Detailed photocurrent spectroscopy of the semiconducting group VIB

Aleksander A. Tedstone , Edward A. Lewis , Nicky Savjani , Xiang Li Zhong , Sarah J. Haigh , Paul O'Brien , and David J. Lewis. Chemistry of Materials...
2 downloads 0 Views 565KB Size
J. Phys. Chem. 1982, 8 6 , 463-467

463

Detailed Photocurrent Spectroscopy of the Semiconducting Group V I Transition Metal Dichaicogenldes K. K. Kam and 6. A. Parklnclon' Ames Laboratwy' and Department of Physlcs, Iowa State Unhwslty, Ames, Iowa 5001 1 (Received: July 15, 1981; In Flnal Form: October 9, 1981)

The photocurrent spectra of single crystals of the semiconducting group VI transition metal dichalcogenides (MoS2,WS2, WSe2,and MoSe2)were measured as a function of crystal orientation and surface morphology as well as the polarization and the angle of incidence of the incident radiation. The spectra were analyzed with a band edge analysis revised to include the diffusion of carriers, and estimates of the transition energies were obtained. Structure corresponding to the excitonic transitions was also observed. The results were discussed with respect to the applicability of these materials to solar energy conversion.

Introduction The semiconducting transition metal dichalcogenides, which crystallize with a layered structure, have demonstrated respectable sunlight to electrical energy conversion efficiencies and remarkable stability to photocorrosion when employed in a photoelectrochemical cell.'+ In light of this fact, a systematic and detailed study of the wavelength-dependent photoresponse of a variety of these materials would provide useful information about how the photoresponse of these compounds matches the wavelength distribution of the solar spectrum. The photocurrent spectroscopy of these materials may also provide clues to the nature of the photoexcited transitions which has important consequences when the thickness, and thus the total amount of material needed in a device to absorb most of the solar radiation, is considered. We report an investigation of the effect of composition, orientation, and surface morphology of the crystal, the polarization, and the angle of incidence of the incident radiation as well as the solution composition and the electrode potential on the photocurrent spectra of WSe2, MoSe2, WS2, and MoS2. Experimental Section Single crystals of WSe2, MoSe2, WS2, and MoS2 were grown by chemical vapor transport with chlorine as described in ref 2 with the only difference being that the hot zone for the sulfides was a t 1100 "C. The crystals were cleaved with Scotch tape or mounted as grown to give the desired surface morphology. All crystals were determined to be the two-layer hexagonal polytype (2H) by X-ray diffraction and all were n type as determined by the direction of the photocurrent. Before being mounted, the crystals were cleaned with trichloroethane, acetone, and methanol and then rinsed thoroughly with triply distilled water. The ohmic back-contact was made by applying highly conducting silver paint (G. C. Electronics) between the crystal and a copper disk machined to fit in an 8-mm glass tube. The crystal was sealed with epoxy (Tra-Con) so that only the desired surface was exposed. Electrolyte was prepared by using reagent-grade chemicals (Fisher Scientific) as received and triply distilled water. Unless otherwise specified, the solution is 1 M NaI/O.l mM IF Photocurrent measurements were made in an electro-

* Address correspondence to this author at the following address: Solar Energy Research Institute, Golden, CO 80401. 'Operated for the US Department of Energy by Iowa State University under contract No. W-7405-Eng-82. 0022-3654/82/2086-0463~0~ .25/0

chemical cell under potentiostatic control. The light source was either a 10o0-W tungsten halogen lamp or a xenon arc lamp. The light was focused on the entrance slit of a high-intensity 0.25-m grating monochromator (Schoeffel) with interchangeable gratings. Both the entrance and exit slits were set at the same width no wider than 1.5 mm so that resolution better than 100 A in the near-infrared region and 50 A in the visible and ultraviolet region was obtained. Corning glass filters were used to remove the unwanted second-order radiation. A polarizing filter was used in some of the measurements to study the effect of a s- and p-polarized light on photocurrent in these materials. An electrochemicalcell constructed from Teflon with a quartz window was designed so that it could be rotated about a vertical axis which lies along the crystal surface so that the angle of incidence of the incoming radiation could be varied easily. Standard phase sensitive detection was used to measure the photocurrent signal from the potentiostat (Princeton PAR 174A) by chopping the light at 47 Hz. The spectral distribution of the lamp was measured either with a photoacoustic cell with a microphone (General Radio) or by a thermopile (Dexter Research Center). The agreement between the two methods was very good. A minicomputer (Digital Equipment Corp.) was used to control the experiment and average data at each wavelength to minimize noise.

Results and Discussion The fact that the surface morphology is an important factor in controlling the performance of a device constructed from the layered compounds was realized by several groups rather early in their studies of these systems.6s When the edge of a layered crystal is exposed to the electrolyte, the dangling bonds associated with the unsaturated transition metal or chalcogenide atoms introduce surface states within the band gap. These surface (1)H. Tributach, Sol. Energy Mater., 1, 257 (1979). (2) D. Canfield, K. K. Kam, G. Kline, and B. A. Parkinson, Sol. Energy Mater., 4,301,1981. (3) Fu-Ren F. Fan, Henry S. White, Bob Wheeler, and Allen J. Bard, J . Electrochem. Soc., 127,518 (1980). (4) Lynn F. Schneemeyer, Mark S. Wrighton, Angelica Stacy, and Michael J. Sienko, Appl. Phys. Lett., 36, 701 (1980). (5) W. Kautek, J. Gobrecht, and H. Gerischer, Ber. Bumenges. Phys. Chem., 84,1034 (1980). (6)H.J. Lewerenz, A. Heller, and F. J. DiSalvo, J. Am. Chem. SOC., 102. ~ .1877 - (1980). (7,H.. J. Lewerenz, A. Heller, H. J. Leamy, and S. D. Ferris, ACS Symp. Ser., No. 146,17 (1981). (8) D. Canfield, T. Furtak, and B. A. Parkinson, J. Appl. Phys., 61, 6018 (1980). \----,-

0 1982 American Chemical Society

464

-

k

The Journal of Physical Chemistry, Vol. 86,No. 4, 1982

0.5t

*

Kam and Parkinson

1

D

I

a 0 W

c1 t

0

I. I

-

1.0WAVELENGTH

I

o

0.3

0.8

0.7

(pm)

Flgurs 1. Normalized photocurrent spectra of WSe, In the near-infrared region: (0)sample with well-cleaved van der Waals surface exposed to the soiutlon; (A)sample with surface defects; (0)as grown crystal mounted with an edge exposed to the solutlon; (0) sample cut wlth a razor blade and mounted with the cut edge exposed to the solution.

levels act as potent recombination centers for photogenerated electrons and holes. The recombination processes, which severely degrade the efficiency of light to electrical energy conversion, have been studied with an electron beam induced current (EBIC)7technique and in situ with a scanning laser spot technique.8 We reported differences in the photocurrent spectra between a van der Waals surface and the edge of a WSez crystal? Figure 1shows a more detailed look at the effect of crystal orientation and surface morphology on the long-wavelength region of the photocurrent spectra for WSez crystals from the same growth ampule. The spectra have been normalized to better illustrate the relative contribution of photocurrent in the l.l-O.9-rm region. However, the actual quantum yields on the van der Waals surfaces are much greater than on the samples mounted with the edge exposed to the electrolyte, due to the greatly increased surface recombination velocity associated with exposed edges. The region of the spectra shown in Figure 1is the region where the absorption coefficient of WSez is relatively small but increasing so that the penetration depth of the light is decreasing as the photon energy increases. The relatively large penetration depth of the light at these wavelengths, when compared to the space charge layer thickness of the photoelectrodes, suggests that the diffusion of carriers to the space charge region will dominate the spectra in this region. The mobility of carriers in the layered semiconductors has been shown to be higher perpendicular to the c axis than along the c axis because of the structural anisotropy.1° With this fact in mind, one might expect to collect a relatively larger number of the carriers formed very deep in the sample with only the edge exposed to the electrolyte as is the case in Figure 1. An estimation of the space charge layer thickness when compared with the absorption depth of the light from optical measurements reveals that a significant number of carriers are created (9) D. Canfield, T. Furtak, K. K. Kam, G. Kline, and B. A. Parkinson, Faraday Discuss.Chem. SOC.,in press. (10)W. Y.Liang, J. Phys. C , 6 , 551 (1973).

\,

I

'

b

Y

Flgurs 2. Schematic diagram of the photoexcited e-h pairs and their subsequent motion when crystals at different surface morphology are illuminatedwith photon of energy close to the band edge: (a) e-h pah generated within the space charge region (SCR) will be separated by the space charge field; (b) canlers generated within the diffusion region diffuse toward the SCR; (c) carrlers generated below SCR diffusion re@n will eventually recombine; (d) a carrier generated below diffuskn regbn of illuminatedsurface may be within diffusion region of adjacent surface if step height is comparable to the diffusion length.

in the space charge layer only for wavelengths shorter than 700 nm. Figure 2 shows a model of a step on the surface which demonstrates the same phenomenon for a van der Waals surface with a high density of steps with heights on the order of the diffusion length. In Figure 2 the parallel lines represent the layers of the crystal with the solution at the top of the figure and the wavy lines represent the penetration depth for a photon with energy slightly above the indirect band edge. Electron-hole (e-h) pair a is created in the space charge layer (W) and is separated and collected. Electron-hole pair b is created within the diffusion region (LD) and can diffuse to the surface. Electron-hole pair c is created deep in the material and eventually recombines. A carrier formed very deep in the sample (d) can still be within the diffusion length of an adjoining van der WaaLS surface and can then be collected, yielding a larger relative long-wavelength response than an almost perfect van der Waals surface. In fact the larger diffusion length along the layer planes will assist this process. The sample in Figure 1 with the largest relative longwavelength response (and the lowest overall quantum yield) was cut with a razor blade along the c axis and illuminated on the cut edge. This procedure apparently produces a large number of damage states which extend into the band gap of the semiconductor and can then be excited with lower-energy photons. The same trends are observed with crystals of MoSez, but we were not able to repeat the same experiments with

The Journal of Physical Chemisrfy, Vol. 86, No. 4, 1982 465

Transition Metal DichalcogenMes

0.6

r

r

a

9 y0.4

e

-

a

3c

8

a

z 3

8

a

0

0.2

-

a2

a

I

a a

0.0 1100 io00

2 .o

c

900

800 700 600 WAVELENGTH (nm)

500

400

3x1

8 8

8

TI-

9&

Sdo

8bO 7& WAVELENGTH ( n m )

5bO

3b0

4&

..

I

n.2 0

n=1

.

n.2 o

n-1

0

I

. e

0

0

J

.

0.5

0.0

1

1.15

I

1.25

-".o;:,.'

I

1.35

1.45

1.55

hv (eV)

Figure 3. Top: Quantum yield spectra of WSe, with welkleaved van der Waals surface. Bottom: Plot of { [4 /( 1 - 4 ) ] ( hu))" vs. h u for n = '/*, 1, and 2.

WS2or MoSobecause the crystals of these compounds that we grew were very thin. Figures 3-6 show the wavelength dependence of the quantum yield for the four materials employed in this study. The photocurrent spectra are for samples with as perfect a van der Waals surface as we can obtain by careful cleaving of the sample before mounting. The samples whose spectra are shown exhibited very good junction properties such as low dark currents, high photovoltages and fill factors, properties which are characteristic of samples with a high degree of surface perfecti~n.l*~r' The quantum yields have not been corrected for the reflection losses from the shiny crystal surfaces because of the wavelength, polarization, and surface morphology dependencies of the reflection. Also, we would like the spectra that we measure to be indicative of the actual response that a cell would have in real sunlight. The reflection losses may be as high as 30-40%1° so the quantum yield per absorbed photon would approach 1at the spectral response maximum for all of the materials that we have studied. Index of refraction matching by the solution would not be expected to lower the reflection to below 20%. A feature which is evident in all of the spectra is several dips in the quantum yield which correspond to the positions of the A and B exciton absorption maxima at room

1.1

I

.2

1.3

1.4

1.5

h v (eV)

Figure 4. Same as Figure 3, but for MoSe,.

temperature-l1 The positions of the excitons have been indicated with an arrow in the figures, and in every case a small dip is observed. There are at least two possibilities to explain this observation. One is that the binding energy of the excitons is large enough that the electric field in the space charge layer is not sufficient to dissociate the excitons and so they eventually recombine.12 Another possibility is that the excitons are created very close to the surface, because of the high absorption coefficient associated with an excitonic transition, and thus are subject to a surface recombination process associated with less than perfect surfaces. We tend to favor the later explanation. The low quantum yields at 350 nm and the peak at 315 nm are the result of the absorption of these wavelengths by the photogenerated iodine at the semiconductor surface. The lamp spectra, for which the photocurrent spectra were corrected, were obtained with the same path length of the same electrolyte in a cell in front of our standard photodetector. Despite this precaution the extinction coefficient of 1, is high enough at these wavelengths to block much of the incoming radiation with even the small amounts of photogenerated.,I The peak in the quantum yield at 315 (11)J. A. Wilson and A. D.Yoffe, Adu. Phys., 18, 193 (1969). (12)A. R.Bed and W.Y.Liang,J.Phys. C,9, 2459 (1976).

466

Kam and Parkinson

The Journal of Physical Chemistry, Vol. 86, No. 4, 1982

I M N a 1 l O . I mM I 2 o 1M NaBr/O.2mM Br2

#

fs

a

2

ff

*a"

?a

a a

z

.

a

"0.

f

2a2t

i

0 .

.

WAVELENGTH ( n m i

WAVELENGTV I

I

( nm) I

1

I

1.6

1.7

1.8

n.2 o

x4

n= I

.

d

a

1.3

1.4

I .5

I .6

1.7

1.8

1.9

hv (ev) U.V

Flgwr 5. Same as Figure 3, but for WS,: (0)wlth 1 M NaIlO.1 mM I, solution; (0) with 1 M NaBr10.2 mM Br, solutlon.

nm corresponds to a dip in the I< absorption spectrum.13 If bromide is used in place of iodide, the spectrum remains flat to 300 nm (Figure 5 ) . We chose to measure most of the spectra with the iodide solutions however because these solutions have demonstrated the highest solar to electrical conversion' efficiencies despite the high ultraviolet absorption. The plots shown below the respective photocurrent spectra are an analysis of the band edge for the various materials based on an approach used by Butler14 and other workersls but modified to include the fact that the carrier diffusion length and the light penetration depth may be of the same order in this region of the spectra. The equation used to fit the band edge is derived from the photocurrent equation for a given wavelength16

1 - exp(-aw)

+

ffLP exp(-aw) aL, + 1

1

1.2

'.3

1.5

1.4 hv

1.9

ieV)

Flgure 6. Same as Flgure 3, but for MoS,: ( 0 )for sample with well-cleaved surface; (0) for sample with a few surface defects.

where the total photocurrent (Jbd)is the s u m of the space charge generated current (JscR) and the diffusion current (JDIF). R is the reflectivity of the semiconductor, q is the elementary charge, Io is the incident photon flux,a is the absorption coefficient, Lp is the minority carrier diffusion length, and w is the space charge layer width. For the transition metal dichalcogenides used in the study, a w is close to zero in the region of the photocurrent onset. With this condition eq 2 simplifies t o Jwbl = qIo(1 - R ) [aLp/(aLp + 111 (3) The quantum yield (a) is then given by (4)

(2)

(13) Trammimion measurements of path length 1.5 mm of solution containing different concentrations of Is- were obtained with a Cary spectrometer. (14) M.L. Butler, J.Appl. Phys., 48,1914 (1977). (15) W.Kautek, H.Gerischer, and H. Tributsch, J.Electrochem. Soc., 127, 2471 (1980).

Solving for a and neglecting the reflection loss, we obtain

.=(")'1 - a

L,

(16) W.W.Gartner, Phys. Reu., 116, 84 (1969). (17) J. I. Pankove, 'Optical Processes in Semiconductors", Dover Publications, New York, 1971.

The Journal of Physical Chemistty, Vol. 86, No. 4, 1982 467

Transition Metal Dlchalcogenides

TABLE I: Summary of Band Gaps Measured for Various Compounds in This Studya EdiD

E ~ ~ z D

WSe. MoSk,

1.35 1.39 1.20 1.35 1.38 1.09 1.74 1.79 1.35 ws 2 MoS, 1.69 1.74 1.23 a Direct and indirect band-gap energy (eV) of WSe,, MoSe,, WS,, and MoS, determined from the intercept of the graphs of [ hu@/(1 - @)Isn vs. !w. The superscript 2D (3D)means two- (three-) dimensional material, and the subscript dg (ig) means direct (indirect) band gap.

Near the absorption edge of a semiconductor the absorption coefficient is given byI7 (Y

= A(hv - E,)*/hv

(6)

where A is a constant, Eg is the band gap of the semiconductor, and the exponent m is determined by the nature and dimensionality of the transition. Combining eq 5 and 6, we obtain

where n = l / m . The value of n is for an indirect transition in a three-dimensional material, and for a direct transition a value of 1 or 2 is expected for a two- or three-dimensional material, respectively.18 A plot of the left side of eq 7 with the various n values against the photon energy should yield a straight line with a slope proportional to the diffusion length of minority carriers and an x-axis intercept equal to the band-gap energy for the transition. As is evident from the figures, straight-line portions can be seen over selected energy ranges for virtually all of the n values. Table I lists some values obtained by fitting straight lines to determine the direct and indirect transition energies for the materials that we have studied. Care should be taken that the values in Table I are not taken as absolutes because of the many problems which could plague the treatment described above. The major problem is due to the large variations in current voltage behavior and spectral response from sample to sample. The surface morphology differences are obvious and can lead to behavior described earlier, but bulk properties of the crystals can also vary widely. Unlike most of the technologically important semicondudors, with which most of the previous photoelectrochemical studies have been done, the transition metal dichalcogenide crystals are not cut from a single boule, and thus each crystal has nucleated and grown independently. The types of variations from crystal to crystal that can be expected with these and similar compounds are (1)deviations from stoichiometry, (2) stacking faults, screw dislocations and polytypism, and (3) variations in the type and homogeneity of impurities. The impurities can have a major affect on the space charge layer thickness and the minority carrier diffusion lengths but also could induce impurity absorption which could severely effect the low quantum yields associated with the indirect transitions. Indeed, we can detect small photocurrents (@ = 4 X 10"') at wavelengths as long as 1.3 pm with most of the samples that we have studied. There may also .be problems in applying the band edge analysis developed for three-dimensional materials with s or p tran(18) P.A.Lee, Ed.,"Physics and Chemietry of Materiale with Layered Structures", Vol. 4, Reidel, Dordrecht, Holland, 1976.

sitions to a pseudo-two-dimensional system with d-d transitions. It is unclear whether the parabolic band assumption which is used in the three-dimensional materials is valid for these systems.lg The angle of incidence and the polarization of the incident light were varied in hopes that the change in the electric vector of the incoming radiation would couple the radiation more or less to the electronic transitions in the layered compound. Gerischer has observed an effect of the polarization and the angle of incidence on the photocurrent spectxa of GazSePm The angular dependence was studied for p-polarized light with a normalization to spolarized light at normal incidence for several doping levels in the semiconductor. The large angular dependences that were observed were explained by a large anisotropy in the absorption coefficient and by different diffusion lengths associated with the impurity doping. We were unable to measure significant changes in the photocurrent spectra as a function of the polarization or angle of incidence of the incoming photons for either faces or edges of crystals of layered materials, used in this study. The small differences that were observed could be accounted for by the changes in illuminated area and the polarization-dependent reflectivity of our materials.10v21 Changes in the electrode potential had little effect on the shape of the photocurrent spectra, but the quantum yield decreased as the electrode potential was made more negative. This behavior is expected as the electrode approaches the flat-band condition. In conclusion the photocurrent spectra of these semiconductors can provide information about their surface morphology and transport properties, thus allowing one to determine how the photoresponse of the material matches the solar spectra. Our results suggest that the diselenides would be more efficient in a solar cell, a prediction which is verified by solar efficiency measurements.2*22Another interesting finding is that structure due to excitonic transitions is visible in the photocurrent spectra. Low-temperature optical measurements are underway to provide a better understanding of the solid-state optical transitions within the semiconductor and how these transitions influence the junction formed at the solidelectrolyte interface. The transport properties of the layered crystals are also extremely important to the photocurrent yields especially in the long-wavelength regions of the spectra. Independent transport measurements are underway to determine the transport mechanism, the carrier diffusion lengths, and the carrier lifetimes. It is obvious that not all of the impurities present in these semiconductors contribute to states within the band gap which limit the diffusion 1engths.O If the specific impurities which contribute to the bulk recombination process could be identified and eliminated, the production of crystals with long diffusion lengths and thus high solar to electrical conversion efficiencies would be facilitated. Acknowledgment. The crystals used in this study were grown by Gary Kline. Many helpful discussions with and suggestions from Dave Lynch are greatly appreciated. Mike Butler also provided helpful insights. This research was supported by the Solar Energy Research Institute under contract No. XP-9-8198-1. (19) M.Butler, private communication. (20) H. Gerischer, ACS Symp. Ser., No.146,l (1981). (21)K. Saiki, M. Yoshimi, and S. Tanaka,Phys. Status Solidi B, 88, 607 (1978). (22) K. K. Kam,G. Kline, B. A. Parkinson, under preparation.