Photoion-pair formation in Cl2

Chemical Physics 123 (1988) 317-328 North-Holland, Amsterdam

P H O T O I O N - P A I R F O R M A T I O N I N C12 J. B E R K O W I T Z , C.A. M A Y H E W and B. RUSCIC Chemistry Division, Argonne National Laboratory', Argonne, IL 60439, USA Received 24 February 1988

The relative photoion yield curves of CI~, C1 and CI÷ from C12 (normalized to absolute cross sections) are presented, from threshold to 740 &. A structured spectrum involving the formation ofCl+-Cl - ion pairs, and amounting to ~ 1% of total ionization, extends from its threshold to ,~ 750 &. The spectrum is divided into four distinct bands. The two lowest in energy must correlate with CI+ (3P) .}_C1 (tS). The other two are inferred to have C1+ (~D) and CI+ (~S) asymptotes, respectively. The fine structure observed implies predissociation of photoexcited Rydberg states by ion-pair states. The Rydberg states can be assigned by their absolute energies, vibrational spacings and vibrational envelopes. The lowest-energyion-pair band can be observed at least 0.16 eV below its thermochemical threshold at 300 K. This band has a pronounced temperature effect, which can be interpreted as an example of heterogeneous predissociation. The contribution to CI+ from dissociative ionization rapidly overwhelms the ion-pair component below 802.5 ~, the dissociative ionization threshold.

1. Introduction Considerable attention has been directed to the study of molecular chlorine in the ultraviolet, and especially the v a c u u m ultraviolet, in the last few years. This has included m e a s u r e m e n t s of the absolute photoabsorption and p h o t o i o n i z a t i o n cross sections [ 1 ], a higher-resolution photoelectron spectrum [2], the partial cross sections for formation of various states of CI~ [ 3 ], the fragmentation cross section [ 1 ], detailed spectroscopic studies of the electronic spectrum in the VUV [ 4 - 6 ] below the ionization threshold using c o n v e n t i o n a l light sources [4], synchrotrons [ 5,6 ] a n d lasers with two- or three-photon excitation [ 7 - 9 ] . Additional i n f o r m a t i o n has been forthcoming from electron energy loss spectroscopy [10,11]. The interpretation of several of these experimental results has been based on the ab initio calculations of Peyerimhoff and Buenker [ 12 ]. These authors, using their M R D CI method, have calculated potential energy curves for the ground state and a n u m b e r of excited singlet and triplet states (but without s p i n - o r b i t interaction). A striking feature of these calculations is the n u m b e r of avoided crossings between valence and Rydberg states, which give rise to adiabatic double-well potential energy curves. Quite a few of these 0 3 0 1 - 0 1 0 4 / 8 8 / $ 03.50 © Elsevier Science Publishers B.V. ( N o r t h - H o l l a n d Physics Publishing Division )

curves correlate with ion pairs (C1 ÷ - C I - ); they have an outer well with ion-pair character, and an i n n e r well with valence or Rydberg character. A remarkable example of this double-well character is provided by Moeller et al. [5]. These authors examined the fluorescence consequent upon excitation of the XIZ + - , 11Zu+ electronic transition, with slightly different excitation energies (73440 and 73560 c m - ~ ) . In the region of the fluorescence spectrum corresponding to 0 ~
318

J. Berkowitz et al. I Photoion-pair forrnaHon in Cl:

examination of their figure shows that the observed threshold for ion-pair formation occurs significantly below the calculated thermochemical threshold. We shall have occasion to return to this observation in the analysis of our own results. It can readily be shown from well established thermochemical data that ion-pair formation in C12 has a threshold above the ionization potential of C12. The region from just below the ionization threshold to higher energies is much less well understood than the electronic spectrum to lower energies. Just below the ionization threshold, the spectrum is highly congested [4]; above the ionization threshold, it is broadened. However, with the technique of photoionization mass spectrometry, and selection of C1as the observed product ion, it is possible to observe even weak processes in the presence of much more abundant ones. We shall show that the processes giving rise to this product are highly structured, and offer the possibility of obtaining detailed information about excited states above the ionization potential.

2. Experimental arrangement The basic photoionization mass spectrometric apparatus has been described previously [15]. In addition to the ordinary target gas chamber operating at ambient temperature, another target chamber was employed which had provision for cooling the incoming gas to a temperature as low as 78 K. An ironconstantan thermocouple was employed to monitor the temperature of this chamber, which for the present purposes was maintained at ~ 160 K. For the measurement of negative ions, the potentials of the ion repeller and ion focusing plates were reversed, and the ion multiplier was floated at high positive potential, so that the impinging C1- ions would strike the first dynode with 600 eV of kinetic energy. Under these conditions, the sensitivity to C1 ions was ~ 10% higher than to C1+ ions measured under normal positive ion conditions. Both the hydrogen many-line spectrum and the Hopfield continuum of helium were used as VUV light sources, Most of the experiments were conducted with a resolution of 0.28 & (fwhm), but a selected set, described below, was performed with 0.14 (fwhm).

3. Experimental results 3.1. General observations An overview of the experimental results obtained in the present research is shown in fig. 1. The ordinate has been placed on an absolute scale by comparison with the total ionization cross sections given by Samson and Angel [ 1 ]. The spectral dependence of the C1+ ion yield is similar to that reported earlier by Dibeler et al. [13,14], but the present results have been obtained with higher resolution and better statistics, thereby enabling us to observe a large number of autoionizing features. Photoion-pair formation amounts to about 1% of total ionization at 780 &, and even less at longer and shorter wavelengths. The wavelengths of structural features in the ion-pair curves appear to be very nearly the same as in the C12~ curve, but with quite different relative intensities. Above ~ 820 &, the C1+ and C1- photoion yield curves are identical, within experimental error. However, at shorter wavelengths the C1+ photoion yield curve has a contribution from photodissociative ionization, as well as ion-pair formation, whereas the CIphotion yield curve continues to represent only ionpair formation. The C1+ contribution from dissociative ionization rapidly overwhelms the C1+ contribution from ion-pair formation beyond the dissociative ionization threshold. Therefore, the curve labeled C1+ represents predominantly photodissocia-

120

Po ~F, 3

~o CI,

800

900 WAVELENGTh (£)

tOO0

Fig. 1. Spectral dependence of the photoion yield of CI ] , C l (ion-pair formation) and C1 + (photodissociative ionization) from C12. The curves are normalized to the absolute ionization cross section of Samson and Angel [ 1 ]. The thresholds for formation of various ion pairs are also indicated.

J. Berkowitz et al. / Photoion-pairformation in (72

tive ionization. The onset of this process is to somewhat longer wavelength than its calculated thermochemical threshold (802.66 /k), presumably due to the contribution of internal thermal energy at 300 K. Also noted in fig. 1 are the calculated thresholds for ion-pair formation to C1+ (3p2,~,o) + C1- (~So), C I + ( I D 2 ) + C I - ( I S o ) and CI+(~S0)+CI-(ISo). Thus, energy constraints limit the first two bands in the ion-pair spectrum (from threshold to 940/k) to C1+ (3p)+C1-(~So) products. The third band can, and probably does correlate with CI+(~D2)+ CI-(~So), and the fourth band probably correlates with C1+ (~So) + C1- ( t S0). In the ensuing sub-sections, we shall present more detailed analyses of (a) the C1+ photoion yield curve and (b) each of the ionpair bands, in turn. 3.2. The C l { photoion yield curve

In fig. 2, the photoion yield curve of the parent ion, CI ~, is compared with the electron energy loss (EEL) curve of Stubbs et al. [ 11 ]. In the region displayed, dissociative ionization has not yet begun, and ion-pair formation represents < 1% of total ionization. Furthermore, according to Samson and Angel [1] the quantum yield of ionization is unity in this region. Therefore, the absolute cross section and spectral dependence of photoabsorption and photoionization should be identical, within experimental uncertainties and the qualification cited. The EEL curve [ 11 ] was obtained with 100 eV incident electron energy,

"'rWl

/

/ t,..'.'~*-"

, ....

O ON,Z . . . . .

,

/\

(ELECTRON ENERGY LOSS)

JY'. i .:/"L "-~

" v.~,,,.,.~

JJJt _

I 11.6

,

L , _ 12.2 EXCITATION

I 12.8 ENERGY

I 15.4 (eV)

,

I [4.0

•

Fig. 2. A portion of the photoion yield curve of Cl~, compared with the electron energy loss curve of Stubbs et al. [ 11].

319

at an angle of 2°. It should approximate a photoabsorption curve, although the incident electron energy is not as high as necessary for rigorous application of the Born approximation. The broad general features of the two spectra are similar. However, the intensities do not match up as well as one would expect. If we normalize the two spectra near the maximum absorption at ~ 12.8 eV, the photoionization curve lies significantly higher at 11.75 eV, and especially above 13.4 eV. Since our photoionization curve is in rather good agreement with the absolute measurements of Samson and Angel [ 1 ], we believe that the problem lies in the EEL curve. This was obtained with a position-sensitive detector, spanning 0.25 eV, which was then overlapped with successive 0.25 eV ranges until the extended spectrum, covering about 4 eV, was accumulated. The lens system would have to be continually adjusted to ensure constant transmission over this region, and that may have been difficult to achieve. The best energy resolution displayed in the EEL spectra [ 11 ] was 18 meV (fwhm). The corresponding resolution of the photoionization data shown in fig. 2 was 3.6 meV. The higher resolution manifests itself in two ways. We observe additional, finer structure and larger peak-to-valley excursions. In table 1, we compare the energies (in eV) of the autoionization peaks observed in the present experiment with those observed in the EEL experiment. Between threshold and ~ 12.5 eV, many more peaks can be noted in photoionization than in EEL. For the intense band between 12.5-13.0 eV they are comparable, and the agreement between the absolute energies of the two experiments is rather good. Above 13.0 eV, there are again many more peaks in the photoionization experiment. The straightforward interpretation is that the EEL experiment, with 18 meV resolution, involves blending of finer structural details. However, the lowerresolution results may actually simplify the assignment. In the region above the ionization threshold, Stubbs et al. [11] describe four vibrational series (labeled by them J, K, L and M; we include J, L and M in fig. 2 ) which are electric dipole allowed, according to their analysis. From their table 3, the average vibrational separation is 0.045, 0.044 and 0.0435 eV for J, L and M, respectively.

J. Berko~,/itz et aL / Pkotoion-pair formation in 67:

320

Table 1 Comparison ofautoionization peak energies (eV) observed in photoionization (CI~, CI+-C1 - ) with electron energy loss measurements ofStubbs et al. [2 ]. (The conversion factor 12398.52 eVA has been used) Ct + 11.487 11.493 11.513-11.517 11.522 11.540 11.564 11.573 11.584 11.589 11.598 11.607 11.620-11.623 11.634-11.640 11.650 11.665-11.684 11.718 12,526 12.544 12.562

12.604 12.626 12.635 12.660 12.672 12.690 12.701 12,722 12.738

Ion pairs

EEL

12.995 13.006 11.513 13.019 13.037 13.051

11.593

12.840 12.848 12.864 12.879

13.070 13.082 13.109 13.127

11.670

13.168-13.186

12.528 t2.533 13.224 12.561 12.571 12.593 12.605 12.621 12.634 12,656-12.662

12.565

12.605

EEL

CI +

Ion pairs

13.007 13.012 13.018 13.032 13.052

13.005

13.894 13.920

13.897 13.927

13.296 13.338 13.369

12.699 13.431

12.744 12.762 12.777-12.784

13.464 13.477

13.516 13,527

12.797 12.825 12.832 12.841

13.562 13.596 12.834 13.640-13.628

12.863 12.880

13.677

12.882

13.386 13.402 13.425 13.433 13.452 13.462 13.477 13.490 13.507 13.516 13.523 13.547 13.561 13.602 13.635-13.627 13.649 13.667 13.677 13.708

12.889 12.896 12.909 t2.927 12.937 12.956 12.971 12.985

13.934 13.961

12.929 12.939 12.956 12.970 12.986

13.723 13.763 12,926 13.808 12.953 12.971

13.850

13.723 13.760 13,79t 13.806 13.831 13.851 13,881

13.977

13.973-13.988 14.012

14.023

14.020-14.032 14.056

14.066

14.062-14.073 13.112 14,100

13.143 13.154 13.164 13.i73 13.176 13,189 13.213 13.241 13.254 13.276 13.295 13.338 13.367

EEL

13.940

13.170

14.145 14.186 14.209 14.292 14,309

13.224 14.348

14.086 t4.096 14.131 14.182 14.209 14.312 14.334 14,345 14.375

14.382 13.290 13.335 13,380

13.387 12.701 12.718 12,735

13.071 13.083 13.093 13.113

12.655

12.785 12.792 12.806 12.827

Ion pairs

13.063

12.742 12.747 12.753 12.763 12,780

CI~

14.388 14.424 14.464 14.505

13.429 13.466

14.430 14.460 14.501 14.535 14.578 14.617

15.547 15.576 15.615 15.641 15.680 15.706

13.516 15.714

15.733 13.559 13.599

15.742 15.765 15.778 15.796

13.631 15.814

15.828 13.674

15.843

13.715

t5.871

13.757

15.902

13.803 13.844

15.930

15.858 15.890 15.919

13.884

15.952 15.975

J. Berkowitz et al. /Photoion-pairformation in Cl 2

This matches well with the vibrational spacing of the CI~ ( 2Fluj) ionic state given by van Lonkhuyzen and de Lange [2 ]. By contrast, the vibrational spacing of the CI~ ground state ( 2FIg,i ) is 0.079 eV, and that of the second excited state (2E~-) is 0.034 eV. Hence, it seems plausible to conclude that J, L and M are Rydberg states converging to 2Fluj. This conforms with the assignment of Stubbs et al. However, they describe these transitions as •"(~g2 ~u4 ~ g4,, y ~ 2

~

""(~g2 ~ u3 ~g4 nS (~g, I ~ u+ •

The product of the open shells nu × Cygmust be nu, so such Rydberg states would have to be IH u. The effective principal quantum numbers given by Stubbs et al. for the series J, L and M are 2.85, 3.76 and 4.68, respectively, which are roughly consistent with a quantum defect of 2.2, characteristic of an s-like Rydberg electron. However, their method of arriving at these effective quantum numbers needs revision. It is based on their assignment of v' = 0 for the J, L and M series, and a value of 14.252 eV for the adiabatic IP of A 21-Iu.3/2obtained from Huber and Herzberg [ 16 ]. The vibrational progressions of both the Rydberg states and the convergence limit span a very large Franck-Condon region when the transitions originate from the neutral ground state. The v' = 0 components are weak, and difficult to uniquely assign. Thus, van Lonkhuyzen and de Lange [2 ] have concluded from their higher-resolution PES and new emission spectra [ 17 ] that the adiabatic IP is 14.04 eV. This value, together with the v' = 0 assignments of J, L and M given by Stubbs et al. give very irregular effective quantum numbers. A more reliable approach is to combine the largest vibrational feature in J, L and M with the vertical IP, given by van Lonkhuyzen and de Lange as 14.393 eV. In this way, one obtains n*=2.95, 3.94 and 4.98 for J, L and M, respectively, which is consistent with their observation of a quantum defect of 2.04 for ns progressions in C12. The regularities in the vibrational intervals indicate either that one. Rydberg state is dominant in each of the bands J, L and M, or that the vibrational components of two or more Rydberg states mesh. In this context, it is noteworthy that the convergence limit is actually two states, A21-Iu,3/2.~/2, which have equal probability on geometric (angular momentum degeneracy) grounds, but are very difficult to un-

321

ravel in both the photoelectron spectrum [ 2 ] and the emission spectrum [ 17 ] of CI~, ostensibly because two vibrational quanta are nearly equal to the spinorbit splitting. The band designated "K" by Stubbs et al. may be the high-energy portion of the J band. This would be consistent with the breadth of the A2Hu,, band in the photoelectron spectrum of Cl2. The spectral region near the ionization threshold appears to be dominated by the " G " band (see fig. 2). The ionization threshold occurs just below v ' = 4 of this band, preventing our observation of the stronger, lower energy features. Here again, the lowerresolution EEL spectrum is helpful in drawing our attention to the dominant, albeit blended features, which have a spacing of 0.079-0.080 eV. Such a vibrational spacing matches that observed [2 ] for the ionic ground state, and therefore this Rydberg component must be a member of a series converging to X2Hg,i.

It is very difficult to assign vibrational numbering to Rydberg states converging to Cl~ ( 2FIgj) because the vibrational spacing, the spin-orbit splitting (2Hg,3/2-2Hg, I/2 ), and the separation between successive Rydberg members in the vicinity of the adiabatic IP are approximately the same. It seems likely that the allowed transitions in this energy range are ... ~ g4 --~ ~ g3 HpO'u, l I~ u

-,ng3npnu, 1• u

.

(The Rydberg electron could also have nd character, but nSffgis optically forbidden. ) Our best estimate of the quantum defect for an np~u series is obtained by starting with the lowest member of this series assigned by Stubbs et al., which is their P band. If we identify their assigned v' = 0 for the P band with v= 0 for X 21-Ig.3/2 of CI~, then n* = 2.457, and the higher v' will correspondingly converge to higher v' ofX 2Hg, and have about the same n*. If they have failed to see v' = 0, and instead have asigned v' = 0 to what is really v' = 1, then n*=2.415. In either event, the large gap between this Rydberg member and the first IP enables us to make a rough first estimate of this quantum defect. The next member of this Rydberg series is designated A3 (5pou) by Stubbs et al. Two distinct members appear in their fig. 3, the third being very weak. If we identify the 10.406 eV peak with v= 1 of CI~, and the 10.475 eV peak with u= 2, as they have,

322

Z Berkowitz et aL / Photoion-pair fi~rmation in Cl_,

then n*=3.419. By this bootstrapping method, we also arrive at n * = 4 . 3 8 9 for v' = 2 of the E (6pau) band correlated with v= 2 of CI~, which matches their assignment. On this basis, the quantum defect for npcy is ~ 1.6. This is precisely the p quantum defect found by Shaw et al. [ 18 ] in an investigation o f inner-shell excited states of Clz using electron impact excitation, and also by Schwartz [ 19 ] in excitation from the L shell of HC1. Stubbs et al. [ 11 ] state that the effective principal q u a n t u m number corresponding to the " G " band is 6.501, and this leads them to assign this feature to the transition

t

1040 ---+ " " t ~ g 2T t u /4' C g 83 p ( Y u , . . . .t.x.211.4.rr4 g'*u"g

ll-Iu

"

Such an assignment implies a p q u a n t u m defect of -~ 1.5, which is close to the inferred value. It seems likely that both np~u and nprtu series are excited (they should have similar quantum defects), and that such series may converge to both 2Hg,3/2 and 2Hg, l/2. This is suggested by the many autoionizing features observed in the photoionization spectrum, with a varying, but average separation of 0.015 eV, spanning the region from 11.487 to 11.718 eV (see table 1 ). Before leaving this spectral region, we wish to point out for subsequent reference the weak feature just below 11.48 eV (fig. 2), which we attribute to a hot band. It extends about 0.07 eV below the ionization threshold, and has an intensity <0.1 of the nearby peak, which is roughly consistent with a Boltzmann population for v" = 1 in neutral CI2 of ~ 7% at 300 K. 3.3. The ion-pair b a n d at ~ 1 0 4 0 - 1 0 6 0 ~4

The threshold for ion-pair production is readily calculated from three very well established quantities: D0(Cl2)=2.479367 eV [ 16], IP(C1) = i2.9674 eV [ 20 ], and EA (CI) = 3.61272 eV [ 21 ]. The resulting value is 11.834 eV-= 1047.7/~. Fig. 3b is an enlarged version o f this band, obtained at 300 K. About half of this band lies below the calculated threshold. Since the calculated value is deemed to be very reliable, several experiments were performed to determine the cause of this abnormally low experimental onset. Collisional effects were suspected, and tested by measuring the relative intensity of several features in this spectrum, below and above the calculated

1050 WAVELENGTH (.~)

1060

J070

Fig. 3. (a) The lowest-energy ion-pair band from C12,obtained at 160 K. (b) The same ion-pair band, obtained at 300 K, The calculated thermochemical threshold (corresponding to CI + ( 3P2) + C! (tSo) ) is marked.

threshold, as a function of target gas pressure. The relative intensity o f several C1+ features were also examined in this study, which covered about two orders of magnitude in pressure. The relative intensities of all features measured were independent of pressure, essentially ruling out bimolecular effects. Possible electric field effects were explored by measuring the relative intensities o f the aforementioned features as a function of the ion repeller voltage. Once again, no significant electric field effects manifested themselves. Finally, this ion-pair region was rescanned using an ionization chamber which was cooled to ~ 160 K, which is close to the lowest temperature which could be achieved without condensing the C12. The resulting spectrum is shown in fig. 3a. It can readily be seen that most of the structure which appeared below the calculated threshold at 300 K (fig. 3b) has been eliminated at the lower temperature. Hence, thermal internal energy is implicated in this very low experimental onset. At 300 K, k T ~ 0 . 0 2 5 eV, which corresponds to ~ 2 . 2 A in this wavelength region. We had already noted that the Boltzmann population of v" = 1 (C12, 300 K) is only 7%, and this vibrational hot band would have the effect of lowering the threshold by 0.0687 e V = 5.7/~. We can readily see a C I - signal at least as far as 1062/~, which is almost 15 A ( ~ 0.16 eV) lower than the calculated threshold. This seems

323

J. Berkowitz et al. / Photoion-pair formation in Cl2

to be an extraordinary temperature effect, much more extended than that normally seen in primary ionization or dissociative ionization. It may also be present in the corresponding ion-pair formation process in Br2. The figure displaying this process in the papers of Dibeler et al. [ 13,14 ] extends at least 0.1 eV below the calculated ion-pair threshold, and also displays structure (presumably predissociation) below the threshold. A likely explanation of a long tail below the 0 K threshold is based on the rotational Boltzmann distribution, rather than vibrational, although both may contribute. Unlike the vibrational distribution, which declines exponentially with increasing v", the rotational Boltzmann population has a m a x i m u m at J' >> 0 and a long exponential tail. As we shall see later, the high J states, much more significantly populated at 300 K than at 160 K, can give rise to a nontrivial signal well below the nominal threshold. However, if a process which uses the excess rotational energy is to be as efficient as indicated by the experiment, the mechanism must somehow incorporate a probability factor which increases with rotational angular momentum, partially counteracting the attenuation of high rotational states imposed by the Boltzmann factor. A mechanism that has such a probability factor is heterogeneous predissociation (i.e. A / l = + 1 for H u n d ' s cases (a) and (b), Ag2= ___1 for H u n d ' s case ( c ) ) . The theory for this process was developed early on by Kronig [22], who has shown that when AA-- _+ 1 the rate of predissociation is linearly proportional to the rotational energy (i.e. approximately proportional to j 2 ). If each rotational level of the photoexcited state has a probability of predissociating which is approximately proportional to j 2 , then the combined influence of the rotational Boltzmann population and the enhanced rotational predissociation can be simulated by multiplying the rotational distribution by the j2 factor. Fig. 4 illustrates this effect at 160 and 300 K. This mechanism remained largely a curiosity for many years [23, p. 418], but in the more recent period a number of examples have been carefully investigated which reveal the presence of rotationally dependent predissociation [24] rates. Perhaps coincidentally, most of these examples involve the diatomic halogens and interhalogens. Heaven [25] has

~-~+

Xo**~ % k*

÷

×o ~ o4 o +°4 ,~ 44 xx÷ + oo4 k' _

+ + ÷+

o

.o o o4+ + x~÷ 6^6' !

o

+

x x

+ ~o * ~ ~o + ~° o × ~o +÷ a °o + ×x ~4 °o +÷+ •* 4 °% +++ ~× a44 OOo ++ ,×× aa ~ Oooo

0.04

o.~8

o,',2

•

++++

o?16......

o.2o

ROTATIONAL ENERGY (eV)

Fig. 4. The rotational Boltzmann population o f C12 as a function o f rotational energy, at 160 K ( X ) and 300 K ( O ) . M o d i f i e d

rotational distributions, weighted by a factor 7 2 tO simulate the effect of the heterogeneous predissociation probability, are also plotted at 160 K ( A ) and 300 K ( + ). provided a recent review of these studies. They primarily involve excitation of the halogen in the visible region (X ~Y.~ -~B 3H(0+ ) ); the 0~+ state is then predissociated by a repulsive ~Yl( 1u). The conditions for a heterogeneous predissociation exist, and in a number of cases listed by Heaven, the rate of predissociation is proportional to J ( J + 1 ). If the predissociation leads to the formation of an ion pair, as is the case in C12, high J states would not have to penetrate a rotational barrier which could retard the formation of products. The effective potential for a rotating system, including the electronic and centrifugal potentials, may be written as [23, pp. 426, 427 ] Us(r) = Uo(r) + (f72/2cltr 2 ) J ( J + 1 ) .

Usually, Uo(r) depends more strongly on r than r -2, and the effective potential manifests a barrier. This makes it more difficult for higher J states to dissociate. However, if Uo(r) is the potential energy of two ions (e.g., C1 + + C1- ) which varies asymptotically as r - 1, no potential maxima arise [ 23, p. 427 ]. In the case of the ion-pair band in C12, the primary excitation most likely involves Rydberg states, presumably converging to C1J- X 2Hg,3/2 a n d / o r 2Hg,l/2. Judging by the Franck-Condon envelope of the X 21"lg photoelectron band of C12 [2], the probable vibrational distribution of these Rydberg states is rather

324

J. Berkowitz et aL /Photoion-pairformation in C12

wide. Thus, in spite of the fact that the ion-pair threshold lies ~ 0.35 eV above the X 2I-Ig,3/2 ground state of CI~-, one can conclude that the primary excitation will rather densely propulate the region in the vicinity of the ion-pair formation limit. The states which are nominally below the ion-pair limit can still predissociate if the corresponding high J population is significant. The following examples should illustrate how these states could give rise to a significant tail below the ion-pair threshold through a mechanism involving heterogeneous predissociation. If there were a state with its J' = 0 level at the ionpair limit, and if this state were predissociated, the corresponding spectrum would show a rotational band centered at the threshold, in its full intensity. If, for example, there were another state whose J ' = 0 level was 200 cm-~ below the ion-pair limit, one would observe only a portion of the rotational band, corresponding to initial transitions to J' = 2 3 and higher (taking B = 0.179 cm ~, characteristic of CI~ X21-Ig, and hence o f the corresponding Rydberg states). Taking into account both the rotational absorption transition probability and the heterogeneous predissociation probability, J' (J' + 1 ), it can be shown that such a rotational band, centered at 200 c m - ~ below threshold, would still display 76% o f its total intensity at 300 K and 48% at 160 K. The temperature dependence becomes even more dramatic as the difference between the ion-pair limit and the nominal energy of the state which predissociates increases. For example, for a state lying 550 c m - J below the limit, levels with J' >/55 can predissociate. The corresponding rotational band would still display ~ 27% of its nominal intensity at 300 K, but only < 5% at 160 K. Such a rotationally enhanced predissociation tail extends to much longer wavelengths than a simple Boltzmann hot band tail. As mentioned earlier, the Boltzmann population ofv" = 1 C12 IZ+ ( ~ 550 c m - ~above v" = 0) is only 7% at 300 K. At still longer wavelengths, bands centered at 1000 and 1200 cm i below the threshold would display 5% and ~ 2.5% of their original intensities, respectively, if the j 2 dependent predissociation mechanism is invoked, but only 0.8% and 0.3% based on the simple Boltzmann distribution. Thus, both the temperature dependence and the persistence of the experimentally observed signal be-

low the ion-pair formation threshold, can be explained by a mechanism involving overlapping bound states which are populated by the initial excitation and which subsequently are heterogeneously predissociated. On energetic grounds, the predissociating (perturbing) state must asymptotically correlate with C1 + (3pg) d- C1- (ISg), and therefore must have 3H or 3 z - character. The neutral ground state is X JZ +, and therefore the initial photon-induced excitation is nominally spin-forbidden if the bound states are also triplets. However, this need not be a major impediment. Moeller et al. [5 ] had previously noted a significant absorption cross section in C12 from X ~Y~g to the 2 3I] u Rydberg state. Grein, Peyerimhoff and Buenker [26 ] have shown that strong spin-orbit interaction with 2 JFlu provides significant oscillator strength for the X JZ + ~ 2 3Hu transition. Subsequently, Lee et al. [ 6 ] measured an oscillator strength to 2 3Hu which was only a factor 3.7 weaker than that to 2 1H~. Alternatively, if the initial excitation is to an optically allowed singlet state, it may be predissociated by 3H~, which also implies spin-orbit interaction. Thus, the coupling of spin and orbital angular momentum, and probably rotational angular momentum, are somehow involved in the spectral and temperature dependence, and intensity of this ionpair band. Further insight into the nature of the states giving rise to this ion-pair band can be drawn from the following analogy. In fig. 5a, we show the excitation function for a VUV fluorescence band observed by Moeller et al. [ 5 ] between ~ 72000 and 76000 c m - ~. They assign the emitting state as 1 1Eu, although they also find the states 2 lI-Iu and 2 3Hu in the same energy region. These states all have vibrational quanta (inner well) of around 600 cm ~=-0.074 eV. Thus, they are to be associated with electron excitation from the uppermost occupied ng orbital. The dashed curve in fig. 5a is the envelope of F r a n c k - C o n d o n factors connecting the double-well upper state, taken to be 1 IZ u+, and the X JZ+ ground state, as calculated by these authors. In fig. 5b, we show the excitation function for a VUV fluorescence band between 90900 and 94400 c m - 1 observed by Lee et al. [ 6 ], but unassigned. Fig. 5c is our ion-pair band, between 93700 and 96600 c m - l . These three bands have approximately the same breadth, and general structure. This analogy

J. Berkowitz et al. / Photoion-pairformation in CI2

325

C bJ

7-

jl J "

--

I

'

I--'

76,000

I

g o o

72,000

CL

i

'

74,000 cm-I

(b) rY

I

95O 94,000

92,000 cm-/

96,000

94,000 cm-I

Fig. 5, (a) The excitation function for a VUV fluorescence band observed by Moeller et al. [ 51 between 72000 and 76000 cm- i. The dashed curve is the envelope of Franck-Condon factors calculated by these authors. (b) The excitation function for another VUV fluorescence band, between 90900 and 94400 cm-~, observed by Lee et al. [6] but unassigned. (c) The lowest-energy ion-pair band, observed in the present experiment. suggests that they m a y be m e m b e r s o f the same Rydberg series ( o r perhaps group o f series) converging tO X 2I[g#. Moeller et al. [ 5 ] analyzed their excitation function to e m a n a t e from an u p p e r state with " i n n e r well" v i b r a t i o n a l features, which migrates to the " o u t e r well" a n d fluoresces from there. We speculate that our ion-pair b a n d involves excitation to an inner well, which migrates to the outer well, and ultimately dissociates into C1 + + C1-. 3.4. The ion-pair band between 950 and 990 ~ This spectral region is displayed in greater detail in fig. 6. In table 1, we c o m p a r e the energies & t h e peaks in the ion-pair b a n d s with those o b s e r v e d in the CIjp h o t o i o n yield curve, and also with the EEL experim e n t [ 11 ]. The 9 5 0 - 9 9 0 tk b a n d matches the b a n d

960

J

I

_,

I

970 980 990 WAVELENGTH (~)

/

~

"960

I

970

Fig. 6. The ion-pair band between 950 and 990 A, obtained with a resolution of 0.28 tk (fwhm). A portion of the spectrum, as shown at the right hand side of the figure, has been obtained with 0.14/~ resolution. called J by Stubbs et al. We observe m o r e peaks in both the CI~- and ion-pair experiments than are reported in the EEL experiment. Presumably this is just a function o f the energy resolution o f the respective experiments. We have e x a m i n e d a densely structured p o r t i o n o f this b a n d with 0 . 1 4 / ~ - 0.0018 eV resolution ( f w h m ) , a n d found essentially the same widths and structures as with 0.28 A resolution (see fig. 6). Hence, the natural line widths are indeed measured in the p h o t o i o n i z a t i o n experiments. There is good agreement among the three experiments for c o m p a r a b l e peak energies, and the intensity profiles are also similar (see fig. 1 ). Since we have already discussed these general features in section 3.2, we can i m m e d i a t e l y assign the m a j o r part o f this structure to 244 ly~k ~ ' ...l~gT~u~g 234 •..l~g~uT~g, nS(~g, l"lu.

On energetic grounds, the products are still restricted to C1 + (3P) + C I - ( ~S ), and consequently the asymptotic form o f the potential energy curve m u s t be 31-Iu. In section 3.2, we h a d assigned this b a n d to ~Hu in the ClJ- p h o t o i o n yield curve. It is, o f course, possible that the excited state in section 3.2 is 31-Iu, acquiring some singlet character by s p i n - o r b i t interaction. However, the intensity o f this b a n d in the Cl~- curve is about two orders o f magnitude stronger than in the corresponding ion-pair band. Hence, a m o r e attrac-

326

.L Berkowitzet al. /Photoion-pairformation in (72

tive interpretation is that photoabsorption occurs to the ~FIustate, which primarily undergoes autoionization, but partially is predissociated by a 3Hu curve, resulting in the observed ion-pair products. The additional peaks in the photoionization studies not observed in the EEL experiment indicate that more than one excited state is involved. 3.5. The ion-pair bands between 880 and 935 A This spectral region, shown in greater detail in fig. 7, corresponds to the domain of the L and M bands in the nomenclature of Stubbs et al. [ 11 ]. A perusal of table 1 reveals that more features can be identified in the ion-pair spectrum than in either the C1+ photoion yield curve or the EEL experiment. The latter two are in rather good agreement between 13.170 and 13.674 eV, then get somewhat out of step. The ionpair curve also has features that match well with both C1J- and EEL, until about 13.85 eV, beyond which there is some discrepancy. Above 14.10 eV (where no EEL data are presented) there is once again good agreement in the energies of the peaks in the CI~- and ion-pair curves. Apart from the supernumerary features, the L and M bands have been previously assigned (section 3.2 ) to successive numbers of the ns~vg, ll-[ u Rydberg series culminating on A 2rlu,i. The widths of the fine structure features of the ion-pair bands in this spectral region are distinctly broader than those described in the preceding section, between 960 and 990 ~. Also, there is a rather abrupt onset in the ion pair curve, essentially at the thermochemical threshold for formation ~F

i I

~80

890

i

I

_

900

~

L

910 920 WAVELENGTH

I

I

930 (,~)

/

L

~990- 900

Fig. 7. The ion-pair bands between 880 and 935 A.,obtained with a resolution of 0.28 A (fwhm). A portion of the spectrum, as shown at the right-hand side of the figure, has been obtained with 0.14/k resolution.

of C1+ ( I D 2 ) + C1- (tS o ). These two observations may be correlated. Thus, ~D2+ ~So correlate with the molecular states ~Z, JrI and IA. We have assigned the major features in this region to tIIu, which can correlate with the ID 2 + IS 0 asymptote without invoking spin-orbit interaction. The fact that there are discrete features implies that the optical transition is not directly to a repulsive state, but rather to a predissociating state or to the inner well of a double-well state. Since the peaks in the CI• curve are essentially at the same energies as in the ion-pair curve, and the CI~ curve is more intense, the widths o f the peaks may be determined by the rate o f autoionization, rather than the rate of predissociation. There is a rather dramatic reversal of intensities o f the " L " and " M " bands in the ion-pair spectrum, compared to either the CI~ or EEL curves. In the latter two, the peaks in the L band are superimposed upon a broad band, whereas the M band lies in a valley. In the ion pair curve, the M band is at least twice as intense as the L band. Clearly, the efficiency of the ion-pair process for the M band is greater than that of the L band. The L band occurs near the threshold f o r C1 + ( I D 2 ) production. If there is a small barrier in the relevant potential energy curve, it might explain the lower efficiency. In the same vein, the J ionpair band is significantly more intense than the 10401060 ~ band near the CI+(3P) threshold. Both of these bands have the same 3p limits, but the J band lies significantly above threshold, and should be less susceptible to barrier effects. 3.6. The ion-pair band between 760 and 815 A This spectral region in the ion-pair curve has some weak fine structure superposed upon a broad band. Some weak structure can also be detected in the C l f photoion yield curve, but the positions o f the peaks in the two experiments are not correlated well. The average spacing of the fine structure peaks on the C1+ curve is 0.032 eV; in the ion-pair curve, it is 0.030 eV. Both are in reasonable agreement with the average vibrational spacing (0.034 eV) of the second excited state in the photoelectron spectrum [2], 2Zg. In addition, the overall shape of the ion-pair band looks very much like that of the photoelectron band (see fig. 8). Hence, we assign the ion-pair band to a Rydberg state converging on 2E+, i.e.,

J. Berkowitz et al. / Photoion-pair formation in CI2

N2+ (2E~)

[ CI2+(2~:g+)

~qTFFT~ ION-PAIR

L;--\ 16.5

"

.....

16.0 IONIZATION ENERGY (eV)

-2.5

\ _1 15.5

Fig. 8. The He I photoelectron spectrum of CI 2 in the region of the 2Zg' band (from van Lonkhuyzen and de Lange [2] ) and the ion-pair band observed in the present experiment between 760 and 815 ~. The dashed curve is a simulated photoelectron spectrum, discussed in ref. [2].

244,x; ...O'g~uI~g,

44 ,+ ~ •..(~g~u ~g I~u, ~u

•

Our choice of (;~ for the excited level, rather than nu, is not based on quantum defect analysis, since it is difficult to uniquely correlate a particular feature of the Rydberg state with a corresponding feature of the ionic state. Instead, we base our analysis on a correlation with the likely products. This ion-pair band appears rather abruptly at the thermochemical threshold for C I + ( ' S o ) + C 1 - ('So). This combination correlates with 1X, rather than 11-I.The relative intensity of this ion-pair band suggests that it is an allowed process, suffering neither from selection rules nor from barriers.

4. Summary The photoionization of C12 has been studied in three ion channels, CI~, Cl + and C1-. The CI+ photoion yield curve is similar to a pseudo-photoabsorption curve, as measured by electron energy loss spectroscopy, but the present experiments are conducted with higher resolution, displaying more resonance features at their natural line widths. The combination of C1+ and CI- measurements enables us to apportion and distinguish ion-pair processes and

327

photodissociative ionization. The intensity of the ionpair process is two orders of magnitude weaker than total ionization. At least five super excited states are observed to undergo ion-pair formation. These states have been assigned as Rydberg states whose cores are X 2I"Ig,A 2I-[ u and 2Z~-. The ion-pair products of the first two states are confined by energy constraints to C1+ (3p) +C1- ('So); the next two are inferred to be CI + ( 'D2 ) + C1- ( ' So ); and the last one inferred to be CI+ (ISo)-t-C1- ('So). The first ion-pair band displays a remarkable temperature effect. At 300 K, it extends to 0.16 eV below the thermochemical threshold, but at 160 K it extends just slightly below this threshold. The mechanism of incorporating such a magnitude of internal thermal energy is not understood, but it is believed to involve rotational, rather than vibrational energy.

Acknowledgement We wish to thank Dr. John Comer for providing us with a publication-quality copy of his electron energy loss spectrum. We also wish to acknowledge the assistance of Mr. Bruce Gluckman and Mr. Brian Moudry in the analysis of the data. This research was supported by the US Department of Energy, Office of Basic Energy Sciences, under Contract No. W-31109-Eng-38, and by the US-Yugoslav Joint Board for Science and Technology through the US Department of Energy Grant No. PN 561.

References [ 1 ] J.A.R. Samson and G.C. Angel, J. Chem. Phys. 86 ( 1987 ) 1814. [2] H. van Lonkhuyzen and C.A. de Lange, Chem. Phys. 89 (1984) 313. [3] T.A, Carlson, M.O. Krause, F.A. Grimm and T.A. Whitley, J. Chem. Phys. 78 (1983) 638. [4 ] A.E. Douglas, Can. J. Phys. 59 ( 1981 ) 835. [ 5 ] Th. Moeller, B. Jordan, P. Gtirtler, G. Zimmerer, D. Haaks, J. le Calve and M.-C. Castex, Chem. Phys. 76 (1983) 295. [ 6 ] L.C. Lee, M. Suto and K.Y. Tang, J. Chem. Phys. 84 ( 1986 ) 5277. [7] T. Shinzawa, A. Tokunaga, T. Ishiwata and I. Tanaka, J. Chem. Phys. 83 (1985) 5407. [ 8 ] T. Ishiwata, T. Shinzawa, T. Kusayanagi and I. Tanaka, J. Chem. Phys. 82 (1985) 1788.

328

J. B e r k o w i t z et aL / P h o t o i o n - p a i r f o r m a t i o n in Cl 2

[ 9 ] T. Ishiwata, I. Fujiwara, T. Shinzawa and I. Tanaka, J. Chem. Phys. 79 (1983) 4779. [10] D. Spence, R.H. Huebner, H. Tanaka, M.A. Dillon and R.-G. Wang, J. Chem. Phys. 80 (1984) 2989. [ 11 ] R.J. Stubbs, T.A. York and J. Comer, J. Phys. B 18 ( 1985 ) 3229. [ 12 ] S.D. Peyerimhoffand R.J. Buenker, Chem. Phys. 57 ( 1981 ) 279. [13] V.H. Dibeler, J.A. Walker and K.E. McCulloh, J. Chem. Phys. 53 (1970) 4715. [14]V.H. Dibeler, J.A. Walker, K.E. McCulloh and H.M. Rosenstock, Intern. J. Mass Spectrom. Ion Phys. 7 ( 1971 ) 209. [ 15 ] S.T. Gibson, J.P. Greene and J. Berkowitz, J. Chem. Phys. 83 (1985) 4319. [ 16 ] K.P. Huber and G. Herzberg, Molecular spectra and molecular structure, Vol. 4. Constants of diatomic molecules (Van Nostrand-Reinhold, New York, 1979).

[ 17] R.P. Tuckett and S.D. Peyerimhoff, Chem. Phys. 83 (1984) 203. [ 18 ] D.A. Shaw, G.C. King and F.H. Read, J. Phys. B 13 ( 1980 ) L723. [19] W.H.E. Schwarz, Chem. Phys. 11 (1975) 217. [20] R.E. Huffman, J.C. Larrabee and Y. Tanaka, J. Chem. Phys. 47 (1967) 856;48 (1968) 3835. [21 ] R. Trainham, G.D. Fletcher and D.J. Larson, J. Phys. B, to be published. [22] R. de L. Kronig, Z. Physik 50 (1928) 362. [ 23 ] G. Herzberg, Molecular spectra and molecular structure, Vol. 1. Spectra of diatomic molecules (Van Nostrand, Princeton, 1950). [24] M.A.A. Clyne, M.C. Heaven and J. Tellinghuisen, J. Chem. Phys. 76 (1982) 5341. [25] M.C. Heaven, Chem, Soc. Rev. 15 (1986) 405. [26] F. Grein, S.D. Peyerimhoffand R.J. Buenker, Can. J. Phys. 62 (198) 1928.