1812
Anal. Chem. 1987, 5 9 , 1812-1815
Stray Light Rejection in Fiber-optic Probes Kristen J. Skogerboe and Edward S. Yeung* Ames Laboratory-USDOE
and Department of Chemistry, Iowa State University, Ames, Iowa 50011
Stray llght is one of the major limitations In flber-optlc probes toward achieving better accuracy and hlgher sensitlvlty. Natural reflection off surfaces of the optical components can be reduced by using an Index-matchlng fluld. The residual stray llght can be futther suppressed by hlgh-frequency moduletlon and phase dlscrknkralion, since llght returning from the fiber Is temporally separated from the lncldent radlatlon. By use of a pollshed frit to a b w longlludlnal flow, a cell wlth 37-nL volume and 3-mm optical path can be constructed to couple with a single 1 0 0 - m a m e t e r optlcal Mer. Improved performance for absorption measurements is demonstrated for dyes In a flow-tnJectlonmode.
A fast-growing and important area within analytical chemistry is the development of methods that are based on fiber optics. The popularity of optical fibers is due not only to their small diameter (5 pm to 5 mm) but also to the long distances that these fibers will carry light without attenuation. The small diameter of the fiber makes possible new areas of in vivo determinations (1)because they can be inserted into the body without damage to the tissue. In vivo fiber probes have been developed to dynamically monitor pH, blood gases, and glucose (2-4). The low attenuation properties of fibers make them useful for remote sensing in environments where measurement is otherwise difficult. Fiber optics can be used for on-line monitoring of an industrial process (5), for the remote analysis of groundwater (6),or for long-distance monitoring of radioactive materials (7). In general, fiber-optic probes can be classified into three categories: absorption, fluorescence, and reflectance. In each case, light has to be sent t o the region of interest, and light has to be collected and returned to some detector. The three common arrangements are shown in Figure 1. Of these, the single fiber probe (Figure IC)is the simplest and is best suited for miniaturization. It also avoids problems with cross-talk when two fibers (Figure 1b) are used. However, the single fiber probe is also the one most susceptible to stray light interferences. Specifically, natural reflection a t the ends of the fiber will contribute t o the measured light intensity. This is not so serious a problem if fluorescence is monitored, because filters can be used to discriminate between stray light and the real signal. A 600-pm-diameter in vivo fiber probe based on a single fiber has been developed (8). The probe was designed to produce a set path length by using a small mirror inside the needle which guides the fiber into the body. A large negative deviation from Beer's law is observed due to stray light present in the probe. The absorbance signal can be corrected for stray light by substracting the background signal of a highly absorbing reference compound from the sample absorbance (9). This approach is useful but is limited, for several reasons. The background correction decreases the precision of the absorbance measurement due to propagation of error through the subtraction. Also, changes in local environment, such as refractive index, are reflected in the amount of stray light; therefore, background correction is not complete. Finally, it is not well suited for on-line analysis, unless the correction is made simultaneously.
The development of a new absorbance fiber probe is described in this article. This spectrophotometric probe measures real-time absolute absorbance by minimizing the stray light contribution to the signal. This is accomplished via a coupling cell which matches the refractive index of the optical fiber to minimize reflections, and an optical delay line based on high-frequency modulation to further reduce stray light. The probe utilizes a 100-pm-diameter optical fiber so that small volume applications are possible.
EXPERIMENTAL SECTION The experimental arrangement to detect absorbance is shown in Figure 2. A He-Ne laser (Melles Griot, San Marcos, CA, Model 05-LPL-540) is modulated into two beams with a Bragg cell (Coherent Associates, Danbury, CT, Models 304 and 305D). The Bragg cell is driven by a signal generator (Wavetek, San Diego, CA, Model 162) at 850 kHz. An aperture of 0.099 in. diameter, made-in-house, was positioned to pass the first-order beam from in. diameter 3-cm the Bragg cell, blocking all other orders. A focal length lens (Oriel, Stratford, CT) was used to focus the first-order beam into the coupling cell containing one end of a 30-m length of optical fiber. This length of fiber was wound around a spool and the other end was mounted into the absorbance cell. The reflected light was collected with a 1in. diameter 5 cm focal length lens (Oriel, Stratford, CT) and directed to a photomultiplier (PMT) (Hamamatsu, Middlesex, NJ, Model R928), operated a t 500 V by a high-voltage power supply (Cosmic Radiation Labs, Bellport, NY, Model lOOlB Spectrastat). The output of the PMT is sent to a high-frequency lock-in amplifier with 50 52 input termination (EG&G,Princeton Applied Research, Princeton, NJ, Model 5202), and to a voltmeter (Keithley, Cleveland, OH, Model 155). The outputs of the lock-in (with 1 s time constant) and the voltmeter were sent to a computer (Digital Equipment, Maynard MA, Model PDP 11/10 with LPS-11 laboratory interface). All optical components are mounted on a 2 in. thick optical table (Newport Research Corp., Fountain Valley, CA, Model LS-48). Optical Fiber. The 30 m of optical fiber (Newport Research Corp., Fountain Valley, CA, Model F-MLD) had a 100 pm core and 140 pm core-plus-cladding diameter. The refractive index of the core was 1.4815. Approximately 5 in. of the protective coating on each end of the fiber was mechanically removed with a jacket stripper (Newport Research Corp., Fountain Valley, CA, Model F-STP), leaving a fiber of 140 pm diameter. The ends of the fiber were cleaved with a scribe (Newport Research Corp., Fountain Valley, CA, Model F-CL-1). To ensure that a clean, flat surface had been produced, each end of the fiber was examined under a microscope. Coupling Cell. The cell, made in-house, consisted of a brass housing with two windows machined 90' from each other. A 1 in. diameter Inconel beam splitter (Oriel, Stratford, CT) was mounted at 45' in the center of the cell. The cell windows were sealed by attaching microscope coverslips to the brass with epoxy. in. The optical fiber was mounted in the coupling cell with a 'I4 0.d. fiber chuck (Newport Research Corp., Fountain Valley, CA, in. opening, Model FPH-J). The chuck is inserted through a 'I4 and a liquid-tight seal is made with an internal O-ring. The cell was filled with cyclohexane (Fisher, Fairlawn, NJ). AbsorbanceCell. The other end of the fiber was inserted into a second fiber chuck, with a in. Swagelok nut attached to the end. The chuck was coupled to 'Il6 in. 0.d. chromatography tubing with a to ' I l 6 in. reducing union. The tubing, 180 pm i.d. stainless steel (Alltech,Deerfield, IL), forms the absorbance cell, the volume of which is determined by how far the fiber is positioned from the frit. For a 1.5 mm distance between frit and fiber,
0003-2700/87/0359-1812$01.50/0 CZ 1987 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 59, NO. 14,JULY 15, 1987
1813
RESULTS AND DISCUSSION Fiber Optic Coupling. In a conventional absorbance detector, stray light can be reduced to less than 0.1% by careful instrument design. For a fiber-based absorbance detector it is difficult to approach this standard. One source of spurious radiation comes from light reflected at the fiber interface as light is focused into the fiber. For any interface a difference in refractive index also causes scattered light. The amount of reflected light at an interface can be calculated by
C
Figure 1. Configuration for fiber optic probes: F1 and F2, optical fibers; R, reflecting surface; b, path length: (a) linear two-fiber probe, (b) bifurcated parallel probe, (c) single-fiber probe.
Flgure 2. Fiber optic absorbance probe and chromatographic system: P, pump; V, injection valve; CM, microbore column; CL, absorbance cell; FC, fiber chuck; OF, optical fiber; CC, coupling cell; BS, beam splttter; L, lens; A, aperture; M, mirror: OA, Bragg cell, Laser, HeNe laser: D, driver: W, square-wave generator, LI, lock-in amplifier; CP, computer; VM voltmeter; PMT, photomultiplier tube.
the cell volume is 0.037 pL with a path length of 3.0 mm. Standard replacement 2 pm pore size frits (Alltech,Deerfield, IL) included with the column were polished to yield greater reflectivity. The frit is first mounted flat in an acrylic resin and polished with 600 grit paper. Next it is polished with a vibratory polisher using a 0.3-pm aluminum oxide abrasive. Finally, the mount is dissolved with acetone and the frit is washed in an ultrasonic bath of acetone. The original frits were not replaced with polished frits for the chromatography columns used in this study. The chromatographic separation developed did not warrant a smaller cell. Therefore, the polished frit is seated in a 1/16 in. union with 3.5 cm of 180 pm i.d. tubing between it and the column frit. This extra length of tubing increases the total cell volume to 0.9 pL. Chromatography. The chromatographic system used consisted of a syringe pump (Isco, Lincoln NE, Model uLC-500) and a 1.O-pLsample loop coupled to an internal loop injection value (Rheodyne, Berkeley, CA, Model 7410). For reverse-phase chromatography a 25 cm X 1 mm 5-pm adsorbosphere C18 chromatography column (Alltech Associates, Inc., Deerfield, IL) is used with an eluant system consisting of, by volume, 75% MeOH (Burdick and Jackson, Muskegon, MI, HPLC Grade) and 25% aqueous buffer of 4 mM citric acid (Fisher, Fairlawn, NJ). The water is deionized and purified by a commercial system (Millipore, Bedford, MA, Milli-0 System). The pH of the eluant is adjusted to 7.4 with NaOH. Solutions of Bromocresol Green (Hartman-Leddon Co., Philadelphia, PA) were prepared in the eluant. For normal-phasechromatographya 25 cm X 1mm 5-pm microsphere silica column is used with cyclohexane as an eluent (Fisher, Fairlawn, NJ). Azulene (Aldrich Chemical Co., Milwaukee, WI) was dissolved in cyclohexane. All eluants were degassed under vacuum by ultrasonic agitation.
p is the percentage of light reflected and nl and n2 are the refractive indexes of the two mediums through which light is passing (IO). For the coupling interface, where light is focused into the optical fiber from air, reflected light is 4% of the total intensity. Depending on the exact geometry, there may be additional reflections due to the focusing lens and the beam splitter. With the refractive index of the coupling medium matched to the that of the optical fiber, the percentage of reflected light is significantly reduced. For an optical fiber (n = 1.48) in cyclohexane ( n = 1.43) the amount of reflected light, p, is 0.03%. This is small compared to other sources of stray light (scattering off imperfect faces and Rayleigh scattering), which cannot be compensated by index matching. So, the entire coupling cell (Figure 1)is filled with cyclohexane, and the lenses are placed outside the cell to minimize their contributions. It is sometimes possible to cut the entrance face of the optical fiber a t an angle away from the normal so that the reflected light is not retracing the optical path. However, to couple a large beam into a small fiber, focusing is needed, and this gives rise to a range of entrance angles to the fiber. Also, the returning beam from the other end of the fiber is spread over the acceptance angle (typically 60") and spatial discrimination between the two is very difficult. The index-matched coupling arrangement is a convenient solution. Proper design of the probe end of the fiber is also necessary. The key is to return as much light as possible for a given absorption path length. A high-quality reflecting surface (R in Figure 1) is desirable, but it should be compatible with a small probe volume to retain the benefits of fiber probes. For flow injection analysis or microcolumn liquid chromatography, a polished frit as described here allows good reflectivity and longitudinal flow. Figure 3 shows the surface quality of the frit before and after polishing according to the procedure described here. Substantial increase in reflectivity is observed. Polishing still does not allow returning a substantial fraction of the light to the optical fiber. This is because the light exiting the fiber is contained in a cone of about 60" (i.e., f/0.87).One can estimate the fraction of light, F, returned to the fiber, given total reflection by R (Figure 1) placed a t a distance l/zb from the fiber via
F = (&)(12) 4rb2
= 3-r2
b2
where r is the radius of the fiber. For an absorption path length b = 3 mm and a 50 Fm radius for the fiber, less than of the incident intensity is returned to the fiber. This is a much more important factor than the attenuation due to losses in the fiber itself, which is typically around 3 dB/km in optimal wavelength regions. This emphasizes the need to use high-intensity light sources (to produce a measurable signal) and the importance of stray light reduction in the optical design. Temporal Discrimination of Stray Light. To further reduce stray light, we use phase modulation to time resolve the analytical signal from the background. In this arrange-
1814
ANALYTICAL CHEMISTRY, VOL. 59, NO. 14. JULY 15. 1987
I
100
"rn
Flgure 4. Conblbutlng signals processed at the LIA. These signals are (a) the sbgnal h m Um abmbance cell. (b) the background signal due 10 stray light, 90° out of phase due 10 optical delay. and (c) lha sum of signals a and b.
FW- 3. M w a phomgaphs of (a)axwmimd trit and ib)polshed
hit.
menta long length of fiber is used as a delay line so that light that must pass through the optical fiber reaches the detector at a later time than light which did not pass through the fiber. The velocity of light in a fiber is given by Y = c/n, where c and n are the speed of light in vacuum and the refractive index, respectively. In this experimental arrangement we match the length of fiber to the modulation frequency so that the light that passes through the sample reaches the detector at 90' out of phase with the backscattered radiation. Then a lock-in amplifier (LIA) can be used to distinguish the two signals. The modulation frequency, /, can be related to the length of fiber, I, required to produce a 90- phase shift by the following equation:
/ = u/41
(3)
From eq 3 we can calculate that for 100 m of optical fiber (n = 1.48) the signal must be modulated at 500 kHz to produce a quarter-cycle phase shift. The function of lock-in detection in substracting the background signal is illustrated in Figure 4. In simple terms, the function of a LIA is to calculate the difference between separate phase regions within one cycle of a signal. If the LIA is used to differentiate two half-cycles from each other, as in Figure 4. a 90' phase-shifted signal is suppressed because it has equal components contributing to each phase region being compared. A phase shift of more or less than 90" does not give complete background cancellation. A convenient length of fiber for this work was 30 m. T h e u8e of longer lengths was excluded, due to handling difficulties and increased light loss. Shorter lengths of fiber could not be implemented due to pulse dispersion. which limits the maximum frequency of modulation. Since light has a double pass through the fiber, the effective length is actually 60 m. From eq 3 we calculate the theoretical modulation frequency required to achieve a 90° phase delay to be 844 kHz. Experimental verification of the modulation frequency was accomplished by positioning the fiber in the absorbance cell at
least 2 cm from the frit. At this distance light reflected a t the frit and collected by the fiber was not detectable, as predicted by eq 2. In this position the major sources of light arriving at the PMT are from the modulated reflection a t the cyclohexanefiber interface within the coupling cell and from the fiberabsorbance cell interface. This is the waveform that can be used to set the phase relationship in Figure 4. It is important to emphasize that this procedure is different from locking in on the modulated light being coupled into the fiber probe. By moving the fiber away from the frit, while maintaining the same refractive index environment, we are able to lock onto the true background signal. For the fiber in this position the difference of the signals in the Oo and 90" quadrants is indicative of the contribution of stray light. After the LIA is optimized in its phase relationship, the fiber can be moved to its regular position for absorption measurements. Absorption Measurements. The use of a chromatographic system here allows reliable introduction of small samples into the absorption cell without the uncertainties in optical alignment normally associated with cell placement. Noise in the base line can also be continuously monitored to allow optimization. Only single components were injected in these studies, so the chromatographic system functions as a controlled sample introduction device rather than a separation step. In the complete arrangement shown in Figure 2, modulated at 850 kHz, we found that the signal to noise (SIN) decreased for increasing path length. For a 3-mm path length we were able to achieve a base line stability of 2 x This is consistent with a previous report of S I N in fiber optic based probes (8).T h e noise level was not flow sensitive between the flow rates of 10-100 pL/min. We attempted to use a reference photodiode to adjust for laser intensity fluctuations but found that the ratio of the lock-in and reference signals offered no improvement in base-line stability. The intensity stability of the HeNe laser alone was 0.1%. Even though this represents considerable flicker noise, the main contribution to noise in the fiber probe was apparently from mechanical instability of the optical fiber. It is likely that the S / N will improve if a design for rigidly mounting the fiber spool is developed. If this major 8ource of noise is reduced, then flicker noise may become a factor. For measurements in small volumes, even with limited base-line stability, the fiber probe represents an improvement in detectability due to increased path length. By calculating the ratio of absorbance, A, to path length, b, we can assess the concentration detectability of the detector for a given absorber. To estimate the improvement we use as comparison a state-of-the-art capillary absorbance detector, for which the path length is perpendicular to eluant flow. The best detectability for a detector of this type has heen demonstrated with capillary electrophoresis with an absorbance detectablility
ANALYTICAL CHEMISTRY, VOL. 59, NO. 14, JULY 15, 1987
b-. W
0
20,
02
1.0
3 .o
5.0
1.0
3 .o
5.0
CONCENTRATION ( M x 1 o
.
- 1~
Figure 5. Beer’s law plots of (a) modulated and (b) nonmodulated BCG absorbance signals. Plots represent 95 % confidence intervals for three replicate measurements.
of 1 X lo-* absorbance units for an 8 0 - ~ mpath length (11). For that detector A / b is 1.3 X For our modulated fiber-based absorbance detector with a 3-mm path length this ratio is 6.7 X (SIN = 2). Therefore, even though the S/N is smaller for the fiber-based detector, the concentration detectability is actually better by a factor of 2, due to the increased path length. In order to assess the improvement made by the optical delay line in removing stray light, we compared absorbance measurements obtained with and without modulation. This is easily done using the arrangement in Figure 2. Bromocresol green was selected (BCG, 698 g/mol) for reverse-phase chromatographic studies because it was retained on a CIS column. Solutions of BCG in eluant were found to have a molar absorptivity of 38000 L/(mol cm) a t 633 nm. Verification that any difference is not due to the scale factor is accomplished by calculation of the total (integrated) peak absorbance in each case. A plot of total peak absorbance vs. BCG concentration for the LIA signal (with modulation) and voltmeter (without modulation) is provided in Figure 5. This plot indicates that not only are the total absorbance signals different but each type of signal has a different slope. It is important to emphasize that the data used to make these plots were obtained from the same injection and the same detector, with the only difference being how the signal from the photomultiplier tube was processed. The difference between the curves in Figure 5 is due to stray light. Phase-sensitive detection and an optical delay loop minimize negative deviations from Beer’s law. We see that when no modulation is used, stray light diminishes the absorbance signal. So, sensitivity is actually increased with modulation. Similar results are obtained when azulene (t = 333 L/(mol cm)) is used with normal-phase chromatography.
1815
The method we used to mount the optical fiber in the absorbance cell made it difficult to measure the path length accurately. We estimated it to be 3 mm based on translation of the fiber to the point where it is stopped by the frit. By using the slope of the Beer’s law plot for the modulated data obtained by linear regression, together with the experimentally determined molar absorptivity, we can calculate the path length of the absorbance probe. These calculations (for both BCG and azulene) show that b = 3.1 mm. Since the path length was not known exactly, we cannot conclude unequivocally that the modulated probe is measuring the absolute absorbance of the sample without interference from stray light. However, since the path length calculated from the slope confirms the estimated value of the path length, it is rea~ sonable to conclude that the absorbance calculated by the modulated optical delay method is very close to an absolute absorbance measurement. At the very least, this method demonstrates a marked improvement in stray light compensation for fiber probes. In summary, we have demonstrated a unique improvement in fiber optic absorbance probes. The probe minimizes stray light and essentially provides absolute absorbance detection. The combination of index-matched coupling of the light source to the fiber, modulation and phase discrimination of the absorption signal from stray light, and a polished frit to enhance reflectivity allows measurements in a cell volume of 37 nL with a 3-mm path length. Further improvements can be realized if the flicker noise in the light source and the vibration-induced noise in the optical fiber can be reduced. Similar improvements in accuracy and in sensitivity should also be possible in other fiber-optic probes, such as those based on fluorescence or reflection.
LITERATURE CITED Peterson, J. I.; Vurek, G. G. Science 1984, 224. 123-127. Peterson, J. I.; Goldstein, S. R.; Fitzgeraid, R. V.; Buckhold, D. K. Anal. Chem. 1980, 52, 864-869. Johnson, C. C. Biomed. Sci. Instrum. 1974, 1 0 , 45-50. Schultz, J. C.; Mansouri. S.; Goldstein, I. J. Diabetes Care 1982, 5 , 245-253. Eckbreth, A. C. Appi. Opt. 1979, 18, 3215-3216. Chudyk, W. A.; Carrabba, M. W.; Kenny, J. E. Anal. Chem. 1985, 5 7 , 1237-1 242. Klainer, S.; Hirschfeld, T.; Bowman, H.; Milanovich, D.; Johnson D. Earth Sciences Division Annual Report, LBL 11981, November 1980; Lawerence Berkeley Laboratory, Berkeley, CA. Coleman, J. T.; Eastham, J. F.; Sepaniak, M. J. Anal. Chem. 1984, 56, 2249-2251. Sharpe, M. R. Anal. Chem. 1984, 56, 339A-356A. Olson, E. D. Modern Optical Methods of Analysis; McGraw-Hill: New York, 1975; p 22. Walbroehl, Y.; Jorgenson, J. W. J . Chromatogr. 1984, 315, 135-143.
RECEIVED for review January 5,1987. Accepted April 16,1987. The Ames Laboratory is operated for 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.
