J. Chem. SOC., Faraday Trans. I, 1988, 84(2), 657-664 Spectrochemistry of Solutions Part 20.l-The Infrared, Near Infrared and Visible Spectra of Liquid Ammonia Joyce C. Dougal, Peter Gans* and J. Bernard Gill* Department of Inorganic and Structural Chemistry, The University, Leeds LS2 9JT The infrared, near-infrared and visible spectra of liquid ammonia at 293 K have been recorded. The fundamental and first six overtone regions have been assigned on the basis of the existence of four overtone progressions. The spectral data have been interpreted, together with other published spectral data, in terms of a model in which the local environment of the ammonia molecules is the predominating influence. Evidence is presented for the existence, in the liquid, of ammonia molecules with 0, I, 2 or 3 hydrogen bonds.In 1936 Costeanu and Barchewitz reported the spectrum between 600 and 950 nm of liquid ammonia at 198 K. They found two series of bands which could be assigned to overtones in anharmonic progressions, pointing to fundamental vibrations at 3333 and 3407 cm-1.2 Although many studies of the vibrational spectra of liquid ammonia have been reported ~ubsequently,~-ll this remarkable observation appears to have been neglected. Roberts and Lagowski examined one of the bands with the aid of curve resolution and difference spectroscopy, but they admitted that their analysis was not completely satisfactory, l2 Having already built a cell in order to examine the infrared spectra of electrolytes dissolved in liquid ammonia, we have now constructed a cell suitable for obtaining spectra in the visible and ultraviolet regions of the spectrum.By using both these cells, we have been able to measure the spectrum of liquid ammonia at room temperature from 200 to 8000 nm (50000 to 1250 cm-l). The infrared spectrum between 3000 and 3600cm-l has been analysed by curve resolution, but in the other region the band maxima were identified in the second or fourth derivative of the spectrum with respect to wavelength. We have found at least four overtone progressions. In the discussion we have attempted to collate all vibrational spectral observations by means of a model of liquid-ammonia structure which shows a significant and variegated degree of hydrogen bonding. Experiment a1 The cell used for recording infrared spectra has been described previo~s1y.l~ The cell used for recording the near-infrared and visible spectra was constructed to the same design as that of the infrared cell, the main differences being that the path length is fixed at 10 mm, and the windows are of quartz.A path length of 1 mm was obtained by inserting a 9mm quartz spacer into the cell. All spectra were recorded at ambient temperature. Infrared spectra were recorded as before.13 Below 4000 cm-l a path length of 12.8 pm was used. For the region 3048-3600 cm-l 25 independent scans were run at 15 min intervals, the infrared beam being blocked off between measurements to minimise sample heating. The 25 scans were co-added. The infrared spectrum was also recorded on a Nicolet 5-MX interferometer from 1200 to 4800 cm-' with a 50 pm path length.Between 4500 and 12 500 cm-I the spectra were recorded in sections on a Cary 14H 657658 Spectrochemistry of Liquid Ammonia spectrophotometer equipped with digital paper-tape output. The sample path length was 1 mm below 7200 cm-l and 10 mm above 7200 cm-'. The paper tapes were read into a BBC microcomputer recently programmed by us for use as an infrared data station with the SP3 infrared spectrophotometer. The spectra were co-added using the software facilities provided by that program. Between 2 and 6 scans were co-added for each group of bands. Spectra in the region 12500-50000 cm-l were recorded on a Pye-Unicam PU8800 spectrophotometer, connected to a BBC microcomputer for which we have written software with facilities similar to those available for the analysis of infrared spectra.Results The infrared spectrum between 3048 and 3600 cm-' is shown in fig. 1, together with the fit obtained with our VIPER ~r0gram.l~ The bands all have Lorentzian shape, and the second derivative of the fitted spectrum was in good agreement with the second derivatives obtained from the observed spectrum by the convolution method, except that the band at 3369 cm-' appeared as a shoulder in the convolution but as a resolved band in the derivative of the sum of Lorentzians. The band at 3339 cm-l is more than twice as broad as any other band in the fit. The spectrum observed can be classified into seven regions, corresponding to the fundamental and the first six overtones of the N-H stretching vibrations.In each region there is a fundamental or overtone complex, and a complex of combination bands of those vibrations and each of the two fundamental bending vibrations. Also, there are ternary combinations involving a stretching mode and both bending modes. The presence of underlying band maxima was ascertained by means of our MULTISMOOTH program,15 using the fact that the derivative of the spectrum with respect to wavelength is apparently more well resolved than the spectrum itself. Where possible the positions of the band maxima were determined by locating, using a least-squares method, the zero crossing points in first or third derivatives of the spectra. These peak positions are given in table 1. The spectrum in the first overtone region and its fourth derivative is shown in fig.2; Roberts and Lagoswki reported a band of similar general appearance.12 The spectrum and fourth derivative of the region involving the combinations of first overtone and bending modes is shown in fig. 3. The intensity of absorption decreases by a factor of ca. 20 for each step up the overtone sequence. Thus, whilst a 1 mm path length gave excellent spectra in the first overtone region, a 10 mm path length did not result in such a high signal-to-noise ratio in the second overtone region, and only the second derivative could be interpreted with confidence. Because the spectra obtained in the third and fourth overtone regions yielded rather noisy second derivatives, these regions could not be examined in detail. Whilst the higher overtones were only just detectable, as shown in fig.4, their positions could be established. Four series of overtone and combination bands have been assigned, designated by the subscripts a, b, c and d. 13 bands in the (a) series fundamental, overtones and combination bands were fitted by the method of unweighted least-squares to eqn (l), yielding the results shown in table 2. v = v, 0, +v: x,, +v2 v2 +v2 v, XZ3 +v, v, + v, v, X,,. The fundamental and overtone wavenumbers of the (b), (c) and ( d ) series were fitted to (2) eqn (2) : The values for the (a) series agree within 3 cm-l with those reported by Costeanu and Barchewitz,2 whilst for the (b) series the 198 K harmonic wavenumber is some 43 cm-l lower than the room-temperature value. v = vcu,+v2x33.J .C. Dougal, P. Gans and J . B. Gill 659 3600 3400 32 00 wavenum ber/ cm -' Fig. 1. Infrared spectrum of the fundamental N-H stretching region of liquid ammonia at 293 K, together with five Lorentzian bands used to fit the spectrum. For the combination bands with v, in the (b) series the two equations v = v v ~ ~ s + v , + v x , , (v = 2,3) (3) were solved for the two unknowns v, and A',,; v, and X34 were calculated in a similar manner. The band wavenumbers derived from these parameters are therefore equal to the observed wavenumbers. Bands at 6414 and 9404 cm-l have not been assigned; they could either belong to a fifth overtone progression or they could be combination bands derived from symmetric and antisymmetric stretching vibrations. Also, the combination bands at 4458 and 4914 cm-l do not fit into the progressions of combination bands with overtones, implying that they derive from different N-H stretching fundamentals.Discussion Very little is known about the structure of liquid ammonia. In crystalline amrnonial6* '' each molecule forms three N-H ... N bonds and three H - - . N-H bonds. However, although the hydrogen bonds are approximately linear, they are unusual in that the nitrogen atom's lone electron pair cannot lie on the bond axis. The structure presumably arises because the hydrogen bonding is relatively weak, so that packing requirements determine the overall structure. The structure is in marked contrast to that of crystalline water, in which there are an equal number of protons and lone pairs on each water molecule, resulting in a fully hydrogen-bonded structure.The ammonia molecule, with three protons and one lone pair, cannot form a regular hydrogen-bonded structure in three dimensions. X-Ray diffraction data obtained from the liquid have been interpreted'' in a way that indicates that each nitrogen atom has two distinct nearest nitrogen neighbours. Unfortunately, molecular-dynamics simulations have not yet yielded a model which reproduces the diffraction data adeq~ate1y.l~~ 2o All that could be saidlg is that the models examined agree in predicting a significant degree of association in the liquid. This is consistent with the quantum-mechanical calculations on the ammonia dimer,lg which suggest that the configuration with the linear N-H ... N bond is the most energetically favourable configuration, with a dimerisation energy of ca. 15 kJ mol-l.We propose a new interpretation of the vibrational spectroscopic data, not in terms of the overall liquid structure, but in terms of the immediate environment of the ammonia molecule. There is, of course, a link between these two descriptions, but we do not think it can be derived from the vibrational data.660 Spec t rochem is t ry of Liquid Ammonia Table 1. Observed and calculated peak maxima (in cm-l) in the spectrum of liquid ammonia at room temperature obs calc assignment notes 1640 1750 3 203 3 247 3 339 3 369 3 386 4 360 4458 4914 6071 6 309 6 384 6414 6 541 6 637 6704 677 1 7 622 7 704 7 757 7 843 8 122 8 230 8316 8 379 9 194 9404 9581 9 864 9 978 10111 10 702 10943 11 171 11 450 12515 12 889 13635 14037 15 299 15 822 16431 17 934 'Id 3337 v3, 3373 VSc - - 6535 2v3a 6663 2v3, 6700 2~~~ 6764 2v3, 7624 v,+2v3, exact v, + 2v3, v, + 2v3c - '2 + 2v3d - 8135 v,+2v3, exact vp + 2v3, v4 + 2v3c v, + 2v,, + v, - '4 + 2v3d - - - - 9594 3v3ar 9979 3V3c 9823 3v3, 10113 3v3d 10 699 v, + 3v3, exact v, + 3v3, 11 156 v,+3v3, exact v, + 3v3, 12513 4v3, 12888 4v3, 13 635 v, + 4v3, 14044 v,+4v3, 15294 5v3, 15847 5v3, 16423 v, + 5v3, 17935 6,, 1046 (Raman)3 broad 32 17 (Raman)3 3260 (Raman)3 3303 (Raman)3 c - - - broad - - (1046 + 3386 + 1640 = 6072) - - (1046 + 6704 = 7750) (1046+6771 = 7871) - (1640+6704 = 8344) (1640+6771 = 8411) (1046+ 1640+6541 = 9227)J .C. Dougal, P . Gans and J . B. Gill 661 1 I 1500 1600 w avelengh/nm Fig. 2.Absorption spectrum of liquid ammonia in the first overtone region (a) and fourth derivative of the absorbance with respect to wavelength (b). w avenum ber/ cm - I I I I 1 1 1 1 wavelength/nm Fig. 3. Absorption spectrum of liquid ammonia in the region of the combination bands with the first overtones (a) and fourth derivatives of the absorbance with respect to wavelength (b).662 Spectrochemistry of Liquid Ammonia 0.3 w avelength/nm Fig. 4. Absorption spectrum of liquid ammonia (10 mm cuvette) in the third, fourth and fifth overtone regions. The same spectrum is shown on two different absorbance scales, (a) and (b). Table 2. Calculated harmonic wavenumber and anharmonicity constants for overtone band progressions ; fundamental wavenumbers and anharmonicity constants for combination band progressions" band o,/cm-' X,,/cm-' v,/cm-' XZ3/cm-l v,/cm-' X,,/cm-' a 34O7( 4) -70(1) 1056(17) 16 (4) 1658 (21) -32(6) b - 3441 (15) -55(4) 1043 12 1607 -7 C 3397 ( 7) -23 (3) - d 3404(ll) -11 (4) - - - - - - - " Least-squares values of the error in the last digits are given in parentheses.The isolated ammonia molecular has Csv symmetry, giving rise to stretching vibrations of A; and E' symmetry, both of which are infrared and Raman active. There is also the possibility of Fermi resonance between 2v, and vl. The ammonia molecule in the liquid may have various environments. For example it may donate to 0, 1, 2 or 3 N-H ..- N bonds, and it may form H N-H bonds of the conventional kind or of the kind that exists in the solid. We suggest that the effect of these relatively weak interactions can, as a first approximation, be considered as a perturbation.Corset and Lascombe have calculated the effects of a perturbation consisting of a change in the N-H stretching force-constant. 21 The various environments cause different degrees of shift in the A; and E' wavenumbers, and in the cases of 1 and 2 N-H --. N bond structures, separation of the components of the E' vibration with the lifting of degeneracy. Thus, although the species with 0, 1, 2 or 3 N-H -.. N bonds should give rise to eight polarised and four depolarised Raman bands (all infrared active), some bands will be at almost the same wavenumber. Changes in the relative absorption intensities may also occur. In the region of the fundamental stretching vibrations the infrared spectrum showsJ.C. Dougal, P. Gans and J. B. Gill 663 remarkably few coincidences with the Raman spectrum, in which it is now generally agreed that there are three polarized bands.3* The infrared band at 3247 cm-l does not coincide with any of the polarised Raman lines, but it has been classed as a symmetric vibration v,, because of its low wavenumber. It therefore appears that there are four distinct symmetric vibrations ; ignoring Fermi resonance, that could mean that there are four ammonia environments. We believe that Fenni resonance should be ignored, following the interpretation by Gardiner et aL3 of data obtained from solutions of ammonia in carbon tetrachloride and acetonitrile. In dilute solutions they observed a strong band at ca.3314 cm-' and a very weak band at ca. 3206 cm-l (both polarised). As the ammonia concentration is increased the strong band shifts to lower wavenumber, the weak band shifts to higher frequency and grows in intensity relative to the strong band, new bands grow in relative intensity at ca. 3387 and 3217 cm-l (pure-liquid values), and v, shifts from 989 to 1046 cm-l. These changes are consistent with changes in the relative proportions of various ammonia environments, i.e. increasing aggregation, with concentration. Since both CH3CN and CCl, are weak hydrogen-bond acceptors it is reasonable to suppose that the strong band observed in the dilute solutions is predominantly that of monomeric ammonia molecules hydrogen-bonded to the solvent. In that case the weak band belongs to an ammonia aggregate such as a dimer, and is not the result of Fermi resonance.Lemley et aL4 have suggested a two species model for the solvent structure. Their most convincing evidence is the observation of two polarised N-D stretching vibrations in the Raman spectrum of NH,D. However, two bands would be predicted with the local environment model. One environment would have a N-D N bond and one would not; perturbations due to other hydrogen bonds in the molecule would be small. From the intensities of the two bands we would conclude that there are roughly equal proportions of the species with and without the deuterium bond. Our observation of four symmetric stretching fundamental bands together with four overtone progressions is highly suggestive of the existence of four local environments. These could well be the environments (IWIV): I I1 I11 IV N N N N I I I I I In the local-environment model hydrogen bonding involving an adjacent ammonia molecule and the nitrogen atom would be expected to perturb the spectrum less than hydrogen bonding involving the nitrogen atom of another ammonia molecule.The packing constraints that result in unusual hydrogen bonding in solid ammonia are not present in the liquid. It is therefore likely that the normal hydrogen bond with the lone pair on the bond axis will predominate. Only one environment can exist in an infinite structure; the species which has one N-H . - - N bond and one H ... N-H bond can exist in an infinite chain. The other environments can only exist in small aggregates.Various workers have studied the effect of temperature on the Raman-active N-H stretching vibration^.^. 5, 6 q lo All have observed large changes in the relative intensity of the component bands. Perhaps the most significant observation is growth of the depolarised band at ca. 3385 cm-l at low temperatures. This band, which we assign as664 Spectrochemistry of Liquid Ammonia v3d, is not found in the spectrum of a dilute solution of ammonia in carbon tetrachloride or a~etonitrile.~ It is thus a band characteristic of aggregates, and would be expected on that basis to be more prominent at low temperatures, where the liquid is expected to be more structured. Kruh and Petz have reported radial distribution functions calculated from X-ray diffraction data obtained at 277, 228 and 199 K.22 They concluded that the structure is more highly ordered at low temperatures.Buback has reported infrared spectra of the fundamental regions as a function of density, between 303 and 573 K; above the critical temperature, 405.5 K, there is of course only one phase. At 423 K the wavenumber of the band maximum in the v3 region increases from ca. 3385 to ca. 3405 cm-l at the lowest den~ity.~ This suggests to us a change in the direction of less aggregation, with the smaller aggregates having a higher absorption wavenumber. We have reported that the v4 bending region in the Raman spectrum is more complex than a single band.'' In addition to the peak at 1640 cm-' another band was found at 1750 cm-l, with five times the integrated intensity at 293 K, and approximately equal intensity at 196 K.We have now also found a band at 1750 cm-l in the infrared spectrum (table 1). This is yet another clear indication of a temperature-dependent aggregation process existing in liquid ammonia. The precise form of the aggregation is as yet unknown, but the new data presented here on the fundamental and overtone absorptions, taken together with the published data on Raman scattering, are consistent with a model in which any one ammonia molecule may form between 0 and 3 N-H .-- N bonds with neighbouring ammonia molecules. We are grateful for the support of the S.E.R.C. both for the grant for infrared and computer equipment (GR/B/00817), and for the provision of a CASE studentship for J.C. D. (cooperating body Johnson Matthey Technology Centre). We also thank Prof. B. L. Shaw F.R.S. for the use of the interferometer, and Dr T. R. Griffiths for the use of the Cary 14H spectrophotometer. References 1 Part 19. P. Gans, J. B. Gill and L. H. Johnson, J. Chem. SOC., Dalton Trans., 1987, 673. 2 G. Costeanu and P. Barchewitz, C. R. Acud. Sci., 1936, 203, 1499. 3 D. J. Gardiner, R. E. Hester and W. E. L. Grossman, J . Raman Spectrosc., 1973, 1, 87. 4 A. T. Lemley, J. H. Roberts, K. R. Plowman and J. J. Lagowski, J. Phys. Chem., 1973, 18, 2185. 5 M. Buback, Ber. Bunsenges. Phys. Chem., 1974, 78, 1230. 6 J. W. Lundeen and W. H. Koehler, J. Chem. Phys., 1975, 79, 2957. 7 C. A. Plint, R. M. Small and H. L. Welsh, Can. J. Phys., 1954, 32, 653. 8 T. Birchall and I. Drummond, J. Chem. SOC. A , 1970, 1859. 9 B. Bettignies and F. Wallart, C. R. Acad. Sci., 1970, 271, 640. 10 M. Schwartz and C. H. Wang, J . Chem. Phys., 1973, 59, 5258. 11 P. Gans and J. B. Gill, J. Chem. Soc., Dalton Trans., 1976, 779. 12 J. H. Roberts and J. J. Lagowski, Electrons, Fluids, Nat. Met. Ammonia Solutions, 3rd Colloq. Weyl, ed. 13 P. Gans, J. B. Gill, Y. M. MacInnes and C. Reyner, Spectrochim. Acta, Part A , 1986, 42, 1349. 14 P. Gans, Comput. Chem., 1977, 1, 291. 15 P. Gans and J. B. Gill, Appl. Spectrosc., 1983, 37, 515. 16 I. Olovsson and D. H. Templeton, Acta Crystallogr., 1959, 12, 832. 17 J. W. Reed and P. M. Harris, J . Chem. Phys., 1961, 35, 1730. 18 A. H. Narten, J . Chem. Phys., 1977, 66, 3117. 19 A. Hinchcliffe, D. G. Bounds, M. L. Klein, I. R. McDonald and R. Rhigini, J. Chem. Phys., 1981,74, 20 M. L. Klein and I. R. McDonald, J . Chem. Phys., 1981, 74,4214. 21 J. Corset and J. Lascombe, J. Chim. Phys., 1967, 64, 665. 22 R. F. Kruh and J. L. Petz, J . Chem. Phys., 1964, 41, 890. J. Jortner (Springer Verlag, New York, 1973), p. 39. 121 1. Paper 71901 ; Received 19th May, 1987