lOiiilt m i L B M DflS l.b3 i INTERIOR iU i3j.AX£;i3 vjrxiiV^juv^GICAL SURVEY GEORGE OTIS SM1.TH, DlBBcroE BUKLETIN 588 THE CONSTITUTION OF THE NATURAL SILICATES BY FRANK WIGGLESWORTH CLARKE WASHINGTON GOVERN ENT PRINTING OPFIOB 1914 MEMCAL ^Clnl®(DL Digitized by the Internet Archive in 2007 with funding from IVIicrosoft Corporation http://www.archive.org/details/constitutionofnaOOclarrich DEPARTMENT OF THE INTERIOR UNITED STATES GEOLOGICAL SURVEY GEORGE OTIS SMITH, Dibkctoe BUtiLETIN 588 THE CONSTITUTION OF THE NATURAL SILICATES BT FRANK WIGGLESWORTH, CLARKE WASHINGTON GOVERNMENT PRINTING OFFICE 1914 CONTENTS. Chapter I. Introduction 1 5 Chapter II. The silicic acids 10 Chapter III . The silicates of aluminum I9 General relations , 19 The nephelite type 21 The garnet type 24 The feldspars and scapolites 34 The zeolites 40 The micas and chlorites 51 The aluminous borosilicates 65 Miscellaneous species 74 Chapter IV. Silicates of dyad bases 87 Orthosilicates 87 Metasilicates 94 Disilicates and trisilicates 107 Chapter V. Silicates of tetrad bases, titanosilicates, and columbosilicates 113 Appendix 124 Index 125 3 frr» I I THE CONSTITUTION OF THE NATURAL SILICATES. By Frank Wigglesworth Clarke. CHAPTER I. INTRODUCTION. In the solid crust of the earth the sihcates are by far the most important constituents. They form at least nine-tenths of the entire known mass and comprise practically all the rocks except the sand- stones, quartzites, and carbonates, and even these exceptions are com- monly derivatives of the silicates, which break up under various condi- tions, yielding new bodies of their own class, together with free sihca and limestone. From the geologist's point of view, therefore, the sihcates are of fundamental importance, and a study of their inner constitution may be reasonably expected to shed light upon many serious problems. For example, every primitive rock or eruptive mass contains an aggregation of silieates, each one of which is capable of undergoing chemical change in accordance with limitations imposed by the structure of its molecules. When these changes take place secondary compounds, alteration products, are formed, and in time the rock becomes transformed into new substances, quite unlike those which originally existed. A knowledge of the processes which thus occur should be apphcable to the study of the rocks and should ulti- mately render it possible so to investigate a metamorphosed mass as to clearly indicate its origin. These processes are dependent on chemical structure, and the study of this with regard to the sihcates is the pur- pose of the present memoir. From the standpoint of the chemist the problem under consideration is one of great importance but also of great difficulty. Some of the difficulty is real, some only apparent. At first sight the natural sih- cates appear to be compounds of great complexity, but this difficulty becomes much less serious after careful examination. Few of the nat- ural sihcates exist in even an approxunately pure condition; many that seem fresh have undergone traces of alteration; isomorphous mix- tures are exceedingly common; and much confusion is due to defective analyses. By multiphed observations these difficulties can be ehmi- nated from the problem, but others yet remain to be disposed of. The 6 THE CONSTITUTION OF THE NATUEAL SILICATES. organic chemist, to whom most of our knowledge of chemical structure is due, deals mainly with bodies of known molecular weight, which can be measured by the density of a vapor or by cryoscopic methods. To the mineral chemist such knowledge is not available, for the com- pounds which interest him are neither volatile nor soluble, and their molecular weights can only be inferred. The simplest empirical for- mula of a siHcate is not necessarily its true formula ; the latter may be a multiple or polymer of the former; and here we find a difficulty which is at present almost insuperable. Strong evidence can be brought to bear upon this side of the question, but it is only partial evidence and not finally conclusive. The case, however, is by no means hopeless, for even the partial solution of a problem is better than no solution at all. An approximation is some gain, and it is possible so to investigate the constitution of the sihcates as to bring many relations to light, developing formulae which express those rela- tions and indicate profitable lines for future research. The problem is open to attack along several lines, and various methods of investigation can be brought to bear upon it. First, of course, the empirical formula of each sihcate must be definitely ascer- tained, which involves the discussion of sufficiently numerous analyses and the efimination of possible errors due to impurity, alteration, and isomorphous admixtures. In this work the microscope renders impor- tant service to the analyst and makes his results much more certain. By the aid of the microscope many supposed mineral species have been proved to be mixtures, and the problem of the silicates has been thereby simpHfied. Indeed, the final outcome of such investigation generally indicates, for any given natural sihcate, simplicity of compo- sition, and this is what should be expected. These compounds are, as a rule, exceedingly stable salts, whereas complex substances are com- monly characterized by instabifity. The mineral sihcates are formed in nature under conditions of high temperature or are deposited from solutions in which many reactions are simultaneously possible, and these circumstances are strongly opposed to any great comphcations of structure. Furthermore, they are few in number, only a few hun- dred at most being known; whereas, if complexity were the rule among them, sHght variations in origin should produce corresponding variations in character, and millions of different minerals would be generated. That few variations exist is presumptive evidence that only few are possible, and hence simphcity of constitution is reason- ably to be inferred. In fact, we find the same small range of mineral species occurring under the same associations in thousands of widely separated locahties, a few typical forms containing a few of the com- monest metals being almost universally distributed- The longer the evidence is considered, the stronger the argument in favor of simple silicate structures becomes. INTRODUCTION. 7 The empirical formula of a silicate having been estabUshed, its physical properties may next be considered, and of these the crystal- line form and the specific gravity are the most important. From identity of form, or complete isomorphism between two species, we infer similarity of chemical structure, and the inferences thus drawn are often of the highest value. On the other hand, dissimilarity of form and identity of composition indicate isomerism, as for example in the cases of andalusite and kyanite, and here again we obtain evidence which bears directly upon the study of chemical constitution. From the specific gravity the so-called molecular volume of a species may be computed, and that datum gives suggestions as to the relative con- densation of a molecule in comparison with others of similar empirical composition. For instance, leucite and jadeite are empirically of similar type, but the latter has by far the greater density, together with superior hardness. It is therefore presumably more complex than leucite, and this supposition must be taken into account in con- sidering its ultimate formula. From what may be called the natural history of a mineral still another group of data can be drawn, relating to its genesis, its con- stant associations, and its alterabihty. In this connection pseudo- morphs become of the utmost interest and, when properly studied, shed much light upon otherwise obscure problems. An alteration product is the record of a chemical change and as such has weighty significance. The decomposition of spodumene into eucryptite and albite, the transformation of topaz into mica, and many like occur- rences in nature are full of meaning with reference to the problem now under consideration. Just here, however, great caution is nec- essary. Mineralogic literature is full of faulty records regarding alterations, and many diagnoses need to be revised. Pseudomorphs have been named by guesses, based on their external appearance, and often a compact mica has been called steatite or serpentine. Every alteration product should be identified with extreme care, both by chemical and by microscopical methods; for without such precau- tions there is serious danger of error. Each supposed fact should be scrupulously verified. Closely allied to the study of natural alterations is their artificial production in the laboratory. The transformation of leucite into anal- cite, and of analcite back into leucite, is a case in point, and the admi- rable researches of Lemberg furnish many other examples. Work of this character is much less difficult than was formerly supposed, and its analogy to the methods of organic chemistry renders its results highly significant. Atoms or groups of atoms may be spHt off from a molecule and replaced by others, and the information so gained bears du-ectly on the question of chemical structure. With evidence of this sort relations appear which could not otherwise be recognized, and 8 THE CONSTITUTION OF THE NATURAL SILICATES. these relations may be closely correlated with observations of natural occurrences. Evidence of the same or similar character is also furnished by the thermal decomposition of silicates, a line of investigation which has been successfully followed by several investigators. Thus garnet, when fused, yields anorthite and an olivine ; talc, on ignition, Uberates silica; and the prolonged heating of ripidolite produbes an insoluble residue having the empirical composition of spinel. All such facts have relevancy to the problem of chemical constitution, and their number could easily be enlarged by experiment. As yet the field has been barely scratched on the surface; upon deeper cultivation a goodly crop may be secured. The artificial S3nithesis of mineral species, with the alUed study of crystalline slags and furnace products, furnishes still more evidence of pertinent utility. But here again caution is needed in the interpreta- tion of results. A compound may be produced in various ways, and it does not follow that the first method which is successful in the labora- tory is the method pursued by nature in the depths of the earth. The data yielded by synthesis are undoubtedly helpful in the determination of chemical constitution, but they furnish only a small part of the proof needed for complete demonstration, and their apphcability to geologic questions is extremely limited. For the latter purpose they are only suggestive, not final. Suppose now that the empirical formula of a siUcate has been accu- rately fixed, and that a mass of data such as I have indicated are avail- able for combination with it. Suppose the physical properties to be determined, the natural relations known, the alteration products observed, its chemical reactions and the results of fusion ascertained ; what then? It still remains to combine these varied data into one expression which shall symboHze them all, and that expression will be a constitutional formula. To develop this, the established prin- ciples of chemistry must be intelligently applied, with due regard to recognized analogies. The grouping of the atoms must be in accord with other chemical knowledge; they must represent known or probable sihcic acids; and any scheme which fails to take the latter consideration into account is inadmissible. Not merely composition, but function also is to be represented, and the atomic linking which leaves that disregarded may be beautiful to see but is scientifically worthless. A good formula indicates the convergence of knowledge; if it fulfills that purpose it is useful, even though it may be supplanted at some later day by an expression of still greater generality. Every formula should be a means toward this end, and the question whether it is assuredly final is of minor import. Indeed, there is no formula in chemistry to-day of which we can be sure that the .last word has been spoken. INTEODUCTION. 9 In the development of constitutional formulae for the silicates it sometimes happens that alternatives offer between which it is difficult to decide. Two or more distinct expressions may be possible, with the evidence for each so strong that neither can be accepted or abandoned. In such cases nothing can be done but to state the facts and await the discovery of new data, to which, however, the formulae themselves may give clues. This sort of uncertainty is pecuHarly common among the hydrous siUcates, and often rises from the difficulty of discriminating between water of crystallization, so called, and constitutional hy- droxyl. This difficulty is furthermore enhanced by the common occurrence of occluded water or water in so-caUed ''solid solution," and also by the adsorption of water when a mineral is pulverized for analysis. The serious nature of the latter compHcation was not rec- ognized until quite recently. In discriminating between rival formulae one rule is provisionally admissible. Other things being equal, a sjnnmetrical formula is more probable than one which is unsymmetrical. Sjrmmetry in a molecule conduces to stability; most of the sihcates are exceedingly stable; and hence symmetry is to be expected. This rule has presumptive value only, as an aid to judgment, and can not be held rigidly. It expresses a probabifity but gives no proof. In a problem like that of the sili- cates, however, even a suggestion of this kind may render legitimate assistance. There is an extensive literature relative to the constitution of the sihcates, which, however, has been well summarized by Doelter,^ whose summary need not be duphcated here. When necessary suit- able reference will be made to the different authorities. I Handbuch der Mineralchemie, vol. 2, pp. 61-109, 1912. CHAPTER II. THE SILICIC ACIDS. If aU the silicates were salts of a single silicic acid the problem of their constitution would be relatively simple, but this is not the case. Many silicic acids are theoretically possible, and several of them have representatives in the mineral kingdom, although the acids them- selves, as such, are not certainly known. Their nature must be inferred from their salts, and especially from their esters, and this side of the problem is the first to be considered. As sihcon is quadrivalent, its orthoacid is necessarily represented by one atom of the element united with four hydroxyl groups, thus — Si(0H)4, or, structurally: H I O H— O— Si— O— H I A To this acid, orthosilicic acid, the normal silicic esters and many common minerals correspond. Its normal salts, reduced to their simplest expressions, may be typically represented as follows: Types. Examples. R^SiO, (C,H,),SiO, R",SiO, M&SiO, R^^^CSiOJe Al,(SiO,)3 R-SiO, ZrSiO, Any silicate in which the oxygen atoms outnumber the silicon atoms by more than four to one, as, for example, the compound Al2Si05, must be regarded as a basic salt. By eUmination of water orthosihcic acid may be conceived as yield- ing, first, metasilicic acid, HgSiOg, and, secondly, the anhydride, SiOj, thus; Si(OH), 0=Si=(0H)2 0-=Si=-0 10 THE SILICIC ACIDS. H Many salts that correspond to metasilicic acid are known, but no esters have yet been certainly obtained. The esters first described by Ebelmen were supposed to be metasilicates, but aU recent investiga- tions have shown them to be ortho compounds, possibly more or less impure. Troost and Hautefeuille, however, have described an ester having the formula (C^HJgSi.O.^, which is a polymer of a metasiUcate, but its true nature has not been determined. The simplest formulae for typical metasilicates are as follows : Types. Examples. R^SiOa Na^SiOa • R^SiOg MgSiOg R^"2(Si03)3 Al,(Si03)3 K-CSiOs)^ ZrCSiOe)^ The last two examples, AI2 (8103)3 and Zr (8103)3, are salts not actually known, but theoretically possible. By the coalescence of two molecules of orthosilicic acid and suc- cessive elimination, molecule by molecule, of water, a series of disihcic acids may be produced, thus : Si(OH), 8i=(OH)3 0=Si— OH 0=81— OH — H,0 6 — H,0 O — H,0 Si(OH), 8i=(OH)3 8i=(OH)3 0=8i— OH The first of these new acids, orthodisiHcic acid, Hg8i207, is a sex- basic acid of which several esters are known. It is therefore well established, and a number of minerals appear to be salts of it. The second acid is a polymer of metasilicic acid, dimetasilicic, and its formula is H48i206. The third compound, metadisilicic acid, H28i205, is represented by no esters, but among its salts are the minerals mordenite, ptilolite, milarite and petaHte. By removing the last molecule of water the group 8120^ would remain, a multiple of 8i02. By a similar process, that is, by the elimination of water from three or four molecules of orthosilicic acid, a series of trisilicic and quadri- silicic acids may be theoretically developed. These higher acids offer many possibihties for isomerism, just as we know to be the case among the hydrocarbons. For the present, however, only the tri- silicic series need be considered, for above that series the long chains of atoms would presumably be unstable. At all events the higher series are at present unnecessary for the interpretation of known minerals. 12 THE CONSTITUTION OF THE NATURAL SILICATES. The trisilicic acids are important and develop as follows : Si=(0H)3 I O Two isomers. Two isomers. 0=Si— OH Si=(0H)3 0=Si— OH 0=Si~OH A i A i Si=(0H)2 Si=0 A A Si==(0H)2 Si=0 A A Si=(OH), A Si=(0H)3 Si=(0H)3 Si=(0H)3 0=Si— OH Si=(OH), HsSigO.O HeSigO^ HeSigO^ H^SigO^ H.SigO^ Still another acid is possible to complete the series, HgSigO^, to which, however, no known minerals correspond.^ The first acid of the series, orthotrisiUcic acid, has several representatives in the mineral kingdom. The second and third, the trimetasilicic acids, are polymers of metasilicic acid and make, with the similar acids of the previous series, four of the same general formula, TiHgSiOg. To these acids the four known modifications of magnesium metasilicate may perhaps correspond. The fourth and fifth acids are most impor- tant, for they represent the feldspars and appear also in some micas, the scapolites, and several other species. Their isomerism is most suggestive and possibly accounts for such pairs of minerals as ortho- clase and microcline, or eudidymite and epididymite, although the latter case is doubtful. The simple name trisihcic acid may be assigned to them, for in abundance their salts outrank all the other acids of the series. Now, by including the quadrisilicic acids for the moment, ignoring isomers, and tabulating the several compounds,^ some interesting relations appear: Dehydration derivatives of orthosilicic acids. Orthoacids. First anhydride. Second anhydride. Third anhydride. Fourth anhydride. " Fifth anhydride. H4Si04 H^SiOs-..- H4Si206---- HeSi30,.... HsSi^O,^-- Si02 HfiSioO, H^Si^Os-..- g^SisOs..-. HeSi^On-- Sio04 HoSLOin H^SiaO^--.. H4Si40,o-- SisOe H2Si409-.-. HioSiAa siA- 1 This acid is assumed by Tschermak to be the acid of albite. 2 This form of tabulation has also been employed by Tschermak, Zeitschr. physikal. Chemie, vol. 53, p. 350, 1905. THE SILICIC ACIDS. 13 This table can be extended indefinitely, with the result that in each vertical column every member below the first differs from the one preceding it by the addition of HjSiOg. Furthermore, the first anhydride in each series is either HgSiOg or a multiple thereof. That is, we have a number of homologous series, quite similar to those with which oi^anic chemistry has made us familiar. The final anhy- dride in every series will be a multiple of SiOg, and that fact seems to shed some light upon the possible differences between quartz glass, tridymite or cristobaUte, and quartz. The commonest associates of quartz are the trisihcate feldspars, to which quartz may be related in respect to its molecular magnitude. Tridymite and cristobalite, with lower specific gravity, are less condensed than quartz and may belong in the disihcic series. The still lighter quartz glass is perhaps the simplest molecule of all, SiOg. This is hardly more than pure specula- tion, but the observed relations are certainly suggestive. The denser forms of siUca are surely polymers of SiOa. In the foregoing discussion the silicic acids have been represented by ^' chain" formulae, analogous to the formulae of the aliphatic hydro- carbons. But ''ring" formulae of several types are also possible, and some authorities prefer them. For example, one type is as follows: (OH), (OH), (0H)3 ^i o Si Si V ^\ (0H)3 Such formula can be extended indefinitely, but no matter how many siUcon atoms are introduced into the ring the saturated compound wiU be a metasiUcic acid, nU^SiO,. The successive anhydrides wiU correspond empirically but not structurally to some of the acids of the previous scheme, although none can be equivalent to the higher orthoacids. This limitation makes the ring system less general than the linear or chain system of expressions. Such acids as H,SiO„ H^Sifi^, and HgSigOio are impossible under it. By the coalescence of two or more rings, such as is common among the aromatic hydrocarbons, still more complex acids are conceivable, thus: O O (OH),=Si'^ ^Si-^ ^Si=(OH), vv 14 THE CONSTITUTION OP THE NATURAL SILICATES. or H^SigOg, isomeric with the important trisilicic acids of the chain series. Again, ^ (0H)2==Si Si Si Si=(0H)2 ^y ^y \y or HaSi^Oio, a polymer of the disihcic acid HaSigOg, and so on indefi- nitely. Here again the limitation holds that the acids with a higher oxygen ratio than appears in the formula H4Si30g are excluded from the scheme. With triple linkings of oxygen only one siUcic acid is immediately possible, namely, O H— O— Si- O— Si— O— H V or HgSiaOg, another isomer of the disiUcic acid in the chain series. Two such rings, however, may be linked together by an oxygen atom, thus: O O H _-0— Si— O— Si— O— Si— O— Si— O— H O O or HaSi^Og, an acid which corresponds to no known compounds. All possible acids which appear in the ring formulae are included in the chain system, at least so far as their empirical formulae are considered. It is evident, therefore, from what has been already demonstrated, that the chain system is the most complete and general. It is not necessary, in the present state of knowledge, to go beyond it, although this conclusion should only be held tentatively. It is possible that some of the simpler rings may help to interpret some cases of isomerism.^ So far, then, there are only a moderate number of siHcic acids whose salts appear to need consideration in interpreting the natural silicates. They are: Orthosilicic acid H4Si04 Metasilicic acid HgSiOg Orthodisilicic acid HeSigOy Dimetasilicic acid H4Si206 Metadisilicic acid HaSigOg Orthotrisilicic acid HgSigOio Trimetasilicie acid HeSigOg (two isomers) Trisilicic acid H4Si308 (two isomers) 1 Ring formulae, like some of those given in the text, are used by Vemadsky, Zeitschr. Kryst. Min., vol. 34, p. 37, 1901. THE SILICIC ACIDS. 15 Many other acids are theoretically possible, and one of them, HgSi^Oi^, is perhaps represented by Troost and Hautefeuille's ester^ (C2H5)8Si40i2. Salts of such acids may occur in the mineral kingdom, but so far as present evidence goes the probability of their existence is very small. If the natural siHcates were simple normal salts of a few siUcic acids the problem of their constitution would not be difficult. But rela- tively few of the known species are of this description; the greater number are double salts, and even triple replacements are not uncom- mon. Furthermore, there are acid and basic salts to be interpreted, and the latter class offers the most serious difficulties. A basic meta- siHcate, for example, may have the same empirical composition as an orthosihcate, so that its ratios, studied apart from other evidence, tell nothing as to the class in which it belongs. For instance, the formula AljSiOg, which represents the composition of three distinct minerals, andalusite, sillimanite, and kyanite, admits of several different structural expressions. As a basic metasificate it may be written /SiOa Al — O — ^Al and as an orthosihcate it becomes either 0=A1— SiO^^Al, or SiO^^ If its molecular weight is a multiple of that indicated by the formula AlaSiOg, then the possibilities of isomeric structure become still more complicated. Its composition alone does not give its molecular struc- ture, and other evidence, as shown in the introduction to this memoir, must be brought to bear before the problem can be even approxi- mately solved. This evidence is sometimes available, sometimes not, as will be seen in the systematic discussion of the individual species later. A similar but less troublesome difficulty arises from the common occurrence of mixed salts, which may represent one or more silicic acids. For example, a well-crystallized silicate on analysis gave the following empirical formula: Na2CaAl4Si8024. At first sight this appears to be a rather complicated metasilicate, but microscopic evidence shows that the mineral is a plagioclase feldspar, and the formula then is resolvable into 2NaAlSi308 + CaAl2(Si04)2, or, in petro- graphic notation, Ab2Ani . A trisihcate and an orthosihcate have crys- tallized together in isomorphous mixture and simulated a metasilicate. 16 THE COKSTITUTION OF THE N-ATUKAL SILICATES. In the interpretation of any silicate, therefore, it becomes impor- tant to determine which acid it represents, and that is not often so easy to do as in the case just cited. With some minerals the evidence seems to be very clear, with others it may be misleading. It is only by careful study of a mineral in relation to other species, and with regard to the alterations or chemical reactions of which it is capable, that this phase of the constitutional problem can be solved. A noteworthy attempt in this direction has been made by Tschermak,^ who in a series of researches has studied the decomposition of silicates by hydrochloric acid and sought to identify the silicic acids so liber- ated. Tschermak has been followed by Himmelbauer, Baschieri, Silvia Hillebrand, and others. The validity of his method is sharply criticized by Miigge.^ Some of Tschermak's conclusions are in har- mony with the generally accepted views ; but others are at least siu*- prising. For example, anorthite and olivine are made to be basic metasilicates; albite is a salt of the acid HaSigOy; and garnet and its congeners are derived from trisilicic acid, H^SigOg. Even if it be admitted that the acids obtained by Tschermak are definite com- pounds, which has been seriously questioned by Miigge and Van Bemmelen, what evidence is there to show that they represent in any proper sense the original minerals? In the decomposition of the latter many reactions may occur, and the acids finally studied are not necessarily those which were first set free. It is safe to say that the validity of Tschermak's method is not well established. Siliceous jellies, obtained by decomposing alkaline silicates with acids or by the hydrolysis of SiCl4 or SiF4, have been studied by many chemists with varying results. Norton and Roth^ claim to have prepared a definite orthosilicic acid from SiF4, but the compound lost water steadily on exposure to air. Their memoir contains numerous references to the literature of the subject. A hasty glance over the entire field of the natural silicates will show, first, that many of them are most easily interpreted as orthosilicates, and, secondly, that by far the greater number are salts of aluminum. As regards both abundance and variety the aluminous silicates out- rank all the others, and from the wide range of composition which they exhibit we can obtain clues to their constitution. In other words, they furnish the most evidence, and some of it is of the highest import. Their relations to one another are oftentimes clear and unmistakable, so that the constitution of one salt is the key to that of a second, and thus generalization becomes possible. 1 K. Akad. Wiss. Wien Sitzungsber., vol. 112, Abth. 1, p. 355, 1903; idem, vol. 115, Abth. 1, p. 217, 1906; Zeitsehr. physikal. Chemie, vol. 53, p. 349, 1905; Centralbl. Mineralogie, 1908, p. 225; Zeitschr, anorg. Chemle, vol. 63, p. 230, 1909; idem, vol. 66, p. 199, 1910. 2 Centralbl. Mineralogie, 1908, pp. 129, 325; and Van Bemmelen, Zeitschr. anorg. Chemie, vol. 59, p. 225, 1908. See also Serra, R. accad. Lincei Atti, vol. 19, p. 202, 1910. 8 Am. Chem. Soc. Jour., vol. 19, p. 832, 1897. THE SILICIC ACIDS. 17 The constitution of these aluminous silicates has been studied from various points of view. Thsy may be regarded as ordinary silicates, m which the function of the aluminum is entirely basic, or as salts of complex acids containing aluminum as part of the negative radicle. That is, the existence of alumosilicic acids is assumed and the various minerals are classed as alumosilicates. This mode of interpretation has been much in vogue of recent years but is open to the objection that it IS purely hypothetical. Aluminum may so combine with silicic radicles as to form complex ions, but that it does so is quite improved. Some writers have argued that the aluminum of the silicates is unreplaceable by other basic radicles, and that the com- pounds in question are thus different in character from the silicates of dyad bases. When, however, andalusite or topaz alters to mus- covite, one-haK of the aluminum is replaced by the group KH2, and so the argument breaks down. The term "alumosilicate" may, nevertheless, be used as one of convenience, provided that we remem- ber its limitations. The complex ions may exist, but they should not be taken too strenuously for granted. The fact that alumina combines with silica is alone certain. That the alumosilicates are double salts, with all of the aluminum basis, is just as probable as the alternative hypothesis. A good summary of the diverse views rela- tive to the alumosilicates is given by Doelter.* Some of them will be considered later as regards their bearing on individual mineral species. A novel interpretation of the alumosilicates has recently been put forth by W. and D. Asch,^ and, as it has received considerable atten- tion, it may be briefly noticed here. The authors have developed what they call the "hexite-pentite theory," in which rings of silicon hydrates are represented as coalescing with similar aluminous rings. These rings, as the terminology indicates, may contain either six or five atoms of silicon or aluminum alternating with oxygen atoms, as in the ring formulae already considered here, and they have a super- ficial analogy with the benzene ring of organic chemistry. The silicon hexite acid, HigSigOig, is evidently a multiple of HjSiOg; the aluminum ring is HgAlgOig, equivalent to diaspora. By the coales- cence of four such hexite or pentite rings, either two silicic and two aluminous or three of one to one of the other, the authors develop formulae for 17 alumosilicic acids, having from 6 to 24 replaceable hydrogen atoms, and with molecular weights ranging between 873 and 1,693. From such complicated acids and their successive anhydrides the alumosilicates are derived. The acids themselves with few exceptions, have no representatives in nature and are purely hypo- » Handbuch der Mineralchemie, vol. 2, pp. 61-109, 1912. • Die Silicate in chemischer und technischer Beziehung, Berlin, 1911. 43633°— BuU. 588—14 2 18 THE CONSTITUTION OF THE NATURAL SILICATES. thetical. From them, with so many replacements possible an enor- mous number of salts can be predicated, and isomorphous mixtures, altered or impure minerals, and even bad analyses may easily be given place in the system. Whether, however, silicates of corre- sponding complexity could exist at the temperatures of even the coolest magma is most questionable. A generalization which does too much may be worse than no generalization at all. Even for such substances as glass the authors of the hexite-pentite theory write structural formulae. CHAPTER III. THE SILICATES OF ALUMINUM. GENERAL RELATIONS. A strictly logical investigation of the natural silicates might well begin with those of magmatic origin, for from them all others have been derived. Such a procedure, however, can not be adopted exclusively, for the various compounds, primary and secondary, are connected by so many interlocking relations that neither class should be considered alone. This point is well illustrated by the primary alumosilicates, some of whose derivatives are more sug- gestive than the original species. In this class the only simple sihcate which has so far been crystalUzed from a molten mixture of sihca and almnina is silUmanite, Al2Si05. More complex salts are easily generated from dry fusions, as, for example, nephehte, leucite, and anorthite, and each one is the progenitor of many secondary minerals. The problem of their structural correlation is the problem now in hand. If we consider first the orthosihcates of aluminum one general rela- tion is easily made apparent. By a general relation I do not mean a universal relation, for exceptions are possible, but only that many of the salts are connected by a simple regularity or even a law. To make this clear it is necessary to recognize the fact that aluminum is now properly regarded as a trivalent metal, its formerly supposed quadrivalency being no longer admissible. Formulae in which alimai- num appears as a tetrad are not vaHd, and ferric iron, which replaces aluminum, follows the same rule. This point has been established by the vapor density and cryoscopic investigations of recent years, and is sustained also by the position of aluminum in the periodic classifica- tion of the elements. This much admitted, the general relation of which I have spoken is as follows: Many, perhaps all, of the orthosihcates of aluminum are most simply represented as substitution derivatives of the normal salt AI4 (8104)3. To illustrate this rule for present purposes, the following examples will suffice : Aluminum orthosilicate . .Al4(Si04)3 Eucryptite Al3(Si04)3Li3 Nephelite Al3(Si04)3Na3 Kaliophilite Al3(Si04)3K3 Muscovite Al3(Si04)3HK3 Paragonite Al3(Si04)3NaH2 Kryptotile Al3(Si04)3H3 Andalusite Al3(Si04)3(A10)8 Topaz Al3(Si04)3(AlF2)8 19 20 THE CONSTITUTIOlir OF THE NATURAL SILICATES. These formulae express not only the composition of the minerals but also many facts concerning their relations, such as their association, their alteration one into another, and so on. Thus, topaz and anda- lusite are crystallographically akin ; both minerals, as well as others in the series, alter easily into muscovite, and these facts become intelhgible in the light of the formulae given. In the use of the formulae, however, one possible misconception must be avoided. They express a rela- tionship of constitution but do not imply that nature first generated the normal salt and then actually developed the other compounds from it. To emphasize this point an analogy may be drawn from organic chemistry. Alizarin, derived constitutionally from anthracene, was originally obtained from a glucoside contained in madder root. But nobody supposes that the madder plant took anthracene as a starting point from which to produce the dye. The constitutional or struc- tural derivation is one thing; the natural origin is quite another. Whether aluminum orthosilicate as such occurs in nature is still a matter of doubt. At all events its existence has not been definitely established. It is theoretically possible, and an artificial hydrate of the formula Al4(Si04)3.6H20 has been described by Pukall.^ As regards its ultimate constitution or chemical structure there is much uncertainty. Its formula can be written structurally in several ways ; as, for instance, with each aluminum atom linked with all three Si04 groups, or with only one atom so connected. In a sense this problem is analogous to that offered by the benzene ring, prism, or nucleus, a conception of which the utiHty is fully recognized, in spite of outstanding uncertainties. For practical purposes, that is, for the coordination of known facts, expressions like the following are sufficient: .Si04=Al ^i04=KH2 ^i04=CaH Al— Si04=Al Al— Si04=Al Al— SiO^^CaH \si04^Al \si04=Al \si04=Al These expressions indicate the observed replaceability of aluminum atoms by other atoms or groups and have no ulterior significance. So long as their limitations are kept in mind they are useful, but beyond this it would be unreasonable to go. With prolonged discussion and more evidence we may get a deeper insight into the nature of the fundamental molecule; at present, speculation concerning it would be premature. The relations expressed are clear, no matter what others may be revealed in the future. As a working hypothesis, the concep- tion of substitution from a normal salt may be a^pHed to many non- aluminous silicates, as in the magnesian series, the silicates of quadriv- alent metals, and so on. These points wiU be developed in subsequent 1 Deutsche chem. Gesell. Ber., vol. 43, p. 2098, 1910. THE SILICATES OF ALUMINUM. 21 chapters. For the present we need only to consider the alumosih- cates, group by group. THE NEPHELITE TYPE. If, now, we start out from the normal aluminum orthosilicate the first and simplest replacement possible is that of a single aluminum atom by three monads, giving a compound of the general formula AlgCSiOJgR'g. This formula represents several well-known minerals, and I propose to designate it the nephehte type. At first sight it seems to be reducible to the simpler expression R'AlSi04, but that expression, as will be seen later, does not indicate all the known rela- tions of the group. The first three representatives of this type are as follows: Eucryptite AlgCSiOOgLig Nephelite ..Al3(Si04)3Na3 Kaliophilite Al3(Si04)3K3 These species are all hexagonal, are nearly equal in density, and all gelatinize with hydrochloric acid. The second and typical member of the series has been made synthetically, and is then found to have the composition indicated hj the formula. The natural nephehte, however, has a composition which is more exactly represented by the complex formula R'3Al3Si9034, in which a little potassium appears among the components of E,', and the sihca is in excess of the amount required by theory. The potassium is doubtless due to an isomor- phous admixture of kahophilite, and the excess of sihca can be explained by the presence of a salt isomeric with albite and having the composi- tion Al3(Si308)3Na3. This replacement of Si04 by SigOg appears to be common among the sihcates, and its recognition clears up many discrepancies. In this case one molecule of the trisilicate commingled with fifteen of the ortho salt will produce the divergence from normal composition shown in the analyses of natural nephehte. This view of the constitution of nephehte has been adopted by Schaller,^ and is also favored by Bowen,^ who has studied the fusion diagram of the system silica, alumina, and soda. In that investiga- tion Bowen found that at about 1,550° nephehte is transformed into a trichnic isomer, a soda anorthite, to which the name carnegieite has been given. Foote and Bradley,^ however, have advanced a shghtly different interpretation of the anomalous composition of natural nephelite, ascribing the excessive sihca to ''solid solution. '^ These three researches represent the most recent and most conclusive work. Morozewicz * has explained the divergent analyses of nephehte by assuming a series of different nephehte molecules, derived from a 1 Washington Acad. Sci. Jour., Sept. 19, 1911. » Am. Jour. Sci., 4th ser., vol. 33, p. 439, 1912. 2 Am. Jour. Sci., 4th ser., vol. 33, pp. 49, 551, 1912. * Acad. Cracovie Bull., 1907, p. 958. 22 THE CONSTITUTION OF THE NATURAL SILICATES. number of alumosilicic acids. That interpretation seems to be no longer tenable. A normal mineral (Na;K)AlSi04 has been described by Zambonini ^ under the name of pseud on ephelite. It is evidently a mixture of nephelite and kaliophilite. The equivalency of these species is also clearly proved by an experiment of Lemberg,^ who heated nephelite (elseolite) with a solution of potassium silicate and obtained a product having the composition of kahophihte. Eucryptite and nephelite both alter with great ease into muscovite, a potassium salt of which paragonite is the sodium equivalent. Fur- thermore, C. and G. Friedel,^ by heating finely divided muscovite to 500° in a solution of alkali, obtained nephehte in crystals. From this evidence the formulae of muscovite and paragonite become directly related to those of the nephelite series, thus : Nephelite Al3(Si04)3Na3 Muscovite Al3(Si04)3KH2 Paragonite Al3(Si04)3NaH2 Physically, the two micas have no resemblance to nephehte, being different in form, sHghtly denser, and refractory toward acids. The relationship is purely one of chemical type, and is estabHshed by the fact of alteration from one into another. Kryptotile, according to UliHg,* is an end member of the mica group. If so, its formula becomes Al3(Si04)3H3. The clayhke mineral leverrierite has apparently the same composition and may be the same compound, and another clay, rectorite, is similar but with one additional molecule of water, which is lost at 110°. The compound may be regarded as an alumosihcic acid, with three replaceable hydro- gen atoms, although such an interpretation of it is not necessary. Through muscovite a connection is recognizable between the forego- ing species and the two minerals andalusite and topaz, whose simplest formulae, tripled, may be written as follows : Topaz Al3(Si04)3(AlF2)3 Andalusite Al3(Si04)3(A10)3 Here we encounter the evidently univalent atomic groups F — A1=0 and — Al/ \f both of wliich play an important part in various other minerals. The two species, topaz and andalusite, are closely alHed crystallograph- ically. They have sensibly identical molecular volumes, and both undergo alteration into muscovite mica.^ In topaz, as shown by the 1 Chem. Soc. Jour., vol. 98, pt. 2, p. 1078, 1910. Abst. from Accad. Napoli Rend., 1910. 2 Deutsche geol. Gesell. Zeitschr., vol. 37, p. 966, 1885. 8 Min. Soc. Bull., vol. 13, p. 183, 1890. * Zeitschr. Kryst. Min., vol. 47, p. 215, 1910. 6 For a good example of this alteration see Clarke and Diller, U. S. Geol. Survey Bull. 27, p. 9, 1886. THE SILICATES OF ALUMINUM. 23 investigations of Penfield and of Jannasch, hydroxyl commonly replaces a part of the fluorine, hydroxyl and fluorine being clearly isomorphous. The formula given is that of normal topaz, entirely free from alteration. Obviously the formula of muscovite is the key to all other formulae in this group of siHcates. Its minimum molecular weight is represented by the expression AlgKHaSigOia and to that the others must conform. The general formula Al3(Si04)3R'3 is the lowest possible, and the formula NaaAIjSiaOg, which is often assigned to nephelite, is too small. It may represent the isomeric carnegieite, which being stable at high temperatures is perhaps molecularly less condensed than nephelite. But of this there is no clear evidence. The tripled formulae are also sustained by an experiment of Silber,^ who heated an artificial nephe- lite silicate in a sealed tube with a solution of silver nitrate and replaced one- third of the sodium by silver: that is, one of the three sodium atoms seems to be differently combined from -the others. This sub- stitution can be expressed structurally in several ways, but its con- sideration must be deferred until later. Nephelite yields some zeoHtic derivatives, especially hydronephelite and natroUte, but their discussion belongs to another section of this chapter. To sum up, we have now eight definite species represented by the fundamental type Al3(Si04)3ll'3, the first substitution from the hypothetical normal orthosilicate of aluminum, and these compounds may be divided into three subtypes as follows: Nephelite. Muscovite. Topaz. ^iO^^Naa ^iO,=KH2 .SiO,={AlF,\ Al— SiO,=Al Al— SiO,=Al Al— SiO,=Al \siO,=Al \siO,=Al \siO,^Al symbols which clearly indicate the known chemical relations between the several minerals. In six of the eight examples the simplest pos- sible formulae have been tripled, for otherwise the relationships which exist could not be structurally shown. The correctness of this pro- cedure will appear stiU more definitely in the consideration of the groups which foUow. The species silUmanite is isomeric with andalusite, but the structural character of the isomerism is not clear. The two species have nearly the same molecular volumes, and presumably the same molecular weights, but a third isomer, kyanite, is much denser and therefore not so easily correlated with the others. It is commonly regarded as a basic metasiUcate, although that is not its only conceivable structure. It is easy to write constitutional formulae for all these minerals, but they would be of httle real significance except in so far as they repre- 1 Deutsche chem. GeseU. Ber., vol. 14, p. 941, 1881. 24 THE CONSTITUTION OF THE NATURAL SILICATES. setited possibilities. Sillimanite is the most stable compound of the three, and the only one which has been obtained magmatically. At high temperatures kyanite and andalusite are transformed into sillimanite.^ Structural formulae for andalusite and kyanite have been proposed by Zulkowski,^ but they are based upon the minimLum molecular weight of AlgSiOg and are therefore inadmissible. THE GARNET TYPE. By this title I propose to designate the second series of derivatives from the normal salt, Al4(Si04)3, in which two atoms of aluminum have been replaced. The general formula of the type obviously is Al2(Si04)3ll'6; and in this series bivalent elements or radicles fre- quently appear. In lagoriohte, an artificial soda garnet,^ Il'6 = Na6; in prehnite B.\ = 00,2^.2} ^^^ i^i normal garnet and epidote R'g = 3R''. There are, therefore, three subtypes to consider — one in which all the replacing atoms are univalent, one in which all are bivalent, and one intermediate between the other two. Under the first subtype two species may be definitely placed, namely, lagoriolite, Al2(Si04)3Na6, and zunyite, which is more com- plicated. In zunyite R'g is composed of the univalent radicles — ^A1=F2, — Al=Cl2, and — A1==(0H)2, but the species has been found in only one locaHty, and needs further study. At present, if we unite the chlorine in it with the fluorine, it may be provisionally represented by the expression .SiO,=(A102H2)2.AlF2 Al— Si04=(A102H2)2.AlF2 \si04=Al . This formula expresses the facts which are now available but is not conclusive. Its isometric character, however, helps to connect zunyite with the garnet and sodalite groups, as has been shown by Brogger.'^ In the second subtype, when R'q is partly composed of bivalent and partly of univalent atoms, two species may be placed, thus: Prehnite Al2(Si04)3Ca2H2 Biotite Al2(Si04)3Mg2HK Possibly the tetragonal sarcolite, which has the general formula of a garnet with the lime partly replaced by soda, may fall here also, but the analyses of this mineral are unsatisfactory, and its relations are 1 Vemadsky, Sex;. Min. Bull., vol. 12, p. 447, 1889; vol. 13, p. 256, 1890. 2 Monatsh. Chemie, vol. 21, p. 1086, 1900. 3 See Morozewlcz, Min. pet. Mitt., vol. 18, p. 147, 1898-99. The formula here given to lagoriolite is that of the ideally pure mineral. The actual product contains a notable admixture of the corresponding lime compound. * Zeitschr. Kryst. Min., vol. 18, p. 209, 1891. THE SILICATES OF ALUMINUM. 25 still uncertain. Biotite will be more fully considered in the section devoted to the mica group. Under the third subtype of this series we find the garnet group itself, together with epidote and several related species. The sodalite group is also akin to garnet and to the second subtype and will be considered in this connection a little later. The generic term garnet covers several species, all isometric and strictly isomorphous, in which magnesium, calcium, and ferrous iron replace one another, and chromium, aluminum, and ferric iron are also equivalent terms. Thus we have: Grossularite Al2(Si04)3Ca3 Pyrope Al2(Si04)3Mg3 Almandite Al2(Si04)3Fe^^3 Spessartite Al2(Si04)3Mn3 Andradite Fe2(Si04)3Ca3 Ouvarovite Cr2(Si04)3Ca3 To these may be added schorlomite, a garnet in which titanium occurs both as part of the acid, that is, with Ti04 replacing Si04, and also as Ti'^' among the triad bases, equivalent to aluminum. The monoclinic partschinite, isomeric with spessartite, also falls into this group. The several species of garnet occur in a great variety of isomorphous mixtures and some of them contain small quantities of alkalies, due to the presence of compounds like lagoriohte.^ In the epidote group several species appear, one, zoisite, being orthorhombic, whereas the others are monoclinic. These species are characterized by the bivalent group of atoms =A1 — OH or =Fe — OH, thus: Zoisite Al2(Si04)3Ca2(A10H) Epidote a rAl2(Si04)3Ca2(A10H) Epidote 6 lFe2(Si04)3Ca2(FeOH) Piedmontite (Al,Mn)2(Si04)3Ca2( AlOH) Allanite (Al,Ce,Fe)2(Si04)2(Ca,Fe)2(A10H) Hancockite Al2(Si04)3(Ca,Pb,Sr)2(FeOH) or, in general, as compared with garnet. Garnet R^^^2(Si04)3ll^^3 Epidote W,{SiO,),WyW^'OB) A chromium epidote, ''tawmawite," containing 11.16 per cent of Crfia has also been described. ^ The facts that garnet alters into epidote and that the two minerals are often associated give emphasis to the formulae. 1 For elaborate studies of the garnet group see Brogger and Backstrom, Zeitschr. Kryst. Min., vol. 18, p. 209, 1891; Weinschenk, idem, vol. 25, p. 365, 1S96; Uhlig, Naturh. Ver. preuss. Rheinl. u. Westfalens, Verb., vol. 67, pt. 2, p. 307, 1910; and Seebach, Centralbl. Mineralogie, p. 774, 1906. 2 See Bleeck, India Geol. Survey Records, vol. 36, p. 254, 1907-8. 26 THE CONSTITUTION OF THE NATURAL SILICATES. Although garnet as a rule is unattacked by acids, and epidote is only in part decomposable, both species are so broken up by strong ignition as to be readily acted upon by hydrochloric acid, with separa- tion of gelatinous silica. According to Doelter and Hussak,^ garnet yields upon fusion sometimes anorthite and an olivine; or meionite, augite, and olivine; or melilite and anorthite; and occasionally spinel. Epidote, says Doelter,^ yields lime-augite and anorthite, and prehnite behaves like garnet. These facts are interesting, but they give no direct information regarding chemical structure. By fusion the molecules of a silicate are broken down, and on cooling the. melt a complete rearrangement of the atoms may take place, although not necessarily. When calcium alumosilicates are fused they may, as in the case of anorthite, recrystallize unchanged, or they may solidify as compounds having little or no structural relations to their pro- genitor. When, however, silicates are broken down by mere calcina- tion and without fusion the reaction may be highly instructive. Examples of this kind will be noted later. In the four species sodalite, haiiynite, noselite, and lazurite we have a group of minerals which Brogger has classified as alkali gar- nets.^ Like garnet, they are all isometric, and they are characterized by the presence of the bivalent groups =A1 — CI, =A1 — SO4 — Na, and =A1 — S — S — S — Na. There are also artificial products, ultra- marines, in which the groups =A1 — S — S — Na and ^Al — S — Na appear. By adopting Brogger's formulae, which are preferable to those formerly proposed by myself,^ these species may be written as follows : Sodalite Al2(Si04)3Na4(AlCl) Haiiynite •- Al2(Si04)3Na2Ca(AlS04Na) Noselite Al2(Si04)3Na4(AlS04Na) Lazurite Al2(Si04)3Na4(AlS3Na) They fall therefore properly under the second subtype, but are con- sidered at this point on account of their analogies with garnet. The formulae just assigned to these minerals represent, of course, the ideally pure compounds, which rarely, if ever, occur in nature. The four species are all evidently derived from nephelite, with which sodalite is commonly associated, and their composition varies in the same manner as that of the parent mineral. Like nephelite, sodalite yields natrolite, hydronephelite, and muscovite by alteration. Furthermore, C. and G. Friedel,^ on heating powdered muscovite with soda solution and sodium chloride at a temperature of 500°, obtained 1 Allg. chem. Mineralogie, p. 182, 1890. 2 Idem, p. 183. » Brogger and Backstrom, Zeitschr. Kryst. Min., vol. 18, p. 209, 1891. * U. S. Geol. Survey Bull. 42, p. 38, 1887. 6 Soc. min. Bull., vol. 13, p. 183, 1890. THE SILICATES OF ALUMINUM. 27 sodalite artificially, although nephelite was probably first formed as an intermediary, and the two species were commingled in the product. The two hexagonal species, cancrinite and microsommite, are also, like sodalite, undoubtedly derivatives of nephelite, but their formulae are rather uncertain. At Litchfield, Maine, cancrinite often occurs in intimate mixture with nephelite (elseolite). A cancrinite described by Zambonini ^ corresponds very closely to a mixture of nephelite and the compound Al2(Si04)3Na3Ca(AlC03). In its purest varieties cancrinite approximates to the formula Al2(Si04)3Na4H(AlC03), in which a little soda is replaced by lime, and the univalent group — A1=C03 may be partly substituted by — Al= Si03. Mcrosommite, according to the published analyses, varies widely in composition, invariably containing potassium and having a notable proportion of chlorine and SO3 among its constituents. If, however, we assume in it the univalent radicles — Al=Cl2 and — A1=S04, its composition reduces easily to the form Al2(Si04)3 (NaK)3Ca(Al(S04Cl2)), like cancrinite, both species having then the composition of the general type Al2(Si04)3R'e. The theory as pro- posed, then, assumes univalent complex radicles for cancrinite and microsommite, and bivalent radicles for the sodalite group, thus : In cancrinite group. In sodalite group. — Al=-Cl2 =A1— CI — A1=S04 =A1— SO4— Na — A1=C03 =A1— S3— Na — Al=Si03 and the typical structures are as follows: Cancrinite. Sodalite. .Si04=Na2(AlC03) /Si04^^1__Cl Al^Si04=Na2H At— Si04=Na2 \si04=Al \si04=Al The best analyses of microsommite give very nearly ^i04=Na2-AlS04 ^i04=Na2.AlCl2 1 Al"Si04=NaCa + 2 Al— Si04=NaCa \si04=Al \si04=Al with nearly half the sodium replaced by potassium; the radicle AICO3 is also sometimes present. There are arguments both for and against these formulae and the pecuhar univalent and bivalent radicles assumed in them. The assumption of a group -Al=Si03, equivalent to and replacmg 1 Appendice alia mineralogia vesuviana, p. 35. 28 THE CONSTITUTION OF THE NATURAL SILICATES. — ^A1=C03, is clearly suggested by the experiments of Lemberg/ who by the action of sodium silicate solution upon elseoHte, obtained a compound which he designates as a cancrinite containing Na2Si03 in place of NagCOg. By similar reactions with sodium carbonate he pro- duced a substance having the composition of true cancrinite. Hence, whatever the ultimate molecular structure of cancrinite may be, we are amply justified in assuming in it the replaceability of CO3 by SiOg. These experiments fairly represent a large number of like kind which are due to Lemberg, and which are recorded in his papers. Some of these will be cited later, but a reference to the work of his colaborer,^ Thugutt, is in place at this point. Starting from a hydrated nepheUte, artificially prepared from kaolin, Thugutt succeeded in producing a large series of compounds analogous to sodahte, in which the original silicate had taken up, at moderately high degrees of heat and pressure, various other salts of sodium, such as the chlorate, selenate, formate, oxalate, and so on. These compounds, however, are all hydrated, and so differ from the natural minerals of the sodalite group, and they are regarded by Thugutt as formed by molecular union. Following Lemberg, he regards sodalite as a molecular compound of nephelite with sodium chloride, and taking his series of compounds throughout, he looks upon the sodium salts which have been added to the funda^ mental sihcate as equivalent in function to water of crystaUization. In favor of this view he cites many arguments, some of which are entitled to considerable weight. Thus, when sodahte is ignited NaCl is driven off, whereas if the chlorine were united with aluminum AICI3 should be expelled. Similarly, by the action of water alone, sodium chloride can be spht off from the sodahte molecule, thus indi- cating a looser form of union than the proposed structural formulae show. But what is molecular union ? To this question there is no satisfac- tory answer, and even in the case of water of crystallization the term is only a confession of ignorance. Unless we assume the existence of two kinds of chemical union, it means merely that the structural link- ing is unknown, and that the problem is laid on one side, conveniently labeled for future reference. The constitutional formulae here adopted for sodalite and cancrinite are intended to give a provisional solution of the problem in their particular cases and to express the genetic relationships with nephehte on the one hand and the crystallographic analogy with garnet on the other. The objections to them raised by Thugutt are serious but not absolutely conclusive. When sodium chloride is spht off from sodalite the mechanism of the reaction is 1 Deutsche geol. Gesell. Zeitschr., 1885, p. 962. 2 Mineralchemische Studien, Dorpat, 1891. See also Thugutt on cancrinite, Neues Jahrb., 1911, p. 25. Zambonini (Contributo alio studio del silicati idrati, Napoli, 1908) regards cancrinite as a mixtiu'e of silicates, with all the water extraneous— that is, not essential to the molecules. On the sodalite group see also Silvia Hillebrand, K. Akad, Wiss, Wien Sitzungsber., vol. 119, p. 775, 1910. THE SILICATES OF ALUMINUM. 29 quite unknown, and the relative affinities in the molecule are quite unstudied. Until these are understood the objections raised by Lemberg, Thugutt, and others are not fatal. Furthermore, the pres- ence of a group =A1 — CI does not imply, as Thugutt supposes, the splitting off of AICI3 by heat. To effect such a decomposition three molecules of sodaUte would have to be broken up, and there is no probability that such a disintegration would occur. At all events the formulae proposed fulfill a definite purpose, even though they are not finally established. They express known relations but not necessarily all the relations which the future may reveal. The facts that the sodalite-cancrinite minerals are derivable from nephelite and that nephelite is again derivable from them are unquestionable. The question of the molecular structure of a typical garnet, Al2(Si04)3Ca3, remains to be considered.^ If it is regarded as a derivative of the normal salt Al4(Si04)3 it may be written in at least two ways, thus: 1. 2. .SiO,^l SiO,=Ca Ai:lsio,^r^^ ^\siO,=Ca \siO,^l ^SsiO,=Ca That is, isomerism is possible, and of the two species, partschinite and spessartite, one may belong to one type and the other to the other In the first expression there is still a replaceable atom of aluminum, but in the second expression none; in the first at least one calcium atom must link two SiO^ groups, whereas in the other no such linkage occurs; and these facts may be connected with others. For example, garnet alters into mica, and the mica group, as will be seen later, con- tains members in which the third aluminum atom is replaced. This points at once to the first type of formula as preferable, and the alterability of garnet into epidote brings the latter mineral into the same category. Zunyite and sodalite, being isometric, should also foUow garnet, but derivatives of the second type are theoretically possible and may exist. Even under the first type alone isomerism is conceivable, and the orthorhombic zoisite may be contrasted with the monoclinic lime epidote as f oUows : Si04=Ca SiO,=Al-OH / >A10— H / >Ca Al— Si04==Ca Al— Si04=€a \si04=Al \siO,=Al » Tschermak (K. Akad. Wiss. Wien, Sitzungsber ., vol. US, p. 217, 1906) regards garnet, epidote, zoisite, and prehnite as salts of the acid KShOs, but his formula do not well show the relations of these minerals to other species. 30 THE CONSTITUTION OP THE NATUKAL SILICATES. even though we can not assign either species to either formula defi- nitely. My obj ect here is merely to show that the formulas have prop- erties by virtue of which they are able to express known differences. Additional evidence for the formula assigned to garnet is supplied by the composition of vesuvianite, which is most simply represented by the coalescence of two garnet molecules with partial hydration, thus: Garnet. Vesuvianite. SiO = 1 SiO,= 1 / Cag / Caa Al— SiO,= Al— Si04= \siO,=Al \siO,=Al— OH Ca .Si04=Al— OH Al— 810^= ] The formula agrees well with many analyses of vesuvianite, but actually, as with other species, its composition varies. About one- seventh of the calcium is commonly replaced by magnesium, and in some varieties boron, presumably in the group =B — OH, replaces in part the corresponding aluminum radicle.^ Fluorine is also often present in small amount as the equivalent of hydroxyl. In short, a variety of isomorphous replacements or comminglings are possible without affecting the essential structure as shown by the formula. Such replacements are too well known to need detailed discussion here. The true molecular weights of silicates, however, are unknown, and it is therefore conceivable that the formulae of garnet and epidote should be doubled. These minerals and vesuvianite have nearly the same specific gravities, 3.3 to 3.5, for the purely calcic varieties. The specific gravities in the cancrinite-sodalite group, on the other hand, are about a unit lower, a fact which favors the simpler, less- condensed molecular structure. On doubling the formulae of garnet and epidote the following comparison with vesuvianite and anorthite is interesting: Garnet. Anorthite. .Si04=Ca3=Si04v .Si04=Ca3=Si04v Al— Si04=Ca3=Si04— Al Al— SiO^^Al^^SiO^— Al \si0,^Al2=Si0/ \si04=Al2=Sio/ 1 On boron in vesuvianite see Wherry and Chapin, Am. Chem. Soc. Jour., vol. 30, p. 1684, 1908. Wein- garten (Centralbl, Mineralogie, 1902, p. 726) represent"? the mineral by the formula A10H=Si207=Ca2. In the former edition of this memoir, U. S. Geol. Survey Bull. 125, vesuvianite was given the formula Al2(Si04)6R6- AlOH. See also U. S. Geol. Survey Bull. 262, p. 72, 1906, for variations in the present formula. THE SILICATES OF ALUMINUM. 31 Epidote. Vcsuvianite. ^i04=Ca3=SiO,s^ ^SiO,=Ca3=SiO,v Al -SiO,=Al2=SiO,— Al Al-SiO,=Ca3=SiO,— Al \siO,- Ca-Sio/ \siO,-Ca -Sio/ AlOH AlOH AlOH AlOH In these new formulae the essential character of the former ex- pressions is unchanged, but the presumably greater condensation is indicated, with the derivation from two molecules of aluminum orthosilicate instead of from one. They are also sustained by the facts that garnet, epidote, vesuvianite, scapolites, and in some locaU- ties anorthite often occur in hmestones as products of contact meta- morphism; that vesuvianite alters into garnet, garnet into epidote and scapolite, and that all four minerals alter into micas and the magnesian varieties into chlorites also. The species are connected constitutionally and genetically, the analogies connecting them are remarkably suggestive and complete, and the formulae here proposed render those analogies intelligible. In the Swedish '^mangan- idocrase" a salt occurs which is doubtless the vesuvianite eqidvalent of spessartite, but the compound in a pure state is unknown. Kyanite, an isomer of andalusite, but of much higher specific gravity, may perhaps be represented as a basic member of the garnet series, although it is morphologically very different. The formula .SiO,=Al=Si04. Al— SiO,=Al2=SiO — Al \sio, sio/ (aIo)3 (A10)3 expresses this relation and also its comparative instabihty. Both kyanite and andalusite, at very high temperatures, are transformed, with disengagement of heat, into a third isomer, siUimanite,^ which probably has the simplest formula of the three. This, however, is so purely hypothetical that it would be useless to discuss the several species further. Two more species, meHlite and gehlenite, which, like vesuvianite, are tetragonal, may perhaps be best considered now. Both species are very variable in composition, and neither seems to admit of one definite formulation. They appear to be mixed silicates, like the intermediate plagioclase feldspars and scapoHtes, but the end inem- bers of each series are difficult to identify. An artificial sihcate recently obtained by Shepherd an d Rankin,^ in the Geophysical 1 See Vernadsky, Soc. Min. Bull., vol. 12, p. 447, 1889; vol. 13, p. 256, 1890. 2 Private communication. 32 THE CONSTITUTION OF THE NATURAL SILICATES. Laboratory of the Carnegie Institution, is probably one end of the gehlenite series and has the formula Al2Ca2Si07, which is that of a basic metasilicate. Structurally it can be written ^^^^XAlO^Ca, in which the basic radicle — ^Al<^ ^Ca is analogous to the more familiar — ^A1=(0H)2. The other end of the series is probably a silicate of the vesuvianite type, .SiO=Ca3=Si04v Al— SiO,=Ca3=SiO— Al \si04=Ca3=Si6/ in which the three replaceable aluminum atoms of the normal ortho- silicate are substituted by calcium. If this supposition is correct we have the following complete series: Ale(Si04)6Ca3, anorthite. . Al4(Si04)6Ca6, garnet. Al3(Si04)6Ca9, in gehlenite. In the Mexican gehlenite analyzed by Wright ^ there is an approxima- tion to Shepherd and Rankin's silicate, namely, eight molecules of that compound commingled with one of the other. The comparison is as follows: Found. Reduced. Calculated. SiOa 26.33 .03 27.82 L43 .50 .01 2.44 39.55 .21 .10 1.85 1 26. 69 J 1 29. 12 44. 19 26 51 TiOg ALO, 28 98 Fe'a FeO.. MnO MgO CaO .. 44.51 NajO . . K2O HoO 100. 27 100.00 100. 00 The second column is recalculated to 100 per cent after uniting iron with alumina and recomputing the other bases to their equivalent in 1 Wright, F. E., Am. Jour. Sci., 4th ser., vol. 26, p. 545, THE SILICATES OF ALUMINUM. 33 lime. Other gehlenite analyses reduce equally well but with much larger proportions of the orthosihcate compound. The Orawitza gehlenite, for example, is very nearly 2 Al^da^SiO^ + 1 Al^da^Si A4. This commingUng of an orthosihcate with a very basic metasiUcate is not easy to explain, but it seems to fit the actual evidence. It is furthermore sustained by an observation of Cathrein,^ who has reported pseudomorphs after gehlenite of fassaite, a metasiUcate, and grossular garnet. The formula commonly assigned to geh- lenite, Ca3Al2Si20io, is inadmissible. In gehlenite the oxygen is always in excess of the orthosihcate ratio, but in melihte the reverse is generally true. The Vesuvian melilites agree nearly with Groth's formula, Rg" R2''' Si fi^^, but the mineral from other localities exhibits quite different ratios. An artificial ''mehlite" obtained by Bodlander ^ from Portland cement is very nearly ^iO^=Ca3=Si04v Al— SiO,=Ca3 =SiO — Al \si04=Mg3^SiO,/ as the following comparison shows : Found. Reduced. Calculated. SiOj -- - 37.96 9.46 2.93 12.77 34.75 1.53 .64 38.63 1 1L51 12.99 1 36. 87 39.22 ALO, ' 11.11 FegOg MgO .... 13.07 CaO 36.60 K2O NajO 100. 04 100. 00 100. 00 Other melilites seem to be mixtures of this type of compound with the corresponding trisihcates — that is, with SigOg in place of Si04, but the evidence is not conclusive. Such mixtures are found in the feldspar, scapohte, and mica groups and are well known. Just as the calcic anorthite crystaUizes with the sodic albite so probably in melilite two compounds, one calcic or magnesian, the other alkahne, replace each other isomorphously. Gehlenite and mehhte, how- 1 Min. pet. Mitt., vol. 8, p. 408, 1886-87. 43633°— Bull. 588—14 3 2 Neues Jahrb., 1892, vol. 1, p. 53. 34 THE CONSTITUTION OF THE NATURAL SILICATES. ever, seem to have one end compound in common, and that com- pound belongs in the anorthite-garnet series as already shown. Its synthesis in the pure state is yet to be effected. Whether the formulae here proposed are true or not they are useful for purposes of correlation. They are, moreover, emphasized by an experiment made by Lemberg,^ who has shown that gehlenite, when heated to 200° with a solution of potassium carbonate, gives calcium carbonate and a product having the composition of a potash mica, whereas similar treatment with sodium carbonate converts the mineral into cancrinite. Gehlenite, garnet, cancrinite, and musco- vite are therefore related to one another, and this fact is expressed by the formulae proposed. Furthermore, at Orawitza, in the Banat, gehlenite is found in rolled pebbles containing grains of vesuvianite, a fact which indicates a common genesis for both minerals. An alternative interpretation of the relations between meUhte and gehlenite is offered by the hitherto generally accepted theory of Vogt,2 who regards both species as varying mixtures of two silicates, one the *' gehlenite silicate," Rj^'I^s^SiaOio, and the other akerman- ite, R^'^SigOiQ. The last compound is found in slags and has recently been identified by Zambonini among the minerals of Vesu- vius. In it B,/^ is principally Ca, but with a notable proportion of Mg also. A purely calcic siHcate of that type has not yet been found. Vogt's theory has been seriously questioned by Bodlander and Zambonini. The new interpretation now offered seems to be more general. The little-known mineral arctohte is possibly another member of this group, with affinities toward prehnite. Its composition is fairly expressed by the formula Al^CSiOJeCaMgH^ which is that of prehnite with CaMg in place of Csl^. The integrity of this species, however, is not yet fully established. THE FELDSPARS AND SCAPOLITES. Although orthosilicic and trisilicic acids are technically distinct and from the chemist's point of view should be studied separately, their salts containing aluminum occur in such a variety of mixtures that in sev- eral groups of minerals the two acids must be considered as mutually equivalent and their compounds discussed together. Two such groups, closely allied, are the feldspars and the scapoHtes. For each of these groups the theory developed by Tschermak has met with general acceptance. In the case of the feldspars, Tschermak 1 Deutsche geol. Gesell. Zeitsch., p. 237, 1892. 2 Neues Jahrb., 1892, vol. 2, p. 73. THE SILICATES OF ALUMINUM. 35 was undoubtedly anticipated in great part by Hunt, Waltershausen, and others, but to him full recognition is due. More recently it has been put upon a thorough quantitative basis by the synthetic experi- ments of Day and AUen.^ According to this theory the trichnic plagioclase feldspars consist of albite, AlNaSigOg, and anorthite, CaAlaSiPs, wliich, commingled in various proportions, give the inter- mediate oligoclase, labradorite, andesite, and so on. There are also the trichnic microcline and its monochnic equivalent, orthoclase, both represented by the formula AlKSigOg, the monochnic barbierite, isomeric with albite, and the recently described carnegieite, or soda anortliite, already mentioned as an isomer of nephelme. The mineral celsian, BaAlaSiaOg, is empirically the barium equivalent of anorthite, but it is monochnic and isomorphous with orthoclase.^ Hyalophane and other barium feldspars are mixtures of orthoclase and celsian. The exact nature of the isomeric equivalencies among the feldspars is not clear; they may be due to the structure of the salts inde- pendently of the acids which they represent, or to isomerisms among the acids themselves. The latter possibihty was discussed in the section on the silicic acids and seems to be the more probable, at leaj^t so far as the trisilicates are concerned, but for present purposes the problem may be left outstanding. In the discussion later of the spe- cies eudidymite and epididymite the question of isomeric trisihcates wiU be considered. For the scapolite series Tschermak has elaborated a theory wliich is closely parallel to that of the feldspars. These tetragonal minerals are shown to be most easily interpretable as mixtures of two end compounds, meionite, AlgCa4Si6025, and mariahte, Al3Na4Si9024Cl. Neither end compound has yet been found in nature quite free from the other, but the variations in composition, in optical character, etc., are all accounted for, and the theory, so far as it goes, is satisfactory. I have tentatively examined some possible alternative hypotheses, and none of them fulfills all necessary conditions so well as tliis scheme of Tschermak' s. Upon studying the feldspars and scapoHtes more closely, certain analogies appear other than those indicated by the parallehsm of the two series. Both groups of minerals are easily alterable, and both yield kaolui as a final product of the change. Furthermore, both alter to muscovite, or to pinite, which is only an unpure pseudomorphous mica, and kaolin crystallographically has close relations with the mica family. Feldspars, scapohtes, muscovite, and kaolin are therefore presumably connected, and the structural formulsB of the minerals should render the relationship apparent. 1 Am. Jour. Sci., 4th ser., vol. 19, p. 93, 1905. 2 Strandmark, Zeitschr. Kryst. Min., vol. 43, p. 89, 1907. 36 THE CONSTITUTION OF THE NATURAL SILICATES. The typical empirical formulae to be considered are now as follows : Albite AlNaSigOg Anorthite AlgCaSigOg Meionite Al6Ca4Si6025 Marialite Al3Na4Si9024Cl Muscovite AlgKHaSiaOia Kaolin Al2H4Si209 For muscovite the constitution has already been indicated, and this clue, together with the general hypothesis of derivation from normal salts, enables us to correlate all six of the formulae given. To do this it is necessary to triple the formulae of albite and anorthite, and we have the following expressions: Albite. Anorthite. ySigO^Nag .Si04^Al2=Si04v Al— SigOs^Al Al— SiO,=Al2=SiO— Al ^SigOs^Al \siO,=Ca3=Sio/ Marialite. Meionite. Si303^Na, /SiO,=Al,=SiO,^ / )>A1— CI Al— SiO,=Al2=SiO,— Al Al-Si^OsdNa, \sio_Ca,=Sio/ \Si3O3SAl I J Ca O da Muscovite. Kaolinite. ^iO,=KH2 /OH Al-Si04=Al Al— SiO,=H3 \siO,=Al ^SiO^^Al On this basis mariahte becomes the trisilicate equivalent of soda- lite, although the two species are quite unlike in form. Anorthite is the calcium salt corresponding to nephelite, which is also alterable into kaolin. Again, garnets are known to alter into feldspars and scapolite, and, according to Brauns, in the alteration of diabase, prehnite and epidote are sometimes derived from anorthite. . These species, therefore, are all connected by numerous cross relations, all emphasizing one another and pointing to a community of molecular type. So far the formulae are highly suggestive, but as yet they do not indicate the mechanism of the reaction by which a trisilicate feldspar breaks down into kaolin, and they need development in that direction. Tschermak,* from the composition of the siUcic acids derivable from the several minerals, infers that albite is a salt of the acid HsSigOy, and that anorthite is a basic metasiHcate. Himcmel- 1 Zeitschr. physikal, Chemie, vol. 53, p. 349, 1905. THE SILICATES OF ALUMINUM. 37 bauer/ by the same method, makes meionite a metasilicate and mariahte a derivative of a new acid, HioSigOjg. Such formulae fail to express the known relations of the minerals at all clearly. Closely aUied to the feldspars in its petrographic relations is the isometric mineral leucite, AlKSiaOg. Empirically it seems to be a metasihcate and is commonly so regarded, but it may easily be con- ceived as a mixed salt, containing orthosilicate and trisilicate mole- cules. By alteration it yields orthoclase, nephehte, muscovite, and kaolin, and the pseudoleucite of Magnet Cove has been shown by J. F. Wilhams to consist of orthoclase and elseolite intimately com- mingled.^ This case probably represents the typical breaking up of leucite, the formation of kaoHn or of muscovite in other instances being due to secondary reactions. On the other hand, C. and G. FriedeP have obtained leucite synthetically from muscovite as a starting point, orthoclase and nephehte being produced at the same time, and Lemberg,* in his experiments, has transformed leucite into sanidine, anorthite, and microsommite and also into andesine. In a later paper ^ Lemberg describes^the action upon various sihcates of the salt NagSiOg.SHjO, at 200° under pressure, kaolin, albite, elseolite, leucite, and analcite all yielding a siHcate-cancrinite contain- ing SiOg in place of CO3. These facts connect the several species together, but to their explanation the empirical expression AIKSijOg gives no clue. A formula for leucite, to be satisfactory, must be a multiple of this, and several such multiples fulfill the conditions of the problem. The isometric form of leucite suggests at once a relation with the sodahte group, and this can be indicated by the quadrupled formula Al4K4Si8024. We then have, as a distinct possibihty, the following series of molecules, including for comparison the tetragonal mariahte. Sodalite. Marialite. Leucite. / >A1-C1 / >A1-C1 / >Al-SiO,=Al Al-SiO^^Na^ Al— SisOs^Na^ Al— SigOs^K^ \siO,=Al \si308^Al \si30«=Al On this basis leucite is clearly reduced to the uniform type of the minerals to which it is apparently related, and also of those into which it alters, but the formula proposed can not be regarded as final. It is offered here only as a first approximation toward answering the questions which are suggested and is therefore subject to modifica- tion in the future. 1 Sitzungsber. K. Akad. WLss. Wien., vol. 119, p. 115, 1910. « Arkansas Geol. Survey Aim. Kept., 1890, vol. 2, pp. 267 et seq. 3 Soc. min. Bull., vol. 13, p. 134, 1890. < Deutsche geol. Gesell. Zeitschr., 1876, pp. 611-615. 6 Idem, 1885, pp. 961, 962. 38 THE CONSTITUTION OF THE NATURAL SILICATES. Closely allied to leucite is another isometric mineral, analcite, which empirically has the composition AlNaSi206.H20. This species alters, like leucite, into feldspar, and an alteration into prehnite is also recorded. Furthermore, Lemberg has shown, in the papers already cited, that leucite, by the action of soda solutions, may be transformed into analcite, and that analcite, by similar treatment with potash, yields leucite again. With these facts in view, analcite may be written SiO^ =Na2 / >A1— SiO,=Al Al— SigOg^Na^ +4H2O ^SigOs^Al exactly equivalent to leucite. That the water is entirely nonconstitu- tional has been shown by the experiments of G. Friedel,^ who found that it could be expelled continuously without change in the crystal nucleus. The dehydrated mineral, moreover, could take up water again, or instead of water various vapors and gases. These sub- stances seem to be occluded, much as water is held in a sponge, only the ratio between the water and the silicate is definitely molecular. This phenomenon seems to be peculiarly characteristic of the zeoUtes, which will be considered in the next section of this bulletin. The close relationship between leucite and analcite is shown not only by the work of Lemberg but also by investigations carried out in the laboratory of the Geological Survey by Clarke and Steiger.^ When these minerals are heated with dry ammonium chloride in sealed tubes to 350° C, double decomposition takes place and ammonium is substituted for the fixed alkaUes. In each case the new compound has the empirical formula NH4AlSi206; that is, an ammonium analcite or leucite precisely equivalent to the original sili- cates is formed. A volatile base has replaced the fixed bases, and the substance so formed splits up on ignition in such a way as to shed light on its constitution. If, now, ammonium leucite is a true metasilicate, a salt of the acid HjSiOg, it should break up, when ignited, in accordance with the following equation: 2 NH,A1 (8103)2 = Al2 (8103)3 + 2 NH3 + H20 + 8i02 and one-fourth of the silica ought to be set free, measurable by extraction with sodium carbonate solution. No such liberation of silica occurs, and we may therefore conclude that analcite and leucite are not metasilicates, but more probably mixed orthosiUcates and trisilicates, as shown in the constitutional formulae assigned to them here. The evidence against their being metasilicates at least seems to be conclusive. 1 Soc, min. Bull., vol. 19, pp. 94, 363, 1896. a U. S. Geol. Survey Bull. 207, 1902. THE SILICATES OF ALUMINUM. 39 One other isometric mineral, pollucite, may perhaps be considered here. Its empirical formula, as established by the analyses of Wells ^ and Foote,2 is that of a metasihcate, Hfis.Al^Sfi^, which, however, may also be written as a basic trisilicate, thus: ^i30«=Cs,H Al— Si308=Cs2H \si30,=(A10)3 If further investigations should show that the water of pollucite is not an essential part of the molecule, its empirical formula would be very close to AlCsSigOg, and the mineral would become the caesium equivalent of leucite and analcite. Pollucite, however, differs from those minerals in one notable respect, namely, on heating with dry ammonium chloride only one-third of its caesium is replaced by ammonium instead of the entire amount. This observation needs to be checked by experiments on pollucite from new localities before any safe conclusions can be drawn from it. That pollucite is a true metasilicate is very doubtful. Although kaohn miner alogically is not a member of the feldspar group, it is properly discussible here as a derivative. The formula assigned to it in the foregoing pages is not unimpeachable, but it sug- gests its relations to the feldspars and micas and also represents the fact that the water in it is wholly constitutional. In fact the mineral is stable far above the ordinary temperatures of dehydration, so that the water can he regarded only as an essential part of the molecule. In addition to the formula proposed for kaohn the following expres- sions are possible without assumption of any higher molecular weight: (1) SiA.H4(AlO)3 (2) SiA-H2(A10H)2 (Brauns's) (3) SiA.CAlHp^)^ (Groth's) Si03— AIHA (4) H-0-Al.«« or,^ 3 See Rinne (Neues Jahrb., 1896, vol. 1, p. 139; idem. 1897, vol. 1, p. 41), on heulandite and stflbite^d Grandjean (Soc. min. Bull., vol. 33, p. 5, 1910) on the replacement of zeolitic ^^f/'^ ^t^^^^^f ^^^"fj'P^": Zambonini (Contributo alio studio dei silicati idrati, 1908) has also done much to show that zeolitic water is essentially absorptive. 4 U. S. Geol. Survey Bull. 125, 1895. 42 THE CONSTITUTION OF THE NATURAL SILICATES. work, as just cited, renders a- complete revision of the subject neces- sary. Now, regarding water as not belonging to the true silicate molecules, we may discuss the zeolites with reference to their genetic relations, beginning with the obviously related starting points, the formulae of nephelite and albite. These minerals, as we have already seen, are compounds of the same type, one an orthosilicate, the other a trisilicate, Al3(Si04)3Na3 and Al3(Si308)3Na3. From these species hydronephelite, natrolite, scolecite, mesolite, analcite, and faujasite appear to be derived, either directly or indirectly, as the formulae to be proposed clearly show. In nearly every case the simplest empirical formula is discarded as not fairly representing the known relations between the minerals, and to only one of the above-named species does the rule not apply. That species is hydronephehte,^ an obvious derivative of its original type and of its more direct parent, sodalite. Its formula is Al3(Si04)3Na2H.3H20. For natrolite alternative formulae have been proposed. One, Al2(Si04)5Na2H4, regards the water of the mineral as constitutional. But when natrolite is heated with dry ammonium chloride in a sealed tube it is transformed into the compound Al2(NH4)2Si30io, an ortho tri- silicate. The simplest formula for natrolite, then, is Al2Na2Si30io. 2H2O, which is not obviously related to nephelite or to its near relatives among the zeolites. Nephelite can be written as a basic orthotri- silicate, but that involves more difficulties than the one it might seek to avoid. Natrolite, then, with other species, is best repre- sented in a less immediately obvious manner by doubling its formula and bringing it, into line with its congeners, especially with sodalite, scolecite, mesolite, and edingtonite. MesoHte, however, is only a crystalline mixture of natrolite and scolecite and needs little consid- eration. Edingtonite is rather doubtful, but Lemberg,^ by the action of barium chloride solution upon natrolite, obtained a silicate which appears to be that mineral. This species is included here on account of its chemical analogy to scolecite. So much premised, the formulae now offered are as follows, beginning with the anhydrous type species: Nephelite. Albite. Sodalite. Si04=Na3 SigOg^Nag Si04=Na2 / / / >A1-C1 Al— SiO^sVl Al— SigOgSVl Al— Si04^Na2 \si04^Al \si3O3SU \si04^Al 1 Thugutt (Neues Jahrb., 1910, vol. 1, p. 25) regards hydronephelite as a mixture of iiatrolite, gibbsite, and diaspore; but the miaeral analyzed by me was purified by means of Thoulct solution, was homogeneous under the microscope, and apparently hexagonal as judged from its optical behavior. 2 Deutsche geol. Gesell. Zeitschr., vol. 28, p. 553, 1876. THE SILICATES OF ALUMINUM. 43 Hydronephelite. SiO.^Na^H Al— SiO.^Al 3H2O Natrolite. / >Al— SigOs^Al Al— SiO^^Na^ ^SiO^^Al 4H2O Edingtonite. SiO^^Ba / >Al-Si30s=Al Al— SiO^^Ba \siO,=Al 6H2O Analdte. SiO,=Na3 , / >A1— SiO,=Al 4H,0 Scoledte. SiO^^Ca / >A1— Si303=Al \siO,^Al 6H2O Faujasite. SigOg^Nag / >A1— SiO,sVl Al— SigOs^^Ca 19HoO The formula assigned to faujasite is quite unlike that usually given, but it best fits Damour's analysis, as the subjoined comparison shows: Found. Calculated. SiOa 46.12 16.81 4.79 5.09 27.02 47 46 ALO. 16 14 CaO 4 43 NaoO 4.91 HoO 27.06 99.83 100.00 The isometric character of faujasite relates it to analcite and leucite, but its immediate derivation was probably from albite. A group of monoclinic zeolites, closely related in structure to the foregoing species, is that formed by wellsite, phillipsite, harmotome, and stilbite, to which Pratt and Foote ^ assign the following general- ized formulae: Wellsite : RAlaSigOio.SHaO Phillipsite RAl,Si40i2-4H20 Harmotome - RAloSisOn-SHjO Stilbite RAlgSieOio.eHaO 1 Am. Jour. Sci., 4th ser., vol. 3, p. 443, 1897. 44 THE CONSTITUTION OF THE NATURAL SILICATES. These formulae make a beautifully regular series, but unfortunately they do not represent the wide variations in composition exhibited by some of the species. Harmotome, a barium salt, and stilbite, a calcium compound, are fairly constant, except for variable replace- ments of the dyad radicle by sodium or potassium. Wellsite rests on a single analysis, in which calcium, barium, strontium, potassium, and sodium appear. In phillipsite R is principally calcium, but with varjdng replacements by potassium and sodium, and the ratio of silicon to oxygen is by no means constant. A phillipsite reported by Zambonini ^ is very near R"Al2Si30io.4H20, with alkalies partly replacing lime. Other phillipsites are much richer in silica and approach a trisilicate in composition. In fact, all phillipsites appear to be mixtures of orthosilicates and trisilicates, ranging between SSigOg : lSi04, and iSigOg : 3Si04. Such a range and even a greater one is only to be expected when we remember that many zeolites are derivatives of plagioclase feldspars. The zeolites vary as the feld- spars vary between end products, which may or may not be definitely known. Stilbite, for instance, represents a hydrated calcium albite or trisilicic anorthite, which in the anhydrous condition is yet to be discovered. In this series of silicates, then, we have the plagioclase variation in the ratio Si : O, whereas the ratios R : Al and the degree of hydration for each species are constant or nearly so. The formulae being qualified by recognizing the common replacements of lime or barium by alkalies, the four ''species" may be assigned the following general expressions, that for phillipsite representing a fair average between its extreme variations : Wellsite. Phillipsite. SiO^^R Si04=Ca / >A1— Si308=Al / >A1— SigOg^Al Al— SiO^^R Al— SiO^MZJa ^SiO^^Al \si308=Al 6H2O 8H2O. Harmotome. Stilbite. SisOg^Ba Sifi,=Csi / >A1— SiO,=Al / >A1— SigOs^Ai Al— SisOs^Ba Al— Si308=Ca \si308=Al ^SisOg^Al IOH2O 12H2O In phillipsite as much as half the calcium may be replaced by potas- sium. That replacement is characteristic of the species. 1 Contributo alio studio dei silicati idrati, p. 114, 1908. THE SILICATES OF ALUMINUM. 45 The hydration of these silicates increases in a very regular and remarkable manner, and proportionally to the number of silicon atoms in the molecule. For every sihcon atom one molecule of water is retained. This rule holds true for the typical formulae, but if incU- vidual analyses are studied considerable variations will be found. Phillipsite nearly always contains an excess of water. Moreover, the low molecular weight of water is the cause of apparent irregularities when formulae are deduced from analytical data. A small error in the determination of water is exaggerated in the computed ratios. The rule is not universal, but it certainly apphes to the wellsite-stilbite series. Stilbite, at one end, is entirely trisihcate; at the other end there should be a pure orthosihcate R2^l4(Si04)4.4H20, but no such zeoUte is known. The orthorhombic thomsonite approaches the required composition but not quite closely enough. Empirically either lawsonite or its isomer hibschite would complete the series, but their hydration appears to be constitutional, and crystallographicaUy they belong elsewhere. Among the plagioclase zeolites, if such a term is admissible, there are two, essentially orthosihcates, which may be regarded as hydra ted anorthite. These species are thomsonite and gismondite, and they may be represented as anorthite plus water, using the tripled formula for anorthite as developed in the preceding section of this work. It is better perhaps to treat them as less condensed molecularly than anorthite, because of the loose crystalline structure which permits the retention of zeolitic water. On this basis their formulae fall in line with those of the other zeoUtes, as follows: Thomsonite. Gismondite.^ SiO,=Ca SiO^^Ca / Vl— SiO.^Al / \a1— SiO,=Al Al-SiO,^Ca Al-SiO,^Ca \siO,=Al \siO,-Al 5H2O SUfi In each of these minerals variations are common, just as among the feldspars from which they are probably derived. Gismondite con- tains some potassium replacing calcium, or, in other words, an admixture from orthoclase or microcUne. Thomsonite may have as much as haK its Ume replaced by soda, due perhaps to onginal carnegieite, and it often carries an excess of siUca, either in ^'sohd solution" or else representing trisilicate groups. Carnegieite, it should be remembered, is an isomer of nephelite, a species that, under • some conditions, alters into thomsonite. I See Zambonini (Neues Jahrb., 1902. vol. 2, p. 79) for the composition of gismondite and also of phiUipsite. See also Sachs, Centralbl. Mineralogie, 1904, p. 215. 46 THE CONSTITUTION OF THE NATURAL SILICATES. Three more zeolites, like stilbite, are entirely trisilicate, namely, heulandite, epistilbite, and brewsterite. Heulandite and epistilbite are isomeric, or in crystallographic terminology the compound is dimorphous. They differ from stilbite in containing less water, 11 molecules instead of 12. The formula commonly assigned to them, if doubled as is done here, assumes only 10 molecules of water, but all trustworthy analyses give a larger proportion. Brewsterite differs from them in its dyad bases, having barium and strontium w^ith only a httle lime. In the formula to follow presently the calcium is united with barium. In heulandite strontium is often present, and soda to a small extent replaces hme. The two formulae, identical in type with those which have preceded them, are as foUows : Heulandite. Brewsterite. SigOs^Ca SigOg^-Ba / >A1— SigOs^Al / >A1— SiaOg^Al Al— SigOs^Ca Al— SigOs^Sr ^SigOs^Al ^SigOg^Al IIH2O IOH2O Brewsterite evidently is derived from an unknown feldspar con- taining strontium. Edingtonite and harmotome may represent original celsian or hyalophane. Doelter's synthesis of heulandite from anorthite has already been mentioned, but an interesting observation by Rinne ^ remains to be noticed. On decomposing heulandite with sulphuric acid he obtained a crystalline form of silica, which appeared to be cristobalite. This fact may shed some light on the relative molecular magnitude of the zeolite, as was suggested in Chapter II, on the silicic acids. Baschieri,^ working by Tschermak's method, regards heulandite and stilbite as salts of an acid, HioSigOiy. Natrolite and laumontite yielded him orthosilicic acid, and analcite he formulates as a dimeta- silicate. Erionite, an orthorhombic zeolite described by Eakle,^ probably belongs as a trisilicate with stilbite and heulandite. Its formula, which accurately reflects Eakle's analysis, is as follows: SigOs-^NaK / >A1— SigOs^Al Al— SigOs^Ca ^SigOs^Al I2H2O 1 Neues Jahrb., 1896, vol. 1, p. 139. 2 Zeitschr. Kryst. Min., vol. 46, p. 479, 1909. 3 Am. Jour. Sci., 4th ser., vol. 6, p. 66, 1898. THE SILICATES OF ALUMINUM. 47 Laimiontite, chabazite, gmelinite, and levynite are plagioclase zeolites in which the ratios are empirically metasilicate or nearly so. That is, Si04 and SigOg groups appear in equal or approximately equal numbers. Laumontite is essentially calcic, with insignificant alkaline replacements. In gmelinite alkalies predominate and lime is quite subordinate. Chabazite varies widely from a calcic variety to one which is mainly alkaline, and levynite is a lime zeolite with Si04 to SigOg as 3 to 2. The variations in composition are quite like those which occur among the feldspars, and the crystalUne comminglings are of the same order. If regarded as a zeolite, anal- cite is of similar constitution to these species but of lower hydration and less variabiUty.^ Lemberg's syntheses of analcite from three of them have already been mentioned. Now, repeating the formula of analcite to facilitate comparison, the several zeolites can be well represented as follows: Analcite. Gmelinite. / >Al-SiO,^Al / >Al-SiO,^Al Al— Si308=Na, Al— SigOg^Na^ \Sig08^Al ^SigOg^Al 4HP I2H3O Laumontite. -. ■ Caldum chabadte. SiO. ^Ca SiO, =Ca / .>Al-SiO,sAl / >Al-SiO,=Al All-Si,0«^Ca Al-Si30,^Ca \si30,=Al \si30,=Al 8H3O 12H,0 Sodium chabazite is empiricaUy identical with gmelinite but different in form. This sihcate therefore appears to be dimorphous or isomeric, and the possibUity of isomerism is easily shown. One acid radicle is represented in what may be caUed the side cham of the molecule as SiO,. Let that exchange places with an SijO, group and an isomeric arrangement is at once given. There are other possibihties, but the one is enough for present purposes. As for the last of these plagio- clase zeohtes, levynite, the best analyses represent a mixture of three orthosihcate and two trisihcate molecules of the same type as the other members of the group, with calcium for the dyad radicle, and 10 molecules of water. The formula commonly assigned to levynite, CaAl,Si30„.5H,0, does not fit the facts. From Hillebrand s analysis of the Table Mountain levynite we get the foUowmg companson between observation and theory: . ■ on variation, in thocompositioaofanalclteseoroote and Bradley, Am. Joar. Soi., 4th ser.vol. 33, p. 433, 1912. 48 THE CONSTITUTION OF THE NATURAL SILICATES. Found. Calculated. SiOa 46.76 21.91 11.12 1.34 .21 18.65 46 55 AloO, 21 98 Cab ] 12 07 NagO KoO 1 HoO 19 40 99.99 lOO! 00 So far all the zeolites considered, except hydronephelite, are assigned one type of formula with varying hydration. If we unite SiO^ and SigOg under the indiscriminate symbol X, the general formula becomes and this covers all variations of composition accurately. R'' may stand for calcium, strontium, or barium, and E,' for either sodium or potassium. The derivation of the zeolites from feldspar and lenads, however, is not always direct. In many cases it can be observed and verified, but in others the zeolites seem to have been formed in cooling magmas from feldspathic material rather than from the feld- spars themselves. Inclosed bubbles of water, perhaps magmatic water, have helped to generate the zeolites, especially in amygdaloid rocks. The zeolitic amygdules can hardly be explained otherwise, and in such a process the Si04 and vSigOg radicles may easily be sup- posed to change places, forming the silicate nuclei corresponding to calcium albite on the one hand and sodium anorthite or carnegieite on the other. Regardless of the vahdity on nonvalidity of the foregoing sugges- tions, which, by the way, are not new, the constitutional and genetic connection between the normal zeohtes and the feldspars seems to be perfectly clear, and it ought to be easily confirmed by petrographic investigation. Data of this kind, in addition to those alrec.dy cited, are even now available, and many alterations of the most pertinent kind have been observed. Thus laumontite, heulandite, stilbite, and analcite alter into albite or orthoclase; laumontite and stilbite into analcite ; chabazite into natrolite ; and gismondite into phillipsite. So also alterations into prehnite are recorded on the part of laumontite, scolecite, mesolite, natrolite, and analcite, and the identity of chemical type seems to be ahnost unquestionable. From the formulae here developed all these alterations become intelligible, and the theory of substitution from normal salts is very emphatically sustained. Several other zeolitic minerals are known, which, however, do not belong in the normal series. The two closely allied species mordenite THE SILICATES OF ALUMINUM. 49 ajQd ptilolite, for example, are to be classed as metadisnicates, and their constitution, which I have fully discussed elsewhere ' is easily expressed by regarding both minerals as mixtures of the two molecules Al— S12O5 Al— Si^Os— R' ySi,0, +6H2O and \si2O5 +6H3O Al— Si^O— H Al— Si^Os— H \siA-H \siA-H in which R^ =Na or K. In one occurrence of ptilolite the water is lower than is required by these formulae, and it seems probable that a trihydrate may exist. The metasilicate zeolite, laubanite, is the precise equivalent of ptilolite and mordenite and is easily interpreted thus: Al— S1O3 \siO3 +6H3O Al— SiOg \sio;>^^ Pilinite, a similar mineral, seems to be Al2(Si03)5Ca2.H20, a monohy- drate corresponding to the hexhydrated laubanite. Unlike laubanite, pilinite is undecomposed by hydrochloric acid, but physically all four of the species here grouped together resemble one another very closely. Possibly bavenite,2 CaaAlzSigOig.HaO, is to be classed with lau- banite as a metasilicate, although different in structure. Foresite and the manganese zeolite ganophyllite are two more species of unusual form. Their formulae can be written thus: Foresite. ' GaTwphyllite. Al--SiO,^Al Al~-SiO, = J ' \siO4=0a +IOH3O ^i04= Mn 4-7H30 Al-SiO, =A1 Al-SiO, = I ^^ \si308=Al \si30s= 1 Am. Jour. Sci., 3d ser., vol. 44, p. lOX, 1882. 2 See Artini, R. accad. Llncei Attl rondicionti, vol. 10, p. 139, 1901. 43633"— Bull. 588—1 50 THE CONSTITUTION OF THE NATURAL SILICATES. These expressions represent fairly well the actual composition of the two minerals, the analysis of foresite by Manasse ^ and that of ganophyllite by Hamberg ^ being taken for comparison. Foresite. GanophylUte. Found. Calculated. Found. Calculated. SiOa 48.93 27.56 49.91 28.28 39.67 7.95 .90 Lll .20 35.15 2.18 2.70 .20 9.79 42 69 ALO, } 8.60 Fe!or CaO 5.16 5.18 MffO MnO 39.29 NagO L14 KoO PbO? HoO 16.66 16.63 9.96 99.45 100. 00 99.85 100.00 Both species need additional study, especially with reference to their possible variations. A still more unusual type of zeolite is stellerite, recently described by Morozewicz.^ To this the empirical formula CaAlgSiyOig.THaO is assigned, which constitutionally is equivalent to Al— Si^Os^ +7H2O \si308=Al Grattarola's pseudonatrohte seems to have nearly the same formula, but with only SHjO. The structure proposed for these species is, of course, only tentative, and may be set aside at some future time. A considerable number of other zeolitic minerals have been described, but their nature is by no means clear. Offretite may be a variety of phillipsite, unusually rich in potash, and gonnardite may belong with natrolite and scolecite. References to other imperfectly known zeolites may be found in Dana's Mineralogy. The morphological characteristics of the zeolites probably depend in great part upon their mode of hydration, but this point needs to be developed. So also does the relation between zeohtes and kaolin, into which the minerals of this group sometimes alter. Furthermore, zeolitic substances of indeterminate nature are believed to exist in soils and clays, and it is conceivable that such bodies may be inter- mediately formed during the transition from feldspar into kaolin. In 1 Zeitschr. Kryst. Min., vol. 35, p. 514, 1902. a Geol. Foren. Forh., vol. 12, p. 586, 1891. Acad. Cracovie Bull., 1909, vol. 2, p. 344. THE SILICATES OF ALUMINUM. 51 studying the mechanism of that change this possibiUty ought to be considered. THE MICAS AND CHLORITES. On account of their wide distribution, their variety of composition, and their genetic relations to other species, the micas and chlorites form one of the most instructive and interestmg families of minerals. Two of the micas, muscovite and biotite, have already been noted among the members of the first and second of the preceding groups; and we have seen how frequently they are produced by the alteration of other silicates, some of which have been synthetically derived from micaceous material. As regards the substitution theory, the minerals of this family are pecuHarly suggestive, for the reason that they form a series of the most complete character. Thus, starting from the normal aluminum ortho- sihcate, we have Normal orthosilicate AL(Si04)3 Muscovite Al3(Si04)3KH2 Normal biotite Al2(Si04)3Mg2KH Normal phlogopite Al(Si04)3Mg3KH2 No further substitution of the same order is possible, for the reason that it would remove the Unking atom of aluminum, and break up the fundamental molecule. Muscovite, the first species in the foregoing series, occurs in nature as an independent mineral, and also as an alteration product of nephehte, eucryptite, topaz, andalusite, the feldspars, the scapolites, and various other natural silicates. All these alterations become inteUigible in the Hght of the formulae adopted in this memoir. In its more typical occurrences muscovite agrees sharply with the formula given, but it varies in composition within well-defined limits. First, it ranges toward its sodium equivalent, paragonite, which has the cor- responding formula AlgCSiOJgNaHj. Secondly, in fuchsite, the chromic mica, a chromium salt partly replaces the aluminum com- pound, and similar ferric replacements are also known. The chromic replacement is generally quite small, and so, too, is that of iron, although one sericite (a secondary muscovite), analyzed by Senn- hofer,! is very nearly represented by AlaFeCSiOJgKHj. Much larger replacements of aluminum by vanadium are found in the mineral roscoehte, in which as much as 24 per cent of V2O3 has been determined. An ideal roscoehte should have the formula AIV2 (SiOJgKHs, requiring 33.6 per cent of V2O3, but the pure compound is yet to be discovered. In kryptotile, A^CSiOJaHg, we have probably the extreme hydrogen end of the muscovite series, and leverrierite may be the same species but of different o rigin. The presence of 1 Min. pet. Mitt., vol. 5, p. 188, 1882-83; Dana, E. S., System of mineralogy, 6th ed., table, p. 618, 1892, analysis No. 43. 52 THE CONSTITUTION OF THE NATURAL SILICATES. magnesia or of ferrous iron in a muscovite is attributable to small admixtures of biotitic molecules. The most important variation in muscovite is in the direction of increased silica. Normal muscovite contains 45.3 per cent of SiO^j but varieties exist in which the percentage rises to nearly 59. Mus- covites of this class have been designated by Tschermak as phengites, and they are most easily explained upon the supposition of trisilicate admixtures. The molecule Al3(Si308)3KH2 is identical in type with ordinary orthosilicate muscovite, and its presence completely accounts for all excesses of silica over the normal amount. In Sandberger's lepidomorphite, for instance, the orthosilicate and trisilicate mole- cules occur in nearly equal proportions. All known muscovite may be represented by the general formula Al3(Si04)3E,'3 + Al3(Si308)3R'3, in which the latter molecule varies from to 50 per cent, and with ferric iron, chromium, or vanadium sometimes partly replacing alumi- num. The authenticity of this trisihcic variation is fully confirmed by certain of the lithia micas, in which the ratios are entirely trisili- cate. AlUed to muscovite and paragonite there is also the basic mica euphyUite, in which the univalent group — A1=(0H)2 appears. The formula of euphyllite appears to be Al3(Si04)3KH(A102H2), which agrees closely with the best analyses. With the biotites and phlogopites the variability of composition is much greater than in the muscovite series. Typical or normal biotites may be represented by th^ subjoined formulae, the actual minerals, however, being commonly mixtures. .SiO=MgK Al— SiO^^MgH \si04=Al .SiO=Fe''K Al— SiO,=Fe''H \siO,^Al .SiO,=MgK Al— SiO,^MgH \siO,=Fe''' .SiO,=Fe''K Al— SiO,=Fe''H \siO,=Fe''' These formulae correspond to the following c ompositions : 1 2 3 4 SiOs 43.06 24.40 37. 35 21.16 40.27 11.41 17.89 17.89 36.22 ALO, 9.98 Fe^Oo 15.66 ^ 19.14 F&).;.;;.".. ..::::. :::::::: 29.88 9.75 1.86 28.18 K2O 11.25 2.15 10.52 2.02 9.20 HoO 1.76 100. 00 100. 00 100. 00 100.00 THE SILICATES OF ALUMINUM. 53 The siderophyllite of Lewis agrees very closely with No. 2 of these formulae. Haughtonite is near an equimolecular mixture of Nos 1 and 2, with some ferric replacement of aluminum. Sodium, as in muscovite, often partly replaces potassium. In the normal phlogopite series four typical compounds may occur but the entirely magnesian variety is the only one which is found even approximately pure. These compounds are — .SiO,=MgK Al— SiO,=MgH \siO,=MgH 7 /SiO,^MgK Al— SiO=Fe''H \siO,=Fe"H /SiO,=MgK AI— SiO,=MgH \si04=Fe"H 8 ySiO,=Fe''K Al— SiO,=Fe"H \siO,=Fe"H equivalent to the following percentage compositions: 5 6 7 8 SiOo 43.27 12.26 28.85 40.18 1L38 17.86 16.07 10.49 4.02 37.66 10.67 8.37 29.70 9.83 3.77 35.16 9.96 ALO, MgO FeO 42 19 KoO 1L29 4.33 9 18 HoO.. 3 51 100. 00 100. 00 100. 00 100.00 A sodium phlogopite containing no potassium has been described by Griinhng/ but his analysis is not altogether satisfactory. Aspidolite is another mineral which is probably sodium phlogopite, but it needs reexamination. In manganophyll, a manganese mica of variable composition, with from 9.7 to 17.1 per cent of MnO, the molecule Al(Si04)3Mn3KH2 seems to occur. It is intimately associated with the manganese zeolite, ganophylhte, which is itself micaceous in appearance. The two minerals are closely related. To the typical biotite and phlogopite molecules few natural micas actually correspond, although intermediate mixtures are very com- mon. Many of the analyses, moreover, are difficult to interpret with any degree of accuracy and for several reasons. The state of oxida- tion of the iron is frequently uncertain, because in grinding a mineral for analysis ferrous compounds may be partly oxidized to the ferric condition. In fine grinding, furthermore, some water is adsorbed 1 Zeitschr. Kryst. Min., vol. 33, p. 218, 1910. 54 THE CONSTITUTION OF THE NATURAL SILICATES. from the atmosphere, and an error by no means small is thereby incurred. Titanium is also present in many micas, and its exact function in them is quite unknown. It may be present as TiOg replacing silica, as TigOg replacing alumina, or, which is probably more common, as inclusions of rutile. An unusual type of mica is Breithaupt's alurgite, which, as analyzed by Penfield,^ corresponds to a mixture of molecules — 2Al3(Si303)3KH3 SAl^CSiOJaK^H, 3Al3(SiOj3Mg,KH with a slight excess of H over K in the last compound. The second of these molecules, an alkaline biotite, is the characteristic feature of alurgite. Similar compounds, parallel to phlogopite, seem also to exist, having the general formula Al (8104)311 'g, but all of these bodies conform sharply to the general theory of the micas and are substitu- tion derivatives of the normal aluminum salt. In many of the magnesian micas fluorine is found, and the iron micas frequently contain oxygen in excess of the amount necessary to convert all the silicon into the radicle SiO^. When this excess is real, that is, not ascribable to defective analysis, it may be due either to alteration or to the replacement of univalent radicles by such groups as AIO2H2, and R^'OH. Replacements of this kind indicate a transi- tion toward the chlorites, as will be seen later. Fluorine in the ferromagnesian micas may represent either a group like — A1=F2 or — R" — F, and these appear most conspicuously in the lithia micas. An average lepidohte, for example, agrees well with the formula ^Si303=(AlF2)3 ^SiO=H,K,Li Al— Si30=K3 + Al— SiO=AI ^SigOg^Lig \siO=Al That is, the mineral is a mixture of a trisiUcate with a muscovite, the actual proportions varying on both sides of the ratio 1:1. Lepido- lite sometimes forms borders on plates of muscovite, and Baumhauer ^ has shown that lepidohte often contains inclusions of muscovite recognizable only under the microscope. Another Uthia mica, zinnwaldite, is a similar mixture of the same trisilicate with a ferrous biotite, thus: /Si30s^(AlF3)3 ^SiO^^FeK Al— SigOg^Kg + Al— SiO =FeH ^SigOg^Lig ^SiO^^Al 1 Am. Jour. Sci., 3d ser., vol. 46, p. 289, 1893. Penfield's interpretation of alurgite is quite different from that adopted here, a Zeitschr. Kryst. Min., vol. 51, p. 344, 1913. THE SILICATES OF ALUMINUM. 55 which represents its composition very closely. The calculated com- position of the two micas is as follows : Lepidolite. Zinnwal- dite. SiOa 5L43 25.50 46.54 2L58 10.15 13.27 3.17 .63 9, 04 A1203 FeO K2O 13.43 4.29 .64 8.13 LioO HoO F Less 103. 42 3.42 103. 38 3 38 ^ 100. 00 100.00 In both minerals a little potassium is commonly replaced by sodium, and a little fluorine by hydroxyl. Cryophyllite is near zinn- waldite but more complicated. It forms borders on plates of the iron mica annite, which is a mixed silicate between a biotite and a phlogopite, and its derived lithia mica exhibits similar complexity. Other known lithia micas are varying mixtures of the typical mole- cules found among the other members of this group of minerals, and in irvingite, which is fully two-thirds trisilicate, an alkalLue biotite appears. Irvingite is well represented by the formula 8Al(Si308)3(AlF2)3K3Li3 + 9Al2(Si308)3K2H, + 1 1 Al^CSiOJgLigNag. Polylithionite, which is entirely a trisilicate mica, has a quite differ- ent type of formula from those already given. The typical mineral as shown by Lorenzen's analysis has the formula F ^SieOs^Na^K 5A1— F + lAl— Si303=Na2K \si308=Li3 \si308=Na3K which leads to the subjoined comparison: AI2O3- FeO... K2O.-. NaaO-. LiaO.. F Less O Found. Calculated. 59.25 12.57 .93 102. 11 3.08 99.03 59.79 12.74 5.85 7.72 9.34 7.88 103. 32 3.32 100.00 56 THE CONSTITUTION OF THE NATURAL SILICATES. In Flink's analysis of polylithionite from another locality potassium is in excess of sodium, but otherwise the ratios are nearly the same. In the clintonite group or so-called ^'brittle micas" we have a series of highly basic compounds commonly free from alkahes. They are morphologically like biotite and are characterized by the presence of the univalent radicle AIO2R'', R'' being either Mg, Ca, Fe'', or Mn. The most basic mica of the group, the end member of the series, is xanthophyUite, which has approximately the formula AlCSiOJgCAlOaR'Og, with Rg^MggCag. In seybertite three of the univalent radicles are replaced by hydrogen, and in chloritoid there is still more replacement of a different kind. The ideal formulae are as follows: Xanthophvllite. Seybertite. Chloritoid. ^iO^— (A102Ca)3 ySiO,— H3 .SiO,=(A102H2)H2 Al— Si04=(A102Mg)3 Al-Si04=(A102ll)3 Al— SiO,— (AIOH)H \si04=(A102Mg)3 \siO,=(A102R)3 \siO,=(A102Fe)3 In xanthophyUite there is always some hydration, and in the other species there are various small replacements of Al by Fe''', of Fe'' by Mn, etc., as in all the other micas. Ottrelite, a fourth member of the series, is like chloritoid, but contains SigOg instead of SiO^; that is, its formula is trisihcate The pure theoretical compounds have the following composition : Xantho- phyUite. Seybertite. Chloritoid. Ottrelite. SiOo 16.45 46.37 21.53 42.70 23.81 40.48 28.57 48 38 AI2O3 27 42 FeO 19.36 MgO 21.91 15.27 19.14 13.40 3.23 CaO HoO 7.14 4.84 100. 00 100.00 100. 00 100. 00 XanthophyUite decomposes on ignition into two portions — one soluble and the other insoluble in hydrochloric acid. The insoluble portion has the composition of spinel, a mineral which generally accompanies cUntonite micas, and of which the formation is ren- dered inteUigible by the formulae.^ The actual decomposition may perhaps be represented by the subjoined equation: XanthophyUite. Spinel. Garnet. Ca3MgeAl,o(SiO,)30i3 = 4MgAl20, + Al,(SiO,)3Ca3 + 2MgO. 1 For a different interpretation of these micas see the former edition of this memoir, U. S. Geol. Sur- vey Bull. 125, 1895; also for details see Clarke and Schneider, U. S. Geol. Survey Bull. 113, p. 27, 1893. THE SILICATES OF ALUMINUM. 57 The lime-alumina garnet and the free magnesia would constitute the soluble portion of the ignited mineral. The equation, of course, is purely hypothetical and would be difficult to verify experimentally. One other mineral, willcoxite, an alteration product of corundum, seems to be best classified with the clintonite micas. It appears to be a basic analogue of the mixed biotite-phlogopite kind, and its analysis gives quite sharply the subj oined formula : ^i04=(A102Mg)2Na 1 Al— SiO,=(A102Mg)2H \siO,=(A102Mg)3H ^iO,=(A102Mg)Na2 + 3 Al— SiO,=(A102Mg)H2 \siO,=Al A little iron is present in the mineral and the sodium is partly replaced by potassium. From some points of view kaolin may be regarded as a member of the mica series, especially when its crystalline form is considered. With it the calcium mica, margarite, which is commonly classed as a member of the clintonite group, can be conveniently correlated. Furthermore, margarite yields an alteration product, dudleyite, which falls into line with the other two species, thus : Kaolin. Margarite. Dudleyite. .OH yOn yOR Al_Si04=H3 Al— Si04=CaH Al— SiO^^CaH \si04^Al \si04=(A10)3 \siO,^A10H.AlHA Cookeite, a micaceous mineral found associated with lepidolite and encrusting lithia tourmalines, also seems to belong here. Its formula is simply written •OH Al— Si04=Li,H(A102H2) + H3O \si04=Al which agrees well with Penfield's analysis of the cookeite from Maine and Schaller's analysis of the California mineral. SiOs- AI2O3 Fe^Oa CaO. LijO. NaaO. K2O.. H2O.. F Penfield. Schaller 34.00 45.06 .45 .04 4.02 .19 .14 14.96 .46 ).32 35.53 44.23 Trace. 2.73 2.11 .31 14.18 L46 100.55 Calculated. 35.08 44.73 4.40 15.79 100.00 58 THE CONSTITUTION OF THE NATURAL SILICATES. These expressions for kaolin and its analogues are suggestive but not altogether conclusive. They represent the known facts fairly well, however, and so serve their purpose for the time being. By hydration, and sometimes by oxidation, the micas undergo alteration, yielding a great variety of products which are known in general as vermiculites. This is especially true as regards the ferro- magnesian micas, which lose alkalies and take up water with the greatest ease, in accordance with what seems to be a well-defined law. Thus we have Biotite. Hydrohiotite. ^iO^^MgK ^iO,=MgH Al— SiO,=MgH Al— SiO^^^IgH + SH^O ^SiO^^Al \si04=Al Phlogopite. Hydrophlogopite. ^i04=MgK ^iO,=MgH Al— SiO,=MgH Al— Si04=MgH + SH^O \si04=MgH \si04=MgH These micas occur in nature in great variety of admixture, and the corresponding vermiculites show a parallel complexity. In the normal series, however, the alteration commonly follows the line indicated by the formulae, and the vermiculite is simply the mica with H in place of K or Na, plus 3 molecules of loosely combined water. Two of these molecules are, as a rule, given off at 100°, and regained in moist air, suggesting an analogy between the vermiculites and the zeolites. Some vermiculites are only monohydrated, and many of the so-called species which have received names are mere mixtures of altered and unaltered micas, representing stages of transition between the original mineral and the final product. Maconite, lucasite, and phila- delphite are incompletely altered micas of this kind. Jefferisite is quite near the normal hydrohiotite, kerrite approxi- mates to a hydrophlogopite, and lennilite is a mixture between the two, but in all three of these minerals the hydration is somewhat irregular, and there are the usual replacements of aluminum and mag- nesium by ferric and ferrous iron. In some of the vermiculites basic radicles appear, corresponding to the excesses of oxygen over the normal ratios that are found among the micas themselves. For example, roseite and protover- miculite may be written: Roseite. Protovermiculite. .SiO,=(MgOH)2H /SiO,=H3 Al— Si04=(A10H)H + 2H2O Al— SiO,^(MgOH)3 + SH^O \siO,=(A10H)H \siO,^Al THE SILICATES OF ALUMINUM. 59 expressions which fit the actual analyses fairly well. Such formulae, however, must be interpreted with much caution. They do not necessarily imply that these micas are definite compounds; they merely symbolize one kind of alteration to which the minerals of this group are subject. They mark a transition between the micas and the chlorites, and similar but more complex examples are foimd in hallite, painterite, pyrosclerite, vaaHte, and pattersonite. Cas- wellite is another altered mica, rich in manganese and lime but of uncertain formula. Indeed, it is hardly worth while to write formulae for these minerals, for none of them seems to be a single definite compound. Between the micas and the more basic chlorites the relations are exceedingly close. All the species are foliated, all or nearly all are monoclinic, and to each of the ferromagnesian micas one or more chlorites, higher in magnesia and water, seem to correspond. The exact formulation of the chlorites, however, is not a simple matter. Some of the so-called '' species" are not homogeneous; others are isomorphous mixtures; and in all of them replacements of one dyad base by another, or of aluminum by iron, occur. In the chlorites the basic univalent and bivalent groups — Mg — OH, — Fe" — OH, — A1=(0H)2, =-Al— OH, — Fe'''=(0H)2, and ==Fe"'— OH appear, but their precise identification is complicated by uncertainties in the hydration of the minerals. In general, the water shown by the analyses is constitutional, but in some chlorites it may be extraneous, like the water of the zeolites. An exact study of the hydration of the chlo- rites by modem methods is yet to be made. Some so-called chlorites, containing alkalies, are obviously mix- tures of chlorites and micas, but the true species are all referable to the types of molecule represented by biotite and phlogopite except a few that fall more nearly into line with margarite and kaolin. Typical chlorites, which, however, are rarely if ever found pure, may be repre- sented thus: Biotite-chlorite. Phlogopite-chlorite. .SiO,=(MgOH)2H ^SiO,=(MgOH)2H Al— SiO,=(MgOH)2H Al— SiO,=(MgOH)3H \siO,=Al \siO,-(MgOH),H An average pennine contains these two molecules commingled in the ratio 1:1, whereas in clinochlore and leuchtenbergite the ratio is 2 : 3, with the second Al of the biotite formula replaced by 3A10. That is, the formula of a typical clinochlore is represented by 3 Al(SiOj3(MgOH)eH3 + 2 Al(SiOj3(MgOH)A(A10)3. These expressions give the subjoined compositions for the two chlorites. 60 THE CONSTITUTION OF THE NATURAL SILICATES. Pennine. CUno- chlore. SiOo 34.35 14.60 38.17 12.88 31. 58 AloO, 19. 65 MgO 36.49 H2O . . 12 28 100. 00 100. 00 The mixtures may occur in other proportions; the magnesium may be partly replaced by iron, and in the varieties kammererite and kotschubeite some chromium, equivalent to aluminum, appears. A glance at the tables of analyses in the textbooks of Dana and Hintze will show how variable in composition these and other chlo- rites really are. Three chlorites, rumpfite and the ferrosoferric minerals cronstedtite and melanolite, seem to conform very nearly to the biotite type of formula, thus — Rumpfite. Cronstedtite. ^SiO,^(MgOH),H ^SiO=(Fe''OH)2(Fe'''0,H3) Al_SiO,=(A103H2)3 \siO,=Al Fe— SiO,=(Fe"OH)2(Fe'''02H2) \siO=Fe'" Melanolite. .SiO,=(Fe''OH)H2 Fe— SiO,=(Fe''OH)H2 \siO,^Fe''' The composition of the three minerals, as given by these formulae, is as follows : Rumpfite. Cronstedt- ite. MelanoUte. SiOa 30.20 42.78 20.93 33.70 ALOo fiK :::. 37.21 33.48 29.63 FeO 26. 67 MgO . . - 13.42 13.60 H2O 8.38 10.00 100. 00 100. 00 100. 00 In melanolite a little ferric oxide is replaced by alumina. The two species brunsvigite and delessite may be given either biotitic or phlogopitic formulae, according to the character of the THE SILICATES OP ALUMINUM. 61 hydration. If one molecule of their water is zeolitic in character- that IS, nonessential-the minerals are of the biotite type If all th« water is constitutional, they are related to phlogopite Brunsvigite IS essentiaUy a ferrous chlorite, delessite is ferromagnesian, and the two species correspond to the two formulae— Brunsvigite. ^iO,=(FeOH)2H Al— SiO=(FeOH)2H \siO,=(A10H)H Delessite, ^iO,=(MgOH),H Al-SiO,=(MgOH),H \siO =(A10H)H In delessite the magnesian end compound is sometimes nearly approached, but usually about one-fourth of the magnesium is replaced by iron, and a little ferric replacement of alumina is ako common. Delessite is a very variable mineral. When computed with MggFei, its average composition is as follows, m comparison with brunsvigite: Delessite. Brunsvig- ite. SiOs 32.97 18.68 13.19 21.98 13.18 28.04 15.89 44.86 AloO, FeO MgO H2O ii.'2i 100.00 100.00 Closely related to the two preceding species are prochlorite and grochauite, to which the following structures may be assigned: Prochlorite. ^iO,^(MgOH),H Al— SiO,=(FeOH)2H "^SiO^^CAlOH) (AIO2H2) Grochauite. ^iO,=(MgOH)3 AI^SiO=(MgOH)3 \siO,=(A10H) (AIO2H2) The composition of these two minerals is as follows: Prochlorite. Grochauite. SiOo - 28.28 23.90 22.50 12.66 12. 66 27.53 ALO, 23.39 FeO MgO 36.69 H2O 12.39 100.00 100.00 62 THE CONSTITUTION OF THE NATURAL SILICATES. This prochlorite is fairly typical, but a so-called " prochlorite " from Culsagee, North Carolina/ is quite different in its ratios and contains very little iron. It is allied to metachlorite, as the following formulae show: Culsagee prochlorite. ^iO,=(MgOH)3 Al— SiO,=(MgOH)2H \si04=(A10H)H The percentage compositions are as follows : Metachlorite. ^iO=(FeOH)3 Al— SiO=(FeOH)2H \siO,=(A102H2)H Culsagee prochlorite. Metachlo- rite. SiOa 28.48 24.21 22.73 AloO, 19.32 FeO 45 46 MgO 3L65 15.66 H2O 12.49 100.00 100. 00 Corundophilite is a chlorite with still higher alumina, which prob- ably has the following constitution: ^SiO,=(ROH)3 Al~SiO=(ROH)3 \siO,=(A10,H,)3 SiO^ , 22.68 AI2O2 '. 25. 69 :^eO 17.88 MgO 20.15 ^20 13.60 100. 00 The calculated composition gives Rg = Mg4Fe2. The water is a little too high, and may be partly vermicuHtic. If so, (A102H2)3 should be replaced by (A10)3, ^ replacement which may be desirable in some of the other formulae already given. The need of investigating the hydration of the chlorites has already been pointed out, but it is well to emphasize it here. Diabantite and thuringite appear to be mixed sihcates, inter- mediate between the vermicuUtes and the chlorites, thus : 1 Analyses 17-20 on p. 654 of Dana's System of mineralogy, 6th ed., 1892. THE SILICATES OF ALUMINUM. 63 1 Al— SiO,=RH Diabantite. ^iO=(ROH),H + 2 Al— SiO,=(ROH)2H \siO,=RH Thuringite. ^iO,=FeH /SiO=(FeOH)3 Al— SiO,=FeH + Al— SiO,=(FeOH)3 \siO,=(A102H2)3 \siO,=.(A10,H,)3 In diabantite Ri2 = Mg7Fe5, and in thuringite one-third of the aluminum is replaced by ferric iron. Hence the following composi- tions : Diabantite. Thuringite. SiOa 34.93 13.19 22 78 AloO, 19 37 FeoOo 10 13 FeO : 23.29 18.11 10.48 36.33 MgO HoO 1L39 100. 00 100.00 The talc-chlorite of Traversella is another intermediate compound simply formulated as shown below. In the actual mineral one-fifth of the dyad portion is Fe. .SiO,=(ROH)2H Al— SiO,=RH \siO,=RH SiOa ^ 39.00 AlA 11.04 FeO ^ 12.47 MgO 27.75 H2O 9.74 100. 00 Stilpnomelane has a truly chloritic formula, but is distinct from the others in being a trisihcate. This is probably true also of the very uncertain ekmanite, which has a most variable composition. The two formulae may be written thus: Stilpnomelane. Ehmanite. ^i308=(FeOH)2H ^iO,=(FeOH)3 Al— Si30=(FeOH)2H Al— SiO,=(FeOH)3 \si,0,=(FeaH)3H \siO,=(FeOH)3 64 THE CONSTITUTION OF THE NATURAL SILICATES. The theoretical composition follows : Stilpnome- lane. Ekmanite. SiOa 48. 91 4.62 39.13 7.34 40 91 AloOo 3 86 FeO 49 09 HoO 6 14 100. 00 100. 00 Epichlorite appears to be a mixture of stilpnomelane with the equiv- alent magnesian orthosilicate, Al(Si04)3(MgOH)6H3, in nearly equal percentages. The recently described minguetite ^ is apparently a mixture of stilpnomelane and a lepidomelane. So far all the chlorites conform to the mica type of formula as represented by biotite and phlogopite. In general they may be regarded as salts of an alumosilicic acid, Al(Si04)3H9, and its corre- sponding trisilicate, in which the nine hydrogen atoms are replaceable by a variety of basic radicles. The apparent complexity of the chlorites vanishes and the relations between them become clear and simple. There is, however, a group of chlorites of distinct character from the normal series. Like margarite they model after kaolin in the following manner: Kaolin. Al— SiO^^Hg \siO,^Al Strigovite. /OH Al— SiO,=(Fe''OH)H2 \siO,=Fe''' / Aphrosiderite. OH Daphnite. yOR Al— Si04=(Fe''OH)3 \siO,=(A10H)H Al— SiO,=(Fe''OH)3 ^SiO^^Al Sheridanite.^ .OH Al— SiO,^(MgOH)3 \si04=(A10H)H The theoretical composition of the four chlorites is therefore as follows : Strigovite. Aphrosid- mte. Daphnite. Sheridan- ite. SiO, 33. 61 14.17 22.22 20.00 25.32 2L52 24.40 20.73 30.30 ALOo 25.76 Fe'o' .... .. fS).....: .. .;.:...:. . :.:: 45.57 43.90 MgO :. 30.30 H2O 10.00 7.59 10.97 13.64 100.00 100. 00 100.00 100.00 1 Lacroix, Soc. min. Bull., vol. 33, p. 273, 1310. 2 See Wolff, J. E., Am. Jour. Sci., 4th ser., vol. 34, p. 475, 1912. THE SILICATES OF ALUMINUM. 65 Strigovite is the least satisfactory of these species, for the reason that the analyses are discordant. The formula given it in Dana may be written A10H(SiO,)2R.(ROH)3H, which, however, is of the same general type as that employed here. Several other chloritic minerals have received names, such as chamosite, klementite, eurahte, pycnochlorite, epiphanite, huUite, and chlorophseite, but they are not very well characterized. The analyses can all be interpreted in harmony with the other chlorites but it is hardly worth while to discuss them in detail until the several minerals shall have been more thoroughly studied. Enough has been done to show that the micas, clintonites, vermiculites, and chlorites form one systematic group of minerals, and all the valid evidence is satisfied. The facts that garnet and vesuvianite alter into chlorites and that chloritic pseud omorphs after feldspar are known serve to connect still more closely the formulae here adopted with the similar formulae of the preceding groups of minerals. When clinochlore or leuchtenbergite is strongly ignited, it yields, like xanthophylhte, a product insoluble in hydrochloric acid, having the composition of spinel. This reaction establishes still more defi- nitely the relationship between the chlorites and the cHntonite group, and it is readily intelUgible in the light of the structural expressions. The splitting up, under influence of heat, of mixtures containing such groups of atoms as MgOH, AlOH, and AIO2H2 ought to generate spinel, and the appearance of a compound of this character is evidence in favor of the formulae. An interpretation of the chlorites proposed by Tschermak has had some acceptance, but the scheme is complicated and subject to very serious objections. According to Tschermak the ''orthochlorites,'' which include pennine, cHnochlore, prochlorite, and corundophilite, are molecular or crystalline mixtures of serpentine, H^MggSijOg, and an uncertain substance, amesite, H4Mg2Al2Si09, with the equivalent molecules containing iron. Now serpentine, on strong calcination, breaks up into water, olivine, and enstatite, the enstatite being insoluble in acids. But the magnesian chlorites, which, if Tscher- mak's theory were true, should yield about 18 per cent of enstatite on ignition, yield none at all. Spinel is formed instead of enstatite, and in quantities proportional to the excesses of oxygen over the ortho- siHcate ratio. That is, serpentine molecules are not present in the chlorites, and the Tschermak hypothesis breaks down. THE ALUMINOUS BOBOSILICATES. In this group of minerals, of which tourmaline is the most impor- tant, there are five species, namely, tourmaline, axinite, dumor- tierite, serendibite, and manandonite. There are also several 43633°— Bull. 588—14 5 66 THE CON&TITUTION OF THE NATURAL SILICATES. borosilicates of the rare earths, which may properly be studied with them even though they contain no aluminum. Although tourmaline in its several varieties is apparently quite complex, the evidence for its interpretation is abundant and ample. Its variations in composition are shown by numerous good analyses, its associations are well known, and its alteration products have been observed in a sufficient number of examples. From the minerals which have been discussed in the preceding chapters it differs essen- tially in that it contains boron, and the part played by this element is a new question to be interpreted. When tourmaline undergoes alteration, the commonest product is a mica, and between the micas and the tourmalines there are very strik- ing analogies. With the lithia micas, Uthia tourmalines are generally associated; with muscovite and biotite, iron tourmalines occur; and magnesian tourmalines accompany phlogopite. In each case the composition of the tourmaline seems to bear a relation to that of the associated mica. Furthermore, the varieties of tourmaline shade one into another through an unbroken series of gradations, and this may happen to some extent in one and the same crystal. The genus tourmaline, in short, represents a series of compounds, and these are parallel to the normal mica series. Upon the constitution of tourmaline there have been many essays written and much controversy. It is not necessary to discuss here the Hterature of the subject in detail, for only two types of formula are now seriously considered, and they differ principally in regard to the ratio between silicon and oxgyen. Penfield and Foote ^ represent the mineral as a salt of the acid HuAlgEgSi^Oji, which may be written structurally thus: ^SiO,=H, Alf >SiO,=H2 Al— B^Os^H, In this formula the boron appears as the radicle of an acid H4B2O5, which is a rational compound. So far as its atomic ratios are con- cerned, the formula fits most of the published tourmaline analyses very weU, although not aU of them, but it does not suggest the well- known alterability of the mineral into mica. It is also difficult to apply in detail, that is, to apportion the several bases replacing hydrogen in any symmetrical manner, so as to indicate clearly the 1 Am. Jour. Sci., 4th ser., vol. 7, p. 97, 1899. Criticized by Clarke, idem, vol. 8, p. Ill, 1899. Reply by Penfield, idem, vol. 10, p. 19, 1900. THE SILICATES OF ALUMINUM. 67 end members of the series of compounds which are commingled in tourmahne. This difficulty has been avoided in part by later writers, notably by Schaller' and Reiner,^ who triple the Penfield-Footeformulai so making the ultimate tourmahne acid HgoSiiaBgOgg. From this both Schaller and Reiner derive a number of distinct compounds, which, as mixed crystals, make up tourmahne. The tripled formula, how- ever, is unwieldy, and not easy to represent by a simple and presum- ably stable series of structures. This objection is entitled to some weight but is not necessarily fatal. In the Penfield-Foote type of formula the siUcon-boron-oxygen ratio is represented by Si^BPai, which is equivalent to SigBgOgi.g. In the tourmahne formulae proposed in the former edition of this memoir the nucleus SigBgOgi appears, with an oxygen ratio sHghtly lower than that of Penfield and Foote. In general the Penfield-Foote formula fits the actual analyses a little better than mine, but the difference is very slight and suggests a possible constant error. Heretofore all analyses of tourmahne have been made upon finely ground material, and fine grinding is accompanied by more or less adsorption of water from the air. It also leads to some oxidation of ferrous iron, a change which takes place very easily in tourmaline, and these two almost certain sources of error tend to raise the oxygen ratio in the minerals as analyzed. These sources of error were not known when the pub- lished analyses were made and must now be taken into account. Now, the foregoing considerations being kept in mind, and als<> the alterabiUty of tourmaline into mica, the following formula for the fundamental tourmaline acid seems to be probable: .SiO,=H3 Al— Si04=H3 \si04=Al— BO2 Al— B03=H2 ySiO^^Al— BO3 Al— SiO,=H3 This involves the lowest observed ratio between aluminum and silicon, and also the constant ratio between sihcon and boron. The boron is shown as partly metaborate and partly orthoborate, which may be regarded as improbable. As an alternative the boron may be represented as the quinq uivalent group Bfi,, the radicle of a possi- 1 Zeitschr. Kryst, Min., vol. 51, p. 321, 1913. ,:„hiflcl 3 Inaugural DLertation, Heidelberg, 1913. Reiner-gives a good summary of earlier work on the subject. His formijse are those of Wiilflng. 68 THE CONSTITUTION OF THE NATUEAL SILICATES. ble acid H5B3O7, a rational compound. But many tourmalines con- tain fluorine, and some analyses show a deficiency of boron, which is not always ascribable to analytical error. Fluorine in such tourma- lines may perhaps replace the BO2 group, a supposition for which there are good arguments, although it may not be absolutely proved. To this point reference will be made later. In the proposed tourmaline acid various replacements of hydrogen are possible, by means of which the individual tourmalines, as mixed crystals, can be quite accurately formulated. At one end of the series, with the group AlgSig, we find some magnesia and iron tourma- lines, at the other end, approaching the ratio AlgSi^, are the colored lithia tourmalines. Of the lithia tourmaHnes the extreme member known is the rubellite from Elba, which approximates in composition to the following mixture of molecules : ^iO,=Al Al— Si04=Al \siO,=Al— BO, ^SiO,=H3 /SiO,=H3 Al— Si04=Al Al— SiO,=Al \siO,=Al— BO, \siO,=Al— BO2 3A1— BO,=Na, + 4 Al— B0,=Na2 + 10 Al— BO,=Li, ^iO,=Al— BO2 Al— Si04=Al \si04=Al Al— SiO,=Al \si04=Al ySi04=Al— BO2 Al— Si04=Al \siO— Al This may be compared with Schaller's analysis of the mineral thus : Calculated. Si02. B2O3. T1203 FeO- MnO. CaO. NagO LiaO. H2O. F... 38.08 11.06 43.98 100. 00 The differences here are mainly due to the water and to the small neglected impurities. The water found is probably too high, because 1 All the analyses of tourmaline cited here were made in the laboratory of the United States Geological Survey. THE SILICATES OF ALUMINUM. 69 of fine grinding. The comparison is also influenced by the fact that the mixed molecules are not exactly represented by such simple numbers as are assumed in this computation. The latter considera- tion applies throughout the discussion. At the other end of the series we have the magnesian tourmaline from Pierrepont, New York, as analyzed by Riggs. The mixed molecules are nearly as follows : SiO,=FeH Al— SiO^^FeH \si04=Al— BO 'SiO,=MgH /SiO=MgH Al— SiO,=MgH \si04=Al— BOj 3A1— B03=NaH + 2 Al— BO,=NaH + 5 Al— BO,=Ca Al— SiO,=MgH \siO,=Al— BO, xSiO^^Al— BO2 Al— Si04=FeH \siO=FeH .SiO^^Al— BO2 Al~SiO=MgH ^SiO^^MgH /SiO,=Al— BO2 Al— SiO,=MgH \si04=MgH Found. Calculated. SiOo 35.61 .55 10.15 25.29 .44 8.19 1L07 3.31 L51 .20 3.34 .27 35.91 TiOo B2O3 10.48 ALO, 25.44 FeoO, FeO... -. 8.62 MgO n.i7 CaO 2.79 NaoO L55 KoO HoO 4.04 F * 99.93 100.00 An interesting intermediate tourmahne is the black variety from Lost VaUey, Cahfornia, analyzed by Schaller. Its formula is that of a mixed crystal, as follows : .SiO=Al .SiO=FeH ^iO,=MgH Al— SiO,=Al Al— SiO=FeH Al-SiO,=MgH \siO,=Al-BO, \siO,=Al-B03 \siO,=Al-B03 Al— BO,=H, + 1 Al— B03=H2 2A1— B0,=Na2 + 2 Al— BOe^H^ + /SiO,: =A1-B03 ^lo,=Al-B03 /SiO,=Al-BO, Al'_SiO=Al Al-SiO,=FeH Al-SiO=MgH \siO,=Al 1 ^*v^4 \siO,=FeH \siO,=MgH 70 THE CONSTITUTION OF THE NATURAL SILICATES. Found. SiOa- B2O3. AI2O3 Ti^Og FeO.. MnO. CaO.. MgO. Na^O. H2O.. 35.96 10. 61 33.28 .36 1L04 .13 .42 3.48 2.16 3.31 100. 75 Calculated. 35.87 10.42 33.43 1L43 3.18 2.46 3.21 100. 00 It is not necessary to give more examples of tourmaline formulae, for all tourmalines are represented by the type of structure proposed here. The alkali tourmalines tend toward the more aluminous end of the series, the magnesian and iron tourmalines toward the less aluminous. All the tourmalines now known are evidently mixed crystals, but the complexity of the mixtures is more apparent than real. Whether any tourmalines contain ferric iron is still uncertain, but the probabilities are adverse to its presence. It appears in many analyses, but that is probably due to oxidation, so that the minerals analyzed were not quite normal. Ferric tourmalines are theoretically possible, but their existence is unproved. The iron of tourmaline is at least predominatingly ferrous. The formulae for tourmaline adopted here not only express the composition of the mineral but also indicate its obvious relation to the micas, and its ready alterability into them. A molecule of tourmaline, with elimination of boric acid and One atom of aluminum, splits into two molecules of the mica type, and the transformation is easily understood. Potash is of course taken up. Certain experiments by Ijemberg,^ who investigated the action of alkaline solutions upon tourmaline, are in accord with these suppositions. Although otherwise interpreted by Brogger, the minerals cappelin- ite, melanocerite, karyocerite, and tritomite seem to be chemically akin to tourmaline. This view of their nature has abeady been sug- gested by Wiik,2 and it is sustained both by chemical and by mor- phological considerations. Cappelinite is hexagenal, and the other species, like tourmaline, are rhombohedral. They are silicates of rare earths, which are mostly trivalent, like aluminum ; all contain boron, and all but cappelinite contain fluorine also. Furthermore, all four species, considered together, illustrate the reciprocity between boric acid and fluorine, which has been suggested in the discussion of tourmaline. Thus, if we compute the atomic ratios from the analyses cited by Brogger,^ the following relation appears: 1 Deutsche geol. Gesell. Zeitschr., 1892, p. 239. » Zeitschr. Kryst. Min., vol. 16, pp. 462-469, 1890. 2Zeitschr. Kryst. Min., vol. 23, pp. 421, 422, 1894. THE SILICATES OF ALUMINUM. 71 Cappelinite . Melanocerite Karyocerite. Tritomite Si. 236 218 216 226 488 92 134 210 304 296 226 B+F. 488 396 430 436 That is, Si : B + F : : 1 : 2 nearly, variations being due to the fact that in the first three minerals the boric acid was determined by difference, and also, probably, to the occasional replacement of fluorine by hydroxyl. Another source of variation is found in the presence of tetrad bases, as will be seen later, but for the moment the relation indicated seems to be reasonably clear. The first number of the group, cappelinite, is a borosilicate of yttrium and barium and approximates in composition to /BO3 BO3 Y— BO2 +4Y— BO2 \siO,^BaH \siO,=Y With the earths of uncertain molecular weight, designated as^'yttria," are a little lanthanum oxide and trifling quantities of ThOg and CeOj, and with the barium are some calcium and alkalies. The other three members of the group are all more complicated than cappelinite, and vary from it in t5rpe by containing tetrad oxides, such as CeOj, ThOj, and ZrOg. In eudialyte and catapleiite we have two rhombohedral silicates of zirconia, which help to explain these com- pounds. Catapleiite probably has the constitution (OH)3Zr.Si308.R'3. If we regard the tetrad bases in the cappelinite group as forming orthosilicates of this same type, the remaincTer of each mineral may be written as a mixture of molecules like those already designated, but with cerium earths predominating over yttrium, and fluorine replacing some boric radicles. Thus, melanocerite is not far from .OH .B03=Ca .F R^^< OH \si04=CaH \siO,=R' -F \siO,=R' Karyocerite may be written similarly, and tritomite becomes .OH ^BO^ R^^Fe SiO,=(A10)2 which expresses a partial relation to the micas, andalusite, and so on. The theoretical percentage composition calculated from this formula agrees well with the results of analysis, except that it gives the sihca nearly 1 per cent too low, a discrepancy which Penfield and Pratt attribute to inclusions of silica in the minerals analyzed. By means of a slightly different formula the relations of staurohte to the other species can be much more clearly shown, but it assumes 1 Am. Jour. Sci., 3d ser., vol. 47, p. 81, 1894. THE SILICATES OF ALUMINUM. 75 that the ideal staurolite is not yet known. The expressions proposed are as follows: Andalusite. Staurolite. /Si04=(A10)3 ^iO,=(A10)3 Al— SiO,=Al Al— SiO,=(A10)3 \siO,=Al \siO,=Fe Fe ^iO^^Fe Al— SiO,=(A10)3 \siO,=(A10)3 This formula, in contrast with that of Penfield and Pratt, and with their reduced analysis of staurolite from Lisbon, New Hampshire, gives the following percentage composition: Lisbon. Penfield and Pratt. New formula. SiOs 27.44 55.16 15.72 L68 26.32 55.92 15.79 1.97 27.90 55 35 ALO, Feb.. . 16 75 H2O... . 100. 00 100.00 100.00 If, now, we assume that the actual staurolite is slightly altered by hydration, some Fe being replaced by Hj and by FeOH + H, the dis- crepancies between formula and analyses are sufficiently accounted for. The new formula is more symmetrical than the old one; it better expresses the alterability of staurolite into muscovite; and it seems to satisfy the evidence with sufficient completeness. When we remem- ber that staurolite is excessively liable to inclusions and alterations, a very sharp agreement between analysis and theory is not to be expected. Still another orthorhombic species, harstigite, has a formula analo- gous in some ways to that of staurolite. For harstigite there is but one analysis extant, which gives nearly ^i04=CaH Al— SiO,=CaH \siO,==Mn Ca Al— SiO,=CaH \siO,=CaH 76 THE CONSTITUTION^ OF THE NATUBAL SILICATES. This, in comparison with Flink's analysis, gives the following per- centage composition: Found. Calculated. SiOo. AI2O3 MnO. MgO. CaO. K2O. Na^O H2O. 38.94 39.13 10.61 11.09 12.81 3.27 15.43 29.23 .35 30.44 .71 3.97 3.91 99.89 100. 00 This result is fairly satisfactory. More data relative to harstigite are evidently needed. Two closely related silicates, the calcic lawsonite and the manga- nous carpholite, may perhaps be analogous in structure to staurolite and harstigite. For both minerals the simplest empirical formula is H^R^'AlgSijOio, which, tripled, can assume the following form: Lawsonite. ^i04=(A10A)3 Al— Si04=H3 \si04=Ca Ca ^iO^^^Ca Al— SiO,=H3 \si04=Al Carpholite. .SiO,=(A10A)3 AI— Si04=H3 \si04=Mn ^iO,=Mn Al— Si04=H3 \si04=Al Lawsonite is orthorhombic. The isometric hibschite seems to have the same composition, but is too incompletely known to be satis- factorily discussed here. An interesting pair of ultrabasic silicates is furnished by the species kornerupine (or prismatine) and grandidierite. Both are ortho- rhombic; kornerupine alters into kryptotile, and grandidierite yields a similar and perhaps identical derivative. To kornerupine the empirical formula AlaMgSiOe has been commonly assigned, which, tripled, may be written: ^iO,=(A10,Mg)3 Al— SiO^^il \siO,^Al THE SILICATES OF ALUMINUM. 77 in close analogy to andalusite and kryptotile. The recent investi- gation of prismatine by Uhlig/ however, shows that the mineral is more complex and in better agreement with the formula NaEgMgeAl^^SiAo. Grandidierite,2 described by Lacroix in 1902, has the empirical formula R'\(A1, Fe)22Si7056, with aluminum largely predominant over iron. R" is mainly magnesium, but with some replacement by iron, calcium, and a little sodium. That is, it is a mixed crystalline mineral, but of fairly definite type. Both species, with the empirical formulae now given them, are expressible by analogous structures, thus • Komerujnne. ySiO,^Al Al— SiO,=(A102Mg)3 \si04=NaH Al— SiO^^Al '>SiO,=H3 Al— Si04=(Al02Mg)3 \si04=Al Grandidierite. ^SiO,=(A102R)8 Al— SiO,=(A10)3 \siO,=(A10)2 Al— SiO,=(A102R)3 a1— Si04=(A10)3 \siO,=(A102R)3 The nacreous or micaceous mineral batavite ^ is perhaps of similar type, although Weinschenk has assigned it the relatively simple formula, 4H20.4MgO.Al203.4Si02. Its composition, however, is equally well expressed as follows : ^iO^^MgH Al— SiO,^MgH )>SiO,=Mg Al— Si04=MgH + 4H2O \siO,=Mg Al— SiO,=MgH \si04^MgH Batavite, like many other minerals, needs further investigation, as the following comparison between analysis and formulae clearly shows : Found. Calculated. Wein- schenk. New formula. RiO 42.33 16. 35 28.17 13.19 4L81 17.77 27.88 12.54 43.30 A 1 15.78 MaO 28.86 H20 12.06 100.04 100. 00 100.00 1 Zeitschr. Kryst. Min., vol. 47, p. 215, 1910. 8 Soc. min. Bull., vol. 25, p. 85, 1902; vol. 27, p. 259, 1904. 78 THE CONSTITUTION OF THE NATURAL SILICATES. A still more unusual type of silicate is presented by didymolite, a mineral recently described by Meister/ who assigns it the formula 2Ca0.3Al203.9Si02. Constitutionally this seems to be a basic deriva- tive of a trisilicic anorthite, with the following structure: .SigOg^AlO (AlO^Ca)^ Al— SigOs^Al \si308=Al To bityite, a mineral from the tourmaline region of Madagascar, described by Lacroix,^ may be assigned the empirical formula H7Li2GlCa3Al9Sie034. This leads to a structural formula analogous to that of tourma- line, which suggests that bityite may be an intermediate compound between tourmaline and cookeite. ^iO,=(A10A)3 Al— SiO^^I^iR" \siO,=R'' Al— OH ^iO,=R" Al— SiO^^LiR'' \si04=(A10)3 The comparison between analysis and formula is as follows: Found. Calculated. SiOo 3L95 4L75 14.30 .13 2.27 2.73 .40 .16 6.50 32^58 ALO, 4L54 Cab.. 1 15. 20 2.26 GIO LioO ] 2.72 NaoO KoO H2O 5 70 100. 19 100. 00 The water found is probably a little too high, due either to alteration or perhaps to fine grinding. Tourmaline, manandonite, serendibite, and bityite seem to be closely related species. The vanadiosilicate or arseniosilicate, ardennite, is a compound of quite different t5rpe from any so far considered. It is a mixed silicate Neues Jahrb., 1912, vol. 1, ref. 403. * Soc. min. Bull., vol. 31, p. 241, THE SILICATES OF ALUMINUM. 79 in which sometimes the vanadic radicle predominates and sometimes the arsemcal group, but the structure of the molecule is the same in both varieties. According to Prandtl ^ its empirical formula is HeMn5Al5(As,V)Si5028, which can be written structurally as follows: /SiO^^MnCAlO^H^) Al— Si04=MnH ^i04=Al— As04=Mn Al— Si04=MnH \siO,=Mn(A102H2) The vanadic ardennite is strictly isomorphous with the arsenical compound and in it V replaces As. Among the silicates of aluminum, salts of orthodisilicic acid are very rare. The only one which seems to be thoroughly defined is iolite, which agrees best with the formula ^i^O^-AlMg Al— SiA-AlMg ^iA.(A10H)2 Al— SiA-AlMg ^Si^O^-AlFe which requires the following percentage composition: SiOg 49. 26 AI2O3 33.50 MgO 9.85 FeO 5.91 H2O 1.48 100.00 in close concordance with the best recorded analyses. By alteration iolite passes into mica, going through an intermediate stage, however, known as chlorophyllite. This substance may be regarded as formed by hydration, in which the linking group of SizOy in iolite is split into two orthosilicic radicles, yielding two molecules of the type Al— SiAm^Mg \si2O =AlMg from which the final transition into a mica is easy. If we take Ram- melsberg's analysis of chlorophyllite, recalculate the ferric oxide into 1 Zeitschr. Kryst. Min., vol. 40, p. 392, 1905. 80 THE CONSTITUTION OF THE NATURAL SILICATES. alumina and lime into magnesia, reducing afterward to 100 per cent, we get the following comparison between observed fact and the com- position of chlorophyllite computed from the foregoing formula: Found. Reduced. Calculated. AI2O, Fe^O MgO. CaO. H2O. 46.31 25.17 10.99 10.91 .58 6.70 100. 66 47.99 33.34 11.74 6.93 100. 00 48.39 32.90 12.90 5.81 100.00 The agreement is as close as could be reasonably expected. The replacement of a little magnesia by a little water in the original altera- tion product accounts for the discrepancies. Two other magnesian alumosilicates, of rather uncertain character, may possibly be related to iolite; namely, the lasallite of G. FriedeP and the pilolite of Heddle.^ Both minerals are highly hydrous and are fairly represented by the subjoined formulae: Lasallite. Al— SiA-Hs Al— SiaO^.Hs ^Si^O^-AlH^ + 7H,0 PHolite. ^Si^O.-MgHe Al— SiA-Hs ^iA.Mg2 Al— SiA-Hs +7H20 \ Si,0,.MgH, The exact hydration is doubtful but not far from that shown in the formulae. There are a number of other questionable minerals recorded,^ of generally similar composition, but their consideration in detail would hardly be profitable. Some of them are probably mixtures of magnesian silicates with clays. Two more aluminous minerals are probably to be classed as ortho- disilicates. One barylite, is near ySijO^.AlBa Al— Si207.AlBa \si3O7.AlBa 1 Soc, min. Bull., vol. 24, p. 12, 1901; idem, vol. 30, p. 80, 1907. 2 Min. Mag., vol. 2, p. 206, 1878. 8 See Dana's System of mineralogy, 6th ed., pp. 705-711, 1892. THE SILICATES OF ALUMINUM. ' g^ which requires the following percentage composition: SiOa Al^O, 35.19 BaO. 19.94 44.87 100.00 The other silicate, sphenoclase, is approximately A\,(Sifi,) fi^^, which may be analogous to barylite in structure or written as a calcium salt similar m type to okenite. Both barylite and sphenockse, how- ever, are uncertain, and their relations are not definitely known Well-defined metasihcates of aluminum, or alumometasiHcates seem to be few in number. It has already been shown that many of the species classed as metasihcates are really mixed salts, trisihcates and orthosiUcates being commingled. With beryl, however, GlgAl^SieO^g, the evidence in favor of a meta- siHcate structure is fairly good, although the composition can also be expressed as that of a basic trisiUcate. There are thus two alternatives, Ai4i03^^^ \sio. Gl Gl and Gl C> But beryl alters into mica, a fact which is favorable to the first of these formulae, and all of its commoner alterations seem to take place by replacement of glucinum. In the trisiUcate formula the alumina should be equally replaceable, and so far the evidence is adverse to it. Furthermore, Traube ^ has effected the synthesis of beryl by precipi- tating a mixture of glucinum and aluminum sulphates with a solution of sodium metasiHcate, and then crystalUzing by fusion of the precipi- tate with boron trioxide. Since the starting point was a metasilicate, there is a fair presumption that the product was a metasilicate also. Beryl can be written as a pseudometasihcate, but there are no data to justify doing so. The mineral astroUte, described by Eeinisch ^ may perhaps be a metasiHcate. If so its formula is (Na,K)Fe''(AlFe"0 (Si03)5.H20, but the species needs further study. Spodumene and jadeito also have apparently metasiHcate ratios, but on mineralogic grounds it is best 1 Neues Jahrb,, 1884, vol. 1, p. 275. * Centralbl. Mineralogie, 1904, p. 108. 43633°— Bull. 588—14 6 82 THE CONSTITUTION OF THE NATUKAL SILICATES. to defer their consideration and to take them up in connection with the pyroxenes and amphiboles. The alumosiUcates derived from metadisihcic acid, HgSigOs, are few in number. Two of them, ptiloHte and mordenite, have already been discussed as zeolites; two others, petaUte and milarite, demand attention now. The formula of petaUte, empirically, is AlLi (81205)2, similar to the metasilicate formula for spodumene, and the two species are commonly associated. Petalite, however, has far the lower density of the two and is therefore presumably composed of smaller molecules. An alteration product of petaHte, hydrocastorite, approximates roughly AlCa ^8104^ Ca— CO3— Ca but with no evidence to go upon other than its composition. To the associated hillebrandite the formula Ca2Si04.H20 has been given, but the water in it may be constitutional. In that case hillebrandite becomes a basic metasihcate, (CaOH)2Si03. Molybdophylhte, PbMgSi04.H20, may be similarly constituted. The three orthorhombic species, bertrandite, calamine, and ilvaite, are most conveniently represented as derivatives of the bipolymer R4(Si04)2, hke chondrodite. Crystal lographically ilvaite resembles 1 See Wright, F. E., Am. Jour. Scl., 4th ser., vol. 26, p. 647, 1908. 90 THE CONSTITUTION OF THE NATURAL SILICATES. humite, but bertrandite and calainine are related to each other. The simplest analogous formulae for the three minerals are as follows : Bertrandite Gl ySiO,. Gl< >G1 H^ X^l— OH Calamine. Zn< >Zn II (ZnOH), Ilvaite. Ca Al— OH Calamine may also be written as a metasilicate, (ZnOH)2Si03, with half of the formula indicated above, but then the analogy with ber- trandite disappears. The structure proposed is therefore prefer- able, at least until more evidence has been accumulated. CHnohe- drite is equivalent to calamine, but with half the zinc replaced by calcium. Zamhonini ^ interprets calamine as a basic orthodisiHcate. Baschieri^ regards ilvaite also as an orthodisilicate. It is not easy to write constitutional formulae for these minerals and to show their relation to other species on that basis. Further investigation is evidently needed. To the datolite group a similar constitution is ascribable. The species, reduced to their simplest empirical expressions, are these: Datolite HCaBSiOg Homilite CasFeBaSigOio Euclase HGlAlSiOg Gadolinite GlgFeYgSiaOio By doubUng the formulae of datoHte and euclase all four of the minerals become similar in constitution. Hydrogen here is evi- dently basic, and boron must play the same part as aluminum and yttrium. Assuming these elements to be present in the univalent groups BO, AlO, and YO, the subjoined formulae follow: Datolite. Euclase. Hg H2 (BO), (AlO), 1 Contributo alio studio dei silicati idrati, 1908. 2 Zeitschr. Kryst. Min., vol. 49, p. 112, 1911 . SILICATES OF DYAD BASES. 91 Eomilite. Gadolinite. Fe Fe (BO), (yI)), Gadolinite alters with great ease, passing by hydration into a brownish-red earthy substance. The analyses of this substance, whicli is probably never pure and definite, are not altogether satis- factory, but they indicate in a general way a transformation into the compound Fe'^OH <:o;> II (YO,H,), to which the alteration product very roughly approximates. It will be observed that all of the foregoing structural formulae in this group of compounds are rings or series of rings. From them, however, chainlike molecules are derivable, and these seem to exist in friedelite, pyrosmalite, and dioptase. The last-named mineral, simply written, is CuH3Si04, but it is morphologically related to the two other species, which have a much greater complexity of composi- tion. The following expressions derived from the polymer RgCSiO^)^ are probably the best to represent existing evidence : Friedelite. Dioptase. /Si04=H2(MnCl) ySiO^^Ha Mn< Cu< \SiO,=H2 >SiO,=H2 Mn< Cu< \Si04=H2 ^SiO^^H^ Mn< Cu< \SiO,=HMn \Si04=HCu Pyrosmalite is like friedelite, but a large part of the manganese is replaced by iron. Possibly karyopilite may be similar, having the formula /SiO,=H2(MnOH) Mn< NSiO^^H^ Mn< \Si04=HMn 92 THE CONSTITUTION OF THE NATURAL SILICATES. These formulae are purely tentative and need additional support. By synthetic and genetic investigations they may be supported or overthrown. That they sustain one another and fit in well with the formulae of the preceding species is all that can be said in their favor. Palache ^ and Zambonini ^ assign more complex formulae to friedelite. The chlorine in that mineral seems to be partly replaced by hydroxyl. Bementite is still another silicate of similar t3^e. In accordance with the latest analysis by Steiger,^ its formula is simply written: /Si04=MnH Mn< >SiO,=H2 Mn< >iO,==H3 Mn< \SiO,=MnH For serpentine, H4Mg3Si209, several formulae are possible, and con- cerning them there has been much discussion. The species commonly originates in nature from the alteration of olivine .on the one hand and from pyroxene or amphibole on the other, and it is therefore conceiv- able that it may include two or more isomeric compounds. In favor of this supposition there is some evidence but nothing conclusive. Massive serpentine, chrysotile, antigorite, and other allied varieties differ in their physical properties and suggest the existence of isomers, but much more investigation is needed in order to decide whether the supposition is true or not. By some authorities serpentine is regarded as an orthosilicate and by others as a salt of the acid HeSigOy. On the latter supposition it becomes Mg=Si207^H2(MgOH)2, which may be derived either from 2Mg2Si04 or 2MgSi03, with loss of magnesium in one case and gain in the other. On the orthosilicate basis it is simply derivable from the polymer Mg4 (8104)2 and is related to the intermediate alteration product, villarsite, as follows: Mg^{SiO^)2- Villarsite. Serpentine. Mg Mg H, II II . If .Si04V .Si04. .Si04V Mg<; >Mg Mg< >Mg Mg<; >Mg \SiO/ ^8104/ \Si04^ Mg H Mg— OH H Mg— OH 1 Am. Jour. Sci., 4th ser., vol. 29, p. 177, 1910. » See Palache, idem, p. 182. 2 Contributo alio studio dei silicati idrati, 1908. SILICATES OF DYAD BASES. 93 On this scheme the formula for serpentine corresponds with that of chrondrodite; and the fact that the latter mineral alters readily into serpentine is strong evidence in its favor. In short, that formula best indicates the genetic relationships of serpentine, and on such grounds is preferable to the alternative diorthosihcate expression. The latter is not disproved; it is simply rendered less advantageous as regards existing evidence, which is the evidence now to be inter- preted. In some former investigations, carried on jointly with Dr. Schnei- der,i I sought to obtain experimental data in support of the ortho- siHcate formula here assigned to serpentine. By acting on ser- pentine with dry gaseous hydrochloric acid we found that a part of the magnesium could be removed as chloride, whereas olivine and the magnesian micas were not attacked. At first it seemed probable that the reaction would give a quantitative measure of the magne- sium combined as MgOH; but our later experiments and those of Lindner ^ have shown that the expectation was not well founded. I stiU beheve, however, that the reaction discriminates between those magnesium siUcates which contain MgOH and those which contain Mg and H combined othenvise, for only the members which must belong to the first class are acted upon by the reagent. Brauns's objections ^ to this supposition, on the ground that the dry hydro- chloric acid becomes moist, are not weU taken, for the reaction always took place at temperatures lower than those at which water is given off. His criticisms may apply to the later stages of the reaction, after it has once fairly begun, but not to its initiation. The magnesian micas which contain several per cent of water are aU decomposable by aqueous hydrochloric acid, but are scarcely touched by the dry gas; whereas, on the other hand, serpentine and the chlorites are strongly attacked. After the gaseous acid has acted it becomes moist, but very slowly, and most of the moisture is carried past the mineral under investigation before it has had time to pro- duce an appreciable effect. It is possible, however, that a slow stream of the acid may act differently from a rapid current, and that the discordant results of observation may be due to differences of this kind. When serpentine is ignited water is expelled, and a residue having the composition MggSigOy is left behind. According to Kammels- berg ^ the water is given off in two portions — one-half upon weak 1 U. S. Geol. Survey Bull. 78, p. 11, 1891; Bull. 90, p. 11, 1892; and Bull. 113, pp. 27 and 34, 1883. 2 A. Lindner, Inaugural Dissertation, Breslau, 1893. •' Neues Jahrb., 1894, vol. 1, p. 205. « Handbuch der Mineralchemie, 2d ed., p. 506, 1875. 94 THE CONSTITUTION OF THE NATURAL SILICATES. ignition, the other after heating more strongly. On the orthosili- cate theory these stages may be represented thus : Serpentine. First stage. Second stage. H^ \Mg— OH ^^ At the end of the second stage, if the ignition has not been too intense, the residue is still decomposable by hydrochloric acid, but by prolonged heating it is broken up quantitatively into soluble olivine and insoluble enstatite. Allied to serpentine is the somewhat doubtful picrosmine, to which the formula Mg2H2Si207 is commonly assigned. Although this expres- sion suggests a diortfhosilicate, it may also be written /SiO.,- SiO,- M< Mg i which represents picrosmine as a dehydrated serpentine altered sub- sequently by rehydration, with replacement of one magnesium atom by two of hydrogen. This mode of interpretation brings the mineral into line with serpentine, and all the known relations of the species are adequately expressed. Several other hydrous silicates seem to belong in this group, but their nature is altogether doubtful. Thus we have Aphrodite Mg2H4(Si04)2 Kerolite Mg(SiO,)2H5(MgOH) Nepouite, H4Ni3Si209, seems to be the nickel equivalent of ser- pentine. Garnierite and noumeite, other Jiydrous silicates of nickel, are too variable in composition to be definitely classed. METASILICATES. Although the metasilicates appear at first sight to be extremely simple, they are actually rather difficult to interpret. It is easy enough to deduce their empirical formulae and to write them after- wards in structural terms, but this is not sufficient. The structural formulae must express all known relations for each species, and in attempting to satisfy the established conditions the difficulties begin to appear. In the first place, metasilicic acid itself is defectively SILICATES OF DYAD BASES. 95 known, and no ester of the form RjSiOg h^s yet been certainly obtained. Troost and Hautefeuille's ester (CaHgJaSi^Oiz suggests the possibility that metasilicic acid, like metaphosphoric acid, may poly- merize, but an attempt to draw general conclusions on so important a question from one datum only would be most unwise. The possi- bility of polymeric acids, however, was clearly shown in the general discussion of the silicic acids in Chapter II of this bulletin. It is also emphasized by the existence of four distinct modifications of mag- nesium metasilicate, as proved by Allen, Wright, and Clement.* Whether these modifications represent different metasilicic acids or not is yet to be discovered. Again, as we have repeatedly seen, a mineral m,ay be apparently a metasilicate and yet really a mixture of orthosilicates and trisilicates. Even a basic trisilicate can have seemingly metasilicate ratios. All these considerations complicate the identification and study of thp true metasilicates to such an extent that only provisional conclusions can be drawn from the data now on hand. A crystallized silicate of sodium, NagSiOa.SHgO, is well known. A solution of this salt added to a solution of calcium chloride precipi- tates a compound which, dried over sulphuric acid, has, according to my own observations, the composition CajSigOg.SHgO. This, minus the water, is analogous to the mineral wollastonite, from which another mineral, pectolite, is derived. If wollastonite, instead of the formula CajSijOe be given the formula CagSigO^, it may be compared structur- ally with pectolite, as follows: Wollastonite. Ca< \ NSiOg >Ca Ca< / It must be remembered that the molecular weights of the inorganic silicates are not known but only assumed; and the problem suggested by the foregoing expressions is to find a set of structural formulae which shall represent all the available evidence. Now, wollastonite is commonly classed with the pyroxenes, on crystallographic grounds; and so too is pectolite. But both species are very easily decomposed by even dilute and weak acids, whereas the normal pyroxenes are quite refractory, and, furthermore, wollastonite has a lower density than any pyroxene which approaches it in composition. Chemically, then, these species are dissimilar, and it is very doubtful whether they can properly be grouped together. That wollastonite and pectolite are true metasilicates, however, is sustained by the fact that when 1 Am. Jour. Sci., 4th ser., vol. 22, p. 385, 1906. 96 THE CONSTITUTION OF THE NATURAL SILICATES. pectolite is ignited one-sixth of the silica, proportional to the hydrogen of the mineral, is split off in the free state and can be determined quan- titatively. That is, two molecules of pectolite are decomposed with liberation of metasilicic acid, HjSiOg, which, in turn, divides into SiOz + HjO. In some varieties of pectolite manganese replaces part of the calcium, and schizolite is a mineral of similar composition but with different ratios, its formula approximating H3Na3Ca3Mn2(Si03)8. Schizolite, however, may be an isomorphous mixture of two silicates and not a simple compound. It is in no true sense equivalent to pectolite. Empirically the nonaluminous pyroxenes resemble wollastonite in their ratios. Thus we have, according to the commonly accepted formulae developed by Tschermak, Doelter, and others : Enstatite, orthorhombic MgaSigOg Diopside, monoclinic CaMgSigOe Hedenbei^ite, monoclinic CaFeSiaOg Rhodonite, trlclinic MngSigOg There is also a great number of other intermediate species or isomor- phous mixtures in the pyroxene series, such as bronzite, hypersthene, schefferite, sahlite, jeffersonite, and fowlerite, in which we find, variously replacing one another, salts of magnesium, calcium, iron^ manganese, and zinc. All these minerals, however, conform to the general formula ESiOg, or 112^1206, which adequately expresses their constitution so far as they alone are concerned. This formula can be written structurally — Si03 SiO r/ >R > which would be satisfactory if the pyroxene series ended here and if the amphiboles were unknown. Going a step further we find in augite a pyroxene containing aluminum and having an oxygen ratio greater than in the group SiOg. In place of aluminum ferric iron also occurs, and alkalies are sometimes present. Leaving these variations out of account, for consideration later, we have in augite, as interpreted by Tschermak, together with the normal compound 112^1206, the basic salt RAlgSiOg. This substance, however, is not known by itself, unless it is repre- sented by kornerupine, although the latest analysis of that mineral tends to negative the supposition. An artificial silicate of similar type, K2Al2Si06 has been prepared by Weyberg,^ and may serve to strengthen Tschermak' s assumption. This salt constitutionally may be regarded as a basic orthosilicate — 1 Centralbl, Mineralogie, 1911, p. 326. SILICATES OF DYAD BASES. 97 /Al— 0~K SiO,< \.A1— O— K and its magnesian equivalent as ^Al-0. but neither expression is to be taken as final. The excess of oxygen in the pyroxenes may be due, with equal probability, to the equiva- lent group Al< _ >Mg which has been assumed in preceding sections of this memoir. Still another series of silicates containing triad bases and also alka- lies are classed with the pyroxenes, as follows : Spodumene AlLiSiaOg Jadeite AlNaSiaOg Acmite .Fe^^^NaSiaOg and their empirical formulae are fairly satisfactory. Structurally these expressions become, as metasilicates, \SiO3— R' and babingtonite, which contains no alkalies, is similar, thus: ^SiOg SiO, Fe" + 3 R" /SiO, C>" Fe-Y ^SiO E,'' being =Ca, Fe'', and Mn. The ferric molecule is evidently equivalent to two' acmite molecules, with Nag replaced by a linkmg atom of iron. So far, except partially in the comparison between wollastomte and pectolite, the formulae cited for the pyroxenes express composi- tion and composition only. But spodumene, as shown by the elaborate research of Brush and Dana, spHts up on alteration into a 43633°— Bull. 588—14 7 98 THE CONSTITUTION OF THE NATURAL SILICATES. mixture of eucryptite, an orthosalt, and albite, a trisilicate. This observation suggests two alternatives: Either that spodumene is derived from a polymetasilicic acid, or else that it is a pseudometa- silicate, a mixed ortho- and tri-salt, like some of the species which have already been explained. An analogy with leucite, for example, will at once be inferred, and that species, empirically, is strikingly like spodumene, thus : Leucite AlKSiaOg Spodumene AlLiSi^Oo Like spodumene, leucite alters into a feldspar and a member of the nepheline group, but it differs from spodumene in form and in density. The specific gravity of the isometric leucite is 2.5, that of the monoclinic spodumene is nearly 3.2, and hence we may reasonably infer that spodumene has the larger and more condensed molecule. In order to explain the relations of leucite, its empirical formula was quadrupled, and in that way a relation with the garnet group was brought out. For spodumene, regarding it also as a mixed siHcate, a sixfold multipUcation of its formula indicates its greater density, and its splitting up into eucryptite and albite, with partial replace- ment of lithium by sodium, is representable as follows: 1^2 Li /SigOs— Al=Si04\ A;feSi308— Al< >A1— Si04=Al XSigOs^Al— SiO/ I. I! Li Lig From this grouping of atoms the transition into Al3(Si308)3Na3 + AI3 (Si04)3Li3 is hardly more than simple cleavage, and the relations between the three species are intelligibly expressed. Acmite, which yields pseudomorphs of analcite, and jadeite, also, probably follow the same rule, the formula of one being typical of the others. The ferric molecule in babingtonite should be still another instance of the same kind, with Fe''3 in place of Lig, and Fe'^' instead of Al. Pos- sibly pyrophyUite, HAlSijOg, is related to spodumene and jadeite, much as kaoHnite is related to the feldspars. PyrophyUite is prob- ably a pseudometasilicate, for silica is not liberated from it upon ignition, at least not to any noteworthy extent. If the formula just developed for spodumene should be sustained, it would seem necessary to adjust the other pyroxenes with it. For Tschermak's aluminous constituent of augite this adjustment is easily made by taking the formula AlaMgSiOe six times, as was done SILICATES OF DYAD BASES. 99 for spodumene. The paralleUsm between the two species is then representable as follows : Spodumene. . . Al,(Si30«)3(SiO,)3Li« Alummum-augite Alo(SiO,)3(SiO,)3(A102Mg), the univalent A102Mg having been recognized among the micas. This formula serves to explain the weU-known alterabihty of augite into epidote and into mica, and so far at least is useful. I do not, however, feel incHned to put very much stress upon it, for as yet it i^ only an expression of analogy, which may or may not prove to.be valid. It would seem to require the recognition of all the pyroxenes as pseudometasiiicates, in which case the normal series, containmg only dyad bases, would become R R II II /SigOs— R— SiO,V R< >R \Si3O— R— SiO/ I R where four atoms of R are given linking functions, and the other four are, so to speak, replaceably combined. On this basis we should write Diopside Mg4Ca4(Si308)2(Si04)2 Hedenbergite Fe4Ca4(Si308)2(Si04)2 The formula R "4 (AlOg (8104)2 (Si04)2 would be exactly parallel with these, and affords another expression for Tschermak's compound, Al2RSi06. Or the ground of simplicity this is preferable to the more complex expression based on the formula of spodumene. If to mon- ticelUte we assign the quadrupled formula Mg4Ca4 (8104)4, and to for- sterite the similar formula Mgg (8104)4, diopside becomes equivalent to them in structure, with one-half the orthosilicic radicle replaced by the trisiUcic SigOg. The pyroxenes and the oUvines thus appear to be curiously related compounds, although they are unlike morpho- logically. This mode of interpreting the pyroxenes is so remote from our usual conceptions that I offer it with great diffidence. It unifies the group, however; it expresses the observed alterations of the several species; and despite its complexity it will be found to be sustained and strengthened by evidence brought out in the study of the amphiboles. This last-named group of highly important minerals resembles the P3rroxenes in composition and is explained by Tschermak in essen- tially the same way. Their mJleiaii^' weights, ^howeyei-, are taken 100 THE CONSTITUTION OF THE NATURAL SILICATES. as double those of the pyroxenes, for the reason that the atomic replacements seem to occur by fourths rather than by halves. This point is exemplified by a comparison between diopside and tremoHte, which, reduced to their simplest empirical formulae, become Diopside CaMgSigOe Tremolite CaMg3Si40i2 The pyroxenes, however, are somewhat heavier than the amphiboles and from their greater density we may suppose them to have the larger molecules. Hence the formula of diopside should be a multiple of that just cited and presumably greater than Ca2Mg2Si40i2. Upon this point the phenomenon of urahtization has definite bearing. In this process pyroxene is converted into amphibole, with increase of volume and little or no change of composition. In other words, a complex molecule has been dissociated into simpler molecules — a phenomenon the direct opposite of polymerization. In the face of this evidence it is difficult to see how the current views as to the relative molecular magnitudes of pyroxene and amphibole can be maintained. The pyroxenes must form the more complex group and the amphiboles the simpler. In the amphibole group the orthorhombic anthophylHte is the equivalent or isomer of enstatite and hypersthene. Then follows a monoclinic series, containing tremolite, actinolite, cummingtonite, daimemorite, and other minerals, all represented by the general empirical formula RSiOg, with calcium, magnesium, iron, or man- ganese as the bivalent metal. In griinerite the salt FeSiOg exists by itself, and in richterite and astochite alkaline siUcates appear. If we regard the minerals as pseudometasiHcates, having molecular weights lower than the pyroxenes and with the bases replaceable by fourths, the typical amphiboles are most simply represented by formulae like the following: AnthophylHte. Tremolite. Mg Mg II II Mg< >Mg Mg< >Mg Mg Ca Richterite becomes a mixture of salts, R R /SiO,. /SiO,\ R< >R + R< >R \Si,0/ \Si3O, SILICATES OF DYAD BASES. 101 commingled in ratios near 1 : 1 , and astochite is similar, but with NaH in place of Na2. Potassium may also partly replace sodium. Another alkaline amphibole of doubtful character, waldheimite, approximates R .A i which is the formula of a trisiHcate pure and simple, with R = Ca, Fe Mg. The existence of this compound is strong evidence in favor of the pseudometasihcate theory, and, as will be seen later, it does not stand alone. Among the amphiboles, as among the pyroxenes, aluminous and ferric compounds are common, and with these the minerals approach orthosihcate ratios. Tschermak's interpretation of these ratios is practically the same as in the pyroxene series, namely, by the assump- tion of molecules of the form AlgRSiOg or Al^RgSijOij. An alterna- tive to this view is offered by Scharizer,^ who shows that the horn- blendes can be explained as roixtures of actinolite, R4Si40i2, with an orthosihcate called syntagmatite (R'2R'03Al2(SiO4)3, whose ratios are similar to those of garnet. An amphibole from Jan Mayen Island approaches very nearly to syntagmatite in composition, and so also does the Canadian hastingsite. If a compound of this type is present in the amphiboles it would explain at once their alterabihty into epidote, micas, and chlorites, but so far as the composition of the group is concerned neither Tschermak's view nor Scharizer's is abso- lutely necessary. The Tschermakian molecule, however, can be written either as A102Mg (A10)2 /SiO,. \sio/ ^io/ A102Mg (A10)2 the latter form resembling that of tremoHte, and also connecting the group still more closely with the oh vines. It is also parallel to the last formula suggested for the corresponding pyroxene compound, being one-half of that formula and identical with it m type. No amphibole is yet known which corresponds precisely in consti- tution to acmite and spodumene. In glaucophane we find a species, M{ >A1 or as 1 Neues Jahrb., 1884, vol. 2, p. 143. 102 THE CONSTITUTION OF THE NATURAL SILICATES. which^ as a metasilicate, may be written AlNaSijOe + (MgFe) SiOg, and in crocidolite we find anotlier similar salt; Fe'^'NaSisOe + FeSiOg. Crocidolite alters easily, and one of the products of alteration, which has been named griqualandite,- is very near Fe^'HSigOg, the equiva- lent of acmite in general type. This last compound can be written H ^SiO,. ■ i and so adjusted as an amphibole-hydrogen-acmite to the remainder of the group, but glaucophane and crocidolite are best formulated as follows : Glaucophane. Crocidolite. (AIO), (Fe"'0), .11 , .11 Mg Fe< >Fe Si,0/ \Si,0/ ^SigOg Ka^ Na, which makes them, as trisilicates, precisely equivalent in structure to the normal amphiboles. These compounds, and their corresponding orthosilicates, commingled with salts Hke tremolite or actinolite, give mixtures which conform in composition to the aluminous hornblendes. Rhodusite, which is alUed to glaucophane, appears to be a mixed silicate of the following constitution: Na2 (Fe'''0)2 .11 .11 2 Fe< >Mg + 1 Mg< >Mg ^SigOs^ ^Sifi/ Mg - Mg In rhodusite the ratio of Si to O is distinctly less than 1 to 3, and the same is true of crossite, which may be written as a mixture of the two molecules R (Fe-'O), >R + iR<: >] i,0/ VgO/ Naj Nag with some Al in place of Fe, and R being = Fe, Mg, Ca. SILICATES OF DYAD BASES. 10^ Arfvedsonite, in which W is mainly Fe'', is represented quite closely as a mixture of this order: Fe Ca Fe 2 Fe< ' '\Fe + 4 Fe< ' "\Fe + 6 Fe/^''^'\Fe \sio/ \sio/ ^ao/ (AlO)^ Na^ . Na, Hastingsite, soretite, and philipstadite may all be formulated in a similar way, except that they tend toward the orthosilicate end of the series. In philipstadite SigOg and Si04 appear in equal propor- tions; in the two other amphiboles SiO^ is largely in excess. Gas- taldite and riebeckite are similar species, but the analyses are not perfectly conclusive. It is, perhaps, necessary to assume the presence in these minerals of acmite-Uke molecules, riebeckite being empirically near 2 Fe'^NaSigOg + FeSiOg. The formula of barkevikite is also uncertain. Several minerals belonging to the group of amphiboles are char- acterized by the presence in them of noteworthy quantities of titan- ium. If this is assumed to represent an orthotitanate, the formulae all reduce to expressions exactly similar to those given in the preceding pages. For example, the type mineral of this group is senigmatite, to which the following formula may be assigned: Fe Fe Fe /TiO.v /SiO,\ ySisOg. I' I i in which W is f Na and f (ALFe''0O. Cossyrite is a similar mixture of molecules, and very near 10 Fe3(Ti04)2Na2 7 Fe3(Si30s)3(Fe'''0)3 10 Fe3(Si308)2Na2 with the usual variations due to the replacement of Fe by Ca, Mg, etc., of Na by K, etc. 104 THE CONSTITUTION OF THE NATURAL SILICATES. The other titaniferous amphiboles are also evidently variable mixtures, which, in the best analyses, are representable as follows: Linosite. Kxrsutite. 1 Ca^FeCTiOJ^Na^ 2 CaFe^CTiOJ^Na^ 1 Ca^MgCSiOJ^CFe'^'O^ 2 Ca^MgCSiOJ^HCAlO) 2 Mg3(Si303)2(A10), 3 Mg3(Si303)2(A10)3 Anaphorite. Rhonite. 1 R3(TiO,)2Na2 2 Mg,Ca(TiOj2(RO)2 3 R3(Si30s)2Na2 5 Mg^CaCSiOJ^CRO)^ 1 R3(Si303)2(Fe'''0)2 5 Fe2Ca(SiOj2(RO)3 In anaphorite R=Ca, Fe, Mg, in the ratio 1 : 2 : 5; and in rhonite RO is about § AlO to J Fe'^'O. Rhonite represents the orthosilicate end of the series ; anaphorite shows the nearest approach to trisilicate ratios. It is evident that an indefinite number of similar molecular mixtures are possible, but all are likely to be of the same general type. A careful study of the best analyses in the pyroxene and amphibole groups will strengthen very materially the view here developed that the species are not true metasilicates. Although in most of the analyses the approximation to metasilicate ratios is very close, there are distinct variations toward orthosilicates on one side and toward trisilicates on the other, and it is only by assuming that we have mixed silicates to deal with that all the anomalies can be made to disappear.^ On this theory, if we represent SigOg, SiO^, and Ti04 groups indiscriminately by the general symbol X, all the amphiboles are covered by the following typical symbols, in which R" stands for any dyad metal, and R' for K, Na, H, AlO, or Fe'"0: R''A R"3X,R', or, structurally, R"AR', R" A II R" A E'. 1 II r'^' R', A In a similar way all the pyroxenes, except the acmite-spodumene group, which has the special formula discussed previously, may be represented as formed by mixtures of R^^X^R'', R",X,(R-'0)s 1 This view'of the constitution of the pyroxenes and amphiboles was first advanced by G. F. Becker, Am. Jour. Sci., 3d ser., vol. 38, p. 154, 1889. SILICATES OF DYAD BASES. 105 which is i^ accordance with the theory developed by Tschermak except as to the molecular magnitude of the compounds— that is, the pyroxenes are essentially bipolymers of the amphiboles, and the character of the structure is the same for both groups. The analogy between these formulae and those of the oUvines has already been pointed out, and it is emphasized by still more evidence. Pseudo- morphs of pjrroxene (fassaite) after monticellite have been found at Monzoni and are well known. Furthermore, Becke ^ has described pseudomorphs of anthophyllite and actinolite after olivine, so that a connection between the two groups is clearly indicated. The tracing of this connection in a more general way would seem to offer a profit- able field for investigation. Many amphiboles contain water and some contain fluorine. These constituents are easily accounted for, being present either as univalent radicles like AlOgHj and AIF2, or with H replacing Na. Allen and Clement,^ however, in their study of tremolite found that the mineral contained water which could be expelled continuously and therefore behaved as if it were not constitutional but in "solid solution." A study of their analyses leads to some doubt as to this conclusion. The water as determined is needed to completely satisfy the silica, and so to give rational formulae to the different samples of tremoUte which they studied. Additional investigation is plainly needed here. Another view of the amphiboles has been developed by Penfield and Stanley,^ who assume the presence in them of such bivalent groups as Al^OF^, AipCOH)^, AlA^^'', and Al^O^R^Na^, and also the univalent group MgF. With the aid of such assumptions they are able to formulate all the amphiboles as metasificates. Such an interpretation of these minerals is evidently more complicated than the formulation adopted here, and it does not provide for the varieties in which the ratio Si to O is below the metasilicate requirements. By the hydration of pyroxene or amphibole either serpentine or talc may be generated. Talc has the composition HaMggSiPia, and may be written structurally like amphibole either Mg ^ . H, A A Both expressions are in accord with the fact recorded by Schneider and myself,^ that upon the ignition of talc one-fourth of the silica is 1 Min. pet. Mitt., vol. 4, pp. 355, 450, 188l-«2. 2 Am. Jour. Sci., 4tli ser., vol. 26, p. 101, 1908. a Am. Jour. Sci., 4th ser., vol. 23, p. 23, 1907. 4 U. S. Geol. Survey Bull. 78, p. 13, 1891. 106 THE CONSTITUTION OP THE NATUEAL SILICATES. set free quantitatively. This would give the ignited residue the com- position shown by the subjoined alternative formute: Mg II /SiO,. ^O Mg^ >Mg and Mg\ >Mg . \Si,0/ ^hO/ II Mg and of these the first would seem to accord the better with the remark- able stability and insolubility of the material. A metasilicate for- mula, H2Mg3(Si03)4, is also admissible, and accords equally well with the evidence concerning talc. The pseudometasilicate expression, however, seems to be preferable in view of what is known as to the genesis of the species. The nickel silicate, alipite, HgNig (8103)3, also has metasilicate ratios, but there is no further evidence as to its constitution. ChrysocoUa is probably a metasilicate, and perhaps empirically CuSi03.2H20. It can not be well regarded as impure dioptase, for that mineral gelatinizes with hydrochloric acid, whereas chrysocolla does not. The species, which may be a mixture of compounds, needs careful investigation. The same is true of plancheite, to which Lacroix^ assigns the complex formula HgCuyCCuOH) 8(8103)13. Too little is known of this species to admit of any more definite formu- lation. The lead silicate alamosite, Pb8i03, is analogous to wollastonite and is therefore, in all probability, a true metasilicate. Agnolite, HaMug (8103)4.1120, resembles talc in its ratios, and may also be classed here. Leucophanite, NaCaGlF8i206, is another definite species which is easily figured thus : /SiOg— Gl— F /8i03— Ca— F Ca< or Gl< \8iO3— Na \8iO3— Na but between the two alternatives there is no way of deciding. The mineral hillebrandite was interpreted by its discoverer, Wright,^ as an orthosilicate of calcium, Ca28i04.H20. The water, however, is not given off at low temperatures, and is probably con- stitutional. If so, the species is a basic metasilicate, /Ca— OH OH 8i03< 1 Compt. Rend., vol. 146, p. 722, 1908; Soc. min. Bull., vol. 31, p. 250, 1908. 2 Am, Jour. Sci., 4th ser., vol. 26, p. 551, 1908. SILICATES OF DYAD BASES. 107 This conclusion is borne out by the fact that the mineral gives an immediate and deep rose color when its powder is moistened with phenolphthalein solution. That reaction is suggestive of the alkahne group CaOH, and a new fluosihcate, custerite, recently discovered by Umpleby and SchaUer, of the Geological Survey, gives the same coloration. Custerite is represented by the formula /Ca— F Si03< x:;a— OH which is analogous to that proposed for hillebrandite. Several other species are possibly metasilicates, although the evi- dence is not sufficient to warrant a definite conclusion. They are empirically — Weinbergerite NaFe^^gAlSi^Oig Spodiophyllite (Na,K)2(Mg,Fe)3(Al,Fe)2Si8024 TaramelUte Ba4Fe^^Fe^^^4Siio03i PhoHdohte, a hydrous magnesian sihcate, containing some potassium and aluminum also, seems to admit of no simple and satisfactory formula. DISILICATES AND TRISILICATES. Although the existence of the sexbasic acid H6Si207 has been well estabhshed by the preparation of its esters, its metallic salts are Httle known and uncertain. I have already shown (see discussion of serpentine, p. 92) that a mineral may be apparently an orthodisilicate and yet be equally well explainable otherwise; and what is true for that mineral may be true for others. For the following species the orthodisihcate formulae seem to be the best and simplest, even, though they are not wholly free from objection. They fit existing evidence but are not absolutely conclusive. The typical member of this group of minerals is the hexagonal lead silicate, barysihte, PbgSiaO^. The artificial compound from the slags of Bonneterre, Missouri, described by Dana and Penfield, is near this in composition, and may be PbaCaSiaOy. Hardystonite, Ca2ZnSi207, is also typical. The pseudodeweylite of Zambonini,^ which has the formula Mg3Si207.3H20, is another compound of this type. The formulae assigned to these species are merely their simplest empirical expressions. The true molecular weights are not known nor is there any available evidence to show whether or no the formulae should be doul)led or tripled. Two other lead-calcium silicates appear to belong here. Nasonite, described by Penfield and Warren,^ contains ch lorine, and ganomalite 1 Contributo alio studio dei silicati idrati, p. 88, 1908. 2 Am. Jour. Sci., 4tli ser., vol. 8, p. 346, 1899. 108 THE CONSTITUTION OF THE NATURAL SILICATES. is regarded by them as the corresponding hydroxyl compound. The formulae are simply written as follows : Nasonite. Ganomalite. /Si^O^.Ca^CPbCl) ySiA-Ca^CPbOH) Pb< Pb< >iA.Pb2 >i207.Pb2 Pb< Pb< \SiA-Ca2(Pba) \SiA-Ca2(PbOH) The univalent groups — Pb — CI and — Pb — OH are similar to other groups which have been assumed elsewhere. The group of zeolitic calcium silicates, okenite, gyrolite, and apo- phylUte, are unquestionably related to one another and are best represented as salts of IleSiaOy. In nature gyrolite may be derived from apophyllite and apophyllite also from gyrolite, and Doelter has generated apophyllite from okenite by artificial means.^ First, by heating okenite with potassium silicate and water to 200°, crystals of apophylhte were obtained. Secondly, by heating okenite with aluminum chloride, sodium carbonate, and carbonated water at 220°, apophyllite, analcite, and chabazite were produced. The most satis- factory general formulae for the three species are these: Okenite. Gyrolite. Apophyllite. /SiA-Hs /Si^O^.CaHg /Si207.H4(CaOH) Ca< Ca< Ca< ^Si^O^.CaH^ >Si207.H4 ^Si^O^.H^ Ca< . Ca< Ca< ^Si^O^.Hs \Si2O7.CaH3 \Si207.H,(CaOH) In apophyllite fluorine may partly replace hydroxyl, and K may replace the univalent CaOH. With K : CaOH : : 1 : 1 the composition of apophyllite becomes SiOg: - 52.03 CaO 24. 27 K2O 6. 79 H2O 16. 91 100. 00 The uncertain mineral plombierite may be a fourth member of this group, with the formula .Si^O.-CaHa Ca< >SiA-Ca2 +9H2O Ca'^ \si2O7.CaH3 1 Neues Jahrb., 1890, vol. 1, p. 118. SILICATES OF DYAD BASES. 109 To the calcium-manganese silicate, inesite, various formulse are assignable. By Flink it is regarded as 2(CaMn)Si03.H20. But part of the water is stable at temperatures above 300°, and this fact is expressed by Schneider's formula (CaMn)Si308(MnOH)2.H20, Both formulae agree with the analyses approximately, but the analysis by Lundell is better represented by the following mixture: /SigOy.CaHg Ca< Mn< ^Si^O^.Ca^ +2H2O, and Ca< Mn< \Si2O7.CaH3 ioO^.MnH, 5i207.Mn2 + 2IL0 ^iaOy.MnHg which requires the following percentage composition: Found, Lundell. Calculated. SiOo 42.92 .73 36.31 .37 8.68 10.48 42 18 PbO 1 37.44 1 9.84 10.54 MnO .. MffO... CaO H2O 99.49 ■ 100.00 To inesite from Mexico, which contains less water than is shown above, Farrington ^ assigns the formula HaC^In, Ca)6Si60i9.3H20. This formula is difficult to represent constitutionally and does not fit the analysis as sharply as is to be desired.^ A similar structure probably belongs to the magnesian spadaite, as follows: .Si20,.MgH3 Mg< >Si207.MgH2 Mg< \Si20,MgH3 Another magnesian silicate, saponite, is perhaps normally H4(MgOH)3 Si207, although the analyses all show admixtures of some aluminous compound. To cuspidinc, a calcium fluosilicate, Zambonini ^ assigns the sub- I'oined formula: , _ ^ /Ca— F 1 Field Columbian Mus. Bull., Geol. ser., vol. 1, p. 221, 1900. 2 See also Zambonini, loc. cit. » Mineralogia vesuviana, p. 273, 1910. 110 THE CONSTITUTION OF THE NATURAL SILICATES. An analysis of cuspidine from Franklin, New Jersey, by Warren leads to the same formula, although Palache ^ interprets it differently. Zambonini, following Groth, regards bertrandite as similar, and assigns to it the formula Gl^Sip^.CGlOH)^. In accordance with Hillebrand's analysis,^ the rare mineral row- landite should have an orthodisilicate formula as follows: Y— F /Sip,=Y Fe< \Si207=Y Y— F which expresses the composition of the species very closely. Minerals corresponding to metadisilicic acid, HaSisOg, are rare. A few of them have already been mentioned among the alumosilicates, but only two belong here. Rivaite, described by Zambonini,^ is a metadisihcate of calcium and sodium and is probably to be figured ^iaOg— Na which suggests a relationship withpectohte. The rare mineral hyalotekite, which contains boron, agrees very closely with the empirical formula R^BFSieOjy, if we regard the water in it as hydroxyl replacing fluorine. This can be interpreted either as a metadisihcate or as a trisiHcatC; with the following alternative formulae : NSiPs or Il< yR \Si2O — R— F 0=B/ \R— F Its association with feldspar and schefferite rather favors the trisili- cate formula, but the two are empirically identical. If we reduce 1 Am. Jour. Sci., 4th ser., vol. 24, p. 185, 1907 3 Appendice alia mineralogia vesuviana, p. 16. 2 U. S. Geol. Survey Bull. 113, p. 45, 1893. SILICATES OF DYAD BASES. Ill Lindstrom's analysis to 100 per cent, after calculating the water (ignition) into its equivalent of fluorine, rejecting as impurities the traces of Mfi, and Fefi,, and consolidating hke bases, we get the following comparison with theory: Found. Reduced. Calculated. SiOa 39.47 3.73 25.11 .09 .29 .75 .17 7.82 20.08 .09 .89 .18 .06 .99 .06 .59 38.60 3.63 1 25. 71 J- 9.43 1 21. 32 . B2O3 38. 10 pbo : ■••■ 3.71 26.22 CuO MnO GIO • Na20 CaO 9.22 21.5^ BaO MgO k;o :: A1203 FeA F.!... 2.26 CI 2.01 Ignition LessO 100.37 100. 95 .95 100. 85 .85 100.00 100.00 In computing, R" has been regarded as Ca : Ba : Pb : : 7 : 6 : 5; that is, hyalotekite is a mixture of isomorphous calcium, barium, and lead salts in the indicated ratio. The agreement between analysis and theory is as close as could be reasonably expected. Salts of the octobasic orthotrisilicic acid HgSigOio seem to be few in number, at least so far as present evidence goes. Two of them are silicates of nickel, namely, Connarite HaNijSiaOio Genthite MgaNiaSiaOio.eHjO Deweylite, according to Zambonini,^ is equivalent to genthite, its formula being MgSi30io.5-6H20. He shows that the so-called dewey- lite reaUy represents two species, one having the formula just given, the other, pseudodeweylite, being an orthodisilicate. Pseudodewey- Hte has already been considered. The complex fluosihcate, meliphanite, is also probably an orthotri- silicate, although other formulae have been proposed for it. Zeophyl- 1 Contributo alio studio dei silicati idrati, p. 88, 1908. 112 THE CONSTITUTION OF THE NATURAL SILICATES. lite, which is in some respects aUied to apophylUte, belongs in this class, and the two minerals appear to be similar in structure, thus: Meliphanite. /O— Na Si— O. Si i \Ca Zeophijllite. Si— Ov I /O— Ca— F Si< +H,0 I \0— Ca— F Gl sko>- \, Si— o. O— Ca— F O— H Trisilicates of the form H^SigOg are numerous in the mineral kingdom and are especially represented by the alkali feldspars and their derivatives. They are common in isomorphous admixture with orthosilicates, forming the minerals which I have classed as pseudo- metasiUcates, and only two remain to be noted here. These two silicates, eudidymite and epididymite, have the same empirical for- mtda, HNaGlSigOg, which, doubled, becomes Na2 /Si.O, Gl ^.. %g: Si,0/ 1 a structure conformable to the type of several orthosilicates. The isomerism between eudidymite and epididymite is explainable by giving one species the constitution just written, whereas the other, in place of Na2 and H2, would have the two groups NaH. It can also be ascribed to a different linking with the oxygen of the acid, and the empirical formula HNaGlSigOg can be put in two forms, thus: 0=Si— O— Na Si— O 0=Si— O— H and o r°- -Na 0=Si— O— H a third isomer being also conceivable. The doubled formula, however, brings out analogies with bertrandite and other species and therefore, in default of evidence, is to be preferred. CHAPTER V. SIIilCATES OF TETRAD BASES, TITANO SILICATES, AND COLUMBOSILICATES. On account of their relatively small number and general scarcity, the sihcates containing the tetrad metals, tinanium, zirconium, and thorium, are difficult to interpret in any satisfactory manner. Evi- dence exists, however, which seems to show that they are explainable by the same principles which apply to aluminum and the dyads, and that the theory of substitution from normal salts is a good working hypothesis to start from. One definite normal salt is known in this series, the mineral zircon, ZrSiO^. As with the other inorganic sihcates, the true molecular weight of this compound is unknown, and it can be inferred only from a study of its derivatives. If we assume it to be represented by the polymeric expression Zr4(Si04)4, it contains replaceable basic atoms, and a number of other zirconium sihcates fall naturally into series derivable from this as the fundamental member. In this connection the mineral auerbachite is pecuharly suggestive, for its composition is best indicated by the formula Zr4(Si308)(Si04)3, which goes to show an important analogy between this group of silicates and those which have been previously considered. This formula, compared with Hermann's analysis of auerbachite, gives the foUowing results: Hermann. Calculated. SiOo 42.91 55.18 .93 .95 42.45 ZrOa 57.55 FeO HgO 99.97 100.00 Although zircon is a very stable and defuiite mineral, it alters by hydration into malacone, cyrtohte, and a variety of other indefinite substances which can not as yet be interpreted with any clearness. At the same time other bases, such as lime and the rare earths, are taken up, producing mixtures of great complexity. Malacone 43633°— Bull. 588—14 8 113 is 114 THE CONSTITUTION OF THE NATURAL SILICATES. probably the first hydration derivative, and its relations to zircon, regarding the latter as Zr4(Si04)4, may possibly be as follows: Zircon. Malacone. Si04=Zr OH /siO / /siO ~^ Zr<^.^^^Zr Zr< ^^Zr V S1O4. \ SlO^K Si04=Zr Si04=Zr The original cyrtohte from Rockport is near malacone, but the cyrtolite from Colorado, analyzed by Hillebrand, approximates to Zr OH OH SiO^^HCZrO^H^) Si04=H(Zr02H2) with part of the ZrOgHg replaced by other bases. A number of other altered zircons or derivatives of zircon have received specific names, but their nature is more or less doubtful. Two of them, however, are interesting and may represent distinct species. The anderbergite of Blomstrand, for instance, is very near 'SiO.=Zr— OH ^ /Si04=Y .Si04=Zr— OH and the alvite of Nordenskiold, analyzed by Lindstrom, is well repre- sented as a molecular mixture of two compounds, Zr(SiOj4(ZrOH)4 + Zr(SiOj4(G10H)i2. The alvite of Forbes and Dahll seems to be quite different, but the analysis of it is unsatisfactory. In Lindstrom' s analysis one ZrOH group is replaced by R'", mainly Fe, Y, and Ce, and a little GIOH is replaced by Ca. Many altered zircons, like these minerals, contain rare earths in greater or less proportion, and this fact counts heavily against a common interpretation of zircon as a mixed oxide having the formula ZrOg 4 SiOj, or 0=:Zr/ \si=0 This view, which is prevalent in Germany, is based upon the close morphologic relations between zircon, rutile, and cassiterite. The SILICATES OF TETRAD BASES. 115 chemical relations, however, are at least equaUy important, and a good formula must express or suggest them all. The formula here proposed for malacone is not absolutely certam. The dehydration experiments of Zambonmi ' seem to show that it may be merely a hydrated zircon, that is, a zircon contamuig dissolved water in variable amounts. A broader study of malacone from several distinct locaH- ties seems to be desirable. The foregoing formulae, taken by themselves, are entitled to little consideration, but they become more significant when studied in connection with other compounds later. It wHl be noticed that one atom of zirconium is represented as linking four groups or radicles together ,^ just as one aluminum atom has a similar triple function in the aluminous orthosilicates. This is practically equivalent to regard- ing the minerals under consideration as derivatives of a complex dode- cabasic zircosihcic acid, H^^ZrSifi^^, which is at least as probable an interpretation as any other which has so far been advanced. In eudialyte, elpidite, and catapleiite we have a group of zirconium sihcates which form a highly suggestive series. Taking the simpler members first, they may be represented thus: Elpidite. Lime catapleiite. Soda catapleiite. /OH /VOH /OH //OH /OH //OH Zr Zr Zr XXSigOg^Nall^ \si303=NaH, \\ OH \Si308=CaN"a \\0H \Si303=Na3 Connecting these formulae with zircon, we have the facts that Si308 occurs in auerbachite, and that at Laven, according to Brogger,^ zircon is found both intergrown with catapleiite and pseudomorphous after it. Here again a different interpretation of the minerals has been advanced by Zambonini,^ who regards their hydration as extrinsic and not constitutional. To catapleiite he assigns the formula Na2ZrSi309.2H20, which, interpreted constitutionally, becomes 0=Zr=Si308=Na2.2H20. The calcium-tin sihcate, stokesite, is similar, and it has the formula CaSnSi309.2H20. Elpidite, according to Zambonini, is a metadi- silicate, Na^ZrSieOis.SHp. These formulae are legitimate and rational, and the only valid objec- tion to them is that they fail to bring out clearly the relations between the several species and the fundamental compound, zircon. 1 Contributo alio studio dei silicati idrati, p. 72, 1908. 3 Zeitschr. Kryst. Min., vol. 16, p. 105, 1890. » Op. cit., pp. 54-64. 116 THE CONSTITUTION OF THE NATUKAL SILICATES. Eudialyte and eucolite are commonly regarded as metasilicates, with the compound ZrOClj as an admixture. But that compound is not found in nature by itself, and both minerals, unlike most of the true metasilicates, gelatinize with acids. Furthermore, the analyses of eudialyte and eucolite show a considerable range of variation in the ratio Si : O, although approximating somewhat nearly to the assumed SiOg. If now we treat eudialyte and eucolite as mixtures of trisilicates and orthosilicates, like the feldspars, scapolites, and some micas, all difficulties vanish, the chlorine becomes equivalent to hydroxyl, and the minerals fall into line with catapleiite and elpidite as the firet members of the series. All varieties of eudialyte and eucolite are then interpretable as mixtures of the two molecules CI CI /siO,=CaNa /sigO^-CaNa ^SiO,=CaNa ^^ "". SigO^^CaNa Si04=CaNa SigOg^CaNa commingled in ratios nearly but not exactly 1:1. Hydroxyl replaces chlorine to some extent, while iron and manganese partly replace calcium; but the ratios shown by the formulae are constant, and the structural analogies with the aUied species are perfectly clear. The so-called ''zircon pyroxenes," rosenbuschite, wohlerite, laven- ite, guarinite, and hiortdahUte all conform to the type of expressions adopted here, although the analytical data are too scanty to yield positive conclusions. They can be given metasihcate ratios, follow- ing Brogger, by regarding the zirconium present in the form of a meta- zirconate. If this explanation is correct we should expect to find zirconates in nature, free from admixtures, but no such minerals are yet known. Artificial zirconates have, indeed, been prepared; but zirconium is more markedly basic than acid in its functions, and the analogy furnished by the orthosiHcate zircon has been my guide in the interpretation of these species. Furthermore, the ratio Si to O in each mineral is 1 to 4 or nearly so, which places them among the orthosiUcates. The simplest of these species, and probably the most definite, are rosenbuschite and guarinite. Guarinite, which originally was described as an isomer of titanite, has recently been reexamined by Zambonini and Prior ^ and found to be identical or nearly so with hiortdahlite, although the two minerals differ somewhat in composi- tion. Zambonini and Prior represent guarinite as a mixed salt of calcium metasilicate and sodium zirconate, with some calcium fluoride, but the following formula, which gives the same ratios, is more probable: J Mineralog. Mag., vol. 15, p. 259, 1909. SILICATES OF TETRAD BASES. 117 ^i04=Ca3=Si04v F— Zr— Si04=Ca3=SiO,— Zr— F \si04=:Ca2=Si0/ Na Na In hiortdahlite, as represented by Cleve's analysis, the same molecule appears, plus a less condensed molecule : /SiO,=Ca3=SiO,. F^^Zr/ >Zr-F3 \SiO— Ca— Sio/ Na k NaH the two being commingled in the ratio 1:1. Hydroxyl may replace fluorine to some extent in either case. These formulae compare with the actual analyses as follows: Guarinite. Hiort^a-hlite. Found. Calculated. Found. Calculated. SiOa 30.53 19.70 31.81 21.44 31.60 21.48 1.50 30 56 ZrOo 1 24. 77 TiOs CboO.. 1.68 .21 TaoO- FeO: ...... -. .34 .94 .96 32.53 .10 Fea. 1.91 1.56 35.80 .57 .43 6.13 r 39. 36 MnO CaO f 34. 11 MffO K2O NaoO 5.45 6.53 .58 5.83 6.30 H,0 .91 F._: 1.28 3.34 5.78 Less 99.80 .54 101. 40 1.40 102. 39 2.45 102. 43 2.43 99.26 100.00 99. 94 100.00 To rosenbuschite the following formula may be assigned: .Si04=Ca3=Si04v F_Zr— Si04=Ca3=SiO— Zr— F \siO— Ti— Sio/ ) i J Nao Fj Na2 118 THE CONSTITUTION OF THE NATUKAL SILICATES. This composition compares with Cleve's analysis as follows : Found. Calculated. SiOa 31.36 20.10 6.85 1.00 .33 1.39 24.87 9.93 5.83 30.30 ZrOs 20.54 TiOa 6.73 FegOg LaoOg MnO • CaO 28.29 NaaO 10.44 F .. 6.40 Less 101. 66 2.47 102. 70 2.70 99. 19 100. 00 The formulation of wohlerite is complicated by the presence of the quinquivalent element columbium, which may be either in the dyad group, =CbOF, or in the trivalent, ^Cb=0. This alternative leads to two possible types of structure, one of them Hke those immediately preceding, the other resembling that of eudialyte, thus: SiO^^Ca Si04=Ca / \cbOF / \zrF2 I. 4 Zr— SiO^^Ca + 1 Zr— SiO^^Ca \\Si04=CaNa \NsiO,=CaNa SiO^^CaNa SiO^^CaNa with small amounts of the groups W'^ replacing ZrFg. .Si04=Ca3=Si04. 4 0=Cb— Si04=Ca3=Si04— Cb=0 \si04=Zr =SiO / Na Na n. + .SiO^^Cag^SiO^v 1 F— Zr— SiO^^Cag^SiO — Zr— F \siO4— Zr— Sio/ Na2 F2 Na2 On comparing these alternative formulae with Cleve's analysis of wohlerite we have: SILICATES OF TETRAD BASES. 119 Found. Calculated. I. II. SiO. 30.14 .42 16.12 12.85 .48 .66 1.26 1.00 .12 26.97 7.50 .74 2.98 30.27 } 18.46 13.52 30.48 TiOa ZrOa 19.20 CboO, 12.87 FeaOa CegOg FeO 28. 26 7.83 MnO MgO 28.39 CaO NaoO 7 47 H2O F. 2.87 2 75 LessO 101. 24 1.24 101. 21 1.21 101. 16 1.16 100.00 100.00 100.00 These comparisons between observation and theory are as close as could be expected when we consider the evident impurities of the several minerals. Lavenite can be formulated in a similar manner, but the result is doubtful. The mineral contains impurities which can not be easily disposed of by mere calculation. New analyses are needed on purer material. It is worth noting, before going further, that these new formulae for the "zircon pyroxenes" are curiously similar to those developed for garnet, epidote, meionite, spodumene, and similar minerals. In these zircosilicates the trivalent radicles^Zr — F and ^Cb=0 play the same part that aluminum and ferric iron play elsewhere. The analogy is suggestive but may be nothing more. The typical sihcate of thorium, thorite, or orangite is an unsatis- factory species on account of its wide variations in composition. It is commonly supposed that the mineral, as it exists in nature, has been derived from an original ThSi04 by hydration, and that ThSi04 was isomorphous with zircon. The nearest approach to the type is found in orangite, which may perhaps be regarded as a thorium cyrtolite or thorium malacone. Yttriahte is another thoriferous mineral, which, however, seems to be a mixture of two salts. Its emphical formula is that of an ortho- trisihcate, and as given by Hillebrand's ^ analysis, it is not far from ThSi207R + 4Y2SiA; U. S. Geol. Survey BuU. 262, p. 61, 1905. 120 THE CONSTITUTION OF THE NATURAL SILICATES. K, being Fe, Mn, Pb, and Ca. The second of the two compounds is analogous to thortveitite, Sc2Si207. The actual jnineral contains many impurities, so that its assumed constitution needs to be verified by analyses of better material, if that should ever be found. Steen- strupine is stUl another silicate containing variable quantities of thorium, but the analyses are discordant and unreducible to any simple formula. Mackintoshite, a silicate of thorium and uranium, may perhaps be represented by the formula UO2.2ThO2.3SiO2.3H2O, and thorogum- mite appears to be a hydration derivative of it. Both minerals are closely related to thorite, which sometimes contains noteworthy quan- tities of uranium. The corresponding salts of quadrivalent uranium and thorium are probably isomorphous. Uranium functions as a hexad element in uranophane, CaO.2UO3.2SiO2.6H2O, and the mineral is possibly a basic orthodisilicate, containing the dyad radicle UO2. Its formula then becomes Ca(U02)2Si207.6H20. Nsegite is a complex zircosilicate, containing uranium, thorium, yttrium, and columbium, to which no definite formula can as yet be assigned. To the titanium silicates astrophyllite, Johns trupite, and rinkite formulae, Uke those given to zircon and its derivatives, are assignable. Indeed, this has already been done for astrophyUite by Brogger, who writes the formula Ti(Si04)4R' '411^4. This seems to be the dominant molecule in astrophyUite, which, however, varies in composition. To the Colorado mineral we may more precisely give the formula Si04=FeH OH /si04^FeK . /si04=FeH ^ ''''^Si04=FeNa + ^ '^'^Si04^FeH Si04=FeH Si04^FeH which requires the following percentage composition: SiOa 34. 30 TiOa 12.20 FeO 4L 16 K2O : 5.36 NaaO 3. 55 H2O 3. 43 100. 00 Some iron is replaced by manganese, and ferric iron, perhaps as =Fe — OH, is also present. In the fluoriferous astrophyUites the fluorine should replace hydroxy!. SILICATES OF TETRAD BASES. 121 Johnstnipite and rinkite are both fluoriferous, and both contain earths of the cerium groups. In johnstnipite, a little ZrOj, ThOj, and CeOa replace some TiOa- For Johns trupite the expression /siO,=CaJ^a ^^SiO.^CaNa Si04=CaH agrees well with the ratios given by analysis. . In rinkite we have, with great probability, the mixture Si04=CaNa SiO,=Ti— F /si04=CaNa ^ /siO,=Ca(CeF2) v^Si04=CaNa v SiO^^CaCCeF^) Si04=CaNa Si04=Ca(CeF2) Even the complex mosandrite reduces to the same general type, agreeing very closely with OH F /siO^^HNaCCeO^H,) . /siO^^CaH ^Si04=HNa(Ce02H2) "^ ^ ^Si04=CaH Si04^HNa(Ce02H2) Si04=CaH in which R =Ce^^ : Zr : Ti : : 1 : 2 : 2. For each of these species the pubhshed analyses agree well with the composition calculated from these formulae. Amo^g the titanosiUcates, neptunite appears to be a meta-com- pound, analogous in general structure to astrophyllite. Its formula, which is sharply in accord with the analysis, is SiOg— R' /siOg 'r^^" SiOa— R' where R' = Na, K, and R" = Fe, Mn. There is still another group of titaniferous silicates which seems to be unconformable with the foregoing scheme of interpretation. Titanite, the typical member of the group, has the empirical formida CaTiSiOg, for which two distinct structures have been proposed. One regards the mineral as the calcium salt of an acid, HgTiSiOg, analogous to HaSigOg, and the other treats it as a basic orthosilicate, Ca==Si04=TiO, 122 THE CONSTITUTION OF THE NATURAL SILICATES, A careful study of the recorded analyses of titanite leads me to prefer the orthosilicate expression, for the actual ratios vary in a way which indicates a replacement, sometimes of Ca and sometimes of TiO by other bases. According to the other formula, only the calcium should be replaceable. This variability of ratio is well shown by some of the varieties of titanite, such as grothite, alshedite, and eucolite- titanite, but the data are not absolutely conclusive. If, however, titanite is a basic orthosilicate, it should be classed with the orthosalts of dyad bases. On the other hand, the acid character of the titanium is suggested by the remarkable hydration derivative of titanite, xanthitane, of which the composition is approximately represented by the formula Al^TiA^Hg. For tscheffkinite, as shown by Eakins's analyses, the composition is approximately (FeCa)3Ce6Ti4Si6032, but the constitution of the mineral is very doubtful. Keilhauite appears to be like titanite, with Ca or TiO replaced by R'"OH or K^'OgHj, but the analyses are widely discordant. Benitoite, having the simple and definite formula BaTiSigOg, may be either a metasilicate or else a trisilicate, Ba = Si308 = TiO. The trisilicate formula is analogous to that of titanite and is there- fore to be preferred. Nasarsukite is a mineral of rather imcertain relations, but appar- ently a trisilicate < i308 = N"a2 i,0« = Na2 x^b^3w 8 — ^.11*2 in which about one-sixth of the sodium is replaced by the univalent group — Fe=0. The ideal sodium salt compares with the analysis of the actual mineral by Christensen as follows: Found. Calculated. SiOo 6L63 14.00 6.30 .28 .47 .24 16.12 .29 .71 63.83 TiOo 14.19 FeoOo AI2O3 MnO . 2L98 MgO . NaoO. H^O f! .:;:....:.. Less 100.04 .30 100. 00 99.74 SILICATES OF TETRAD BASES. 123 Leucosphenite, which is related morphologically to eudidymite, is probably also a trisilicate, although its formula has been written Na,Ba(TiO)2(SiA)5. This formula is difficult to interpret structurally, and a more rational expression, which also fits the analysis fairly well, is as follows: Na3Ba^Ti,(Si303),. Lorenzenite, an orthotrisihcate, appears to be a crystalline mixture of two salts, thus : 4 Na2(TiO)2Si207 + Na2(ZrO)2Si207. Several other minerals of more obscure character remain to be mentioned here. Molengraafite ^ is essentially a titanosilicate of sodium and calcium, which may perhaps have the empirical formula HNa3Ca4Ti3Si402o, if replacements and obvious impurities are left out of account. Its true character is quite uncertain. Epistolite is a complex columbosihcate, which is not far from Hi3Na6Cb3Si5027. Chalcolamprite and endiolite are also columbosiHcates of doubtful constitution. Possibly the phosphato-sihcates erikite and britholite, and perhaps also steenstrupine, should be classed with them. » See Brouwer, Centralb. Mineralogie, 1911, p. 129. APPENDIX. A number of well-defined silicates, of uncertain constitution, are difficult to place in any of the classes covered by preceding chapters. They naay be briefly sunamed up as follows: Thaumasite. — Empirical formula: CaSi03.CaS04.CaC03.15H20. This can be written structurally, regarding the water as extrinsic, as in some respects analogous to woUastonite, but such a formula would lack the evidence necessary to sustain it. Tseniolite. — ^A silicate of magnesia and the alkalies of uncertain composition. The one analysis of it is incomplete. Bakerite. — ^A borosilicate of calcium. The empirical formula is 8CaO.5BA.6SiO2.6H2O. Roeblingite .' — This species is unique in that it contains a sulphite radicle. It is regarded by Penfield and Foote as a mixed siUcate and basic sulphite, 5 H2CaSi04 + 2 CaPbOCSOg). This can be given a structural formula, which, however, would have little real significance. Langhanite.' — A silicate of manganese and iron containing antimony. AlHed on crystallographic grounds to hematite and ilmenite. A multitude of other silicates have been described as species, but without, as yet, securing fall recognition. Some of these are doubt- less mixtures or impure varieties of well-known minerals, but others may be ultimately established as good and definite compounds. A discussion of the analyses, without, experimental investigation of the various minerals, would have very uncertain value. I therefore omit these doubtful species from consideration. 124 INDEX TO SPECIES. A. Page. Acmite 97, 98, 103 Actinolite 100, 106 iEnigmatite 103 Agnolite 106 Agricolite 86 Akermanite 34 Alamosite 106 Albite 36, 37, 40, 42, 43, 48, 98 Alexandrolite 84 Alipite 106 AUanite 25 AUophane 84 Almandite 25 Alshedite 122 Alurgite 54 Alvite 114 Amesite 65 Amphibole 99, 100, 101, 102, 103, 104, 105 Analcite 37, 38, 39, 40, 41, 42, 43, 46, 47, 48, 98, 108 Anaphorite 104 Andalusite 15, 19, 22, 24, 31, 51, 72, 74, 75, 77, 84 Anderbergite 114 Andradite 25 Annite 55 Anorthite 26,30,32,36,37,41,46 Anthophyllite 100, 105 Anthosiderite 84 Antigorite 92 Aphrodite 94 Aphrosiderite 64 Apophyllite 108 Arctolite 34 Ardennite 78, 79 Arfvedsonite 103 Arizonite 86 Aspidolite 53 Astochite 100,101 Astrolite 81 Astrophyllite....- 120 Auerbachite 113, 115 Augite 26, 96, 99 Axinite 65, 72 B. Babingtonite 98 Bakerite 124 Barkevikite 103 Barsowite 74 B arylite 80, 81 Barysilite 107 Batavite 77 Bavenite. 49 Beckelite 85 Bementite 92 Benitoite 122 Page. Bertrandite 89, 110 Beryl 81 Biotite 24,51,52,53,54,58,66 Bityite -. 78 Brewsterite 46 Britholite 85 Bronzite 96 Brunsvigite 60,61 C. Calamine 89, 90 Cancrinite 27,28,34 Cappelinite 70, 71 Carnegieite 21, 23, 45, 48 Caswellite 59 Catapleiite 71, 115, 116 Celadonite 86 Celsian 35, 46 Cenosite 85 Cerite 84 Chabazite 41, 47, 48, 108 Chalcolamprite 123 Chamosite 65 Chloritoid 56, 57 Chloropal ' 84 Chlorophaeite 65 Chlorophyllite 79,80 Chondrodite 88, 93 Chrysocolla 106 Chrysotile 92 Cimolite 84 Clinocblore 59, 65 Clinohedrite 90 Clinohumite... 88 Clintonite 56 Coimarite HI Cookeite 67,78 Corundopbilite 62, 65 Cossyrite - 103 Crocidolite 102 Cronstedtite 60 Crossite 102 Cryoi^yllite 55 Cummingtonite 109 Cuspidine 109, 110 Custerite 107 Cyrtolite 113,114 D. Danalite 89 Danburite 74 Dannemorite 100 Daphnite ^ Datolite 90 Delessite 60' 61 Deweylite H^ 125 126 INDEX TO SPECIES. Page. Diabantite. 62, 63 Didymolite 78 Diopside : 96,99, 100 Dioptase 91 Dudleyite 57 Dumortierite 65, 73 E. Edingtonite 42, 43, 46 Ekmanite 63,64 Elaeolite 22, 27, 28, 37 Elpidite 115, 116 Endeiolite 123 Enstatite 94, 96, 99 Epichlorite 64 Epididymite 112 Epidote 24, 25, 29, 31, 36 Epiphanite 65 Epistilbite 46 Epistolite 123 Erikite 85,123 Erionite 46 Euclase 90 Eucolite : 116 Eucolite-titanite 122 Eucryptite 19,21, 51, 98 Eudialyte 71, 115, 116 Eudidymite 112 Eulytite 86 Euphyllite 52 Euralite ^. 65 F. Fassaite 33, 105 Faujasite 42, 43 Fayalite 87 Foresite 49,50 Forsterite 87, 88 Fowlerite 96 Friedelite 91 , 92 Fuchsite 51 G. Gadolinite 90, 91 Gageite 88 Ganomalite 107, 108 Ganophyllite 49,50,53 Garnet 24,25,26,29,30,32,34,56,65 Gamierite 94 Gastaldite 103 Gehlenite 31, 32, 33, 34 Genthite Ill Gismondite 45, 48 Glaucochroite 87 Glauconite 86 Glaucophane 101, 102 Gmelinite 41, 47 Gonnardite 50 Grandidierite 76, 77 Griqualandite 102 Grochauite 61 Grossularitc 25, 33 Grothite 122 Griinerite 100 Guarinite " 116, 117 Gyrolite 108 H. Hallite 59 HaUoysite 83 Page. Hancockite 25 Hardystonite 107 Harmotome 41, 43, 44, 46 Harstigite 75 Hastingsite 101, 103 Haughtonite 53 Haiiynite 26 Hedenbergite 96,99 Hellandite 85 Helvite 89 Heulandite 41, 46, 48 Hibscbite 45, 76 Hillebrandite 89, 106, 107 Hiortdahlite 116, 117 Hisingerite 84 Homllite 90, 91 HuUite 65 Humite 88 Hyalophane 35, 46 Hyalotekite 110,111 Hydrobiotite 58 Hydrocastorite 82 Hydronephel ite 26, 40, 42, 43 Hydrophlogopite 58 Hypersthene 96 T. Tgelstromite 87 Ilvaite 89,90 Inesite 109 lolite 79 Irvingite 55 J. Jadeite. 97,98 Jefferisite 58 JeffersoBite 96 Jobnstrapite 120, 121 K. Ksersutite 104 'Kainosite 85 Kaliopbilite 19, 21, 22 Kaolinite 36, 37, 39, 40, 57, 58, 64, 83, 84 Karyocerite 70, 71 Karyopilite 91 Keilbauite 122 Kentrolite 85 Kerolite 94 Kerrite 58 Klementite 65 Knebelite 87 Komerupine 76, 77,96 Kotscbubeite 60 Kryptotile 19, 22, 52, 76, 83 Kyanite 15, 23, 24, 31 L. Lagoriolite 24, 25 L^ngbanite 124 Lasallite 80 Laubanite 49 Laumontite 41, 46, 47, 48 L&venite 116, 119 Lawsonite 45, 76 Lazurite 26 Lennilite 58 Lepidolite 54,55 Lepidomorpbite 52 Leuchtenbergite 59, 65 INDEX TO SPECIES. 127 Leucite 37,38,40,43,98 Leucophanite 106 Leucophcenicite gg Leucosphenite 123 Leverrierite ^ 22 51 Levynite 47' 4g Linosite IO4 Lorenzenite 123 Lucasite 5g M. Mackintosliite 120 Maconite 5g Malacon 113,114,115 Manandonite 65, 73, 78 Manganidocrase 31 Manganophyllite 53 Margarite 57 Marialite 36, 37 Meionite 23, 36, 37 Melanocerite 70, 71 Melanolite 60 Melanotekite 85 Melilite 26, 31, 33 Melite 84 Meliphanite 112 Mesolite 42, 48 Metachlorite 61 Microsommite 27, 37 Milarite 82 Minguetite 64 Molengraafite 123 Molybdophyllite 89 Monticellite 87, 99, 105 Montmorillonite 83 Mordenite 48 Mosandrite 121 Miillerite 84 Muscovite 19, 22, 23, 26, 34, 36, 37, 51, 52, 55, 66, 72, 74 N. Naegite 120 Nasarsukite 122 Nasonite 107, 108 Natrolite 26,40,42,43^46,48 Nephelite. 19, 21, 23, 27, 28, 36, 37, 40, 42, 46, 51 Nepouite 94 Neptunite 121 Newtonite 83 Nontronite 84 Noselite 26 Noumeite 9^ p. Ofiretite 50 Okenite 108 Olivine 26, 94*, 99, 105 Orangite 119 Orthoclase 37,48 Ottrelite 56 Ouvarovite 25 P. Painterite 59 Paragonite 19, 22,51, 52 Partschinite 25,29 Pattersonite 59 Page. Pectolite 93 Pennine .....59,60,65 Petalite •;. g2 Phenakite §7 gg Phengite '52 Philadelphite sg Philipstadite 103 Phillipsite 41,43,44,45,48 Phlogopite 51,63,64,59,66 Pholidolite 107 Picrosmine 94 Piedmontite 25 Pilinite 49 Pilolite ,. gQ PlancMite igg Plombierite iQg PoUucite 39 Polylithionite : 55 Prehnite 24, 29,34,36,38,48 Prismatine 73 77 Prochlorite gl, 62^*65 Prolectite gg Protovermiculite sg Pseudobrookite gg Pseudodeweylite 107, m Pseudoleucite 37 Pseudonatrolite 59 Pseudonephelite 22 Ptilolite 49 Pycnochlorite 65 Pyrope 25 Pyrophyllite gg Pyrosclerite 59 Pyrosmalite 91 Pyroxene 95,96,97,99,100,104,105 R. Rectorite 22, 83 Rhodonite 96 Rhodusite 102 Rhonite 104 Richterite 100 Riebeckite 103 Rinkite 120,121 Rivaite 1 10 Roeblingite 124 Roscoelite 51 Roseite 58 Rosenbuschite 116, 118 Rubellite 68 Rumpfite 60 S. Sahlite 96 Samoite 84 Saponite 109 Sapphirine 73 Sarcolite 24 Scapolite 31, 33, 34, 36, 51 Schefferite... 96,110 Schizolite , 96 Schorlomite 26 Schrotterite 84 Scolecite , , 42,43,48 Serendibite 65,73,78 Sericite 51 Serpentine 65,92,93,94 128 INDEX TO SPECIES. Page. Seybertite 56 Sheridanite 64 Siderophyllite 53 SiUimanite 15, 23, 24, 31, 83 Sodalite 26, 27, 28, 29, 37, 40, 42 Soretite 103 119 25,29 Sphenoclase 81 Spinel 56,65 Spodiophyllite 107 Spodumene 97,98,99 Spurrite 89 Staurolite 74, 75 Steenstrapine 123 Stellerite 50 Stilbite 41, 43, 44, 45, 48 Stilpnomelane 63,64 Stokesite 115 Strigovite 64 Syntagmatite 101 T. Taeniolite 124 Talc... 105,106 Talc-chlorlte 63 Taramellite 107 Tawmawite 25 Tephroite 87 Termierite .' 84 TMlenite 85 Thaumasite : 124 Thomsonite 45 Thorite 119 Thorogummite 120 Thortveitite 85 Thuringite 62, 63 Titanite 121, 123 Page. Topaz 19,22, 23,51 Tourmaline 65,66,67,68,69,78 Tremolite 100, 101, 105 Trimerite 87,89 Tritomite 70, 71 Troostite 89 Tschefifkinite 122 U. Uranophane 120 V. Vaalite 59 Vermiculite 58, 59 Vesuvianite i30, 31, 34, 65 Villarsite 92 W. Waldheimite 101 Weinbergerite 107 Wellsite 43,44 Westanite 83 Willcoxite 57 Willemite 87, 89 Woerthite 83 Wohlerite 116, 118, 119 Wolchonskoite 86 WoUastonite 95, 96 X. Xanthophyllite 56,65 Y. Yttrialite 119 Z. Zeophyllite Ill, 112 Zinnwaldite 54, 55 Zircon 113, 114, 115 Zoisite 25,29 Zungite 24, 29 O' 5{)4"to