Jason
Dickens
and
Michael
Sepaniak
*
Department of Chemistry, University of Tennessee, Knoxville, TN 37996-1600, USA
First published on 14th January 2000
A modular separation-based fiber-optic sensor (SBFOS) with an integrated electronically controlled injection device is described for potential use in remote environmental monitoring. An SBFOS is a chemical monitor that integrates the separation selectivity and versatility afforded by capillary electrophoresis with the remote and high sensitivity capabilities of fiber-optic-based laser-induced fluorescence sensing. The detection module of the SBFOS accommodates all essential sensing components for dual-optical fiber, on-capillary fluorescence detection. An injection module, similar to injection platforms on micro-analysis chips, is also integrated to the SBFOS. The injection module allows for electronically controlled injection of the sample onto the separation capillary. The design and operational characteristics of the modular SBFOS are discussed in this paper. A micellar electrokinetic capillary chromatography mode of separation is employed to evaluate the potential of the sensor for in situ monitoring of neutral toxins (aflatoxins). The analytical figures of merit for the modular SBFOS include analysis times of between 5 and 10 min, separation efficiencies of approximately 104 theoretical plates, detection limits for aflatoxins in the mid-to-low nanomolar range, and controllable operation that results in sensor performance that is largely immune to sample matrix effects.
The work described herein focuses on the fabrication and fundamental evaluation of a modular SBFOS that circumvents the aforementioned limitations of earlier designs. Rather than adapting existing components, the SBFOS is designed from the ground up to maximize performance and flexibility. This was accomplished with a modular design. The detection module was specifically designed to accommodate all essential sensing components for dual-fiber and on-capillary detection. Dual-fiber sensors have exhibited improved detection over single-fiber sensors.8 Moreover, the necessary critical alignments of the associated optical apparatus for dual-fiber sensors are much simpler than the critical optical alignments for single-fiber sensors. This former attribute is especially desirable in remote field monitoring applications, where it is necessary to align optical components reliably and routinely. The dual-fiber approach also allows for the implementation of different types of fibers, lens, filters, etc., to independently optimize the excitation and emission elements. An on-capillary detection approach was also implemented to minimize detector-related losses in separation efficiency.
Quantitative reproducibility and separation efficiency are often limited in CE by the nature of the sample introduction procedure. Sampling with the modular SBFOS described herein is accommodated by an integrated injection module. This module allows for electronically controlled sample injection onto the separation capillary of the SBFOS. This injection module is very similar to the injection platforms that are described in many CE microchip devices.9–14 Although these types of injection platforms are more difficult to employ in remote applications, their integration to an SBFOS is more desirable for in situ sensing over alternative sampling approaches (i.e. frontal mode or a flow injection approach). Because the injection module is a “fixed-loop injector”, a constant volume of sample can be injected. This capability has implications when considering sample matrix effects that are often problematic in in situ monitoring. A micellar electrokinetic capillary chromatography (MEKC) separation mode, wherein separations are based on differential association with a charged micellar phase,15 was employed to evaluate the potential of the sensor for in situ groundwater monitoring of neutral toxins (aflatoxins).
![]() | ||
Fig. 1 Schematic of rugged detection module and associated components of the modular dual-fiber SBFOS (A) and photograph of a constructed dual-fiber SBFOS detection module (B). |
The custom in-house fabricated detection module (1 in × 1 in × 1 in) consisted of two sections: a base plate (1 in × 1 in × 1/2 in) and a covering plate (1 in × 1 in × 1/2 in). In the base plate, numerous slits (refer to Fig. 1) were fabricated along the diagonal of the block that were 380 µm wide and 330 µm in depth. The slits served to house the fiber-optics and separation capillary and allowed for easy alignment of the fiber-optics with the detection window of the separation capillary. The depths of these slits were designed so that the separation capillary and the capillary-housed optical fibers could be secured in place with the covering plate. The slits for the fiber-optics were made along the diagonal of the square sensing module and thus a 45° angle relative to the separation capillary was created. This results in a 90° angle between the excitation and emission-collection optical fibers.
A 1/4 in diameter hole (1/8 in depth) was also made at the center of each section of the sensing module. This void space was created to minimize background noise contributions from module wall reflections. Additionally, this space facilitated the use of an index matching optical gel to minimize laser light scatter from the outer capillary walls. The final sensor component was the sensor-side reservoir/electrode assembly that was secured onto the detection module. This assembly consisted of a small 1 ml vial with an embedded platinum wire.
![]() | ||
Fig. 2 Electronically controlled injection device (A) and injection loading and separation processes (B). |
Although bench-top CE instrumentation with fiber-optic-based LIF detection has been established and is commercially available,6 our goal was to develop a robust SBFOS for remote monitoring in potentially hostile environments. To develop a robust yet sensitive SBFOS, the following design factors were implemented: (i) on-capillary detection to minimize detector-related losses in separation efficiency; (ii) dual-fiber detection to improve detection as previously discussed; (iii) ELP separation capillaries to improve detection sensitivity; (iv) refractive index matching gel to minimize optical background; (v) evaluation of the overlap of the excitation cone of radiation with the viewing cone of the signal-collection fiber within the detection zone for optimized detection; (vi) evaluation of collection efficiency for a variety of signal-collection fibers; and (vii) a detection module to ensure easy assembly and alignment of critical components.
Optical background and associated measurement noise are due to inadequate spectral rejection of the laser scatter and LIF or Raman emission of the two fibers and the capillary. The use of on-capillary detection can exacerbate the background problem. Fiber and capillary fluorescence are particularly problematic when UV lasers are necessary for appropriate analyte excitation, as is the case in this work. Optical scatter of excitation radiation from the capillary wall also results in smaller fluorescence signals as less excitation radiation impinges onto the detection volume. While conventional focusing optics external to the excitation fiber-optic could be used to focus the excitation radiation onto the detection volume, accommodating these optics within the detection module would be difficult and probably compromise the ruggedness of the sensor. Moreover, such optics are not readily available for UV operation.
In this work, it was determined that refractive index matching gel was very effective in minimizing background levels resulting from laser scatter at the capillary and, if present, scatter of fluorescence that is generated in the excitation fiber. For example, when measurements of the test solute Kiton Red were performed using He–Ne laser excitation (see Experimental), optical background levels were typically over 250 nA without gel while less than 20 nA with gel. The filter used here may not have been ideal in terms of rejecting laser scatter. Nevertheless, the reduction in background is substantial. Although not studied, the reduction in background levels with gel could be greater when UV lasers are employed and excitation fiber fluorescence is involved.
Three sensors were constructed that varied in the type of signal-collection fiber-optic (see Experimental). Calibration plots were constructed for injections of Kiton Red over the concentration range of 10−5–10−8 M. Response factors and LODs (S/N = 2) obtained from these plots are shown in Table 1. For sensors with signal-collection optical fibers that differed only in numerical aperture (SBFOS A, NA 0.22; SBFOS B, NA 0.66), it is not surprising that the response factors are only slightly better for the larger NA fiber. Although the acceptance cone of SBFOS B is much greater than that of SBFOS A, visual inspection of Kiton Red flowing through the ELP bubble, with excitation radiation impinging alternately from the excitation or emission fibers, revealed excellent overlap of the excitation radiation with the acceptance cones of both the NA 0.22 and 0.66 collection fibers. Moreover, the positioning of both fibers relative to the excitation zone within the ELP bubble is such that the collection efficiency was limited mostly by the small solid angle of fluorescence emission impinging on the distal end of the collection fibers and not by the critical angle requirements for transmission within the fibers (the latter being NA dependent).16 The converse is true for the large diameter fiber (SBFOS C) with its large solid angle of collection (see Table 1).
Sensor (NA and core diameter of collection optical fiber) | SBFOS A (0.22; 100 µm) | SBFOS B (0.66; 100 µm) | SBFOS C (0.22; 400 µm) |
Response factor (slope of calibration plot) | 25.2 | 32.2 | 267 |
LOD/M | 2.3 × 10−7 | 1.1 × 10−7 | 1.1 × 10−8 |
In most bench-top CE-LIF detection schemes, emission radiation is collected 90° from the excitation radiation. A 45° angle between the excitation or signal-collection optical fibers and the separation capillary was used in the current SBFOS. This configuration is easily and accurately fabricated by machining alignment slits along the diagonals of the square detection module block (see Fig. 1). Moreover, the 45° configuration results in a ∼35% increase in the optical path length across the detection window relative to collecting 90° to the separation capillary. ELP capillaries are often used to enhance detection sensitivity in CE. When collecting perpendicular to a 75 µm id capillary with a 3× ELP detection window, the optical path length across the center of the detection window is ∼225 µm. When collecting at 45° across the ELP detection window, the optical path length is ∼300 µm. Another advantage of the ELP capillary is that the wall thickness is reduced and hence there is less attenuation and/or fluorescence emission from the capillary itself.
Using SBFOS C, a calibration plot for fluorescein using the Ar-ion laser (see Experimental) was generated for injections over the range 1.0 × 10−7 to 6 × 10−10 M. The LOD (S/N = 2) was 5.1 × 10−10 M and the regression constant for the plot was 0.998. Based on an assumed injection volume of 10 nl, this corresponds to an absolute LOD of 5.1 attomol injected. This LOD compares reasonably well with most bench-top CE-LIF systems and is substantially better than previously reported single-fiber SBFOS prototypes.1,2
![]() | (1) |
Simple experiments were performed in order to validate the above treatment and to fine tune the system to account for the fact that the injection module includes an offset junction and not a true cross-junction. This was accomplished with a simple CE-LIF system consisting of 5∶3 mM phosphate–borate buffer, an He–Ne laser for excitation, and the laser dye Kiton Red as analyte. This simple system was used for initial characterization so that optimum conditions could be established prior to pursuing a more difficult separation mode (i.e. MEKC). Fig. 3 illustrates the peak response when the voltages across channels A and D are varied from 1.4 to 2.4 kV. As the potential increases, the volumetric flow rate into the offset injection loop increases (due to side channel flow) resulting in dilution of the injection volume and a decrease in the peak response. Based on these results, with a 3 kV primary voltage across the sample and sample waste channels, a secondary voltage of 1.4 kV across both the separation and running buffer channels was used for sample loading.
![]() | ||
Fig. 3 Effect of loading conditions. Loading conditions: VBC = 3 kV; voltage at channels A and D varied. Separation conditions: VAD = 11 kV; VB = 4 kV; VC = 4 kV. |
Examination of the separation process was also performed. Problems with sample leakage and inadequate flow are also solved with appropriate secondary potentials. When no secondary potentials are applied to side channels B and C (i.e. left floating) during the separation step, significant flow into these channels occurs and a lower effective flow rate results in the separation channel. Fig. 4A illustrates the resulting electropherogram that exhibits a long migration time. For proper fluid control, a secondary voltage across the side channels (channels B and C) must also be applied during the separation process. Although a side channel voltage that exactly matches VJ may produce an ideal flow rate in the separation channel, diffusive and hydrodynamic effects result in some sample leaking from the side channels into the separation channel during the separation. Under these conditions, peak shapes as seen in Fig. 4B often occur. Secondary voltages across the side channels that are slightly less than VJ will prevent this effect by causing a slight flow from the junction into the side channels while still maintaining adequate flow in the separation channel. Fig. 4C demonstrates the resulting electropherogram when the optimum side channel secondary voltages are applied. Under established optimum conditions for the loading and separation steps, the standard deviation for peak area reproducibility for seven consecutive injections of Kiton Red was 1.8%. Because the pinched injection experiments were conducted manually (i.e. without high voltage switching), improved reproducibility should be possible if the system is fully automated.
![]() | ||
Fig. 4 Effect of floating channels B and C during separation (A), conditions where side channel leakage occurs during separation step (B), and conditions where applied side channel voltage is approximately optimum (C). Loading conditions: same as in Fig. 3 (VAD = 1.4 kV). Separation conditions: VAD = 11 kV; VB = float; VC = float (A); VAD = 11 kV; VB = 5 kV; VC = 5 kV (B); and VAD = 11 kV; VB = 4 kV and VC = 4 kV (C). |
The established optimum loading and separation conditions were initially employed for the MEKC separation of aflatoxins B1, B2, G1 and G2. The aflatoxin samples were prepared in the SDS-based running buffer to maintain a uniform buffer system throughout the integrated SBFOS system. Upon establishing adequate separation of all components, these conditions were then used to construct a calibration plot from 3 × 10−6 to 1 × 10−8 M and to determine LODs for each of the aflatoxins. The LOD for each aflatoxin was based on a peak height that resulted in an S/N ratio that was twice the baseline noise (see Table 2). These LODs are well below the ppb level concentrations allowed in the feedstocks of most countries21 and thus should be adequate in trace environmental monitoring applications.
Aflatoxin | Minimum detectable concentration | Minimum detectable amounta | Correlation coefficient (r2) |
---|---|---|---|
a Minimum detectable amount based on 15 nl injected. | |||
G2 | 8.6 × 10−9 M (2.8 ppt) | 43 fmol (14 pg) | 0.997 |
G1 | 3.1 × 10−8 M (10 ppt) | 158 fmol (50 pg) | 0.998 |
B2 | 7.2 × 10−9 M (2.3 ppt) | 33 fmol (10 pg) | 0.997 |
B1 | 1.5 × 10−8 M (4.7 ppt) | 71 fmol (22 pg) | 0.998 |
In real world in situ monitoring, a sample such as groundwater will obviously not contain the SDS-based running buffer. In previously described electronically controlled injection studies, the running buffer and sample solvent were equivalent, and thus resistance per unit capillary length was constant throughout the SBFOS. A large difference between the sample's conductivity and the running buffer's conductivity, as is the case in in situ groundwater monitoring, will result in a significant change in the resistance along the sample and sample waste channel during the loading step. This will result in a change in the junction voltage during the loading step.
Applying the same circuit analysis as described above, the integrated SBFOS system can be modeled under the in situ conditions to determine appropriate secondary potentials. In establishing a realistic model, the ionic strength of common regional groundwater samples was determined to be roughly equivalent to a 1∶0.6 mM phosphate–borate running buffer.
Additionally, a typical 5∶3 mM phosphate–borate running buffer is assumed. When a 1.5 kV secondary voltage is applied during the loading step under in situ conditions (i.e. groundwater monitoring), VJ decreases from 1.5 kV to 0.8 kV as channels B–C fill. Under these conditions, significant leakage of running buffer from the separation and running buffer channels is expected as observed previously (refer to Fig. 3). If the secondary voltage is decreased to 1.0 kV and the equivalent circuit is modeled again, the junction voltage decreases from 1.32 kV to 0.75 kV. In this situation, upon initial application of the primary and secondary voltage, running buffer leakage from the injection loop initially occurs. However, by the time the sample reaches the offset injection loop, VJ decreases to a voltage approximately matching the secondary voltage and negligible leakage of sample should occur, i.e. a constant volume of sample should be injected.
Figs. 5A–5C illustrate the monitoring of aflatoxins in two regional groundwater samples and an approximately equivalent ionic strength running buffer (i.e. 1∶0.6 mM phosphate–borate buffer) spiked with 3 × 10−7 M of each aflatoxin. An essential factor in determining appropriate conditions for in situ groundwater monitoring is the approximation of the ionic strength of the groundwater. Because the groundwater sources investigated above were all collected within the same geographical region, their ionic strengths are relatively equivalent, and thus the same operating conditions were utilized for each sample. From source to source, the peak areas are approximately the same (within 3%). The main source of error from sample to sample is the variability in the resulting junction voltage. Because the ionic strength of each sample will not be perfectly matched to the approximate ionic strength of the sample used in the developed model for in situ monitoring, a mismatch between the secondary voltages and the actual junction voltages is likely. This would result in either leakage or dilution of the sample loop, depending upon the difference in the ionic strengths between the approximate equivalent running buffer and the sample. To determine appropriate in situ operating conditions for operation in a variety of environmental sites, a conductivity probe followed by the determination of the appropriate secondary voltage would be necessary to avoid the aforementioned effect. Additionally, generating calibration data based on standard solutions prepared at typical ionic strengths for groundwater would be necessary for adequate analyte quantification.
![]() | ||
Fig. 5 Monitoring of aflatoxins in local river samples and equivalent buffer spiked with 3 × 10−7 M G2, G1, B2, and B1. Limestone Creek (Washington County, TN) (A), Tyson Creek (Knox County, TN) (B), and equivalent phosphate–borate buffer (C). |
This journal is © The Royal Society of Chemistry 2000 |