for the determination of silicon at ng/g levels in gallium arsenide*. Ru-Shi Liu I and Mo-Hsiung Yang. Institute of Nuclear Science, National Tsing Hua University, ...
OrigWna| papers
Spectrophotometric and polarographie methods for the determination of silicon at ng/g levels in gallium arsenide* Ru-Shi Liu I and Mo-Hsiung Yang Institute of Nuclear Science, National Tsing Hua University, Hsinchu, Taiwan 300, Republic of China Spektralphotometrische und polarographische Bestimmung von Silicium in Galliumarsenid im ng/g-Bereich Zusammenfassung. Die beschriebenen Verfahren beruhen auf der Si-Bestimmung als Heteropolys/iure. Sic umfassen den AufschluB der Probe mit Salzsfiure und Brom im PTFEGef/iB, Eliminierung der Matrixelemente (Verdampfung von Arsen als Arsentrichlorid und Extraktion von Ga als Tetrachlorokomplex mit Diisopropylether) sowie anschlieBende spektralphotometrische Bestimmung als Silicomolybdfinblau oder polarographische Bestimmung als fl-Silicomolybdfinsfiure. Die Optimierung der Vorbehandlungstechnik sowie der instrumentellen Analysenparameter wird beschrieben. Die Nachweisgrenzen liegen bei 7 ppb bzw. 5 ppb. Die vorgeschlagenen Methoden wurden auf verschiedene Sidotierte Proben angewendet. Die Ergebnisse werden verglichen und Unterschiede diskutiert. Summary. Spectrophotometric and differential pulse polarographic determinations of silicon in gallium arsenide as heteropoly acid have been established. The analysis comprises decomposition with a mixture of hydrochloric acid and bromine in a PTFE vessel, elimination of matrix elements by evaporation of arsenic as arsenic trichloride and extraction of gallium as gallium tetrachloro-complex anion by di-isopropyl ether, and finally spectrophotometric determination of silicomolybdenum blue or polarographic determination of fl-silicomolybdic acid. Optimization of sample pretreatment procedures and instrumental determination have been carefully elaborated. The detection limits of the developed methods were found to be 7 ppb and 5 ppb, respectively, for spectrophotometry and polarography. The proposed methods have been practically applied to the analysis of various Si-doped samples. The results obtained by the chemical methods are compared with those from the electrical measurement and the discrepancies found are discussed.
* Presented at the Colloquium Spectroscopicum Internationale XXIV, September 15- 20, 1985, Garmisch-Partenkirchen, FRG 1 Present address: Materials Research Laboratories, Industrial Technology Research Institute, 1021 Kuang Fu Road, Hsinchu, Taiwan, Republic of China Offprint requests to: M.-H. Yang Fresenius Z Anal Chern (1986) 325:272 9 Springer-Verlag 1986
277
1. Introduction Gallium arsenide is an important compound semiconductor of the type of AnIBv. The crystals have quite high resistivity at room temperature but this is reduced in the presence of impurities even below 10 - 9 g/g levels [1]. Silicon is known to be one of the important doping elements which can render the GaAs crystals either to become n-type or p-type. For the purpose of quality control in the preparation and purification of the materials and for monitoring the doping concentration in the crystals, it is urgently required to develop an analytical method which can accurately and sensitively determine silicon of both impurity and dopant in the GaAs crystals. Though electrical measurement can provide a simple means for evaluating the carrier concentration in the crystals, the information so obtained is restricted only to impurities of electroactive form, while those of electroinactive or electrically compensated species cannot effectively be determined. Chemical methods on the other hand, though tedious and time-consuming, can be totally unaffected by the electrical property and can thus provide absolute concentration of the specific element of interest. For the determination of traces of silicon in GaAs a range of instrumental methods have been proposed, e. g., spark source mass spectrometry [2, 3], emission spectrometry [4], neutron activation analysis [5] and spectrophotometry [6]. Most of the published instrumental procedures call for a preliminary preconcentration step. Undoubtedly these pretreatment procedure can enhance the sensitivity, but they are tedious to carry out, and contamination from the reagents is always a possibility. Most of the pretreatment procedures involve prior separation of matrix elements by extraction of gallium as gallium tetrachlorocomplex anion [7, 8, 9], and distillation or extraction of arsenic as arsenic trichloride [10]. One alternative approach has recently been reported by separation of traces of silicon from the dissolved sample solution by volatilization of silicon tetrafluoride [11]. The present study follows basically the existing principles of matrix separation by evaporation of arsenic as arsenic trichloride and extraction of gallium as gallium tetrachlorocomplex anion by di-isopropyl ether. The traces of silicon in the residual solution are subsequently transferred to silicomolybdic acid and determined both by spectrophotometry and polarography. Some modifications have, however, been made in order to decrease the analytical blank and to facilitate the optimal combination of preconcentration procedure and instrumental determination. The develop-
Or g naJarbe eH ment of the polarographic method, which has heretofore not been reported for the determination of silicon in GaAs, is aimed to be a cross check of the results obtained by spectrophotometry. The purpose of this study is twofold, one is to establish routine quality control of the Si-doped GaAs crystals, the other is to pursue the minimum detection limit achievable by this established procedure. In order to attain a low detection limit, the principles of ultratrace analysis, which include meticulous control of laboratory air, container surface, reagents etc., is strictly obeyed throughout this work.
2. Experimental
Apparatus The determination of Si was made both by a Beckman Spectrophotometer (Model 26) with 10 m m glass cells, and a Princeton Applied Research Polarographic Analyzer (Model 384) in connection with a Model 0082 X-Y recorder and a Model 308 S.M.D.E. A pH-meter was used for pHadjustment. A thermostatically controlled heating block was used for decomposition of samples and evaporation of sample solutions. Sub-boiling distillation equipment of fused quartz (Berghof) was used for the purification of HC1, H2S04, and diisopropyl ether. The radiotracers of 76As and 72Ga used in this study were obtained by irradiating As203 and Ga metal in the Swimming Pool type reactor of National Tsing Hua University, Taiwan (1 Megawatt, neutron flux 3 • 1012 n/cm 2- s). A N a I (T1) y-scintillation counting system was used. Electrical measurement was performed by van der Pauw method.
Containers and pipets Containers and pipets made of teflon and polypropylene were used exclusively throughout the work and were cleaned by immersing in 1:1 HNO3 overnight and steaming successively with HNO3 vapor and water vapor before use.
Reagents All chemicals used were of analytical grade. Double-distilled water (quartz still) was used in all instances. HC1, H2SO4 and di-isopropyl ether were purified by sub-boiling distillation. Ammonium molybdate solution: Dissolve appropriate amounts of the reagent in water and dilute the solution to two different concentrations. Molybdate solutions of 0.1% and 1.2% were used for polarographic and spectrophotometric measurements. Tartaric acid solution, 5% (w/v) reducing agent: Dissolve 0.8 g of unhydrated sodium sulfite in 20 ml of water and add 0.16 g of 1-amino-2-naphthol-4-sulfonic acid. The whole solution is then added to a 100-ml polypropylene flask containing 70 ml of water with 10 g of sodium bisulfite; diluted to the mark and stored in a refrigerator. Dilute the solution 10 times before use. Etching solution: HzSO 4 (96%)/H202 (35%)/H20 (4:1 : 1). Triton X-100 solution: 0.2%. Citrate buffer: Dissolve 8.75 g of citric acid monohydrate in 50 ml of I M sodium hydroxide, dilute to 250 ml and adjust the pH to 2.5 with 1 M hydrochloric acid.
Silicon standard solution, 1000 gg/ml: Prepared from commercially available standard solution.
Analytical procedure The flow chart for the determination of silicon in GaAs is shown in Fig. 1. The GaAs wafer is partly used for electrical measurement, and partly for chemical analysis. GaAs wafer (each about 0.1 g) was treated sequentially by ultrasonic washing for 5 min each with trichloroethane and acetone, rinsing with water, etching with a solution containing conc. H z S O 4 : H 2 0 2 : H 2 0 (4:1:1) for 5 min, and finally rinsing with water and drying. The treated sample was then transferred to a 50 ml Teflon beaker and 2 ml of conc. HC1 and I ml bromine were added. The whole system was inserted in a heating block and heated gently to 45~ for dissolution. After that the sample solution was heated to 60 ~C and kept at that constant temperature for 10 rain, during which time practically all bromine is evaporated. For the removal of the arsenic matrix, 1 ml of 12 N HC1 and 1 ml of 10% hydroxylammonium chloride solution were added in order to reduce the last traces of bromine to bromide and to reduce arsenic(V) to arsenic(III), and subsequently evaporated to dryness by raising the temperature to the boiling point of the solution ( ~ 110~ A time of 40 min is normally required for this process for complete evaporation of arsenic as arsenic trichloride. For the removal of gallium, two 2 ml portions of 8 N HC1 were added to dissolve the residue, and the resulting solution was transferred to a 10 ml PE tube. 2 ml of di-isopropyl ether was added to the solution, this was extracted twice with 2 ml portions of the organic solution and the organic layer which contains gallium was discarded. The final aqueous solution was transferred to a 10 ml Teflon beaker and evaporated at 100~ to a small volume ( ~ 50/xl). Subsequently, after adding 2 ml of water and adjusting with 1 M NH4OH to pH 1.1, the resulting solution was subjected to the analysis by spectrophotometry and polarography.
Spectrophotometric determination Pipette 1 ml of the resulting solution into a 5 ml PE tube, add 0.4 ml of 1.2% ammonium molybdate, mix and allow to stand for 5 min. Subsequently, add 0.4 ml of 5% tartaric acid solution and 0.2 ml of reducing agent. Shake the tube thoroughly and leave to stand for 20 min. Measure the optical density of the reduced silicomolybdic acid at 810 nm in 1 ml cells. The blank value is established by the same procedure with a sample.
Polarographic determination Pipette 1 ml of the resulting solution into a 15 ml PE tube, add 1 ml of 0.1% ammonium molybdate and mix (the pH being slightly changed to 1.3). Add 0.6 ml of methyl ethyl ketone, then mix and set aside for 2 h. Add citrate buffer (pH 2.5) to make up to 10 ml and transfer the solution to a polarographic cell, add 0.1 ml of 0.2% Triton X-100 solution and deoxygenate the solution by passing nitrogen through it for 15 rain. Record the differential pulse polarogram between - 0 . 2 and - 0 . 6 V vs. SCE, using a pulse height of 50 mV, a scan-rate of 2 mV/s and a drop time of 2 s. Measure the peak height at - 0.35 V (_+ 0.04 V) on the polarogram relative to that of the blank sample. 273
Original papers GaAs Sample (Wafer)
GaAs cleaning and decomposition
+ Evaporation of As
Extraction of Ga (8NHC1, di-isopropyl ether)
Enrichment of Si (pH adjusted to about 1.1)
I /~-silicomolybdic acid formation
Electrical measurement
Differential pulse polarography (Current measurement at -0.35 ___0.04 V)
Molybdenum blue formation
Spectrophotometry (absorbance measurement at 2 = 810 nm)
3. Results and discussion
Optimization of sample pretreatment procedures To ensure accurate and sensitive determination of trace amounts of silicon in GaAs, a complete separation of matrix elements prior to the instrumental determination, as well as meticulously control of blank value throughout the analytical procedures is of primary importance. The dissolution of GaAs can be done by aqua regia [5] or a mixture of concentrated HC1 and Br2 [12]. Both decomposition solutions comprise oxidizing constituents (HNO3 and Br2) which could convert arsenic in the dissolving solution to arsenic(V) and would therefore prevent the volatilization of arsenic as arsenic trichloride. The choice of HC1-Brz mixture as decomposition solution is based primarily on the consideration that bromine can be easily eliminated as compared to HNO3 from the dissolving solution due to its higher volatility (b. p. = 58 ~C). After heating the solution to 60 ~C, the last traces of bromine can be easily reduced by the addition of hydroxylammonium chloride to bromide, and arsenic(V) to arsenic(III). From the radiotracer study (As-76), it reveals that arsenic can be completely eliminated as arsenic trichloride in concentrated HC1 medium through evaporation at 110 ~C. Further removal of gallium matrix is accomplished by extraction with di-isopropyl ether in HC1 medium [13]. The tracer study indicates that the extraction yield can be as high as 99% at 8 M HC1 for 1 min extraction. No appreciable loss of silicon can be found throughout the separation procedures. For sensitive determination of silicon, further optimization should be made by adjusting the residual solution 274
Fig. 1 Flow chart for the determination of Si by spectrophotometry, differential pulse polarography and electrical measurement
containing traces of silicon to a medium suitably used for spectrophotometric and polarographic determination, while avoiding introduction of high blank from reagents, container surface and laboratory air. Traces of silicon remaining in the HC1 solution after a series of pretreatment steps are subject to gentle evaporation at 100 ~C, with no other addition of reagents. As the volume is reduced to about 50 ~tl, 2 ml of water is added to it and further adjusted to pH 1.1. The resulting solution can be used for the subsequent instrumental determination.
Spectrophotometrie determination of silicon The most widely used method for the colorimetric determination of traces of silicon utilizes the silicomolybdic acid complex formed by the reaction of silicic acid and ammonium molybdate in acid solution. Strickland discovered the existence of two forms (c~ and /~) of silicomolybdic acid and shows that in the visual region of the spectrum the absorption of the r is approximately twice that of the a-form [14]. The/~-form complex is, however, not stable but will change spontaneously into the a-form in several hours [14]. For the determination of silicon as yellow heteropoly acid some organic solvents like acetone and methyl ethyl ketone can be used to stabilize the/~-form [14, 15]. The yellow complex, however, cannot be used for the determination of silicon lower than in Ixg/g levels. To increase the sensitivity, the yellow silicomolybdic acid upon reacting with suitable reducing agents can yield intensely colored molybdenum blue which forms the basis of sensitive method for determining silicon.
OriliHalarl 0.20
ile
f.6 PH=JI 5 x ] O 17
2.3-t-10 -3 1.8+10 -3 --
2,250 1,126 --
spectrophotometry, p o l a r o g r a p h y and electrical measurements by van der Pauw method. The concentration of silicon determined by chemical methods are expressed both in gg/g and a t o m / c m 3 and those by electrical measurement only by carrier concentration/cm 3. The analytical d a t a obtained by s p e c t r o p h o t o m e t r y and p o l a r o g r a p h y are seen in g o o d agreement each other, but they are significantly higher than those obtained by electrical measurement. The discrepancy found between them can be attributed to the fact that the electroinactive fraction o f silicon in the crystal is not determinable by electrical measurement, while b o t h o f electroactive and inactive fractions can, however, be equally determined by the chemical method. The difference in silicon concentration observed between different samples are also w o r t h to be noted. The silicon concentration o b t a i n e d by chemical methods in sample 2 (2.4 x 1018 a t o m / c m 3) is seen to be higher than that in sample I (2.1 x 10 TM atom/cm3), while the resistivity and mobility o f sample I are, however, higher than those o f sample 2. This seems to agree well with the theoretical prediction that the sample with higher concentrations o f foreign a t o m in the crystals will basically result in lower mobility and resistivity, and this agreement m a y further be used, to a certain extent, as an indication o f the reliability o f the present d a t a obtained by the chemical methods.
Acknowledgements. The authors thank the Materials Research Laboratories, ITRI for providing single crystals of GaAs samples. We also thank Mr. C. Y. Nee, T. P. Chen and R. S. Tang of ITRI for very helpful discussion. The financial support of the National Science Council of the Republic of China is highly appreciated.
References l. Sze SM (1981) Physics of semiconductor devices. Wiley-Interscience, New York 2. Gauneau A, Ruppert A, Minier M, Regreny O, Coquille R (1982) Anal Chim Acta 135:193 3. Kane PF (1966) Anal Chem 38:29A 4. Geilmann W, Tolg G (1960) Glastechn Bet 33:295 5. Green DE, Heslop JAB, Whitley JE (1963) Analyst 88:522 6. Otmakhove Z], Chashchina OV (1971) Trudy tomsk gos Univ 204:140 7. Roberts JA, Winwood J, Millett EJ (1965) Proceeding of the SAC Conference, Nottingham. W Heifer & Sons Ltd, Cambridge, p 528 8. Lysenko VI, Kim AG (1965) Tr Kom Anal Khim 15:200 9. Owens EB (1959) Appl Spectrosc 13:105 10. Kristaleva LB, Kristalev PV (1963) Sb Nauch Tr Permsk Politekh Inst 14:68 11. Kikuchi R, Kaiser G, T61g G (1983) Bunseki Kagaku 32:E231 12. Maenhaut W (1966) Anal Chem 38:103 13. Koch OG, Koch-Dedic GA (1974) Handbuch der Spurenanalyse, vol I. Springer, Berlin Heidelberg 14. Chalmers RA, Sinclair AG (1965) Anal Chim Acta 33:384 15. Grasshoff K, Hahn H (1959) Fresenius Z Anal Chem 168:247 16. Shell HR (1958) Treatise on analytical chemistry, part 2. Interscience Publishers, New York 17. Chalmers RA, Sinclair AG (1966) Anal Chem 34:412 18. Fogg AG, Osakwe AA (1976) Anal Lett 9:23 19. Fogg AG, Osakwe AA (1978) Talanta 25:226
Received November 19, 1985
277
