CaO, and this would seem to account for the failure of a few of the cements to pass the acceler- ated tests. That this compound was CaC0 3 and not MgC0 3 was definitely ascertained by the maximum index which was found to be about 1.66, that of the MgC0 3 being 1.717. The cements were exceedingly fine grained, the largest crystal- lites being those of the /32CaO.Si0 2 which never exceeded 0.03 mm. in length and few attained that. This coupled with the fact that the cements when received for examination were finely ground, rendered the determination of the optical constants and the identification of the constituents a very difficult and pains- taking matter. The orthosilicate and tri-aluminate agreed very closely with those described by Shepherd, Rankin and Wright in all determinable properties except the index of refraction which was lower in both instances and furthermore was not constant. This lowering of the refractive index and the absence of MgO and MgO.Al 2 O 3 led^to the investigation of the possibility of the re- placement of CaO by MgO in these compounds. The program outlined consisted of the burning and microscop- ical examination of 2CaO.SiO 2 and 3CaO.Al 2 3 in which the CaO was partly replaced by MgO according to molecular pro- portion. Also in view of the fact that in the cement industry CaO and MgO are calculated together, a small series of burns were made in which definite percentages of CaO were replaced by MgO. The materials used were a flint containing but 0.2%Pe 2 3 , a light calcined MgO containing but 0.5% Si0 2 , C. P. CaCO 3 and Al oxide containing no Si0 2 or CaO. The various mixes were moulded into small hollow cylinders and burned in a pot furnace, using natural gas and compressed air, the samples resting on a magnesite base which offered practically no contamination at the temperatures employed. The effort was made to produce homogeneous samples with satisfactory crystal growth for optical examination at temperatures as close as possible to rotary kiln v] Congress of Applied Chemistry 75 practice. In cases where homogeneity was not satisfactory, the samples were reground and reburned several times until satis- factory. The burning temperatures in no case exceeded 1500 and the time varied. The /32CaO.Si02 were first made according to the formula 2 (mCaOnMgO) Si02 in which m and n were varied to give com- pounds ranging from 2CaO.SiO 2 to CaO.MgO.SiO 2 . The first effect noted was that burning 2CaO.SiO2 at tempera- ture not exceeding 1500 was followed by dusting; remoulding and reburning the sample several times had no preventive effect on the dusting. The additions of MgO in increasing amounts diminished the dusting. Up to 4%, however, it was necessary to remould and reburn the specimens to produce non-dusting samples. With increasing MgO the clinker became harder and more vitreous, the softening point falling so that a sample con- taining 15% MgO could not be burned with a 5% sample for the same length of time on account of the melting of the former. The microscopic examination of CaO.MgO.SiO2 showed it to be homogeneous and composed of crystallites showing a prismatic development. The indices of refraction were determined as f olio ws:L = 1.649+ .003, = 1.60+.003 and y = 1.664+. 003, X-y being 0.016 approximately. Occasional cleavage was observed and parallel extinction noted. The optical character was nega- tive. These properties agree closely with those of the mineral Monticellite (CaO.MgO.Si0 2 ) of the Olivine series. The examination of the dusted specimens revealed y orthosili- cate with occasionally a small amount of ft. It was determined that MgO does not form a homogeneous product with CaO to any appreciable extent in the y orthosilicate because of the pres- ence of the MgO in the dust. Reburning of the dusted specimens eliminated the dusting. The specimens now showed an absence of y orthosilicate and MgO and the presence of ft orthosilicate with a minimum mean refractive index of about 1.70, that of the pure ft 2CaO.Si0 2 being about 1.725. This homogeneity was observed in specimens containing up to 6% MgO. At this point traces of unhomogeneity were noted. The new compound con- sisted of fine irregular bands surrounding some of the orthosilicate grains and of a lower index and double refraction. This was 76 Original Communications: Eighth International [VOL. probably the compound CaO.MgO.Si0 2 although accurate obser- vations on the bands could not be made owing to their minuteness. It is not surprising that homogeneous compounds of /3 orthosilicate containing such large amounts of MgO should exist since the optical properties of CaO.MgO.Si02 and ft 2CaO.SiO2 are fairly analogous and both crystallize in the orthorhombic system. (See Table II- A). TABLE II-A. Orthosilicates with Molecular Replacement Formula %MgO Temp. Time Characteristics 2(.98 CaO, .02 MgO).S:O 2 .93 1500 12hrs. Dusting Homogeneous; elimin- consists of B or- ated thosilicate with with mean refractive reburn- index lower than ing 1.725 that of the B orthosilicate. 2(.95 CaO, .05 MgO).SiO 2 2.65 1500 12 hrs. 2(.934CaO, .066MgO). SiO 2 3.10 1500 12 hrs. 2(.91 CaO, .09 MgO).SiO 2 4.82 1500 12 hrs. The minimum re- fractive index was approxi- mately 1.70. 2(.894CaO, .106 MgO).SiO 2 5.01 1450 6 hrs. Non- Dust- ing. 2(.882 CaO, .118 MgO).SiO 2 5.59 1450 6 hrs. 2(.87 CaO, .13 MgO).Si0 2 6.18 1450 6 hrs. Unhomogeneous consists of B or- thosilicate and CaO. MgO. SiO 2 2(.86 CaO, .14 MgO).SiO 2 6.66 1450 5 hrs. 2(.78 CaO, .27 MgO).SiO 2 10.64 1400 3 hrs. 2(.50 CaO, .50 MgO).SiO 2 25.57 1400 2 hrs. The tri-silicates were not extensively studied since in these cements under examination it was present in very small quanti- ties. However, substitution of MgO for CaO in the tri-silicate formula gave non-dusting products. The clinkers in every case Congress of Applied Chemistry 77 were soft and porous and lacked the hard vitrification of the orthosilicate containing MgO. Jesser 1 also states that MgO re- places CaO in the silicates and observed the formation of Monti- cellite. (See Table II-B). TABLE II-B. Tricalcium Silicates with Molecular Replacement Formula %MgO Temp. Time Characteristics 3(.962 CaO, .038 MgO).SiO 2 2.01 1500 5hrs. No Unhomogeneous ; dust- consisted mainly ing. of B orthosili- 3(.924 CaO, .076 MgO).SiO 2 4.05 1500 5hrs. Clinker cate with low- soft and dere index and porous. considerable a- 3(.814 CaO, .186 MgO).SiO 2 10.17 1500 5hrs. mount of 3 CaO. 2 CaO. MgO. SiO 2 18.8 1500 5 hrs. SiO 2 . Free CaO was noted in all specimens and free MgO in only slight amounts, except in the specimens con- taining 10.17 and 18.8% MgO. In those mixes in which percentage replacement was practiced, dusting was reduced, vitrification increased considerable ft 2CaO.- SiOa and a little 3CaO.Si02 formed. In every case both free CaO and free MgO were present in the dust. (This is to be expected since CaO and MgO do not combine with equal amounts of SiO 2 and the replacement of any per cent, of CaO in a silicate by the same per cent. MgO would give an excess of bases over acids.) Campbell and White 3 state that the more reactive CaO enters into combination before the inert MgO leaving to the latter the task of displacing CaO till equilibrium is reached. Equilibrium is not reached at 1500 but at higher temperatures more tri-sili- 'Zent. Hydraul. Zement. 1, 41. 2 J. Am. Chem. Soc. 28, 1273. 78 Original Communications: Eighth International [VOL. cate would be formed and no free CaO or MgO remain. More- over, in a cement raw mix formula in which the MgO and CaO are calculated together, free CaO and MgO would be present in the clinkers were it not for the fact that additional acids in the shape of A^Os and Fe20s are present to combine with them. (See Table H-C). TABLE II-C. Orthosilicates with Percentage Replacement %MgO Temp. Time Characteristics 2 CaO. SiO 2 , 1500 20 hrs. Dusting y orthosilicate. persist- ed 2(CaO 99% MgO 1%) SiO 2 1 1500 8 hrs. Dust- Unhomogeneous; 2(CaO 97% MgO 3%) SiO 2 3 1500 8 hrs. ing re- consisted of B 2(CaO 94% MgO 6%) Si0 2 6 1450 4 hrs. duced orthosilicate with with lowered in- reburn- dex and small ing. amount of 3 CaO . SiO 2 . 2(CaO 90% MgO 10%) SiO 2 10 1450 4 hrs. (Both free CaO and MgO in dust.) Continuing with the aluminates, mixes were made corres- ponding to 3(mCaO. nMgO.) A1 2 O3 in which m and n were varied to give MgO in % varying from 1-15. None of the 5-3 alumi- nates were made since this compound was present in small amounts in the cement, and its formation seems to depend partly on the temperature of burning. For instance, an underburned sample of 3-1 aluminate would contain 5-3 aluminate and free CaO due to the latter not entering into combination, or the two compounds might be present in a high burned sample due to the dissociation of the 3-1 aluminate, it being unstable at its melting point. Shepherd and Rankin. 1 No dusting phenomena were noted, simply increased vitrifi- cation and lowering of the fusing point. . Ind. Eng. Chem. 3, 212. v] Congress of Applied Chemistry 79 The microscopic examination revealed homogeneity up to 10% MgO. The crystals were more or less cubic in development, isotropic and the index of refraction varied from 1.710 that of C3aO.Al 2 O 3 to 1.67, that of the specimen containing 10% MgO. Beyond this % spinel, MgO.Al 2 O 3 , and free CaO were noted to be present. Our efforts to produce homogeneous products of the formula 2MgO.Al 2 3 or 3MgO.Al 2 O 3 were unsuccessful which is in agreement with Shepherd and RankinV statement that only one compound MgOAl. 2 O 3 is formed in the binary mixtures of MgO and A1 2 O 3 . (See Table III). No ternary systems con- taining MgO were investigated since the work on the cements examined seemed sufficient to predict its actual state in cements of usual MgO content. TABLE III. Tricalcium Aluminates with Molecular Replacement %MgO Burn. Temp. Time Burn. 3(.95 CaO, .05 MgO).Al 2 O 3 2.24 1420 2hrs. 3(.91 CaO, .09 MgO).Al 2 O 3 3.84 1420 2hrs. 3 (.89 CaO, .11 MgO).Al 2 O 3 4.98 1420 2hrs. 3(.87 CaO, .13 MgO).Al 2 O 3 5.91 1420 2hrs. Homogeneous crystals of 3(.845 CaO, .155 MgO).Al 2 O 3 7.09 1400 2hrs. cubical development with 3(.82CaO, .18MgO).Al 2 O 3 8.26 1400 2hrs. refractive index o *o a . li a -^ "1 *o o o. I "M 1 !? 51 o o 8l 8 Q o 28 Cement A 3 .00 1 21 .00 .1 30 .0436 29 3 .00 1 21 .00 .5 10 .0397 30 3 .00 1 21 .00 .1 30 .0419 31 3 .06 1 21 .00 .5 30 .1186 32 3 .09 1 21 .00 .5 30 .1448 33 3 .15 1 21 .00 .5 30 .1909 34 3 .00 1 21 .00 .5 30 .0796 * 35 Cement B 3 .00 1 21 .00 .5 30 .0332 36 3 .03 1 21 .00 .5 30 .0558 37 3 .06 1 21 .00 .5 30 .0797 38 3 .15 1 21 .00 .5 30 .1272 *The cement and lime were finely ground together before being used in the experiment. v] Congress of Applied Chemistry 87 w e i Gr. of cement and lime used a o -j d 11 CC. of absolute alcohol fi sj g" F Grams of CaO extracted -2 2 1 39 Cement B 3 .15 1 21 .00 .1 30 .1110 40 Cement C 3 .00 1 21 .00 .5 30 .1055 41 3 .03 1 21 .00 .5 30 .1318 42 3 .06 1 21 .00 .5 30 .1390 43 3 .09 1 21 .00 .5 30 .1767 44 3 .15 1 21 .00 .5 30 .1880 45 3 .00 1 21 .00 .5 30 .1070 * 46 Cement D 3 .00 1 21 .00 .5 30 .0893 47 3 .03 1 21 .00 .5 30 .1173 48 3 .06 1 21 .00 .5 30 .1089 49 3 .09 1 21 .00 .5 30 .1631 50 3 .15 1 21 .00 .5 30 .1586 51 3 .00 1 21 .00 .5 30 .0789 * 57 3 .03 6 21 .00 .5 30 .1106 58 3 .06 6 21 .00 .5 30 .1292 59 3 .09 6 21 .00 .5 30 .1632 60 3 .15 6 21 .00 .5 30 .1663 61 3 .15 1 21 .00 .5 60 .1850 64 3 .15 1 21 .00 .5 30 .1500 * 65 3 .15 1 21 .00 .5 60 .1336 * 66 3 .15 1 21 .00 .5 60 .2025 1* 67 3 .15 1 42 .00 .1 60 .2078 1* 70 3 .15 1 00 .21 30 .1941 1* 71 3 .15 1 00 .21 60 .2049 1* 72 3 .15 1 00 .50 60 .2132 1* 73 Cement B 3 .00 1 00 .21 60 .0557 1** 74 3 .15 1 00 .21 60 .1722 1** ir The residue on the filter was washed with 75 cc. of water. *The cement and lime were finely ground together before being used in the experiment. **The cement and lime were ground together only enough to break up the small lumps of lime. 88 Original Communications: Eighth International [VOL. The best extractions were obtained by adding 21 c.c. of ab- solute alcohol, 30 drops of para cresol and 0.5 c.c. of water to 3 grams of cement to be tested and proceeding as above outlined. As may be seen from the table the results were hardly satisfac- tory from a quantitative standpoint. Where only small addi- tions of lime had been made, the increased amount of lime ex- tracted . corresponded fairly well with the amount of lime added. But with increasing additions of lime the percentage of extrac- tion fell off so that it did not give the total amount present. It was also found difficult to get duplicate determinations to check very closely. The best that can be said for the method as used is that under uniform conditions the extraction will be approximately quanti- tative when small amounts of lime are involved. MODIFICATION OF WHITE'S PHENOL TEST FOR FREE LIME Because of the rapidity with which B-naphthol and p-cresol reacted with free lime, it was decided to try a modification of White's test using these substances instead of phenol. Equal weights of b-naphthol were added to nitro benzene, xylene, a- bromnaphthalene and machine oil. A drop of water was added to each of these mixtures. The mixtures were then tried upon both pure calcium oxide and a cement to which 5% calcium oxide had been added. In no case did crystals form, which resembled calcium phenolate crystals; however, when the solvent evapo- rated crystals of b-naphthol were formed. The test was next tried with p-cresol. It is to be understood that any of the following reagents, which are named, contain equal parts of p-cresol and the solvent together with a trace of water, p-cresol with nitrobenzene did not give crystals. With xylene, crystals resembling those of calcium phenolate were formed when the reagent was added to a cement, which contained 5% free lime. These crystals resembled the crystals of calcium phenolate excepting that they were a pale red color. On account of this latter characteristic they could be distinguished very easily with the ordinary microscope, thus making their identifi- cation possible without the use of the polarizing microscope. A v] Congress of Applied Chemistry 89 reaction with pure calcium oxide took place immediately as was shown by the formation of the red salt, but no crystals were formed. When kerosene was used as a solvent, crystals were produced both in pure lime and in the cement, which contained 5% lime; crystals were also formed when machine oil was used. By far the best solvent used was found to be absolute alcohol. One c.c. of p-cresol, 3 cc. of absolute alcohol and 4 drops of water were used to make this reagent. The following tests were made : The cement containing 5% lime gave crystals within 2 minutes. Cements B, D, and C, and cement E ; samples No. 1, 3, 4, all gave crystals within 2-3 minutes. A sample of cement A, which had been exposed to the air for about three months was tested. This gave a few crystals at the end of 15 minutes, showing that free lime was very scarce. Cement E, sample No. 6, gave a very few crystals within 10 minutes, while sample No. 5 gave no crys- tals within 20 minutes. A cement, which had once contained free lime, but which had been exposed to carbon dixoide for two months was tested. It gave a very few crystals within 15 minutes. As stated above, cement E, sample No. 5, and the exposed cement A did not give crystals for a long time. These two cements were ground in a mortar and again tested for free lime. Cement A gave crystals within 1.6 minutes and cement E, No. 5, gave an abundance of crystals within 4 minutes. Free lime must have been liberated by the grinding, otherwise the time it took the crystals to form would have been the same in both cases. The experiment therefore shows the effect of aging on free lime, i. e., free lime on the outside of the particles of cement being attacked, while the free lime on the inside was not affected. Aging of cement does not necessarily do away with free lime. Obviously a coating was formed on the outside of the particle, which was not penetrated by carbon dioxide and moisture. This accords with the results obtained by Reibling and Reyes. 1 ibling and Reyes, Loc. cit page 398. 90 Original Communications: Eighth International [VOL. CONCLUSIONS From the above experimental facts the following conclusions may be drawn: First. That calcium phenolate and the calcium salts of homo- logues of phenol are very difficultly soluble in any of the common solvents. Second. That phenol and its homologues do not at all act with the same rapidity, or to the same degree of completeness with free lime. P-cresol apparently reacts immediately; b-naphthol arid phenol do not react so rapidly and the remaining homo- logues, which were tried, react only slightly. Third. That a roughly quantitative extraction of free lime is possible by means of a properly proportioned mixture of para cresol, with absolute alcohol, and a trace of water, when the total quantity of lime is small. The method does not give good results with large quantities of free lime. Fourth. That p-cresol with alcohol and a trace of water gives a more rapid test for free lime than White's test. Also the test is more delicate for it will give an abundance of crystals where White's test gave only a few. Apparently the test is too delicate for practical use. Fifth. That the aging of cement for a period of time does not necessarily destroy all the free lime, which it contains. THE PHYSICAL AND CHEMICAL PROPERTIES OF PORTLAND CEMENT BY W. C. REIBLING AND F. D. REYES Bureau of Science, Manila, P. I. ' INTRODUCTION The main effort in this work was directed toward a study of those characteristics of Portland cement regarding which there exists the greatest amount of misconception and diversity of opinion, our object being to assist in the universal effort to formu- late cement specifications so drawn as to guarantee the manufac- ture and use of Portland cement of the quality sought for. Our investigations were conducted on many grades and brands of material, careful consideration being given to the conditions of burning, grinding and seasoning to which the various products had been subjected. After working on commercial products, the investigations were continued on non-aerated clinker received from manufacturers in Europe and China. Finally, one of the authors visited a few large cement factories where every courtesy was shown him and where he was enabled to secure valuable information and to collect special material for this work. The endeavor at first was to ascertain the significance of the ultimate chemical composition. Within wide limits no apparent relationship between the ultimate chemical composition and the physical properties of the various cements could be discovered. This made necessary a more comprehensive investigation which for the sake of brevity and clearness was presented, as far as possible, as an abstract of the results, a general summary of which follows. PART I Free Lime in Portland Cement (1) As our investigations progressed it soon became evident that the Portland cements examined all contained free lime and 91 92 Original Communications: Eighth International [VOL. that their physical properties were influenced to a marked degree not only by the amount of free lime, but also by its condition; that is, whether this calcium was present as the hydroxide, oxide, or as the latter heated to a degree of incipient fusion. (2) In order to obtain reliable information about free lime the method for its detection first discovered by A. H. White 1 was employed. However, in order to make this test efficient and accurate, a chemical and microscopic investigation of the forma- tion of calcium hydroxide-phenol crystals was necessary. The result of this investigation and the manner in which it is possible to determine relative amounts of free lime and to distinguish between that which is sintered, non-sintered, or slaked, are fully described. (3) The application of this test gave conclusive proofs of the presence and the effects of free lime in commercial Portland cements, and a study of this free lime under different condi- tions of burning, grinding and seasoning showed the following results. A. Concerning the degree of burning: (a) That as the kiln temperature increased, non-sin- tered calcium oxide gradually became converted into a sintered state having different physical properties. (b) That this conversion may occur at temperatures far below those necessary for the proper burning of the cement. (c) That underburned cement may contain both sin- tered and non-sintered lime. (d) That all of the free lime in hard-burned cement is sintered. (e) That Portland cement clinker can be burned per- fectly so as neither to contain free lime nor have lime liberated in the ordinary process of cooling and grinding. 'Journ. Ind. & Eng. Chem. (1909), 1, 5. v] Congress of Applied Chemistry 93 B. Concerning the effects of seasoning: (a) That non-sintered lime hydrates more readily and quickly than sintered lime. (b) That sintered lime may hydrate so slowly by mere exposure to the atmosphere that the action takes place essentially on the outer exposed parts of the particles and only gradually penetrates to the interior. (c) That aeration tends completely to convert at least the surfaces of the hydrated lime into carbonate. (d) That the penetration of air into a mass of ground cement is limited approximately to a thin outer layer. (e) That when cement is aerated in thin layers the conversion into carbonate goes on practically as fast as hydration. (f ) That aeration tends to coat the particles of sintered lime with an impermeable film of calcium car- bonate so that even thoroughly aerated, finely ground cement may contain unslaked free lime. C. Concerning the effect of free lime upon the soundness: (a) That the usual cause of unsoundness is unslaked free lime. (b) That the effect of free lime upon the soundness is influenced by the cohesive properties of the ce- ment, the "speed of slaking," fineness, the tem- perature and amount of water used in gauging, and the effect of impurities and retarders. (c) That the test for soundness as indicating the pres- ence of free lime are relatively crude as compared with the microscopic study of calcium hydrox- ide-phenol crystals. (d) That slaked lime does not cause failure in sound- ness tests. (e) That non-sintered lime must be present in quan- tity to cause unsoundness, and if it is so present, the disruption is likely to occur in water and air, as well as in steamed pats. 94 Original Communications: Eighth International [VOL. (f) That a fair amount of fused sintered lime is liable to cause unsoundness in the accelerated tests. (g) That fineness assists soundness. (h) That the agreement between the microscopic evi- dence and the result of tests for soundness was very close. PART II The Seasoning of Portland Cement (4) The preliminary work had shown that although seasoning improves the soundness of unsound cements, it often works seri- ous injury to the strength and setting properties of sound prod- ucts. Also, that different methods of seasoning and storing the same material usually produced different effects upon the physical and chemical properties, and that the different cements were influenced in unlike manner by the same conditions of sea- soning. Cements stored in air-tight receptacles underwent little or no change. Therefore, a thorough study was made of the absorption of volatile constituents by both clinkers and pulver- ized cement. (5) Working with the finished product the following became evident : (a) That air penetrates very little (less than 13 millimeters) into a mass of undisturbed cement; however, water absorbed from the atmosphere may slowly penetrate farther. (b) That cements exposed to the air in small, repeatedly re- mixed quantities show the greatest increase in the per- centage of carbon dioxide. This confirmed the micro- scopic evidence that the free lime in Portland cement exposed to the atmosphere changes to carbonate soon after hydrating, the tendency being to coat the individ- ual particles with calcium carbonate. (c) That cements stored in air-tight receptacles show a slow decrease in the percentage of moisture, due as evinced by the microscopic evidence, to the slaking of free lime. v] Congress of Applied Chemistry 95 (d) That the rate of absorption decreases very rapidly as the reaction proceeds, fine particles absorbing proportion- ally more than coarser ones. (e) That, atmospheric conditions being the same, the quan- tity and rapidity of absorption depends largely upon the quality and quantity of free lime. (6) Contrary to the results obtained with ground material, it often proves more difficult to remove the cause of unsoundness from hard-burned than from underburned clinkers. The reason for this is that perfectly sintered clinker is practically inert to water and atmospheric influence and consequently the free lime imbedded in this hard, dense, inert magma is more thoroughly protected than the free lime in underburned clinker. (7) This work together with a study of manufacturing prac- tices and the strength developed by hard-burned and underburned cements lead to the following conclusions: (a) That aeration is the least efficient practical method of seasoning Portland cement. (b) That a high loss by ignition and a corresponding low spe- cific gravity are not characteristic of commercial cements made from well-burned clinker. (c) Cement made entirely from underburned clinker seldom appears on the market except as hydraulic limes. It would fail to pass the test of strength. Underburned cement usually comes to the consumer mixed with the harder burned material from the same mill and our present specifications are such that a mixture of 45 per cent, of disintegrated clinker and 55 per cent, of sound clinker passed all requirements except the percentage loss by ignition. This emphasizes the real importance and great value of the test for volatile constituents. (d) The best set kiln process yields a considerable amount of underburned clinker. Some manufacturers sort this out very carefully. Others do very little or no sorting and their finished product is not true Portland cement but a mixture of seasoned underburned and well-burned cement containing sintered, nonsintered and hydrated 96 Original Communications: Eighth International [VOL. free lime, calcium carbonate, and fused and sintered compounds of many kinds. (e) The rotary kiln is capable of producing a more uniformly burned clinker than the set kiln. Extreme fineness in the grinding of the raw material is necessary to produce a perfectly sintered product. Few manufacturers grind fine enough, the majority producing a hardburned clinker but one which still contains a considerable per- centage of free lime some of which fails to slake before induration and causes the much-discussed character- istic drop in the strength of rotary cements. PART III The Setting Properties of Portland Cement (8) For determining the time of initial and final set the method employing the Vicat needle was found to be reliable, impartial and accurate, abnormal results being obtained only with cements of very poor quality. (9) Preliminary experiments on the setting properties of commercial products demonstrated: (a) That although cements of different chemical composition have different natural setting tendencies, changes in the setting properties were due primarily to the absorp- tion of water or water and carbon dioxide. (b) That changes in the setting properties were independent of the ultimate chemical composition, the fineness, the amount of retarder, or the quantity of water and carbon dioxide absorbed. (c) That further consideration of the nature of the raw ma- terials used or the burning process employed could give no more definite information. (10) However, recourses to the microscopic test for lime showed that changes in the rate of set were brought about largely by altenations in the condition of the free lime and that the work on the ordinary commercial cements would have to be supple- mented with a study of known and selected material ground, v] Congress of Applied Chemistry 97 plastered and seasoned under known conditions. For this pur- pose large quantities of fresh, nonseasoned clinkers were obtained from three manufacturers who were selected in order to obtain standard cement with characteristically different chemical and physical properties. (11) Manipulated in the ordinary way, the setting properties of these non-seasoned, non-plastered cements were essentially the same, the plasticity being poor and the set abnormal and ap- parently slow and erratic. Further study showed that in reality these products set so quickly that they became regauged by any ordinary process of manipulation and that there was an imme- diate generation of much heat as soon as the water was added. Part of the heat generated was due to the hydration of free lime and part to the setting of the cement. (12) A study of the effects of plaster on the nonseasoned material showed the following: (a) As the amount of plaster used was increased step by step in each cement, the initial set took place earlier, then later, and finally again earlier. Also the amount of water required for normal consistency at first decreased decidedly, and then later slightly increased. The same quantity of plaster did not effect the set and plasticity of the different cements to the same extent. (b) The small amount of plaster used had no appreciable effect upon the slaking of ignited lime, and therefore, did not prevent the generation of heat due to the slaking of the free lime in this material. (c) The effect of minute quantities of plaster was to prevent regauging and to produce in reality a slower setting cement but one which after ordinary manipulation set more and more quickly until all regauging was entirely eliminated. (d) Further additions of plaster retarded the set, from 1.5 to 3 per cent, producing a maximum effect. Beyond this amount the natural tendency of plaster of Paris to set quickly manifested itself in the combined results ob- tained and the time of initial set again approached a maximum. 98 Original Communications: Eighth International [VOL. (13) A study of the effects of various methods of seasoning either the non-plastered ground cement or clinker showed that as soon as the free lime had become thoroughly hydrated less water and less plaster were required to produce a normal paste of standard consistency and set. The efficiency of the different methods of seasoning depended entirely upon the relative amounts of calcium hydrate produced and maintained, the con- version into carbonate acting so as to decrease the plasticity and the retarding influence of the sulphate. (14) A study of the effects produced by seasoning plastered cements showed that no radical difference was manifested if plaster was added before the cement had seasoned. (15) Additional consideration and experiments proved: (a) That the compounds which react with water and cause cements to set had not been appreciably affected by the atmospheric influences which altered the condition of the free lime. (b) That all hydraulic compounds which cause cement to set are not affected to the same extent by heat or retarders. (c) That the normal rate of the reaction of the compounds which cause cements to set varies considerably. (d) That when free calcium oxide is present and water is added the heat of hydration tends to increase the natural rate of set of the hydraulic compounds. (e) That a preliminary hydration of the free lime adds to the efficiency of the retarder and increases the plasticity of the cement. (f) That the substitution of calcium carbonate for slaked lime tends to reduce the plasticity and decrease the efficiency of plaster of gypsum as a retarder. (g) That ultimately, after prolonged exposure, the cement itself may finally become practically inert, (h) That four changes may occur in the rate of set and two in the plasticity, namely: 1. An acceleration of the set and an increase in plasticity as regauging is eliminated as free lime slakes. 2. Then a retardation of the set and a further in- v] Congress of Applied Chemistry 99 crease in plasticity as more free lime be- comes slaked. 3. An acceleration of the set and a decrease in plasticity as the quantity of slaked lime becomes reduced by conversion into calcium carbonate. 4. A retardation of the set and a further decrease in plasticity as the cement after prolonged exposure tends to become inert, (i) That these changes account for all of the variations met with in the action of commercial Portland cements. (j) That the results and conclusions derived from the selected material by the method of investigation resorted to, proved truly characteristic of the general nature of the commercial products represented. (16) This knowledge led to very definite results and conclu- sions concerning the practical control of the setting properties. It became evident: (a) That, whatever the influence of chemical composition, the real phenomena entering into the reactions, the natural activity of the setting compounds and the quan- tity and condition of the free lime, and analysis, such as is outlined, of the cement before it is packed is an accurate means of ascertaining the possible effects of storage on its setting properties. It will not only in- form the manufacturer if the set of the product of his kiln is capable of being kept within normal or desirable limits during the process of ordinary storage, but also, will fix the minimum amount of retarder required to do so. (b) That sometimes a cement must be seasoned before its set can be controlled, but in most instances this is not necessary, and in all instances the necessity for season- ing can be avoided by proper burning. (c) That it is the effects of hydration of the free lime and not its conversions into carbonate to which the manufacturer must give special consideration. (d) That proper packing is necessary for best results. 100 Original Communications: Eighth International [VOL. (e) That almost all commercial cements which failed to pass standard specifications only because of their rate of set would have proven satisfactory in all respects had they been seasoned or plastered properly. (17) A study of the influence of fineness upon the rate of set introduced no new or unsurmountable factors into the problem of the control of the set. Obviously, if a cement is reground and tested before the calcium oxide newly liberated from coatings of slag or calcium carbonate has become converted into hydrate, the heat produced by the slaking of this line will tend to quicken the set. This influence can be removed simply by seasoning the cement. One other influence, namely, that due to the increase in the amount of active cement, can not be removed without seriously injuring the strength of the material. However, in most instances it required only the use of a small additional quantity of retarder to overcome this influence, and with few exceptions, it is certain that manufacturers can control the set of their product even though it is all ground to an impalpable powder. (18) A brief summary of all of the results and conclusions derived from this investigation could not be made, but special attention was directed to the following: (a) That manufacturers especially should give the subject of partial regauging due consideration as such cements although apparently slow setling at the mill are apt to be quick setting when tested at their destination. (b) That the policy of using a minimum amount of gypsum has often resulted in quick setling material where 0.5 per cent, of additional retarder would have prevented all trouble. (c) That this work establishes many definite facts and clears up much misconception concerning the efficiency of various methods of treating the clinker. Taking, for instance, the investigation of H. Spencer Conover, 1 his results, far from indicating that the more quickly the clinker cooled, the more slowly the cement set, only iCement Age (1905), 3, 479-86. v] Congress of Applied Chemistry 101 offer additional corroboratory evidence to the conclu- sions arrived at by our own experiments. On the other hand, adequate reasons are given for the beneficial re- sults obtained by H. K. G. Bamber's method 1 of grind- ing the clinker in the presence of a limited amount of live steam. Taking into consideration only the hydra- tion of the free lime this method is more efficient than aeration. It is suggested that even greater efficiency might be obtained by dropping the red-hot clinker into water as soon as it leaves the rotary kiln. PART IV The Strength of Portland Cement (19) A study of the divers conclusions, uncertainty and con- fusion concerning the cause and significance of results obtained from testing the strength of Portland cement, resulted in the following observations. (a) That were it not for their characteristic drop in tensile strength, high testing rotary cements would meet with universal approval and the problem of cement testing and standardization would be greatly simplified. (b) That the failure to establish a more definite relationship between the tensile and compressive strength can be attributed partly to differences in the size of test speci- mens, the large cubes or cylinders usually used for test- ing the strength under compression being less apt to show the peculiarities of cement than the small bri- quettes used for tension test. (c) That results obtained by crushing small specimens show the up and down values common to tension test curves, but, that even when all specimens are practically the same size, the values do not go up and down coinci- dently. (d) That tension tests are as useful as determinations of com- pressive strengths, and that a retrogression in the values Concrete and Const. Eng. (1909), 4, 196. 102 Original Communications: Eighth International [VOL. of either occurs only when due to the development of undesirable contending influences. (20) It was thought that a study of the hardening properties of the numerous commercial Portland cements at our disposal together with the available knowledge concerning their chemical composition would give valuable and more definite information on several essential points. This preliminary investigation soon led to the following results and conclusions. A. Concerning the typical curve of strength as described by W. P. Taylor: 1 (a) That the characteristic fluctuations in strength could not be attributed, as suggested by Taylor, to the different rates of hardening of the different constituents of the cement and to a deterioration of the sulphates and aluminates. B. Concerning the noteworthy deductions and conclusions arrived at by W. A. Aiken 2 which relate to the influence of ulti- mate chemical composition and early gain in strength: (a) That it soon became evident that the development and the maintenance of the early strength did not depend upon the early gain in strength, or upon a narrow limit- ation of the percentage of calcium oxide, the value of the silica-alumina ratio, the hydraulic modulus, or on any other available chemical information. In fact, in most instances it was found possible to change the natural effect of all of these factors by altering the de- gree of burning, the fineness of either the raw material or the finished product, and by seasoning the cement or the clinker in a different manner or to a different extent. (b) That as stated by E. B. M'Cready, 3 "the rate at which the disruptive strain due to free lime increases in dif- ferent samples under various conditions of burning, grinding and testing, is the kernel of the nut which we . Soc. Test, Mat. (1903), 3, 413. 2 Cement Age (1903), 1, 75. demerit Age (1905), 5, 339. v] Congress of Applied Chemistry 103 ought to crack before placing too much reliance on rules deduced simply from analyses or percentages of gain." (c) That a consideration of the physical and chemical prop- erties of free and combined calcium oxides provide in itself sufficient proof that here, as in the setting phenomena, different degrees of grinding, burning, and seasoning produce marked effects to which can be attrib- uted the primary influences which operate to cause uncertainty as to the development of strength. (21) Concerning the effects of anhydrous free lime, a consid- eration of its chemical and physical properties makes the follow- ing evident: (a) That it is the lime which fails to slake before the cement has set that is apt to cause disintegration or weakness. (b) Lime burned at a white heat hydrates much more slowly and expands 50 per cent, more than lime burned at a low heat. (c) That owing to the impermeable nature of indurated neat cement the expansion due to free lime will not develop equally in specimens of different shape, size, volume, or density. (d) That this expansive force manifests itself less in lean than in rich mortars and that the action of free lime is rendered less harmful by using sands containing poz- zuolana. (e) That when the cement is submerged in cold water "this disruptive force may not develop its full value in a month or a year." 1 (f) That in air, sintered lime slakes much more slowly than in water. (g) That owing to the impermeable coating of slag and cal- cium, carbonate boiling water may fail immediately to attack the free lime in aerated cement unless regrind- ing has been resorted to. (h) That the force which operates to cause cements and mortars to disintegrate will not become apparent to . B. M'Cready. Cement Age (1905), 5, 339. 104 Original Communications: Eighth International [VOL. the eye until it has overcome the strength of cohesion developed by other constituents in the cement. That, nevertheless, if a hardened cement contains anhydrous free lime the strength, which otherwise it is capable of developing and maintaining, is certain sooner or later to be affected when this lime is permitted to slake. (22) A similar consideration of the properties of slaked lime makes it evident that its influence upon the strength also depends upon the nature of the cement and the manner in which it is used. It does not possess the ability to harden in water, but hydrau- licity may be slowly imparted to it by the presence of pulverized pozzuolana. If permitted to absorb carbon dioxide it has ce- mentive properties of its own, but the process of induration is slow and confined more or less to exposed surfaces. Also, the presence of slaked lime may affect the strength of a cement in a mechanical manner, inasmuch as it shrinks very much if per- mitted to dry, increases the plasticity, and in certain conditions decreases the permeability. (23) Concerning the effects upon the strength of various combinations of calcium oxide with silica, alumina, or iron oxide much remains to be ascertained. However, the comprehensive and thorough work of 0. Schott, 1 E. D. Campbell, 2 S. Keissman 3 and others, has been established the following: (a) That the pulverized, fused or perfectly sintered calcium compounds in calcareous cements possess the property of hardening under water without an appreciable change in volume and without the necessity of prelim- inary curing. (b) That with proper mixtures, the nearer we approach a thoroughly combined and fused clinker the less is the expansion of the resulting cement. (c) That since with the silicates the strength increases as the lime increases while with the aluminates the oppo- site is true, and since the high silicates and the low 1 Cement and Eng. News (1910), 22, No. 9-12. 2 Amer. Chem. Soc. (1904), 26, 1273. 3 Cement and Eng. News (1911), 23, No, 1-3. v] Congress of Applied Chemistry 105 aluminates require the greatest heat, it is evident that high temperatures produce high strengths. (d) That the differences in the physical properties of the various calcium compounds account for the failure of ultimate chemical analyses to reveal the true nature of commercial cements. (e) That combined magnesia like combined lime has no in- jurious effect in cement. (f) That the hardening process of hydraulic calcium com- pounds is to a large extent not limited to exposed sur- faces, but takes place throughout the mass, this being one of the main reasons that good Portland cements give more constant and reliable results than other cal- careous cements. (24) A general consideration of the physical and chemical properties of hydraulic limes of all classes offers a remarkable demonstration of the properties of free and combined lime which have been pointed out in the preceding pages. It shows: (a) That from the low gravity and unreliable hardening prop- erties of the natural pozzuolane cements, we rise in efficiency and usefulness through the hydraulic lime, slag and natural cements to the great and reliable strength of good Portland cement, solely by reason of the condition of the free and combined lime character- istic of each. (b) That everburning natural cement is essentially under- burning Portland cement, and that the evil effects re- sulting from either process are due to the production of a maximum amount of slow slaking, sintered free lime and free magnesia. (c) That a more efficient natural (or Roman) cement than that produced at present by burning cement rock in set-kilns could be manufactured by blending clay and limestone in proper proportion and then burning the mixture at a low temperature in a rotary kiln. (d) That rapid cooling is an essential to the efficiency of slag cement on account of the fact that its low lime and high silica and alumina contents are heated to liquefaction. 106 Original Communications: Eighth International [VOL. (e) That rapid cooling is not an essential to the preservation of high limed silicates and low limed aluminates in Portland cement, because the fusion has only reached the incipient state and the percentage of calcium oxide is high. (f) That owing to the unequal degree of burning to which the raw material in different parts of the set kiln is sub- jected and to fuel contamination the aggregate produced by this process usually consists of a mixture of all five classes of hydraulic cements in which the well sin- tered Portland cement clinker, typical of good rotary practice, predominates. (g) That the characteristic low early strength of set kiln cement is due to the presence of the underburned ma- terial and not to the slow cooling of the clinker. (25) These general considerations brought us to the point where we were prepared to investigate more specifically the occurrence and the cause of changes and differences in the hard- ening properties of commercial Portland cements. Obviously, the soundness of a cement is a first consideration, and the fact that any tendency toward unsoundness develops more slowly in cold than in hot water, and still more slowly in air, verifies the previous statements concerning the properties and influences of free lime. Obviously, also, cements so bad as utterly to dis- integrate when subjected to the normal tests for soundness need no further consideration until additional seasoning enables them to remain sound in cold water or air. (26) A study of the effects of seasoning on the hardening prop- erties of cement produced from soft, decidedly underburned Portland clinker revealed such erratic action that few definite statements, concerning the strength could be made. However, the points worthy of emphasis about this material are: (a) That nonseasoned, underburned Portland cement is unsound. (b) That the underburned clinker seasons readily. (c) That seasoned to soundness the cement hardens very slowly but shows a steady increase apparently for an indefinite length of time. v] Congress of Applied Chemistry 107 (d) That ordinarily any expansion due to free lime occurs soon after the set, and therefore, if the percentage of magnesia is low it is safe to assume the ultimate sound- ness if the neat mortar does not disintegrate after the usual 28-day tests. (27) A study of the effects of seasoning on the strength de- veloped by hard-burned Portland cement showed entirely dif- ferent characteristics, namely: (a) That the hard-burned clinker may, or may not, produce a cement which requires no curing to enable it to pass the accelerated test for soundness. (b) That seasoned or nonseasoned, provided the cement passes the hot tests, it may be used with reasonable certainty of its ultimate soundness; but that with hard- burned cement no reliance can be placed on the normal 28-day tests for strength or soundness, a fact which ought to be sufficient cause for the rejection of all Port- land cements which fail to pass the hot tests. (c) That under normal conditions the expansion due to free lime in indurated cement develops very slowly, so slowly in fact that it may not effect the early strength. (d) That hard-burned Portland cement hardens very rapidly and attains a great early strength, but unlike the soft- burned product the strength at the end of years is usually less than at the end of 28 or sometimes even 7 days. (28) A study of the effects of seasoning on the hardening prop- erties of the numerous sound commercial Portland cements at our disposal resulted as follows: (a) That the early strength of even hard-burned cements can be reduced to a low figure by thoroughly aerating the ground product. (b) That even prolonged aeration of the ground product has no marked effect on the ultimate strength. (c) That this effect of aeration on the strength results from the conversion of slaked lime into calcium carbonate and that the tendency of this reaction to confine itself to the surfaces of the individual particles of cement 108 Original Communications: Eighth International [VOL. fully accounts for the manner in which the hydraulic properties are retarded rather than eliminated. (d) That owing to this coating of carbonate, much depends upon the permeability and exposure of the mortar as to how quickly the inner active constituents of aerated cement become indurated. (e) That the original hardening properties of an aerated, hard- burned cement are restored almost entirely by regrind- ing; but, that while such treatment would tend to in- crease the efficiency both in sand carrying capacity and in constancy of strength and volume, it is too expensive to be practical. (f) That the characteristic drop in strength may not be elim- inated even by prolonged aeration. (29) This knowledge and a little further study of commercial products showed: (a) That, although the strength and even the character of the curve of strength of any given cement may be affected to a considerable extent by the method of molding employed, the quality of water or quality of sand used, and the exposure or seasoning resorted to, still in spite of these variable factors certain characteristics influence the strength to such an extent as to be readily apparent. (b) That a low early strength always results from premature partial regauging or caking, coarse grinding, the pres- ence in quantity of foreign substances such as clay, sand or slag, underburning, or excessive seasoning. (c) That a low early strength, provided the cause was not due to coarse grinding, is always associated with a low specific gravity (dried at 110) and a corresponding high loss by ignition. (30) These facts concerning the hardening properties of underburned and hard-burned Portland cements, etc., enable us fully to designate the general character of all the commercial cements examined and to account for the general nature of their curves of strength. Thus: (a) Adulterated, coarsely ground or caked cements were v] Congress of Applied Chemistry 109 readily detected by means of chemical or physical exam- inations. Such cements showed poor early and ulti- mate strength. (b) Underburned cements when sound have a low gravity and the microscopic and chemical examination proved, as in most instances, that the low gravity was due to underburning rather than to adulteration or excessive aeration. Such material hardened in the manner char- acteristic of seasoned, underburned Portland cement. (c) If the cement was sound, a gravity above 3.10 was posi- tive proof of hard-burning, and such material always showed the characteristic hardening properties of a well sintered Portland cement. (d) Intermediate products of the cement kilns were repre- sented by a specific gravity which was neither high nor low as evinced by the fact that usually then consisted of a mixture of soft and hard-burned cement, such materials showed no definite character in hardening properties. (31) These and similar results cleared up all misconceptions regarding the development and the significance of the early gain in strength. They showed conclusively: (a) That the efficiency in the early tests of Portland cements depends primarily upon the thorough sintering or fusing of proper raw materials and that the best commercial practice in this respect produces a sound, finely ground product with a low content of volatile constituents and a specific gravity above 3.10 (dried at 110, but not ignited) . (b) That unfortunately, it is characteristic of such hard- burned material to show considerable fluctuation and a decided decrease in strength after induration. (c) That this, and the slowness with which soft-burned cements harden, make it necessary to depend upon perfections in the manufacture of hard-burned, sintered or fused products for greater economic efficiency and certainty in concrete construction work. (d) That, in order to secure the desired permanency in 110 Original Communications: Eighth International [VOL. strength, the destructive force which apparently oper- ates after hard-burned cements have become thoroughly indurated must be discovered and eliminated. (32) Accordingly, in continuing this investigation, we con- fined our studies to a consideration of hard-burned Portland cements. We had measured the changes in volume of bars of neat and sand mortars of the commercial cements and found a very suggestive coincidence in the failure of prolonged aeration entirely to eliminate free lime, the drop in strength and an ulti- mate abnormal expansion in a hard-burned product. This led us to believe that the fluctuations in strength might be due to the same internal strains which tended to produce abnormal changes in volume. Campbell and White 1 were the first to measure the changes in volume due to free lime in hard-burned cement, and although our results verified their conclusions in most respects, we found some radical differences, namely: (a) That owing to its longer confinement in the clinkering zone, the free lime in hard-burned commercial products slaked much less rapidly than that prepared in their small experimental kiln, as long as 7 months being re- quired to slake all of the free lime in commercial prod- ucts submerged in running water. (b) That, like free lime and an ultimate drop in strength, abnormal expansion is not eliminated entirely by pro- longed aeration. (c) That, like the changes in the strength and the effects of free lime upon the soundness, much depended upon the density and permeability of the mortar as to the extent and nature of the changes in volume. (d) That owing to greater permeability and more uniform hydration, sand mortars changed in volume more than the action of the neat mortar indicated. (e) That contrary to the general belief, our results indicated that Portland cements do not expand in water unless they contain free lime (or free magnesia). (33) Our next experiments were made for the purpose of iJour. Am. Chem. Soc. (1906), 60, 273. v] Congress of Applied Chemistry 111 ascertaining the effects of changes in volume on the strength. The results obtained showed: (a) That within the limits of perfect elasticity, Portland cements may show a gain in strength in spite of consid- erable expansion. (b) That nevertheless, the strength can be increased by re- ducing the expansion. (c) That, as was to be expected, there is no direct relationship between tensile and compressive strengths during the period of time in which the cements show marked changes in volume. (d) That there is a marked relationship between the strength in tension and in compression after the volume has become constant and apparently no internal stresses are operating. (e) That fluctuations in strength are caused by internal strains resulting from the hydration of free lime and that within the elastic limit these internal strains affect the strength in tension and in compression in a dissim- ilar manner. (f) That the early strength developed by hard-burned Port- land cement is the more reliable as an indication of the ultimate strength the less free lime the material contains. (g) That these results are not contradicted by the experience of the Society of German Portland Cement Manufac- turers which failed to establish any definite relationship between the durability of strength and any of the so- called accelerated tests for constancy of volume, be- cause our results apply only to hard-burned cement. (h) That, taking as an outside instance the work of the St. Louis Testing Laboratories, 1 there seems to be a definite relationship between the durability of strength and the results of accelerated tests of hard-burned Portland cement. (i) That, as the soundness test fails to measure and often even to detect free lime or to designate its condition, the *U. S. Geol. Survey (1908), Bull. No. 331. 112 Original Communications: Eighth International [VOL. microscopic test offers much more reliable information on this point. (j) That we have found this effect of free lime so character- istic that no 28-day, or longer, tests of strength are needed to determine the fitness of cements for use. (k) That, provided that an otherwise satisfactory cement hardens in a desirable manner for, say, 7 days, and furthermore, that the microscopic tests show no more than a little free lime after regrinding this cement, then there need be no doubt of the ability of this material to harden in a satisfactory manner with age. In short, the study of a large number of commercial products has convinced us that we have here the much desired means of ascertaining the true quality of Portland cements without resorting to the prolonged tests for strength. (34) Throughout this work it was necessary to take into consideration the effects of the degree of fineness to which the cements had been ground. Obviously, the coarse, inactive par- ticles may be considered as clinker, the finer grinding of which produces a cement whose properties depend upon the same con- ditions of composition, burning, seasoning, etc., as that of the powder produced from large clinker. Therefore, the only new consideration which the subject of fineness introduces concerns itself with the permanency of the strength developed by the finest and most active particles. The results obtained proved the dur- able nature of the indurated impalpable powder, and taking into consideration, also, the fact that free lime hydrates more readily the finer its state of subdivision, the great benefits derived from commercial fine grinding is readily apparent. (35) The foregoing results and observations concerning both the temporary and ultimate strength lead to very definite con- clusions, namely: (a) That for greatest temporary efficiency it is necessary to grind fine and to burn at a high temperature. (b) That the endurance of the great early strength thus ob- tainable, the increase in strength with age and the con- stancy in volume will be the more satisfactory the less free lime (or magnesia) the indurated cement contains. v] Congress of Applied Chemistry 118 (c) That Portland cement of the desired quality can be obtained by proper mixing, hard-burning, and fine grinding of both the raw materials and the finished product. (d) That, as greatest efficiency is only obtainable at a cor- responding expense to the manufacturer, cements should be purchased on a basis of quality rather than upon a mere consideration of quantity. CONCLUSIONS (36) A brief summary of all the important conclusions arrived at can not be made, the interdependent nature of such conclu- sions preventing a brief statement of facts. However assuming that the quality in Portland cement which we need is constancy in volume and setting properties, and reliability in strength, and that it is of vital importance that this material both hardens rapidly and maintains a great strength, we believe that the enforcement of the following recommendations will increase the efficiency of the present standard cement specifications of the American Society for Testing Materials. A. Concerning the constancy of volume : (a) We can not hope to secure the desired efficiency in Portland cement unless the manufacturer is induced to burn his materials so that no season- ing is required to produce a sound cement. Therefore, it is necessary to demand perfect soundness in conjunction with a high specific gravity, and we recommend : (b) That failure to meet the requirements of the ac- celerated tests shall (in place of " need not " as now specified) be sufficient cause for rejection. B. Concerning the specific gravity: (a) That the best burning and proper storing produces a product which has a high specific gravity (or a low loss by ignition) . Therefore, (b) That the specific gravity of the cement as received (i.e., dried but not ignited) shall not be less than 8 114 Original Communications: Eighth International [VOL. 3.10 unless the loss by ignition is less than 2.00 per cent. (c) That the above recommendation provides for the possibility of a well burned cement with a lower specific gravity provided the low gravity is not due to subsequent absorption of volatile constitu- ents; but our experience does not include such a possibility. (d) That the clause " Should the test of cement as received fall below this requirement a second test may be made upon a sample ignited at a low red heat " be omitted. (e) That the clause " A low specific grayity in con- junction with a high loss by ignition is positive proof of undesirable burning, adulteration or seasoning " be substituted for the present para- graphs concerning the significance of the specific gravity. C. Concerning the fineness : (a) As the specifications now stand, there is little incentive to induce the manufacturer to grind to the degree of pulverization that modern im- provements in grinding machinery has made practicable unless his cement is so poor that extreme fineness is necessary to enable it to pass the requirements for strength and soundness. Therefore, (b) That the cement shall leave a residue of not more than 5.0 per cent, by weight on the No. 100, and not more than 20 per cent, on the No. 200 sieve. D. Concerning the tensile strength : (a) That the average of at least four briquettes repre- senting at least two separate mixtures of the same sample shall be taken for each test, excluding any results which are manifestly faulty. E. Concerning re tests : (a) Manufacturers should be impressed with the fact that these are minimum requirements; that v] Congress of Applied Chemistry 115 ample provision already has been made in the specifications for lack of uniformity in testing as well as in real quality; and that we demand a quality o superior that, regardless of the vari- able factors, the ability of the cement to pass all requirements shall be a certainty. Therefore, (b) That the results obtained from the original test shall be considered as final unless it becomes evi- dent that a serious error in sampling or testing has resulted in totally misrepresenting the quality of the cement. In other words, that " border- line " cements should be avoided as much as possible. F. Concerning the practical significance of the above recom- mendations : (a) Manufacturing conditions are such that we can not hope to secure Portland cement which con- tains no free lime. Also, it is realized that the proposed specifications are not perfect. How- ever, we believe that the enforcement of the above recommendations will support and pro- mote the best practice in grinding and burning, and accordingly, secure greater uniformity and efficiency than the present specifications. (b) Without the hearty, honest cooperation of both manufacturer and user little can be accom- plished. The degree of fineness and burning are important financial considerations to the manu- facturer, and the consumer should buy on a basis of quality. (c) The testing of a great number of commercial Port- land cements from many parts of the world has convinced us of the feasibility of these recommen- dations from both an economic and practical standpoint, and the results obtained have repu- diated all claims to the contrary. For instance, a certain manufacturer in America stated that owing to a long sea voyage he could not guar- 116 Original Communications: Eighth International [VOL. antee his cement to pass the 3.10 requirement for specific gravity. Our work showed conclusively that cement stored in good barrels undergoes very little change due to atmospheric influences and many cements imported from Europe and America show consistently a gravity above 3.10 and a low loss by ignition. There are cements which as stated in the " Introduction " of our work show the most remarkable uniformity in physical properties. (37) We desire to emphasize the importance of the calcium hydroxide-phenol microscopic test for free lime, as in every instance the physical and chemical properties of the different products examined demonstrated the accuracy and usefulness of this test. As stated, we believe that in the hands of an expert it gives more definite and reliable information regarding the constancy of strength and volume than the usual 28-day or even 3- or 6-month tests. However, there is one undesirable feature to this test; namely, that it requires considerable experience and ability correctly to interpret the significance of the phenolate crystals formed on the microscopic slide. Therefore, in order to make this test generally practicable and universally dependable it must be simplified or made quantitative. Certainly, its possi- bilities and importance warrant much more extended research in this direction than we have had opportunity to accomplish. THE CONTROL OF DUST IN PORTLAND CEMENT MANUFACTURE BY THE COTTRELL PRECIPI- TATION PROCESSES BT WALTER A. SCHMIDT Los Angeles, California The control of the dust arising from the rotary kilns in the manufacture of Portland Cement is continually becoming a more serious problem. This is partly the result of the enormous growth of the Portland Cement industry which now demands factories of such magnitude that the large volumes of gases leav- ing the stacks carry enormous quantities of dust into the atmos- phere, but is probably more directly attributable to the present trend of public opinion, which continually demands a more thorough control of fumes and smokes. The question of the proper relationships which should exist between the factory and the surrounding inhabitants has become a very important social problem. At the present time a large amount of hardship is being caused by improper action on one side or the other, often substantiated by our Courts on the ground of mere technicalities. This is a problem which should receive the closest and most thorough study by a competent body in an endeavor to establish such laws as will draw a line of equity between the different parties coming into contact through the development of our modern industries. Should every smelter, refinery and like industry emit poisonous and noxious fumes from its stacks and all other factories permit smokes and dusts to escape in unrestricted quantities, life in any large industrial center would be unbearable. On the other hand, however, most industrial furnace processes cannot be conducted without the production of large volumes of gases which usually carry from the furnaces volatilized materials and solid particles, these solids being swept along by the heavy rush of the gases. As the factories grow in size and number the damage and " nui- 117 118 Original Communications: Eighth International [VOL. sance " caused by the fumes and dusts usually assume most aggravating magnitudes and, after a certain mark is reached control of the annoying material in the stack gases becomes a necessity. At the outset we are, therefore, confronted with two conflicting factors, first the one from within the factory, which in many instances makes it utterly impossible to prevent the formation of dusts, fumes and smokes in the manufacturing proc- esses; and, secondly, the one from without, which makes it equally impossible to permit these materials to escape into the air, due to the damage to the surrounding country. A further complexity arises from the fact that the surrounding country very often demands the maintenance of the factory for its own continued prosperity. It is readily seen, that to permit a problem of such conflicting interests to take its course without thorough study and guidance, is bound to work hardships upon one party or the other, and, of necessity, will result in such unjust proceedings as have at times closed some of our most important industrial establishments or permitted other factories to continue causing damage to the territory surrounding them. A vital question confronts the people to-day in this regard, in the establishment of a rational relationship between the different parties, taking into considera- tion the various questions entering into the rapid changes of our present day industrial development. The problem of relationship between factory and farmer has become extremely critical in Southern California with the Cement Industry where the dust arising from the two large factories, the Riverside Portland Cement Company at Riverside, California, and the California Portland Cement Company at Colton, Cali- fornia, settled upon the Orange and Lemon groves in the vicinity, causing some damage. It would be out of place in this paper to comment upon the actual damage done, as all of the court testi- mony presented has so far thrown little light upon actual results. Whether the dust really does cause any injury to the trees is a question of minor importance in this particular locality, as the dust makes the groves unsightly and the farmers are convinced that actual damage has been done. v] Congress of Applied Chemistry 119 The Riverside Portland Cement Company experimented with numerous methods in an endeavor to control the dust arising from their factory, with little encouraging results, and two years ago the writer undertook the work of applying the Cottrell Elec- trical Precipitation processes to this new problem. The Cottrell Processes were invented and developed by Dr. F. G. Cottrell, now of the United States Bureau of Mines, but until recently of the Chemistry Department of the University of California. These processes were first developed in connection with the problems arising in the manufacture of sulphuric acid by the Contact Process. After the successful control of these acid mists, the processes were applied to smelter fumes, aiming directly at the control of the sulphuric acid, in these gases. The first plant has now been in steady operation for over five years, operating upon the Parting Flue in the refinery of the Selby Smelter on San Francisco Bay. The processes were later applied to the larger problem of general smelter fume control. No at- tempt will here be made to go into detail regarding the work with these processes in this field, as the early history of the processes was quite thoroughly discussed in an article published a year ago in the Journal of Industrial and Engineering Chemistry 1 and has since been extensively abstracted in other journals. 2 Further, Mr. Linn Bradley is presenting a paper to the Metallurgical Section of the Congress upon the recent work done in this field of application. In undertaking the application of the Cottrell Processes to the problem of collecting the dust in cement mills, a large number of new factors had to be dealt with, as all work done previously had been conducted upon cool moist gases either containing large quantities of water vapor, acid fumes or similar materials, all of which gave distinct electrical characteristics to the gases. The J The Electrical Precipitation of Suspended Particles, Journal of Industrial and Engineering Chemistry, Aug. 1911. 2 Mining and Scientific Press, Aug. 26, 1911; Scientific American Supple- ment, Sept. 30, 1911 ; Engineering and Mining Journal, Oct. 14, 1911 ; Engineer- ing News, Oct. 26, 1911; Metallurgical and Chemical Engineering, March, 1912; Cement and Engineering News, April, 1912; Rauch und Staub, April, 1912. 120 Original Communications: Eighth International [VOL. primary factors in the cement work are dry non-conducting gases, non-conducting dust particles, intense temperatures, large volumes of gases and large quantities of solid material carried by these gases. It is a fair estimate to assume as an average figure that a rotary kiln 100 feet long and 7 feet in diameter, oil fired as on the Pacific Coast, has a volume of stack gases of 50,000 cubic feet per minute; the gases above the combustion zone, in the stack, having a temperature of about 450 Centi- grade and carrying dust aggregating between four and five tons per day of twenty-four hours. The plan which was first suggested for handling these gases was to turn all of the gases from the entire factory into one general flue and conducting them to such a point as would bring the tem- perature within the region of our past experiences. It was soon found, however, in our first tests that the dry gases presented entirely different electrical characteristics than were encountered in treating the gases from smelter stacks and it was later found that the high temperature facilitated the uniformity of the elec- trical discharge. The plan ; therefore, adopted was to treat the gases at as high a temperature as was permitted by the properties of the structural material used in the treating apparatus. By using ordinary steel, it is quite possible to work at the average temperature of the stack gases, namely 450 Centigrade, pro- vided however, that the abnormal temporary rises in the tem- perature in these gases are prevented. The question necessarily arose regarding the possibility of maintaining satisfactory factory operation if we attempted to regulate the stack temperature too closely and extensive tests were undertaken in pyrometric control of same. Work done during the past year has shown that by in- stalling recording pyrometers in the proper place in the stacks and placing the instruments so that the " Burners " can keep close watch upon them, it is an easy matter to regulate the fires in such a way as will give practically a uniform temperature in the stacks. It was the first plan at the Riverside Portland Cement Com- pany to connect all ten stacks by a common flue which should conduct the gases into a general treating apparatus, but the factory engineers decided that it would be preferable to maintain individual control of the kilns. As the factory design was such as v] Congress of Applied Chemistry 121 to prevent installing this form of apparatus upon the ground, it was decided to allow the entire stack structure to remain intact and treat the gases after they left the existing stacks. These stacks are eighty feet high and a platform has been constructed at this level upon which the entire apparatus has been placed. In order to test out what effect the apparatus has upon the kiln conditions, we have passed the gases through the treater and directly into the air alternately without having the " Burners " know of the direction of flow of the gases. Sensitive recording pyrometers failed to indicate any effect, either in the temperature of the stack gases or the temperature of the burning zone in the kiln, as given by a radiation pyrometer. It might be well here to say a few words regarding the underly- ing principle of the Cottrell Electrical Precipitation Processes, for those who are not acquainted with the processes or who have not the time to read the references cited. Stated in a few words, the principle consists of bringing the dust laden gases under the influence of a series of electrodes some of which maintain a " si- lent " or " glow " discharge. By virtue of the discharge, the space between the electrodes becomes filled with gaseous ions and the dust particles passing through this space become charged by having these ions impinge upon them, imparting their ionic charges to the particles. The charged particles are then passed through an intense electric field which causes them to migrate in the direction of the field, which in the commercial apparatus is arranged transverse to the direction of the flow of the gases. The dust particles are by this action drawn out of the gases and deposited upon the electrodes, the gases being permitted to go their way unaffected and emerge from the treating apparatus freed from the solid particles which had been held in suspension. The distinct advantage which this process has over the me- chanical processes arises from the fact that the gases themselvei play no part in the action in the treater, except as carriers of electricity. In all mechanical processes the entire volume of gases must be acted upon in such a way as to take advantage of the differences in specific gravity between the suspended par- ticles and the gases themselves, or some similar action. In the electrical processes the gases are permitted to pass through the 122 Original Communications: Eighth International [VOL. apparatus unaffected while the particles themselves are taken hold of individually by the electrical field and acted upon in such a manner as to draw them out of the current of advancing gases and so as to precipitate them upon the collecting electrodes. In order to maintain a definite direction of migration of the dust particles, the electrodes must be given an unidirectional elec- trostatic charge, which in the commercial apparatus is generated by rectifying a high tension alternating current. In the com- mercial apparatus the potential varies under different conditions from 20,000 to 40,000 volts. The electrode system consists of two forms of electrodes; first, the discharge electrodes, which are made in various forms, de- pending upon conditions, but are always of a light construction and are so chosen as to maintain a heavy electrical discharge from them; secondly, the collecting electrodes upon which the solids are precipitated. These are usually of a heavy construction and the form and arrangement is so chosen that no discharge takes place from their surface. The two forms of electrodes are alter- nated across the apparatus with an electrode spacing of from two to six inches, this distance varying with the conditions to be met. A series of these rows of electrodes is placed in the treater so that the dust particles are brought under the successive action of this series of electrodes. The length of the treater is so chosen as to affect the desired cleaning of the gases. The cross section of the apparatus is made such as to bring a balance between the two forces acting upon the suspended particles, namely, the fric- tional force tending to carry the solid particles along with the gases and the electrical force tending to draw the suspended particles out of the advancing current of gases. In the installation at the Riverside Portland Cement Company, the treater has a cross section of 12 x 16 feet, and an over-all length of 20 feet. As stated above, the apparatus is placed upon a platform constructed at a height corresponding to the top of the original stacks, namely, 80 feet above ground. Upon this plat- form is placed a short stack extension which extends through the roof of the building structure and is supplied with a damper of special design. Upon either side of the stack is placed a complete electrical treater separated from the stack extension by a large v] Congress of Applied Chemistry 123 louvre damper. By means of these three dampers the gases can either be conducted through one or the other or both treating chambers or emitted directly into the atmosphere as occasion may warrant. Each stack is equipped with two treaters so as to have an auxiliary apparatus for each kiln. In case one treater should be shut down for repairs, cleaning or the like, the other treater will be able to take care of the gases with moderate effi- ciency. Under normal conditions the gases will be permitted to pass through both treaters, insuring a thorough cleaning of the gases. The electrode spacing is here chosen at 6 inches and there are twenty rows of discharge electrodes in series. The dust is precipitated upon the collecting electrodes which are cleaned once every three or four hours by being given a mechanical rapping, this action being made automatic in so far that the operator merely puts an electric motor into operation. The dust falls into hopper bottoms from which it is again conducted into the bins feeding the rotary kilns. Each treater is supplied with a small outlet stack 20 feet high which is sufficient to compensate for the resistance offered to the gases by the treater. As stated above, sensitive pyrometers do not indicate any change of temper- ature in the kilns or stacks when the gases are permitted to escape directly into the atmosphere or are passed through the treater. The operating costs of the apparatus are low. A complete treater of the size described consumes approximately 1\ kilowatt hours. This includes electrical energy for all motors. A 5,000 barrel mill will, therefore, consume approximately 75 kilowatt hours in an entire installation. The manual labor required con- sists of one man 'per shift, of the character of men ordinarily employed to run electrical mill machinery. It is, however, usually advisable to have an extra man on duty. There is no deterioration in the apparatus under steady running and the machinery is subject to exactly the same wear as any other piece of electrical machinery. The progress of the work at the Riverside Portland Cement Company can probably best be illustrated by the series of photo- graphs shown herewith in which No. 1 shows the factory as it appeared in 1910 with the experi- 124 Original Communications: Eighth International [VOL. mental flue No. 1 and laboratory on the ground in front of the building. No. 2 shows experimental treater No. 2 on roof of factory. No. 3 shows experimental treater No. 3 on top of stack. No. 4 shows experimental treater No. 4 also on top of stack. Nos. 5 and 6 show experimental treater No. 4 with power "on" and " off ", (pictures taken 1 minute apart). No. 7 first part of permanent installation. No. 8 view across permanent treater. Other views will be shown at the Congress as lantern slides, showing the details of the apparatus, and the details of con- struction will be described in conjunction with these views. One important question which has grown out of the present work at the Riverside Portland Cement Company lies in the possible utilization of the collected material as a source of potash for fertilizer purposes. This factory does not use clay in its raw mix but uses a decomposing feldspar which has a considerable potash content. In the .burning of the cement the potash is volatilized and condenses again in passing up the stack. The greater part is caught in the electrical treater along with the dust which gives a dust containing considerable potash value. Ex- periments have been conducted for some little time in the en- deavor of utilizing this material either directly as a fertilizer " filler " or extracting the potash from the material with an aim of obtaining a concentrated potash salt. This work is not suffi- ciently far advanced to permit publishing definite figures. Dr. F. G. Cottrell, inventor of the process, is presenting a paper to this Congress before the Section of Conservation and Political Economy upon what he has termed " An Experiment in Public Administration of Patent Rights." This deals with the recently organized " Research Corporation " to which Dr. Cot- trell and his associates have given the Patent Rights to the proc- esses for all of the United States except the six western states and the application of the processes to the Portland Cement industry. These latter rights, along with the foreign holdings, are retained by the parent Companies, the Western Precipitation Company and the International Precipitation Company, with offices at Los Angeles, California. No. 1. No. 2. No. 3. No. 4. No. 5. *"tansfr2 No. 7. No. 8. (Abstract) GLASS FORMULAS. A CRITICISM BY ALEXANDER SILVERMAN, M. S. Professor of Analytical Chemistry and Lecturer on Glass Manufac- ture, University of Pittsburgh, Pittsburgh, Pa. The writer shows the tendency toward publication of many incorrect formulas, citing special cases from a number of books and journals, and asks co-operation of authors for elimination of such errors so as to raise standard for manufacture of glass, and of literature on this subject, to the high plane occupied by other exact sciences. 125 THE VISCOSITY OF MOLTEN GLASSES PAPER I, THE VISCOSITY OF BORATK GLASSES BT HOMER F. STALBT Ceramic Engineering Laboratories, Ohio State University INTRODUCTION The object of this research was to establish, in a qualitative manner at least, the laws of viscosity for molten glasses. An understanding of these laws is of fundamental importance to the silicate industries. The viscosity of molten borate glasses is treated in the present paper; investigations on silicate glasses are now in progress. METHOD The method used for measuring the relative viseosities of the glasses was a modification of that of Tamman, which has been used by Greiner, 1 Arndt, 2 Doelter and Sirk, 3 and others. The principle involved is that of measuring the time of fall through a certain distance of an immersed body of given volume and known effective weight. If the weight is varied in the same liquid, the product of " weight X time" gives a constant known as the " fall-product." The " fall-product " has been shown by Lan- denburg 4 to be proportional to viscosity. In this investigation, an analytical balance, sensitive to 0.10 mg., was fitted with a beam 24 inches long. To one end of the beam was attached a small scale pan, to the other a platinum wire about 5 ft. long. Inserted near the upper end of this plantinum wire was a small rod having two deeply ground grooves exactly 1 inch apart. These grooves served as markers for getting the J N. Jahrb. f. Mineral, 1906. II, p. 152. z z. Electrochemie, XIII, p. 578. 'Monatshefte, XXXII, p. 643. 4 Ann. d. Physik. XXI, 287. 127 128 Original Communications: Eighth International [VOL. time of fall. The wire contained also a small turn-buckle, for use in adjusting the height of the plunger. To the lower end of the platinum wire was attached a platinum plunger consisting of a hollow cylinder 16.5 X 19 mm. outside dimensions, with walls 0.7 mm. thick, and weighing 16.5 grams. The hollow cylinder shape was adopted on account of its giving a large area for inter- nal friction in a small volume, and at the same time being easy to clean. The charge was contained in a platinum crucible 4 c. m. in diameter by 6.5 high. The heating was done in part in a platinum resistance furnace and in part in a small gas-fired pot furnace. We were able to get check readings on the same glass heated by the two methods. The temperatures were determined by a pt pt rd couple, immersed naked in the melt, and a Hartmann and Braun gal- vanometer. The couple and galvanometer were standardized against a similar outfit calibrated by the U. S. Bureau of Stand- ards and kept at the University for standardization purposes. The time of fall was obtained by means of a telescope, sighted on the small marked rod, and a stop-watch, reading 0.10 seconds. Care was taken to have the rod well in motion before the first mark crossed the hair line of the telescope. Parallax was avoided by having a scale fixed rigidly to a table behind the small rod. This method of obtaining the time of fall was considered superior to an automatic recording device in conditions such as these in which the time interval varied largely and in which there was considerable variation in temperature that would be liable to throw out of order any delicate mechanism. To an experienced observer, any variation in rate of fall, due to sticking, etc., dur- ing one reading is easily discernible. The charges were melted in the platinum crucible and carried, with frequent stirring, to a temperature above that at which any determinations were made and sufficiently high to render them fluid enough to insure homogeneity. In all cases beautifully transparent liquids were secured. In a number of cases, after a series of determinations a charge was allowed to cool, was re- ground, and another series run the next day. In these cases check results were obtained. We were also able to get good check results on separate charges of the same composition, the deter- v] Congress of Applied Chemistry 129 minations being run at intervals of several months. All readings were taken as the temperature was lowered step by step. On changing the temperature, an interval of half an hour, during which the charge was frequently stirred, was allowed to elapse in order to bring the whole charge to a uniform temperature. MATERIALS The materials used were B 2 3 3 H 2 0, Ba CO 3 , Ca CO 3 , Sr CO 3 , Mg CO 3 . The purest materials obtainable were used. Analyses showed that none of them contained more than a few thousandths of one per cent, impurities. VOLATILIZATION OF B 2 3 In order to allow for differential volatilization of B 2 3 , the various glasses were analyzed after the determinations were finished. It was found that differential volatilization has taken place only in the mixtures containing less than 0.5 formula weights of base to 1 B 2 3 . The composition listed as 0.0272 Ba O 1 B 2 O 3 was weighed out to contain 0.25 Ba 0, and the one listed as 0.1787 Ba O 1 B 2 3 was weighed out to contain 0.16f Ba O. Any loss that took place in compositions containing 0.5 formula weights or more of base must have consisted of volatilization of borates and not of B 2 3 alone. THE RESULTS A summary of the results obtained is given in the accompanying tables and curves. The individual determinations checked as closely as could be hoped for, considering the nature of the liquids and the temperatures employed. In no case did an individual reading vary from the mean more than 5%, and in the large majority of cases, the maximum variation from the mean was less than 2%. 130 Original Communications: Eighth International [VOL. B 2 3 TABLE I. 0.0272 BaO, 1 B 2 O 3 Temperature Degrees Centi- grade Overweight in Grams Number of Readings Fall Product in Grams x 0.10 Seconds Fall Product Average Temperature Degrees Centi- grade _a Is 11 o " & Jl |l Fall Product in Grams x 0.10 Seconds Fall Product Average 708 7. 4 1799 1799 1000 4. 5 248 722 10. 4 1445 1445 6. 4 254 251.0 750 5. 4 1110 1110 1138 2. 5 110 764 7. 4 973 3. 4 111 110.5 10. 4 968 970 1180 2. 4 85 805 3.75 5 754 3. 4 85.5 85.0 3 755 754.5 1236 2. 4 70 820 2.5 3 715 715 3. 4 69 69.5 860 2.5 4 610 1333 1. 4 49 3.75 4 613 1.5 4 48 48.5 5. 4 612 612 935 7. 4 435 10. 4 441 A or\ er 438 0.1787 BaO, 1 B 2 O 3 940 2.5 3.75 4 4 430.5 431 903 3. 5 227 5. 4 432 431 4. 4 228 227.5 955 7. 4 401 401 917 3. 4 209 972 5. 4 375 375 4. 4 210 209.5 1000 2.5 5 340 1000 1.5 10 76 3.75 326 2. 4 77 76.5 5. 5 328 331 1042 1. 4 40 1077 4. 4 247 1.5 5 39.2 39.6 5. 8 245 246 1152 0.5 4 17.12 1111 3. 5 214 0.7 4 17.15 17.13 4. 4 216 215 1265 0.4 5 10.0 1194 2. 4 146 0.5 4 10.0 10.0 3. 4 144 145 1360 0.3 4 7.35 1250 1.5 7 127.5 0.4 4 7.20 7.27 2. 4 127 127 1347 1.5 4 104.5 2. 4 102.5 103.5 v] Congress of Applied Chemistry 131 TABLE II 0.5 BaO, 1 B 2 S 1 BaO, 1 B 2 3 73 ft .a j n || O 00 M || Product in ams z 0.10 Seconds B Qi *> imperature rees Centi-N grade n arweight in N Grams *O 03 Product in ams x 0.10 Seconds 11 Product Average P| K 15 5T s| 6 * 2 915 1.5 4 163.00 1165 0.3 5 3.51 3.51 2.0 4 164.50 164.00 1220 0.2 4 3.20 1025 0.3 5 18.00 0.3 3 3.10 3.15 0.4 6 18.06 18.03 1275 0.2 5 2.88 1140 0.2 5 5.22 0.3 4 2.88 2.88 1165 1220 1275 0.3 0.2 0.2 0.3 0.2 3 5 5 4 4 5.15 4.20 3.80 3.90 3.60 5.18 4.20 3.85 1.56 BaO, IBaO, 0.3 4 3.68 3.64 985 0.2 4 5.25 5.25 1330 0.2 4 3.50 1010 0.2 4 4.05 0.3 4 3.60 3.55 0.3 4 4.20 4.12 1065 1165 1275 0.2 0.3 0.2 0.3 0.2 4 4 4 4 5 3.60 3.72 3.10 3.12 2.80 3.66 3.11 2.80 0.661 BaO, 1 BzOs 1100 1125 1165 0.2 0.2 0.3 0.2 0.3 4 6 4 4 4 7.25 5.00 5.18 3.75 3.72 7.25 5.09 3.73 2 BaO, 1 B2O S 1220 0.2 5 3.40 1165 0.2 4 3.02 0.3 4 3.45 3.42 0.3 4 3.00 3.01 1292 0.2 5 3.12 1220 0.2 4 2.85 0.3 4 3.30 3.21 0.3 4 2.85 2.85 1275 0.2 4 2.65 * 0.3 4 2.66 2.655 132 Original Communications: Eighth International [VOL. TABLE III 0.5 CaO, 1 B 2 3 0.25 CaO, BaO, 1 B 2 O 3 Temperature Degrees Centi- grade Overweight in Grams Number of Readings .9 if -8 3