Mechanisms of electron-stimulated desorption of protons from water: Gas, chemisorbed and ice phases

119

Surface Science 157 (1985) 119-150 North-Hoifand, Amsterdam

MECHANISMS OF ELECTRON-STIMULATED DESORF’TION OF PROTONS FROM WATER: GAS, CHEMISORBED AND ICE PHASES *

Received

YOOctober

1984; accepted

for publication

7 February

1985

The stimulated desorption of ions from gas phase and condensed phase HsQ on Ni(lll) has been examined theoretically and experimentally for the near threshold excitation regian, 15 to 40 eV. The excited state potential energy curves have been calculated using configuration interaction for Ha0 and a restricted Hartree-Fock (RWF) approach for a variety of small clusters including {ii.@), and NiH,O. Both proton yield and kinetic energy distributions have been measured for ehemisorbed, ice phase, and gas phase water and are discussed in terms of specific electronic excitations corre~nd~ng to possible desarption pathways. For condensed phase water, the major proton desorption threshold occurs at 20-21 eV and is due to surface predissociation. The finai state potential energy curves reached in this process are, in general, described by two electron excitations from the ground state and are thus not dipole allowed. At threshold, these potentiat energy curves correspond to the excited states of the neutral rather than the ionized molecule. Above 28-29 eV, predissociation or shake-up involving excitations from the 0 2s orbital contributes to the ion yield and can give rise to protons of high (7-8 eV) kinetic energy,

1. lntmduction

Eiectron- and Fhoton-stimulated desorption (ESD and FSD) of adsorbates from metal surfaces have engendered the interest of a growing community of workers in recent years fl]_ The stimulated desorption of hydrogen in particular has drawn considerabhe attention. This concern has been fueled both by the technical significance of the hydrogen/metal interaction and by the promise that hydrogen would represent a comparatively simple system which might afford a more thorough understanding of the desorption mechanism. Promise has been shown in both these areas. Stimulated desorption has proved to be an especially sensitive probe for detecting the presence of hydrogen on a surface [2]. While it has not yet been definitively demonstrated, ESD and PSD appear * This work was supported by ibe US Department of Energy. ** Present address: Miles taberatory, E&art, indiana 46514, USA. **+ Comments and requests should be addressed to RHS.

~39~~~28/g5/~~3.3~ 0 Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

J. 0. Noelf

et al. / ESD of protons from water

M+N

X \

Interatomic

Separation

Fig. 1. Schematic drawing illustrating the various near threshold stimulated desorption mechanisms for a metal surface denoted by M and an adsorbed atom denoted by N. The superscripts of M or N refer to possible charge states at infinite separation (asterisk denotes excited state configuration). The potential energy curves are described as follows: X is the ground state curve, A is a weakly repulsive excited state described by a single electron excitation (solid vertical line), 6 is a bound excited state described by a single electron excitation, and C is a strongly repulsive excited state. The dashed vertical excitation X + C represents a direct shake-up process which can occur if C contains some single electron excitation character. Surface predissociation is depicted by the path X + B = C involving a curve crossing from B to C and will occur when the excited state C is described only by two electron excitations.

to be site-specific tools which would enable one to obtain detailed information regarding the bonding preferences of the hydrogen. In general, the mechanisms associated with near threshold ESD and PSD which have been discussed in the literature are of one of three types as shown schematically in fig. f. Here we have drawn the potential energy curves for a two component system consisting of speciese M and N. Curve X is the ground state, curve A is a repulsive state reached through a one-electron excitation, curve B is a bound state reached again by a one-electron excitation, and curve C is a strongly repulsive state reached through a two-electron excitation process. The transition X + A is characteristic of the MGR model as it was originally proposed by Menzel and Gomer [3] and independently by Redhead [4]. Transition X -+ C describes a shakeup process leading to a highly repulsive potential energy curve which diabatically leads to ions. For this transition probability to be large, there must be a significant overlap between this state

J. 0. Noell et al. / E..W of protonsfromwater

121

and another state which is dipole allowed from the ground state. If this overlap is very small then transitions to this repulsive curve may instead be dominated by a predissociation mechanism involving first the transition to a singly excited state (B) followed by a curve crossing event to the doubly excited repulsive state (C) leading to the ionic asymptote (denoted X --, B =j C). We have shown previously [5,6] that this predissociation mechanism should be important for proton desorption from nickel. In the present paper, we extend our prior work to hydrogen desorption from molecularly adsorbed water in an effort to determine both the threshold energy for H’ desorption and the mechanism associated with it. Conceptually, proton desorption from chemisorbed water on a metal surface presents us with a very different situation than for hydrogen on nickel. The hydrogen atom is no longer in immediate contact with the metal surface but rather is bound to the oxygen atom and only feels the perturbation of the metal in a secondary way. An immediate question posed is whether this different environment for the hydrogen makes any substantive difference in the desorption process. It is not evident whether the predissociation mechanism which we presented for hydrogen on a metal surface might also explain desorption from chemisorbed water. A second question which we will address is whether there is a similarity between the process of stimulated desorption of water or ice layers on a metal substrate and the processes of gas phase photodissociation of water. These latter processes have been quite thoroughly studied [7] and we shall apply the results of these studies, insofar as possible, to the present problem. As we did in our earlier work on the hydrogen-metal system [5,6], we have coupled stimulated desorption experiments with ab initio, quantum chemical calculations. The focus of the ESD experiments on H,O was the determination of the relative proton yield and kinetic energy as a function of the incident electron energy. These studies were carried out on a nickel (111) surface for a range of water exposures spanning the sub-monolayer and multilayer regimes. In addition, a more limited set of experiments was performed for thick ice layers in which the temperature of the samples was ramped through the thermal desorption region. The intent was to measure the stimulated desorption of gas phase water using the same technique employed in the adsorbed water studies. This affords a direct comparison between the gas and condensed phase studies and provides a frame of reference for comparing these results with the more detailed gas phase results available in the literature. The theoretical approach was designed to determine the electronic states involved in the stimulated desorption process. The calculations focused primarily on the isolated water monomer. Potential energy curves were calculated for the near-threshold region (35 eV) as a function of the dissociative reactioncoordinate for both the ionized and neutral molecule. Calculation of these potential energy curves as a function of the charge state of the molecule is

important since it is not a priori evident whether ionization competes with or actually leads to desorption. In general, this concept has not yet been well investigated for near threshold stimulated desorption phenomena. By concentrating on the monomer, it was possible to include the necessary electron correlation effects to provide a semiquantitative picture of the electronic states important in the desorption. The importance of the perturbing environment on the bond breaking process was qualitatively evaluated by employing Restricted Hartree-Fock (RHF) calculations for (H,O), and NiH,O. The first of these molecules represents the influence of a hydrogenbonded ice structure, while the second approximates the effects one might expect upon adsorption of water to a metal surface. Finally we would like to point out that many of the ideas and concepts which we discuss here for water - such as the adiabatic versus diabatic nature of the desorption process and the role of ionization in desorption - are important to all cases of near threshold stimulated desorption phenomena.

2. Experimental details The experiments were performed in an ultra-high vacuum chamber with a base pressure of 1 x lo- lo Torr. Both the submonolayer and multilayer measurements were done by dosing a 1 cm diameter nickel single crystal whose surface normal was oriented along the (111) direction, using timed exposures to water vapor emerging from a glass microchannel capillary array. The sample temperature was maintained with a continuous transfer line using either liquid nitrogen or liquid helium. During gas dosing, an auxiliary cryopump was kept fully open to further minimize residual gas contamination. With the sample approximately 2 cm from the doser, the system pressure during this procedure generally rose to the low 10m9 Torr range. The sample temperature was typically in the range 100-110 K during both dosing and ESD yield measurements. Surface coverage (S) was determined using thermal desorption spectrometry (TDS). After each series of ESD measurements the sample was flashed to determine the quantity of water adsorbed on the surface. The area of the mass 18 peak, measured as a function of temperature, was compared to a standard value obtained from a separate set of thermal desorption experiments. By exploiting the fact that above one monolayer coverage, a second peak appears in the TDS spectrum due to the formation of ice multilayers, we are able to equate the area associated with the mass 18 TDS spectrum just prior to the appearance of this second multilayer peak, to the standard for exactly one monolayer. All subsequent coverages were determined using this standard. Initial preparation of the specimen consisted of argon ion sputtering followed by exposure to oxygen. The sample was subsequently annealed at high

J. 0. No& ef al. / CliANNEL

ESD

ofprotonsfrom

123

water CHOPPED

PLATES

ELECTRON BEAM

\

50

SAMPLE

FEEDTHROUGH ELECTRON

BEAM

TRIQOER PULSE

AMPLtFlER OISCRIMINATOR

-

MULTICHANNEL

STOP

TAC

ANALYZER

START d

Fig. 2. Schematic

of the time-of-flight

ion detection

apparatus.

temperature (5 SOO’C) to establish surface order and remove residual carbon contamination through CO or CO, formation with the adsorbed oxygen. Cleanliness and order were checked using low energy electron diffraction (LEED) and Auger spectroscopy. The el~tron-stimulated desorption measurements utilized a time-of-flight method similar to that described elsewhere [8]. A schematic of the experimental apparatus is shown in fig. 2. The incident electron beam was chopped by applying a fast 15 V pulse to the emission cup of a standard electron gun which was normally biased in the off condition. Typical electron beam pulse widths were on the order of 50 ns with a repetition rate of = 100 kHz. With these operating parameters one has approximately 10 ps to collect all of the desorbing ions. This is quite adequate for the low mass species; the higher mass ionic species such as H,O+ require a small accelerating voltage in order to bring them in within the 10 ys window. If this is not done, they are still detected but appear as “wrap-around” peaks. The incident electron beam current density was typically kept below appro~mately 2 x 1O-6 A/cm’. At higher current densities depletion effects become apparent particularfy at energies well above threshold. The ion detector shown in fig. 2 is a drift tube which is attached to a channel electron multiplier array (CEMA). The CEMA assembly consists of two identical 25 mm diameter channel plates oriented in the well known chevron configuration. The drift tube has at each end an 83% transmitting screen and can be biased to accelerate the incoming ions. The total ion flight path for the detector used in these studies was 7.6 cm.

J. 0. NoeN et al. / ESD

124

ofprotons fromwater

6000

4000

2000

0

10

15

20

25

Electron

30 Beam

35 Energy

40

45

50

(eV)

Fig. 3. Proton yield as a function of incident electron beam energy for three coverages of H,O on Ni(ll1): 0.7, 1.2, and 16 equivalent monolayers. The inset is an enlargement of the threshold region for B = 16. Each yield curve is normalized to the incident electron beam current.

The output pulses of the detector are fed first to a preamplifier and discriminator and then finally to the start input of a time-to-amplitude converter (TAC). The stop input of the TAC comes from the initial electron gun trigger pulse after having gone through a suitable delay. The amplitude of the output TAC pulse is thus proportional to the complement of the ion flight time: the ions which arrive first correspond to the pulses of largest amplitude from the TAC output, The time-of-flight spectrum was then acquired on a Tracer Northern 1710 multichannel analyzer fed by the TAG output. In this configuration, determination of the kinetic energy for an ion of a given mass requires only a knowledge of the flight path distance [9].

3. Experimental results The proton desorption yield as a function of the incident electron energy is plotted in fig. 3 for water coverages of 0.7, 1.2 and 16 equivalent monolayers on the nickel (111) surface, The most heavily dosed sample - which represents a condensed water/ice phase - shows an onset at 20-21 eV in energy which

J.O. Noel1 et al. / ESD of protons fromwater

4

6

Proton

Kinetic

8 Energy

10

125

12

14

(eV)

Fig. 4. The H+ kinetic energy distribution curves for f? = 16 monolayers at four separate electron beam excitation energies. The curves are not normalized to each other but rather are plotted so that their essential features can be ~stinguish~.

can be more clearly seen in the inset to fig. 3. This is in agreement with early PSD results of Rosenberg et al. [lo] for proton desorption from condensed water at = 15 K and with the more recent high-resolution PSD results of Stulen and Rosenberg [ll]. The yield curves at lower coverage are similar, though with decreasing coverage the proton yield becomes smaller. It is important to note that the threshold is well above the thermodynamic minimum for the dissociative ionization process for molecular water defined as the sum of the dissociation energy and the ionization potential of hydrogen: H,O-+OH+Ht+e-;

E,,,=E,,,+E,,=18_76eV.

We were able to obtain a measurable Hi+ count rate below 20 eV for high coverage (8 = 16) though this was several orders of magnitude below the count rate at 22 eV. This is significantly different from gas phase studies where a prominent proton threshold is observed at Et,,,. followed by a drop in the yield at higher energies [7a,7b]. The kinetic energy distribution of the protons for several incident electron beam energies is plotted in fig. 4 for the coverage of sixteen monolayers. There are several notable features. First, even at 25 eV the mean proton kinetic

2.0. Nod

126

et ai. / ESD of protons from water

45 m m m m

eV

incident

be am energy

m

0

2

4

6

Proton

Kinetic

8 Energy

10

12

14

feV)

Fig. 5. The H+ kinetic energy distribution curves for B = 16, 2 and 0.3 monolayers at an incident electron beam energy of 45 eV. The curves are not normalized to each other but rather are plotted so that their essential features can be distinguished.

energy is significant (3.5-4.0 eV). As one increases the incident electron energy, a bimodal character in the kinetic energy distribution becomes apparent. The higher kinetic energy fragments appear to result from a mechanism whose threshold is near 30 eV. The relative peak height associated with this second high energy portion is a strong function of the ion collection efficiency. The data in fig. 4 were taken with a low accelerating voltage (5 V) on the drift tube. At larger acceleration voltages (20 V or more) where more ions are swept into the detector, the high kinetic energy peak appears only as a shoulder on the main 4-5 eV peak. Below 21 eV excitation where the yield is extremely low, the H” kinetic energy is expectably peaked at less than 1 eV. It should be noted again that the cross-section below 20 eV is insignificant compared with that above 22 eV. In fig. 5, the kinetic energy distributions at an incident electron energy of 45 eV are plotted for several coverages ranging from submonolayer to the thick ice layer (8 = 16) discussed above. The salient feature of these data is that at lower coverages the higher kinetic energy component, which was so evident in the high coverage case, becomes much less apparent. The implication, of course, is that whatever m~h~isrn is responsible for the higher kinetic energy

peak is considerably less facile when the water molecules are in more intimate contact with metal. There is no apparent change in the mechanism giving rise to the lower kinetic energy peak as the coverage is lowered, neither its threshold nor measured kinetic energy is significantly altered. The kinetic energy data presented above are in marked constrast with published, gas phase results [7] which show that the protons are kinetically cold (= 1 eV) regardless of the energy of the incident radiation. In view of the marked differences between the condensed phase results presented here and the gas phase results reported in the literature, we performed gas phase ESD experiments using the same time-of-flight apparatus used in the condensed phase studies. This was accomplished by monitoring the proton desorption yield as the sample was rapidly heated through the sub~mation point. In this way, the ESD of protons from water could be measured just as the molecules left the metal surface. A series of such measurements were carried out at discrete incident electron beam energies ranging from 19 to 35 eV. The results of these measurements are shown in figs. 6 and 7. The first of these figures shows a sequence of time-of-flight curves, each taken for 1 s, with an incident electron beam energy of 20 eV, at different sample temperatures as the sample was heated through the sublimation temperature. For the heating rates employed, a 1 s integration time corresponded to less than a three degree temperature span. As the water desorbs, one observes the appearance of both low mass (AMU = 1) and high mass ionic species. The flight time and ion accelerating potential for these experiments were not sufficient to unequivocably identify the higher mass species, however the spectra are consistent with the appearance of one or more species with AMU ranging from 17 to 19. These higher mass fragments could be OH’, H,O’ or H(H,O)+, all of which are consistent with an earlier ESD study of condensed H,O near the sublimation point [12]. Fig. 7 shows the proton kinetic energy distribution obtained from the data in fig. 6. These gas phase results show several important features. First, significant desorption of protons is observed at lower incident energies than in the case for the solid. The first significant H+ signal was observed at 19 eV. This is consistent with gas phase results where protons are observed at the thermodynamic threshold for dissociative ionization (18.76 eV) [7]. The second important result is that at all energies, the peak of the proton kinetic energy signal remained at less than 1 eV. This is also consistent with earlier, gas phase work [7a] but is in marked contrast to the condensed phase results. Based on these results, the key points which the theory must address are as follows. For a submonolayer of water adsorbed on a Ni(ll1) surface or for a thick ice layer deposited on the surface, the dominant proton threshold occurs at = 21 eV, well above the thermodynamic threshold for either ion-pair formation (16.93 eV) or dissociative ionization (18.76 ev). At this threshold, the protons leave with significant kinetic energy (distribution peaked at 3 eV,

J.O. NoeN et al. / ESD of protons from water II

“F

a

6

4

110K

2

C

75

5C

160K

25

0

150

100

H+ 50

1

2

TIME

3

4

5

6

(microseconds)

Fig. 6. A sequence of gas phase time-of-flight ESD spectra taken during rapid heating of a thick ice layer through the sublimation point. In each case the spectra were acquired over a time period of 1 s. The approximate temperature at the start of each acquisition period is denoted above each curve. The incident electron beam energy was maintained at 20 eV.

J.O. Noel1 ei al. / ESD of protons from water

.

25

.

*’

.

.

.

129

I

.

+ 20

+ + + Cc+++*

++.+/Q + pa.+ + A;+ + r

15 +

+

l

2 + L+ +.*

10

.

H-y

+

5 -r4

+.. .-

+ 4+ . + ++ . +

+

4.

Y

+

:.+ +

+

+

*

+

I

0 0

+

++A.+

++

*

+ +

.**

.

+

+

+A.++ + 4-l. +++ + + A&‘d +SW +++ + +-+ ++ .

.

.

Proton

I

.

.

.

2

1 Kinetic

Energy

.

3 (eV1

Fig. 7. The kinetic energy distribution for H+ obtained from the gas phase measurements fig. 5. The triangles represent a least square smoothing of the data.

shown in

at 22 eV excitation energy). For the thick hydrogen-bonded ice layer, additional features become apparent. A mechanism yielding high kinetic energy protons (distribution peaked at = 7 eV) is observed. The onset for this mechanism appears to occur at an incident electron energy between 30 and 35 eV. For gas phase water, a completely different behavior is observed. Desorption of protons occurs above the 18.7 eV thermodynamic threshold and the kinetic energy of the protons remains at less than 1 eV for all excitation energies. We shall return to these issues after presenting the results of the calculations.

4. Calculational details The intent of the calculations on water was to provide a quantitative adiabatic description of the excited state potential energy curves of the water molecule and the molecular cation which lie roughly within a 35 eV region above the ground state. Before discussing these results, it is important to clearly differentiate between adiabatic and diabatic states since it will be useful to switch from one representation to the other. An adiabatic curve is defined as the path which the system will follow if, at all times, the separating fragments

are moving infinitely slowly. In this representation, the wavefunction for each state is optimized at each internuclear separation. Consequently, the character of a given root may change along the reaction coordinate. Within this framework, no two curves of the same electronic symmetry will cross and transitions between states are forbidden. The correlated molecular orbital calculations which we will present were performed in this representation. In contrast, for a diabatic representation, the character of the wavefunction does not change. As a simple example, a state of ionic character at large internuclear separation may be followed continuously to a state of the same ionic character at small separation. It is not an easy matter to say whether a given system will behave adiabatically or diabaticalIy 1131. In general, however, the faster the particle is moving, the more likely it will be to behave diabatic~ly. In the case we will consider, the departing particle, a proton, is very light. Since some of the potential energy curves are very repulsive, the proton could be moving extremely rapidly at small internuclear separation, favoring diabatic character. During the discussion we will point out where some of the diabatic curves lie and what the probability is for curve crossing. To interpret the desorption results, it was necessary to provide a consistent description of the adiabatic states as one of the oxygen-hydrogen bonds is stretched from its equilibrium value to an infinite separation. The states of particular interest are those involving excitations from the bonding O-H orbitals (3a, and 1 b, in ground state C,, H,O). It is these excited states which one would expect to be most repulsive as a function of the internuclear separation and hence should be responsible for any direct dissociation process resulting in high kinetic energy protons. The calculation scheme was based upon using an open shell, RHF calculation (141 to define the reference orbitals. From these orbitals, a POLI-CI [15] was performed including all configurations expected to be dominant for any low lying root either at the equilibrium geometry of water or at the asymptotic dissociation limit. The basis set used was the (9s, 5p) primitive set of Huzinaga [16] contracted to [3s,2p] per Dunning [17] for the oxygen atom and the double zeta, segmented contraction [18] of the (4s) primitive set of Huzinaga for the hydrogen atom. Additionally, a single gaussian (S = 0.032 a.u.) [IS] was added to describe the 3s Rydberg orbital of water. The reference orbitals for the POLl-CI calculations were defined by an RHF calculation for the lowest ‘A’ state [19] of the molecule, in C, notation: la’* 2a’2 3a” 4a” 5a” 6a” la”‘. There are several key features which this wavefunction possess which are essential for a proper description of desorption. First, the orbitals correlate smoothly from the molecule to the low lying asymptotic, dissociated limits. Second, this state defines each of the orbitals which are occupied in the states one would suspect as being critical in the desorption process. The 5a’ and the

J.0. Noel1et al. / ESD of protons from water

131

6a’ orbitals respectively represent the 3s Rydberg orbital on the water and the lowest lying oxygen-hydrogen, antibonding orbital (this orbital will become localized onto the bond of interest as the oxygen-hydrogen separation is increased beyond its equilibrium value). These are the lowest lying of the virtual orbitals in the ground state of water and hence one would expect most of the low lying excited states to involve excitations into these orbitals. By defining the orbitals at the RHF level, a treatment of the excited states of interest which is consistent with that of the ground state is afforded at the POLl-CI level [20]. The same orbitals were used as a reference for states of both the neutral water molecule and the molecular cation. The reference list included all configurations satisyfing the following three constraints: (a) they involved only the seven orbitals defined by the RHF calculations; (b) they involved excitations with respect to the ground state from only the 2a’, 4a’ and la” orbitals; (c) they made a dominant contribution to one of the electronic states lying within 35 eV of the ground state. The configurations included in the CI were generated by allowing all single substitutions from each of the members of the reference list with the constraint that the la’ orbital (the Is orbital on the oxygen) remain doubly occupied. This procedure defined a problem of a size for which we could solve for the many roots required. As an example, for the ‘A’ symmetry of neutral water we had 11 reference configurations from which 223 spatial configuration and 297 spin eigenfunctions were generated. The procedure outlined above facilitated solution of the resultant CI matrix within C, symmetry and consequently the excited state curves could be followed smoothly as the oxygen hydrogen bond was stretched. The method clearly was neither as cleanly defined nor as unrestricted as might be desired and hence a check was necessary to insure that our intuition in devising the procedure did not introduce an unphysical bias to the results. This was done by comparing results for the vertical excitation energies with those from a more reliable calculation using the full C,, symmetry of the equilibrium molecule. These more general configuration interaction calculations were performed in the same spirit and framework as those discussed above with several notable enhancements. First, the reference orbitals were defined by an RHF calculation for the lowest ‘A’ state: la:

2a: lb: 3a\ lb: 4a’, 2b\ Sat,.

Second, polarization functions were added to the basis set. These included a set of d functions (S = 0.85 a.u.) centered on the oxygen and a set of p functions (Z = 1.0 a.u.) on each of the two hydrogen atoms. The CI technique was also made less restrictive. The dominant configuration of all low lying excited states was included in the reference list. Also, both single and double substitutions with respect to these reference configurations were allowed with the restriction that at most one electron could be outside the space defined by

132 Table 1 Comparison

J. 0. Noel1 et al. / ESD of protons from water

of calculated

and measured

vertical

excitation

energies (in eV) in H,O

Excitation

POLI-cr

POL-CI

lb, + 4a, 3a, -) 4a, lb, --L 4a,

7.62 10.02 15.47

7.34 9.69 13.89

7.44 c, 9.85 c’ 13.45 c,

Ib,-+oo 3a,+oc Ib,-*cc 2a, --*co

11.98 14.04 19.82 33.27

11.72 14.14 18.35 32.29

12.6 14.7 18.5 32.2

lb:-,4a: 3a,lb, + 4a:

20.77 22.25

20.45 21.87

lb: + 4a,co lb,3a, + 4a,zo lb,3a, + 4a,w lbalb, + 4a,oo

26.64 25.94 32.45 31.66

26.36 26.02 32.01 30.42

” The ground state configuration is la{ 2a{ lb: mixture of O-H antibonding and 0 3s Kydberg b, This represents the higher level CL ‘) Ref. [38a]. d, Ref. [38b].

lY)

3a< lb:; character.

a)

Experiment

the lowest virtual

d’ d, d, d,

orbital

(4a,)

is a

the seven orbitals which were occupied in the RHF calculation. This is very much in the spirit of the POL-CI as it was initially defined [21}. This level of calculation has been well tested and has been proven reliable for a wide range of problems. In table 1, vertical excitation energies predicted for selected excitations by the two levels of CI (POLl-CI and POL-CI) are compared with each other and where possible with experimental data. The more accurate POL-CI calculations reliably predict the various excitation energies. The mean error in the excitation energies, with respect to experiment, is only 2% with the largest error being 7% (lb, ionization). In addition, it is clear that in most cases the POLl-CI also does an excellent job of representing the excited states. For the fifteen excitations presented in table 1, the average difference between excitation energies predicted by the two sets of CI calculations is only 3.9%. In fact, the agreement is generally much better than that; excepting excitations involving the lb, orbital (these excitations were not included in the reference list for the smaller CI) the average error is only 2.4%. 5. Theoretical results and discussion The reaction coordinate for hydrogen desorption to be a simple stretching of a single oxygen-hydrogen

from water was assumed bond. As the hydrogen

moves along this coordinate, the molecule maintains only C, symmetry and hence, as has already been discussed, it was in this symmetry that potential curves were calculated for both the neutral and ionized molecules. However, throughout this paper C,, symmetry notation is used to describe the vertical excitations from the ground state. This convention was used since the initial excitation of the molecule takes place in its equilibrium geometry where C,, is appropriate.

40

-

24

-

1

2

3

4

5

R

6

7

6

9

10

(bohrI

Fig. 8. Potential energy curves of ‘A’ symmetry for ihe neutral water monomer. The dashed lines denote those curves characterized by a single ekctrcm excitation at eq&ibrium separation. Pius signs indicate Ii l asymptotes.

J.U. Noeil et

134

\

0

al. / ESD of protonsfrom

water

H

r------e

R

$i

32

-

28

-

24-

t 15

2016 _ 12 -

\

/’

‘_--

a-

‘.

.

‘-__ 4-

--a--------s--v---

01 1

2

3

4

5

R

8

7

8

9

10

(bohr)

Fig. 9. Potential energy curves of ‘A” symmetry for the neutral water monomer. denote those curves characterized by a single electron excitation at equilibrium signs indicate H + asymptotes.

The dashed separation.

lines Plus

Curves of ‘A’, ‘A’, ‘A’ and 2A” symmetry (in C,) were evaluated [22] resulting in the four sets of curves displayed in figs. 8-11. An initial look might suggest that the implications of mechanisms derived from these curves would be hopelessly lost in their complexity. In an effort to disentangle the information, the discussion will be partitioned more or less in the following manner. First, we will discuss separately the potential curves for neutral (figs. 8 and 9) and ionized (figs. 10 and 11) water. We will then compare these results to those

135

J.O. Noel1 et al. / ESD of protons from water

48

H \

44

0 -----_--

,,

40

36

32

+ v--m_

:

--------__-__-+

28

24

-+

20

\

\ \

16

__---_--_____--___--$ ______--__----_/--

__--

\ _A’

+

YMO

12

8

4 0 12

3

4

5 R

6

7

8

910

(bohr)

Fig. 10. Potential energy curves of ‘A’ symmetry for a singly ionized water molecule. The dashed lines denote those curves characterized by a single electron excitation at equilibrium separation. Plus signs indicate H+ asymptotes.

obtained for NiH,O and (H,O), and then will finally discuss the implications of the calculations and experiments regarding the details of likely desorption pathways. 5.1. Neutral H,O In close analogy to our earlier results for the nickel/hydrogen system [5,6], the excited states for neutral water can be banded into groups of different

f. 0. NoeNet ai. / ESD of protons from water

136

2gs

48

H 44 r

\

0 ~--.W---.. H R

k

4036-

f@_-

322824-

.-\i-

_____-_-__-___

/_--

16t t

\

/’

,’

\ La’

12t i 81 1 4 i

-I

1

2

3

4

5

R

6

7

a

3

10

(bohr)

Fig. 11. Potential energy curves of *A” symmetry for a singly ionized water molecule. The dashed lines denote those curves characterized by a single electron excitation at equilibrium separation. Plus signs indicate H+ asymptotes.

character. The nature of the low-lying curves in figs. 8 and 9 at the equilibrium geometry are given in table 2. The lowest states (dashed curves) represent single excitations at equilibrium separation from the valence orbitals to the lowest lying, unoccupied orbitals. The + signs near R = 10 bohr indicate which states lead to protons. These states are not likely to contribute to proton desorption for several reasons. First, most of these excitations lie below the thermodynamic thresholds for producing protons. Only the lb, -+ 2b, (20.48

Table 2 Calculated vertical excitation energies for Hz0 ‘A“ Symmetry

‘A’ Symmetry Root

Energy ‘)

Excitation

Root

Excitation

1 2 3 4 5 6 7 8 9 10 11 12 13 14

la22a21b23azlbz 1 1 2 1 3a, + 4a, 3a, + 2b, lb, -+ 4a, 3a, -, 5a, lb, -t Zb, lb: -) 4az lb, + 5a, lb: -+ 4a,2b, 3a: -f 4at lb: -+ 2b: 3a,lb, + 4a: lb, -+ 2b, 2a, + 4a,

1b,

0.0 10.0 14.1 15.5 15.7 20.5 20.8 21.2 25.0 25.5 25.8 28.0 28.6 29.6

Energy ‘) (ev)

(cV 1 2 3 4 5 6 7 8 9

lb, + 4a, lb, + 5a, lb, + 2b, lb,3a, -+ 4a: lb,3a, + 4a,Zb, lb,3a, - 4a,5a, 1 b,lb, + 4a: lb,3a, * 4a,2b, lb,3a, + 4a,5a,

7.6 11.0 13.2 22.3 22.5 25.1 25.4 26.4 26.7

‘) With respect to ground state ‘A, water. b, Ground state configuration of neutral water.

eV) and the lb, + 5a, (21.17 eV) excitations (and the excitations from the 0 2s which will be discussed separately) lie above the ion pair formation (z-i,0 4 H++ OH-) threshold of 16.93 eV. However, as was discussed previously, the POLl-CI level of calculation does not accurately represent the lb, -+ 2b, and lb, -+ 5a, excitations. They should, in fact, be somewhat lower in energy relative to the double excitations than is shown in fig. 8. At the more accurate POL-CI level, for example, the lb, + 2b, (19.47 eV) and the lb, + 5a, (19.15 eV) transitions lie one volt below the lowest double excitation. Consequently, the single excitations from the lb, do not lead adiabatically to protons nor are there strong adiabatic couplings to curves leading to protons. Furthermore, these states are in the continuum with respect to ionization of the lb, orbital. For these reasons, we do not expect direct photodissociation following a lb, -+ 2b, or lb, + 5a, tr~sition to be a significant ~ont~butor to proton desorption. As noted already, lying above the singles - and overlapping with them to some extent - are states which involve double excitations with respect to the ground state. The lowest of these states of ‘A’ symmetry, lb: + 4a:, has a vertical excitation energy of 20.8 eV. While the adiabatic curve (root 7 in fig. 8) dissociates asymptotically to protons, the kinetic energy predicted by this pathway is only G 0.1 eV. The diabatic pathway starting from this state would involve a possible coupling to the lowest asymptotic proton state - however, due to the large energy splitting between the curves, this coupling is expected

to be small implying only a small number might reach this 17.81 eV asymptote (experimentally observed at 16.93 eV). Before turning to a discussion of the higher lying ( > 24 eV) excited states of ‘A’ symmetry, let us examine the corresponding low lying ‘A” excited states. As one can see from fig. 9 and table 2, there are no excited states reached through a single excitation in the near threshold region (16-24 eV). In this energy range there are only two states of importance, both characterized by double excitations from the ground state, and are given as follows:

lb, 3a, -+ 4ai,

E = 22-3 eV,

lb, 3a, -+ 4a, 2bz,

E = 22.5 eV_

Neither of these states leads adiabatically to protons. Ifowever, protons could be produced by virtue of the curve crossing (roots 4 and 5 with 6) at R = 5-6 bohr. The asymptotic energy would be 20.6 eV. Protons could be produced by crossing from curves of ‘A” symmetry to those of ‘A’ symmetry. This would provide a means of reaching the lowest ionic asymptote at 17.8 eV but for diabatic behavior such curve crossing events between states of different symmetries is very unlikely 1231. We now turn to the higher lying excited states on the neutral curve manifold. At approximately 2.5 eV there is an onset of a more dense band of doubly excited and strongly repulsive states of both ‘A’ and ‘A” symmetry. Again, the nature of these states at ~uiIib~um separation is given in table 2. Many of these states are significantly more repulsive than the lower lying doubly excited states discussed above and should couple diabati~ally both with curves leading to the 20.7 eV asymptote and, to a lesser extent, with the 17.8 eV asymptote. From these states one would expect most proton kinetic energies to be near 5 eV. These states are analogous to those which we suggested were responsible for proton desorption from a metal surface [5,6]. At higher energies, embedded in the repulsive, doubly excited states are excitations from the 2s orbital of the oxygen atom (2a,): the lowest excitation, to the 3s Rydberg orbital, lies at 29.6 eV. While this state is not repulsive in a diabatic sense, it mixes strongly and contributes to several of the repulsive states in this region which can lead to highly energetic protons, We will discuss this further in section 7.

We now turn to desorption processes in which ionization of the molecule occurs before the dissociation event. The potential curves on which these occur are shown in figs. 10 and 11. The asymptotes which correspond to protons are again labelled with a plus sign. For each symmetry, the first 10 vertical ionization energies are given in table 3. The lowest lying states involve one

J. 0. Noel et al. / ESR of protons from water

Table 3 Calculated vertical ionization

energies for H,O *A” Symmetry b,

2A’ Symmetry Root

Excitation

Energy a’

Root

Excitation

3a, + 00 lb,+w lb: --) 4a,m lb: + 2b,w 3a: + 4a,w 3a,lb, 4 4a,w 2a, +w 3a: + 2b,w lb< -+ Sa,w 3a,lb, 4 4a,w b,

14.0 19.8 26.6 29.0 31.2 32.5 33.3 33.5 34.1 34.4

a) With respect to ground state H,O. b, Two roots of the same spatial configuration

Energy ‘) (ev)

eV 1 2 3 4 5 6 7 8 9 10

139

1 2 3 4 5 6 7 8 9 10

lb, - 00 lb,3a, + 4a,w lb,3a, + 4a,w b, lb,3a, + 2b,w lb,3a, + 2b,cc b, lb,lb, + 4a,w lb,lb, + 4a,w b, lb,lb, -+ 2b,w lb,3a, + Sa,cc lb,3a, --) 5a,w b,

but different spin configuration

12.0 25.9 28.0 28.1 30.0 30.9 31.1 32.2 32.6 34.9

are possible.

electron excitations (ionizations), again denoted by dashed lines. Above these are more repulsive doubly excited states in which one of the two electrons is ionized. These are denoted by the solid line curves in figs. 10 and 11. Also, analogous to the case of the neutral curve manifold, excitation (ionization) of the 2a, (oxygen 2s) orbital, with a vertical excitation at the POLl-CI level of 33.3 eV, lies embedded within the band of more repulsive double excitations. Each of the one-electron ionization processes is approximately 4 eV higher in energy than the lowest analogous nonionizing excitation. For the near threshold region this results in the excited state reached by ionization of the lb, orbital (root 2 in fig. 10) occurring virtually at the thermodynamic threshold for proton desorption. As will be discussed in the next section it is the excitation to this state that is the first step in the dominant predissociation process of gas phase water. Due to the charge state of the molecule there is no longer a long range coulombic attraction [24] as there was for the neutral system. As a result none of the adiabatic curves rise in energy at large interatomic separation. This results in a larger overall potential energy drop during desorption. This could in part account for the larger ion kinetic energies observed at higher excitation energies.

6. NiH,O and (H,O), Until now, we have centered our attention on a description of an isolated, gas-phase water molecule. Since the experiments were primarily focused on

condensed phases, it is important that we make an assessment of how transferable the conclusions drawn from molecular calculations for the monomer are to the chemisorbed water and ice regimes. In order to compare these different regimes on an equal footing, Hartree-Fock calculations were performed for II@, (H,O), and NiH,O. In (H,O),, a central molecule was tetrahedraIly coordinated to four other water molecules in an ice-like configuration. Shifts in the orbital energies for the central water with respect to the monomer should then be representative of changes one could expect in ice. Similarly, the influence of the nickel atom in NiH,O should roughly indicate the magnitude of the effect one would expect chemisorption to have on the water orbitals. These latter conclusions are more uncertain since the water orbitals on a surface might interact implicitly with a large number of water molecules or Ni atoms. Koopmans’ theorem eigenvalues for the valence orbitals of water in H,O, (I-I,O), and NiH,O are compared in fig. 12. In addition, the binding energy of the 3s Rydberg orbital for each of the three systems is shown. The primary point to be obtained from these calculations is that the orbitals clearly retain their integrity as water orbitals in each of the environments. The changes in the eigenvalues with respect to the monomer are on the order of tenths of volts for (H2Qj5 and a few volts for NiH 20. For the occupied valence orbitals of water, this result was to be expected. Experimental data for water chemisorbed on Ni [25], Pt [26], Ru (2’71,Au [28] and for molecular ice ]29-331 all imply only small shifts in the lowest three valence ionization energies. The largest changes are on the order of 2 eV. As shown in fig. 12, our results are consistent with these observations. Additionally, the calculations imply that the oxygen 2s (2a,) orbital also is not much perturbed. What was somewhat less clear, a priori, is that the oxygen Rydberg orbital should remain essentially unchanged. This is an extremely diffuse orbital (the !: = 0.032 gaussian function describing this state has a maximum in its radial density at R = 3% bohr) extending mueh beyond the molecule. One might expect that this orbital would be considerably destabilized in a spatially crowded en~ronment since it must be orthogonal to each of the occupied valence orbitals. As seen in fig. 12, however, this destabil~ation is relatively modest in both (H&3), and NiH,O: in neither case is it larger than 2 eV. The conclusion to be drawn from the c~culational results is that the desorption mechanisms presented for II,0 should be essentially unchanged for chemisorbed and ice phases. There is however, one possible exception: as will be discussed in the following section, the onset of gas phase proton desorption depends upon having the lb, ionization at an energy very near the thermodynamic threshold for dissociative ionization. In the experiments cited above for H,U on Ni, Pt, Ru, Au and ice, there was agreement that each of the valence orbitals was made less stable compared with gas phase ionization values. In particular, relative to the vacuum level, ionization of the lb, orbital was shifted

141

0.

(4.95)

(3.12) ,.. ~,,,,,,.......~'~~' 4,....1.

(3.54! ~~.....-.....__...,,...,~,",,*

4a1

zat

-10.

lb?

*r c9

(13.81?

3a 1 f’s.421

t

iz -20.

.._.I. Pf13.48) ..__ ,..,_.‘,l.t.‘.....~~.~~~ c 1 3.47) -~~~~~~~~.....,‘,, _...,..,.~...,...~..~~........ _'...,_ Jl6.4@) I'.....,, '..._ -.._,_ ‘k..,

“4

(ts.5s)

_..._....... ... . ........ . (tS.B8f

I..., f fS.401 -.._,_ .‘...._.._ ....._ (2 1.83f

-30.

*+ -40.

(37.08) ._.‘,."HI

+

(36.91) ,..,....,.........Lt-ll -..._ '...__ .+.......,_, '.... (39.39)

Fig. 12. Comparison of Koopmans’ eigenvalues and the 3s Rydberg orbital (4a, ) binding energy obtained from Hartree-Fock calculations for H,O, (H,O),, and NiH,O.

from its gas phase value of 18.6 to 17.5 eV for water chemisorbed on the (110) face of nickel [25] and to 18.0 eV for ice [30]. What is significant, is that these small shifts may be important in the differences observed between gas phase and condensed phase proton desorption at energies near the thermodynamic threshold. 7. Proton deswption pathways Possible desorption pathways for water based on our calculations are presented in table 4, listed by increasing excitation energy, In this table, the vertical excitation energies quoted are those of the more accurate POL-CI calculations. In essence, the pathways involve one of the three different types of mechanisms as discussed in the introduction and described by fig. 1. These

142

J.O. Noel1 et al. / ESD

Table 4 Proton stimulated

desorption

pathways

Pathway

Initial state

Final state

l:X-+B=C 2:X-A 3:X&B-C

lb,-+ce _

3a,lb, lb, + lb: --t lb,3a, lb,3a,

4: X+B=C 5:X-+B-C x-rc

GSW”‘-,+cbb’

GSW+$+

lb,3a, lb,3a,

2a,-+4a,,co

+ 4a,m Zb,, 5a, 4a: -4a: + 4a,2b, + 4a,Sa, I -+ 4a,a,

DERS ‘) DERS

-

a) Ground state water. b, Continuum orbital. ‘) Doubly excited repulsive

ofprotonsfromwater

Vertical excitation energy (eV) la.4 19.2, 19.5 20.1 21.6 22.7 > 25

> 29

Proton kinetic energy (eV) cl <2 O-2 5-6 6-8 9-10

state.

involve (i) single excitation to an excited repulsive state, denoted X -+ A; (ii) direct double excitation or shake-up to an excited repulsive state denoted X -+ C; and (iii) a prediss~iation mechanism involving first a single excitation to an excited non-repulsive state followed by a curve crossing event to a doubly excited repulsive state, denoted X -+ B =j C. The lowest energy pathway is a predissociation (type iii) involving a bound intermediate state in which the molecule is initially ionized before the curve crossing event. Specifically, this intermediate state is that associated with the ionization of the lb, orbital (calculated: 18.4 eV [34]; measured: 18.5 eV) which lies almost at the thermodynamic threshold of 18.76 eV. This state, as is evident from fig. 10 (root 2), is bound and hence cannot directly dissociate. However it does cross with the second 2A” state [35] in fig. 11 (described by the excitation lb, 3a, + 4a,oc) which asymptotically yields protons. The relationship between these states is more evident in fig. 13 where we have plotted only the lowest 4 curves for the ionized system. This predissociation path provides a mechanism which facilitates proton desorption immediately upon reaching the thermodynamic threshold. Protons produced by this mechanism will possess very little kinetic energy. This pathway for the dissociative ionization of water has been presented earlier [36] and is now generally believed to be the dominant process at all incident energies above 18.76 eV for gas phase water. Our results are consistent with this and provide calculational corroboration for the mechanism as it was presented by Lorquet and Lorquet [3’7]. It is important to point out that this mechanism is dependent upon the unique situation of having both a bound state lying very near the the~~yn~c threshold for dissociative ionization

14.3

1

2

3 f?

4

5

6

(bohr1

Fig. 13. Potentiai energy curves of ‘A’ and 2A” symmetry for a singly ionized water mdecule which. are pertinent to the threshold pho~~~ss~ia~i~~ mechanism in gas phase ESD.

and a repulsive state, which asym~toticalIy dissociates to protons, cmssing the bound curve, As a consequence, one cannot necessarily expect a similar process for other molecules. Indeed, in the gas phase photodissociation experiments for W,O, NH,, and CH, by Kronebusch and Berkowitz [7bJ, only I-I,0 exhibited a proton yield peaked near the ~~~modynamic threshold energy. In the above process, the initial state which is reached has a very long lifetime - in the absence of an available molecular reneutralization process in comparison with the time required to dissociate once a curve crossing event

144

has occurred (lo-i6 s). Requirements of this mechanism suggest that slight changes in relative excitation energies and in energy transfer, which could occur due to chemisorption or ice formation might significantly affect the cross-section for the process. Specifically, the shift of the excitation energy with respect to the dissociative ionization threshold is not known, but the ionization data suggests the possibility that for ice and chemisorbed water the lb, ionization could drop below the thermodynamic desorption threshold. If so, the cross-section for pathway 1 could be markedly reduced. Alternatively, the B’B, ionized state could be ~brationally or electronically quenched in ice or chemisorbed water before the curve crossing can occur. Our experimental results for the chemisorbed and ice phases, in which we find virtually no protons produced at the 18-19 eV threshold, indicates that this dominant gas phase mechanism is indeed quenched for the condensed phase. The next energetically accessible pathway, depicted by 2 in table 4, involves a single excitation to the repulsive state lb, + 2b, or 5a,. However, as discussed earlier, these curves are not likely to couple strongly with curves leading to protons and the predicted excitation threshold and proton kinetic energy distribution do not match our experimental measurements. We therefore conclude that pathway (2) is not important in either the gas or condensed phase proton desorption. Pathway (3) is the next energetically accessible pathway in table 4 and involves double excitations to repulsive curves with a predicted thre’8hold of approximately 20-22 eV and a proton kinetic energy distribution of O-2 eV. This is consistent with our experimentally measured threshold and kinetic energy distributions (figs, 2 and 3). The lb: + 4a:, lb,3a, + 4,: and lb,3a, + 4a, 2b, curves involved here are double excitations which do not contain any significant single excitation component in the wave function. Thus a direct shake-up process is not likely. We therefore suggest a predissociation mechanism similar to what we proposed for proton desorption of hydrogen on nickel [5,6], in which the intermediate state involves a single excitation to a localized continuum orbital. While the mechanism is similar to the predissociation pathway (l), it differs in that the intermediate here is not long lived but can decay rapidly, particularly through ionization. Thus, while the probability of excitation to the intermediate state may be large, the curve crossing to the repulsive curve will be small due to the competing decay processes. However, is not important; the hydrogen once on the repulsive state, “reneutralization” nucleus will leave the molecule, forming H+, HP, H or excited H* depending on additional curve crossing events as indicated in figs. 8-11. Recent high-resolution PSD experiments of Stulen and Rosenberg [ll] are consistent with the two lower energy excitations being responsible for the threshold proton desorption for water multilayers. The next energetically accessible pathway presented in table 4 is (4) and has an onset of appro~mately 25 eV. The mechanism is essentially the same as for

pathway 3 just discussed, involving a predissociation mechanism to doubly excited repulsive curves. As in pathway 3, most of these states cannot be reached directly through a strong shake-up process therefore suggesting an intermediate excitation to a localized continuum orbital. However, since the excitation is higher in energy, a channel (4a) involving curve crossing directly to an ionized repulsive state becomes available. This excitation energy pathway can produce protons with greater kinetic energy. The onset of this pathway is consistent with the experimental observation of a gradual shift in proton kinetic energies as the incident electron energy is increased above threshold. Finally, the last and highest energy pathway presented in table 4 involves excitations out of the oxygen 2s (2a,) orbital. The states reached by single excitations out of the 2s are mixed with the doubly excited repulsive states in the same energy range, thus giving rise to the possibility of significant curve crossing events between the singly excited states and repulsive states {pathway 5a) or to the possibility of a direct shake-up process (pathway 5b). Indeed, the mixing between states is so great that it is not possible here to distinguish between these two mechanisms. The initial excitation can involve the promotion of the 2s electron to a bound Rydberg orbital, a continuum orbital, or, at energies greater than 32 eV, ionization. On the one hand, experimental photoabsorption crossections indicate that excitations from the 2s orbital near 30 eV are small [7a,7b]. On the other hand, the overall mechanism is more direct than pathways 3 and 4. We cannot as yet make an a priori assessment of how these two counterposed effects balance one another. Since the minimum excitation energy for the 2s electron is greater than 28 eV, such a pathway cannot be responsible for the experimentally observed 20-21 eV threshold for proton desorption from chemisorbed water or ice. However, it is likely that pathway 5 involving the 2s excitations is responsible for the behavior of the ion kinetic energy distribution curves above 30 eV. The more repulsive nature of the doubly excited states coupled with the 2s excitations in this energy range would give rise to proton kinetic energies up to 10 eV. This is consistent with our experimental results for ice which show the onset for production of higher kinetic energy protons (see fig. 4) at energies above =I 30 eV. It has recently been suggested by Stockbauer et al. [39] and Weng and Kammerer 1401 that excitation of the oxygen 2s orbital is responsible for the near threshold (20-25 eV) proton desorption for H,O on titanium, niobium and tantalum. Our calculations as discussed above show that this is not likely. Again, for the monomer the 2s ionization is at 32-33 eV [41], a result corroborated by gas phase photoionization experiments [7a,42,43]. As has already been noted, upon chemisorption, the orbitals of the molecule are relatively unchanged. These results are in good agreement with XPS experiments on ice [32,33] which show that the 2s ionization is at 32 eV relative to the vacuum level; basically unchanged from gas phase data. Even excitations

146

J.O. Nod

et al. / ESD of protons from water

from the oxygen 2s to the lowest lying oxygen Rydberg require on the order of 28-29 eV and are well above the observed threshold. Excitation to the Rydberg orbital is analogous to excitation to the Fermi level for the solid surface. For typical work functions in the 4 to 5 eV range, excitation from the 2s to the Fermi level would require 27-28 eV, a value well above the observed proton desorption threshold. The improper assignment of the 0 2s ionization potential for water by Stockbauer et al. [39] and Weng and Kammerer [40] may be due to an implicit assumption that this orbital should be the same in both gas phase oxygen and water, which is not the case. Weng and Kammerer [40] also examined the ion energy distribution for protons desorbing from Ta and Nb surfaces which had been at least partially exposed to H,O. They discovered that above = 30 eV, the energy distribution curves were bimodal, exactly analogous to what we have seen here for water on Ni. They proposed that this arises from the initiation of a new mechanism involving excitation of the metal core levels, Ta 4f (25 eV) (or 5p) and Nb 4p (34 eV). This seems rather tenuous since it would predict that the onset of the high energy portion of the proton kinetic energy curves should differ by about 9 eV for the two metals - what they in fact observe is that in both cases the onset or threshold is the same, = 30 eV. Based on our results we suggest that the high proton kinetic portion of the energy distribution curves observed by Weng and Kammerer is due to the same pathway as we have presented here: excitation of the 0 2s leading to doubly excited repulsive states (pathway 5).

8. Summary and conclusions Based upon the above analysis of excited state potential energy curves and experimental data we can now draw a consistent picture for proton desorption from gas phase water, ice and water chemisorbed on a metal substrate. In general, our results show that the orbital energies of the water molecule remain basically unchanged going from the gas phase to molecular ice. As a result, the threshold energies associated with the various H+ desorption mechanisms should remain relatively unchanged for the different phases. We also expect that the characteristics of proton desorption from water chemisorbed on Ni(ll1) should be essentially the same for any surface on which molecular adsorption occurrs. The dominant features of gas phase ESD of protons from H,O are the 18.7 eV threshold and proton kinetic energies of less than an electron volt regardless of the incident excitation energy. This clearly implies that pathway 1 of table 4 represents the dominant mechanism for desorption at all energies. In view of earlier discussions, it is of course not surprising that it should be the dominant, indeed only, mechanism near threshold. It is interesting, however,

J.0. hell

et af. j ESD ofproions from water

147

that one does not observe hotter protons coming off at higher incident energies when pathways 2-5 become energetically possible since there should be no physical phenomena precluding them. Failure to detect them must represent a limit in sensitivity. The cross-section for these higher energy processes is expected to be much lower than that for pathway 1. As a result the H” kinetic energy profile is completely dominated by the low energy protons (= 1 eV) produced by this threshold predissociation mechanism. The situation for water chemisorbed on a Ni(ll1) surface is very different. In that case, we observed no strong evidence for the occurrence of pathway 1. This failure to see a strong 18.7 eV threshold with less than 1 eV protons must be attributable to the interaction of the molecule with the solid. How the effect can be so large as to totally negate the dominant gas phase effect can be understood as follows. The critical feature of the gas phase threshold mechanism is the long lifetime of the B*B, state. Since this is a bound state, it can live for many vibrational periods (lo-l3 s) before predissociating. This extended lifetime is necessary for this to be a facile mechanism since the curve crossing occurs at a rather long internuclear separation (3.5 bohr). When the water is chemisorbed on a nickel surface, this time is not afforded the molecule in the BZBz state following ionization. The metal serves as an energy sink so that the excitation can be transferred to the solid on a time scale of lo-‘” s. This essentially precludes pathway 1 as a major cont~butor at any excitation energy to the stimulated desorption of protons from adsorbed water. What one does observe for the chemisorbed case is a threshold of 21-22 eV with the desorbed protons carrying 2-3 eV of kinetic energy at threshold. This corresponds very well with what our calculations suggest for pathway 3. At somewhat higher energies, pathway 4 becomes possible and gives rise to protons of up to 6 eV kinetic energy. Above 30 eV, excitation of the oxygen 2s becomes potentially important. This is represented by pathway 5. The 2s ionization pathway gives rise to very high energy protons, up to 10 eV. Additionally, one now has that instead of competing with the stimulated desorption process, production of ions follows as a result of the ionization of the oxygen 2s electron. However, this process is still weak for chemisorbed water as was shown in fig. 5 where only a very small high H+ kinetic energy contribution is observed above 30 eV excitation for B = 0.3 monolayers. The final coverage regime to be considered is the thick condensed layer, where in essence substrate supported ice is being sampled. The nickel surface plays no role in the desorption process: the features observed here should be identical regardless of the substrate. The orbital energies of the water molecule are virtually unchanged upon condensation. One might thus expect to see more similarity between gas phase water and ice than is actually the case. Below 21 eV an extremely weak H+ desorption is observed with the protons carrying on the order of 1 eV or less kinetic energy. These are likely produced through pathway 1 but again as in the case of chemisorbed water, the B2B, state

148

J.O. Noel1 et al. / ESD

of protonsfromwater

reached after ionization of the lb, orbital does not live long enough to effect a curve crossing to the C*B, repulsive curve. Thus this pathway is also strongly quenched in ice as it was in the chemisorbed phase. Protons which were attributed to the continuum predissociation for chemisorbed water are also observed for the ice layer with no substantive changes between the two regimes. Neither the threshold nor the kinetic energy distribution is significantly altered. Above 30 eV the ESD spectrum of ice has an additional feature which is not as markedly apparent in the chemisorbed phase; there is the onset of a mechanism which yields protons of 7-8 eV kinetic energy. This we have assigned as being due to pathway 5. The primary step of this pathway is ionization of the 2s orbital which has a vertical excitation energy of 32.2 eV and a Franck-Condon region from 30-36 eV [5a]. As was pointed out earlier, the excited state reached by this process is represented diabatically by a bound state curve, a result of the fact that the 2s electron is not involved in hydrogen bonding. This state is embedded within a very repulsive band of doubly excited states and hence through curve crossing events can lead to high energy protons. Finally, we have shown here that by comparing experimenta:ly determined threshold data with calculated excited state potential energy curves it is possible to identify the generic form of excitations and mechanisms important to ESD and PSD processes. We have not as yet concentrated on predicting the exact distribution of ion kinetic energies. Such studies however, are important since they must ultimately focus on the adiabatic versus diabatic nature of the fragmentation process. This important issue has already been the center of controversy between Brenig [44,45] and Bell et al. [46] in a series of papers dealing with the theory of electron-stimulated desorption. Also, we have seen here for water that it is important to examine the excited states of both the neutral and ionized molecule. We expect that such an analysis will in general be crucial for the understanding of ESD and PSD in molecules and molecular solids. For the study of stimulated desorption of adsorbates on metals such an analysis is unlikely to be as important since the addition or subtraction of an electron has very little change on the total energy of the system.

Acknowledgement The authors would like to thank S.J. Haney assistance throughout these experiments.

for

his expert

References [l] See, for example,

D. Menzel, J. Vacuum

Sci. Technol.

20 (1982) 538.

technical

J. 0. Noell et ai. / ESD of protons

fromwater

149

[2] Cf. (a) H on W: T.E. Madey, J.J. Czyzewski and J.T. Yates, Surface Sci. 49 (1975) 465; D. Menzel, Surface Sci. 47 (1975) 370; (b) H on Pt: J.H. Craig and J.L. Hock, Surface Sci. 100 (1980) L435; (c) H on Pd: R.H. Stulen, J. Vacuum Sci. Technol. 20 (1982) 846; R.H. Stulen, T.E. Felter, R.A. Rosenberg, M.L. Knotek, G. Loubriel and C.C. Parks, Phys. Rev. B25 (1982) 6530; (d) H on Ni: see ref. [3], and F.N. Simon, D. Lichtman and T.R. Kirst, Surface Sci. 12 (1968) 299; (e) H on Ce,O,: B.E. Keel, G.M. Loubriel, M.L. Knotek, R.H. Stulen, R.A. Rosenberg and CC. Parks, Phys. Rev. B25 (1982) 5551; (f) H on Si: H.H. Madden, D.R. Jennison, M.M. Traum, G. Margaritondo and N.G. Stoffel, Phys. Rev. B26 (1982) 896; (g) Condensed phase H: R. Clampitt and L. Gowiand, Nature 223 (1969) 815. [3] D. Menzel and R. Gomer, J. Chem. Phys. 41 (1964) 3311. f4] Redhead, Can. J. Phys. 42 (1964) 886. [5] CF. Mehus, R.H. St&en and J.O. Noefl, Phys. Rev. Letters 48 (1982) 1429. [6] C.F. Melius, J.O. NoelI and R.H. Stufen, J. Vacuum Sci. Technol. 20 (1982) 559. [7] (a) K.H. Tan, C.E. Brion, Ph.E. van der Leeuw and M.J. van der Wiel, Chem. Phys. 29 (1978) 299; (b) P.L. Kronebusch and J. Berkowitz, Intern. J. Mass Spectrom. 22 (1976) 283; (c) J.H.D. Eland, Chem. Phys. 11 (1975) 41; (d) N. Bose and W. Sroka, Z. Naturforsch. 28 (1973) 22; (e) R.B. Cairns, H. Ho&on and RI. Schoen, J. Chem. Phys. 55 (1971) 4886. (81 K.M. Monahan and V. Rehn, Nucl. Instr. Methods 1.52 (1978) 255. [9] This assumes no significant contact potential difference between the sample and the detector. [lo] R.A. Rosenberg, V. Rehn, V.O. Jones, A.K. Green, CC. Parks, G. Loubriel and R.H. Stulen, Chem. Phys. Letters 80 (1981) 488. [ll] R.H. St&en and R.A. Rosenberg. J. Vacuum Sci. Technol. A2 (1984) 1051. 1121 R.H. Prince and G.R. Floyd, Chem. Phys. Letters 43 (1976) 326. [13] R.D. Levine and R.B. Bernstein, Molecular Reaction Dynamics (Oxford University Press, London, 1974) p. 175. [14] F.W. Bobrowicz and W.A. Goddard III, in: Modern Theoretical Chemistry, Vol. 3. Ed. H.F. Schaeffer III (Plenum, New York 1977). [15] J.O. NoeII, M.D. Newton, P.J. Hay, R.L. Martin and F.W. Bobrowicz, J. Chem. Phys. 73 (1980) 2360. [16] S. Huzinaga, J. Chem. Phys. 42 (1965) 1292. 1171 T.H. Dunning, Jr., J. Chem. Phys. 53 (1970) 2823. [18] T.H. Dunning, Jr. and P.J. Hay, in: Methods of Electronic Structure Theory, Ed. H.F. Schaeffer III (Plenum, New York, 1977) p. 1. 1191 This wavefunction was chosen such that the 4a’ and Sa’ orbitals (water Rydberg and oxygen-hydrogen antibonding orbitals respectively) would be uniquely defined rather than determined merely by or~ogon~ization to the occupied orbitals of ground state H,O. 1201 Since each of the orbitals was defined by what is essentially a generic state, each of the states of interest is treated equally with respect to the correlation which is included. [21] P.J. Hay and T.H. Dunning, Jr., J. Chem. Phys. 67 (1976) 5077; H.F. Schaeffer and F.E. Harris. Phys. Rev. Letters 21 (1968) 1561. 1221 The ‘A’ and 3A” states were not calculated since they are not dipole allowed. Their potential energy curves will be slightly lower than their corresponding open shell singlet counterparts. We do not expect these states to be important at high incident electron energy. [23] Assuming an interaction region of 0.9 bohr. 7 eV kinetic energy and a 1.2 eV energy gap, Landau-Zener theory predicts a probability of 0.024. [24] In both figs. 7 and 8 it is important to note that the asymptotes leading to protons are 2.7 eV higher than the value at R = 10 bohr due to the coulombic interaction.

[25] C. Benndorf, C. Nobl, M. Rusenberg and F. Thieme, Surface Sci. 111 (1981) 87. [26] G.B. Fisher and J.L. Gland, Surface Sci. 94 (1980) 446; G.B. Fisher and B.A. Sexton, Phys. Rev. Letters 44 (1980) 683. (271 P.A. Thiel, F.M. Hoffmann and W.H. Weinberg, J. Chem. Phys. 75 (1981) 55. [28] CR. Brundle and M.W. Roberts, Surface Sci. 3X (1973) 234. [29] B. Baron and F. Williams, J. Chem. Phys. 64 (1976) 3896. [30] T. Shibaguchi, H. Onuki and R. Onaka, J. Phys. Sot. Japan 42 (1977) 152. 1311 M.J. Campbell, J. Liesegang, J.D. Riley, R.C.G. Leckey and J.G. Jenkin, J. Electron Spectrosc. Related Phenomena 15 (1979) 83. 1321 P.J. Page, D.L. Trimm and P.M. Williams, J. Chem. Sot., Faraday Trans. I, 70 (1974) 1769. 1331 I. Abbati, L. Braicovich and B. De Michelis, Solid State Commun. 29 (1979) 511. [34] The value of 18.4 eV is taken from the more accurate POL-CI calculations reported in table 1. At the POLl-Cl level - those corresponding to the results presented in figs. 7-10 - a value of 19.8 eV is calculated. [35] A quartet state of the same electronic configuration (la: 2a: lb: 3a: lb: 4a’,) would also exist. This state will lie somewhat lower than the doublet state presented here. It is this quartet state which Lorquet and Lorquet (see ref. 1371) postulated as being responsible for proton desorption. This is not possible. however, since the lowest OH + H+ asymptote must be a doublet. [36] F. Fiquet-Fayard and P.M. Guyon, Mol. Phys. 11 (1966) 17. 1371 A.J. Lorquet and J.C. Lorquet, Chem. Phys. 4 (1974) 353. [38] (a) P. Gurtler, V. Saile and E.E. Koch, Chem. Phys. Letters 51 (1977) 386; (b) A.W. Potts and W.C. Price, Proc. Roy. Sot. (London) A236 (1972) 181. [39] R. Stockbauer, D.M. Hanson, S.A. Flodstrom and T.E. Madey, Phys. Rev. 826 (1982) 1885. [40] S. Weng and O.F. Kammerer, Phys. Rev. B26 (1982) 2281. [41] The value of 33.3 eV is that calculated at the POLl-CI level. The actual ionization energy, as was presented in table 1, is somewhat less: 32.2 eV experimentally and 32.3 eV at the POL-Cl level. [42] K. Siegbahn, J. Electron Spectrosc. Related Phenomena 5 (1974) 3. 1431 M.B. Robin, N.A. Kuebler and C.R. Brundle, in: Proc. Intern. Conf. on Electron Spectroscopy, Ed. D.A. Shirley (North-Holland, Amsterdam. 1972) p. 351. [44] W. Brenig, 2. Physik B23 (1976) 361. [45] W. Brenig, Surface Sci. 61 (1976) 659. [46] B. Bell, M.H. Cohen, R. Comer and A. Madhukar, Surface Sci. 61 (1976) 656. [47] J. Appell and C. Kubach, Chem. Phys. Letters 11 (1971) 486. [48] This should be degenerate with the 20.7 eV ‘A’ asymptote. The slight difference is due to asymmetries introduced in the CI. A more accurate value for this asymptote is 19.8 eV. This is taken from thermochemical data for the bond breaking process. Also for both cases, the OHfragment could decay during the desorption process leading to an asymptote 1 eV lower.