Surface concentrations of indium, phosphorus and oxygen in indium

J. A. Leavitt and L. C. McIntyre, Jr. Department of Physics, University of Arizona,Tucson, Arizona 85721. The surface of Indium phosphide single cryst...
1 downloads 0 Views 600KB Size
Anal. Chem. 1992, 64, 2929-2933

2929

Surface Concentrations of Indium, Phosphorus, and Oxygen in Indium Phosphide Single Crystals after Exposure to Gamble Solution Tami B. Dittmar and Quintus Fernando' Department of Chemistry, University of Arizona, Tucson, Arizona 85721

J. A. Leavitt and L. C. McIntyre, Jr. Department of Physics, University of Arizona, Tucson, Arizona 85721

The surface of lndlum phoqhlde single crystals exposed to synthetic lung fluid (Gamble rolutlon) has been Investigated. An oxygen depth proflle, obtalned by employlng the 3.034MeV resonance In the elastk scattering of CY partkles from loo,detected oxygen at maxlmum concentratlonsof 24% In a layer 1000 A thick. O I n and O P ratloe were found to Increasewlth prd0ng.dexpoeuretbne as detennlnedby angleresolved XPS, although Mndlng energles and peak fwhm's remained relatively constant. The surface concentration of lndlun was found to be decreased In thlr 1000-&thlck layer. Ina+was detected at parts per Mlllon kvelr In the Gamble sdutlon, conflrmlng that lndlum was leached from the InP surface. The sbnllarlty of P I n ratlos at all lengthsof exposure suggests that phosphorus was also leached by the Gamble sdutlon.

INTRODUCTION Indium phosphide is an important 111-V semiconductor for the fabrication of a variety of semiconductor devices. The properties of these devices are determined to a large extent by the nature of the dielectric oxide films that are present on the surface of the indium phosphide. Numerous studies have been undertaken in order to understand the factors that influence the formation of the surface oxide films and to characterize these oxides.'-' The surface oxide composition was found to vary within wide limits. Upon exposure to air, a nonhomogeneous surface oxide layer was formed and has been reported by various investigators to consist of one or more of the following components: In& In(OH)3, and InPOr.lo The growth of native oxides on 111-V semiconductor surfaces is, therefore, a complex process which is influenced by the differences in the physical properties of the oxides of the 111and V components. The formation of surface oxides may also have important consequences in the workplace where particulates are formed during the cutting, grinding, and polishing of semiconductor wafers. Inhalation of these particulates may be an occupa-

* To whom inquiries should be addressed.

(1) Hofmann, A.;Streubel, P.; Miesel, A. Surf. Interface Anal. 1988, 12,315-319. (2)Hoflund, G.B.;Corallo, C. F. Surf. Interface Anal. 1986,10,319323. (3)Bergignat, E.; Hollinger, G.; Robach, Y. Surf. Sci.1987,189/90, 353-361. (4)Wilmsen, C. W.; Geib, K. M.; Gann, R.;Costello, J.; Hryckowian, G.; %to, R. J. J. Vac. Sci. Technol. 1985,3,1103-1106. (5)Franz,G. J. Appl. Phys. 1988,63,500-505. (6)Korotchenkov, G.S.;Nikhailov, V. A.; Tsvitainsky, V. I. J. Vac. Sei. Technol. 1986,B3,981-984. (7)W i h e n , C. W.; Kee, R. W. J. Vac. Sci. Technol. 1978,15,15131517. 0003-2700/92/0364-2929$03.00/0

tional hazard that should be evaluated.ca We have previously reported that inhalation of gallium arsenide particulates is a potential health hazard.u*26We arrived at this conclusion by measurement of the rate of dissolution of GaAs in synthetic lung fluid (Gamble solution) and by a parallel determination of the progressive change in the surface composition of GaAs by X-ray photoelectron spectroscopy (XPS) and by Rutherford back-scattering spectrometry (RBS). We report here a similar study with indium phosphide; single crystals of indium phosphide were immersed in Gamble solution and the indium phosphide surface was examined at periodic intervals by X P S and RBS. Our expectationwas that surface oxide formation would occur more readily when the indium phosphide surface was in contact with the Gamble solution and that oxide formation would result in the formation of a surface layer of the oxides of phosphorus and indium. Because of the relative insolubility of the indium oxide in the Gamble solution, the surface of the indium phosphide crystal should, therefore, show a depletion in phosphorus. This expectation can be c o n f i i e d experimentallyby the determination of the surface concentrationof indium, phosphorus, and oxygen. In this study we have exposed indium phosphide crystals to synthetic lung fluid (Gamble solution) for a maximum of 8 weeks. The gross changes in the indium phosphide surface have been deduced by RBS, and the oxygen depth profile has (8) NIEHS. Executive Summary of Data: Indium Phosphide; Contract No. N01-ES-85219;NIEHS Reeearch Triangle Park, NC, 1988. (9)Fowler, B. A. In Handbook on the Toxicology of Metale, 2nd ed.; Freiberg, L.,et al., Eda.;Elsevier Science Publishers: New York, l9W, pp 267-275. (10)Brakhnova, I. T. Modif. Siluminov 1970,174-178. (11)Brakhnova, I. T. Poluch Suoistva A i m e n Fosfidou 1977,96-99. (12)Roshchina, T. A. Gig. Tr.Prof. Zabol. 1966,10,3Cb33. (13)Stokinger, H. F. In Patty's Industrial Hygiene and Toxicology, 3rd ed.;Clayton, G. D., Clayton,F. E., Eda.;John Wiley and Sone: New York, 1981;pp 1654-1661. (14)Fadeev, A. I. Gig. Tr. Prof. Zabol. 1980,3,45-47. (15)Fadeev, A. 1.Met.: Gig.Aspekty Otsenki Ozodorovleniya Okruh. Sredy 1983, 246-251. (16)Tarawnko, N. Y.;Fadeev, A. 1. Gig. Sanit. 1980,45,13-16. (17)Leach, L. J.; Scott, J. K.; Steadmen, L. T.; Maynard, E. A.; Armstrong, R. D. Uniu. Rochester At. Energy Proj. Rep. No. UR-690, 1961. (18)Podosinovekij, V. V. Gig. Sanit. 1965,30,28-34. (19)Morrow,P. E.;Gibb, F. R.;Cloutier, R.; Caaarett, L. J.; Scott, J. K.Univ. Rochester At. Energy Proj. Rep. No. UR-508, 1958. (20)Castronovo, F. P.;Wagner, H. N. Br. J. Exp. Pathol. 1971,52, 543-559. (21)Smith, G.A.;Thomas, R. G.; Scott, J. K. Health Phys. 1960,4, 101-108. (22)Smith,G.A.;Thomaa,R.G.;Black,B.;Scott, J.K. Univ.Rochester At. Energy Rep. No. UR-500,1957. (23)Isitman, A. T.;Manoli, R.;Schnidt, G. H.; Holmes, R. A. Am. J. Roentgenol., Radium Ther. N u l . Med. 1974,120,716-781. (24) Pierson, B.; Van Wagenen, S.; Nebemy, K. W.; F e m d o , Q.; Scott, N.; Carter, D. E. Am. Znd. Hyg. Assoc. J. 1989,50,455-469. (25)Webb, D.R.;Sipea, I. G.; Carter, D. E. Toxicol.Appl. Pharmacol. 1984, 76,96-104. Q 1992 American Chemical Society

2930 ANALYTICAL CHEMISTRY, VOL. 64, NO. 23, DECEMBER 1, 1992

been obtained by using the 3.034-MeV resonance in the elastic scattering of a particles from lSO. The 0:In and 0:P ratios in the near-surface and subsurface regions have been determined by angle-resolved XPS. Finally, graphite furnace atomic absorption spectrophotometry (GFAA) was used to measure the concentration of indium that was leached by the Gamble solution during the 8-week period.

EXPERIMENTAL SECTION Indium Phosphide Single Crystals. An indium phosphide singlecrystal was obtained from AT&Tas a 5-cm-diameterwafer cut at (100) 6' (110) with a measured density of 1.9 X 1On atoms/cm3. No attempt was made to clean or modify the native oxide layer present on the single crystal surface. The InP singlecrystal wafer was broken into nine smaller chips, each approximately 1 cm2,for the solubility study. Synthetic LungFluid (Gamble Solution). Gamblesolution was prepared according to the procedure published by Diem.2B The componentsof the Gamble solution were added sequentially with stirring to approximately 900 mL of distilled, deionized water (DDI). A 95% C02/5% 02 v/v gas mixture was bubbled through this solution at 150 mL/min for 1h. The solution was filtered f i i t through an 8-rm fiiter followed by a 0.45-pm filter to remove any undissolved particles, and the filters were washed with three 1-mL aliquots of DDI. At this stage, the pH of the Gamble solution was 7 a the pH was raised to 7.41 with 1 M NaOH in order to approximate the physiological pH and then diluted to 1.00 L with DDI. Solubility Measurements. The reaction vessels consisted of 125-mL wide-mouth Nalgene bottles fitted with rubber stoppers through which gas inlet and outlet tubes were inserted. A plastic bubbler was placed at the end of the inlet tube to provide an even distribution of gas over the surface of the solution. The reaction vessels were covered with foil to exclude light and were secured upright in a water bath maintained at 37 OC. Sufficient Gamble solution was added to each vessel with a buret to give a ratio between solution volume and InP surface area of 45 cms to 1om2. Each InP chip was wrapped loosely in thin cotton gauze and suspended with Nichrome wire from the bottom of the rubber stopper; then each stopper was wrapped with Teflon tape and placed tightly into the mouth of a reaction vessel. The gas inlets were adjusted to provide a gentle agitation of the solution surface by the 95% Cod5% 02gas mixture which was bubbled at 120 mL/min through the solutions for 30 min each day. Eight InP chips were exposed to Gamble solution; the ninth chip was immediately stored in a desiccator for use as a blank. One chip was removed weekly from the solution, washed for 30 s with DDI, dried with Nz,and stored in a desiccator for analysis by XPS and RBS. The reaction vessels containing the Gamble solution were sealed and frozen until analysis for indium by GFAA. A blank containing only Gamble solution was also subjected to the same environment; the pH of this solution was measured weekly to ensure that there was no fluctuation in pH during the course of the study. Surface Analysis of the I n P Single Crystals. X-ray Photoelectron Spectroscopy (XPS). The InP single-crystal surfaceswere analyzedby X-ray photoelectron spectroscopywith a Vacuum Generators ESCAlab MK I1 photoelectron spectrometer. The incident Al Ka X-rays (1486.6 eV) were provided by an Al source operated at 300 W. The emitted electrons were detected at a pass energy of 50 eV, and the spectrum was scanned repetitively to improve the signal-to-noise ratio. The carbon 1s peak at 284.6 eV was used as the reference for energy calibration.n The atomic density ratios were calculated from the scattering cross sections of indium, phosphorus, and oxygen by

-

where N ~ N isBthe atomic density ratio, ZJZB is the integrated (26)Diem, K.,Lenter, C., Eds. Documenta Geigy Scientific Tables, 7th ed.; Ciba-Geigy Ltd.: Basle, Switzerland, 1970; p 523. (27) Muilenberg, G. E., Ed.Handbook of X-ray Photoelectron Spectroscopy; Perkin Elmer Corp.: Eden Prarie, MN, 1979.

peak intensity ratio, u is the photoionization probability, and AM is the inelastic mean free path at E,,, the kinetic energy of the detected electrons.28 No attempt was made to determine the true areas of the peaks. Angle-resolved XPS was performed at 8 = loo and at 80° where 8 is the angle formed between the detector axis and the plane of the sample surface. Considering the effective inelastic mean free paths, these takeoff angles correspond to electrons emitted from depths below the surface of the InP single crystal of approximately 10 and 50 A, respectively.2B Rutherford Back-Scattering Spectrometry. Back-scattering analysis of the indium phosphide near-surface region was performed using a 6-MV Van de Graaff accelerator (HighVoltage Engineering Corp.). The energies of the back-scattered 4He+ ions, which were normally incident on the sample were measured at a back-scattering angle of 170.5O by a silicon surface barrier detector that subtended a solid angle of 0.78 msr and had a resolution of 15 keV. Depth profiling of oxygen was performed using the 3.034-MeV resonance in the inelastic scattering of a particles from l60. Computer simulation was used for quantitative determination of oxygen content in these f h s . A description of the elemental composition of the target as well as experimental parameters such as detector solidangle,number of incident particles (derived from the collected charge), and ' H e 4 scattering cross sections"J was information that was input to the program. The program then divided the sample into layers and calculated the backscattering yield from oxygenas a function of incident beam energy. This calculation was absolute and did not require a standard for calibration. The layers were chosento be much thinner (inenergy loss) than the oxygen resonance width; for the InP samples analyzed in this study, 10 layers, each about 2.5 keV thick, were found to be sufficient. The program was checked against two A l 2 0 3 films with known thicknesses and was found to correctly simulate measured yield curves. Determination of Indium in Gamble Solution. The synthetic lung fluid was analyzed for indium by graphite furnace absorption spectroscopy. The frozen samples were thawed and brought to room temperature, and the volume was measured to assure that no loss occurred during the course of the experiment. Due to the complex composition of synthetic lung fluid, it was necessary to separate the In3+species from the Gamble solution before analysis by GFAA. This was accomplished by f i t digesting the samples with 2 mL of concentrated Has04 and hydrogen peroxide which was added dropwise throughout the digestion. In3+was extracted as an ion pair of InC&-FtJV+where &N+ is the methyltricaprylammonium ion (Aliquat 336) into a mixture of hexane and methyl isobutyl ketone. The indium was then back-extracted into the aqueous phase with a 5 5 9 0 vJv/v mixture of HNO~:CH~COOH:HZO?~ A calibration curve was obtained with indium standards prepared in Gamble solution and subjected to the same digestion and extraction procedure. An aliquot of a solution containing lo00 ppm each of PdClz and NHSIZPOIwas added as a matrix modifier to each of the extracted standards and samples. The optimum matrix modifier concentration in the sampleswas found to be 250 ppb. A 50-pL aliquot of each sample was pipetted onto a graphite microboat; the microboata were placed on a Tefloncoated drying tray and evaporated to dryness on a hot plate. The drying temperature was carefully maintained as low as possible in order to prevent sample loss by spattering. GFAA analysis was performed on a Thermo Jarrel Ash Smith Hieftje 12spectrometer equipped with a ThermoJarrel Ash CTF 188 controlled-temperature furnace atomizer fitted with a rectangular furnace cuvette that accommodated pyrolytic graphite microboat sample holders (Nos. 124544-04 and 041629-01). (28)Nebesny, K.W.; Maschhoff, B. L.; Armstrong,N. R.A w l . Chem. 1989,61,469A-MA. (29)Briggs, D.,Seah, M. P., Eda. Practical Surface Awlysia by X-ray and Photoelectron Spectroscopy; Wiley: New York, 1983. (30) Leavitt, J. A.;McIntyre L. C., Jr.; Ashbaugh, M. D.; Oder, J. G.; Lin, Z.; Dezfouly-Arjomandy,B. Nucl. Znstrum. Methods 1990,B44,26065. (31)Scott, N.;Carter, D. E.; Femando, Q. Anul. Chem. 1987,59,888890.

ANALYTICAL CHEMISTRY, VOL. 64, NO. 23, DECEMBER 1, 1992

4000

2031

30t

I\ t

'Oo0

0 0

200

400

600

800

1000

a

i

750

800

850

3020

-

3040

3060

3080

Flgure 2. Measured 'H+O laboratory cross section near the 3.034MeV resonance at a scattering angle of 170.5'.

0

0 700

3000

E o ~ o a( k e " )

Channel Number

500

2980

900

950

Channel Number

Flgure 1. Spectra of l.9-MeV 'lie+ Ions back-scattered from Indium phosphide samples exposed to Gamble solution for 0 (-) and 6 (-) weeks. The sample normalwas set 7O away from the beam axis, and the sample was rotated about the normal direction to avow channeling effects. The full spectrum (a) shows indium and phosphorus slgnals near channels 901 and 617, respectively;the oxygen signal expected near channel 419 is burled In the background Continuum. Spectrum b shows an enlargement of the Indium signal near channel 901; the &week sample has a depletion or dilutlon of indium In the near-surface region as indicated by the notch in the spectrum.

RESULTS AND DISCUSSION Back-Scattering Spectrometry. The superimposed random back-scattered spectra of 1.9-MeV 4He+ions on the unexposed indium phosphide chip and the chip exposed to synthetic lung fluid for 6 weeks are shown in Figure 1. The indium and phosphorus signals are evident beginning near channels 901 and 617, respectively; no oxygen step near channel 419 is apparent in either sample owing to the low sensitivity of back-scattering analysis for detecting low Z elements. The back-scattered spectra of the unexposed InP sample is typical for a sample of infinite thickness; the signals for indium (channel 901) and phosphorus (channel 617) appear as steps with heights proportional to atomic number and are superimposed on a broad continuum covering the energy range scanned. In contrast, the 6-week sample shows a decrease in the indium signal which indicates that a depletion or dilution of indium has occurred in the near-surface region. Due to the weak phosphorus signals, no definite statement can be made about the concentration of this element. However, upon enlargement of the phosphorus peak at channel 617, it does appear that the phosphorus signal has become somewhat more rounded; this discontinuity suggests that the phosphorus atoms are also depleted from the surface. Conventional RBS is inherently insensitive to elements of low atomic number, such as oxygen. The basic problem with

using 4He+back-scattering for detecting oxygen in a matrix containing heavy elements (e.g. indium) is that the oxygen signal is superimposed on a high count rate caused by scattering from the heavy element matrix. In almost all cases for the InP samples considered, the oxygen back-scattering signal would not be detectable at this 4He+energy (1.9 MeV); however, the sensitivity of back-scattering spectrometry to oxygen can be greatly enhanced when a resonance at 3.034 MeV in the elastic scattering of 4He+by lSO is employed. This resonance is approximately 10 keV wide and has a peak back-scattering cross section about 25 times greater thanthe Rutherford value at that energy, as shown in Figure 2.S' Resonanceback-scattering is therefore a sensitive method of obtaining the oxygen profile in the near-surface region of indium phosphide. An example of a back-scattering spectrum at an offresonance energy (3015keV) is shown for the sample with the highest oxygen content (the sample exposed for 8 weeks) in Figure 3a. Figure 3b shows the corresponding spectrum at 3040 keV, which is just above the resonance energy. The small peak just below channel 300 is caused by scattering of 4He+by oxygen near the samplesurface. The scattering from surface In and P appears at approximatelychannels 700 and 480, respectively. Again, this is caused by scattering from In and P below the surface, reflecting the energy loss of the'He+ ions entering and leaving the sample. The oxygen content as a function of depth below the InP crystal surface was obtained by measuring the yield in the oxygen peak as a function of incident 4He+energy. At the resonance energy, 3.034 MeV, enhanced scattering occurs for oxygen present at the surface of the InP sample. The incident beam energy was then raised beyond the resonance energy. The 3.034-MeV resonance now occurs at an increased depth below the surface. Thus, enhanced scattering now occurs from oxygen present in this deeper region and the resulting signal reflects the oxygen at this depth in the InP single crystal. In this manner, a "depth profile" of oxygen content was obtained. The results of this procedure are shown in Figure 4 for InP samples exposed for 5-8 weeks. The uncertainties of the observed yields are quite large due to the large background which was subtracted from the small observed peaks. The width of the yield curve is relatively constant at about 25 keV (3060-3035 keV) for samples exposed to Gamble solution for 6-8 weeks. The stopping power of InP for the 4He+beam was calculated to be approximately 24 eV/A from the density of InP and the tabulated stopping cross section values of 3.034MeV 'He+ on indium and phosphorus. On the basis of the calculatedstopping power of 24 eV/A and the measured energy width of 25 keV, this corresponds to an oxygen depth

2932

ANALYTICAL CHEMISTRY, VOL. 64, NO. 23,DECEMBER 1, 1992 4000 r

I I

3000

1000

4000 r

1

I

3

8 30001

e

I

0 0

600

400

200

800

~~

0

1

Channel Number

2

3

Flgure 5. Relattve oxygen content

4000

3000

4

5

6

7

8

WEEKS EXPOSED

In the near-surface (-1000 A)

region of InPslngle crystals exposed to Gamble solution for 0-8 weeks. 7 Data were obtained by back-scattering of 4He+using the 3.034-MeV

1\

I

i , ' 0 0

600

400

200

800

Channel Number

Flgure 3. Back-scattered spectra of InP in Gamble solution (a)at an off-resonance energy of 3.015 MeV and (b) at 3.040 MeV. The enhancement of the oxygen signal at about channel number 300 can be Seen In the spectrum taken at 3.040 MeV, whlch Is Just above the resonance energy of leg. Depth

6)

0

4000

I

1000 I

Resonance Width

'He+ BEAM ENERGY (keV)

Figure 4. Relative yleld as a function of 4He+ incident energy about the3.040-MeV lBOresonanceforInPslnglecrystalsexposedtoamble

solution for 5-8 weeks. Percent oxygen was calculated from these data to be 10% , 13 % , 2 0 % , and 24 % at these respective exposure times. distribution of about lo00 A; the uncertainty in thickness of this oxygen-containing layer is &20%. The minimum depth resolution is defiied by Sx = I'/(dE/dx) where r is the resonance width and dE/dx is the stopping power. Since the oxygen resonance is 10 keV wide, a stopping power of 24 eV/A corresponds to a depth resolution of approximately 420 A. Because the oxygen-containing layer is only about lo00 A thick, it was impossible to discern details of the oxygen distribution in the surface region of these samples. Hence, only a rough measure of the oxygen content and an estimate

l6O resonance.

of the oxygen penetration depth can be obtained by this method. The yield maxima near 3.040 MeV, plotted as a function of exposure time in Figure 5,indicate that the oxygen content in the InP crystal surface was significantly enhanced as the exposure time to synthetic lung fluid increased. It is noted that data collected for the 4-week sample was anomalous for all analyses and was well above the trends observed in the oxygen profile of all samples. Although the faces of the crystals that were exposed to the 4He+beam were clearly marked, it is conceivable that the 4-week sample was inverted a t some point during the course of the experiment. This anomalous behavior was observed in the X P S results as well. Computer simulations were used to provide a better idea of the oxygen content in the oxygen-containing layer. Satisfactory simulations for samples exposed for 6-8 weeks were obtained assuming a uniform oxygen areal density and a 10004 oxygen depth distribution. A 500-Alayer thickness provided the best fit for the InP sample exposed for 5 weeks; the oxygen depth for samples exposed for leas than 4 weeks could not be detetermined by this method. Using the results of the computer simulations, the oxygen content in this nearsurface region was found to steadily increase from 10% in the sample exposed to Gamble solution for 5 weeks to a maximum of 24% at &weeks exposure. At exposure times less than 4 weeks, the oxygen content of the InP samples could not be determined. A resonable upper limit, however, is that the 1000-A surface layer of these indium phosphide crystals contains 10% oxygen. X-ray Photoelectron Spectroscopy. X-ray photoelectron spectroscopy was employed in an attempt to obtain a better elemental profile of the near-surface region of the indium phosphide crystale. XPS is more sensitive and gives a better elemental profiie than RBS of the uppermost 100 A of the crystal surface. Furthermore, depth p r o f i i g of this uppermost surface region was possible by changing the angle formed between the electron analyzer and the plane of the sample surface. The indium 3d5p (443.6 eV), phosphorus 2p (128.3 eV), and oxygen 1s (531.2 eV) peaks were monitored and the peak ratios were calculated according to eq 1. Figure 6 shows the relation of 0:P and 0:In peak ratios as the exposure time to synthetic lung fluid was increased. The 0:P and O I n ratios increased rapidly after the first 2-3 weeks of exposure to synthetic lung fluid. Measurements of O I n and O P atomic ratios were made a t takeoff angles of 10 and 80° (-10 and 50 A, repectively)in order to obtain a comparieonbetween surface

ANALYTICAL CHEMISTRY. VOL. 64, NO. 23. DECEMBER 1, 1902

1.c. 0

5

I

I

2

3

5

6

7

8

1

2

3 4 5 WEEKS EXPOSED

6

7

a

4

. 0.8 I 0

I

1

1

Flgure 6. Angle-resoked XPS atomlc denslty ratlos for O h (a)and O P (b)at80" (=50 A, -)and 10" ( d o A, -)in InP single crystals exposed to Gamble solution for zero to eight weeks.

--

and near-surface elemental composition. The 0:In and 0:P ratios obtained at the surface-sensitive angle are initially greater than the ratios obtained when the detector was oriented to monitor electrons being emitted from the subsurface region. This effect is more pronounced for the 0:P ratios than the 0:In ratios and indicates a possible depletion or dilution of phosphorus at the surface of the InP single crystal. These results show that the InP single crystals have a surface region initially rich in oxygen. At lengthened periods of exposure to Gamble solution, however, further oxidation of the entire oxygen-containing layer occurs as illustrated by the steep rise in the O P and O I n curves in Figure 6. By 8 weeks, 0:P and 0:In ratios in the surface and near-surface regions of the oxide layer become similar and the curves in Figure 6 begin to coalesce. This confiims that the oxygencontaining layer of the InP single crystals was chemically altered by the Gamble solution; this layer became enriched in oxygen and notably more homogeneous during the course of the experiment. The P:In ratios were also calculated;these

201111

ratios, however, remained relatively constant in all samples, indicating that these elements, if they dissolved in the Gamble solution, did so in similar concentrations. Although the oxygen content in the near-surface region of the InP crystals showed a notable increase during the course of thisstudy, no consistent shifta in the binding energies of any peaks were found. Furthermore, no significant deviation in the peaks from a normal Gaussian shape was observed, nor was any increase in the fwhm of these peaks detected. On the basis of these results, no attempts were made to determine the true nature of the oxygen-containing layer formed on the Surface of the InP crystals. No indication of surface etching by the Gamble solution was observed in the X P S spectra obtained. Determination of Indium. The samples containing Gamble solution exposed to the indium phosphide single crystals were analyzed for total indium by GFAA. Trace levels ( 4 5 ppb) of indium were found in the Gamble solution samples exposed from 0 to 3 weeks, beyond 3 weeks, indium levels increased to a fiial concentration of 60ppb in the &week sample. This result confiirma the conclusion that was reached from the RBS results that indium was in fact depleted from the surface of the InP crystal and was not merely diluted by the formation of surface oxides. Because phosphate isamajor component of the Gamble solution, it was impossible to c o n f i i experimentally whether phosphorus was also leached out from the surface;however, becausethe P:In ratios obtained by XPS remained relatively constant and because the 0:In and OP ratios were similar, it appears likelythat phosphorus ia indeed leached by the Gamble solution.

CONCLUSIONS This study has shown that a complex dynamic interaction occurs when indium phosphide is exposedto Gamble solution. In our previous work, GaAs was found to dissolve in Gamble solution to give As levels approaching 1 ppm after exposure periods of only 10 days. In contrast to that report, indium phosphide was found to dissolve at a much slower rate; however, the trends observed in this study support the conclusion that the extent of dissolution of InP increases significantly with time. In an industrial environment, the predominant form of exposure is through inhalation of InP particulates. Although InP does appear to be relatively stable, in comparison with GaAs, it must be emphasized that the total surface area of inhaled InP particulates is enormous and can result in very high leveb of dissolved InP. RECEIVED for review June 11, 1992. Accepted September

10,1992.