Anal. Chem. 1982, 54. 2111-2113
precise to within the experimental error, but the sensitivities, especially that for AA, vary significantlly. For quantitative in vivo measurement, it; is essential to calibrate each electrode before implantation and to recalibrate it after the experiment, if possible.
ACKINOWLEDGMENT The electron microscope facility time was provided by the Institute for Physical Science and Technology, University of Maryland at College Park. This work is in partial fulfillment of the requirements for a Ph.D. degree in chemistry for L.F.
Wightman, 13.M., Chemistry Department, Indiana University, personal communication, 1982. Huff, R. M. PhD. Dissertation, University of Kansas, 1980. Lindsay, W. S.;Kizzort, B. L.; Justice, J. B.; Saiamone, J. D.; Neil, D. B. Chem. Blomed. Environ. Instrum. 1980, 10, 311-330. Hawiey, M. D.; Tatawawadi, S.V.; Piekarski, S.; Adams, R. N. J. Am. Chem. SOC. 1987, 8 9 , 447-450. Blaedel, W. J.; Mabbott, G. A. Anal. Chem. 1978, 5 0 , 933-936. Brown, A.-P.; Anson, F. C. Anal. Chem. 1977, 4 9 , 1589-1595. Cheng, H.-Y.; Falat, L.; LI, R.-L. Anal. Chem. 1982, 5 4 , 1384. Wightman, R. M. Anal. Chem. 1981, 53, 1125A-1134A. Evans, J. F.; Kuwana, T. K. J . Electroanal. Chem. 1977, 8 0 , 409-416. Ewing, A. G.; Dayton, M. A.; Wightman, R. M. Anal. Chem. 1981, 53, 1842-1847.
LITERATURE CITED Tse, D. C.-S.; Kuwana, T. Anal. Chem. 1978, 5 0 , 1315-1318. Lane, R. F.; Hubbard, A. T.; Blaha, C.D. Anal. Chem. 1978, 4 8 , 1287-1 293. Cheng, H.-Y.; Strope, E.; Adams, R. N. Anal. Chem. 1979, 51, 2243-2246. Evans, J. F.; Kuwana, 7’. J . Nectroanal. Chem. 1977, 80, 409-416. Faccl, J.; Murray, R. HI. Anal. Chem. 1982, 5 4 , 772-777. Wang, J. Anal. Chem 1981, 53, 2280-2203. Gonon, F. G.; Fombariet, C. M.; Buda, M. J.; Pujoi, J. F. Anal. Chem. 1981, 53, 1386-1389 Gonon, F.; Buda, M.; C:espugllo, R.; Jovet, M.; Pujol, J. F. Braln Res. 1981, 223, 69-80. Adams, R. N., Chemislry Department, university of Kansas, personal communlcatlon, 1982.
2111
Ladislav Falat H.-Y. Cheng* Department of Chemistry University of Maryland College Park, Maryland 20742 RECEIVED for review April 21, 1982. Accepted July 6, 1982. This work was supported by grants from the donors of the Petroleum Research Fund, administered by the American Chemical Society, and the Cottrell Grant, administered by Research Corporation.
Solid-state Standard Addition Method in Secondary Ion Mass Spectrometry for Improvement of Detection Limits Sir: There has been considerable interest in quantifying secondary ion mass spectrometry (SIMS) results. Techniques utilizing theoretical and semitheoretical models of secondary ion emission have been proposed for this purpose but typically yield only “order of magnitude” results (1-3). Empirical calibration methods employing external or internal standards have also been shown to give excellent results ( 4 , 5 ) . However, the required SIMS standards are difficult to obtain since they must be homogeneous on the microscale and must closely approximate the chemical composition of the material to be analyzed. Recently, the technique of ion implantation has been adopted to fabricate SIlMS standards (6). Ion implants are well understood, and by utilization of the known implanted dosage, accurate converoion of secondary ion intensity to concentration can be achieved (7). Moreover, these ion implant standards can be tailor-made for a particular analysis requirement. Solid-state standard addition, the use of ion implantation to perform a conventional standard addition analysis to a solid samplle, has been used to determine bulk dopant concentration in steel and semiconductor samples with good results (8). One handicap of this approach is that the residual concentration to be measured must be relatively high. Most realistic dopant concentrations in semiconductor materials are too low for the direct use of this standard addition approach. In order to improve the oensitivity of this technique, we have investigated the use of (computerized signal integration to increase the signal-to-noise ratio. This technique utilizes the known Poisson statistics of the ion counting circuitry. For this case, the noise assoaiated with a particular number of counts, n, is equal to the square root of 1%. Therefore, the signal-to-noise ratio can lbe improved by a factor of 10 by increasing the number of counts by 100 times. This has been experimentally demonstnated in the field of vidicon spectrometry (9). A computeriz,edintegration method was adopted
by our laboratory for the analysis of gelatin standards for quantitative SIMS analysis of biological tissues and was found to give improved detection limits (10). In this study, the method of solid-state standard addition combined with computerized signal integration is applied to the quantitative analysis of a silicon wafer doped with a very low concentration of boron (approximately 1015atoms/cm3). Because of the existence of systematic errors in ion counting, there is deviation from the linear signal-to-noise vs. square root of time improvement. The experimental procedures are modified accordinglyto obtain an optimal signal-to-noiseratio. The method is found to yield an accuracy of 12% and precisions of 17-25 % depending upon measurement conditions. This boron concentration is almost an arder of magnitude below the detection limit found previously by the usual solid-state standard addition method.
EXPERIMENTAL SECTION Instrumentation. Ion implantation was performed with an Accelerators Inc. 300R ion implanter (11). The SIMS analysis was carried out with a Cameca IMS-3f ion microscope equipped with an electron multiplier for signal detection (12). The instrument is interfaced to a HewletbPackard 9845T microcomputer for control and data acquisition. The IMS-3f was also interfaced with a Digital PDP-11/34 minicomputer for data processing. A 5.5-kV 02+ primary beam was rastered over a 250 X 250 pm2area at a current density of 1.60 mA/cm2. The sampled area was resticted to 150 X 150 gm2 and positive secondary ions were monitored. All analyses were performed at 2 X torr. Software. The integration routine used to acquire the residual dopant signal repeatedly scans across a given mass for an operator-defined number of cycles, a cycle being one scan. The time allotted to each cycle is also under operator control. The data from each scan are added to a buffer containing the summation of all previous scans until the specified number of scans have been completed. At this time, the contents of the buffer are output. Depth profiling of the ion implanted region was performed with the standard software supplied by the manufacturer. The depth
0003-2700/82/0354-2111$01.25/00 1982 American Chemical Society
2112
ANALYTICAL CHEMISTRY, VOL. 54, NO. 12, OCTOBER 1982 -IMPLANT, PROFILE
POST
SIGNAL
SPUTTERI-SUMMED~ FOR
BORON IMPLANT PROFILE
10000
5000 DEPTH
Figure 1. Schematic of the experiment. profile data were then transferred to the PDP-l1/34A for processing by a peak integration program. This program allows the operator to subtract out the background level of the unimplanted region from the implant signal and then outputs the integrated ion intensity of the implant peak. Procedure. A commercially available polished boron doped silicon (100) wafer was used in this experiment. The boron concentration in the wafer was determined by four point probe electrical measurement to be 2.6 X 1OI6 atoms/cm2. The wafer was cut, cleaned ultrasonically with trichloroethylene, acetone, methanol, and deionized water, and mounted on an aluminum aisk with conductive silver paint. The sample was then implanted with 1 X 10l5atoms/cm2of llB at 100 kV. The single crystal Si was tilted at an angle of 7 O to reduce implant channeling. After implantation, the sample was depth profiied for l1B. The depth profile was terminated and stored when the boron signal dropped to the constant residual level. The sample was then sputtered for another 5 min, thus removing an additional 200 nm of material to ensure complete removal of implanted boron. At this point the integration routine was run with an integration time of 3 s/cycle. The schematic of the experiment is shown in Figure 1. The depth of the sputtered crater was measured by a Taylor-Hobson Talystep stylus type surface profiler with a resolution of 5-10 nm to enable conversion of the implant fluence to concentration. To determine the noise of the electron multiplier, we ran the integration program with the exit slit of the mass spectrometer closed completely and the primary ion extraction voltage turned off. To calculate the concentration of the background boron dopant, we divided the total residual signal accumulated by the integration routine by the total time of integration. The noise signal of the electron multiplier (1count/cycle) was then subtracted from this value to give the residual signal, SI,in counts/s. The following equation was used to calculate the residual concentration: S, TF c, = IDA where C, is the residual elemental concentration to be determined in atoms/cm3, S, is the residual isotopic signal in counts/s as defined previously, T is the time of analysis in seconds, P is the fluence of the implant (1 X 10l5atoms/cm2),I is the integrated sum of the implanted ion intensity in counts (typically about 1 million counts), D is the depth of the crater in centimeters (about 2X cm typically), and A is the isotopic abundance,
RESULTS AND DISCUSSION Figure 2, a depth profile of the llB implant, shows the residual boron signal obscured by noise. The usual solid-state standard addition technique using the same primary ion current density and a 5-9 acquisition time per analysis point yielded a noise limited detection limit for boron in silicon of approximately 2 X 10l6atoms/cm3, which is the detection limit of B in Si published in literature (7). With the present computerized integration approach, integrating each analytical point for 600 s (200 cycles), a boron concentration of 2.3 X
DEPTH (AI
Figure 2. Depth profile of "B in silicon.
IO
20 Tin,
30 (SEC'")
Figure 3. Signal to noise ratio plotted against the square root of the integration time. lOI5 atoms/cm3 was calculated with a standard deviation of h4 X 1014atoms/cm3. This value agrees within 12% with the electrical measurements and is an order of magnitude below the detection limits using the conventional technique. This increase coincides with the expected SIN improvement using a 100-fold increase in counting time. The degree of improvement in sensitivity possible with the use of signal integration was evaluated by repeating the experiment using increasing integration times. Figure 3 gives the SIN ratio plotted as a function of the square root of the integration times. As shown the SIN ratio begins to level off beyond 300 s integration, instead of continuing along the straight line. The time of 600 s (200 cycles) was chosen in the analysis as this time gave the best results within a reasonable analysis time. The leveling off of the curve in Figure 3 suggests that the maximum improvement to be gained by using this technique is approximately an order of magnitude. The signal-to-noise ratio becomes constant with an increasing number of cycles for two reasons. For low-level signals, a long integration time tends to result in deterioration of the apparent signal, but the most significant reason is that a systematic error derived from computer ion counting is not accounted for. Owing to the logarithmic nature of the data display, any signal below 1 count/cycle is counted as 1 count/cycle. However, it is not necessary to remove this systematic error to obtain accurate results. Figure 3 shows that a constant signal-to-noise ratio of about 1is obtained for 200 scanned cycles or more. Statistically, the signal is reliable because it is twice the background noise (systematic and random). Further comparison with the electrical value confirms the accuracy of this methodology. Since any signal with a lower signal-to-noise ratio would be unreliable statistically, the concentration measured in this experiment is concluded to be the detection limit of boron in silicon by the present method. Further increases in sensitivity beyond that will have
2113
Anal. Chem. 1982, 54, 2113-2114
to come from improved ionization efficiency, increased collection of secondary ions by the mass spectrometer, and improved detectors. The technique of computerized signal integration however has t3hown its ability to improve the sensitivity of the solid-state standard addition method to more realistic concentration levels. An analysis time of several minutes instead of seveiral seconds does not seem unreasonably long to provide an order of magnitude higher sensitivity. One caution, however, is that the sample to be analyzed by this technique must be homogeneous enough in depth to sputter through several hundrled nanometers of material while integrating a reliable signal intensity.
ACKNOWLEDGMENT The authors wish to acknowledge the assistance of Susan Palmateer for measuring the resistivity of the silicon wafer, and William Harris, Jr., for manuscript preparation. The use of the ion implantation facilities of the National Research and Resource for Submicron Structures a t Cornell University is acknowledged. LITEXATURE CITED (1) Andersen, C. A,; Hinthome, J. R. Anal. Chern. 1976, 45, 1421-1438. (2) Rudat, M. A,; Morrison, G. H. Anal. Chem. 1979, 51, 1179-1187.
Simon, D. S.;Baker, J. E.; Evans, C. A., Jr. Anal. Chem. 1976, 4 8 , 1341-1348. Ganjel, J. D.; Leta, D. P.; Morrison, G. H. Anal. Chem. 1978, 5 0 , 285-290. Havette, A.; Siodzian, G. J. fhys. Lett. (Orsay, F r . ) 1980, 4 7 , 555-558. Gries, W. H. I n t . J. Mass Spectrom. Ion Phys. 1979, 3 0 , 97-112, 113-125. Leta. D. P.; Morrison, G. H. Anal. Chem. 1980, 52, 514-519. Leta, D. P.; Morrlson, G. H. Anal. Chem. 1980, 52, 277-280. Busch, K. W.; Howeii, N. G.; Morrison, G. H. Anal. Chem. 1974, 4 6 , 575-581. Zhu, D.; Harris, W. C., Jr.; Morrison, G. H. Anal. Chem. 1982, 5 4 , 419-422. Leta, D. P. Ph.D. Thesis, Corneli University, Ithaca, NY 1980. Ruberol, J. M.; Lepareur, M.; Autier, B.; Gourgout, J. M. V I I I t h International Congress on X-Ray Optics and Microanalysis and 12th Annual Conference of the Microbeam Analysis Society, Boston, MA, 1977, pp 133A-133D.
Paul K. Chu George H. Morrison* Department of Chemistry Cornell University Ithaca, New York 14853 RECEIVED for review February 19, 1982. Accepted July 19, 1982. Funding of this project was provided by the National Science Foundation and the Office of Naval Research.
Determination of Ethylenediamine in Aqueous Solutions by Ion Chromatography Sir: In a recent publication, Wimberley suggests an alternative eluent for the determination of divalent cations by ion chromatography (I). As an extension of this work, we have investigated the determination of ethylenediamine in aqueous solutions. Using ZnC12and HC1 in the eluent, we were able to successfully chromatograph ethylenediamine and demonstrate a linear response with conductomietric detection. Our experiments substituted ZnC1, and HC1 for ZII(NO~)~ and HNOBbecause the chloride causes less suppressor column degradation (2). EXPERIMENTAL SECTION Apparatus. A Dionerr Model 16 ion chromatograph with a 0.1-mL sample loop was used for all experiments. A 4 X 50 mm cation precolumn (Dionex part no. 030830) and a 7 X 60 mm cation suppressor column (Dionex part no. 030834) were used for the separation. A Shimadzu C-R1A integrating recorder was used to record and integrate the signal. A Hewlett-Packard 5995A gas chromatograph/mass spectrometer was used to confirm that the eluting species was ethylenediamine. Chemicals. All chemicals were reagent grade and dissolved in grade 3 (18 MQ/cm) deionized water. RESULTS AND DISCUSSION Attempts to separate ethylenediamine by using the conditions established by Small et al. were unsatisfactory because ethylenediamine is stronglly retained to cation exchange resins (3). Retention times were too long to be practical and peak shapes were unsuitable for quantitation. The chromatographic system, as described by ,Jaworski, is also unsuitable because a 2 N HC1 mobile phase would expend the suppressor during the course of one deternnination (4). In a solution containing 0.004 N HC1 and 0.0025 M Zn2+ the ethylenediamine in complexed to form the Zn(NH2CHzCHzNH2)2+ chelate (5,6).This complex changes the
free energy of solvation for the ethylenediamine and changes the partition coefficient between the complex in solution and that bound to the resin. This change allows the ethylenediamine to be chromatographed with acceptable peak shapes. After separation, the complex enters the suppressor column where Zn2+ is precipitated and HC1 is stripped by the hydroxide form resin; hence, the ethylenediamine enters the conductivity detector as the free base in a background of deionized water. We determined that suppressor column regeneration as described by Wimberley proved generally adequate (1). However a more thorough regeneration was achieved by first washing the exhausted suppressor for onehalf hour with a solution containing 0.0025 M HC1 and 0.0025 M rn-phenylenediamine dihydrochloride, followed by a regeneration cycle with 1N NaOH. Several experiments were run in an attempt to optimize the chromatogram. Initially both a precolumn and an analytical column were used for the separation. This produced acceptable results with retention times of approximately 25 min. Methanol and 2-propanol were separately added to the eluent in 5% and 20% concentrations (v/v) to reduce adsorption. The addition of organic solvents to the eluent produced tailing of the peak and reduced the detector response. Finally, the analytical column was removed and the separation performed on the precolumn only. This decreased the retention time of ethylenediamine to approximately 7 min. Using only a precolumn, all alkaline earth cations were eluted before 4 rnin and were only slightly resolved from one another. All monovalent cations and amines tested (Li+,Na+, NH,', K+, Rb+, Cs+, methylamine, trimethylamine, ethylamine, diethylamine, triethylamine, ethanolamine, diethanolamine, and triethanolamine) were found to elute near the dead volume. Nine standard solutions of ethylenediamine (0, 1.0,5.0,10,15, 20,50, 75, and 100 ppm) were prepared from
This article not subJect to US. Copyright. Published 1982 by the American Chemical Society
