Anal. Chem. 1996, 68, 2255-2258
Characterization of a Simple Raman Capillary/Fiber Optical Sensor Vincent Benoit and M. Cecilia Yappert*
Department of Chemistry, University of Louisville, Louisville, Kentucky 40292
The characterization of a simple, dual-fiber quartz capillary/fiber optical sensor (C/FOS) for remote excitation and collection of Raman signals is presented. The Raman signals acquired with the C/FOS exhibit a 70-fold sensitivity enhancement and a 50-fold improvement in detectability relative to those obtained with the corresponding conventional dual-fiber sensor without the capillary. A background spectral feature at 790 cm-1 is related to the optical fiber background and is not due to the capillary tube. With no focusing lenses or filters needed at the sample site, the remote Raman C/FOS is easy to assemble and use, and it is relatively inexpensive compared to other designs. The development of fiber-optic sensors suitable for Raman scattering measurements has received increasing attention lately. The power of Raman spectroscopy in spectrochemical analysis lies in the richness of structural, qualitative, and quantitative information available in a single spectrum. However, the poor sensitivity of spontaneous Raman scattering spectroscopy poses great challenges in the design of sensitive, small, remote Raman sensors. Although the collection efficiency of single-fiber sensors is maximal due to the complete overlap of the excitation and emission cones,1,2 the use of this configuration for Raman sensors is limited by the high, often overwhelming, level of background scattering originating from the silica fibers.3 Multiple-fiber sensors in which the excitation and collected radiations are transmitted by separate fibers are less prone to background problems.4 Schwab and McCreery5,6 have reported the use of multiple-fiber sensors to collect Raman signals. The use of several collection fibers improves the sensitivity of the measurements. However, cost considerations could limit the use of these sensors in routine Raman analysis. Recent papers have discussed long-fiber-length, double-fiber remote Raman sensors.4,7,8 Due to the long lengths of the fibers, these designs require additional optical components such as lenses and filters to improve the sensitivity of the measurements and to reject strong background signals. This addition of optical components in the proximity of the sample not only reduces the (1) (2) (3) (4) (5) (6) (7)
Zhu, Z. Y.; Yappert, M. C. Appl. Spectrosc. 1992, 46, 912-918. Zhu, Z. Y.; Yappert, M. C. Appl. Spectrosc. 1992, 46, 919-925. Huy, N. Q.; Jouan, M.; Dao, N. Q. Appl. Spectrosc. 1993, 47, 2013-2016. Myrick, M. L.; Angel, S. M. Appl. Spectrosc. 1990, 44, 565-570. Schwab, S. D.; McCreery, R. L. Anal. Chem. 1984, 56, 2199-2204. Schwab, S. D.; McCreery, R. L. Appl. Spectrosc. 1987, 41, 126-130. Sharma, S. K.; Schoen, C. L.; Cooney, T. F. Appl. Spectrosc. 1993, 47, 377379. (8) Schoen, C. L.; Cooney, T. F.; Sharma, S. K.; Carey, D. M. Appl. Opt. 1992, 31, 7707-7715.
S0003-2700(95)01237-6 CCC: $12.00
© 1996 American Chemical Society
simplicity and convenience of the sensor but also increases its size. Early papers6,9,10 reported the enhancement in Raman signals obtained from samples contained in a capillary when the excitation was launched coaxially along the capillary. We also reported sensitivity enhancements of almost 2 orders of magnitude in the fluorescence signals acquired with capillary/fiber optical sensors (C/FOS) as compared to those obtained with conventional optical sensors.11,12 The earliest reports9,10 did not employ optical fibers. Schwab and McCreery,6 on the other hand, used 18 fibers concentrically arranged around the single excitation fiber and a 1-mm-diameter capillary up to 1 m long. Our sensor merely consists of a parallel double-fiber sensor coupled to a capillary tube. The capillary was shown to act as a conduit for the excitation and emission radiations, according to our partially reflective waveguide (PRW) model.12 In this paper, we report and characterize the analytical figures of merit of this simple, dual-fiber C/FOS for the remote excitation and collection of Raman signals. Because the PRW model developed for fluorescence emission is still applicable to the case of Raman signals, a sensitivity enhancement with respect to conventional sensors was expected. As the relative magnitude of the background signal is relevant to the detectability, the signalto-background ratio and the limiting sources of noise are presented and compared to those limiting the measurements obtained with conventional fiber sensors. EXPERIMENTAL SECTION Sensor Setup. The double-fiber optical sensor was constructed as previously described12 from 100-µm-diameter core, 140µm-diameter cladding optical fibers. The collection fiber was 1 m long, while the length of the excitation fiber varied between 1 and 15 m. The capillary used in the C/FOS was a 13-cm-long piece of the same 320-µm-i.d. fused silica tubing used previously.12 The brown polyimide coating was removed, and the resulting clear capillary was cleaned as reported previously.12 For the collection of Raman signals, the sensor was introduced inside the capillary, which was then filled with the sample solution through the use of a syringe connected to the distal end of the capillary by a piece of Teflon tubing. After spectral acquisition, the sample was removed by suction from the same end of the capillary. The conventional sensor measurements were carried out with the very same double-fiber sensor, which was retrieved from the capillary, held in a vertical position, and immersed in the sample solution contained in a 30-mL glass vial. (9) (10) (11) (12)
Walrafen, G. E.; Stone, J. Appl. Spectrosc. 1972, 26, 585-589. Ross, H. B.; McClain, W. C. Appl. Spectrosc. 1981, 35, 439-442. Zhu, Z. Y.; Yappert, M. C. Anal. Chem. 1994, 66, 761-764. Benoit, V.; Yappert, M. C. Anal. Chem. 1996, 68, 183-189.
Analytical Chemistry, Vol. 68, No. 13, July 1, 1996 2255
Figure 1. Raman spectra of PNP solutions and blank (water) obtained with a dual-fiber sensor without (A) and with (B) the capillary. See text for details.
Instrumentation. The 488-nm line of an argon ion laser (Innova 90-6, Coherent Inc., Auburn, CA) was used as the excitation radiation. The laser beam was focused onto the excitation fiber with a condensing lens. The radiant power at the tip of the sensor was about 15 mW. Signals transmitted to the distal end of the collection fiber were focused to the focal point of an f/4.2 spectrograph (Instruments SA Inc., Edison, NJ; Model HR-320) equipped with a 1800 grooves/mm holographic grating (Rd ) 0.75 nm/mm). The slit width was fixed at 100 µm. The Raman signals were detected by an intensified diode array (Tracor Northern TN-6122A) interfaced to a computer. Chemicals. The analyte used throughout the studies was p-nitrophenol (PNP) obtained from Aldrich. Aqueous solutions of concentrations ranging from 10-3 to 10-1 M were prepared by successive dilutions. The solutions did not absorb significantly in the spectral region of interest. RESULTS AND DISCUSSIONS Figure 1 shows the Raman spectra corresponding to the blank (distilled water) and PNP solutions of different concentrations, obtained with both sensor configurations (dual-fiber sensor without [A] and with [B] the capillary) with 1-m-long fibers. The traces represent the averages of 20 spectra. The acquisition time for each spectrum was 30 s with the conventional sensor (bottom traces) and 3 s with the capillary sensor (top traces). All the spectra are displayed on the same intensity scale. The spectrum of the 10-1 M solution of PNP obtained with the capillary sensor exhibited several bands corresponding to different vibrational modes of PNP. In particular, the strong band at about 1300 cm-1 was assigned to the symmetric stretching vibration of the nitro group. The intensity of this band was used as analytical signal throughout the quantitative analysis. Background Studies. To identify the source of the relatively strong band at about 790 cm-1 in the blank spectral traces, spectra were acquired using the conventional optical sensor without the 2256
Analytical Chemistry, Vol. 68, No. 13, July 1, 1996
capillary and with air as the sample. The band at 790 cm-1 was still detected. Furthermore, the spectral output of the excitation fiber was detected directly and revealed the same band. It is therefore concluded that this spectral feature is related to the optical fiber background and is not due to the aqueous sample or the capillary tube. Its intensity was related not only to the laser power but also to the amount of scattering taking place in the sample, as was observed by probing suspensions of calcium carbonate. The silica fiber background also comprised broad features close to the Rayleigh line. The rest of the background spectrum was fairly flat over the spectral region used. The values for the analytical signals used in the characterization study were obtained by baseline subtraction of the raw spectral traces. As long fiber lengths are required in remote applications, and since most of the background in the C/FOS described herein originated from Raman scattering of the excitation radiation within the excitation fiber, the effect of fiber length on the signal-tobackground ratio (S/B) was investigated. Raman spectra were acquired with excitation fiber lengths of 1, 5, 10, and 15 m and with the same 1-m-long collection fiber. No new background features appeared, even when a 15-m-long excitation fiber was used. Instead, when the fiber length was increased to 5, 10, and 15 m, the intensity of the silica background band was found to decrease to 64, 40, and 28%, respectively, of the signal obtained with the 1-m-long excitation fiber. This behavior suggests that the possible length-dependent increase in fiber background was not realized due to the attenuation of both the excitation radiation and the background radiation as they traversed the longer lengths of silica fibers. The decrease in the total excitation radiational power with fiber length was measured by detecting the output of the excitation fiber with a Si photodiode. The measured signal decreased to 89, 82, and 68% of the intensity measured with the 1-m-long fiber when the length was increased to 5, 10, and 15 m, respectively,
Table 1
capillary sensor conventional sensor
relative sensitivity
relative background
relative noise
S/N
S/Ba
S/Bkb
detection limit (M)c
73 1
26 1
6.4 1
62 5.4
10 3.7
6.6 0.2
1 × 10-3 5 × 10-2
a Signal-to-background ratio. b Signal-to-blank ratio. c Calculated as DL ) kσ /m, in which k ) 3 (z ) 3/21/2; 98.3% confidence level), σ bk bk is the blank standard deviation, and m is the slope of the calibration curve.
and was equivalent to 1.2 × 10-2 AU/m at the excitation wavelength (488 nm). This reduction in the overall excitation power was less pronounced than that observed in the background feature collected by the collection fiber. This is attributed to the larger degree of attenuation affecting the rays effectively collected by the collection fiber as compared to the entire cone of excitation. The overall absorbance of the excitation radiation represents the weighed attenuation of rays with different launching angles and thus with different number of reflections. The greater contributions correspond to those rays with very small entrance angle and thus less reflections.1 As the rays more effectively intercepted by the collection fiber are those with greater angle and thus more attenuation,2 it is reasonable to expect a more significant decrease with fiber length in the intensity of background feature than in that of the entire excitation radiation. Because of these attenuation effects, a decrease in the intensity of the analytical Raman band was also observed. The S/B for a 5.0 × 10-2 M PNP solution decreased from about 10 to 3.9 and 4.7 as the excitation fiber length was increased from 1 to 5 and 10 m, respectively. The precision of these values was affected by poor reproducibility in the experimental coupling of the laser radiation onto the excitation fiber as the fiber was cut to the different lengths. As this coupling affects the angular distribution of the radiation and consequently the collected signal, relative deviations of up to 35% in the S/B values were observed. Studies are underway to develop a quantitative model that relates these attenuation effects with fiber length and angular distribution of the excitation radiation. Analytical Figures of Merit. The analytical figures of merit for the conventional sensor and the C/FOS (1-m-long fibers) are shown in Table 1. The signals used in this comparison were obtained under identical experimental parameters, except for the acquisition time. Because of the lower sensitivity of the conventional sensor with respect to the capillary sensor and the limited dynamic range of the IDA, longer acquisition times (30 vs 3 s) were needed with the conventional sensor to obtain reasonable signals. Signal and noise levels in the spectra obtained with the conventional configuration were then corrected to compensate for the difference in acquisition times. The sensitivity enhancement provided by the capillary was found to be equal to 73. This value is close to that previously observed in the case of fluorescence signals.11,12 This was expected since the PRW model developed to describe the sensor behavior in fluorescence measurements is readily applicable to spontaneous Raman scattering. Thus, an increase in sensitivity by almost 2 orders of magnitude can be achieved when a capillary is coupled to the double-fiber sensor. It is important to mention that inner filter effects were negligible in this case, as the measured molar absorptivity of PNP was not greater than 0.13 L M-1 cm-1 over the spectral region of interest (488-530 nm). In the case of samples exhibiting significant absorption of the
excitation and/or emission radiations, the sensitivity enhancement is expected to be dramatically reduced. The blank (sum of background and dark signals) measurements obtained with the C/FOS were also characterized by a relatively higher level of background than in the case of the conventional sensor. In the latter, most of the blank signal (>95%) was due to the dark current, and the level of background was small. The greater background observed when the capillary was used originated from scattering of light at the capillary wall. However, the increase in sensitivity provided by the capillary was about 3 times greater than the corresponding increase in background level, and thus the capillary sensor exhibited better S/B. As expected from the different levels of analytical signal and background, both configurations also differed in the nature and level of limiting noises. In the conventional sensor, the limiting noise was that of the dark current. This seemingly ideal situation would suggest that, by increasing the laser power, one could achieve better signal-to-noise ratio (S/N). Although this is true theoretically, the increase in laser power could lead to sample photodecomposition and heating. In the C/FOS, it was the noise from the analytical signal that limited the sensitivity of the measurements. This noise included contributions from both shot and flicker noise associated with the laser source. Nevertheless, the C/FOS configuration showed an almost 8-fold improvement in the S/N versus that of the conventional sensor. A calibration curve was determined for the C/FOS. After collection of spectra from a solution of a given concentration, the capillary was emptied by suction and then rinsed with several capillary volumes of the subsequent solution. The calibration curve showed good linearity (r ) 0.9999) for concentrations ranging from 1.0 × 10-3 to 1.0 × 10-1 M. Theoretical detection limits (DLs) for the capillary/fiber and conventional fiber sensors were 1 × 10-3 and 5 × 10-2 M, respectively, and were calculated as the concentration that gave a signal equal to thrice the standard deviation or noise of the blank. For the capillary sensor and the conventional sensor, 20 spectra of distilled water were acquired using integration times of 3 and 30 s, respectively. The standard deviation in the signal corresponding to the diode at the analytical frequency was evaluated and used as the blank noise. To check the experimental detectability of the capillary sensor, a solution of PNP with a concentration of 1 × 10-3 M, equivalent to the DL, was prepared. The analytical signal could be easily differentiated from the blank, and its magnitude was about 3 times greater than the blank noise. This ability to experimentally reach the theoretical detection limit is, indeed, remarkable and suggests that, unlike in most fluorometric determinations,13 the sources of background signal can be effectively subtracted out from the total signal, thus leading to calibration curves with minimal losses of linearity near the limit of detection. (13) Ingle, J. D., Jr.; Wilson, R. L. Anal. Chem. 1976, 48, 1641-1642.
Analytical Chemistry, Vol. 68, No. 13, July 1, 1996
2257
In summary, the addition of a capillary to a conventional dualfiber sensor leads to improvements in Raman measurements of about 1 and 2 orders of magnitude in detectability and sensitivity, respectively. Furthermore, the sample volume is reduced by a factor of 6. CONCLUSIONS This study clearly shows the advantages of coupling a capillary to a fiber-optic sensor for the collection of Raman signals from liquid samples. The capillary serves as both the sample container and a conduit for the excitation and emission radiations. This results in an improvement in sensitivity and detectability, along with a reduction of the sample volume. The probe (sensor plus capillary) is of reduced size compared to designs described previously and could be further miniaturized through the use of smaller capillary and fibers. With no focusing lenses or special filters and with no need of alignment, the simplicity of the design makes this sensor ideal for in situ applications as well as for
2258
Analytical Chemistry, Vol. 68, No. 13, July 1, 1996
conventional analysis of liquid samples, normally performed with a right angle geometry. As stated earlier,12 trapping of light inside the capillary is most efficient when the outside medium has a low refractive index. If the sensor is immersed in water, then “leakage” of the radiation will occur and sensitivity will be lost. To alleviate this, we propose an alternate design in which the glass capillary is enclosed inside a concentrical metallic capillary of slightly bigger diameter so that both capillaries are separated by a layer of air. The sensitivity of the sensor would then be independent from the outside medium in which it is used, and the sensor would also become more rugged. Received for review December 22, 1995. Accepted April 16, 1996.X AC951237U X
Abstract published in Advance ACS Abstracts, June 1, 1996.
