EXCHANGE 
 
 "A 
 
The Radiating Potentials of 
 Nitrogen 
 
 HENRY DEWOLF SMYTH 
 
 vK 
 
The Radiating Potentials of 
 Nitrogen 
 
 A DISSERTATION 
 
 PRESENTED TO THE 
 
 FACULTY OF PRINCETON UNIVERSITY 
 
 IN CANDIDACY FOR THE DEGREE 
 
 OF DOCTOR OF PHILOSOPHY 
 
 BY 
 HENRY DEWOLF SMYTH 
 
 Reprinted from 
 (THE PHYSICAL REVIEW, N. S., Vol. XIV., No. 5, November, 1919.) 
 
Accepted by the Department of Physics 
 June 1921 
 
 PRESS OF 
 
 THE NEW ERA PRINTING COMPANY 
 LANCASTER, PA. 
 
 * * **"/!' 
 
Reprinted from the PHYSICAL REVIEW, N.S., Vol. XIV^N^o,, y v Noveipj&t, 
 
 THE RADIATING POTENTIALS OF NITROGEN. 
 
 BY H. D. SMYTH. 
 
 SYNOPSIS. 
 
 1. A formula is derived by which accurate corrections allowing for the distribution 
 of velocities of the impacting electrons may be applied to the observed values of the 
 radiating potentials. 
 
 2. Measurements on nitrogen revealed 
 
 (a) a very strong effect at 8.29 0.04 volts. 
 
 (&) a very doubtful effect at 7.3 volts. 
 
 (c) an effect appearing only at lower pressures but strongly at 6.29 .06 volts. 
 
 3. These results are explained as follows: 
 
 (a) The wave-length X corresponding to 8.29 .04 is 1490.7 10 and this effect 
 is .therefore, identified with the doublet found by Lyman at 1492.8 and 1494.8. 
 
 (6) This gives X = 1700 and may be identified with the second doublet 1742.7 and 
 1745.3 attributed to nitrogen but by some thought due to silicon. 
 
 (c) The value X = 1965 20 from 6.29 .06 volts is taken to correspond to the 
 beginning of the band spectrum at 1870.9. The discrepancy is attributed to the pres- 
 ence of nitrous oxide. 
 
 According to another theory the effect at 6.29 is considered due to lines in the region 
 2,000-3,000 A.U. coming from neutral atoms. The value 6.29 in this case is taken as 
 the speed necessary before the electrons can split up the molecules. 
 
 4. From 3 (a) assuming line spectra to come from atoms, we have as an upper 
 limit to the heat of dissociation of a gram molecule of nitrogen 190,000 calories. 
 
 From the second theory in 3 (&) we have as a possoible actual value for this heat of 
 dissociation 145,000 calories. 
 
 5. Qualitative evidence was obtained supporting Davis and Goucher's discovery 
 of true ionization in the neighborhood of 18 volts. 
 
 I. INTRODUCTION. 
 
 THE substances whose minimum radiating and ionizing potentials 
 have been investigated fall naturally into two classes, first, the 
 monatomic gases and metallic vapors, in which collisions below a certain 
 velocity are elastic, and second, the diatomic gases, in which collisions 
 at all velocities are inelastic or at least partially so, as in the case of 
 hydrogen. . 
 
 The investigations for substances of the first class have recently been 
 summarized and discussed by McClennan. 1 The conclusions drawn are, 
 briefly, as follows. A vapor, when bombarded by electrons of velocity 
 V emits a radiation eV = hv. The first radiation produced is the line 
 
 1 J. C. McClennan, "The Origin of Spectra," Proc. London Phys. Soc., XXXI., Part I., 
 pp. 1-29, 1918. 
 
 44472,3 
 
4IS H. D. SMYTH. 
 
 at the head of the single t principal series, and thereafter shorter and 
 shorter wave-length radiations set in as the bombarding electrons attain 
 correspondingly higher velocities. Finally, when the speed of the elec- 
 trons reaches a value corresponding to the frequency v = (1.5, S), the 
 convergence frequency of the singlet principal series of the element in 
 question, ionization occurs. 
 
 For diatomic gases the mechanism of ionization or the production of 
 radiation is obviously more complicated than where only a single atom 
 is involved. Either it is necessary first to dissociate the molecules and 
 then, by a second impact, to cause radiation or ionization, or only one 
 collision is necessary. .Again, if the effect is the result of one impact, 
 it may occur in two ways; an electron may be displaced and the molecule 
 broken up simultaneously, or an electron may be displaced without 
 affecting the bond between the atoms. 
 
 Possibly as a result of the above, the spectra of the diatomic gases 
 are very complex and, for the most part, not resolved into series. This 
 makes the prediction of ionizing and radiating potentials difficult. An- 
 other source of error especially serious in the case of diatomic gases arises 
 from the uncertainty in applying the necessary correction for velocity 
 distribution, as discussed later in the paper. 
 
 In consequence of these complications, although values of the ionizing 
 and radiating potentials have been determined for a number of different 
 diatomic elements and compounds, little has been accomplished toward 
 connecting them with the spectra. In spite of the simple structure of 
 the hydrogen molecule and the extensive data on the spectrum avail- 
 able, the experimental results obtained have not yet been reconciled 
 with those predicted by Bohr's theory or by the quantum relation 1 
 eV = hv. 
 
 Though the spectrum of nitrogen has not been resolved into series, in 
 the region of the ultra-violet with which we are concerned the line 
 spectrum is very simple. Lyman, 2 in his investigation of the extreme 
 ultra-violet, found for nitrogen no lines except two doublets, one con- 
 sisting of the lines 1492.8 and 1494.8 and the other 3 of the lines 1742.7 
 and 1745.3. Applying the quantum relation to these, we have for the 
 corresponding radiating potentials, 8.28 and 8.27, and 7.04 and 7.08 
 volts. The longest wave-length of the band spectrum which Lyman 
 found in this region is 1870.9 corresponding to 6.6 volts. Now the 
 generally accepted value for the radiating potential of nitrogen is 7.5 
 
 1 Bergen Davis and F. S. Goucher, PHYS. REV., No. 10, p. 101, 1917. 
 
 2 Lyman, " Spectroscopy of Extreme Ultraviolet," p. 113, 1914. 
 8 Possibly due to silicon. 
 
No L 's XIV "] RADIATING POTENTIALS OF NITROGEN. 4! I 
 
 volts while Davis and Goucher 1 found a second more intense radiation 
 at 9 volts. 2 
 
 In view of the discrepancy between these experimental values and those 
 calculated from the spectrum, it was thought worth while to attempt a 
 more exact determination. 
 
 II. APPARATUS. 
 
 The apparatus used had been designed for a somewhat different purpose 
 but was found fairly satisfactory. 
 
 As shown in Fig. I, it consisted of a glass tube 1.5" in diameter, with 
 a filament F sealed in at one end, a gauze G in the middle and a disc P 
 sealed in from the other end, all of platinum. Electrical connections 
 
 Ji 
 
 Fig. 1. 
 
 were so arranged that the electrons coming off from the hot filament F 
 were accelerated by a field between F and G but met a stronger retarding 
 field between G and P. The difference between accelerating and retard- 
 ing fields remained constant. The gauze G was connected to a Leeds & 
 Northrup high sensitivity galvanometer which measured the electronic 
 current between F and G. The disc P was connected to a Dolazalek 
 electrometer of sensitivity about 3,000 millimeters per volt, making the 
 capacity of the electrometer system of the order of magnitude of 50 cm. 
 The potentials were read on a'Robt. W. Paul voltmeter. 
 
 By coating the filament with barium oxide increased emission was 
 obtained. 
 
 The pressure in the apparatus was regulated by a Gaede pump and 
 measured with a McLeod gauge. 
 
 Nitrogen was prepared by heating sodium nitrite and ammonium 
 chloride with distilled water and was introduced through a drying tube. 
 The apparatus was evacuated and washed out several times with nitrogen 
 before measurements were made. 
 
 The setting in of radiation was detected in the usual manner by the 
 photoelectric effect on P causing an increase in the speed of deflection of 
 
 1 Bergen Davis and F. S. Goucher, PHYS. REV., No. 13, pp. 1-5, 1919. 
 
 2 The fact that these are radiation effects and not ionization seems to have been proved by- 
 Davis and Goucher and is taken for granted in this paper. 
 
4 I2 H. D. SMYTH. 
 
 the electrometer. It was found impossible to eliminate all zero drift 
 from the electrometer. This made it necessary to start readings far 
 enough below the break point to give a good zero. The accelerating 
 potential was run from 3 or 4 volts up to n or 12 and then down again, 
 the intervals between readings near critical points being as small as .2 
 of a volt. The values going up and coming back were averaged. The 
 electronic current measured by the galvanometer was kept constant 
 in the earlier runs by adjusting the temperature of the filament but, in 
 the later runs, the temperature of the filament was maintained constant 
 and the galvanometer current allowed to vary with the accelerating 
 potential, thus tending to accentuate the sharpness of the break. 
 
 III. VELOCITY DISTRIBUTION CORRECTION. 
 
 The most serious source of error in experiments of this type is due to 
 the fact that the electrons reaching the gauze will not all have exactly 
 the velocity corresponding to the accelerating field. 1 The factors causing 
 this trouble are the potential drop along the filament, initial velocity of 
 emission and, in the case of diatomic gases, inelastic impacts. 
 
 Now, in the case of monatomic gases the elastic impacts make possible 
 a very effective method of eliminating this error. In this case, the 
 electrometer current rises rapidly as the accelerating field passes the 
 critical value, reaching a maximum when all the electrons have attained 
 the critical speed, and then falls off again as more and more of the ionizing 
 collisions take place on the filament side of the gauze. When the 
 accelerating potential becomes great enough to allow two ionizing colli- 
 sions by one electron another maximum occurs, and so on. Thus, by 
 measuring the intervals between successive maxima the true value of the 
 ionizing potential can be found. 2 
 
 With diatomic gases, however, this method is impossible since, with 
 every increase of the accelerating field, some electrons which have lost 
 energy by inelastic impact will attain the critical speed and there will 
 be a continuous increase in the electrometer current. It becomes neces- 
 sary, therefore, to actually measure the velocity of the electrons coming 
 across and then make a correction. 
 
 Franck and Hertz did this but felt so uncertain as to the right method 
 of making the correction that they claimed an accuracy of only one volt 3 
 for their results. 
 
 Goucher 4 eliminated errors due to the potential drop along the filament 
 
 1 J. Franck and G. Hertz, Verb. d. D. Phys. Ges., 15, p. 37, 1913. 
 
 2 J. Franck and G. Hertz, Verh. d. D. Phys. Ges., 16, p. 457, 1914. 
 
 3 J. Franck and G. Hertz, Verh. d. D. Phys. Ges., 15, p. 39, 1913. 
 
 4 F. S. Goucher, PHYS. REV., No. 8, p. 561, 1916. 
 
No L ' S XIV '] RADIATING POTENTIALS OF NITROGEN. 413 
 
 by introducing an equipotential electron source consisting of a platinum 
 thimble surrounding a tungsten heating element. With this arrangement 
 in mercury vapor he found that over 70 per cent, of the electrons had 
 velocities corresponding to the applied voltage and, further, that the 
 number having a velocity corresponding to .5 volt greater than the 
 applied field was too small to be measured. He therefore made no 
 correction for velocity distribution. In the later work of Davis and 
 Goucher 1 on nitrogen, the same type of electron source was used and 
 the correction considered unnecessary. As before stated the values 
 obtained were 7.5 and 9 volts. 
 
 Bishop, 2 working with a filament, took the point of maximum slope of 
 his electron current curve for his velocity distribution correction. That 
 is, he took the most probable velocity of the electrons. He found the 
 value of 7.5 volts for nitrogen and the same for N 2 O. 
 
 Hughes and Dixon, 3 who obtained the values 7.7 volts for nitrogen 
 and 9.3 for nitric oxide, took as their velocity correction the highest speed 
 detectable on their velocity distribution curve. 
 
 Attempts to apply velocity distribution corrections in preliminary 
 tests of the present apparatus proved that the methods mentioned 
 above give quite different results, and that the discrepancy between them 
 varies with the filament temperature and gas pressure. Furthermore, 
 when correcting by the method of Hughes and Dixon, the relative 
 sensitivities of the apparatus for radiation and for velocity distribution 
 measurements and, also, the scale to which the measurements were 
 plotted could be altered so as to vary the value of the corrected "break" 
 point by as much as a volt, without any obvious way of selecting the 
 correct value from among the various possible ones. 
 
 This experience led to a closer study of the problem with a view of 
 determining the method of handling data from the two types of measure- 
 ment which would give the most nearly correct result. The variation 
 of zero drift sets a practical limit to the scale of plotting of either curve. 
 Therefore, it remains to determine an appropriate scale for the more 
 sensitive measurement and this is done by finding the relation between 
 the sensitivities of the apparatus for the two types of measurement. 
 Obviously, the "break" point in the radiation curve is due to the radia- 
 tion from the smallest number of electrons which can produce a large 
 enough radiation effect to be detected by the apparatus. Since not all 
 electrons capable of producing radiation do so, owing to failure to collide 
 or for other reasons, this "smallest" number of electrons is larger than 
 
 1 Bergen Davis and F. S. Goucher, PHYS. REV., No. 13, p. i, 1919. 
 
 2 F. M. Bishop, PHYS. REV., No. 10, p. 244, 1917. 
 
 8 A. LI. Hughes and A. A. Dixon, PHYS. REV., No. 10, p. 495, 1917. 
 
H. D. SMYTH. 
 
 FSE COND 
 
 [SERIES. 
 
 
 G 
 
 
 
 E 
 
 
 
 
 
 
 
 p 
 
 ^T 
 
 
 
 xt 
 
 p *' 
 
 
 
 A 
 
 the least number which can be detected if they are permitted to strike 
 the receiving electrode, as in the velocity distribution measurements. 
 The problem, therefore, is to find how many electrons must pass the 
 gauze with sufficient energy to cause radiation in order that one may 
 produce radiation. In other words, how many times less sensitive is the 
 
 apparatus for radiation than for ve- 
 locity distribution measurements and 
 how should the latter be plotted in 
 order that the greatest velocity shown 
 should give the correction appropriate 
 to the "break" point in the former? 
 
 The following analysis leads to a 
 very usable expression for the ratio 
 of the sensitivity of the electrometer 
 Fig. 2. system for radiation to that for veloc- 
 
 ity distribution experiments. 
 Let: 
 
 N = number of collisions per centimeter path at I mm. pressure; 
 X A and X R = the electric intensities in the accelerating and retarding 
 
 fields, respectively; 
 n = number of electrons per unit time reaching the gauze G 
 
 from the filament side; 
 f(V)dV = the probability of an electron reaching the gauze with a 
 
 speed between V and V + d V\ 
 VQ = the minimum radiating velocity ; 
 
 A and B be two planes parallel to G distant d and d' on either side of 
 it, where 
 
 V- V 
 
 d = 
 
 and 
 
 d' 
 
 V-V G 
 X R 
 
 - the probability of an electron reaching A with velocity 
 between F and V Q + d.V. 
 
 In all cases, velocities are expressed in terms of equivalent volts. 
 Evidently 
 
 nf(V )dV(i - e- Nd ) = n(^ Nd - i)f(V)dV 
 
 is the number of electrons which pass A with velocities between Fo and 
 VQ + dV and collide before reaching G. 
 After passing the gauze the electrons will go a distance 
 
 d' = (V - V )JX S 
 
No L ' S XIV '] RADIATING POTENTIALS OF NITROGEN. 415 
 
 before losing their ability to cause radiation. From these we have 
 
 nf(V)dV(i - e-*> Nd> ) 
 
 as the number of radiating impacts after passing the gauze by electrons 
 which reached the gauze with speeds between V and V + d V. 
 
 We have, then, for the total number of impacts at a speed greater than 
 VQ, by electrons going in the direction away from the filament, 
 
 - 
 (e P *A -e P **)f(V)dV t (l) 
 
 where V m is the maximum speed of any electron reaching G. 
 
 There will also be a small number of electrons which will go through, 
 be stopped, and acquire sufficient acceleration in the opposite direction 
 to produce radiation. For these we have 
 
 as the number which do not collide before their velocity is retarded to 
 and 
 
 ne 
 
 v 2F 
 
 '~ P 
 
 as the number which will regain a velocity VQ in the opposite direction 
 before collision. 
 Then 
 
 is the number which, having passed G with velocity between V and 
 V + dV, escape collision until they have reversed their direction and 
 regained a speed greater than V , finally colliding before their speed is 
 again reduced to VQ. 
 
 On substitution we have 
 
 dM 2 = n'e~ pN ^(i - 
 
 2F 
 
 Therefore 
 
 Xr m T F+F O / 2F v F O \ 
 
 [c X R e * x * X A ]f(V)dV (3) 
 
 
 is the number of collisions by electrons coming in the reverse direction 
 and colliding while they have a velocity between F m and VQ. 
 
 We have, for the total number of collisions by electrons having veloci- 
 ties greater than VQ, 
 
41 6 H. D. SMYTH. 
 
 V (A} 
 
 X Vm V ( ZV V ~ V <>\ ^T/ 
 
 -0 
 
 Now if we take the probability of the production of radiation and its 
 detection by photoelectric effect to be ( F V )/k F , where the value 
 of k Fo may be found from the results of Johnson 1 we have for the effective 
 number of radiating impacts when the maximum electronic speed is F TO , 
 
 M' = 
 
 We now turn to the velocity distribution measurements. If D is 
 the distance from the gauze G to the plate P and V R is the retarding 
 potential, the number of electrons getting to the plate will be, for given 
 values of V m (i.e., VA, the accelerating potential) and V R , 
 
 (6) 
 
 Now, the form of the function /( F) will vary with V m (or VA) and its 
 value will depend also on the particular value of F substituted. 
 For a given value of V m we will have, from (6), 
 
 (7) 
 
 Consider the part near the foot of a typical velocity distribution 
 curve, as shown by Fig. 3. The curve is a section of a parabola, while 
 actual experimental measurements are indicated by dots. It is evi- 
 dently sufficiently accurate, for the present purposes, to consider the 
 lowest part of the distribution curve to be parabolic, so that its slope 
 is taken to be proportional to the horizontal distance from the foot. 
 Use of this property gives the graph shown in Fig. 4. 
 
 Let V R > be some particular value of V R ; then 
 
 l = R' 
 
 Also, we have, for any value of V R less than F TO , 
 
 dM P _V m -V R (dM P 
 dV R " V m - F 
 
 1 J. B. Johnson, PHYS. REV., X., p. 609, 1917. 
 
No L 6 XIV 'j RADIATING POTENTIALS OF NITROGEN. 
 
 Combining these two equations, we have 
 
 (M P ) VK . - 
 
 2 F w - V R 
 
 dV R 
 
 
 
 O.4-.0 \ .1 
 
 1'LI-- : - ' 
 
 > 
 
 
 _^> . \ - ./.j 
 
 
 
 
 II '\ - ft; 
 
 ) <- rrv pt^tT, fc - __ u. - 
 
 
 
 . 7 r \ -p 
 
 WCily JiSLTLDU. .Ipn 
 
 
 
 o 6.5 \ 
 
 J y, 71__ *n 1 ** 
 
 
 
 
 / | p p"| 
 
 , , 
 
 
 . i e 
 
 xcerL TL^LaL J^ulu. i 
 
 jS ^1. 
 
 
 < -z nH V 
 
 
 
 
 G.0.0 "V 
 
 ruro.Pt to. j . ! 
 
 
 
 
 T 
 
 
 
 g 
 
 "* lj~ .- f IL 
 
 (ylfll.x^f|=D. 
 
 Jifi^l 
 
 
 
 J 
 
 c V ! 
 
 ok-Mp =. OJ364 
 
 Pg* 
 
 P^ 2.0 . 
 
 
 T 
 
 il= _ 'QJ-ff 
 
 
 
 
 
 
 
 O ~ '- . - -.- - V 
 
 J - ;: i ^ 
 
 
 
 +- 2.0 y 
 
 l -! i 
 
 
 
 2 - : \ 
 
 i i . 
 
 
 
 jjj 
 
 -t - 
 
 
 
 <u i c 
 
 i ; - i - - 
 
 
 
 C^ /.O 
 
 \j H L_ _4 
 
 
 
 L 
 
 \ ~ 
 
 
 
 - _4_ ' -, 
 
 i\ _ . i_ 
 
 k~ 
 
 If 
 
 
 ^r ~T ~^~ 
 
 S *^ I i ' 
 
 
 
 
 
 ; "Xj 
 
 i 
 
 
 
 TE^" -. ' . 
 
 
 
 
 
 
 
 -2j 0.5 
 
 - - ^ 
 
 I 
 
 
 LU i .:- 
 
 v 
 
 
 
 
 1 >^ ' 
 
 
 
 
 O.C ! 
 
 
 
 417 
 
 u -/.o -0.8 -as -0.4 -0.2 -o.o 
 
 Fig. 3. 
 
 Substituting for dM P /dV R from (7), and solving for f(V) v=VR , we have 
 
 2(M P ) F /F TO - 
 
 *(Vm ~ V*? 
 We are endeavoring to find the ratio of corresponding values of M' 
 
 10 
 
 8 
 
 4- 
 
 
 
 \ 
 
 
 
 
 
 
 
 
 1 | 
 
 
 
 - 
 
 1 
 
 E 
 
 
 
 A^ 
 
 7^/r 
 
 P 
 
 A ! 
 
 
 
 
 
 
 
 \ 
 
 
 
 
 
 
 
 Ifi 
 
 Y 
 
 c 
 
 
 
 
 
 
 
 
 br 
 
 ?a' 
 
 r-j- 
 
 
 
 
 
 q 
 
 
 / 
 
 rr/tr 
 
 p 
 
 & Pa 
 
 rfl 
 
 
 
 : 
 
 s 
 
 
 
 
 
 
 "ft 
 
 n: 
 
 re 
 
 
 
 
 
 
 
 
 1 
 
 
 \ 
 
 
 
 
 
 
 i 
 
 ^ 
 
 
 
 
 J 
 
 
 
 \ 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 V 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 K 
 
 
 
 
 ; 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 r 
 
 j 
 
 j 
 
 
 
 
 
 
 
 
 
 | 
 
 
 
 
 
 
 
 | 
 
 \ 
 
 
 
 
 
 
 
 
 
 
 
 
 
 | 
 
 - 
 
 
 
 
 
 
 
 
 \ 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 V 
 
 
 
 
 
 
 
 
 
 
 I 
 
 
 
 
 
 
 i 
 
 
 \ 
 
 
 
 
 
 
 
 
 
 
 
 
 
 i 
 
 . 
 
 \ 
 
 
 
 
 
 - 
 
 
 
 
 
 
 
 
 
 
 
 
 1 
 
 
 
 
 S : 
 
 
 
 
 
 
 
 . 
 
 
 
 
 
 
 fi^l 
 
 . 
 
 
 
 
 < 
 
 
 
 
 
 
 
 - 
 
 
 
 
 1 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 /\ 
 
 n' 
 
 
 
 i IV 
 
 
 
 
 
 
 
 
 
 
 
 
 
 1 
 
 
 
 
 
 \ 
 
 
 
 
 
 
 
 1 
 
 
 
 
 
 
 
 , 
 
 
 
 
 
 \ 
 
 
 
 
 
 
 
 
 
 
 
 
 
 1 
 
 4rf 
 
 CL= 
 
 My>) 
 
 
 \ 
 
 
 
 
 
 
 
 
 
 
 
 Li 
 
 
 
 
 * ' 
 
 H 
 
 
 
 \ 
 
 
 
 
 
 , 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 ' 
 
 \ 
 
 
 
 
 
 i 
 
 : 
 
 
 
 1 
 
 
 4 
 
 I 
 
 
 
 
 
 
 
 
 
 V 
 
 
 , 
 
 
 
 
 
 ^ 
 
 - 
 
 ^L']"" 
 
 --- 
 
 
 
 
 
 4,S ' 5.0 
 
 I i2 
 
 5^- 
 
 56 
 
 ,*? 
 
 5 
 
 &0 
 
 ** 
 
 Fig. 4. 
 
4 1 8 H. D. SMYTH. 
 
 and M P , i.e., of values at equal distances from the break point. For 
 this condition we must have 
 
 V m - Fo = V m - V R , 
 
 and ( V Fo) in the case of radiation corresponds to ( V m V R ) in 
 the velocity distribution. We can therefore substitute the value found 
 above for/(F) in the expression for M' if we write V m VQ for V m V R 
 and V VQ for V m V R . If at the same time we divide through by 
 (M P }v Rl , we have the following expression for the required ratio: 
 
 M r C 
 
 IF = 2 < pm \ 
 
 Mp Jv, 
 
 (V m - F ) 2 kV 
 
 for corresponding values of M r and M P in the neighborhood of the break 
 point. 
 
 If, in the above expression, we put the constant factors outside the 
 integral and set pN/X A = a, pN/X R = a' and F F = x, and if we 
 take a = a', which is very nearly true we have 
 
 M 1 2 pND 
 
 C Xm 
 
 I X 2 \ ax ~ ax 4- 
 F L ' 
 
 Jo 
 
 K/T J, T7 / T7 T7 
 
 Mp kVQ(V m V o 
 
 By integration and the expansion of the powers of e (either before or 
 after integration), this equation takes the following form, which includes 
 all terms as far as those in a 3 : 
 
 + -V-a 2 F (F w - Fo) + 2a 2 F 2 +]. 
 
 Practically, the terms in a 2 are negligible, since a is of the order of magni- 
 tude of 0.03. 
 
 If, therefore, we call the sensitivity of the apparatus for radiation R 
 and for velocity distribution S, we have, for practical purposes, 
 
 R _M'_ e ND a(V m - F ) 2 / 
 S ~ioM m ~ ' 
 
 
 where V m is the maximum speed detectable in velocity distribution 
 measurements and the factor i/io is introduced because, in our apparatus, 
 the dimensions were such that P receives only about that proportion 
 of the radiation starting at the gauze. 
 
 Equation (8) is evidently not exact, owing to the approximations 
 
VOL. XIV. 
 No. 5. 
 
 RADIATING POTENTIALS OF NITROGEN. 
 
 419 
 
 introduced in order to obtain a solution. It is believed, however, that 
 none of these approximations is seriously in error, so that the results 
 given by (8) should be of the right order of magnitude. 
 
 The necessary procedure, therefore, is to plot the velocity distribution 
 on the same scale on which the radiation curve is to be plotted and deter- 
 mine V m from this. Then multiply the values of the electrometer current 
 for each value of V R by the corresponding values of R/S and replot. 
 The break point in this new curve determines the correction to be used 
 on the radiation curve. The necessity of plotting a corrected velocity 
 distribution curve is made evident by Table I. which gives the values 
 of S/R calculated by equation (8) for different values of V R V m , taking 
 k = 2 for nitrogen. 1 
 
 TABLE I. 
 
 0.015 mm - 
 
 7-5 volts. 
 
 fc-Fft. 
 
 SIR. 
 
 (p->j& 
 
 SIR. 
 
 .1 
 
 795,000 
 
 .1 
 
 6,500 
 
 .3 
 
 87,500 
 
 .2 
 
 5,500 
 
 .5 
 
 31,000 
 
 .3 
 
 4,700 
 
 .6 
 
 22,000 
 
 .4 
 
 4,050 
 
 .7 
 
 16,000 
 
 .5 
 
 3,500 
 
 .8 
 
 12,500 
 
 .6 
 
 3,200 
 
 .9 
 
 9,500 
 
 .7 
 
 2,750 
 
 1.0 
 
 8,000 
 
 
 
 IV. EXPERIMENTAL RESULTS. 
 
 It was found impossible to eliminate all zero leak from the electrometer. 
 Consequently readings had to be taken at low accelerating potentials to 
 establish a zero line. Errors due to a variation in this zero leak as well 
 as errors of observation were pretty well eliminated by the method of 
 taking a series of readings with fields increasing and then decreasing 
 and plotting the mean. A typical set of readings is given in Table II., 
 below, 
 where 
 
 Vf = P.D. across filament in volts; 
 G = current to gauze in amperes X io~ 8 ; 
 T = time in seconds for electrometer to deflect one cm.; 
 T = mean of T and T' observations as VA was increased and 
 
 then decreased respectively; 
 
 I IT = rate of deflection of electrometer in cms. per second. 
 The values of VA and i/T, that is the accelerating potential and the 
 current to the electrometer, were used in making the curves. The first 
 
 J J. B. Johnson, loc. cit. 
 
420 
 
 H, D. SMYTH. 
 
 [SECOND 
 [SERIES. 
 
 P = 
 
 TABLE II. 
 
 Observations for Run No. 13. 
 o.oi mm., VR VA = 2.5 volts, V/ = 0.8 volts. 
 
 r+ 
 
 G. 
 
 G'. 
 
 C 
 
 T'. 
 
 T. 
 
 I IT. 
 
 l 
 
 6 
 
 6 
 
 81 
 
 81 
 
 81 
 
 .0124 
 
 2 
 
 16.4 
 
 17 
 
 79 
 
 86 
 
 82.5 
 
 .0121 
 
 3 
 
 21.2 
 
 21 
 
 76 
 
 86 
 
 81 
 
 .0124 
 
 4 
 
 23 
 
 22 
 
 82 
 
 8'0 
 
 81 
 
 .0124 
 
 5 
 
 23.7 
 
 22.4 
 
 73 
 
 89 
 
 81 
 
 .0124 
 
 5.5 
 
 24.1 
 
 22.5 
 
 71 
 
 74 
 
 72.5 
 
 .0138 
 
 5.75 
 
 24.4 
 
 22.4 
 
 76 
 
 71 
 
 73.5 
 
 .0136 
 
 6 
 
 24.7 
 
 22.3 
 
 66 
 
 61 
 
 63.5 
 
 .0158 
 
 6.25 
 
 25 
 
 22.1 
 
 63 
 
 60 
 
 61.5 
 
 .0163 
 
 6.5 
 
 25.2 
 
 22.1 
 
 60 
 
 56 
 
 58 
 
 .0172 
 
 6.75 
 
 25.5 
 
 21.9 
 
 55 
 
 55.6 
 
 55.3 
 
 .0181 
 
 7 
 
 25.8 
 
 21.7 
 
 50 
 
 53 
 
 51.5 
 
 .0194 
 
 7.25 
 
 26.2 
 
 21.6 
 
 46.4 
 
 45.4 
 
 45.9 
 
 .0218 
 
 7.5 
 
 26.6 
 
 21.5 
 
 44 
 
 44.6 
 
 44.3 
 
 .0220 
 
 7.75 
 
 26.9 
 
 21.3 
 
 41 
 
 41 
 
 41 
 
 .0244 
 
 8 
 
 27.2 
 
 21 
 
 35.2 
 
 35 
 
 35.1 
 
 .0285 
 
 8.25 
 
 27.4 
 
 21 
 
 29.8 
 
 30.2 
 
 30 
 
 .0330 
 
 8.5 
 
 27.7 
 
 20.9 
 
 23.4 
 
 27.5 
 
 25.5 
 
 .0392 
 
 8.75 
 
 28 
 
 20.8 
 
 20.2 
 
 22.1 
 
 21.1 
 
 .0474 
 
 9 
 
 28.2 
 
 20.7 
 
 16.8 
 
 19.4 
 
 18.2 
 
 .0550 
 
 9.5 
 
 28.5 
 
 20.8 
 
 10.6 
 
 12.5 
 
 11.5 
 
 .0870 
 
 10 
 
 28.7 
 
 20.8 
 
 7.5 
 
 9.0 
 
 8.2 
 
 .122 
 
 11 
 
 29.2 
 
 21 
 
 1.8 
 
 3.0 
 
 2.4 
 
 .417 
 
 12 
 
 29.5 
 
 21.1 
 
 .58 
 
 .7 
 
 .64 
 
 1.563 
 
 detectable departure from the zero line was taken as the break point. 
 Thus, in run no. 4, Fig. 5, the break point comes at 9.3 volts. In order 
 to correct this we examined the velocity distribution curve for run no. 4 
 in Fig. 6. Here we saw that the uncorrected values (curve a) gave a 
 break point at V R VA = o and therefore V m VA o. Using this 
 value to get ( V m V R ) , and so to calculate the corrected values of the 
 current by applying equation 8, we got the curve b which has its first break 
 point at V R VA = i.o. We concluded, therefore, that the first 
 electrons which are effective in producing detectable radiation have a 
 velocity one volt less than the applied field, giving us 8.3 as the true 
 value of the radiating potential from this run. 
 
 Similarly for run no. 5 we found the observed break to be at 9.0 and 
 the corrected value to be 8.0 volts. 
 
VOL. XIV.1 
 No. S . 
 
 RADIATING POTENTIALS OF NITROGEN. 
 
 421 
 
 In run no. 6, Fig. 7, on the other hand, we found the corrected value 
 to come at 6.5 (5.9 observed), but there is a sharp increase in the slope 
 
 in volts 
 
 Fig. 5. Fig. 6. 
 
 of the curve at 8.6 (8.0 observed). This run therefore apparently has 
 two breaks corresponding to two critical speeds. 
 
 r 
 
 10 
 
 Fig. 7. 
 
 Fig. 8. 
 
422 
 
 H. D. SMYTH. 
 
 ("SECOND 
 
 [SERIES. 
 
 In runs nos. n and 12 the effect at the lower voltage was so great as 
 to make it impossible to detect any other. For these runs the corrected 
 values are 6.15 and 6.4 volts. 
 
 From these five typical curves shown, it is clear there are two distinct 
 critical points, each curve showing one or both, more or less sharply. 
 The values observed were weighted according to the sureness with which 
 the break point could be picked. The velocity distribution curves were 
 treated similarly and the weight of the corrected value taken as the 
 product of the weights of the two observations. 
 
 In Table III. the results from the curves shown and discussed above 
 are grouped with all other runs which gave results sufficiently definite to 
 
 have weight. 
 
 TABLE III. 
 
 Experimental Results, 
 
 No. 
 
 (mm.). 
 
 G. 
 
 Observed Breaks with 
 Weighting. 
 
 Corrections 
 with Weight- 
 ing. 
 
 Corrected Breaks with 
 Weighting. 
 
 1 
 
 .048 
 
 25. 
 
 9.4 (4) 
 
 
 -0.4 (3) 
 
 8.5 (12) 
 
 
 2 
 
 .038 
 
 25. 
 
 9.2 (3) 
 
 
 -0.9 (3) 
 
 8.3 (9) 
 
 
 3 
 
 .027 
 
 25. 
 
 9.3 (2) 
 
 7.5 (?) 
 
 -1.0(3) 
 
 8.3 (6) 
 
 
 4 
 
 ,015 
 
 25. 
 
 9.3 (3) 
 
 
 -1.0(3) 
 
 8.3 (9) 
 
 
 5 
 
 .015 
 
 250. 
 
 9.0 (4) 
 
 
 -1.0(3) 
 
 8.0 (12) 
 
 
 6 
 
 .015 
 
 2500. 
 
 8.0 (1) 
 
 5.9 (2) 
 
 +0.6 (3) 
 
 8.6 (3) 
 
 6.5 (6) 
 
 7 
 
 .015 
 
 2500. 
 
 
 5.4 (5) 
 
 +0.7 (2.5) 
 
 
 6.1 (8) 
 
 8 
 
 .045 
 
 750. 
 
 9.3 (2) 
 
 
 -.08 (4) 
 
 8.5 (8) 
 
 
 9 
 
 .045 
 
 140. 
 
 9.2 (1) 
 
 
 -1.0 (4) 
 
 8.2 (4) 
 
 
 10 
 
 .026 
 
 250. 
 
 8.8 (2) 
 
 7.4 (1) 
 
 -0.7 (5) 
 
 8.1 (10) 
 
 6.7 (5) 
 
 11 
 
 .01 
 
 500 
 
 4.75 (4) 
 
 + 1.4 (3) 
 
 
 6.15 (12) 
 
 12 
 
 .01 
 
 2.5 
 
 7.75 (?) 5.2 (4) 
 
 + 1.2 (3) 
 
 
 6.4 (12) 
 
 13 
 
 .01 
 
 40 
 
 6.5 (?) 4.2 (3) 
 
 + 1.9 (3) 
 
 
 6.1 (4) 
 
 Mean values 
 
 
 
 
 8.285 
 .045 
 
 6.285 
 .061 
 
 
 
 
 Note. Runs 7, n and 13 and some not given above show uncertain indications of a third 
 break at about 7.4 volts; G is the electronic current in amperes X io~ 9 . 
 
 V. DISCUSSION OF RESULTS. 
 
 Let us now compare the experimental results just presented with 
 Lyman's data on the spectrum of nitrogen in the extreme ultra-violet 
 given at the beginning of this paper. This is best done by writing 
 corresponding values opposite each other in a table. 
 
 I. The Break at 8.29 Volts. In this table we see that the effect which 
 we got at all pressures tried and with various currents, that ,is the most 
 
VOL. XIV.1 
 No. 5. 
 
 RADIATING POTENTIALS OF NITROGEN. 
 
 423 
 
 TABLE IV. 
 
 Spectrum Data. 
 
 Radiation Experiment Data. 
 
 Remarks. 
 
 X (Obs.). fb (Calc.). />o (Obs.). X (Calc.). 
 
 Doublet almost certainly due to nitrogen 1492.8 
 
 ; 1494.8 
 
 8.28 8.29 .04 1490.7 d= 10 
 8.27 
 
 Doublet attributed to nitrogen but pos- 
 sibly due to silicon '> 
 
 1742.7 1 
 
 7.08 
 
 7.3 (?) 
 
 1700 (?) 
 
 
 1745.3 
 
 7.09 
 
 
 
 Beginning of band spectrum i 
 
 1870.9 
 
 6.6 
 
 6.29 db .06 
 
 1965 20 
 
 intense effect, corresponds with greater accuracy than could be hoped 
 for, to the most certain doublet of the line spectrum of nitrogen. 
 
 2. Possible Effect at 7.3 Volts. The failure of the second value to 
 coincide with that for the other doublet is no greater than the uncer- 
 tainties of its determination. This radiation was apparently the weakest 
 of the three and, while showing up well on one or two curves was, on the 
 whole, rather doubtful. Attention should again be called to the doubt 
 concerning the origin of this doublet as determined by Lyman. 
 
 3. Explanation of 6.2p-Volt Break. Two different theories were de- 
 veloped to account for this effect, one depending on the application of the 
 quantum relation to the band spectrum and the other involving the idea 
 of dissociation and the subsequent production of spectral lines. They 
 are discussed at length in what follows. 
 
 (a) To explain the experimental result of a break occurring only at 
 low pressures and coming at 6.29 volts instead of the 6.6 calculated 
 from the band spectrum, we must consider the effect of a probable 
 impurity. Kreusler 1 found that nitrogen prepared in a manner almost 
 identical with that used in the present experiment had a small quantity 
 of N 2 O present as an impurity. He found that this increased absorption 
 at X = 1, 860 from 2.2 per cent, for atmospheric nitrogen tto 14.3 per cent. 
 He also made measurements on pure nitrous oxide and found 88.4 per 
 cent, absorption at X = 2,000, apparently increasing beyond his powers 
 of measurement at 1,930 and 1,860. 
 
 It is probable, therefore, that at the higher pressures, the radiation 
 X = 1,965 corresponding to 6.3 volts loses so much energy by absorption 
 in the I cm. space between the gauze and the plate that it is not detect- 
 able. This explains the fact that there was no trace of this break at 
 pressures above 0.026 and that it was only detected in one case at pres- 
 sures exceeding 0.015. 
 
 1 H. Kreusler, Ann. d. Phys., 6, p. 419, 1901. 
 
4 2 4 H. D. SMYTH. 
 
 The discrepancy between the value 6.3 found and the 6.6 corresponding 
 to the beginning of the band spectrum is not so easily accounted for. 
 It is a well known fact, however, that the presence of small impurities 
 greatly affects the intensity of emission spectra. We have just seen that 
 there is such an effect in the absorption spectrum of nitrogen in this 
 region. Is it not possible then that the presence of N 2 O may cause less 
 refrangible bands up to X = 1,965 to come out with sufficient strength 
 to be detected? The chances for sufficient experimental error to account 
 for a discrepancy of 0.3 volt do not seem great especially in view of the 
 extremely good agreement in the case of the highest break. It is hoped 
 that this point may be cleared up by further work taking every precau- 
 tion to get absolutely pure nitrogen. 
 
 (b) The other explanation of our 6.3 volt break is quite different. 
 There are lines in the spectrum of nitrogen in the region between 2,000 
 and 3,000 which would produce photoelectric effect and the most re- 
 frangible of which is X = 2,052.* In the previous discussion it has been 
 assumed that these arise from systems that are not present in this 
 experiment, such as charged atoms or molecules, since otherwise we would 
 have a photo-electric effect at lower voltages. If we allow the possibility 
 of the production of some of these lines by a neutral atom, as we must 
 for the doublet previously considered, then we must first have dissociation 
 a*nd then a radiating impact. Now the speed necessary for an electron 
 to produce these radiations is between 4 and 6 volts, but the energy 
 necessary to dissociate a nitrogen molecule is unknown. The velocity 
 necessary for an electron to dissociate a hydrogen molecule is found 
 by calculation from Langmuir's 2 results to be about 3.6 volts and it is 
 known that nitrogen is much harder to dissociate. It is possible therefore 
 that 6.3 is the speed necessary to dissociate the molecules and thus make 
 it possible for the electrons of lower speed to produce radiation. This 
 would correspond to a heat of dissociation of a gram molecule of nitrogen 
 of 145,000 calories, whereas Langmuir 3 found for hydrogen the value 
 84,000. 
 
 This theory necessitates a new explanation of the effect of pressure, 
 since these longer wave-length radiations will not be so strongly absorbed. 
 It must be supposed that the chances for recombination of the atoms 
 at higher pressures than 0.026 are so great as to make the radiation 
 negligibly small. 
 
 As to the soundness of the assumption of radiation from neutral atoms, 
 we have the following statement of J. J. Thomson regarding the lines of 
 
 1 Lyman, "Spec, of Extreme Ultraviolet," p. 83, 1914. 
 
 2 I. Langmuir, J. of Am. Chem. Soc., 37, pp. 417-458, 1915. 
 8 Ibid., p. 457. 
 
No L 's XIV '] RADIATING POTENTIALS OF' NITROGEN. 425 
 
 the hydrogen spectrum: "All theories concur in regarding the atom and 
 not the molecule as the source of these lines, but according to Wien's 
 theory the atom radiates when in the neutral state, while Stark maintains 
 that the radiation is emitted when the atom has a positive charge: 
 according to his view the lines emitted by the neutral atom are far away 
 in the ultra-violet." * Stark 2 found no lines attributable to a neutral 
 nitrogen atom but hardly carried his work below 4,000 A.U. From his 
 work on mercury, however, he concluded that the line 2,536.7 was due 
 to a neutral atom. 3 It is evident then that there is no evidence against 
 our assumption but rather indications of its probability. 
 
 A second assumption implied in this theory is that the neutral mole- 
 cules will not set up ultra-violet radiation when struck by electrons with 
 speeds below 6.3. On this point, I have found no evidence. 
 
 4. Upper Limit to Heat of Dissociation of Nitrogen. A necessary con- 
 sequence of the principle that the line spectrum is due to atomic nitrogen 
 is the determination of an upper limit for the energy necessary to dis- 
 sociate a nitrogen molecule. If the effect resulting rom bombardment 
 by electrons with velocities of 8.24 volts is due to atoms, the molecules 
 must be dissociated by the impact of electrons of this or lower speed. 
 The value 8.3 would give as an upper limit for the heat of dissociation 
 of a gram molecule of nitrogen about 190,000 calories. As has been 
 stated, Langmuir 4 found the value for hydrogen at constant volume to 
 be 84,000 calories. 
 
 VI. lONIZATION AT 18.5 VOLTS. 
 
 Goucher and Davis found that what had previously been called 
 ionization in nitrogen was really radiation but that true ionization did 
 occur at 18.5 volts. Although the apparatus used in the present investi- 
 gation was not well suited for testing this point, by greatly reducing 
 its sensitivity a distinct increase in the slope of the electrometer current 
 curve in the neighborhood of 18 volts was observed, thus supporting the 
 more accurate work of Davis and Goucher. 
 
 It is realized that the present results were not obtained under experi- 
 mental conditions ideal for getting sharp break points or reducing cor- 
 rections to a minimum. The attempt was made, on the other hand, to 
 employ experimental conditions in which the corrections would be as 
 varied as possible, in order to test the soundness of the formula. 
 
 1 J. J. Thomson, "Positive Rays," pp. 96-97, 1913. 
 1 J. Stark, Ann. d. Phys., 55, pp. 29-74, P- 73. 1914- 
 1 J. Stark, Ann. d. Phys., 52, pp. 241-302, p. 247, 1913. 
 4 I. Langmuir, loc. cit., p. 457, 1915. 
 
426 H. D. SMYTH. 
 
 Although the corrections in different tests differed by as much as 3 volts 
 in extreme cases, the corrected values of the break point were quite con- 
 sistent. It appears, therefore, that the above analysis is justified and 
 necessary, and that the final values obtained are trustworthy. 
 
 The author wishes to .express his indebtedness to Professor K. T. 
 Compton, whose supervision and assistance have made this work possible. 
 
 PALMER PHYSICAL LABORATORY, 
 PRINCETON, N. J. 
 July, 1919. 
 

 Return to desk from which borrowed. 
 Tins book is DUE on the last date stamped below. 
 
 U) 21-100'.9/47(A 5 702 S 16)476 
 
YC I 1021 
 
 QC 
 
 UNIVERSITY OF CALIFORNIA LIBRARY