Polarization spectroscopy for elemental analysis of trace concentrations

mental analysis In atomic flames. Improvements In polari- zation rotation detection and excellent suppression of flame background noise through polari...
0 downloads 0 Views 579KB Size
70

Anal. Chem. 1985, 57,70-73

Polarization Spectroscopy for Elemental Analysis at Trace Concentrations William G. T o n g and Edward S. Yeung* Department of Chemistry and Ames Laboratory, Iowa State Uniuersity, Ames, Iowa 50011

We report the use of polarization spectroscopy for trace elemental analysls In atomlc flames. Improvements In polarltatlon rotatlon detectlon and excellent suppresslon of flame background noise through polarlzatlon modulatlon enable this method to achleve detectlon limits of 30 ppt (trllllon) sodlum and 37 ppb barium, whlle taklng advantage of the convenlent and fast analyte Introduction of analytical flames. Furthermore, since H provides Doppler-free Informatlon, the spectral resolutlon Is suitable for Isotope ratio analysis, provldlng excellent selectlvlty and mlnlmlzlng spectral interference.

Trace elemental analysis has been performed most frequently by using atomic absorption, emission, or fluorescence spectrometry. The advent of lasers and their use as excitation light sources accelerated the progress for detection a t extremely low concentrations. Laser-induced atomic fluorescence (AFS) especially benefited from many advantages of lasers, such as high photon flux and monochromaticity. The sensitivity of AFS has been demonstrated for detection limits in the range of 10-250 atoms/cm3 (1-5) and even single atom detection (6) in the gas phase. However, atomic fluorescence analysis of elements in aqueous solutions, using continuous sample introduction schemes such as flames, yields detection limits much higher than those for free atoms (7-9). The main contributing factor for this is the flame background noise caused by the emission from molecular species such as OH, C2, CH, and CN. Other background noises include light scattering by particulates and by atomic and molecular species and stray radiation. Various techniques, including gated detection (7,10,11), laser light amplitude modulation (12,13), background noise correction at nonresonant wavelengths (14, 15), and wavelength modulation (16), have been used to suppress background noise. In this paper, we investigate the feasibility of trace elemental analysis in flames based on polarization spectroscopy (17). So far, this technique has been used for high-resolution atomic spectroscopy. Recent developments in improved detectability in polarization rotation in the condensed phase (18) should bring similar improvements in atomic systems. If so, there is much to be gained. In principle, scattered light and background emission should not contribute to the signal in polarization spectroscopy, making it relatively interference free. Furthermore, polarization spectroscopy provides Doppler-free information, and the resulting high spectral resolution allows selective measurements of the atomic hyperfine components for stable isotope analysis (19). All these can be accomplished while taking advantage of analytical flames as the atomizer for convenient sample introduction.

THEORY Doppler-free polarization spectroscopy (17) is related to saturation spectroscopy (20-22) but provides significantly higher sensitivity. The analyte sample in the atomizer is illuminated by a circularly or linearly polarized pump beam and a counterpropagating linearly polarized probe beam of the same frequency. An optical anisotropy, induced by the

pump beam, is experienced by the weak probe beam when the two beams interact with the same atoms whose axial velocity is essentially zero. The polarization of the probe beam is affected by the anisotropic optical medium and hence allowing the probe beam to be transmitted through crossed polarizers. The Doppler-free spectrum is conveniently collected by monitoring this probe signal with a photomultiplier tube while scanning the laser frequency across the center of the absorption line. Two different types of polarization spectra can be collected depending on whether laser-induced dichroism or birefringence is observed. For the polarization spectroscopy arrangement using a weak linearly polarized probe beam and a circularly polarized pump beam, the pump beam induces different changes in the absorption coefficients, Aa+ and Aa-, for the two circularly polarized components of the probe beam. Hence, the probe beam becomes elliptically polarized depending on the difference Aa' - Aa-. Similarly, the difference in changes of refractive index, An+ - An-, induces birefringence anisotropy and hence rotation of the polarization axis. The Lorentzian-shaped dichroism curve \k and the dispersion-shaped birefringence resonance curve @ are related to each other by the Kramers-Kronig relation (23, 24) cp = -x\k

(1)

CP = (2av/c)An*L

(2)

where

Aa*(L/2) (3) and L is the absorption path length and x is the laser detuning from the resonant frequency x = 2a(v - vo)/r (4) \k

where vo is the resonant frequency of a stationary atom and I? is the natural line width. The Lorentzian-shaped absorption-change profile Aa' is described as (17,201 A(u+ = -1 / 2 ~ o l / ~ s * t ( + l x2) (5)

Aa- = dAa+ (6) where a. is the unsaturated background absorption, I is the pump beam intensity, and is the saturation parameter. The parameter d is dependent on decay rates and angular momentum numbers J' and J of the lower and upper levels of the transition, respectively. It measures the strength of dichroism and birefringence anisotropy. From eq 1, 2, and 3, the refractive-index change An* is related to Aa* as An* = -l/zx(c/2~v)Aa' (7) By combining the contributions of both anisotropies, the light intensity transmitted through the analyzer, IT, is described by (17) IT= Io[f12+ B(s/2)x/(l + x 2 ) + (s/4I2/(1 + x 2 ) ] (8) where

s = -72(1 - d)a&I/I,,t

0003-2700/85/0357-0070$01.50/00 1984 American Chemical Society

(9)

ANALYTICAL CHEMISTRY, VOL. 57,

-1-

PHI

-7-i

LT

I

U

Figure 1. Experimental arrangement for Doppler-free polarizatlon spectroscopy. Optical paths are shown as broken lines and electrical connectins are shown as solid lines: A, analyzer; AL, argon ion laser; AM, nanoammeter; B, slot burner; BS, beam splitter; F, filter; HV, high-voltage op amp; L, lens; LA, lock-in amplifier; LM, light modulator; LT, light trap; 0, oscilloscope; P, polarizer; PDP 11, computer; PH, pin-hole aperture; PMT, photomultiplier tube; R, Fresnel rhomb; RE, chart recorder; RL, ring laser; V, voltmeter; W, wavemeter; WG, waveform generator.

and Io is the incident probe beam intensity. When 0 approaches zero, the polarizers are crossed and the last term in eq 8 yields a pure Lorentzian-shaped resonance curve. As 0 is increased by rotating the analyzer in either direction, both dichroism and birefringence anisotropies contribute to the shape of the curve. When the analyzer is rotated far enough so that s is insignificant compared to 0, the last term in eq 8 becomes negligible and hence a dispersion-shaped resonance curve resulting from the birefringent polarization rotation is observed, The observations are similar for a linearly polarized pump beam (17).

EXPERIMENTAL SECTION A schematic diagram of the experimental arrangement is shown in Figure 1. An argon ion laser (Control Laser, Orlando, FL, Model 554A) is used to pump a CW ring dye laser (SpectraPhysics, Mountain View, CA, Model 380A). The ring dye laser is passively stabilized to provide a single-frequency radiation (40-MHzjitter peak to peak) which can be electronically scanned over 30 GHz with selectable starting frequency and scanning rate. A 1/4 in. thick beam splitter is used to distribute the dye laser beam into three directions. One of the two weak beams (5% each) is sent to a wavemeter (Burleigh Instruments, Fishers, NY, Model WA-20) for laser frequency calibration. The other weak beam is sent through a polarizer and used as the probe beam. The third beam is used as the pump beam and overlaps the probe beam in the opposite direction. A pair of Glan-Thompsonprisms (Karl Lambrecht Corp., Chicago, IL, Model MGT-25-E8-90)serve as the polarizer and the analyzer. The analyzer is mounted in a rotational stage (AerotechInc., Pittsburgh, PA, Model ATS-301R) with a resolution of deg, and the polarizer is mounted in a homemade aluminum mount. The prisms are positioned 60 cm apart on an optical table (Newport Research Corp., Fountain Valley, CA, Model RS-410-8). A 6 cm long slot burner (Varian Techtron, Palo Alto, CA), mounted between the prisms, provides a laminar flame of acetylene and air. A Fresnel rhomb (Karl Lambrecht Corp., Chicago, IL, Model FR4-25-580) is used to produce a circularly polarized pump beam from the linearly polarized dye laser beam. A l-m focal length lens collimates the probe beam through the polarizer, the analytical flame, and the analyzer. The probe beam and the counter-propagating pump beam are aligned to cross each other at a few millimeters above the center of the slot burner, with a crossing angle as small as

NO. 1, JANUARY 1985 71

possible. The small collimated probe beam is completely engulfed within the pump beam volume, all along the length of the flame. A light trap catches the pump beam exiting the flame to eliminate undesirable scattered light. Apertures are used along the optical path between the analyzer and the photomultiplier tube to reduce the background light. The probe beam is detected after passing through a line fiiter by a photomultiplier tube (Hamamatsu Corp., Middlesex, NJ, Model R928) operating at 900 V supplied by a high-voltage power supply (Cosmic Radiation Labs Inc., Bellport, NY, Model lOOlB Spectrastat). The output of the photomultiplier tube is monitored by a nanoammeter (Keithley, Cleveland, OH, Model 160B) for the optimization of the extinction ratio by manipulating the analyzer and the probe beam alignment between the two prisms. To obtain a spectrum, the photomultiplier tube is terminated in a 100-kS1resistor, and the voltage is monitored by a lock-in amplifier (Princeton Applied Research, Princeton, NJ, Model HR-8),where a 1-stime constant is used. The analogue output of the lock-in amplifier is digitized via the Laboratory Peripheral System (LPS-11) of a PDP 11/10 minicomputer (Digital Equipment Corp., Maynard, MA). The 1/0 port of the wavemeter is also interfaced to the minicomputer for digitization of the laser frequency values. The computer takes simultaneous readings of both the polarization signal from the lock-in amplifier and the laser frequency value from the wavemeter every 0.5 s, and the real-time spectrum is displayed on a graphics terminal (Visual Technology Inc., Tewksbury, MA, Model 550). All polarization spectra are collected using the scan range of 30 GHz and the scan time of 100 s. Rhodamine 590 dye and Rhodamine 560 chloride dye (Exciton Chemical Co., Dayton, OH) are used for Na and Ba, respectively, to obtain about 250 mW af radiation. Barium stock solutions are prepared by dissolving barium nitrate crystals (Mallinckrodt Chemical Works, NY, analytical reagent) in triply distilled deionized water and stored in polyethylene bottles. Sodium stock solutions are prepared by dissolving sodium hydroxide electrolytic pellets (Fisher Scientific Co., Fairlawn, NJ) in quadruply distilled deionized water, stored in polyethylene bottles, and used the same day as prepared. A solution flow rate of 5 mL/min into the premixed burner is used throughout. The pump beam is polarization modulated by using an electrooptic light modulator (Lasermetrics Inc., Teaneck, NJ, Model 3030). The voltage necessary to control the modulator is supplied by a high-voltage op amp (BurleighInstruments Inc., Fishers, NY, Model PZ-70), which is in turn driven by a waveform generator (Wavetek,San Diego, CA, Model 162) using a 800-Hz square-wave function. The voltage amplitude of the gquare wave applied to the modulator is adjusted to produce 90° rotation of the polarization, and this is determined to be 140 V.

RESULTS AND DISCUSSION Tq show that we are in fact observing polarization spectra, we must verify the signal dependence on laser frequency, polarization direction, modulation voltage, laser power, and concentration. Figure 2 illustrates the experimental confirmation of eq 8 for a circularly polarized pump beam and a linearly polarized probe beam. Eleven experimental polarization spectra with different 0 values are collected while scanning the laser frequency across the Na D1 resonance line. As the analyzer is rotated away from the perfectly cross-polarized arrangement (0 = 0) to either direction, the first two terms in eq 8 become more significant. Hence, the dispersion-shaped reonance curves, produced by the birefringence, become more prominant. Although the dispersion-shaped signal is larger, the S I N ratio is actually poorer than the Lorentzian-shaped signal. So, for the analysis of trace Na and Ba, and for the determination and optimization of sensitivity, the Lorentzian-shaped curve is used. The dispersion-shaped signal, however, is useful for the locking of the laser frequency to some resonance line and for identifying some closely spaced components of the fine structure. Beyond the 0 values plotted in Figure 2, the magnitudes decrease, showing interference among the three terms in eq 8. The magnitude of the polarization signal is dependent on the angle between the probe and pump polarizations. The

72

ANALYTICAL CHEMISTRY, VOL. 57, NO. 1, JANUARY 1985

Table I. Relative Detection Limits Obtained by Polarization Spectroscopy and by Atomic Fluorescence Spectrometry (AFS)” detection limit, ng/mL

ref

Na this work laser AFS (graphite atomizer) laser AFS continuum source AFS

0.03 0.02 0.1 8

9 8 25

Ba this work wavelength-modulated laser AFS laser AFS pulsed hollow-cathode lamp AFS amplitude-modulated laser AFS Figure 2. Experimental Doppler-free dichroism and birefringence resonance curves for Na D1 line. Eleven polarization spectra with different 0 values are displayed with their corresponding laser frequency values.

Figure 3. Experimental observation of signal dependence on the initial polarization plane of the pump laser.

on-resonance signal can be maximized by supplying an appropriate light-modulator bias voltage so that the pump polarization is rotated to its optimum position. Figure 3 illustrates this characteristic of the polarization signal by using a linearly polarized pump beam and a counter-propagating linearly polarized probe beam., Nine resonance profiles for Na D1 line are displayed with their corresponding modulator voltages used for rotating the plane of the pump polarization to various initial positions. A constant voltage gap of 140 V is used as the modulation amplitude. Figure 3 shows that the magnitude and polarity of experimental profiles vary with the position of the initial pump polarization plane. A 90’ rotation of the initial pump polarization (an increase or decrease of 140 V in modulator voltage) results in the reversal of peak direction. The maximum absolute signal is produced with the polarization direction of the pump beam initially a t 45’ relative to that of the probe beam. Similar results are also observed by using a circularly polarized pump beam. In that case, the proper mqdulation amplitude and bias are needed to produce a purely circularly polarized pump beam to maximize the signal. T o achieve the highest S I N , it is essential to obtain the optimum extinction ratio between the crossed polarizers, pump-probe beam alignment inside the flame atomizer, and pump-probe polarization plane alignment as described above. The S I N improvement of polarization-modulated detection (PMD) over amplitude-modulated detection (AMD) is de-

37 20 12 200 300

16 13 25 16

“All methods used an air-acetylene flame unless otherwise indi-

cated.

termined to be a factor of 5. Amplitude modulation is achieved by using a mechanical chopper while maintaining the same modulation frequency and positions of pump-probe polarization planes as those of the PMD scheme. The S I N advantage of the PMD scheme results from the fact that when polarization is modulated, the background light intensity is not modulated. In addition, the average pump power is increased by a factor of 2 because of the lack of an “ofr‘ halfcycle. The gain through PMD should be even more in highly luminous atom sources or those with significant particulate scattering. The sensitivity also depends on the region of the analytical flame being used. The pump and probe beam positions are optimized by monitoring the on-resonance signal while moving the analytical flame vertically. The optimum region of the flame is found to be at about 4 mm above the burner slot, which is just above the inner cone of the flame. Various flow rates of acetylene and air were also tested for the flame, and the ratio 1:6 (acetylene/air) was determined to yield the highest signal. The polarization signal should depend on the square of the laser power in the experimental arrangement. For the same induced polarization rotation, a higher probe laser power will produce a higher transmitted intensity, as predicted by eq 8. But, a higher pump laser power induces a larger rotation, according to eq 5. The net effect is a quadratic power dependence. Laser power dependence is studied by using relatively high concentrations of analyte solutions (10 ppb for Na and 1 ppm for Ba). Polarization signal is observed even at the lowest dye laser power that could be maintained with lasing stability (Le., less than 10 mW). Quadratic response is observed for the lowest to the highest obtainable power of the dye laser. Linear response of the polarization signal is found for over 3 orders of magnitude in barium or sodium concentration. Similar quadratic dependence on laser power and linear dependence on concentrations are observed for both linearly and circularly polarized pump beams. The improved detectability of polarization rotation (better polarizers) and the excellent suppression of flame emission background (polarization modulation) enable this technique to achieve detection limits of 30 ppt sodium and 37 ppb barium ( S I N = 2). The potential sensitivity for sodium exceeds our ability to avoid contamination during the analysis or to simply find water pure enough to prepare the trace sodium solutions. The quadruply distilled deionized water used in this work is determined to contain 5 ppb sodium by standards addition. The sensitivity of this method is actually competitive with other highly sensitive laser spectroscopic techniques, such as the atomic fluorescence spectrometry. As shown in Table I, the detection limit of this method is com-

73

Anal. Chem. 1985, 57,73-75

parable to those using flame atomizers (16, 25) or graphite furnaces (9). Relative detection limits obtained by different techniques can only be compared roughly because of the many differences in experimental variables and instruments used. For example, the detectability for Ba was found to be 100 ppb in an air-acetylene flame compared to 8 ppb for a nitrous oxide-acetylepe flame (26). The best detectability for Ba AFS, 0.7 ppb, was reported for a nonresonant Ba ion line in the inductively coupled plasma (27). In any case, we have demonstrated that polarization spectroscopy has sufficient sensitivity to be suitable for trace elemental analysis. It is expected that still lower limits of detection can be achieved by improving experimental variables such as using more efficient optical filters and photodetectors, higher laser power, and high modulation frequencies where flame fluctuations and electronic noises of the laser system can be further minimized. Presently, the limiting noise is flicker noise in the probe laser. Ultimately, detectability in polarization spectroscopy will be limited by the signal level when compared to AFS. But, for highly luminous atom sources or those with substantial particulate scattering, the lower noise level in polarization spectroscopy is advantageous when resonant lines are used. In addition to its excellent sensitivity, polarization spectroscopy offers another advantage; Doppler-free spectra. It is especially useful when using atomizers which have relatively high temperatures. The analytical flame here provides a Lorentzian line width (fwhm) of 0.14 cm-’ (for Ba 553.6 nm). It is still narrow enough that this technique can be used for isotope ratio analysis for some hyperfine structures, using the spectral deconvolution scheme described in ref 19. The high spectral resolution also increases selectively and minimizes spectral interference. Finally, to become broadly useful, the laser must operate in the UV region. Pulsed dye lasers may be used in these situations. However, it is necessary to achieve a good beam quality to preserve the excellent extinction ratio of the polarizers. Such experiments will have to wait until the quality of UV polarizers are dramatically improved from those presently available.

Registry No. Na, 7440-23-5; Ba, 7440-39-3.

LITERATURE CITED (1) Fairbank, W. M., Jr.; Hansch, T. W.; Schawiow, A. L. J. Opt. SOC. Am. 1975. 65,199-204. (2) Geibwachs, J. A.; Klein, C. F.; Wessel, J. E. I€€€ J . Quantum Eleciron. 1978 1 4 . 121-125. -~ (3) Omenetto, N. ”Analytical Laser Spectroscopy”; Wiiey: New York, 1979;Chapter 4. (4) Omenetto, N.; Winefordner, J. D. Prog. Anal. At. Spectrosc. 1979, 8 , 1-183. (5) Bolshov, M. A,; Zybin, A. V.; Koloshnikov, V. G.; Vasnetsov, M. V. Spectrochim. Acta, Part 8 1981, 368, 345-350. (6) Pan, C. L.; Prodon, J. V.; Fairbank, W. M., Jr.; She, C. Y. Opt. Left. 1980, 5,459-461. (7) Faser, L. M.; Winefordner, J. D. Anal. Chem. 1972, 44, 1444-1451. (8) Winefordner, J. D. J. Chem. Educ. 1978, 55,72-78. (9) Boishov, M. A.; Zybin, A. V.; Smirenkina, I. I. Sp?ctrochim. Acta Part 8 1981, 368, 1143-1152. (10) Fraser, L. M.; Winefordner, J. D. Anal. Chem. 1971, 43, 1693-1696. (11) Kuhl, J.; Spitschan, H. Opt. Commun. 1973, 7, 256-259. (12) Smith, 8.; Winefordner, J. D.; Omenetto, N. J . Appl. Phys. 1977, 48, 2676-2680. (13) Green, R. 6.; Travis, J. C.; Keiier, R. A. Anal. Chem. 1976, 48, 1954-1 959. (14) Chester, T. L.; Winefordner. J. D. Spectrochlm. Acta, Part B 1978, 378, 21-29. (15) Kolzumi, H.; Yasuda, K. Spectrochim. Acta, Part 8 1976, 318, 237-255. (16) Goff, D. A.; Yeung, E. S. Anal. Chem. 1978, 50, 625-627. (17) Wieman, C.; Hansch, T. W. Phys. Rev. Lett. 1976, 36, 1170-1173. (18) Yeung, E. S.;Steenhoek, L. E.; Woodruff, S. D.; Kuo, J. C. Anal. Chem. 1980, 52, 1399-1402. (19) Tong, W. G.; Yeung, E. S. Talanta in press. (20) Smith, P. W.; Hansch, T. W. Phys. Rev. Left. 1971, 26,740-743. (21) Murnick, D. E.; Feid, M. S.; Burns, M. M.; Kuhi, T. U.; Pappas, P. G.

. -.

“Laser Spectroscopy IV”; Walther, H., Rothe, K. W., Eds.; SpringerVerlag: Berlin, 1979;pp 195-202. (22) Vasconceiios, J. 1. C.; Viiiaverde, A. 6.; Roversi, J. A. J . Phys. 8 1984, 77, 1189-1199. (23) Levenson, M. “Introduction to Nonlinear Spectroscopy”; Academic Press: New York, 1982;Chapters 1, 3,and 4. (24) Deisart, C.; Keller, J.-C. “Laser Spectroscopy 111”; Hail, J. L., Carlsten, J. L., Eds.; Springer-Verlag: Berlin, 1977,pp 154-159. (25) Winefordner, J. D. C/f..MT€CH 1975, 123-127. (26) Weeks, S. J.; Haraguchi, H.; Winefordner, J. D. Anal. Chem. 1978, 50. .., 360-36s ... . (27) Omenetto, N.; Human, H. G. C.; Cavaiii, P. Spectrochlm. Acta , Part B

1984,398, 115-117.

RECENEDfor review August 9,1984. Accepted September 14, 1984. The Ames Laboratory is operated by the U.S. Department of Energy by Iowa State University under Contract No. W-7405-eng-82. This work was supported by the Office of Basic Energy Sciences.

Mode Locking and Instability in a Voltage Thresholded Spark Source Alexander Scheeline* and David W. Kuhns School of Chemical Sciences, University of Illinois, 1209 West California Avenue, Urbana, Illinois 61801

Hysteresis In phase-locking behavior in the charglng of the capacitor in a high-voltage spark source is described. Aithough the mean number of spark flrlngs per power line half-cycle at maximum lnstabillty cannot be precisely determlned, It is demonstrably closer to a half-integer increment of the golden mean than it is to a half-integer. Implications In varlous areas of lnstrumentatlon are mentloned.

In a recent paper (11, the nonlinear behavior of the charging of the capacitor in a voltage-thresholded high-voltage spark source was described. The existence of a strange attractor

was demonstrated. It was suggested that at the point of greatest instability, the number of spark firings per half-cycle of the power line or the firing winding number would be an integer increment of the golden mean, (1 + 5lI2)/2 (=1.61803...). Justification for this assertion can be found in the literature (2,3). This contrasts with the intuitive value of simply a half-integer winding number for- the point of maximum instability (i.e., a random mixture of N and N f 1 firings or “breaks” per half-cycle) when the system is in transition between integer winding numbers. Herein we describe experiments and simulation which remove an experimental uncertainty from the previous paper and which measure the winding number at greatest instability. Hys-

0003-2700/85/0357-0073$0I .50/0 0 1984 American Chemical Society