2124
Anal. Chem. 1982, 5 4 , 2124-2125
Double-Beam Fabry-Perot Interferometry as a Refractive Index Detector in Liquid Chromatography Steven D. Woodruff and Edward S. Yeung” Ames Laboratory and Department of Chemistry, Iowa State University, Ames, Iowa 5001 1
The refractive index (RI) detector will always be a useful tool in liquid chromatography (HPLC) because of its universal response. Problems however remain because of the generally poor detectability, incompatibility with gradient elution, and extreme sensitivity to temperature and pressure fluctuations. We have recently reported the use of Fabry-Perot interferometry for the simultaneous detection of RI and optical absorption in HPLC (I), with improved detectability over the corresponding conventional optical schemes. We now report an optical arrangement designed specifically for RI detection, this time based on double-beam Fabry-Perot interferometry. With this, we are able to further improve the detectability and to remove many of the constraints in the operation of the earlier system (1). Very briefly, the Fabry-Perot interferometer (2) produces maximum constructive interference whenever mh = 2nd, where d is the separation of the plane parallel mirrors, n is the RI of the medium, h is the wavelength of light in vacuo, and m is an integer. For a fixed A, scanning d (e.g., piezoelectrically) results in intensity maxima (governed by the above equation) that can be monitored photoelectrically. The locations of these maxima then trace out the refractive index. In a double-beam arrangement, the same light source provides two parallel optical paths through the same interferometer. The two beams travel through a reference and a sample flow cell, respectively, and are monitored by separate photoelectric detectors. In principle then, the true difference between the two beams can be obtained. In practice, because of misalignment of the mirrors and because of limitations in machining the flow cells, one cannot hope to have identical physical lengths, d, for the two beams. Fortunately, one can easily see that the quantity of interest is not n but An, the change in RI when the chromatographic effluent passes through the detector, and that A n l n = - A d / d , where Ad is the shift in the position for constructive interference. Any fractional mismatch in the physical lengths of the two optical paths will translate only into a fractional error in the amount of correction applied to A n as dictated by the reference beam. Typically, one can expect a mismatch of the order of 0.2 mm over a 10-cm path, so that certain types of noise can be reduced to the order of 0.002 compared to the single-beam case. Examples of these types of noise are laser frequency instability, room-pressure changes, thermal expansion of the interferometer mounts, of the flow cell, and of the optical components, and mechanical vibrations not in the plane containing the two optical paths. On the other hand, certain mechanical vibrations and any temperature and pressure differences in the two flow cells will not be corrected for. Still, one expects that temperature or pressure equilibration is easier to achieve than maintaining a specific temperature or pressure, so that there should be an overall improvement by using two beams.
EXPERIMENTAL SECTION The experimental arrangement is shown in Figure 1. A single-frequency HeNe laser (Tropel, Fairport, NY, Model 100) is split into two parallel beams by a 2.5 cm diameter, 0.6 cm thick uncoated optical flat with the surfaces polished to X/10 and parallel to 2 s (Oriel, Stamford, CN, A-43-143-60). When used in s-polarization and at 45’, this provides laser intensities of 11% and 9% for the reference cell and the sample cell. We note that the return beams to the laser are of the order of 1% and are 0003-2700/82/0354-2124$01.25/0
insufficient to affect the mode stability of the laser. This way, optical isolation used in the earlier scheme ( I ) is not needed here. The flow cells are parallel bores 1.6 mm in diameter machined out of a 10 cm long, 2.5 cm diameter aluminum cylinder. Chromatographic plumbing connected to the sample flow cell is wrapped three times and thermally contacted around the cylinder for added thermal equilibration. The windows are 1.0 cm thick fused silica with a surface finish of h/10 and antireflection coated on the glass-air surfaces (CVI, Albuquerque, NM, PW-0537Q). The use of a metal retainer and a flexible gasket provides a liquid seal with mechanical stability. Normal incidence is used, so that the cell need not be rotated with solvent change, as in the case of a Brewster’s angle geometry ( I ) . Additionally, the interferometer mirrors can be moved closer to the cell for better noise rejection. The interferometer used is of standard design (Burleigh, Fishers, NY, Model RC-110 with RC-670X2.3 mirrors). The two light beams are directed to two separate photomultiplier tubes (RCA, Lancaster, PA, 6342A and 1P28) operated at 800 V and 400 V, respectively, each fitted with an interferencefilter (Rolyn, Arcadia, CA, 66.4540) to reject room light. The optical components are mounted on a rigid table (Newport, Fountain Valley, CA, Model LS-48) without further vibration isolation. In operation, a linear ramp is generated by a minicomputer (Digital Equipment, Maynard, MA, Model PDP 11/10 with LPS-11 laboratory interface) and is amplified by a high-voltage operational amplifier (Burleigh, Fishers, NY, Model PZ-70) to scan the interferometer. For each step in the ramp, the output of each photomultiplier tube is digitized and stored in the computer. After each scan, the computer locates the constructive interference peaks at each phototube. The free spectral range of the interferometer is adjusted so that there is always one peak but never more than two peaks per scan for each beam. It is then easy to use an algorithm to monitor the difference in the two beams, regardless of whether one of the peaks in either beam has drifted off the scan range on consecutive scans. The cumulative difference in the two beams is converted to an analog signal and displayed on a chart recorder (Houston Instruments, Austin, TX, Model 5000). In this manner, the troublesome “resets” ( I ) can be eliminated.
RESULTS AND DISCUSSION The detectability allowed by the present optical scheme is limited by residual differences in the two optical paths, rather than the ability to measure Ad, since Fabry-Perot interferometers have been successfully used to determine laser frequencies to 3 parts in 10l1(3). By removing the cell from the interferometer, one can separately identify the contributions due to vibrations, room acoustic noise, and thermal effects on the interferometer. We found that these cause fluctuations in the 0.1-1.0 s time scale and are of the order of 1 X lo-* RI units. With the cell in place and filled with acetonitrile in the static mode, these short-term fluctuations are reduced by a factor of 5. This indicates that acoustic noise is the main contributor and is reduced when the amount of air space in the interferometer cavity is decreased. With the chromatographic system operating under the same conditions as given in ref 1,we found unpredictable long-term drifts in the base line. These were later determined to be characteristic of the servo control system in the particular pump used. Switching to a different pump (Milton Roy, Riviera Beach, FL, Model 196-0066) avoided this problem. The remaining fluctuations are monotonic and are in the 30-60 min time scale. For chromatographic purposes, correction using a polynomial fit to the base line ( 4 ) can be incorporated if desired. 0 1982 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 54, NO. 12, OCTOBER 1982 D1
0 D2 n
u
M
M
I
r-
I
I
u
Figure 1. Optical arrangement for double-beam refractive Index detector: L, laser; F, optical flat: M, interferometer mirrors; w, cell windows; R, reference flow cell; S,sample flow cell: D1, phototube for reference beam; D2, phototube for sample beam; dashed lines, optical paths.
Figure 2. sucrose, injected; factor of
Refractive Index chromatograms for a mixture of a-giucose, and raffinose, in the order of elution: (a) 14.4 Mg of each (b) 0.72 hg of each injected, with the scale expanded by a 20.
The results for a separation of a-glucose, sucrose, and raffinose with water as the eluent is shown in Figure 2. A five-point running average (equivalent time constant of 3 s) has been applied to the lower trace to further reduce the short-term acoustic noise. Also, base line correction using five points toward the beginning and the end of the chromatograms was introduced to allow comparison between the two chromatograms. This correction amounts to 1%and 25% of the corresponding peak heights for sucrose in the two chromatograms. From the lower trace, we conclude that the detectability for the system is 7 X lo4 RI units (SIN = 3). When acetonitrile was used instead as the eluent, the noise level is lower by a factor of 2. This is presumably because acetonitrile requires a lower pump pressure for the same eluent flow rate when compared to water. Using known concentrations of benzene in acetonitrile, we estimate a detectability of 4 X lo4 RI units (SIN = 3). The top trace in Figure 2 represents the same separation for an injection amount 20 times more than the lower trace and a t a scale expansion of 1/20. A linear response is thus obtained. It is interesting to note that the peaks in the upper trace represent shifts of the constructive interference peak over several free-spectral ranges. Our al-
2125
gorithm provides a direct count in m so that the true An is plotted. In practice, we find that concentrations more than a factor of 3 larger than that in the upper trace result in significant degradation of the beam quality, so that measurement is not feasible. Since the degradation is a direct consequence of the RI gradient at the start and at the end of each peak, a shorter length for the cell and/or a smaller cell diameter will reduce the effect. Naturally, one need not be concerned about the loss of detectability in such cases. With the present cell, the total linear dynamic range is thus 3 orders of magnitude. It is of concern that the above results are obtained with a high-quality interferometer and a sophisticated laser. However, the best detectability mentioned (4 X lo4 RI units) was obtained with a finesse of only 10, which was a compromise for good resolution in both of the light beams. One can thus expect similar performance with interferometers and mirrors of much lower quality than the ones used here. We have also found that spectral drifts and mode shifts in the HeNe laser during the initial warm-up period did not affect the chromatograms, confirming the utility of the double-beam arrangement. In fact, as long as the laser is stable during each scan period (20 ms), An is always monitored properly. We have inspected the spectral properties of several HeNe lasers in our possession, all between 0.5 and 4.0 mW and costing $150 to $700. The output typically contains two to three longitudinal modes, constantly competing with each other for intensity. Each mode, however, is as narrow as that of the single-frequency laser in the short term. The switching between intensities of the modes occurs in the l-s time scale. This means that a simple change in our algorithm will allow a particular mode to be followed for the purpose of determining An between the two optical paths. In fact, there is no reason why all of the modes cannot be used to provide a redundant determination. We fully expect routine HeNe lasers to be adequate for these measurements, especially if the scan rate of the interferometer is further increased to "freeze" the instantaneous mode structure of the laser. Our results demonstrate the clear advantage of the Fabry-Perot geometry vs. other interferometers. The minimum system finesse of 10 is easily achieved compared to the inherent limit of 2 for the Michelson, Mach-Zehnder (5), or Jamin (6) geometries. The use of the location of the interference maximum, as opposed to the inflection point, reduces any problems with the changing finesse during the chromatogram and minimizes the contributions of photon statistics to noise. The present system is limited by pressure and temperature differences in the two cells, but then these will limit any RI detector anyway.
LITERATURE CITED (1) Woodruff, S. D.; Yeung, E. S. Anal. Chem. 1982, 5 4 , 1174-1176. (2) Fowles, G. R. "Introduction to Modern Optics", 2nd ed; Holt, Rinehart and Winston: New York, 1975; Chapter.4. (3) Woods, P. T.; Shotton, K. S.; Rowley, W. R. C. Appl. Opt. 1978, 17, 1048-1054. (4) Cameron, J. M. In "Fundamental Formulas of Physics"; Menzel, D. H., Ed.; Dover: New York, 1960; Vol. I, Chapter 2. (5) Davis, C. C. Appl. Phys. Lett. 1980, 36, 515-516. (6) Cremers, D. A.; Keller, R. A. Appl. Opt. 1982, 27, 1654-1662.
RECEIVED for review May 27, 1982. Accepted July 15, 1982. 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.
