Electronically Excited States of Ammonia BY A. E. DOUGLAS Division of Pure Physics, National Research Council, Ottawa, Canada Received 1 1 th January, 1963 The ultra-violet absorption bands of 14NH3, 14ND3 and 15NH3, which lie between 1400 and 2200 A, have been photographed at high dispersion and the rotation structures of a number of the bands have been analyzed. The Zeeman effect has also been studied for a number of the bands. It is shown that in the lowest excited state, which is of A; symmetry, that the molecule is planar with a bond distance of 1.08 A. The Zeeman studies show the second excited state to be degenerate. A third excited state which is the upper state of parailel bands near 15008, is not fully understood. The electronic states and predissociations of NH3 are discussed.The ultra-violet absorption spectrum of ammonia consists of a large number of band systems extending from 2200A into the far ultra-violet. Earlier work has been reviewed and a number of new aspects of these band systems have been dis- cussed in two recent papers by Walsh and Warsop 1 and by Douglas and Hollas.2 In the present work, absorption bands involving three excited states will be discussed. The first of these excited states, denoted * by A, is the upper state of the 2000 A absorption bands.1 The second excited state is the upper state of the 15OOA bands which have been analyzed by Douglas and Hollas, while the third state gives weak bands in the same spectral region as the B-x system. The rotational structures of a number of the c-2 and 2-x bands have been analyzed and the Zeeman effect of the lines of the B-x bands has been observed.Finally, theoretical interpretations for some of the properties of the excited state are discussed. EXPERIMENTAL All spectra were photographed with a 10-m, concave grating spectrograph3 at dis- persions between 0.18 and 0.28 A/mrn. The continuous background was obtained from a Lyman source and lines from an iron hollow cathode lamp were used as wavelength standards. The absorption tube was designed in such a way that it could be cooled to -80°C or heated a few hundred degrees above room temperature. The Zeeman spectra were obtained at a field of 58,000 oersted in an apparatus described previously.4 The ND3 was obtained from Merck and Co., and the 15NH3 was prepared from K15No3 (supplied by Isomet Corp.) by the action of Devardas alloy in NaOH solution.RESULTS 2-2 BANDS The A-x system of ammonia consists of a long progression of bands near 2000 A. The bands of NH3 are too diffuse to show measurable rotational structure. *It is convenient to designate each electronic state by an empirical sy_mbol. Following the notation adapted for diatomic molecules, the gtound state is denoted by X and excited states of the ssme multiplicity as the ground state by ABC . . . . States of different multiplicity are denoted by Zbi.. . . . The tilde is used to differentiate the empirical symbols from those denoting symmetry species. 158A . E. DOUGLAS 159 From a study of the vibrational structure of the bands, Walsh and Warsop 1 have shown that in the upper state the molecule is planar and that the progression of bands arises from excitation of the out-of-plane vibration v;.In an early paper Benedicts pointed out that some of the corresponding bands of ND3 are discrete and, without giving details, stated that a partial rotational analysis shows that, in the upper state, the molecule has the configuration of a planar equilateral triangle with a bond distance r(N--H) of 1.07A. The present work was undertaken in order to obtain a more complete analysis of the discrete bands of ND3. The spectra show that bands involving 0; = 0 have rotational structures con- sisting of broad lines while those involving v; = 1 have much sharper lines. No measurable rotational structure can be seen in bands with 4 2 2 .Fig. la and lb show the 0-1 and 1-0 bands. (The vibrational quantum numbers given here and throughout this paper are the values of v2, all other vibrational quantum numbers being zero.) The 1-4, 1-1, 0-0, 0-1, 0-2 and 0-3 bands have been measured and the wave-numbers of the lines are given in table 1. The analysis of the bands presented no difficulty. In a parallel band of ND3, the first lines of the Q branches of the sub-bands are the strongest lines in the band and this series of lines Q(J = K ) can be identified readily. Also the first few lines of the R branches are free from overlapping and are easily identified. From these lines, approximate rotational constants can be calculated and thereafter the remaining lines can be identified. The energy levels of the u; = 1 state were established by adding the wave-numbers of the observed lines in the 1-1 and 1-4 bands to the appropriate lower state energy levels which were calculated from the constants given by Benedict and TABLE l.-WAVENUMBERS OF THE LINES OF THE A - 2 BANDS OF ND3 1" BAND v (cm-1) branch(J)K v (cm-1) branch(J) K 47334.18 * 308.57 * 280.42 * 377.13 385-72 * 401.11 * 345-76 334.18 * 366.35 385.72 * 393.35 * 401.11 * 364- 12 393.35 * 3 19.61 306.28 292.60 278.02 263.16 360-45 357-35 398.85 47304.09 * 290-44 275.95 260.63 355.56 351.81 287.75 273.16 257.92 349.35 34509 269.76 * 254.47 * 341.85 336-63 327.09 316-19 31 1.36 304-09 * 298.70 28476 * 269.76 * * overlappedI60 STATES OFNH~ Y (cm-1) 46608.79 * 584.78 * 560.38 * 532.79 * 636.68 652.71 * 596.98 * 584.78 * 572.93 * 560-38 * 546.12 * 532.78 * 617.18 636.68 * 645-32 652.71 584.78 572.93 560.38 546- 12 532.78 614.98 645.32 652.71 570.89 558.10 * * overlappedA.E. DOUGLAS 161 Plyler.6 The upper state levels were fitted to the usual equation for the rotation energy levels of a symmetric top and the rotational constants B and C determined. The accuracy of the data, which is limited by the line width, was insufficient to deduce the three centrifugal stretching constants DJ, DK and DJK. The constants which have been determined are given in table 2. F(J, K ) = BJ(J+ 1) + (C- B)K2, (1) TABLE CONSTANTS DERIVED FROM THE A - 2 BANDS OF ND3 0'- "I' 1" 1-1 1-2 0 4 0-1 0-2 0-3 2 4 v ; = l ? vo (cm-1) 473 67.35 46617.88) 45937.0 4671 3.8 45967.4 45355-1 44887.1 48037 48908 vibronic transition B' (cm-1) C' (cm-1 A;-A," 4.78 2-32 A;-A; A',--A! AZ-A; Ai-AI Ai-A; A;-A; diffuse diffuse The large line widths in bands which have u; = 0 as their upper state, prevent useful rotational constants from being determined for this level.Most of the bands arising from vibrationally excited levels of the ground state have u$ = 0 as their upper state and as a result they yield no useful rotational constants and even their band origins may be in error by a few tenths of a wave-number unit. No attempt was made to measure the diffuse bands with ~ $ 2 2 but there is a weak band of ND3 at 48908 cm-1 which is noL a member of the strong progression of bands. First, spectra photographed at room temperature show the 0-0 and 1-1 bands of ND3 have about the same intensity in spite of the fact that the population of the ug = 1 level is only 0.028 of the population of the t$ = 0 level. This Franck-Condon effect has been noted in the spectrum of NH3 by Walsh and Warsop 1 but the effect is even greater in ND3.Secondly, the relative intensities of the first lines of the R branches are very different in bands with = 0 from bands with u; = 1. In the 0-0 band, for example, R(0) and R(l) are roughly of the same intensity whereas in the 1-1 band the A(1) line is strong and the R(0) line is extremely weak. It will be shown later that this difference in line intensities is very useful in determining the symmetry of the vibronic levels. Two unusual characteristics of the A-r?( system are worth noting. B-2 SYSTEM The &--x system consists of a long progression of perpendicular bands between 1690 and 1400A.The analysis of these bands in NHJ which has been given by Douglas and Hollas,2 shows that the vibronic levels of the upper state are degenerate and there is strong evidence that the degeneracy is electronic.* Furthermore, the * A correction of an error in eqn. (3) of their paper has led to a better fit of the data with only minor changes in the constants. Their eqn. (3) should read 7 F162 STATES OF NH3 molecule in the excited state has a large internal angular momentum, and, since this is assumed to arise from motion of the electrons, it should have a large magnetic moment. The rotational levels of the state should therefore show a strong Zeeman effect and experiments were undertaken to test this point.The theory of the Zeeman effect for a linear molecule has been given by Crawford.8 By a simple extension of his equations, it follows that a magnetic field H will displace a rotational energy level of a singlet E electronic state of a symmetric top by an amount gKMH = 4 6 X l o - - J(J + 1) cnl- '. gefzKMH 2mc J( J + 1) AV = In this equation J, K and M are the rotational quantum numbers, e, k, rn and c have their usual meanings and g is the usual Land6 factor which has a value of unity for a molecule with a magnetic moment of one Bohr magneton. It follows from the form of eqn. (2) that rotational levels with K not very different from J are sensitive to the magnetic field whereas levels with J$ K are relatively insensitive. Experimentally the Zeeman effect at a field of 58,000 gauss was examined in the 1 4 , 2 4 and 3-0 bands of the B---X system and a number of bands of the c-z system.A part of the 3 4 band photographed first with and then without the magnetic field is shown in fig. 2a and 2b. All lines in the bands of the &2? system are broadened by predissociation and therefore it was not possible to resolve the Zeeman pattern of any of the lines. As a result, the experiment is qualitative rather than quantitative. There is no doubt, however, that some of the lines are greatly broadened by the magnetic field whereas others are almost unaffected. Furthermore, qualitatively at least, the broadening of the lines depends on the ratio of K to J in the way one would expect from eqn.(2). A rough value of g has been obtained by measuring the widths of a number of lines in each of three bands. The average of these measurements is g = 0.6. The lines of the c-x system showed no measurable Zeemar effect. e-x SYSTEM Douglas and Hollas 2 have described a progression of weak bands of NH3 which lie in the same region as the stronger B--Z bands and they show on: of the bands in plate 1 of their paper. Some of these bands, here called the c-X band system, fall on bands of the B-2 system but fortunately a number of them are free from overlapping. The bands have the same type of rotational structure as the 2-x bands of ND3 and have been analyzed in the same way. Though the line width in the c-x bands is considerably greater than in the &-z bands, the analysis is quite definite and there is no doubt that they are parallel bands of NH3.The wave- numbers of the lines of the five bands which have been analyzed are given in table 3 and the constants which have been derived from the analysis are given in table 5. Six bands of the corresponding system of ND3 have been measured and analyzed. Here the lines are much sharper. The wave-number of the lines of the ND3 bands are given in table 4 and the constant2 derived from them in table 5. The bands are very similar in appearance to the A-2 band shown in fig. lb. The bands at the long wavelength end of the progressions are overlapped by the stronger &-z bands and experimentally it is impossible to determine the first member of the progression. Thus, there is no direct way of determining the vibra- tional numbering of the upper state vibrational levels.The NH3-ND3 isotope shift is not useful in this respect and the shift is so large that it is not possible to deter- mine which band of ND3 corresponds to a given band of NH3.a b FIG. 1 .-The 0-1 (a) and the 1-0 (6) bands of the 2-2 system. Note that the relative intensities of the R(O), R(1) and R(2) lines are quite different in the two bands. [To face page 162.a b / \ K j l 66 '6 *8 78 *2 48 '6 FIG. 2.-Portions of the 3-0 band of the fi-2 system with (a) and without (b) a magnetic field of 58,000 oersteds.A. E. DOUGLAS 163 170 = 65654.9 cm-1 vo = 66623.6 cm-1 65635.1 584.3 520.0 683.5 698.2 61 1.9 553.9 651.6 693-0 587.0 5564 646.5 635.1 695-5 559.9 525.6 487.8 639.4 624.8 66579.5 5 17.0 437.0 638.9 655-2 579.5 55 1.3 620.1 610.9 650.6 555.5 520.4 614.9 600.8 527.8 489.8 447.5 607-4 588.7 * Much of this band is overlapped by a band of the k-2 system.-f The K structure of each J line is not resolved. The K values listed are those which contribute substantially to the intensity of the line.I64 STATES OF NH3 TABLE 3-contd. v (cm-1) 704.2 709.3 630-2 61 1-9 618-9 596.9 605-3 579.6 589.7 572.1 553.9 Y (cm-1) 566.0 668.4 597.9 547.5 586.6 572.9 540.1 505-1 557.4 540.1 520.4 vo = 67608.9 * v (cm-1) branch (J)K 67604.8 QW 599.4 Q(W 591.7 Q(3)3 582.2 Q(4)4 570.0 Q(515 570.0 Q(413 557-0 Q(6)6 541.8 Q(7)7 532.7 P(3)OJ 524.1 Q(QS 5 12.3 P(4)3 504.4 Q(919 498.2 P(4) 1 * This band is overlapped by a band of CO. The bands of both NH3 and ND3 show a characteristic alternation in that alter- nate bands have strong R(0) lines.As with the 2-3 bands this characteristic is very useful in determining the vibronic symmetry of the excited state levels. TABLE 4.-wAVE-NUMBERS OF THE LINES OF THE c--? BANDS OF ND3 vo = 63581-67 vz = 64301.52 Y branch(J) K branch(J)K 63 571 -50 547.90 520.98 490.16 599.54 613.39 560.1 1 534.69 505-33 580.65 599.54 606.54 618.96 578-56 606.54 64279.69 267.07 253.17 238.16 221.78 204.72 18 6.03 3 18.36 324.97 330.66 334.75 300-32 298.09 294.19 294.79A . E. DOUGLAS TABLE 4-cantd. 165 575.38 571.50 565-60 559.27 551.65 543.21 533.44 65020.94 6499 6.0 1 64965.01 64928.40 65047.37 65057.38 65009.39 64996-0 1 6498 1 -37 64965.01 64947.98 64928.40 65030.02 65026.70 65047.3 7 65053.31 65057.38 64996.89 6498244 64965.68 64947-98 64928.40 65027.72 65023.1 6 65053-98 65057.38 64983.22 64966.96 64949.1 5 66746- 18 71 7.07 776-98 V 289.8 1 283-62 290.53 284.27 285-17 277-77 278.75 269.99 27 1 *27 261-28 262.79 251.57 253.17 240.83 242.72 231.13 YO = 65031.31 64929.93 65 024.43 65018-30 65010.72 65001.68 65059.36 64968-73 64950.78 6493 1.58 6 5020.0 8 650 12-44 65003 a4 1 650 14.74 65005.56 64995.03 64972.1 5 65008.36 64997.73 6498 5-53 65001 -01 64988.80 64975.21 64992.70 64978.85 64983.22 64968-06 64972.79 6496 1 *45 63949.15 vo = 65768.54 cm-1 65705.70 P(5)4 686.73 P(Q4 757.16 Q(4)4166 STATES OF NH3 TABLE 4-contd.789-01 746.63 7 17.07 699.97 767.18 789.01 733-96 7 1 8-49 70 1 -26 682.23 764.79 752.60 793.69 794-25 794.25 792.46 720-47 703.00 683.97 663.42 761.57 75425 735.15 795.43 796.9 1 796.9 1 738.35 72410 800.9 1 793.69 730.13 714-10 696.23 72 1 -02 703.09 683.97 710.79 699.72 66474.87 440.39 526-28 532.88 531-10 52 1 -49 489.84 475-41 459.18 440-91 510.33 506.28 526-9 1 531.10 748.33 737-78 725.44 799.51 799.5 1 796.91 690.57 669-32 751.89 741.25 728-92 714-82 802.89 800.90 796.91 790.80 67367 650.97 745.62 733-05 719.00 805.05 800.9 1 795.43 656-09 632.18 vo = 66511.83 6 6 5 2 9 * 3 5 463.4 1 445.05 504.56 496.47 48632 537.63 537.63 535-71 533.27 448.90 500.26 490.1 2 478.1 3A .E . DOUGLAS 167 TABLE 4-contd. 533.27 533.27 531-60 527.66 521.70 477-07 460.76 442.42 507.93 501.72 493-80 532.88 534.87 53487 533.27 67285.2 275.0 268.6 259.0 256.6 253.3 248.9 243.9 2376 230.2 541.54 539.48 535.71 433.39 495-00 482.69 468.69 544.49 540.44 488.82 474.87 458.39 546.46 48 1 -69 465.25 547.44 473.65 455-10 464-65 455.10 YO = 67260054 TABLE 5.-cONSTANTS DERIVED FROM THE c-? BANDS OF AMMONIA NH3 vt-v*t T;+G'(v2) B' C' vibronic transition (n+4)-0 67 60 8 -9 7-10 5.19 AS-A; (n+3)-0 6 6 6 2 4 - 2 7.615 5.166 Ai-A'; (n + 2) - 0 65654.9 8.1 1 5.12 AS-Ai (n+ 1)-0 64700.4 8.66 5-06 A1-A; n-0 63771.1 9-17 5.03 A"-A 1 N D 3 (n+ 6)- 0 67260.54 4.005 2.67 1 AZ-A; (n+5)-0 6651 1.83 4-122 2.645 Ai-A," (n+ 4) - 0 65768.59 4.246 2.626 A;I--A' (n+2)-0 64301.52 4-517 2.587 A;-A, (n+3)-0 6503 1-33 4.374 2.604 A;-A! (n+ 1)-0 63 5 8 1 *67 4.65 2.573 Ai-4,168 STATES OF NH3 SPECTRUM OF 15NH3 Because the isotope shift of ND3 bands with respect to those of NH3 is so large that the relationship between the two sets of bands cannot be established, the spectrum of 15NH3 was investigated. Here the isotope shift is small and there is no doubt which pairs of bands correspond.Since the rotational structures of the 14NH3 and 15NH3 bands are nearly identical, the isotope shift could be determined by measuring the relative wavelengths of a few corresponding lines in each band. The isotope shifts in the positions of a number of bands of the B-2 and C-2 systems, determined in this way are given in table 6. Once the 15NH3 isotope shift has been established, it was possible to find the correlation between the NH3 and ND3 bands. The NH3-ND3 isotope shift is also given in table 6. TABLE 6.-ISOTOPE SHIFTS OF THE 15NH3 AND ND3 BANDS OF AMMOMA band system 98-9’8 vO(NHSh’O(ND3) v o ( m 3 ) - l ’ 0 ( ’ ~ ~ 3 ) A-X €3-X 0-0 - 6-0 5-0 2 - 0 1 4 0 4 * (n+3)-0 (n+2)-0 (n+ 1)-0 (n- 1)-0 (n-2)-o* c-x n-0 584 cm-1 33-1 28.4 11.5 5.9 o*o* 1592.9 1353.4 11 18.7 896* 683* 488* 35.2 30.4 24.7 19.5* 14.3 * 9*0* These values have been obtained by extrapolating the observed values.OTHER BAND SYSTEMS All bands of NH3 and ND3 lying below 1400 A are so diffuse that even the sharpest of them show no well-resolved rotational structure. It appears that little definite information can be obtained from high resolution spectra and therefore spectra in this region have not been investigated in detail. No triplet states of NH3 are known in spite of the fact that the lowest electronically excited state probably is a triplet state.In an attempt to find absorption to the lowest triplet state, the absorption spectra of NH3 and ND3 were examined in a 5-m multiple reflection cell with absorption paths up to 150 m. The spectrum was examined from 3500A down to the A-2 absorption system and with ammonia pressures up to 1 atm. A number of absorption bands arising from impurities were found but none were found which could, with any certainty, be attributed to the triplet state of ammonia. ELECTRONIC STATES OF AMMONIA The symmetries and statistical weights of vibronic and rotational levels of NH3 and ND3 in the ground state and in excited states of D3h symmetry have been dis- cussed in previous papers.19299 In the following discussion, the symmetric and antisymmetric inversion levels of the ground state of ammonia will be designated by the symbols A; and A,”.For our present purposes, it is important to note that A i - 4 ; vibronic transitions in NH3 will have only odd J lines present in the K” = 0A. E. DOUGLAS 169 sub-band, whereas A;-Ai transitions will have only even lines present. In ND3, a ten-to-one statistical weight alternation in the K = 0 levels results in a similar effect but here the A;--A; transitions have strong odd numbered lines and A,"-A; strong even lines. In the parallel bands which result from these types of transition, the effect can be seen most readily in the R(0) lines. In the 0-0 band of the 2-2 system of ND3, the R(0) line is strong, thus showing that the upper state is vibronically A," and therefore since this is the u" = 0 level, 2 2 - P - 0 g4S FIG.3.-A schematic diagram showing the potential energy curves of NH3. D3h symmetry is assumed throughout since all known excited states of NH3 are of D3h symmetry and the energy required to force the ground state into this symmetry is only 3OOOcm-1. Other singlet states are noted in parentheses. electronically A;. The 1 4 band has a very weak R(0) line from which it follows that the upper state is vibronically A;. This A; vibronic level can arise only from the excitation of a vibration of A," symmetry and this can only be the out-of-plane vibration of a-planar ammonia molecule. Thus, the rotational structures of the bands of the A-;i7? system, in complete agreement with the conclusions of Walsh and Warsop,l show that the upper state is electronically of A; symmetry and that excitation of the v2 vibration gives the long progression of-bands.The diffuse Q head of the 0-0 band of the NH3, 2-X system was found to lie at 46160 cm-1, a value different from and probably more accurate than that given by previous observers. The band origin of the ND3 0 4 band is at 4671343 cm-1.170 STATES OF NH3 The frequency of the 0-0 band of ND3 is 554 cm-1 higher than that of NH3 thus implying that the vibrational frequencies of the upper state are much lower than those of the ground state. If it is assumed that v1 = v3 = 2v4 in both the upper and lower state then the obsezed shift of the 0-0 band corresponds to a value of about 2550 cm-1 for v1 in the A state of NH3 and 1810 cm-1 in ND3.Frequencies cal- culated in this way are quite inaccurate and the observed band of ND3 at 48908 cm-1 may correspond to excitation of v1 = 2194 cm-1 in the excited state. The present work on the &-x system has established two points. First, the 15NH3 isotope shift for the 0-4 band is very small and therefore the vibrational frequencies in the excited state must be approximately the same as in the ground state. Secondly, the observation of a Zeeman effect with a g value of 0.6, establishes that the state is a degenerate electronic state. The high value of g suggests that the e type orbitals responsible for the degeneracy of this state play little part in the N-H bonds. Prior to considering the C-x bands it is useful to review the states expected for ammonia from simple theory.In ammonia, as in all other hydrides, the states derived from the united atom concept should closely resemble the observed states. Thus, one useful method of determining the states of ammonia is to split the states of neon by a field of D3a symmetry. A second method of determining these states is to build up the molecular orbitals from those of N and €3. This has been done by Mulliken 10 and by Walsh 11 who have shown that the electron configuration of the ground state is (ls~)2(a;)2(e')4(a2")2. Orbitals of these symmetries can be derived from the united atom but since the e' and a; orbitals, both of which are derived from the 2p atomic orbital of the united atom, may be very different in energy, a notation which does not involve the united atom is desirable for the ground state. In orbitals with higher principal quantum number, the energy difference between molecular orbitals derived from the same atomic orbital may be expected to be small.This suggests that the orbitals of ammonia can best be written in the following order: ( ls~)(a;)(e')(a~(3sa;)(3pe')(3pa~(4sa;) . . . where the notation implies that (3saf)(3pe') . . . are similar to atomic orbitals. Thus the first few states of ammonia are expected to be : * (1 s ~ ) ~ ( a 1,)2(e1)4(a i), 'A;(C,,) = 3, (1) (2) (1sN)2(a ;)2(e')4(a 2")(3pe') 'E" = 8, (3) ( l ~ ~ ) ~ ( a 1 , ) ~ ( e ' ) ~ ( a i ) ( 3 p a ~ ) ' A ; , (4) (1 qJ2( a ;)2 ( e')3 (a g)2 (3sa 1,) ( 5 ) 1 If - (.1s~)~(a;)~(e')~(a;)(3sa;) A, - 2, E '. Fig. 3 gives a schematic potential energy diagram for ammonia correlating the states of the separated atoms and the united atom.The correlation between the observed and predicted states is satisfactory for the 3, 2 and f3 states. The ground state is, in D3h notation, of 1A; symmetry as pre- dicted. Also, as predicted, the first singlet excited state is of A; symmetry and there seems no doubt that the A state is the A: state resulting from configuration (2)- The state which is certainly a degenerate state could be correlated with either configuration (3) or (5) but the evidence is strongly in favour of (3). If the vibrational analysis given by Douglas and Hollas is correct then the state is E", not E', and * Walsh and Warsop 1 have described an additional state (lsr)2(a;)Z(e34(u~)(a;). There appears to be no way that such an additional state can arise from the united atom and here it is assumed that this state is a different description of the state (2) given above.A.E. DOUGLAS 171 must be associated with (3). Also the high g value observed here could be expected from configuration (3) whereas the e type orbital excited in configuration (5) is strongly associated with the bonding and the state should exhibit a strong Jahn-Teller effect and a corresponding low g value.12 state is not certain. It seems reasonable to associate the state with the 1A; state of configuration (4). Since the 3p electron plays little part in the bonding, this 1A; state should be similar to the state. Indeed, the c and s" states are similar in their B values, in their large cc2 values and in their 0 2 values.There is, however, as discussed below, a major difficulty to this assignment of the c state. The vibrational numbering of the bands of the c-2 system cannot be deter- mined directly from the observed spectra. Alternate bands of this system do have strong R(0) lines, thus indicating that the excited state is planar or nearly so. Since, except for the effects of zero point vibration, B = 2C in the lowest vibrational level of a planar symmetric top we can determine the vibrational numbering by finding the level wherein this condition is fulfilled. This level is not observed directly but can be determined by extrapolation. The large value of cc2 makes the extrapolation reasonably accurate. The extrapolated values of VO, B and C for NH3 and ND3 are given in table 7.Here we see that the level of NH3 which comes closest to fulfilling the relationship B = 2C, is the vibronically A; level at 62010 cm-1 and that this level has a NH3-ND3 isotope shift of 488 cm-1. This may be compared with the state where, from 15NH3, we find the isotope shift is zero in the level where B = 2C. If the vibrational level of NH3 at 61207 cm-1 is considered to be the zero level then it is vibronically A; but the relationship B = 2C is less well fulfilled and there is still an isotope shift of 320 cm-1. This isotope shift would correspond to v1 having the extremely high value of -4000 cm-1 in the upper state (assuming 0 1 = 0 3 = 204). Thus the assignment of the state to electron configuration (4) leads to the unlikely result that the 3pa; orbital has quite different bonding prop- erties than the 3pe' orbital.The assignment of the TABLE 7.-EXTRAPOLATED VALUjS FOR THE CONSTANTS OF CORRESPONDING LEVELS OF THE C STATE OF NH3 AND N D 3 M I 3 ND3 To+ C ( U 2 ) B c To+ G(02) B C ,,, vibronic symmetry (n- l)-Ai 62870 9.67 5-00 62186 4.91 2.56 (n- 2)- A; 62010 10.18 4.97 61522 5.04 2-55 (n- 3) - A 61207 10.68 4.94 60887 5.17 2.53 (n- 4)- A," 60476 11.2 4.91 60289 5.30 2.52 (n- 5)- A ; 59825 11.7 4.9 59738 5.4 2.51 As an alternative assignment, the c-x bands may be attributed to a vibrational progression of the B-X system. If one quantum of a degenerate vibration of the B' state is excited E", A;' and A; vibronic levels result. The E" and A; vibronic levels have symmetry species which can combine with the ground state but unless the intensity of these transitions is enhanced by some interaction, such as Fermi resonance or strong vibrational-electronic interaction, we should expect the transi- tion probability to be low. Considering this possibility and assuming the 62010 cm-1 band of NH3 is the A: band resulting from the excited E vibration, we find values of 2786 cm-1 and -2300 cm-1 for the frequency of the degenerate vibration in NH3 and ND3 respectively. The ratio of these two frequencies is far from that expected from the normal isotope effect and even further from the ratio expected in a molecull172 STATES OF NH3 exhibiting a Jahn-Teller effect.Also these frequencies are much lower than the corresponding frequencies in the ground state and thus are incompatible with the observation that the isotope shift of the 0 4 band is extremely small.Therefore it is most unlikely that the c-x bands are a progression of the &-x system. Probably the most likely explanation of the state is that it is the 1A; state of electron configuration (4) above, but that its B value is affected by Coriolis inter- action of the B' and states. The interaction of a degenerate and a non-degenerate vibrational level of the same electronic state has been treated by Garing, Nielsen and Ra0.13 This interaction may lead to a " giant Z-type doubling " in the K = 1 levels of the degenerate state and a related change in the B value of the non-de- generate state. The interaction of an electronic A and E state should have a similar effect. A large Z-type doubling has been observed in the B-state and it is quite possible that it arises from the interaction of the c and B states.If the NH3 level at 59825 cm-1 (table 7) is considered to be the zero level of the C state this electronic state lies a few hundred cm-1 above the state. This arrangement of levels would, in the notation of Garing, Nielsen and Rao lead to a negative value of a'21la and an effective B value of the state greater than the true value." Thus, this shift of vibrational numbering will give an effective Bo much greater than 2C0 but the true Bo may still equal 2Co. No quantitative agreement can be expected between the perturbation in the c and B states since the a' value of the degenerate state can be the sum of contributions from a number of sources.13~ 14 PREDISSOCIATION All the excited states of ammonia are predissociated. The predissociation is least noticeable in the B-2 bands where the lines of NH3 are about 0-4 cm-1 wide and those of ND3 are too narrow to be measured. The effects of predissociation are greater in the c--2 bands where the lines of the NH3 bands are about 1.0 cm-1 wide and again those of ND3 are much narrower.All lines within a band of the B--Z and C-2 systems show approximately the same width and the line width does not vary noticeably from band to band in the progressions. The predissociation of the 2 state is more complex. This predissociation can best be studied in the ND3 bands where the lines are much narrower than in NH3. The bands of ND3 with I& = 0 as their upper state are diffuse, having line widths of about 2.5 cm-1 while the lines in bands with ZI; = 1 are sharper with line widths of about 0.8 cm-1.All higher levels are much more diffuse. All bands of NH3 are too diffuse to show measurable rotational structure. The = 0 and ZJ; = 1 levels have about the same band width but there is a slow increase in band width with increasing 2~2 beyond the second level. Near 1600A there appears to be a weak continuum which may arise from higher levels of the 2-x system. Aside from the peculiar variation in the degree of predissociation with 212, it is difficult to understand how the 2 state predissociates. It is known that the absorp- tion of light into the 2-x bands leads to the reaction N H ~ + ~ V - - + N H ~ + H .~ ~ ~ 16 The potential energy curves for a hydrogen atom approaching a NH2 molecule to form NH3 are shown in fig. 4. Here there are only two curves lower than that of the A; state which could cause the predissociation. One is the ground state curve and it appears unlikely that two states so far removed from each other will interact strongly. The other lower curve is that of the unobserved 3A; state. Here also it appears unlikely that the interaction between states of different multiplicity would be strong enough to give the observed effect. In addition, the 3 4 state might be * Douglas and Hollas have assumed q = ct'Z/u to be positive but their analysis does not deter- mine the sign of this quantity. If q is negative the B values given by them must be increased by q.A .E. DOUGLAS 173 expected to follow a curve similar to that of the 1Al state and thus in itself be a stable state. state is sufficiently shallow to allow predissociation through the potential barrier without interaction with another state. There is some information on the depth of this curve. As pointed out earlier, the N-H stretching frequencies in the state are probably about 2600cm-1. Though this value is much lower than in the ground state it is still sufficiently high to indicate that there is a marked potential minimum and escape of a hydrogen atom through the potential barrier from the u = 0 level seems most unlikely. It may be assumed that the potential curve of the Bl I 2 -BI -2A, 2 - Bi r(H-NH2) FIG. 4.-A schematic diagram of the potential energy curve of a hydrogen atom approaching a NH2 molecule along its CzV axis.The peculiar shape of the A; curve is assumed to arise from the interaction of the two dashed curves. Only the lowest triplet state (a state which has not been observed) is shown as a dotted curve. There seems only two mechanisms through which the predissociation can occur. First, we can assume that the singlet-triplet interaction is strong and that the triplet state is unstable. Secondly, we may assume that the potential curve of the excited state is far from that expected for a state having simple valence forces. Thus, the potential curve given in fig. 4 may be valid for a normal NH;! molecule but for small changes in the NH2 bond distances and angles the barrier to predissociation may be much smaller. In this way certain relative phases of the two components of the degenerate stretching vibration in the molecule may lead to dissociation. This escape1 74 STATES OF NH3 route does not necessarily alter the vibrational frequencies very greatly. * There is no positive evidence that such a potential actually exists but the observations are not inconsistent with its existence. I wish to thank Dr. G. Herzberg for numerous suggestions during the course of this work and Mr. F. Alberti who photographed all the spectra used here. * Prof. H. C. Longuet-Higgins originated this concept and has developed it further by showing that a two-dimensional oscillator in a potential well entered by three symmetric valleys (i.e., an ash- tray potential) has its vibrational frequencies altered little by these valleys. 1 Walsh and Warsop, Trans. Faraday SOC., 1961, 57, 345. 2 Douglas and Hollas, Can. J. Physics, 1961, 39, 479. 3 Douglas and Potter, Appl. Optics, 1962, 1, 727. 4 Douglas, Can. J. Plzysics, 1958, 36, 147. 5 Benedict, Physic. Reu., 1935, 47, 641. 6 Benedict and Plyler, Can. J. Physics, 1957, 35, 1235. 7 Rao, Brun, Hoffman, Jones and McDowell, J. Mol. Spectr., 1961, 7, 362. 8 Crawford, Rev. Mod. Physics, 1934, 6, 90. 9 Dressler, J. Chem. Physics, 1960, 32, 1682. 10 Mulliken, J. Chem. Physics, 1935, 3, 506. 11 Walsh, J. Chem. SOC., 1953, 2296. 12 Strauss and Coulson, Proc. Roy. SOC. A , 1962, 269, 443. 13 Casing, Nielsen and Rao, J. i’dol. Specfr., 1959, 3, 495. 14 Child, Mol. Physics, 1962, 5, 391. 15 Dressler and Ramsay, Phil. Trans. A, 1959, 251, 453. 16 McNesby, Tanaka and Okabe, J. Chern. Physics, 1962,36, 605.