Design and performance of a light-emitting diode detector compatible with a commercial capillary electrophoresis instrument

Indirect photometric detection in capillary electrophoresis (CE) has been predominantly performed in the UV region, in part due to a lack of suitable high-intensity and low-noise light sources in the visible spectral region. A new photometric detector based on light-emitting diodes (LEDs) as light sources and compatible with a commercially available CE instrument has been designed and constructed and its performance evaluated. The utility of this detector was successfully demonstrated by the indirect photometric detection of anions using a dye as probe and absorbance measured in the visible region. The detector exhibited very low baseline noise (around 0.03 mAU), stable output, and improved upper limit of detection linearity (502 mAU) compared with previously used LED detectors. The detector was tested for indirect detection of anions separated with an electrolyte containing 4m M Orange G as the indirect detection probe, 10m M histidine as an isoelectric buffer, and 0.05% hydrox-ypropylmethylcellulose to suppress the electroosmotic flow. Extremely low detection limits were obtained ranging from 0.16–0.36 m M (excluding chloride 0.56 m M ), with separation efficiencies in the range of 154 000–274000 theoretical plates.


Introduction
Indirect photometric detection (IPD) is a well-established detection mode for capillary electrophoresis (CE) to detect nonabsorbing analytes [1][2][3][4][5][6]. IPD is performed by the addition to the background electrolyte of an absorbing ion of the same charge as analytes, typically referred to as a probe, and monitoring the absorbance at a wavelength at which the probe absorbs strongly. Although some probes are available that absorb strongly in the visible spectral region, IPD has generally been restricted to the use of probes which absorb in the UV region. This can be explained by the fact that the typically used light sources in CE instruments, such as deuterium and mercury lamps, emit strongly in the UV region, but their relative emission intensity in the visible spectral region is lower by about an order of magnitude or more [7]. At low light intensities, the baseline noise is indirectly proportional to the light source intensity [8,9]. As a result, baseline noise in photometric detection in CE depends strongly on the intensity of the used light source.
Several approaches are possible to improve the signal-tonoise ratios and consequently the limits of detection (LODs) of indirect photometric detection. It has been clearly shown [10,11] that the most effective means of increasing sensitivity and thereby decreasing LODs is to increase the absorptivity of the probe. The careful use of highly absorbing probes, such as dyes [11][12][13][14], has been shown to provide highly sensitive and efficient detection. However, because of the light source intensity considerations explained above, detection has usually been performed at lower wavelengths in the UV region where the source light intensity was highest in order to achieve an optimal signal-to-noise ratio. One approach at optimizing the potential benefits of using highly absorbing probes is to use a light source with better characteristics in the visible region. Light-emitting diodes (LEDs) have found some use for this purpose [13,[15][16][17][18] because they emit strongly in the visible region (and for some recent examples, also in the high-UV region [17]), have near monochromatic output, are highly stable, have long lifetimes, and are very affordable. Unfortunately, there are currently no commercially available LED-based detectors for CE. As such, instruments have been modified to allow the use of LEDs as light sources.
Tong and Yeung [19] used a combination of camera and ball lenses to focus light from a LED through a capillary. A camera lens was used to focus the light down to a 1 mm diameter spot. A pair of ball lenses (from a commercial CE instrument) was used to further focus the light source onto the detection window. Inorganic anions were detected indirectly using permanganate as a probe at 565 nm, with LODs in the order of 10 25 M being achieved. Bruno et al. [20] used a gradient index lens attached directly to a LED light source. Noise levels in the order of 10 25 AU were quoted, but no CE separations were provided. Boring and Dasgupta [21] designed a detector for CE that could be fitted with conventional light sources or LEDs. Optical filters were used for wavelength selection and the performance of the detector with an LED light source in terms of noise and linearity was superior to UVemitting light sources.
Macka et al. [15] studied the use of LEDs with emission wavelengths between 563 and 654 nm as light sources. A Waters Quanta 4000 instrument was modified to allow a mercury lamp holder to be adapted to hold a LED light source. Noise levels were lower than with traditional light sources. A baseline noise of 3610 25 AU was observed with a 75 mm capillary filled with water during the application of voltage. Direct detection of metal-Arsenazo I complexes was performed using LEDs as light sources. This instrument was subsequently used for indirect photometric detection with naphthol yellow S [16], Orange G [13], and chromate at 370 nm [17] as probes. Direct detection of a series of metals using metallochromic ligands was also performed. Using this approach, barium and strontium were separated and detected at 654 nm using sulfonazo III as the ligand, with low LODs of 0.35 mM and 0.47 mM, respectively, attributable chiefly to the very low baseline noise level of 2.8610 25 AU [22]. Uranium(VI) and lanthanides were separated by on-capillary complexation with arsenazo III by Macka et al. [23], with detection at 654 nm. A study of solute-wall interactions using this system was also performed [24]. Vachirapatama et al. [18] achieved direct detection of vanadium-PAR (4-(2-pyridylazo)resorcinol) complexes at 568 nm. LODs of 19 ppb were reported and analysis of vanadium in fertilizer samples was performed.
Butler et al. [25] used fiber optic coupling between the LED, capillary, and photodiode detectors (signal and reference), which eliminated the need for focussing optics. Direct detection of metal-PAR complexes at 525 nm was shown and indirect detection of anions and cations was also performed. Chlorphenol red was used as a probe for anions and pyronine G was used for cations, with detection at 525 nm being used in both cases. This detector performed well in terms of stray light and baseline noise levels. A baseline noise level of 30 mAU (100 mm ID capillary filled with water, no voltage applied) was recorded. LODs for indirect detection were not provided.
In this study, we present a LED-based detector that has been designed for use in conjunction with an Agilent 3D CE instrument. The LED light source can be easily interchanged as no modification of the LED or coupling with the capillary is required. This detector provides increased upper detector linearity limits in comparison with previous LED-based detectors. Baseline noise levels and performance during indirect detection are also examined.

Detection cell and electrical circuit
The detection cell is depicted schematically in Fig. 1. A two-piece, black nylon housing was designed and constructed to house an Agilent alignment interface. Each half of the housing had a 10 mm ID sleeve to allow the LED light source or detector to be fixed in line with the detection window of the capillary. The 5 mm diameter LED light source could be mounted in holders with pinhole slits of varying diameter. An Agilent cassette was modified by   adding support columns to hold the housing about 70 mm before the photodiode array detection point. A photograph of the detection cell mounted in the cassette is shown in Fig. 2. The LED current and detector signal (input absorbance) were varied and measured by a controlling unit. An absorbance offset value could also be changed using this unit, and the resultant output absorbance was transferred to a 35900 ADC (Agilent Technologies, Waldbronn, Germany) converter, which interfaced the LED detector and the ChemStation software. The detector housing, electronic circuits, and controller were designed and constructed in-house as follows. A hybrid photodetector consisting of a photodiode and amplifier (Integrated Photomatrix Ltd., Dorchester, England) was employed as a detector. A feedback circuit provided an output voltage which was proportional to the incident light level, i.e. output voltage increases with increasing light level. This voltage was fed into a logarithmic amplifier (LOG101 integrated circuit; Burr-Brown, Tucson, AZ, USA). The output of the logarithmic amplifier was then fed into an offset circuit, which allowed the output level to be adjusted to any desired value. The blue LED used during the majority of this work was obtained from Oatley (NSW, Australia, ELN5B, emission wavelength 476 nm). The intensity of the LED was controlled by a constant current circuit (which was typically set at 30 mA). In order to minimize pickup of electrical noise from the high-voltage power supply, the signal from the detector was connected to the logarithmic converter with a screened cable and the metal body of the detector had a separate drain wire that connected to a copper-shielded box that enclosed the logarithmic converter.

Instrumentation
Separations were performed with an Agilent Technologies 3D CE apparatus. Capillaries of 46.0 cm (total length), 31.0 cm to the LED detector, 37.5 cm to the diode array detector were cut from a 60 m spool of fused-silica capillary (Polymicro Technologies, Phoenix, AZ, USA) of 50 mm inner diameter, 365 mm outer diameter. Capillary detection windows were created by burning a small portion of polyimide coating off the capillary using a butane torch. The capillaries were installed and aligned into an Agilent alignment interface as recommended by the manufacturer, and the interface was mounted in the detector housing. The assembled cell was installed into an Agilent capillary cassette and inserted into the instrument. Spectrophotometric measurements were conducted using a Cary UV-Vis-NIR Spectrophotometer (Varian Australia Pty. Ltd.) with 1 cm pathlength quartz cells.

Procedures
Fused-silica capillaries were semipermanently modified with PEI by flushing with 1 M sodium hydroxide for 30 min, water for 30 min, followed by a 4% PEI solution for 1 h which was left to stand in the capillary for 30 min. The capillary was finally flushed with water for 30 min before use. Orange G was purified as reported previously [13,26] by recrystallization from ethanol-water. A series of standards for detector linearity measurements was prepared by serial dilution by a factor of two of a stock solution of Orange G. A blue LED was installed as the light source (see Section 3.1). Absorbance measurements at 476 nm were performed by flushing the capillary with water or the desired standard solution (approx. 10 capillary volumes), then stopping the flow and measuring the absorbance under static conditions. The absorbance of each test solution was measured in triplicate. Absorbances were measured in order of increasing concentration standards to minimize possible carry-over errors. During measurements used to test the detector stability, the first 10 s of each run (500 data points) were discarded. Separation efficiencies were calculated based on the peak width at half-height. LODs were calculated at three times the baseline noise.

Detector design and general considerations
An Agilent alignment interface was used in this work as it provided a convenient way of holding the capillary in position so that the detection window could be positioned correctly between the light source and detector. It should be noted that by utilizing an original Agilent alignment interface the need for potentially demanding technical design and construction could be avoided and the whole detector was kept relatively simple and straightforward.
In order to characterize any new detector and compare its performance with other detectors, a range of parameters need to be measured. First, it is necessary to investigate the detector linearity. In photometric detection in CE using LEDs it has been shown that it is the stray light (light which reaches the detector without passing through the detection window) that is the key parameter which mostly determines detector linearity, rather than a lack of an ideally monochromatic light source [15]. While a number of approaches can be used to evaluate detector linearity [15,[27][28][29], in this study we followed the approach of Macka et al. [15] by measuring the response (absorbance) of a series of standard solutions in order to calculate sensitivity (absorbance/concentration). Sensitivity was then plotted against absorbance to determine the absorbance value at which sensitivity began to decrease from its maximum value. In this work, the new detector was tested with a 50 mm ID fused-silica capillary which provided a more stringent test of performance than if a 75 mm ID fused-silica capillary were used. Second, it was important to determine the stability and reproducibility of the detector output. Third, performance of the detector under real conditions of usage in a suitable CE separation system needed to be evaluated. The results obtained for the above parameters are presented in the sections below.

Detector linearity
Initial testing of the detector concentrated on determining its upper linearity limit and approaches by which this limit may be maximized. Orange G was chosen as the test absorbing compound as it absorbs strongly at 478 nm, which is an ideal wavelength for use with a blue LED. Data were collected as detailed previously [30] by measuring the absorbances of a series of Orange G solutions of increasing concentration. Sensitivity values (absorbance/ Orange G concentration) were calculated from the measured absorbances and plotted against absorbance. The absorbance at which sensitivity decreased by 5% from its maximum value was used to define the upper limit of detector linearity.
A 5 mm diameter LED light source was mounted in a black plastic housing with a 1 mm diameter pinhole, and a 50 mm ID capillary was installed in a green Agilent alignment interface. Under these conditions, an upper detector linearity limit of 502 mAU was determined from the sensitivity versus absorbance plot shown in Fig. 3. A larger 2 mm pinhole slit allowed more light to pass through, but led to an increased level of stray light level and no measurable improvement in linearity. On the other hand, a smaller 0.5 mm pinhole slit reduced incoming light too severely, resulting in a markedly increased noise level. Similar results occurred when a focusing lens was used in an attempt to narrow the incident light beam. The experimental linearity limit obtained with the LED is less than the upper detector linearity limit of an Agilent 3D CE with a 50 mm ID capillary installed using the in-built deuterium lamp and diode array detector (785 mAU [13]) but is significantly higher than previous results with LED light sources, which are typically around 100-200 mAU [13]. The concentration of Orange G which corresponded to the upper detector linearity limit was calculated to be 6.5 mM. This is a significant increase over a previous use of this probe with a LED detector, for which a maximum concentration of 0.5 mM was possible [13]. This allowed the use of more highly concentrated electrolytes for indirect photometric detection without loss of detection sensitivity. The noise level obtained under static conditions without applying voltage was approx. 0.02 mAU, which is comparable with previously reported noise values. An estimate of the effective optical pathlength can be made by rearranging the Beer-Lambert law (effective pathlength = sensitivity/absorptivity). This resulted in a value of 39.58 mm, which is only a little less than the effective pathlength of 44.99 mm observed with a 50 mm ID capillary and diode array detector.

Detector stability
In order to examine the stability of this detection configuration, a capillary was flushed with 4 mM Orange G at a pressure of 50 mbar (linear flow velocity of 0.1 cm/s) over a 5-min period. The signal was sampled at 50 points/s, with a total of 12 runs being performed. Two parameters were calculated to quantify the detector stability. First, the baseline drift was calculated as the absolute difference in absorbance measured at the start and end of the 5-min period. Second, the baseline fluctuation was calculated as being the maximum absorbance difference between any two points (not necessarily consecutive points) in a run. This parameter therefore represents the largest possible change in the baseline absorbance. These two parameters are plotted in Fig. 4 for each of the 12 runs. Baseline drift varied between 0.006-0.047 mAU (average drift 0.024 mAU) and baseline fluctuations occurred between 0.078-0.104 mAU (average fluctuation 0.094 mAU). Short-term noise (over 0.1 min intervals) was approx. 0.02 mAU. These results indicate that the detection system was both stable and robust.

Performance under real conditions
An electrolyte consisting of 4.0 mM Orange G, 10.0 mM histidine, and 0.05% HPMC was chosen to demonstrate the performance of the detector and light source. This system has been successfully used previously for the indirect detection of anions [13]. The probe, Orange G, has a maximum absorptivity of 19,511 L?mol 21 cm 21 at 478 nm. A PEI-coated capillary was used to minimize adsorption of the probe onto the capillary wall. The addition of HPMC suppressed the electroosmotic flow (EOF) and helped to further minimize adsorption and improve baseline stability. From previous work, it has been shown that maximizing the probe concentration leads to enhanced detection sensitivity [12,13]. A concentration of 4 mM Orange G resulted in a background absorbance of 310 mAU, which is less than the upper detector linearity limit of the detector configuration. Higher probe concentrations resulted in unstable baselines.
A separation of a 20 mM standard mixture of 13 inorganic and small organic anions is shown in Fig. 5. Baseline noise levels using the blue LED (emission wavelength 476 nm) were very low, typically 0.03 mAU, even during the application of 230 kV. This is amongst the lowest noise levels reported using a LED-based detector under real electrophoretic separation conditions (i.e., a high background absorbance electrolyte and applied voltage). LODs and separation efficiencies were calculated (provided in Table 1) from a separation of a 5 mM standard mix shown in Fig. 6.
It should be noted that the slightly fluctuating baselines in Figs. 5 and 6 are due to the use of Orange G as a probe and its associated adsorption behavior. The baseline variation is not induced by the detector, which had been previously demonstrated to be highly stable. Similar baselines have been often observed with the use of dyes in electrolytes for indirect photometric detection. The high absorptivity of the probe also contributed to the baseline signal since a very small change in the probe concentration caused a relatively large change in background absorbance. For example, in Fig. 6 the baseline change of Figure 5. Electropherogram of 20 mM anion mixture with LED detector. Peak identification: 1, chloride; 2, malonate; 3, succinate; 4, phthalate; 5, methanesulfonate; 6, carbonate; 7, iodate; 8, ethanesulfonate; 9, propanesulfonate; 10, butanesulfonate; 11, pentanesulfonate; 12, hexanesulfonate; 13, heptanesulfonate. Conditions capillary, PEI-coated fused silica, 50 mm ID, length 46 cm, 31 cm to the detector; electrolyte, 4 mM Orange G, 10.0 mM histidine, 0.05% HPMC; separation voltage, 230 kV; detection, indirect photometric at 476 nm (blue LED light source); temperature, 257C; injection, 50 mbar for 6 s (injection volume 11.2 nL, 1.24% capillary volume). less than 1 mAU corresponded to an Orange G concentration of approx. 0.013 mM, a relative change (in terms of the Orange G concentration in the electrolyte of 4 mM) of 0.3%. A similar change in concentration of a less absorbing probe, such as chromate, would naturally be less noticeable. LODs were extremely low due to the decreased baseline noise. A similar separation using the same capillary, electrolyte, and voltage with a diode array detector and deuterium lamp would have a baseline noise of approx. 0.07 mAU at 248 nm (UV) and 0.5 mAU at 476 nm (visible). Judged by the typical baseline noise, the LED-based detector performed an order of magnitude better than the diode array detector and deuterium lamp at an equivalent wavelength. The LODs reported here are possibly the lowest reported so far using indirect photometric detection and normal sample size injection. For comparison, LODs obtained with a diode array detector and an optimized Orange G electrolyte [13] are provided. These LODs were calculated at a signal-to-noise ratio of 2, compared with the LED detector values which were calculated at a signal-to-noise ratio of 3. Furthermore, it should be noted that peak shapes were excellent; with no loss of efficiency or symmetry caused by any time response delays introduced from the detector. The relatively higher detection limit and fronted peak shape for chloride can be attributed to the differences in electrophoretic mobility between the analyte and the probe. However, the LOD is still well below commonly observed LODs achieved with the use of indirect photometric detection. Peak widths (at half peak height) were measured using both the LED-based and diode array detectors. The diode array detector and LED detector are situated 37.5 cm and 31 cm, respectively, from the injection end of the capillary and the values of peak width provided in Table 1 show that the peak width measured at the LED detector was generally 5-10% less than that measured with the diode array detector. An additional advantage of this detection system was that the sampling rate could be controlled by the ADC converter. This permitted more data points to be gathered for narrow peaks (baseline width 1.4-2.5 s) such as those shown in these separations. The sampling rate of 20 points/s of the Agilent 3D CE cannot be varied. A sampling rate of 50 points/s using the LED detector and converter was employed throughout this study.
Finally, 30 consecutive separations were performed without electrolyte replacement or replenishment to test the reproducibility of the detector configuration. Migration time, corrected peak area (peak area/migration time), and peak height % relative standard deviation (RSD) were calculated over the 30 runs (Table 2). All parameters showed good reproducibility and were comparable with those reported using the same electrolyte and built-in detection system [13]. The 1 st and 30 th runs are shown in Fig. 7. No decline in peak shapes, baseline stability or migration times were evident.

Concluding remarks
A carefully designed and constructed detection cell using a LED as light source can provide a useful alternative to traditional detection systems. A new detector design utilizing the existing commercial alignment interface simplifies the design and construction significantly, and the detector is highly stable and robust. LED light sources are particularly useful for indirect photometric detection in the visible region and for direct detection of intensely colored metallochromic compounds. The separation and detection of anions using a strongly absorbing electrolyte in the visible region has demonstrated that this detector provides excellent stability and very low noise levels even during the application of high voltages. LODs achieved with this new system are extremely low. Additionally, the design of this detector allows LEDs of various emission wavelengths to be interchanged easily. Importantly, the upper detector linearity limit of the detector is markedly greater than previous LED-based detectors, which allows electrolytes of higher background absorbance to be used without loss of detection linearity. This enables the concentration of highly absorbing probes to be increased, which benefits stacking effects.