PHOTOASSISTED DIMOCIATION OF NITROUS OXIDE
nvolving Electron Transfer at Semiconductor Surfaces.
Photoaseiste ]Dissociation of Nitrous Oxide over Illuminated erric Oxi.de and Zinc Oxides1
Joseph Cunningham,* J. J. Kelly, and A. L. Penny Department of Chemistry, University College, Beljield, Dublin 4, Ireland
(Received July 89, 1970)
Publication costs assisted by the Air Force Ofice of Scientific Besearch, through the European O f i e of Aerospace Research
Qaseow nitrous oxide dissociated when contacted a t 20" with illuminated surfaces of powdered zinc oxides or ferric oxide. The quantum efficiency of photoassisted dissociation over pure zinc oxide was small for all conditions used in this work, being only when high pressures of NzO (-500 Torr) were present during illumination and when photons had energies greater than the zinc oxide bandgap of 3.2 eV. Lower quantum efficiency (4 -10-6) was observed for photons of energies 290 n m Curve e refers to FezOswhich had been treated in the 8ame manner. iii. Wavelength Dependence of Photoassisted Dissociation ouer ZnO. Data for curve b of Figure 3 were
PHOTQA~SIIWED DISSOCIATION OF NITROUS OXIDE 2 r----
621
1
0 P
rrzu
Time (‘mins 10
I
20
ltorrl
Figure 3. Variatior, in rate of photoassisted dissociation of NzO with pressure of N z over ~ (a) ZnO, X > 290; (b) ZnO, X > 364; (c) FepOa, X > 290.
obtained for a ZnO sample which had received the same treatment as that used for curve a except that light reaching the sample was restricted to wavelengths > 364 nm using filter b. The overall light intensity incident on the oxide was reduced by 27% by the filter solution. The extent of photoassisted dissociation was redueed by 75%. Thus, as expected, light inside the band edge of ZnO (A < 385 nm) was more efficient in producing dissociation. Further evidence that photons of en2rgy less than tlie band gap of ZnO were contributing to N70 dissociation was obtained when various solutions OF iYi(CN)42- were used to restrict light incident on the ZnO to progressively longer wavelengths. Dissociation as observed for wavelengths up to 546 nm. Quantum effioiencies for chemical decomposition of KsO(i.e., number off Nz molecules produced per photon) for three speotral regions are listed in Table I. The difference in quantum eflkiency observed using filters a ( h > 290 nan) and b (A > 364 nm) gave the value for quantum efficiency in the region of the band gap (290 < X < 3614 nm). Similarly the difference in quaiiturn efficiency observed for decomposition using filters b and c gave a value for the wavelength region 364 < h < 546 nm. The value for X > 546 nm was obtained by using filter 6 . iv. “ M e m o ~ y.Efect.” A “memory effect” corresponds t o a greater dark dissociation rate over an oxide surface which bad been preilluminated than over the unillumina.ced surface. Such an effect is illustrated for ZnO in Figure 4. ‘The lower dotted line (b) shows the mi) of product Nz detected measured value (2.4 X after a 0.5-hr reaction in the dark over a surface previously equilibrated with 50 Torr N2O. Data on the upper curve (a) were obtained by condensing tlie gaseous N20 in contaci, with the oxide into a liquid nitrogen trap :and illuminatirig the surface for 1.5 hr at pressures of the order oE 1-S >( Torr, corresponding t o Nz product from prior dark reaction. When illumination
Figure 4. Zn0 “memory effect”: (a) MN,the volume of Nz produced by contacting 50 Torr Ne0 with ZnO for 30 rnin following illumination of the equilibrated ZnO under vacuum for 1.5 hr. The time interval between termination of illumination and introduction of NzO is given on abscissa. (h) Volume of Nz produced by contacting 50 Torr NzO with the equilibrated ZnO not preirradiated for 30 min in the dark. (e) V N ,as in (a) but equilibrium pressure 0 2 = 1.0 X 10-3 Torr introduced during interval between termination of illumination and introduction of NaO.
was ended, P I 2 0 was evaporated to contact the oxide surface either immediately or after the interval indicated on the abscissa of Figure 4. The difference between the upper and lower plots exceeds the indicated experimental error, thus confirming a “memory effect” which persisted for more than 20 min after illumination. A similar “memory effect” was observed for Fe20s. Oxygen decreased the magnitude of the “memory effect” in ZnO. For a sample preirradiated for 1.5 hr as described above and then exposed to 1.0 X Torr O2 immediately after irradiation was ended, the volume of N2produced by a subsequent 0.5-hr dark reaction was considerably lower than that observed for the untreated preirradiated oxide, as shown by point e of Figure 4. B. Photoconductivity of Oxides; X > 290 nm. Irradiation of ainc oxides with light corresponding to fundamental absorption produces p h o t o c o n d u c t l ~ i t y . ~ ~ - ~ ~ Section a of Figure 5 illustrates the increase in eurrent observed when an activated Li-ZnO film was illuminated at 20” with light of X > 290 nm. This photocurrent did not attain a constant value within the 3 min shown. Section b shows the decrease in current observed when NzO (0.6 Torr) was introduced to the film during illumination. Section c illustrates the initial sharp decrease in current of the oxide film in the presence of N2O which occurred when illumination was ended. This was followed by a slower decay towards the value in curve d, obtained when KZO (0.6 Torr) was contacted with an activated Li-ZnO film in the dark. The rates of decrease in photocurrent shown in sections b and c of Figure 5 increased EM E20 pressure was increased. The Journal of Physical Chemistry, Vola 76,Na. 6,1071
J. CUNNINGHAM, J. J. KELLY,AND A. L. PENNY
622
10 Time (mins )
0
20
Figure 5 . Variation oi current flow through a Li-ZnO film with illumination and exposure to NzO a t 20'. (a) The increase in currmt z through an activated film under illumination Eollowed Ep (b) the decrease due to exposure of film to 0.6 Torr NzO, ( c ) further decrease in current on termination oE illumination, (d) dark current decrease on exposing an advateci film to 0.6 Torr NzO.
-".I
Oh
OFF
20 40 Time (mins)
60
80
'
Figure 6 . Variation of photoconductivity of ZnO with NzO pressure. A i , the increase in current through ZnO film, previously equilibrated with 0.6 Torr NzO and illuminated under (a) 0.6 Torr NzO, (b) 0.05 Torr NzO, (c) continuous vacuum, current decreased on termination of illumination.
Figure 6 presents results of a sequence of measurements which esta,blished how the change in photocurrent, Ai, 0 1 ZnO depended on N20 pressure. The dark current was first reduced to a constant value by contacting the ZnO film with 0.6 Torr NzO for 40 min. Subsequent illumination (A > 290 nm) of the equilibrated film In the presence of 0.6 Torr NzO produced an increase in current, Ai, as shown in curve a. An equilibrium Photocurrent (&e., A i constant) was established within 20 min of the start of illumination. This establishment of 8x1 equilibrium photocurrent was observed for N20pressures greater than 0.5 Torr. Illumination was ended after 40 rnin and the current decayed rapidly to the vetluei observed prior to illumination (;.e., A i decreased to zero), The half-life for decay of the photocurrenl (time taken for photocurrent to decrease to half the Ai value reached after 40-min irradiation) was 4 min. When MslO pressure was reduced to 5 X Torr, no change occurred in the background dark curThe Journal of Phyaical Chemistry, Vol. 76, No. 6, 1971
rent. Illumination at this pressure produced a faster increase in current as shown in curve b, an equilibrium photocurrent was not established within 40 min and the increase observed after 40-min irradiation was greater than for curve a. When illumination was ended, the current decayed more slowly than for curve a the half-life decay was 5.5 min. Illumination of the ZnO film at a residual pressure of 1 X Torr (curve c) produced the greatest increase in current and, as for curve b, an equilibrium photocurrent was not established within 40 min. Decay of the p ~ o t o c ~ r r e na t this pressure was slower than for curve b and halflife for decay was 7 min. Thus the current remained 2 10% above the dark current value for some hours after the end of illumination. Similar results were found for Li-ZnQ and Fe20a films which had been equilibrated with NzQ. Increasing NzO pressure decreased the rate of increase and the magnitude of the photocurrent and increased the rate of decay of the photocurrent (Le,, decreased the half-life). For NzOpressures greater than 0.5 Torr an equilibrium photocurrent was established within 20 min of the start of irradiation, as for the ZnO case, The greatest relative increase in photocurrent of the zinc oxides ( i e . , the largest ratio of p h o ~ ~ c u r r e ntot dark current) was observed for the oxide of lowest conductivity Li-ZnO, in agreement with the results of other workers.20s21 No photocurrent could be detected for InZnO films in the presence of NzO.
Discussion It was shown in a previous paper that growth of surface potential to a limiting value, &, associated with accumulation of negatively charged surface species via (Ib) and (IC),could explain observed kinetics of NzO dissociation over oxide surfaces in the darka2 The iimited extent of N20 dissociation attainable in the dark and accompanying decreases in conductivity were explained within this framework. Figure 7a illustrates (schematically) electronic-energy states within a semiconductor and close to its initially clean surface. Figure 7c represents electronic energy states after equilibration of the surface with NzO in the dark produces the limiting surface potential, Eo. I n understanding the dark reaction results, the important consequence of Eo was repulsion of electrons into the bulk, thereby suppressing (Ib), the rate-determining step of the ark dissociation a t the oxide surface. As discussed recently by Kwan, et a1.,36and by Fukazawa, et al., 37 photoexcitation can produce important modifications of chemisorption processes at surfaces bearing negatively charged oxygen species. The fo"0llowing processes contribute to such modifications: (II(36) T. Kman, K. M. Sancier, Y. Fujita, M. Betaka, S. Fukazawa, and J. Kirino, J . Res. Inst. Catal., Hokkaido U I E ~ V 16, . , 53 (1968). (37) S. Fukazawa, K. M. Sancier, and T. Mwan, J . Gatal., 11, 364 (1968).
PHOTOASSTBTED DIB~SOCIATION OF NITROUS OXIDE
623
I dt
a
dt d[0-(ads) ] dt
I-
b
I
C
Figure 7. Energy diagram of n-type semiconductor (a) before, (b) after partial, and (c) after equilibrium chemisorption of acceptor species: c.b., conduction band; EF = Fermi level, E* surface barriei potentisl; Eo, limiting surface barrier potential; EA, energy level of adsorbed species.
a'), electrons, promoted by photons from negatively charged surface species, migrate into bulk conductionband states, if they escape electron-hole recombination; and (ICIls') holes, produced by photoabsorption within the depletion layer of the semiconductor particles, migrate 1,owards the negatively charged surface. For an n-type semiconductor surface dark-equilibrated with the electron-at1 aching gas before illumination (see Figure 7c), photoeffects Ira' and b' have the same net effect, reduction of the amount of excess negative charge stored adjacent to the surface. Consequently, the i l l ~ m i n a ~surface e ~ ' may be represented as bearing a surface potential B + ,lower than for the dark-equilibrated surface (cj. Figure 7b). A quantitative calculation of the reduction in surface potential effected by illumination would require detailed treatment of the various absorption and charge equilibria a t the surface. This would involve the species, 0, 0-, 0 2 - , N20, 0 2 , etc., and, since sixface concentrations of these species have not yet been measured in the conditions of our experiments, quantii,ative evaluation of the change in surface produced by itlumination will not be attempted at this time. 'The preceding discussion shows, however, that it, L reasonable to represent the qualitative effect of illumination as a fractional reduction in surface It will be shown in the followpotential (cf. Figure 7') ing paragraphs that photoinduced phenomena involving NsO ai illuminated oxide surfaces can be qualitatively underftood on this basis. The magnitude of increases in conductivity caused by illumination, Ai*,decreased as increasing pressures were mainta,ined over the illuminated surface (6. Figure 6). Conversely, the rate of photoassisted dissociation of N20 over the illuminated surface increased with f",O in the manner illustrated by Figure 3. Processes Ia, Ib, and IC, which were established as the mechanism of N20 clissociation over the same surfaces in the dark, can also account for these photoeffects, since they kave the net effect of converting a mobile carrier into a relatively immobile surface 0- fragment and a gaseous Nzfragment. Process Ib, which is ratedetermining for N20 dissociation in the dark, leads to kinetic expressions 1, 2 and 3. ~
n8 =
I Rd
=
nb
=
kb [N20(ads) ]ne (1)
exp [ - (eEo/kYr)1
k b f NzO(ads)
1nb exp [ - (eEo/kT) ]
(2) (3)
Expression 2 assumes a Boltzmann distribution between the number of carriers in the bulk, nb, and their effective concentration, n,, at a surface which bears the limiting surface potential EO,after dark-equilibration. A corresponding set of expressions, 4, 5, and 6, can be written for the illuminated surface, on the assumptions that (Ib) is still the rate-determining step for K 2 0 dissociation and that illumination reduces the surface potential to a new value E".
n,* = nb* exp[ - (eE*/kT) ]
R"
= k~,[N~O(ads)]nb* exp[-(eE*/kT)]
(5)
(6)
The effective concentration of carriers a t the illuminated surface, n,*, is higher than at the dark-equilibrated surface, (cf. the conductivity results). The same rate constant, kb, is taken for (Ib) a t thei lluminated and at the dark surface. The rates evaluated from experimental results were Rd and (R" - Rd) and experimentally Rd was less than 5% of R* except for Fe203. Expression 7 should then be a valid approximate relationship for zinc oxides and shows that
Rd/R* - Rd
nb/nb* eXp[-e(Eo - E * ) / k T ] (7)
rate of N 2 0dissociation via (Ib) and (IC)are enhanced a t the illuminated surface to an extent depending exponentially on reduction, AE, in surface potential. If nb/nb* is approximated to 1 and the measured rates of Rd and (R" - Ra) are employed in (%>,values of AE = 0.05-0.2 eY can be derived. Undue significance should not be attached to values of LE derived in this manner. Comparison of A E values with Eo value calculated for boundary-layer expressions shows, however, that the derived values are fully consistent with the fractional reduction in surface potential assume present treatment. (See Table 11.) Pressure of undissociated gas phase N2O remained effectively constant while kinetic data were taken on photoassisted dissociations because 15 min after the end of illumination. This accounts for the observed “memory effects,” in which surfaces preilluminated in O ~ C U Oretained significant additional activity for N20dissociation for periods > 15 min after illumination. The presence of another gas known to Iocalize electrons at the surface was expected to decrease the half-life of conductivity and of memory effects after illumination. Oxygen admitted to a preilluminated E n 0 surface produced these effects (ca. point c in Figure 4) Our observations on zinc oxides showing: (a) that quantum eEmcniency of photoassisted IT20dissoThe Journal of Physical Chemistry, Vol. 7’6,No. 6,197’1
ciation was -low5 even with photons of energies greater than the ZnO band gap; (b) that similar quantum efficiencies were obtained for pure, indium-doped, and lithium-doped samples; and (c) that photons with energies lower than the band gap did promote NzO dissociation, can be qualitatively accounted for, as follows, with (Ib) as the rate-limiting step. (a) uanturn eficiency for photoassisted dissociation cannot exceed that for electron promotion to the surface according to the suggested mechanism. Photon flux at the oxide surfaces was ca. IO1’ cm-2 in our work and, with this flux of photons of X = 365 nm incident on ZnO single crystals, Collins and Thomasz0 have measured a ~ u a ~ t ueficieney m of ca. 5 X 10-j. (b) For various zinc oxide samples with dark conductivities differing by four orders of rnagnitude, Collins and Thomas report similar quantum efficiencies for promotion of electrons to the surfaces of their single crystal samples, provided that photon flux was ca. lot7 Similar quantum efficiencies for photoassisted dissociation of N20 over the various zinc oxides are consistent with these observations. (c) Photoconductivity and photodesorption of oxygen has been observed on zinc oxide surfaces for M (500-600 nm) far outside the band edge of crystals (385 nm), Both observations imply reduction of the surface potential which, in turn, would lead to photoassisted dissociation, as observed, A comprehensive qualitative explanation of the observed photoeff ects at the (K20/n-type oxide) interface is thus seen to be possible with a mechanism base reduction of surface potential by illumination and consequent dissociation via (Ib) and (IC). It is pertinent to enquire briefly if other mechanisnis may also account for the observed results. The observed “memory” effects, persisting for minutes after illumination, would not be consistent with mechanisms requiring IT20(ads) to interact with photoproduced excitons, since the latter would disappear within ca. 10-* sec of their formation. It is possible to envisage excitonlike states forming by slow electron-hole recombination after illumination, but their concentrations at the surface small would necessarily be very low. The van~shi~igly probability for random encounter between low concentrations of excitonlike states produced by recombination and the immeasurably small surface coverage by N20(ads), leave process IIb followed by steps outlined below as the exciton-based mechanism possibly consistent with the “memory” effects and other results involving N20a t the illuminated surface : Migration 01 holes towards surface, (s), and electrons into bulk, b (e
+ h)[nlO+ - ()-(ads)]
+h,
+
eb
(I&)
Localization of holes adjacent to adsorbed species, thus reducing surface potential (38) V. A. Crawford and F. C. Tompkins, Trans. Faraday Soc., 46, 504 (1950).
PHOTOASSISTED DISSOCIATION OF NITROUS OXIDE h, -5 faTzQ(aIds)4[h
- N20(ads)],
(IIIb)
- O-(ads)],
(IIIb’)
and -(ads) + [h
Process I b , ~ ~ ~ ~ a t ofi c electrons in to surface against reduced surface potential Electron-hole recombination adjacent to adsorbed species, giving excitoalike states
+ [b --N&&ads)I --+ (e + h)* - N20(ads)
eB
e,
+ Lh - O-(ada)l,,
-
(e 3. h)*
- O-(ads)
(IIIc)
(IIIc’)
Utilization of energy of excitonlike state of the solid, to dissociate N,O (ads) (e
+ h)*
NzO(ads) .-+
N2
+ O(ads)
(IIId)
Although the observed kinetics and the memory effects in phokoassieted NzO dissociation might be explicatble by : ti;uch a,n exciton-based mechanism with
625
(Ib) the rate-determining step, a convincing case for operation of the mechanism cannot presently be made because direct evidence for (IIIb) is lacking. Localization of photoproduced holes adjacent to NzO(ads) has not been established. Several workers have proposed that the competing process, IIIb’, occurs on semiconductor surfaces bearing chemisorbed oxygen16and on this basis it was expected that the presence of low oxygen pressures together with N20 during illumination of the oxide surface would sharply reduce the overall rate of photoassisted dissociation if the exciton-based mechanism were important. Tests revealed, however, that oxygen pressures of Torr did not significantly reduce the rate of photoassisted dissociation of NzO.
Acknowledgments. The authors are grateful to D. 0. Carpenter, New Jersey Zinc Company, for supplying the zinc oxide materials, to Dr. Moruzei and Mr. Austin of Liverpool University for assistance with setting up the quadrupole mass spectrometer, and to Mr. Brady of this department for extensive assistance with the vacuum systems. Support of this work by funds from AFOSR is gratefully acknowledged.
Ths JourmE of Physieal Chmistry, VoE. 76,NO.6, 1971
