Light-initiated surface modification of oxide semiconductors with

Sep 21, 1987 - Department of Chemistry, Mount Holyoke College, South Hadley, Massachusetts 01075 . T. Spitler*. The Polaroid Corporation, 103 Fourth ...
0 downloads 0 Views 968KB Size
Langmuir 1988,4,861-867 initial supersaturation. Oxalate ions strongly adsorb onto calcite, their adsorption conforming with a Freundlich isotherm. As a consequence of their adsorption, oxalate ions render the calcite surface negative, causing an electrokinetic charge reversal at low pH values. Acknowledgment. We would like to express our thanks to Anthony Margaritis for assistance with DSC and

86 1

TGA analyses, Peter Busch of the State University of New York at Buffalo, and Lambros Kombotiatis for the scanning electron microscopy. A grant from the European Communities (NREN3G-0036 GR) in support of this work is gratefully acknowledged. Registry No. CaC03,471-34-1;CaC03.H20,32825-96-0; C204%, 338-70-5; calcite, 13397-26-7.

Light-Initiated Surface Modification of Oxide Semiconductors with Organic Dyes M . A. Ryan Department of Chemistry, Mount Holyoke College, South Hadley, Massachusetts 01075

M. T. Spitler* The Polaroid Corporation, 103 Fourth Avenue, Waltham, Massachusetts 02254 Received September 21, 1987. In Final Form: March 7, 1988 Photoaffinity labels were used as reagents for a light-initiatedmodification of oxide semiconductor surfaces with rhodamine dyes. Illuminating a dye-label complex with UV light resulted in a reactive nitrene which formed an N-0 bond between dye and semiconductor stable for up to 1 year. Intrinsic excitation of the semiconductor was also found to bond rhodamine and xanthene dyes through an ester bridge with the carboxyl functions of the dyes. It was demonstrated that dyes could be attached to electrode surfaces in patterns of 2 5 - ~ mstripes. Work in chemical modification of solid surfaces has demonstrated that extensive control may be exercised over the characteristic reactivity of metal and semiconductor electrodes.1-12 Electrodes have been modified to improve stability against corrosion, to extend the range of their chemical reactivity, and to improve their catalytic prope r t i e ~ . ~Chemical J reagents have been attached to electrode surfaces in order to accelerate interfacial redox reactions and to control diffusion of species to electrode surface^.^ In addition, chemically modified electrode surfaces have been widely used in light-assisted redox reaction~~5 and in photoelectrochemical and electrochromic devi~es.~.' Sensitizing dyes form an important class of molecules used as surface modification agents, primarily at semiconductor electrodes, where a photocurrent can be produced through oxidation of the excited dye. They have (1)Murray, R. W. In Electroanalytical Chemistry; Bard, A. J., Ed.; Marcel Decker: New York, 1984;Vol. 13,p 191. (2)Chidsey, C. E.D., Murray, R. W. Science (Washington, 00 1986, 231, 25. (3)Wrighton, M. W. Science (Washington, DC) 1986,231, 32. (4)Abruna, H. D.;Calvert, J. M.; Denisevich, P.; Ellis, C. D.; Meyer, T. J.; Murphy, W. R.; Murray, R. W.; Walsh, J. L. ACS Symp. Ser. 1982, 192. 133. ---. (5) Murray, R . . Acc. Chem. Res. 1980,13,135.

(6)Reichman, B.;Fan, F. F.; Bard, A. J. J. Electrochem. Soc. 1980, 127,333. (7)Viehbeck, A.; DeBerry, D. J. Electrochem. SOC.1985,132,1369. (8) Moses, P.R.; Murray, R. W. J. Am. Chem. SOC.1976,98,7435. (9)H a m , D.;Armstrong, N. R. J. Phys. Chem. 1978,2,1288. (10)Fox, M. A.; Nobs, F. J.; Voynick, T. A.; J. Am. Chem. SOC.1980, 102,4036. (11)Fujihira, M.; Ohishi, N.; Osa, T. Nature (London) 1977,268,226. (12)Fujihira, M.; Kubota, T.;Osa, T. J.Ekctroanul. Chem. 1981,119, 379.

0743-7463/88/2404-0861$01.50/0

been attached to wide band gap metal oxides such as Ti02, Sn02,and SrTi03.8-12Early work in modification of these electrode surfaces with organic sensitizing dyes generally involved attaching an alkylsilane layer to the surface in a silyl ether linkage and then attaching the photosensitizer to the silyl layer by way of an alkyl chain. The oxidation of the excited dye leading to photocurrent under these conditions was, however, significantly lower than the photocurrent produced with dye adsorbed from an aqueous solution. Fujihira et a1."J2 attributed this difference to the increased distance of the attached dye molecule from the semiconductor surface and to the inhibition of electron transfer by the nonconductive,linking layer on the surface. Other authors have also discussed the extent to which the silyl layer hinders electron transfer to the underlying semiconductor.1° In an effort to improve the sensitization efficiency of the attached dye, Fujihira et al.ll developed a method of attaching rhodamine B directly to the semiconductor electrode surface through condensation of the carboxyl function on the dye with the hydroxylated semiconductor surface. With this ester linkage the sensitized photocurrent was found to be equal to that of adsorbed dye. A similar approach has been taken in the attachment of an inorganic sensitizer, ruthenium tris(bipyridyl), to metal oxide semiconductor electrode surfaces.13 However, esters are subject to hydrolysis in aqueous solution, and this ester linkage is hydrolyzed within hours of its formation. In this work we present a new method for electrode modification that leads to a stronger link between surface (13)Anderson, S.;Constable, E.; Dare-Edwards, M.; Goodenough, J.; Hamnett, A.; Seddon, K.; Wright, R. Nature (London) 1971,280,571.

0 1988 American Chemical Society

Ryan and Spitler

862 Langmuir, Vol. 4, No. 4, 1988

and sensitizer and a more stable modified electrode. We introduce the use of photoaffinity labels, such as are employed in biochemical studies, to the repertoire of electrode modification techniques. These photoaffinity labels are reagents with a functionality that can be reacted with an organic compound such as a dye and another function that becomes labile for a possible surface attachment upon irradiation with the appropriate energy light. The reagent employed in this work has an acyl bromide group, which can be easily reacted with the amine function of the rhodamine dyes, and an azide end, which becomes highly reactive as a nitrene generated through irradiation with 300-nm light. It is this nitrene which seeks the polar hydroxylated surface of an oxide semiconductor to form a stable N-0 link between dye and electrode. Similar techniques have been reported for applications in color proofing on silver halide surfaces.14 A light-initiated attachment technique such as this would overcome many limitations of other attachment techniques. The photoattachment with photoaffinity labeling can be accomplished in conditions open to the atmosphere, in aqueous or other solvents. In addition, the rate and degree of attachment may be controlled by adjusting the intensity of the actinic beam. Of greater significance is the capability to modify an electrode surface in patterns created by illumination of the surface where attachment is desired, with a resolution limited by the size of the actinic beam. The work described here is a study of light-initiated attachment of the organic dye rhodamine 6G to metal oxide semiconductor surfaces with the photoaffinity label p-azidophenacyl bromide as a linking agent. The sensitizer was shown to attach only where the surface was illuminated. Sintered zinc oxide was chosen as the electrode material for initial experimentation because of the availability and ease of preparation of electrodes. After it was determined that the photosensitizer could be attached to the sintered zinc oxide with the linking agent, single-crystal TiOz and ZnO were also modified by attaching rhodamine 6G. However, the energy of light needed to excite the azide to form a nitrene, the labile portion of the photoaffinity label, exceeds the band gap of the semiconductors so that photogenerated charge carriers are also produced. In the course of the experiments with photoattachment, it was found that photogenerated holes in the semiconductor can also initiate attachment of the rhodamine and related dyes covalently to the metal oxide semiconductor surface, although through a different mechanism and a different functional group.

Experimental Section Semiconductor Electrodes. Sintered ZnO disks were made by pressing ultrapure zinc oxide powder from Alfa Produds under 1ton of pressure for 10 min in an evacuated pellet press. They were then heated a t 1000 O C for 2 h.16 Single-crystal TiOz was prepared as electrode material by grinding and polishing the surface of the (001) face to a mirror finish with 1-pm grit and doping in flowing hydrogen a t 1000 "C, followed by an etch in hot concentrated H2SO4.l6 Single-crystal needles of ZnO were used as grown as electrode material. Ohmic contacts were made with indium-gallium eutectic and by attaching an indium wire with Tra-Con conductive silver epoxy. Electrodes made in this manner were characterized by electro(14)Ishida, T.; Kondo, S.; Toel, K. Ger. Offen, DE 3322 703 May 17, 1984;JP Appl. 82/UT 169841 Nov 11, 1982. (15)Inoue, T.;Fuiishima, A.; Honda, K. J. Electrochem. SOC.1980, 127,1582. (16)Cooper, G.;Turner, J. A.; Nozik, A. J. J. Electrochem. SOC.1982, 129, 1973.

cb

C UOCzH

PAD

CYoociH5

&cooFigure 1. Reagents used in attachment experiments. EOSIN

chemical and photoelectrochemical techniques which have been described previously." The band gap of the sintered ZnO was determined to be 3.25 (A.02)eV through the plot of (i - h t ~vs) ~hu from an intrinsic photocurrent action spectrum.18 The flat band potential of -0.50 A 0.02 V vs SCE found from Mott-Schottky plots was verified by the intercept of a plot of i2 vs V from an illuminated current-voltage curve.l8 These values compare well with published values for single-crystal ZnO.l9 The donor density as determined from the slope of the Mott-Schottky plot was found to be 5 X 101*/cm3. Dark current-voltage plots of unmodified electrodes showed little or no current flow in the potential region -0.3 to 1.0 V vs SCE, as expected for this n-type semiconductor. There was no change in the calculated flat band potential or donor density after dye or PAD attachment. Reagents. All reagents were of analytical grade purity and were used without further purification. Laser grade Rhodamine 6G (R6G) and laser grade Rhodamine B (RhB) were purchased from Exciton Chemical Co. Eosin Y was purchased from Aldrich. p-Azidophenacyl bromide (PA) was obtained from Pierce Chemical Co. of Rockford, IL, and 99% pure 1,3-dicyclohexylcarbodiimide (DCC) from Aldrich. Solvents used were spectrophotometric grade. All water used was deionized with a Barnstead ultrapure filter. The structures of these reagents are given in Figure 1. R6G was attached to the nonphotolabile end of PA by reacting a 0.1 mM solution of the dye with a 3.0 mM solution of PA in methanol. For routine use, the dye adduct, PAD, was separated from the reaction mixture by thin-layer chromatography by using a solvent of 35% ethanol and 65% acetone (v/v).'' Through reverse-phase HPLC of the reaction mixture, the labeling was found to go to completion with an observed retention time of 7.5 min for the labeled dye complex with no evidence of R6G, which was found to have a 7.0-min retention time. UV-vis absorption spectra of the adduct eluent from the column showed a one-to-one stoichiometry between the PA label and the chromophore, indicating that the label attached to only one of the amine functions of the R6G. This one-to-one relationship could be confirmed routinely through the absorption spectrum of an orange band eluted from the TLC plate, shown in Figure 2. With an extinction coefficient of 2.02 X lo4cm2/mol at 295 nm for PAz1 and 1.05 X lo6 cm2/mol a t 527 nm for R6G,22this solution was determined to be less than 10% dye and greater than 90% PAD. (17)Kavassalis, C.; Spitler, M. T. J. Phys. Chem. 1983,87, 3166. (18)Butler, M. A. J. Appl. Phys. 1977,48,1914. (19)Gerischer, H. In Topices in Applied Physics; Seraphin, B. O., Ed.; Springer-Verlag: New York, 1979;Vol. 31, pp 115-172. (20)Cramer, L.;Spears, K. G. J. Am. Chem. SOC.1978, 100, 221. (21)Hixson, S.;Hixson, S. Photochern. Photobiol. 1973,18, 135. (22)Drexhage, K. H.In Topics in Applied Physics; Schafer, F. P., Ed.; Springer Verlag: New York, 1977;Vol. 1, p 168.

Light-Initiated Surface Modification

c

25pM PA

A 4.8

-

0 I \ 1

,

1

I

___

ADSORBED DYE PAD

-L A S E R

N -

E

9 e

W

0 2

c

a

a

v)

P0

z 2.4

3 0

0

m

a

1

0.0

-

n \

0.0 I

300

I

I

400

I

I

450

550

650

500 WAVELENGTH (nm)

WAVELENGTH (nm)

Figure 2. The absorption spectrum of the PAD/R6G mixture is depicted as a solid line. The absorption of R6G and PAD differ in the region 280-320 nm where the absorption of R6G is given as a dotted line. The inset shows the spectrum of P A alone. PA alone was found in a different band on the TLC plate. 13C NMR spectra of the PAD revealed the appearance of a carbon in the amide region at a shift of 154 ppm, which was not present in the spectrum of R6G alone, showing that the label is bonded to an amine function. The label displaces a hydrogen on the amine, because it was not found pmible to label rhodamine B with any significant yield. Rhodamine B has the same structure as R6G, but with ethyl functions on the amine groups where R6G has hydrogen. Attachment Conditions. Before photoattachment of PAD or dye to sintered ZnO, the electrodes were rinsed in methanol and water, soaked in 4 M NaOH for 1h to hydroxylate the surface, rinsed in water, and air-dried. Without this soaking in base, the attachment to the electrodes was found to be very ineffective, revealing that the linkage requires a surface hydroxide function. For photoattachment, the electrodes were placed in the dye or PAD solution in a cell with an optically flat quartz window 1mm in front of the electrode and illuminated through the solution. The electrode was not potentiostatically controlled during illumination. All solutions were deoxygenated by bubbling nitrogen through them, and, in the case of photoattachment procedures lasting more than 4 h, nitrogen was bubbled through the solution thoughout the experiment. After photoattachment, the electrodes were removed from the solution, cleaned by immersion 3 times in alternating baths of boiling methanol and boiling acetone, rinsed in water, and airdried. Three sources of ultraviolet light were used for photoattachment: (a) the focused full output of a 150-W or 1000-W xenon lamp or a 200-W Hg-Xe lamp, (b) the 308-nm line of a XeCl excimer pulse laser with a pulse energy of 50-150 mJ and a 10-ns pulse, and (c) the 337-nm line of a nitrogen pulse laser with a pulse energy of 9 mJ and a 10-ns pulse. Photocurrent Action Spectra of Modified Electrodes. The photocurrent action spectra of the modified and unmodified electrodes were recorded as previously described." The electrolyte was 1.0 or 0.5 M KCl in deionized water. Hydroquinone (10 mM) was added to the electrolyte as a sacrificial electron donor or supersensitizerZ3to regenerate the photooxidized dye at the electrode surface. Only KC1 and hydroquinone were in the solution when action spectra of electrodes with photoattached dye were recorded. The electrode was biased at a potential of +.4 V vs SCE. Spatial Scanning Apparatus. A spatial scanning apparatus was set up to allow measurement of the effectiveness of localized attachment of PAD. In this arrangement, a spot or a stripe of monochromatic light is scanned across the electrode surface and

Figure 3. Photocurrent action spectra for adsorbed R6G, attached PAD, and W-attached R6G on sintered ZnO. The maxima for the attached species are seen to be blue shifted from that for the adsorbed dye. the photocurrent recorded as a function of the position of the beam on the elect.rode.% The 514.5-nm emission from an Ar' laser was focused on a 10-pm pinhole, expanded to a collimated beam, and then focused on the electrode surface to make a spot 10 pm in diameter. Electrodes were mounted on a flat piece of Teflon and attached to a cell fitted with a platinum wire counter electrode and a Ag/AgCl reference electrode. This cell was attached flat to a motorized table with the electrochemical face of the electrode up, and the motorized table was moved so that the spot scanned across the semiconductor electrode surface. A plot of photocurrent vs distance was recorded.

Results Photoattachment of PAD to Electrode Surfaces. The photocurrent action spectrum for an electrode placed in a PAD solution and irradiated with 290-310-nm light is similar in shape and magnitude to the action spectrum for a monolayer of R6G adsorbed to the sintered ZnO electrode. These spectra are shown in Figure 3. The similarity of these action spectra is evidence of the attachment of R6G to the ZnO surface through the linking agent PA. However, the photocurrent maximum for PAD attached to the surface is found at 529 f 4 nm, 10 nm blue-shifted from the photocurrent maximum for adsorbed dye. The magnitude of this current will be taken to reflect monolayer coverage of the electrode by PAD since increasing the PAD concentration in the reaction solution from lo4 to M did not increase the photocurrent over this level. The implication of this observation is that the efficiency of current production is the same for both attached and adsorbed dye. The result was the same with a methanolic or aqueous reaction mixture. A number of experiments were run to determine the illumination conditions necessary for monolayer attachment of PAD. With the full output of a 200-W Hg/Xe or 150-W Xe lamp, 8 h of continuous illumination were required to achieve monolayer coverage of attached PAD. The same quantity of PAD was attached to a sintered ZnO electrode by irradiation for 10 min with the 308-nm output of a XeCl excimer laser (150 mJ at 50 hz). Two hours of continuous illumination with a 1000-W Xe lamp attached a monolayer of PAD. Longer illuminations than these did not result in a greater coverage of PAD, which would have

~~~~

(23) Gerischer, H.; Tributsch, H. Ber. Bunsen-Ges. Phys. Chem. 1968, 72, 437.

(24) Furtak, T. E.; Canfield, D. C.; Parkinson, E.A. J. Appl. Phys. 1980,51, 6018.

864 Langmuir, Vol. 4, No. 4, 1988

Ryan and Spitler

I .o In

c a

n L

A DISTANCE

0 Y

.6

W

O \

[L

2 0

0

\0

LT

.2

0-

0

I-

t 0 I

z

w LK ~

a

IL 3

u

0

TIME

I

D ISTANCE

Figure 4. By use of a spot scanner, a 514-nm laser beam is scanned across a stripe of PAD attached to a sintered ZnO electrode. In a the spot is scanned across a single 25-wm-wide stripe. In b it traverses two such peaks designed to be 25 hum wide with a separation of 150 pm. been manifest as increased photocurrent. To establish that the PAD does indeed attach only where the surface has been illuminated, PAD was attached to a sintered ZnO surface in stripes 25 f 2 pm wide by directing the attaching beam through a 25-pm slit and focusing the image of the slit on the electrode surface. The photocurrent was then measured with the spot scanning apparatus. Figure 4a shows a plot of photocurrent vs distance for an electrode with one stripe of PAD attached. In Figure 4b is seen the response with two stripes of PAD attached. The plot shows both stripes to be 27 f 2 pm wide. The optical train of the exposure appratus was designed to yield a separation of 155 pm at the time of the photoattachment. The spot scanning plot of Figure 4 shows a peak-to-peak separation of 157 f 4 pm. The action spectra for adsorbed R6G and for PAD attached to single-crystalZnO and Ti02were also measured. Both single crystals were prepared by soaking in NaOH, 0.5 M for ZnO and 4 M for Ti02. The magnitude at the peaks of the action spectra of both PAD modified electrodes was about half that for the action spectra for dye adsorbed to the same electrodes. Both single crystals required twice the illumination time of sintered ZnO. Increasing the illumination time beyond twice that for sintered ZnO did not increase the magnitude of the action spectrum. UV Attachment of Dye to ZnO Surface. In control experiments it was found that photoattachment of dye to the semiconductor electrode surface can occur without the linking agent PA. When a sintered ZnO electrode was illuminated through a 1.0 X M solution of R6G in methanol, dye attached directly to the surface under illumination with UV light yielding the action spectrum shown in Figure 3. Its shape and magnitude are similar to those of the action spectrum obtained with PAD. Rhodamine 6G, rhodamine B, and eosin Y all attached by this technique, with either methanol or water as a solvent. The same illumination conditions established for the attachment of PAD were required to deposit a monolayer of dye on a sintered ZnO surface. It was also found that 2 h of 337-nm radiation from a nitrogen pulse laser resulted in attachment of a monolayer. When a 380-nm cutoff filter was placed in front of the irradiating beam, no dye attached, indicating that light more energetic than the ZnO

16

8

24

(hours)

Figure 5. The photocurrent from R6G UV attached to sintered ZnO is shown to decline over a 24-h period. The electrode was removed from electrolyte between measurements. band gap is necessary for dye attachment. This method of attachment is referred to as “UV attached dye” in this work. The competition between PAD attachment and attachment by UV irradiation of dye was quantified by using reaction mixtures of PAD and dye which were 0%, 12.5%, 25%, and 50% in PAD. I t was found through stability measurements (vide infra) that in reaction solutions more than 25% in PAD 100% of the photocurrent is caused by attached PAD. For a reaction solution 12.5% in PAD, 33% of the photocurrent is caused by PAD. UV attachment of dye was also effective on the surface of single-crystal ZnO and Ti02electrodes. However, the photocurrent produced with UV attached dye was substantially lower than photocurrent from adsorbed dye and was not increased by longer irradiation times. For Ti02 the photocurrent from attached dye was half that for adsorbed dye. For ZnO the photocurrent from attached dye was 5% of the photocurrent from adsorbed dye. It is the photogenerated hole which initiates the direct photoattachment of dye, as was concluded from the potential dependence of the UV attachment of dye to a sintered ZnO electrode. In a methanol solution of 0.10 mM R6G in 0.1 M LiC1, no dye was found to attach when the electrode was held at -0.75 V vs SCE, 0.25 V more negative than the flat band potential. At this bias holes will be drawn away from the surface. When the electrode was held at +0.4 V vs SCE, a bias which imposes a potential barrier to electrons traveling to the surface during illumination, the creation of dye-sensitized photocurrent showed that dye attached to the surface. Therefore, holes appear to be necessary for the direct attachment of dye to the semiconductor surface. The illumination time to attach dye with the electrode held at a positive bias is about half the time necessary at open circuit. Stability of Attached Species. A major difference between attached PAD and UV attached dye is the greater stability of PAD on the electrode surface. Photocurrent action spectra from a PAD modified electrode were measured 2 months and again 1year after attachment. Two months following attachment, the action spectrum was unchanged, but photocurrent had fallen to 60% of the original magnitude. After 1 year, the action spectrum still remained constant but the photocurrent magnitude had dropped to 20% of the original. The electrode had not been stored under special conditions; it was kept in a clear plexiglass case to keep out dust, but it was exposed to air and room light.

Langmuir, Vol. 4, No. 4, 1988 865

Light-Initiated Surface Modification Figure 5 shows the decrease in the magnitude of photocurrent from UV attached dye over a 24-h period. The initial magnitude of the photocurrent action spectrum for attached dye was the same as for atttirched PAD, but in contrast to PAD attachment, photocurrent from UV attached dye was not stable. After 24 h, the photocurrent action spectrum produced from dye attached to ZnO had fallen to 20% of its original magnitude. Dye-sensitized photocurrent was no longer produced 48 h after attachment. Covalent Attachment of Dye Using a Dehydrating Agent. For comparison of these photoattachment techniques with the method of Osa and Fujihira" for attaching dye to a semiconductor surface, R6G was attached to the sintered ZnO surface by using dicyclohexylcarbodiimide (DCC), a dehydrating agent which forms an ester linkage between the hydroxylated electrode surface and the carboxyl function of the dye. After soaking in 4 M NaOH, a sintered ZnO electrode was immersed in a 0.10 mM R6G and 2.5 mM DCC solution in deoxygenated acetonitrile for 90 min and then cleaned as described above. The photocurrent action spectrum which resulted from dye attached by DCC is similar in magnitude and shape to the action spectra from adsorbed dye, PAD-modified electrodes, and UV-attached dye-modified electrodes. There was no sensitized photocurrent with an electrode treated similarly but without DCC in the solution. In an action spectrum recorded 24 h after the original spectrum, the photocurrent had fallen to 50% of the original magnitude. After 48 h the photocurrent was 25% of the original magnitude, and after 72 h there was no longer a photocurrent action spectrum produced by the ester-linked dye. After each action spectrum, the electrode was rinsed in water, air-dried, and left open to the atmosphere between measurements. Formation of the N-0 Bond. Without electrode pretreatment with base, the photoattachment proceeds weakly. It is clear that the attachment must occur at the negatively charged oxygen sites created by the base. The PAD complex bonds differently with the surface than the R6G alone and with a greater bond strength. Therefore, this stronger bond must be formed with the label; the only function on the label capable of bonding is the nitrene created by decomposition of the azide group. We conclude that the mode of attachment is through the formation of an N-0 bond between a nitrene and a charged oxygen atom on the semiconductor surface. To support this contention, resonance Raman spectra were taken of s-ZnO electrodes with PAD-attached dye and with adsorbed dye. The spectra of PAD-derivatized ZnO electrodes showed a peak at 1015 cm-', which was absent in spectra of the electrode with adsorbed dye. This is within the range of vibrational absorption for a group in which the oxygen bridges a nitrogen and another atom. We attribute this vibrational peak to an -N-0-Zn link formed through the reaction of a nitrene and a surface oxygen site.

Discussion Photoattachment of PAD and of Dye. 1. Attachment of Dye with PAD. It was intended that the photoattachment of PAD to the semiconductor surface proceed through the photoexcitation of the azide to form a nitrene and the insertion of the nitrene into the OH bond on the surface. The nitrene is known to insert into a polar bond,25 which was created on these electrode surfaces through immersion in 4 M NaOH before treatment. In the dilute solutions of PAD and dye in methanol, the OH (25) Bayley, H.; Knowles, J. Methods Enzyrnol. 1977, 46, 70.

on the surface is the most polar bond in the environment of the nitrene with the knowledge that nitrene reacts preferentially by insertion into an OH or NH bond over alternative pathways such as intramolecular rearrangement or insertion into a CH bond.26 This appears to be the case. With an oxide surface pretreated with base, the attachment of the label occurs readily. The high stability of the PAD attachment can be attributed to the strength of the NO bond formed between the photoaffinity label and the oxide semiconductor,which is indicated by the Raman spectra. Unlike the bond linking the dye directly to the surface in UV-attached dye, the N-O bond appears to be attacked by water and air only at a very slow rate. A portion of the 80% decline in photocurrent over 1year may be caused by bleaching the dye in PAD rather than breaking the linking bond. Storage conditions that prevent exposure to light may give greater reproducibility of action spectrum magnitude. 2. UV Attachment of Dye. On the other hand, the UV attachment of dye to the semiconductor surface must proceed through the formation of an ester linkage between the carboxyl group on the dye and the OH on the surface. UV attachment of eosin Y demonstrates that the amine functions of the rhodamine dyes do not play a significant role in attachment since eosin Y has hydroxyl functions in place of the amines in its xanthene structure. This conclusion is further supported by the similarity of the stability characteristics of UV-attached R6G and DCC-attached R6G. Both types of dye attachment showed substantial decrease in photocurrent 24 h after attachment and within 48-72 h electrodes with UV-attached or DCCattached dye no longer produced sensitized photocurrent. The bond which holds UV-attached dye on the surface is not stable, and the photocurrent action spectrum falls to 20% of the original magnitude over a 24-h period, as shown in Figure 5. The ester linkage between the carboxyl group on the dye and the oxygen on the metal oxide surface apparently hydrolyzes when it is exposed to the atmosphere, with the metal-oxygen bond linking the dye to the solid acting as an acid catalyst for ester hydrolysis. However, in a water- and oxygen-free atmosphere the bond between dye and semiconductor surface is stable. The magnitude of the photocurrent action spectrum of an electrode kept in a drybox up to 1month did not decay. After these electrodes were taken out of the water- and oxygen-free atmosphere, their photocurrents decayed with the same dependence as other electrodes with UV-attached dye. The photocurrent action spectrum maximum of UVattached dye is the same as the maximum with attached PAD. These maxima are shifted 10 nm blue of that for adsorbed dye, indicating that the environment of the attached dye is different from the environment of adsorbed dye. On single-crystal electrodes, the absorption maximum of adsorbed dye and the action spectrum maximum are the same.27 As the photocurrent maximum for attached dye has shifted toward the solution absorption maximum, the dye appears to be in an environment similar to that of the solution. Corresponding shifts are also observed in the fluorescence spectra of adsorbed R6G and attached PAD on sintered ZnO. Attached PAD has a fluorescence maximum of 560 nm while adsorbed dye has a maximum of 545 nm. The dye in 0.5 M KC1 solution has a fluores(26) McRobbie, I. M.; Meth-Coh, 0.; Suschitzky,H. Tetrahedron Lett. 1976, 925; 1976, 929. (27) Natoli, L. M.; Ryan, M. A.; Spitler, M. T. J.Phys. Chen. 1985,

89,1448.

Ryan and Spitler

866 Langmuir, Vol. 4, No. 4, 1988

cence maximum of 555 nm and is broad, as is the attached PAD fluorescence spectrum. These shifts toward the solution maxima of absorption and fluorescence indicate that attached dye, whether PAD or UV attached, is not lying flat on the electrode surface, as the adsorbed dye,28but rather projects into the solution. In addition, the correspondence of the PAD action spectrum and the R6G absorption spectrum shows that the chromophore of the dye is unaffected by either the labeling or the surface attachment process. Loss of an ethyl group from the amine of R6G results in a spectral shift of 7-10 nm that is not seen here. Other forms of decomposition result in a complete bleaching of the chromophore. UV attachment of a compound to a semiconductor surface, accomplished by hole oxidation of the compound, has not been previously reported. The organic dyes used in these experiments contain a carboxyl function group attached to a benzene ring. It is this carboxyl group which is presumed to participate in the attachment. Carboxylic acids, including formic, oxalic, and benzoic acids, have been studied as possible agents for photo-Kolbe (or currentdoubling) reactions at metal oxide semiconductor surf a c e ~ . ~However, ~ - ~ ~ no current-doubling effect was observed with oxalic and benzoic acid. The work presented here implies that benzoic acid does react with a photoproduced hole but is not able to double the current because of the bond which is formed between the metal oxide surface and the carboxyl group. The hole plays the role of the acid in an acid-catalyzed transesterification reaction. In this work the chromophore of the dye serves as a tag to prove benzoic acid attachment to the surface and demonstrates that for benzoic acid the photo-Kolbe reaction does not go to completion. Chromophoric tags may then serve as a useful probe of the hole reactivity of currentdoubling reagents a t semiconductor electrodes. Yield of Photoattachment. Defining a yield of one as the case where every photon absorbed in the layer of molecules available for bonding attaches one molecule, the yield of photoattachment for both PAD and UV attachment is low. Within the 3-10-ps lifetime of the nitrene,32 excited PAD has a diffusion length of up to 100 nm, assuming the diffusion coefficient to be cm2/s. About 3 X lOI5 photons cm-2of 300-nm light were provided within one diffusion length of the surface in the total irradiation time. A monolayer of dye, 1 X 1014dye molecules/cm2,28 was attached to the electrode surface. This gives a yield of PAD attachment of 0.033. While the diffusion length of PAD within the nitrene lifetime is long, 100 nm, the nitrene is a very labile reactant and seeks to insert into a polar bond. Failing to find the OH at the surface, the nitrene may react with the solvent and preclude its bonding to the surface. In addition, any photons that are incident on the surface but do not result in attachment may, in fact, be responsible for corrosion at the surface, whereby surface atoms and possible attached species are lost to the solution unless the proper bias is maintained on the electrode. The yield for UV attachment is much lower than that of PAD attachment. Photons are not absorbed by the (28) Spitler, M. T.; Calvin, M. J . Chem. Phys. 1977, 67, 5193. (29) Morrison, S. R.Electrochemistry at Semiconductor and Oxidized Metal Surfaces; Plenum: New York, 1980. (30)Gornes, W. P.; Freund, T.; Morrison, S. R. J . Electrochem. SOC. 1968, 115, 818. (31)Kraeutler, B.; Bard, A. J. J. Am. Chem. Soc. 1978, 100, 4317. (32)Reiser, A.;Terry, G. C.; Willeta, F. W. Nature (London) 1966,211, 410. (33) Salvador, P. J. Phys. Chem. 1985,89, 3863.

reaction solution but by the electrode itself. With little or no absorption of the incident light by the solution, IOzo photons/cm2 are absorbed by the solid. With a concentration of attached dye of 1014molecules/cm2, the yield of UV attachment of dye to the semiconductor surface is 10-6. For UV-attached dye, it is the hole concentration at the surface which is the rate-limiting species. Recombination between the holes necessary for attachment and electrons will decrease the yield of UV attachment. This conclusion was supported by experiments in which dye attached at a greater rate when the electrode was held at a potential of +.4 V than at open circuit, a condition which decreases recombination. The data from the competition experiments between PAD- and UV-attached dye do not lend themselves to interpretation according to a simple kinetic analysis. It is likely that the relative yield of photoattachment in these experiments is a complicated function of the concurrent photocorrosion of the ZnO electrode. The predominance of the yield for PAD attachment may mean that this form of surface modification supresses photocorrosion,although no experiments were performed to explore this possibility. PAD and UV Attachment of Dye to Single-Crystal ZnO and Ti02. Sensitized photocurrents from dye UV attached to single-crystal ZnO and Ti02 are low, particularly with ZnO. It is reasonable to expect that the surface of a single crystal has many fewer surface and defect sites where holes might be trapped than sintered electrodes. In addition, single-crystal ZnO suffers from other conditions that could discourage photoattachment. It is possible that preparation conditions are not favorable to photoattachment of dye. Soaking for 1-2 h in 4 M NaOH was considered to be a condition too severe for the ZnO, which etches readily, so 0.5 NaOH was used in order to avoid damaging the single-crystal surface. All possible sites may not have been hydroxylated. PAD attached to a single-crystalzinc oxide electrode and gave a photocurrent magnitude about half that of adsorbed dye with about half the efficiency of its attachment rate to sintered ZnO. For PAD attachment, as with UV attachment of dye, the polycrystalline surface with its greater surface area and greater concentration of trapping sites appears to make a better attachment substrate than the single-crystal surface. With attachment about 10 times that of the UV-attached dye, PAD appears to be much less discriminating in its selection of polar surface sites to which it may attach.

Conclusion With the technique described here for photoattachment at semiconductor surfaces using a photoaffinity label linking agent, it is possible to control attachment in domains on a surface after one’s own design. As the diffusion length of PAD is 100 nm, the resolution of the patterns made on the surface by photoattachment will be limited by both the illuminating spot size and by diffusion. Therefore, it appears possible to produce domains by interference patterns of laser beams with the resolution limited by the diffusion length of 100 nm. Considering previous work in defined electrode surfaces,34the ability to modify a surface with a light-initiated technique is a powerful tool to have at hand. With recent advances in flash photolysis studies of ~ u r f a c e s , ~combination ~”~ of attachment of different re(34)Kittleson, G. p.; White, H. s.;Wrighton, M. S. J. Am. Chem. SOC.

1984,106, 7389.

Langmuir 1988,4,867-871 active agents to the surface in domains with spectroscopic studies will make it possible to examine photochemistry in the two dimensions of the semiconductor surface and to make time-resolved studies of electron transfer along this plane. (35) h f i n m d , P. A.; Camgrove, T.P.;Stmve,

w.s. J . phys, Chem.

1986,90,5887.

(36) Ryan, M. A.; Spitler, M. T., in preparation.

867

Acknowledgment. This work was funded by the Office of Basic Energy Sciences of the Department of Energy. We wish to thank Prof. D. Dooley of Amherst College for the use of the 1000-W Xe lamp. Reaistrv No. Rhodamine 6G, 989-388: Rhodamine B. 81-88-9: Eosin-Y, i7372-87-1; PAD,114595-93-6;DCC, 538-75-0; PA; 57018-46-9;ZnO, 1314-13-2;TiOz,13463-67-7;Nz, 7727-37-9; Oz, 7782-44-7.

In Situ Optical Structure Factor Measurements of an Aggregating Soot Aerosol H.X. Zhang, C. M. Soremen,* E. R. Ramer, B. J. Olivier, and J. F. Merklin Department of Physics, Kansas State University, Manhattan, Kansas 66506 Received December 1, 1987. I n Final Form: March 22, 1988 We study the agglomerate morphology of a carbonaceous soot aerosol extracted from a premixed methaneloxygen flame during combustion by the use of real-time optical structure factor measurements and subsequent transmission electron microscope examination of the settled agglomerates. We find a fractal morphology is maintained during the agglomeration growth of the aerosol with a fractal dimension D = 1.62 f 0.06. The TEM analysis also shows this fractal morphology with D = 1.72 f 0.10. These values of D are on the low side of the range D = 1.7-1.9 expected from diffusion-limitedcluster aggregation models.

Introduction The aggregation of small particles to form larger particles is a major growth process in colloidal and aerocolloidal systems.l In recent years interest in the clustering process has been heightened by the discovery2 that the random clusters formed during this process display a fractal morphology.M A fractal is a noncompact structure that displays a scale invariance symmetry! A consequence of this symmetry leads to a useful working definition of a fractal, which is that the number of monomers N inside a sphere of radius R centered on the cluster is given by N RD (1) In eq 1, D < 3 is the fractal dimension, the key parameter of the fractal structure. It has been found that D is a function of the spatial dimension3 and the details of the growth process.6 The first experimental work to demonstrate the fractal nature of agglomerates was performed by Forrest and Witten, who studied aerosol particles.2 In this paper we shall use the term "agglomerate" for a cluster of particles held together by weak van der Waals forces, whereas "aggregate" may be applied to clusters held together by partial fusion of the individual particle^.^ Since that time, considerably more experimental effort has been applied to colloidal liquids rather than aerosols. In most aerosols the initial monomer size is less than the mean free path of the gas molecules. Hence, the diffusion is no longer hydrodynamic and is therefore fundamentally different than in a colloidal liquid. Because of this and the obvious technological importance of aerosols, the morphology of aerosols particles should be studied. In the aerosol phase, in addition to the work of Forrest and Witten, studies have appeared regarding fumed silica redispersed either in a liquid or in packed p~wders.~bA variety of results for the

-

*Author to whom correspondence should be addressed.

0743-7463/88/2404-0867$01.50/0

fractal dimension D were obtained ranging from 1.8 to 2.6. Values larger than 1.8 were thought to be a result of the sample preparation method, e.g., the packing process, and thus not inherent to the coagulation process or possibly an indication of nonuniversality. Other aerosol work includes carbonaceous soot agglomerates, which have been collected and examined with a transmission electron microscope (TEM).9 These studies have shown soot agglomerates are fractals with D 1.7. Hence it seems reasonable to conclude that aerosol aggregates and agglomerates do display fractal morphology with D 1.7-1.8. None of the studies above, however, were performed in situ. Thus it is impossible to know the effects of sampling and collection on the resultant morphology. In this paper we study the agglomerate morphology of soot particles obtained from a methane/oxygen flame and captured in a 6-L flask. The morphology was studied both by collection of the resultant agglomerates for examination with a TEM and, most importantly, by in situ optical structure factor measurements. Optical structure factor measurements have been used before to determine the fractal nature of clusters.'JO We have also performed

-

-

(1) Drake, R. L. In Topics in Current Aerosol Research; Hidy, G. M., Brock, J. R., Us.; Pergamon: Oxford, 1972; Vol. 3, Part 2, p 201. (2) Forrest, S. R.; Witten, T. A. J. Phys. A 1979, 12, L109. (3) Kinetics of Aggregatical and Gelation, Family, F., Landau, D. P., Eds.; North Holland Amsterdam, 1984. (4) On Growth and Form, Stanley, H. E., Ostrowsky, N., Eds.;Nijhoff: Boston, 1986. (5) Mandelbrot, B. The Fractal Geometry of Nature; Freeman: San Fracisco, 1983. (6) Waitz, D. A.; Huang, J. S.; Lin, M. Y.; Sung, J. Phys. Rev. Lett. 1984,54,1416. (7) Martin, J. E.; Schaefer, D. W.; Hurd, A. J. Phys. Rev. A 1986,33,

__

3.5__. 41).

(8) Freltoft, T.; Kjems, J. K.; Sinha, S. K. Phys. Rev. B: Condens. Matter 1986,33, 269. (9) Samson,R. J.; Mulholland, G. W.; Gentry, J. W. Langmuir 1987, 3, 273.

1988 American Chemical Society