High-sensitivity electronic Raman spectroscopy for ... - ACS Publications

scopic characterization of bulk gallium arsenide is the work initiated by Wan and Bray (1). They reported the first sharp structure for the electronic...
0 downloads 0 Views 660KB Size
994

Anal. Chem. 1989, 61, 994-998

that the partitioning process may not be the limiting step in the interference.

CONCLUSIONS A model has been developed for the simultaneous determination of analyte-interferent equilibrium constants and degrees of interference. This model is applicable to any system where the measurement of one component of the equilibrium is interfered with by another component of the equilibrium. The model has been applied to the measurement of HOCl by an amperometric membrane electrode in the presence of an organic chloramine. Hydrolysis constants for the N-chloro derivatives of succinimide, cyanuric acid, and 5,5-dimethyl1.6 X and 5.7 x hydantoin were found to be 3.5 x lo4 M, respectively. Interference is also significant with these chloramines. Although response factors for these chloramines relative to HOCl are small, the electrode responses are 1 to 52 times the true HOCl concentrations. Since organic chloramines are generally poor disinfectants, the electrode response is not a good measure of disinfection ability in these model systems. The model has also been used to compare analytical methods for the measurement of free chlorine in the presence of interfering organic chloramines. Registry No. H,O, 7732-18-5; C12, 7782-50-5; N-chloro85424succinimide, 128-09-6;N-chloro-5,5-dimethylhydantoin, 98-2. LITERATURE CITED (1) U.S. Environmental Protection Agency Fed. Regist. 1975, 40(248), 5956649588.

Marks, H.C.; Strandskov, F. B. Ann. NY Acad. Scl. 1950, 53(1), 163-171. Kruse, C. W., Olivieri, V. P.; Kawata, K. Water Sewage Works 1971, 118(6), 187-193. Wolfe. R. L.; Ward, N. R.; Olson, B. H. Environ. Sci. Technol. 1985, 79,1192-1195. Cooper, W. J.; Mehran, M. F.; Slifker, R. A,; Smith, D. A,; Villate, J. T.; Gibbs, P. H. J. Am. Water Works Assoc. 1982, 74(10), 546-552. Isaac, R. A. J.-Water Pollut. Controlfed. 1983, 55(11), 1316. Wajon, J. E.; Morris, J. C. Envlron. I n t . 1980, 3, 41-47. Plnsky, M. L.; Hu, H. C. Envlron. Sci. Technol. 1981, 75(4), 423-430. Johnson, J. D.; Edwards, J. W.; Keeslar. F. J. Am. Water Works AssOC. 1978, 70(6), 341-348. Reinhard, M.; Stumm, W. I n Water Chbrination: Envlronmental Impact and Health Effects; Jolley, R. L., Ed.; Ann Arbor Science: Ann Arbor, MI, 1980; Vol. 3, Chapter 20. Standard Methods for the Examinatbn of Water and Wastewater, 16th ed.;American Public Health Association, American Water Works Association, Water Pollution Control Federation: Washington, DC, 1965; pp 294-315. Morris, J. C. J. Phys. Chem. 1966, 70, 3798-3802. Constantinides, A. Applied Numerical Methods with Personal Computers. McGraw-Hill: New York, 1987; Chapter 7. Box, G. E. P.; Muller, M. E. Ann. Math. Stat. 1958, 29, 610-611. Albert, A.; Serjeant, E. P. The Determination of Ionization Constants; 3rd ed.: Chapman and Hall: New York, 1984; p 82. Jensen, J. N.; LeCloirec, C.; Johnson, J. D. I n Water Chlorination: Environmental Impact and Health Effects; Jolley, R. L., Ed.; Lewis Publishers: Chelsea, MI; Vol. 6, in press. Brashear, G., University of North Carolina at Chapel Hill, unpublished results, 1982. Jensen. J. N.; Johnson, J. D., submitted to Environ. Sci. Technol. Morris, J. C.; Isaac, R. A. I n Water Chlorination: Environmental I m pact and HeaRh Effects; Jolley, R. L., Ed.; Ann Arbor Science: Ann Arbor, MI, 1983; Vol. 4, Chapter 2.

RECEIVED for review August 12, 1988. Accepted January 23, 1989. This work was sipported by the Electric Po& Re' search Institute (EPRI Grant No. RP2300-7).

High-Sensitivity Electronic Raman Spectroscopy for Acceptor Determination in Gallium Arsenide T. D. Harris,* M. Lamont Schnoes, and L. Seibles

AT&T Bell Laboratories, Room 7C-223, 600 Mountain Avenue, Murray Hill, New Jersey 07974

A new method for acceptor determlnatlon by electronic Raman scatterlng in bulk semi-insulating GaAs Is reported. Separate laser wavelengths for photoneutrallzation of acceptors and probing neutral acceptor populations are employed. Sensltlvity is hproved by I O 4 over previous methods. The added sensttlvlty permits a more complete understandlng of charge balance, allows spatial mapping, and Illuminates the varlatlon of shallow donor concentratlon.

INTRODUCTION The determination of trace impurities is central to understanding the growth and processing of semiconductors and devices. Many modern electronic materials have purity levels that place the concentration of unintentional dopants far beyond the detection limit of any currently available analytical method. This condition is particularly true of undoped semi-insulating (SI) GaAs, in that the desired electrical behavior is achieved by a careful balance of the shallow impurities and deep intrinsic defects. Changes in impurity concentrations at the part-per-billion level can have substantial effects on the performance of the material. To compound the problem further, the behavior of any elemental impurity depends on the crystal site it occupies. Consequently, concentrations from bulk chemical analysis must be compared

to crystal site specific methods with caution. While we cannot review solid-state spectroscopy terminology in detail, we add for those unfamiliar with the terms that an acceptor is a lattice site deficient of electrons while a donor contains surplus electrons. For example in gallium arsenide, cadmium (a group I1 element) is an acceptor if located on a gallium lattice site while sulfur (a group IV element) is a donor if located on an arsenic lattice site. Interstitial atoms, impurity clusters, and intrinsic defects do not have predictable electrical behavior. Among the most promising developments in the spectroscopic characterization of bulk gallium arsenide is the work initiated by Wan and Bray ( I ) . They reported the first sharp structure for the electronic Raman (ER) scattering of the excited states of shallow acceptors. They also point out the utility of using the LO phonon intensity as an internal standard for comparing spectra of different samples. This spectrum has the unusual property of being visible only in bulk undoped semi-insulating crystals using 1.064-pm radiation. The observations were explained by the high purity of this material and the unique optical behavior of the midgap defect EL2 ( 2 ) . This spectroscopic process is complex, and the reader is referred to ref 2 for detail. In summary, absorption of 1.064pm radiation converts the midgap defect first to a neutral charge state and then to a nonabsorbing metastable species. Holes released by the photoneutralization of EL2 bind to the normally ionized acceptors, converting them

0003-2700/89/0361-0994$01.50/0@ 1989 American Chemical Society

ANALYTICAL CHEMISTRY, VOL. 61, NO. 9, MAY 1, 1989

to a charge neutral state. Electronic Raman scattering from the acceptor hydrogenic energy levels, present in a nonequilibrium environment of extremely low ionized impurity concentration, exhibit spectra with as narrow a line width as previously observed in any material. Wagner and co-workers expanded on this work in a series of papers beginning with a calibration of the carbon scattering cross section normalized to the 2TA phonon feature (3). This required the use of samples with known carbon concentration. The only method of determining any shallow acceptor concentration in GaAs is for carbon, by local vibration mode analysis (LVM). Subsequently the comparison of selectively excited donor-acceptor pair luminescence (SPL) with ER excited both a t 1.064 pm and a t 0.86 pm with a dye laser showed the complementary nature of each (4). While the dye laser excited ER was detected with better sensitivity using an intensified Si vidicon, interfering luminescence assigned to optical phonon replicas of the D-A pair band, compromised the ability to quantify acceptor concentrations. Unfortunately only two calibration samples with concentrations differing by more than an order of magnitude were used. The inability to detect shallow acceptors compensated by shallow donors, noted previously (I),was ignored. The use of ER to study deeper intrinsic impurities was also demonstrated (5). In two recent reports, this same group has further clarified the usefulness of ER for acceptor determination (6, 7). Multiple sample calibrations for both carbon and zinc are reported and as expected the ER scattering cross section for these two acceptors is the same. Spark source mass spectrometry (SSMS) was used to establish standard zinc concentrations. Given the consistent results for this method and ER, the calibration is almost certainly correct, but the reliability of SSMS for acceptor determination is suspect because of the loss of lattice site information. If zinc were present as interstitials or inclusions, a false high acceptor level would be concluded. These authors also now stress that ER measures the difference between the donor and acceptor concentrations rather than absolute acceptor concentrations. The method using 1.064-pm radiation suffers from the severe disadvantage of poor detector performance. Either the high noise of the Ge diode of Wagner or the low quantum yield and high dark count of the S1 photomultiplier of Bray severely limit both sensitivity and sample throughput. The 1-2 h integration times required make studies of impurity spatial distributions tedious if not heroic. The use of higher energy radiation and a multichannel detector ( 4 ) helped considerably but the compromise from the increased background diminished the advantage. This report stems from the observation in our laboratory that, contrary to the reports of Wagner and co-workers, most samples of commercial SI liquid encapsulated Czochralski (LEC) GaAs did not exhibit any observable ER with 0.85-0.90-pm illumination. Several advantages are available if sensitivity can be increased. First detection limits are improved, measurement time is reduced, or both occur and spectral resolution, from narrower slit widths, can be increased. Improved spectral resolution is vital to the accurate peak area measurement compromised by the spectral congestion present with more than one acceptor. Thinner samples or epitaxial films require large improvements in sensitivity. Since no better detectors exist for 1.064-pm radiation than those noted above, use of higher energy radiation, for which better detectors are available, is necessary for improved sensitivity. The noted absence in our laboratory of ER using 0.85-pm illumination implies that acceptor neutralization is not efficient in all SI GaAs samples a t this excitation wavelength and that 1.064-pm radiation will be necessary. The sensitivity problem can thus be addressed by separating the

QQ5

neutralization and Raman excitation steps allowing optimization of detection within the restriction that the exciting radiation must be subgap, 0.82 pm at 1.7 K. The solution to the excitation problem is not well defined until an optimum wavelength is determined. Current continuous dye laser technology, using argon ion laser pumped styryl9 dye, allows excitation with radiation between 0.79 and 0.91 pm. Detection in this region has been historically difficult because of poor photocathode quantum yield. Both photomultipliers and amplified multichannel detectors will suffer as a result. A new low noise detector technology has recently become available which addresses this wavelength region. Modern Si charged couple device (CCD) camera detectors are available with peak quantum efficiencies near 70% and read-out noise of 10 electrons per spectral channel (8). With some sacrifice in the visible peak response, quantum efficiency in the near-infrared region exceeds all previously available low noise detectors. The combination of this detector with a spectrometer optimized for this spectral region permits detection of scattered radiation with an efficiency approaching the best available a t any wavelength. With this greatly improved detection, many previously impractical studies are possible. First, the dependence of the ER signal on the wavelength and power of the exciting laser and on the power of the neutralizing laser must be understood. We report a system with sensitivity at least 1 x IO4 greater than those previously used. With this system we can confidently study signals on samples with acceptor concentrations typical of currently available commercial material. The use of this detector for ER has been briefly reported in two published conference proceedings (9, 10).

EXPERIMENTAL SECTION The samples used in this study were LEC grown crystals purchased from commercial suppliers. Slabs cut 6 mm thick and polished on both sides were used for infrared analysis. Adjacent to the position of the 13 mm diameter disks used for IR analysis, rectangular pieces 3 mm X 3 mm X 1 mm were cut for ER and selective pair luminescence (SPL). From a large number of commercial samples subjected to IR determination of carbon content, four samples with carbon concentrations ranging from 1.6 X 1015to 1.0 X 1014cm-3 were selected for ER. No transport data were available for these samples. Infrared determination of carbon concentration was performed with a Nicolet 7199 FTIR spectrometer. Spectra were recorded with 0.125-cm-l resolution, at room temperature and at 80 K to maximize sensitivity. Averaging times of 1200 s provided excellent signal to noise. The absorption coefficient of Homma et al. was used to calculate concentrations (11). Excitation of ER was with an Ar ion pumped continuous styryl 9 dye laser. With high dye concentration the tuning range was extended beyond 0.910 pm. Acceptor neutralization was carried out with a CVI YAG Mas Model 95 continuous Nd doped yttrium aluminum garnet laser. Powers of 1.8 W after filtering with a 0.2-nm band-pass filter were directed onto the sample without focus. The diameter of the dye laser was adjusted to half that of the YAG laser, coaxial and counter propagating. Scattered radiation was collected normal to the laser beams and directed to the entrance slit of a Spex 1877 triple spectrometer. The geometry of the experiment is shown in Figure 1. Power densities ranged from 40 to 1 W cm-* for the YAG laser and 4 to 0.1 W cm-' for the dye laser. Dye laser wavelengths of 0.85,0.87, and 0.89 pm were investigated. Since the detection scheme is uniquely sensitive in this spectral region, it will be described in detail. The triple spectrometer is a 0.22 m focal length, f 4.4 subtractive dispersion double spectrometer followed by a Czerny-Turner 0.6 m focal length, f 4.5 single spectrometer. The single spectrometer is fitted with a grating turret, which allows for three gratings with no realignment. The double spectrometer is fitted with 1200 grooves/mm gratings blazed at 0.75 pm. The single spectrometer is fitted with 600, 1200, and 1800 grooves/mm gratings also blazed at 0.75 pm, but all data used for this study was collected by using the 1200

ANALYTICAL CHEMISTRY, VOL. 61, NO. 9, MAY 1, 1989

996

8001

2 400 z

0 05-0.90pm DYE L A S E R

YAG L A S E R

Flgure 1. Block diagram of the instrument. Details are discussed in the Experimental Section.

1

TEKTRONIX CCD #546-6-10

-A

Or

0

I

301 !

w

A

0

l

$A0 200 260 300 RAMAN SHIFT (WAVENUMBERSI

Figure 3. ER spectra: (a) (1) no pump laser, (2) intermediate pump power, (3) full pump power; (b) ER spectrum of sample with 1.6 X 1015 cm3 carbon concentration with probe at 0.85 pm, integration time 30 s; (c)same as b with pump wavelength at 0.87 pm and integration time 60 s; (d) same as b, pump wavelength at 0.89 pm and integration time 200 s. No correction for spectrometer efficiency is made.

a-" 1 o f 0

.

I 0

0

I

I

I

0

n W

'

N zi- .*

YAG L A S E R POWER

Flgure 4. Peak area ratio (carbon 2S3&(T0 power, probe laser at 0.85 p m .

+ LO) vs pump laser

Table I sample

carbon concn, cm-3

1 2 3

1.6 x 1015 7 x 1014 3 x 1014 1 x 1014

4

C/(LO TO)

+

0.05 n.d. 0.02 0.0033

C/Zn

C/Mg

7

n.d. n.d. 0.317

-

n.d. 0.127

of 1.6 X 1015~ m - This ~ . sample was chosen for careful study because the ER spectrum is typical of those shown in ref 1-4. This spectrum was recorded with resolution of 1.5 cm-l. Figure 3b shows the ER lines expanded full scale. Clearly the signal to noise for the 30-s integration time is striking compared to previous reports on samples with similar carbon concentrations. The rising base line can be removed by either reducing probe power or raising pump power, as shown in Figure 3a. Parts c and d of Figure 3 show that for this sample no large dependence on probe laser wavelength is evident, but a drop in scattering cross section at longer wavelengths is apparent.

ANALYTICAL CHEMISTRY, VOL. 61, NO. 9, MAY 1, 1989

I

100

150

200 250 WAVENUMBERS

300

350

Flgure 5. ER spectra of (a) sample 2 with carbon = 7 X 1014 ~ r n - ~ ; (b) sample 3, carbon = 3 X loi4 ~ m - (c) ~ ;sample 4, carbon