THE UNIVERSITY OF MICHIGAN COLLEGE OF LITERATURE, SCIENCE, AND THE ARTS Department of Physics Technical Report No. 3 VIBRATION-ROTATION SPECTRUM OF MATRIX ISOLATED AMMONIA Robert E. Meredith ORA Project 03640 Contract No. AF 19g(604)-6125 Project 8603 Task 860301 under contract with: AIR FORCE CAMBRIDGE RESEARCH LABORATORIES OFFICE OF AEROS-PACE RESEARCH UNITED STATES AIR FORCE BEDFORD, MASSACHUSETTS administered through: OFFICE OF RESEARCH ADMINISTRATION ANN ARBOR November 1963

This report was also a dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in The University of Michigan, 19630

TABLE OF CONTENTS Page LIST OF TABLES iii LIST OF FIGURES iv ABSTRACT vi I o INTRODUCTION 1 Ao Properties of the Matrix 1 Bo Spectra of Trapped Radicals 2 II. EXPERIMENTAL DETAILS 8 IIIo RESULTS OF THE EXPERIMENT 20 IV. INTERPRETATION OF THE OBSERVED SPECTRA 36 A. Some Properties of the Energy Levels of Free Ammonia Molecules 36 Bo Interpretation of the v2 Band of Matrix Isolated Ammonia 43 Co Interpretation of the v3 Band 53 Do The vl Band 57 E. The Concentration Dependent Lines 58 V. ALTERNATIVE EXPLANATION 61 BIBLIOGRAPHY 65 ii

LIST OF TABLES Table Page 1. Line Positions for NH3 Isolated in Argon Using a Copper Substrate. 22 2. Line Positions for the Near Infrared Spectrum of NH3 Isolated in Argon. 23 3. Line Positions for the Near Infrared Spectrum of NH3 Isolated in Krypton. 25 4. Line Positions for the Near Infrared Spectrum of NH3 Isolated in Xenon. 25 5. Line Positions for the Near Infrared Spectrum of NDs Isolated in Argon. 33 6. Line Positions for the Near Infrared Spectrum of ND3 Isolated in Krypton. 33 7. Line Positions for the v2 Fundamental of ND3 Isolated in Argon. 34 8. Line Positions for the v2 Fundamental of NH3 Isolated in Argon Usinga.a Ki Substrate. 34 iii

LIST OF FIGURES Figure Page 1. Diagram of medium infrared spectrometer and fore optics. 10 2o Diagram of exit optics used with the thermocouple detector. 11 3. Diagram of exit optics used with the Golay detector. 12 4. Diagram of dewar tip. 16 5. Diagram of manometer system used for mixing and spraying gases onto the cold finger. 17 6. Spectrum of NH3 isolated in argon with M/A equal to 500 using KC1 (above) and Cu. as substrate. 21 7. Near infrared spectrum of NH3 isolated in argon for an M/A value of 20 (above) and 200. 26 8. Near infrared spectrum of NH3 isolated krypton (above) and xenon for an M/A value of 200. 27 9. Spectrum of V3 of ND3 isolated in argon for M/A equal to 30 (above) and 200. 30 10o Spectrum of vl and 2v4 of ND3 isolated in argon for M/A = 30. 31 11. Spectrum of v2 of ND3 isolated in argon for M/A = 500. 32 12. The lower lying rotational levels of NH3 in a totally symmetric vibrational state. 38 130 The lowest lying rotational levels of NH3 in a doubly degenerate vibrational state. 39 14. Rotational levels of v2 of NH3 showing the effect of inversion doubling. 42 15. Transitions predicted for v2 of NH3. 45 iv

LIST OF FIGURES (Concluded) Figure Page 16o (a) Predicted gas phase spectrum of "A type" NH3, (b) predicted gas phase spectrum of "E type" NH3, (c) composit "A" and "E type" spectra predicted, (d) lines observed in the v2 region for NH3 isolated in argon using a.:Kl substrateo 47 17. Transitions expected for the v3 band of NH3. 54 18. (a) Predicted lines for the v3 band of NH3, (b) observed lines for the v3 band of NH3. 55 19. Energy levels for a harmonic librator. 62 v

ABSTRACT An investigation of NH3 and ND3 isolated in inert gas matrices at 4.2~K has been made in the regions of the vl, v2, and V3 fundamentals, and in the vicinity of the overtone 2v4. The spectra were scanned with spectral slit widths ranging from 1l to.5 cm1, and it was found that this resolution was adequate to insure that all the observed line widths and shapes were free from instrumental broadening. The v2 fundamental has been interpreted in terms of a free rotation model, with the line spacings and inversion splitting having values very close to those observed for gas phase molecules. This interpretation confirms the existence of "E" type and "A" type (i.eo, para and ortho) ammonia, although a substantial amount of nuclear spin conversion from "E" type to "A" type has been found. The interpretation of the vl, V3, and 2v4 spectra was much more difficult, since these bands were too weak to permit observation of the "E" type spectra, and since the inversion doubling of the levels involved cannot be observed due to the broadness of the lines. It was possible, however, to determine the dependence of the lines on the ratio of inert gas atoms to ammonia molecules, and to assign the transitions involved as arising from either single isolated molecules or from ammonia complexes caused by incomplete isolation. Lines were also observed which were attributed to molecules occupying different sites in the matrix, Argon, krypton, and xenon were successively used as matrices, and spectra obtained in each case differed only in that the vibrational frequencies were shifted to longer wavelengths as the matrix was varied, A considerable amount of experimental difficulty was encountered in obtaining spectra using krypton and xenon. This was caused by an increased scattering of the infrared radiation by these matrices, and by a weakening of the lines. vi

I. INTRODUCTION Matrix isolation —the technique of embedding atoms or molecules under study in an otherwise inactive solid medium-was first used to facilitate the study of free radicals. Lewis and Lipkinl reported the successful isolation of a number of organic radicals dispersed in organic solvents. In their work, the radicals were formed by ultraviolet irradiation of the frozen mixture containing the parent molecules. Subsequent workers2'3'4 developed a common practice of producing the radicals in an electrical discharge, and having the gas mixture flow out through a leak into a high vacuum where it hit and condensed on a cold surface whose temperature could be controlled. The condensate was investigated by its emission and absorption spectra. 5,6 A program5' on the study of trapped radicals centered at the National Bureau of Standards from 1956 to 1960 greatly increased the knowledge and understanding of matrix isolation effects. Points relevant to the present work are summarized in the following two paragraphs. A. PROPERTIES OF THE MATRIX The diffusion rate of a radical in a matrix depends strongly upon the size of the radical. The diffusion rate for large radicals is determined almost completely by the ratio of the temperature to the melting point of the matrix-the diffusion becoming rapid when the temperature approached one-half the melting point temperature. Small radicals on 1

2 the other hand were found to diffuse out of the matrix even when the temperature was as low as one-tenth its melting point.7 X-ray diffraction studies8 gave accurate information concerning the structure of condensed gases. Crystals deposited at various rates of speed were found to show a long range order when deposited slowly. Annealing the solid increased the ordering. In all cases it was found that solids formed by condensing gases onto a cold surface were of poor crystal quality, suffering from dislocations, holes, etc. Livingston9 and co-workers concluded that the more imperfect the crystal (or more nearly amorphous), the larger the number of different sites at which impurities could be trapped. X-ray diffraction measurements on condensed 8 argon by Peiser indicated a surprisingly high density of dislocations. Moreover, he observed a slight orientation of the 111 plane parallel to the deposition surface. B. SPECTRA OF TRAPPED RADICALS The spectra studied involved electronic transitions exclusively. Usually, most of the absorption or emission lines could be correlated with the vapor phase spectrum of the radical, with only small frequency shifts arising from the perturbation of the matrix on the impurity. Large shifts were attributed to either polymeric species, or to an entirely different radical. A sizeable number of experiments showed the vibrational structure of the electronic bands. Robinson and McCartylO observed some 20 vibrational transitions in the absorption spectrum of

35 NH2 trapped in solid argon. They observed very small matrix shifts of the vibration frequencies, and the spacings between lines were almost unaltered. These authors also investigated HNO in an argon matrix, discovering again that vibration spacings were nearly the same. These and other workers have concluded that inert gas matrices perturb the vibrational motion of trapped species only slightly. Robinson and McCarty also interpreted the fine structure of their electronic-vibronic bands of NH2 and NH as evidence of free rotation in argon matrices. The rotation constants attributed to this structure were identical with those of the free gas species. Pimentel initiated the study of matrix isolated stable molecules. The usual method of preparation involved spraying the gas mixture through a capillary leak onto a cold Surface. Becker and Pimentell2'13 in a detailed extension of the initial work, investigated the effect of temperature, different matrix materials and active molecules, and the ratio of the number of matrix to active species concentrations (designated by M/A). Chemical inertness, rigidity, low volatility, and transparency in the spectral region of interest were necessary properties of the matrix. Isolation was best achieved at low temperatures. The inert gases and nitrogen were found to be the most suitable matrix materials. They concluded that the matrix lattice is sufficiently rigid to produce isolation as long as the temperature is kept below particular values: 70~K for xenon, 35~K for nitrogen and 30~K for argon. The concentration of the active molecule also had a decided influence on the stability of the

deposit. In general, it was difficult to form deposits with M/A values less than 50. In addition, they found that the intensity of some adsorption lines exhibited a strong dependence upon the concentration of the active molecule-varying directly as the concentration. These lines were attributed to associated active molecules, and they gave evidence of not only dimer formation, but also trimers, tetramers, and polymers.l4 The particular work that led to the investigation here was that reported by Hexter, Milligan, and Dressler15 at the Fourth International Symposium of Free Radical Stabilization held at the National Bureau of Standards in 1959. They had measured the absorption due to the v2 fundamental of NH3 in the 10 micron region for NH3 isolated in argon on a AgCl window held at 4~K and had found some eight lines. The large number of transitions observed seemed to be incompatible with the low temperature at which more than 99X of the molecules would be expected to be in the J = 0 level. Their interpretation, which involved the violation of the selection rule AJ = -1 while retaining the rule AJ = +1, was that the inversion doubling of the v2 vibration had decreased from 36.5 cm-' in the free molecule to about 8 cm-1, and that rotation of the molecule in the matrix was possible only for states having K = J. That is, the molecule could rotate freely about its symmetry axis, but other rotations about other axes were hindered by the matrix. The large number of transitions were explained as being Q(1), R(0), R(1), and R(2) lines. The difficulty in observing transitions from the J = 2 level of molecules at or even close to 4.2~K was not mentioned, although the absence of the correspond

5 ing P branch lines were explained by assuming that K is allowed to be equal to J or very nearly equal to J. The observed spectrum was so far at odds with what would have been anticipated that it seemed worthwhile to look into the problem further. With the spectroscopic facilities of this department several significant advantages could be realized: (a) better resolution, (b) reduction of the implied temperature differential between the sample and the dewar by placing the sample in the exit optics of the spectrometer so that it is illuminated by only the radiation in the band pass of the spectrometer, (c) the possibility of examining the absorptions due to the other fundamentals of the molecule. Becker and Pimentr112 had been the first to publish results of a matrix isolated NH3 investigation. Four strong bands in the 1000 cm-1 region were reported for NH3 isolated in argon. Although no assignments were made, it was concluded that some of the lines observed arose from incomplete isolation of the NH3 molecules. Since starting the work here, Pimentel, Bulanin, and Van Thiell6 have observed the absorption for all four fundamental vibrations of NH3 isolated in nitrogen. They concluded that the rotation of NH3 was strongly hindered, and that there was no inversion doubling in v2. They pointed out that this and other spectroscopic evidence from other molecules indicated that nitrogen matrices trap impurities in a much more restrictive "cage" than noble gas matrices. There are some very striking qualitative similarities between the results of the present work and

6 that of Pimentel et al., even though the latter used nitrogen as a matrix. In particular, Pimentel observed lines which increased in intensity as the M/A ratio was increased. These lines were explained on the basis of a polymer picture, whereas, it will be seen that this work seems to indicate that more than one site may be available to NH3. Considerable work has been done in the interval since 1959 in other laboratories on the matrix isolation spectra of other simple molecules. Schoen17 et al., have investigated HC1 in argon matrices at 40K and 20~K, and have interpreted their results in terms of a hindered rotation of the HC1 molecule. Three absorptions were observed and labeled P(1), R(O), and R(l), in analogy with the notation used for a free rotor. The spacings of the lines were much smaller than observed for HC1 in the gas phase, but for a reasonable value of their hindering potential, the spacing 4B between the P(l) and R(O) lines in the matrix can be accounted for. The spacing 2B between the R(O) and R(1) lines, however, was much smaller than the model predicted. Glase18 and Redington9 et al., have presented evidence of free rotation of water in inert gas matrices. Frequency shifts from gas phase values were measured, and the rotational spacings were found to be almost unchanged in the matrix. An interesting result of their investigations is a proof of the existence of two modifications of matrix isolated water-para and ortho. The para modification has the nuclear spins of the two identical nuclei pointing opposite one another, and the ortho has the spins parallel. It was estimated from the data that the iso

7 lated water molecules were 75% ortho and 25% para. Lines which did not show rotational structure were also observed and were explained as arising from molecules trapped in interstitial sites. The existence of matrix isolated ortho and para species has received considerable attention in recent years. Besides H20, matrix isolated NH2, the methyl radical, and NH3 have been investigated in order to see if ortho and para modifications are both contributing to the spectra. Robinson et al., have concluded that for matrix isolated NH2, almost 100% conversion of the para nuclear species to the ortho has taken place. On the other hand, electron spin resonance measurements on matrix isolated methyl radicals indicated strongly that no such nuclear conversion is taking place. Since in the absence of a spin dependent perturbation a conversion of nuclear spins to one type cannot take place, no adequate explanation for the observed effect has been presented. K. Dressler2O has speculated that the nitrogen atom in NH2 has perturbed the spin functions enough to eliminate all traces of the para molecules. He observed a series of bands in the 1600 to 1900 A region in matrix isolated NH3 and attributed this structure to transitions of the symmetric bending vibration v2. The absence of alternate lines in the structure was interpreted as! arising from an almost 100% conversion from para ammonia to ortho.

II. EXPERIMENTAL DETAILS The problem faced was to determine and account for the fine structure of the vibrational absorption bands of NH3 in rare gas matrices. NH3 has four fundamental frequencies: two non-degenerate parallel type of species Al, vi near 3335 cm-1, and v2 near 950 cm-l, and two degenerate ones of species E, v3 and v4 near 3440 and 1625 cml- respectively. It was difficult to work in the 1600 cm-1 region because of low detector sensitivity and strong water vapor absorption and no observations were made there. It was thought that valuable information could be gained by also obtaining the spectrum of ND3 whose corresponding frequencies lie at 2420, 785, 2555 and 1190 cml —all of which were in accessible regions of the spectrum. The overtone 2v4 lies in the region of vl for both molecules and complicates the situation there somewhat. With respect to matrix materials, the rare gases were selected because of the simplicity of their structure. Of these neon is not satisfactory because of the rapid diffusion of other molecules through the matrix. Most of the deposits were made with argon, and some with krypton and xenon. The basic components of the apparatus involved were: (1) the spectrometer, (2) the low temperature dewar, and (3) the system for preparing the samples. 1. Two spectrometers were used in the course of this work. The symmetric bending vibration v2 of both NH3 and ND3 was investigated with 8

9 a one meter focal length F-5 spectrometer built by Hardy in 1930. This instrument makes use of 8 in. paraboloidal mirrors in a Pfund type mount, and has excellent imaging properties when used in the medium infrared. It utilizes a 6 in. by 8 in. grating. A diagram of the optical components is shown in Figure 1. M1 and M4 are the 8 in. paraboloidal mirrors and M2 and M3 are 8 in. flats which are pierced so that the beam may pass through the entrance and exit slits S1 and S2. A 2400 line/in. grating (G) was used for all measurements made with this instrument. Radiation from a Nernst glower (S) was collected by a mirror ml and rendered parallel by a mirror m2. The parallel beam passed through a KBr littrow prism (P) and then focused on the entrance slit by the mirror m3. This arrangement supplied the desired F-5 beam to the spectrometer proper with no overlapping orders when used in the 10 to 12 micron region. Two sets of exit optics were used for the medium infrared measurements. For spectra of crystals deposited on a copper surface (see Chapter III) it was found most convenient to use the copper substrate as one of the component mirrors. m4 and m5 (see Figure 2) demagnify the image by a factor of two and throw it beyond m5 where there is enough room to place the copper "mirror," m6. After reflecting off the copper, the beam is demagnified again by a 5:1 elliptical mirror and focused onto a thermocouple. The other set of exit optics shown in Figure 3 differed in that the beam was focused through a transparent window being demagnified by m4 and m5, and was focused onto a Golay cell off of a flat me and an elliptical mirror, mg. The radiation was chopped at 10 cps, and

M4I I I Sim FMi M2 m o Figure 1. Diagram of medium infrared spectrometer and fore optics.

11 m4 52 6 m7 Figure 2. Diagram of exit optics used with the thermocouple detector.

G M 4 M 5 S 2 mg of exagtramt optics used with the Golay detector figure 3- Diagram

13 the signal from the detector was amplified, synchronously rectified and recorded. The measurements of the vi, v3, and 2v4 bands were carried out with a 3 meter focal length f/15 ebert vacuum spectrometer designed by H. W. Marshall. Its spectral band pass is customarily less than 0.1 cm-l in the near infrared where photoconductive detectors are available. A 6-in. by 8-in. 300 line/mm grating was used in first order in scanning the NH3 spectra, and a 180 line/mm grating of the same size was used for the ND3 spectra. In all cases appropriate interference filters were used to remove higher spectra order radiations. A Nernst glower was used as source, and a simple set of condensing optics was used in order to supply the spectrometer with an f/15 beam and 2 cm high image. Auxiliary optics in the main spectrometer tank throw an image out far enough to permit passage through a transparent substrate. The beam is then demagnified by a five to one elliptical mirror onto a photoconductive detector placed at the tip of a glass dewar that could be cooled to liquid nitrogen temperature in order to achieve maximum detector sensitivity. The NH3 spectra were scanned using a p-type lead sulfide detector, and the ND3 spectra were scanned using an R-type lead selenide detector, both made by the Eastman Kodak Co. The entire optical path is enclosed in a vacuum housing, and strong atmospheric absorption in the 2v4 and vl region of ND3 made it necessary to pump out the entire system except for the source housing, which was flushed with dry nitrogen. The infrared beam was modulated by a three bladed chopper driven

14 by an 1800 rpm synchronous motor and the signal produced in the detector was amplified by an amplifier tuned to 90 cps. The chopper was also used to modulate a light beam incident on a phototube, thus generating a synchronizing signal. This signal was used to synchronously rectify the infrared signal for display on a chart recorder. 2. The low temperature dewar used for the preparation of the matrix isolated samples except in one case was a typical metal double dewar designed to hold approximately 1.2 liters of liquid helium in the inner container. The machine work was carried out in the Physics Department instrument shop, and the parts were soldered together in the Chemistry shop. The inner containers were made of 1/32 in. brass and were suspended from the outer housing by 1/64 in. stainless steel tubing. A dewar tip in which the cold window was mounted projected downward from the ~helium pot. One-sixteenth in. copper shields were attached to the liquid nitrogen container so that the inner container was completely protected from the room temperature except for a very narrow slit through which the infrared beam and the depositing ammoniarn.(bble gas mixture passed. The inner containers were also polished and chrome plated to minimize radiation loss. The dewar was evacuated before each run by an oil diffusion pump. Through the course of the experiment, it was found that the dewar was capable of holding liquid helium for as long as 12 hr. In as much as most deposits were sprayed on very slowly, it was important that the dewar be at least this efficient. The crystals were deposited by leaking the gas mixture through a

15 10 mil I.D. stainless steel tube which was pointed through the outer dewar housing toward the cold window. The window (sapphire for all near infrared measurements) was forced firmly against the dewar tip in order to be in good thermal.-contact. The dewar tip was of a somewhat special design, and is shown in Figure 4. It was felt to be of importance that the temperature be close to 4.2~K, so the copper tip was made to be hollow all around the window, permitting liquid helium to completely surround it at all times. The window was made flush by recessing the tip, and a snuggly fitting collar was fitted against it to hold it in place. Several checks were made of the thermal contact between the window and copper surfaces. The contact was first coated with a thin layer of vacuum grease, then again with a mixture of vacuum grease with fine copper fillings, and finally with nothing at all. No differences were observed in the spectra of NH3 isolated in argon for the three cases, and it was felt that only a good clean contact was needed. 3. The ammonia and noble gases were mixed in a manometer system shown schematically in Figure 5. The mixing took place in a one liter glass bulb. A mercury manometer attached to the bulb was used to read the pressure of the mixture, and a smaller differential manometer was used to assist in arriving at the desired ratio. The mixing was carried out in the following way:'the differential manometer stopcock was turned to the "open" position and the bulb was pumped out. The stopcock was then closed and ammonia was slowly leaked into the system. The difference between the two columns of the differential manometer accurately

3/4". D. I///I // // 1/16"COLLAR SUBSTRATE 1/16" /C i-~-iHELIUM C3/8P Figure 4. Diagram of dewar tip.

17 Figure 5. Diagram of manometer system used for mixing and spraying gases onto the cold finger.

18 measured the pressure of the ammonia let into the bulb. The stopcock was then turned back to the "open" position, and the desired amount of inert gas was leaked in as measured by the mercury manometer. The liquid used in the differential manometer was chosen to be octoil-s because of its low specific gravity. This made the measurements of ammonia to noble gas ratios much more accurate than could be obtained using only the mercury column. The second section of the manometer system consisted of a smaller bulb and another differential oil manometer. This part served as a let down valve so that a uniform rate of deposition could be established. By first evacuating this section and turning its manometer to the off position and then cracking the stopcock leading to the large bulb, it was possible to maintain a constant pressure of gas mixture entering the low temperature dewar. After a little practice, it was possible to let the mixture leak out of the bulb at a constant rate for several hours. The "let down section" entered into a glass tube leading to the low temperature dewar. The 10 mil stainless steel tube mentioned earlier was fused into the glass tube completing the path from the mixing bulb to the cold window. The hole in the dewar through which the stainless steel tube was aimed was placed as close to the outer window as possible and was approximately 1-1/4 in. from the cold window. The direction of the gas as it sprayed onto the window was not perfect at first, but after several adjustments of the tube the spray seemed to be quite well centered on the window.

19 A spectrum of matrix isolated NH3 was obtained also of a sample deposited on a copper substrate. The dewar used in this case was made of glass and had only one inner container, and had a copper container at its tip. One side of the container was a highly polished flat piece of copper. Ammonia was successfully isolated on this surface, and although the dewar was capable of holding helium for only 2 hr. it was used to observe the first spectra. This copper substrate is of course the copper "mirror" mentioned in the discussion of the exit optics for the medium infrared instrument.

III. RESULTS OF THE EXPERIMENT The work of Hexter et al., on the V2 band of NH3 in argon and nitrogen had just been reported, the results of which were surprising enough to warrant an investigation using higher dispersion. The glass dewar referred to in the previous chapter was modified for this use, and a spectrum of NH3 isolated in argon at a dilution of M/A = 500 was scanned. This spectrum is shown in Figure 6, and the line positions are indicated in Table 1. When compared with the spectrum of Hexter et al., several differences were noted. First, Hexter observed a line at 961 cm-1 and did not observe one at 1032 cm-1. The present investigation does not indicate an absorption at 961 cm-' but does find one at 1032 cm-1. Also, the three lines near 975 cm-l are much better resolved than those in the earlier work. The spectral slit width at which the spectrum in Figure 6 was scanned was approximately 0.2 cm-', so it is felt that the half widths listed in Table 1 are accurate measurements of the natural widths of the lines. A number of attempts were made to bring out a line at 961 cm-1, but with no success. In general, however, Hexter's observations were substantiated, and no light was shed on the purplexing problem of how so many transitions could arise from NH3 molecules that must certainly have been in the ground state at liquid helium temperature. It was decided to examine the other fundamental vibrations of NH3 accessible in the near infrared, vl and v3, which would exhibit a much different inversion splitting, if 20

21 I I I I I I 940 980 1020 CM-I Figure 6. Spectrum of NH3 isolated in argon with M/A equal to 500 using KC1 (above) and Cu. as substrate.

22 TABLE 1 LINE POSITIONS FOR NH3 ISOLATED IN ARGON USING A COPPER SUBSTRATE Frequency, cm-1 Assignment Half Width, cm-1 955 Q(l) 1.4 972 Polymer --- 975 R(O) 2.2 981 Site 1.1 990 Q(1) 1.2 998 R(l) 1.6 1012 Weak --- 1017 Site 1.7 1032 R(l) 1.6 1037 Weak --- nothing else. The new metal dewar was used with the three meter spectrometer to obtain transmission spectra in this region of NH3 isolated in argon. The first measurements were made primarily to ascertain the dependence of the spectra on concentration. Sample spectra of NH3 isolated in argon are shown in Figure 7 at ratios M/A = 200 and M/A = 20 respectively. The lines which differ in dilute mixtures are attributed to polymeric species, and they are indicated in Table 2 as type p. Other lines are attributed to monomeric NH3, and are designated by type m. Table 2 also gives the half widths of the sharper lines, along with the most reasonable vibrational assignment. The spectral slit width used in these scans was approximately.2 cm1, even though the instrument is capable of a resolution of.03 cm-l in this region. It was found that the argon deposit scattered the infrared radiation quite severely, and the loss of

23 TABLE 2 LINE POSITIONS FOR THE NEAR INFRARED SPECTRUM OF NH3 ISOLATED IN ARGON Species Frequency, cm-1 Assignment (Monomer or Half Width, cmPolymer) 3211.6 2v4 m 3234.9 2v4 p 3242.9 2v4 m 1.40 3305.5 V1 p 3310.2 vi m 2.24 3324.2 vl p 3326.2 vl m 3345.7 vl p 3391.5 vs p 3398.7 Vs m 2.8 3433 0 V3 m Broad 3436 0 3s m Broad 3445.0 V3 P Broad energy necessitated opening the slits considerably. The region was scanned with the slits set at approximately.1 cm-~ and again at.2 cmlbut no broadening of any lines were observed. From this it was concluded that the resolution was quite sufficient to assure the measurement of natural line widths. The rate of deposition of the gas mixture was also varied. There was no noticeable difference in the spectra of films grown both slowly and very quickly from the standpoint of line position or strength. It was quite apparent that scattering from the crystals grown very slowly was not as severe as from those grown very rapidly. It was anticipated that if light scattering were to become a problem in subsequent experiments, crystals would have to be grown very slowly. It was also possible

24 to shine a flashlight onto the deposit and view it directly. The only obvious difference in the physical appearance of solids grown at different rates of speed was their transparency. Solids grown very quickly had a white milky appearance, whereas solids grown slowly for several hours had the appearance of a film of ice. NH3 isolated in krypton and xenon was investigated nexto The spectrum of NH3 in krypton for M/A = 200 and NH3 in xenon at M/A = 200 is shown in Figure 8. The line positions and designations as to type are given in Tables 3 and 4- All the gross features of the NH3 spectra in argon, krypton, and xenon are the same, as should be expected. A substantial shift of the line positions toward lower frequencies is evident in going from argon to krypton and to xenon, and the separation between groups of lines is approximately the same. A narrowing and splitting of the broadest argon absorption occur in the krypton and xenon spectra, but this just indicates that the absorptions in question consist of a number of relatively broad lines. Perhaps the most interesting comparison between the three sets of data is the relative strengths of the absorption lines. The apparent weakening of the lines in krypton and xenon as seen in Figure 8 is a real effect. When the number of absorbing molecules on the cold window are taken into account, it turns out that the line strengths of NH3 isolated in xenon are more than an order of magnitude weaker than for NH3 isolated in argono To make matters worse experimentally, the scattering loss is also much greater for krypton and xenon matrices. In

25 TABLE 3 LINE POSITIONS FOR THE NEAR INFRARED SPECTRUM OF NH3 ISOLATED IN KRYPTON Species Frequency, cm'1 Assignment (Monomer or Polymer) 3206.8 2V4 m 3231.2 2v4 P 3236.6 2v4 m 3298.8 vl P 3302.5 vl m 3313.9 v1 P 3317.3 vl m 3334.2 vl p 3384.4 vs p 3392.7 V3 m 3430.7 V3 m 3436.0 vs P 3444.0 vs p TABLE 4 LINE POSITIONS FOR THE NEAR INFRARED SPECTRUM OF NH3 ISOLATED IN XENON Species Frequency, cm-l Assignment (Monomer or Polymer) 3202.2 2v4 m 3228.4 2v4 p 3231.6 2v4 m 3294.0 vi p 3297.0 vi m 3304.6 vi p 3306.8 vi m 3323.8 vi P 3380.3 V3 p 3388.8 V3 m 3413.6 V3 m 3424.6 V3 p

III I I I I I I I I I III I I I I I I I I I 3200 3240 3280 3320 3360 3400 3440 CM' Figure 7. Near infrared spectrum of NH3 isolated in argon for an M/A value of 20 (above) and 200.

I I I I I I I I I I I I I I I I I I I I I I I I I 1 3200 3240 3280 3320 3360 3400 3440 CM' Figure 8. Near infrared spectrum of NH3 isolated krypton (above) and xenon for an M/A value of 200.

28 order to bring out the krypton and xenon lines as shown in Figure 8, it was necessary to grow the crystals for approximately 15 hr. This was accomplished by transferring liquid helium twice before taking down the final scans. The spectral slit width of the instrument during these scans was approximately.5 cm-l. Energy seemed to be scattering so badly that the two pass alignment was taken out, and the slits were opened almost as far as they would go. A mixture of nitrogen and NH3 was also prepared and sprayed onto the sapphire window held at 4.2~K. In a very short time compared to that taken for INNH3-argon mixtures, strong absorptions were observed in the vicinity of 3300 cm-l and 3400 cm-l. There did not seem to be any direct correlation between the NH3-argon spectra and that of the NH3-nitrogen mixture, but there seemed to be only a negligible loss of energy due to scattering in the latter case. It seemed clear also why the spectrum of NH3 in the 10 micron region is so readily obtained. In the first place scattering of 10 micron radiation should be much less severe than for 3 micron radiation, and even more important, the symmetric bending motion has a much stronger dipole change than the other fundamentals. The spectrum of NH3 isolated in nitrogen was the same as that of Pimentel et al., so it was not felt necessary to include it here. In general, the spectra of vl and v3 were in marked contrast to that of v2, having only a few components and showing some broad band absorption. These spectra of the high frequency fundamentals could also not be interpreted in terms of the free molecule energy levels, nor was there

29 a recognizable correlation between the spectra of v1 and v3o A study of the isotopically substituted molecule ND3 was undertaken in an attempt to examine the dependence of the spectra on the mass and moment of inertia of the molecule. Much more difficulty was encountered in obtaining the near infrared ND3 spectra. Not only was the lead selenide detector less sensitive than the lead sulfide, but it was not possible to open the slits wider than approximately 400 microns. It was possible to obtain spectra of ND3 in argon and krypton at several concentrations in the v3 region, but the detector sensitivity fell off so rapidly below 2500 cm-1 that the 2v4 and v1 bands were scanned only for an argon to ND3 ratio of 30. This was not considered a serious matter inasmuch as the extensive NH3 data had shown a very definite pattern as regards the identity of the "monomer" and "polymer" spectra. It is quite evident that for vl, v3, and 2v4, a single narrow "monomer" line always has a single narrow "polymer" line approximately 5 wave numbers toward lower frequencieso This is verified also for v3 of ND3 in argon in Figure 9o For vl and 2v4 of ND3 it is natural to designate the low frequency components of the two pairs of lines as "polymer" lines. The spectrum is shown in Figure 10, and the line positions and designations are given in Tables 5 and 6o It was decided next to scan the V2 band of ND3, This was done with the one meter infrared instrument The deposit was formed on a KC1 windowo The spectrum of the V2 band of ND3 in argon for M/A = 500 is shown in Figure 11, and the line positions are given in Table 7~ The

30 I I I I I 2510 2530 2550 CM' Figure 9. Spectrum of v3 of NDs isolated in argon for M/A equal to 30 (above) and 200.

I I I I I I I 1 2340 2360 2380 2400 2420 CM' Figure 10. Spectrum of Vi and 2v4 of ND3 isolated in argon for M/A =- 30 yy hi

I I I I I 730 750 770 CM-I Figure 11. Spectrum of v2 of ND3 isolated in argon for M/A = 500.

33 TABLE 5 LINE POSITIONS FOR THE NEAR INFRARED SPECTRUM OF ND3 ISOLATED IN ARGON Species Frequency, cm"- Assignment (Monomer or Polymer) 2519 3 V3 P 2527~3 V3 m 2548~8 V3 m 2552,2 V3 m 2555~7 V3 P 2346,0 2v4 m 2373 0 2V4 m 395 0 vi p 239800 vi m 2407~0 vi p 2411o0 vl m TABLE 6 LINE POSITIONS FOR THE NEAR INFRARED SPECTRUM OF ID3 iSOLATED IN KRYPTON Species Frequency, cm-1 Assignment (Monomer or Polymer) 251508 V3 P 2523 0 w m 2543o l V3 m 2549,0 V3 m

34 TABLE 7 LINE POSITIONS FOR THE v2 FUNDAMENTAL OF ND3 ISOLATED IN ARGON Species Frequency, cm'- Assignment (Monomer or Polymer) 7440 Q(1) 753.9 R(O) m 763.7 R(l) m TABLE 8 LINE POSITIONS FOR THE v2 FUNDAMENTAL OF NH3 ISOLATED IN ARGON USING A KC1 SUBSTRATE Frequency, cm-1 Assignment Half Width, cm-' 957.1 Q() --- 976.4 R(O) 7.5 99200 Q(1) --- 999.0 R(1) - 1034.1 R( 1)

spectral slit width used for the measurements was 1/3 cm-. The simple three-line spectrum for V2 of ND3 was completely different from that mentioned earlier for v2 of NH3. A rerun of v2 of NH3 was then tried with the same KCi window as for ND3. This spectrum for an argon to NH3 ratio of 500 is shown in Figure 6, so that a comparison with the spectrum taken earlier can be most conveniently made. Some very striking differences in the two spectra are apparent, and this will be discussed further in Chapter IV. It is evident that the deposition process and/or the substrate used has a very important effect upon the spectra of matrix isolated ammonia.

IVc INTERPRETATION OF THE OBSERVED SPECTRA Ao SOME PROPERTIES OF TH ENERGY LEVELS OF FREE AMMONIA MOLECULES The interpretation of the spectra of molecules freely rotating or nearly rotating as an isolated impurity in a matrix at low temperatures requires a detailed knowledg.e of the energy levels and wave functions of the unperturbed molecule In the case of NH3 and ND3 it is of value to review the following items: (a) The symmetry properties of the rotational levels; (b) The influence of nuclear spin and statistics on these levels; and (c) The sizable doubling of the energy levels of v2 of NH3 caused by the relatively low barrier to inversion of the nitrogen atom through the plane of the three hydrogen atoms. lo The energy of vibration and rotation of a symmetric top molecule for a nondegenerate vibration is: E= G(v... ) + B(JT+l) + (A-B)K2 where G(vV2l2V0~) is the vibrational contribution and is given by Herzberg vo I, po 400. The energy of vibration and rotation of a symmetric top molecule for which one degenerate vibration is excited is given by E = G(viy ooo) + BJ(J+I) + (A-B)K2 + 2AjiK where the term +2A1iK is a correction arising from the interaction of the 36

37 angular momentum due to the ith degenerate vibration with that which is due to rotation about the symmetric top axis. Si has values 0 < j1i < 1. The values of ai for NH3 and ND3 are: NH3 53 =.06 54 = -0.26 ND3 ~3 = o02 =4 = -.36 The ground state rotational constants are: NH3 B = 9.94 cm-1 A = 6.3 cm- ND3 B = 5o1 cm l A = 3515 cm-1 The rotational energy levels for NH3 and ND3 for both parallel (nondegenerate) vibration-rotation and perpendicular (degenerate) vibrationrotation levels are given in Figures:12and 13. With the exception of the T2AtiK correction to the degenerate vibration-rotation levels, the spacings are drawn to scale. Since the symmetry of NH3 and ND3 is C3v, the rotational subgroup is C3, with irreducible representations A and E. The symmetries of the levels shown are the symmetry of the total wave function excluding nuclear spin Th= eve Vr The ground electronic state of an NH3 molecule is totally symmetric. The symmetry of f is thus the product of the vibrational and rotational symmetries. For 4v totally symmetric, the levels have symmetry A or E,

38 J 3 A 3 E 3 E 2 - A 2 -.E 2 E I -A I -E 0 A KTO Kal K~2 Figure 12. The lower lying rotational levels of NH3 in a totally symmetric vibrational state.

J J J J 3 E 3I/- E A ~ 3. ~ E I ~ E _______^ ~E ^0 E K=0 K=l K=2 K=3 Figure 15. The lowest lying rotational levels of NH3 in a doubly degenerate vibrational state.

40 depending on whether the rotatlonaa.l wave fun-ction has symmetry A or E. Si.nce tr = o()JKM p iM G ~Kd 1r is of type A for K = 3q, (q = O,1,2,. ) and type E for K = 3q~l as shown in Figure- 12: If ir is of type E,'must be of type ExA = E for K = 5q or ExE = E+2A for K = 3q.lo For t.he- case K = 3q, there are two levels of type E (except for K 0) for each J, and for the case K = 3ql there are two of type A and. one of type E, as in Figure 13. 2. For NH3 and ND3 and other molecules having identical nuclei of nonzero nuclear spin I there is a nuclear spin wave function TN that must be included i. l the total wave function, ioeo, e= etvrtN The symmetry of $ is thus not given by the product vlr, as in Figures 12 and. 13, but bry *v.'rt*N", In the case of NH3, I is 1/2, and as a conseque. 7n. therFe are four nuclear spin wave functions of symmetry A and two of symmetry Eo The symmetry of * is then 4A or 2E in the case of v Ir = A, or the symmetry of f is 4E or 2E+4A in the case of * v = E., v r It can be seen now that the E rotational levels and the A rotationral level is ha.ve the same statistical weight This is true because there are only two E type spinr functions, and because only A type I

41 are permitted for molecules whether they obey Fermi or Bose statisticso It follows for nondegenerate vibrations that for K > 0, if K doubling is not resolved, the rotation levels for K = 3q have double the statistical weight of those for K = 3q~l. In the case of ND3 the identical nuclei have I = 1, and the results of reasoning similar to the above is that all rotational levels occur, but with K = 3q levels with a weight of 11 and K = 3q~1 with a weight of 8o 3o An important property of the wave functions corresponding to the rotational levels discussed above is the effect on them of inversion of the molecular coordinates through the center of mass. It can be shown that the inversion operator i commutes with the Hamiltonian, H, for a symmetric top and therefore the eigenfunctions of H are simultaneously eigenfunctions of io This is true also for i2 so the eigenvalues of i must be +1 or -1l All energy levels of nonlinear and nonplanar polyatomic molecules must therefore be doubled, and are generally labeled + or -; and all levels in Figures 12 and 13 are actually doubly degenerate This degeneracy can be removed if the barrier to inversion is small enough. In the case of the symmetric vibration v2 this barrier is so low that the overall symmetry of the molecule must be regarded as D3 instead of C3v, and the rotational subgroup consists of species E, Al, and A2o The rotational levels for v2 of NH3, taking into account this removal of the inversion degeneracy are given in Figure 14. The splitting is only.66 cm- in the ground state, but it is 35 84 cm-1 in

42 J J -___ Al 3 A2 A233 E E E. A2 2_____ -A 2 E 2 E I A2 I E A2 Ku 0 KIl Kz2 Figure 14. Rotational levels of v2 of NH3 showing the effect of inversion doubling.

43 the first excited vibrational level; therefore, two Q branches separated by 36.5 cm-1 are observed for v2 for gas phase molecules. For ND3 the separation of the Q branches of v2 is only.4 cm-l' It is seen that not only are there levels of types E, Al and A2, but the + and - levels no longer coincide. When the effects of nuclear spin and statistics are taken into account now, the Al rotational levels for NH3 are no longer permitted (since neither AlxAI nor AlxE has the symmetry A2 demanded by the Fermi statistics). This is indicated by drawing all Al levels as dotted lines B. INTERPRETATION OF THE v2 BAND OF MATRIX ISOLATED AMMONIA The interpretation of spectra of molecules isolated in matrices at low temperatures has relied to a large extent upon models ranging from free rotation to a severely hindered motion of the impurity in its "cage.,, 5,1719 Irregularities in the spectra have often been attributed to the possibility of molecules being trapped in more than one site and to the possibility of interactions between nearest neighbor molecules-i e., the formation of polymers 1416 Since dimer or polymer formation should be strongly concentration dependent, it was imperative that wherever possible spectra be taken for a variety of molecule to inert gas atom ratios (see Chapter III). It was thus possible to pick out lines that were not concentration dependent, and to attribute the transitions giving rise to them to molecules completely isolated in the matrix. All spectra referred to in

44 the following paragraphs will be this "monomer" spectra Some very striking similarities are apparent between the observed spectra of NH3 and ND3 trapped in the inert gases and the spectra predicted on the basis of a free rotation model If the molecules are freely rotating, it must be the case that all selection rules and energy level positions are relatively unchangedo In the case of v2 of NH3, one would expect that several very important features be present: (a) A "two molecule" spectrum, consisting of a nuclear spin "E" NH3 and a nuclear spin "A" NH3B This is an anlaogy with the ortho and para type H20 mentioned in Chapter I; and (b) Removal of the inversion degeneracy~ All spectra were taken at or very close to 4.20K. Therefore, only two rotational levels should be appreciably populatedo The J = 0, K = 0, + level should be the only K = 0 level populated, since at 4o2~ the number of molecules in the J = 1, K = 0 level relative to the lowest level is proportional to e-20/kT e-7o Molecules occupying this level will be called "A type" ammonia The J = 1, K = 1 levels should be populated appreciably if there are no spin dependent perturbations present~ This is true because these two levels exist only for E type NH3, and in the absence of a spin dependent perturbation, the E type NH3 cannot pass over into the A type even though the J = 1, K = 1 level is l6o5 cm-l above the ground state. A free rotation model therefore predicts a superposition of the A and E type NH3 spectra, giving a total of five lines. These five transitions are shown in Figure 150 Since all rotational levels of symmetry

'EHN Jo At OJ, pa^oTpaxd suo]TTsu'J 95 ni nST Z >I I >I O Z>1 3 Ii' i 3 I~-3 3, - 3, 3'VV 3 I~V I 3'- 3 l 3 _ ________ 3F 511

46 Al do not exist in nature, the A type spectrum should not show the doubling feature, and in fact should only consist of one lineo The E type spectrum should consist of four lines, all allowed by the usual selection rules AK = 0, AJ = ~1,0. These are the two Q(1) and the two R(l) lines. This is shown in Figure 16a, b and c. Figure 16d shows the lines observed for NH3 isolated in argon and deposited on the KC1 window. Experiment seems to indicate, then, that the separation between the Q branches has changed to 34.9 cml- in the matrix as compared with 36.5 cm-l in the free molecule, and that the separation between the J = 1, K = 1 and J = 2, K = 1 levels has been changed from approximately 40 cm-l to approximately 42 cm-1. It is possible to rationalize shifts of this size using the results of recent investigations of pressure induced line shifts. Rank2 et al., have investigated shifts in vibration-rotation lines in pure HC1 and hydrogen at low pressures, and in mixtures of these molecules with inert gases. Frequency shifts on the 22 order of a few hundredths of a wave number were observedo May et al., have made Raman measurements on pure hydrogen and hydrogen-noble gas mixtures at very high pressures (up to 1940 atmospheres) in an attempt to measure and explain pressure induced shifts in the vibration-rotation bands. Quite large shifts, amounting to 1.86 cm-1 in the case of the S(1) line of hydrogen mixed with helium, were observed. These shifts were explained as arising from a perturbation on the rotational constant caused by Van der Waals forces. It is suggested here that the matrix "cavity" in which the NH3

0) 20 b) 0 36.5 40 76.5 E(cm') Q(l) R (0 Q(1) R(~) R(M) c) 0 20 36.5 40 76.5 d) 0 19.3 34 I 41.9 7 Figure 16. (a) Predicted gas phase spectrum of "A type" NH3, (b) predicted gas phase spectrum of "E type" NH3, (c) composit "A" and "E type" spectra predicted, (d) lines observed in the v2 region for NH3 isolated in argon using a KC1 substrate.

48 molecule is trapped exerts an effective "pressure" which alters the value of the rotational constant. The work of May et al., lends support to this notion in that the shifts observed in the present work are in the same direction, and in that they show up for the transitions ending at the J = 2 level. As was mentioned earlier, the spectrum of V2 of NH3 was taken using copper as the deposition surface in one case, and KC1 in the other. At first glance the two spectra appear quite different. The line widths measured from the KC1 deposit are noticeably wider, indicating that the temperature of the sample might be considerably higher. If this were the case, P branch lines ought to be present, and the R(l) lines ought to be stronger relative to the main R(O) transition. Since this is not the case it is concluded that the difference in the two spectra are due primarily to the effect of the substrate on the matrix. One can also see that the KC1 deposited spectrum has fewer lines than that obtained on copper-just the opposite of what one would expect at a higher temperature. It is noteworthy that all absorptions on KC1 have their counterpart in the copper spectrum, and that all these lines offer a reasonable fit to the explanation given above-i.e., that the molecule is nearly freely rotating, vibrating and inverting, as though in almost complete isolation. No explanation has been made as of yet of the four unassigned lines appearing in the NH3 on copper spectrum. Two of these transitions are quite weak-absorbing only several per cent of the radiation. The other

49 two are as strong as the "E" type lines discussed above. These four lines are listed in Table 1, but in the interest of clarity it is best to state their positions here in terms of their separation from the lowest frequency Q(1) lineo The weak lines are 57 cm-l and 82 cm-' from Q(l), and the stronger lines are 26 cm"l and 62 cm-l from Q(l). It is proposed that the two stronger lines arise from molecules trapped in sites not present in crystals grown on KC1. Two reasons may be advanced in explanation of this. First, the KC1 window was a good single crystal, whereas the copper "mirror" was not. It is not only reasonable to expect that different substrates should result in crystals of different quality, but work carried out at other laboratories have proved it (see Chapter I)o It is proposed then, that argon films grown on the KC1 window has a more ordered structure and has fewer dislocations and/or fewer available interstitial sites. Moreover, according to Zwanzig,23 the deposition process is a function of the substrate. In the language of Zwanzig, it can be said that if in the case in question the condensing gas has different sticking probabilities, then crystals of different quality may be grown. The two weaker lines may be explained in the same way, of course. There is a possibility however, that the one of lower frequency arises from a breakdown in selection rule. This would be the transition J = 0-+ J = 2. If this is actually the case, only a very small mixing of the wave functions is indicated, since the transition is very weak,

50 and no other selection rule breakdown has been observed. Robinson and McCarty24 have used the notion of multiple trapping sites to explain a complicated spectrum of sodium isolated in argon, and very recently Leroi25 and co-workers have explained a spectrum of CO in argon in the same manner. The CO spectrum was scanned at various dilutions, window temperatures and deposition rates, the conclusion being reached that CO was not rotating in the matrix, and that the observed structure was due to different trapping sites. Leroi also resorted to heating the gaseous CO-argon mixture to a temperature well above room temperature before spraying it on a cold surface. This resulted in a narrowing of the lines and a slight difference in relative intensitieso No reasons were proposed for this change, other than the well established fact that annealing a matrix seems to give more order to the structure. In view of the obvious importance of the deposition process when conducting these experiments, and in view of the widely accepted theory that atoms and molecules sprayed onto very cold surfaces wander about briefly before condensing, the idea that substantially different quality crystals with different trapping sites is not completely out of the question. In concluding the discussion of the v2 band of NH3 it should be emphasized once again that the fact that the separation between the Q branches in the matrix is so nearly the same as that for the free molecule serves as very strong evidence that the matrix isolated ammonia does indeed rotate with nearly complete freedom, and it is difficult to con

51 ceive of the molecule being hindered appreciably and still have so little change in the inversion barrier and rotation constant. Also, the discussion of the v2 band of ND3 to be given below completely substantiates this interpretation. Of course it can be said that one should not be surprised to observe free rotation, since ammonia is a small molecule, and the argon matrix is inert; indeed, these are the primary reasons for choosing the ammonia-noble gas mixture to investigate. It must be remembered, though, that the previous investigations of ammonia in argon matrices had not been extensive enough to permit a satisfactory interpretation. As mentioned above, the spectrum of V2 and ND3 in argon also substantiates the free rotation model. This does not mean to say that the ND3 spectrum has the same appearance as that of NH3. The free molecule inversion doubling of V2 for ND3 is expected to be completely obscured by the widths of the lines, thus reducing the complexity of the spectrum. Only three lines should be observed, corresponding to transitions J = 1, K = 1 + J = 1, K = 1; J = O, K = 0- J = 1, K = 0 and J = 1, K = 1 + J = 2, K = 1. Moreover, the spacing between them should be approximately 10 cm-1. This is exactly what is observed for v2 of ND3 in argon, as in Figure 11. An interesting feature of the v2 lines is their relative intensities. In the case of NH3, one line, R(O), is much more intense that the others. This is the one transition corresponding to A type ammonia. This indicates that there is actually more A type NH3 relative to the E type than

52 one would expect on the basis of statistical weights. Now counting the 2J+1 factor, the- levels should have the weight factors 4 for K = 0 and 4 for K = 1. It would appear that there has been a substantial amount of conversion of E type NH3 to A type, although this should not be the case in the absence of a spin dependent perturbation. The amount of conversion (if indeed this has taken place) is quite large, since the per cent absorption of the stronger line is above 80%, in a range where the absorption is highly nonlinear so that the relative intensity difference of the A and E spectra should be even more drastic than it appears to be. Although difficult to explain, a similar phenomenon has been observed by other workers. Robinson and McCarty have reported substantial evidence for free rotation of NH2 isolated in a matrix. Their spectrum indicated that a very large amount of nuclear conversion has taken place, permitting a near thermal equilibrium to take place at low temperatures. No mechanism was proposed which permits this conversion to take place. K. Dressler20 has carried out absorption experiments in the 1800A region on NH3 isolated in argon interpreting his data as proof of free rotation and almost 100% conversion from nuclear spin species E to A. The relative intensities measured from the three lines observed in the v2 region of ND3 in the present work are.more consistent with the assumption of E and A type ND3. The statistical weights of the K = 0 and K = 1 levels should be in the ratio of 11 to 16 whereas the observed intensities are nearly equal. It should be remembered of course, that

53 the "E type" molecules are trapped at an energy well above the Boltzman energy corresponding to 4.2~K. C. INTERPRETATION OF THE v3 BAND It is most convenient for the discussion of the v3 band of NH3 in the 3400 cm-1-3450 cm-l region to proceed in the same manner as was done with V2. In doing so however, several major differences are expected to be present. These are: 1. Being a prependicular band, v3 has the K selection rule AK = +1 instead of AK = 0 for the parallel vibration v2. Therefore, rotational transitions in this band must undergo a change in the component of angular momentum along the symmetry axis of the molecule, a feature not present in v2. 2. Experiment (see Chapter III) has indicated that the near infrared bands are extremely weak compared to v2. This implies at the onset that the very weak lines (the "E type" spectrum) should be difficult or impossible to observe. 3. Inversion doubling of the v2 band is completely negligible, with the result that fewer lines should be observed. Figure 17 shows the transitions expected if both "A" type and "E" type spectra are observed. The expected lines are compared with the observed lines in Figure 18. The lines are measured relative to the vibrational frequency of the free molecule and the position selected for the observed spectrum in the lower half of the figure is the only one

54 J J J 3 E 3 E 2 ~ E Aif~~ 2~ -- A E'MA I. \ E\ IT E / / / E 3 --— ~ EE _ 2 A / A2 E2 IJ~ E \ \ V2 E / I \\V A I — I A E I~~~~ / I ~ / \\ \ K=O K=l K=2 Figure 17. Transitions expected for the V3 band of NH3.

55 16.5 0 3.5 16.5 28.5 42.5 56.5 Figure 18. (a) Predicted lines for the v3 band of NH3, (b) observed lines for the Vs band of NH3.

56 which fits the upper half. As was expected, only the "A type" spectrum is observed, and the broad absorption seems to correspond to the transition J = 0, K = - J = 2, K = 1. At this point enough information has been obtained to permit some qualitative remarks about the nature of the electric field surrounding the trapped ammonia molecule. First, it must be relatively weak since the molecule evidently is undergoing free rotation. "Weak" as used here does not necessarily mean weak from the standpoint of Stark fields ordinarily applied to a molecule. It means weak in the sense that barriers to free rotation are small. Second, since all evidence points to a strong adherence to the + < K -, + <- / +, - < >, selection rules, the field must be invariant under the inversion operator. In other words, it appears that the molecule is located in a site of cubic symmetry, as should be expected for solid argon. The spectrum of v3 of ND3 agrees with that of NH3. There is no evidence of E type ND3, and the separation between the two absorptions is close to that predicted for a free molecule. The higher frequency v3 absorption begins approximately 21 cm-' from the narrow one, whereas the separation between the two transitions in the gas phase are more like 17.5 cm-1. This is a difference similar to that observed in the V2 band of NH3 for the J = 1, K = 1 + J = 2, K = 1 transition. It is difficult to Justify the frequency difference in this case, unless it is connected in some manner with the angular momentum change AK = +1.

57 D. THE vl BAND If both "E" and "A" type transitions are present in the spectrum one expects to see three lines separated by 20 cml corresponding to a Q(1) line, R(0) line, and an R(l) line, since this is a parallel band. This is shown in Figure 7 in the 3300 cm-' region. Since no "E type" spectrum was observed in the V3 band and only small traces of it were observed in the strong v2 band, it is not expected to be present in the vl region. This means that only the one "A type" line should be seen (i.e., the J = 0, K = 0 - J = 1, K = 0 transition). Actually, two monomer lines are observed, as shown in the figure. The stronger, lower frequency line is assigned to the J = 0, K = 0 - J = 1, K =0 transition, since this is the one that should be strongest. The weaker line 15 cmiaway from the R(0) line does not fit the predicted spectrum at all. This points out the difficulty in trying to interpret a spectrum which is predicted from the start to have only one line present. There is just not enough information to form a basis for comparison. The only reasonable explanation that can be given to the weaker line is that it arises from a different site. That this is a real possibility was pointed out in the discussion of the v2 band, where much more information was available. It must be admitted, of course, that such an explanation in this case is not very desirable. The same situation holds for vl of ND3 Whereas only one strong transition is expected, two are observed, but spaced by approximately 12 cm-1. Once again, there is not enough information available to as

58 sign transitions with certainty. There is, of course, a more fundamental reason why the interpretation of the vl data should be more difficult. It is expected that a strong Fermi resonance should exist between the symmetric component of 2v4 and the fundamental vl. Since v4 of ND3 is approximately 1191 cm-1 2v4 should be at approximately 2382 cm-1. This is very close to vl, which is at 2419 cm-l. To enhance this effect the matrix shifts are such that vl decreases in frequency. The effect of this resonance on the vibration-rotation spacings has not been carried out as of yet even in the gas phase, so it should not be surprising to find anomalies in the spectrum of vl. E. THE CONCENTRATION DEPENDENT LINES The foregoing explanation of the observed spectra tacitly omitted reference to all concentration dependent lines. Whereas it is safe to conclude that lines independent of concentration arise from single isolated molecules-that is, monomers-the origin of the concentration dependent lines is open to speculation. Two kinds of dependence on concentration have been observed; the one is a decrease in intensity with increasing dilution of the sample, and the other is an increase in intensity with increasing dilution. A definite pattern has been observed in connection with the former type line. In every case, these lines appear within 8 cm- or so on the low frequency side of a "monomer" line. The absolute position of these

59 lines does not depend on the dilution, so it can be said that the environment of the molecules giving rise to them does not suffer a change as the ratio M/A is changed. Moreover, it is indicated that only one type of environment is giving rise to them~ That is to say, since the notion of dimer, trimer, tetramer, etc., complexes imply one, two, three or more lines for a given "monomer" line, it can be concluded that these particular lines do not arise from a bunching together of various numbers of molecules. This contention is supported by the fact that no absorptions for high concentrations of NH3 in argon appeared in the region of strong absorption characteristic of solid NH3. It can be concluded from these considerations that the solids with a high concentration of NH3 have two kinds of short range crystal structure; the one having NH3 molecules effectively shielded from any other NH3 molecules and the other having a unit cell consisting of some preferred arrangment of argon atoms and NH3 molecules. As the ratio of NH3 to argon decreases, then, more isolated molecules appear at the expense of the more highly concentrated regions. The other kind of concentration dependence appears in two places; one 55 cm-1 to the high frequency side of vl and the other 45 cm- to the high frequency side of v3. These lines are broader than the "monomer" lines, and grow progressively stronger as the matrix is changed from argon to krypton to xenon. In as much as these absorptions become pronounced with increased dilution, they must surely arise from isolated molecules. That is, even though they are concentration dependent,

60 they must be classified as "monomer" lineso The most reasonable explanation of them, then, is that they arise from molecules trapped in a site different than those giving rise to the spectra discussed in the previous sections. This is in keeping with the belief that several sites (substitutional, interstitial, etc.) are available to the molecules, some being more easily accessible than others. It is also consistent with the model suggested above for the origin of the other type concentration dependent lines, since larger regions of "monomer" type crystal exist at higher dilutions, and more molecules are available to them. This explanation also gains support from the fact that the absorptions in question are broad and that their strength varies with matrix. This is exactly what one should expect since the more unaccessible site should assert a stronger perturbation than the others, and since this accessibility should be strongly dependent upon the quantitative features of the site itself.

V. ALTERNATIVE EXPLANATIONS In view of the modest amount of data on which some of the foregoing arguments were based it is worthwhile to compare the experimental results with models other than that of free rotation. The first of these to be discussed is the librator model. In this model the rotational degrees of freedom are restricted in such a manner that a potential energy of the form V = 1/2kQ2 is assumed. The Schroedinger equation with this potential is identical to that of a two dimensional harmonic oscillator. The energy levels are: 2B2 ^ 1\ E = (n+l) -w + 5 - where n is the librational quantum number, and B ='f2/2I and w =1 /7 and where n = 0,1,2,... and IMI = n,n-2,n-4,... The energy level diagram for the libration levels is shown in Figure 19. The splitting due to the M dependence is small and can be ignored for qualitative discussions of the predicted spectrum. The selection rules for parallel transitions are: An = 0,+2,+4 (Decreasing in intensity) AM = 0 and for perpendicular librations: 61

62 An = +1,3,+5,... AM = +1 Two features which are immediately predicted for a vibration-libration band are the following: first, the line spacings must be equal, and second, the spacings must be proportional to 1/fl, where I is the moment of inertia of the librator. M N AI A2,3 E' E ~1 3 Al o 2 Al 0 0 Figure 19. Energy levels for a harmonic librator. Whether NH3 is inverting or not, the observed spectrum does not seem to fit this scheme. In the first place, the observed spacing of approximately 10 cm-l for v2 of ND3 requires that the equal spacing of the corresponding NH3 lines be approximately 14 cm1l. The only regular

spacing of the stronger NH3 lines are 20 cm —-which is more like 2x10 cm-1. Also, it is predicted for a librator that the An = 0 transition (that of lowest frequency) should be the stronger. This is obviously not the case for v2 for either NH3 or ND3. If NH3 is not inverting while librating, as is most likely the case, this explanation becomes even worse, since no regular spacing is at all apparent. The same is true for the other parallel band vl. In this case, there are two lines, separated by 15 cm-". The spacing between the two observed lines in vl of ND3 is about 12, denying a 1/I2 dependence of the energy levels. The only attractive feature is the relative intensities of the two lines is the NH3 spectrum. The only experimental evidence that comes close to supporting the libration picture is v3 in both NH3 and ND3. Both bands consist of two absorptions, and the ratio of their separations gives a value between 1/2 and 14/. This is what is expected if the actual motion is not completely free, but hindered to some extent. It is safe to say, at any rate, that the librational model does not fit any of the experimental evidence. Another model which can be quickly rejected is the harmonic oscillator model. The molecule is pictured as a three dimensional harmonic oscillator, fastened to the lattice by springs. The energy dependence of the harmonic motion must be proportional to l /h, where M is the total mass of the molecule. As was done with the librator model, it can

64 be quickly seen that the line separations and the ratios of the NH3 and ND3 spacings do not satisfy the required 1/ (17/20) relationship.

BIBLIOGRAPHY 1. G. N. Lewis and D. Lipkin, J. Amer. Chem. Soco 64, 280 (1942)" 2. F. 0. Rice and M. Freamo, J. Amero Chem. Soc. 73, 5529 (1951)o 35 I. L. Mador and M. Co Williams, J. Chem. Phys. 22, 1627 (1954)0 4. I. Norman and G. Porter, Nature 174, 508 (1954). 5. A. M. Bass and H. P. Broida, National Bureau of Standards Monograph 12 (August 1, 1960). 6. Ao M. Bass and H. P. Broida, Formation and Trapping of Free Radicals, Academic Press, New York (1960). 7. E. D. Becker, G. C. Pimentel and M. Van Thiel, J. Chem. Physo 26, 145 (1957). 8. H. S. Peiser in The Formation and Trapping of Free Radicals, edo A. M. Bass and H. Po Broida (Academic Press, Inc., New York, N. Yo, 1960). 9. R. Livingston, H. Zeldes, A. Weinberger, and R. Holmberg, Faraday Society Discussions, p. 26 (Sheffield University, September 1958)> 10. G. W. Robinson and M. McCarty, Jr., J. Chem. Phys. 30, 999 (1959). 11. E. Whittle, D. A. Dows, and G. C. Pimentel, J. Chem. Phys. 22, 1943 (1954). 12. E. D. Becker and G. C. Pimentel, J. Chem. Phys. 25, 224 (1956). 13. E. D. Becker, G. C. Pimentel and M. Van Thiel, J. Chem. Phys. 26, 145 (1957). 14. M. Van Thiel, E. D. Becker, and G. Co Pimentel, J. Chem. Phys. 27, 95 (1957). 15o D. E. Milligan, R. M. Hexter, and K. Dressler, Jo Chem. Physo 34, 1009 (1961) 16. G. C. Pimentel, M. Bulanin, and M. Van Thiel, J. Chem. Phys. 36, 500 (1962). 65

66 17. L. J. Schoen, D. E. Mann, C. Knobler, and D. White, J. Chem. Phys. 37, 1146 (1962). 18. J. A. Glasel, J. Chem. Phys. 33, 252 (1960). 19. R. L. Redington and D. E. Milligan, J. Chem. Phys. 37, 2162 (1962). 20. K. Dressier, J. Chem. Phys. 35, 165 (1961). 21. D. H. Rank, W. D. Birtley, D. P. Eastman, and T. A. Wiggins, J. Chem. Phys. 32, 296 (1960). 22. A. D. May, V. Degen, J. C. Stryland, and H. L. Welsh, Canad. J. Phys. 39, 1769 (1961). 23. R. W. Zwanzig, National Bureau of Standards Monograph 12 (August 1, 1960). 24. G. W. Robinson and M. McCarty, ibid. 25. R. Leroi, paper presented at the Symposium of Spectroscopy and Molecular Structure, Ohio State University (June 1963).