View Article Online / Journal Homepage / Table of Contents for this issue

PAPER www.rsc.org/analyst | The Analyst

UV-absorbance detector for HPLC based on a light-emitting diode

Stefan Schmid,a Mirek Mackab and Peter C. Hauser*a
Received 11th October 2007, Accepted 1st February 2008
First published as an Advance Article on the web 26th February 2008
DOI: 10.1039/b715681b

A flow-through optical absorption detector for HPLC was constructed using a novel deep-UV
Published on 26 February 2008. Downloaded by Aston University on 14/01/2014 15:36:34.

light-emitting diode as radiation source with a peak emission wavelength of 255 nm. For
measuring the transmitted intensity (a property correlated to Transmittance) a special
UV-sensitive photodiode was employed. Besides the power source, no optical or electronic
components other than an inexpensive operational amplifier and a few passive components were
necessary. The performance of the detector was tested with three substances, namely nitrobenzene,
benzoic acid and methyl benzoate, which were separated by gradient elution using an
acetonitrile/water mixture and tetrabutylammonium hydrogensulfate as pH-buffer. Calibration
curves for concentrations between 1.6 lg·mL−1 and 400 lg·mL−1 (nitrobenzene) and 8 lg·mL−1
and 2.5 mg·mL−1 (benzoic acid and methyl benzoate) were determined and coefficients of
determination, r2 , of 0.9945, 0.9972 and 0.9996 were obtained for quadratic curve fits for the 3
compounds respectively. Relative standard deviations (n = 7) for peak areas were determined as
0.35% (nitrobenzene, 80 lg·mL−1 ), 0.27% (benzoic acid, 400 lg·mL−1 ) and 0.83% (methyl
benzoate, 200 lg·mL−1 ). The lower limits of detection were found to be 750 ng·mL−1 , 5.8 lg·mL−1
and 12 lg·mL−1 for nitrobenzene, benzoic acid and methyl benzoate respectively.

Introduction summarized developments in the use of LEDs in analytical

Light-emitting diodes (LEDs) offer a host of advantages to the
analytical chemist. Their emission spectra are typically 20–30 nm
wide, which, while not allowing the scanning of spectra, are
well suited for the absorbance or fluorescence excitation bands
of molecules. Monochromators or other wavelength selection
devices are therefore not needed. LEDs are inexpensive, compact
and robust, and have long lifetimes. Their high efficiency
means that high intensities can be obtained with low power
consumption. High output stability translates into low noise
and thus the limits of detection are often much better then
those obtained with conventional spectroscopic light sources.
These optoelectronic components have therefore been used in
analytical chemistry since the early 1970s, soon after the first,
to todays standards dim, devices in the red wavelength region
became commercially available.1 Many diverse applications have
been described in the literature since, such as detectors for
flow-injection analysis (see for example ref. 2–4), fluorescence
based chemical sensors (see for example ref. 5,6) and detectors
for capillary electrophoresis (see for example ref. 7–13). Other
analytical instruments using LEDs have been commercialised,
such as carbon dioxide gas sensors and portable instruments
for molecular absorption spectrometry. Dasgupta et al. have
a
Department of Chemistry, University of Basel, Spitalstrasse 51, 4056,
Basel, Switzerland. E-mail: peter.hauser@unibas.ch; Fax: +41 61
2671013; Tel: +41 61 2671003
b
Department of Chemical Sciences, Dublin City University, Glasnevin,
Dublin 9, Ireland
instrumentation up to 2003.14,15
Nonetheless, the use of LEDs as light sources in detectors for
HPLC has only been considered in a few instances. Schmidt
and Scott16 reported the ion-chromatographic separation of
heavy metals, which were detected as their complexes with
PAR (4-(2-pyridylazo)resorcinol) using a green LED. O’Toole
et al.17 recently designed a miniature LED based flow cell for
the same application. A similar detector was also used for the
determination of alkaline earth metals in ion-chromatography
as complexes with o-cresolphthalein.18 Berthod et al.19 used a
red LED and an indirect method based on the displacement of
a dye contained in the eluent for the determination of alcohols,
which generally are difficult to detect in HPLC. Wiesufer
et al.20 determined atmospheric hydroperoxides after separation
in HPLC by postcolumn derivatization and detection with an
LED based cell.
The reason for the as yet only limited use of LEDs in the
widespread technique of HPLC is the fact that most analytes
have to be detected via their absorption in the lower UV range
at wavelengths between about 300 and 200 nm. LEDs have
improved over the years and two clear trends are evident.
One is an increase in intensity, the other the development of
devices with ever shorter wavelengths. As shorter wavelengths
have higher energy, the manufacture of devices becomes more
difficult. Blue LEDs were the last ones to become available in the
visible range in the early 1990s. Emitters which peak at 370 nm,
already in the UV-range, have been at disposal for a few years.
However, devices for wavelengths below 300 nm, the so-called
deep UV-range, have only very recently been developed. An
investigation into the use of such a radiation source for the

This journal is © The Royal Society of Chemistry 2008 Analyst, 2008, 133, 465–469 | 465
 View Article Online

construction of a compact and inexpensive absorption detector

for HPLC is reported herein.

Experimental
Instrumentation
The work was carried out on a HP 1100 HPLC system (Agilent,
Waldbronn, Germany). An injection loop of 25 lL and a C-18
column with dimensions of 150 × 4.6 mm and 5 lm particle
Published on 26 February 2008. Downloaded by Aston University on 14/01/2014 15:36:34.

size (Luna C18(2), Phenomenex, Torrance, CA, USA) were

employed. The experimental detector was connected to the
eluent outlet of the commercial detector which is built into the
HPLC system for simultaneous measurements. For graphical
presentation, the output signals from both detectors were passed
to an e-corder 201 data acquisition system (eDAQ, Denistone
East, NSW, Australia) and recorded using the Chart software
package (eDAQ). To produce the quantitative calibration data
for both systems, the signal of the experimental detector
was fed to an auxiliary measurement channel of the Agilent
HPLC system, and the ChemStation software package (Agilent)
was used for processing. To both signals a low pass filter Fig. 1 Schematic drawing of the detector cell. Approximately in real
with a cut-off frequency of 0.25 Hz was applied. The UV- size.
absorption spectra of the test compounds were measured with
a standard commercial UV-Vis-spectrophotometer (UV1-200,
ATI Unicam, Cambridge, UK). The emission spectrum of the
LED was measured with an SD2000 spectrometer from Ocean Delrin, which has a large opening of square cross-section for
Optics (Dunedin, FL, USA). insertion of the cuvette and two smaller bores on the sides to
accept the optoelectronic components. The flat windows of the
LED and photodiode were thus butted directly onto the cuvette
Materials and methods without using any lenses. The holder also effectively shields the
The four standards, benzoic acid, methyl benzoate, benzyl ac- photodiode from stray radiation originating from the LED as
etate and nitrobenzene were obtained as analytical reagents from well as from ambient light.
Acros (Geel, Belgium). Acetonitrile (HPLC grade) was obtained The simple circuit diagram to operate the LED and the
from Fisher Scientific (Pittsburgh, PA, USA) and the water was photodiode is shown in Fig. 2. The system was powered by
purified in a Millipore system (Millipore, Bedford, MA, USA). a regulated split ±15 V supply. The LED was operated with a
Tetrabutylammonium hydrogensulfate (TBAHS)(HPLC grade) current limiting resistor; a trimmer was employed so that the
was obtained from Aldrich (Buchs, Switzerland). The gradient current could be set to 25 mA which is close to the specified
separation was started with a ratio of 25% A (acetonitrile), 65% maximum of 30 mA. An integrated circuit operational amplifier
B (water) and 10% C (a solution of 1 g TBAHS in 200 mL (OPA137, Texas Instruments, Austin, TX, USA) was used to
of acetonitrile and 800 mL water) (v/v/v) which was linearly convert the photodiode current (i) to an output voltage (Vo )
changed to 38% A, 52% B and 10% C over a period of 15 min. according to Vo = -i·Rf . A feedback resistor, Rf , of 10 MX
The total flow rate was a constant 1.3 mL·min−1 . was employed. The parallel capacitor serves to reduce high
frequency noise. The entire assembly, consisting of flow-through
cuvette, holder, and optoelectronic as well as electronic parts (but
Detector excluding the power supply and the data acquisition system)
The LED was obtained from Sensor Electronic Technology were fitted into a aluminium box of outer dimensions of appr.
(Columbia, SC, USA). This device is packaged in a standard 5.5 × 7 × 2.5 cm.
TO39 metal can fitted with a UV-transparent window. The
UV-photodiode with the designation “EryF” originated from
Sglux Solgel Technologies GmbH (Berlin, Germany). The two
optoelectronic components were placed onto opposite sides of
the optical path of a commercially available high pressure flow
through cuvette with small internal volumes suitable for HPLC
detection (178.313-QS, from Hellma, Müllheim, Germany). The
cuvette has a standard optical path length of 1 cm and a round
aperture of 1 mm diameter. In order to fit the parts together
a special holder was constructed. A sketch of the assembly
is shown in Fig. 1. The holder consists of a block of black Fig. 2 Circuit diagram of the detector.

466 | Analyst, 2008, 133, 465–469 This journal is © The Royal Society of Chemistry 2008
 View Article Online

Results and discussion
Spectral considerations
HPLC-instruments are often fitted with a fixed wavelength
detector based on a low pressure mercury discharge lamp
which has a prominent emission line at 253.7 nm.21 Other lines
which are also present are filtered out with a monochromator
or an optical interference filter. This wavelength is popular
as it matches well the absorbance bands of many aromatic
compounds. The new UV-LEDs which have recently become
Published on 26 February 2008. Downloaded by Aston University on 14/01/2014 15:36:34.
solution is simpler. Special UV-photodiodes are available for
measurement of the so-called UV-Exposure Index. These devices
have a glass filter built in to mimic the sensitivity of the human
skin (according to the so-called erythma action curve) and thus
have a significantly reduced response to wavelengths above about
320 nm.
Chromatograms
The chromatograms for the separation of the 4 compounds by

gradient elution are shown in Figs. 4A and 5B for the LED-

available have peak emission wavelengths in the range from
based and the conventional detectors respectively. The overall
350 nm down to 250 nm. The emission spectrum of a specimen
appearance of the two chromatograms is very similar. On close
with a peak wavelength of 255 nm, closely matching the 254 nm
examination of the peaks it is noted that there is a slight variation
line of the mercury lamp, is shown in Fig. 3. The band width (full
in their relative heights between the two chromatograms. In
width at half maximum, FWHM) is approximately 6 nm. Note
particular the peak for nitrobenzene is comparatively tall for the
that the UV-LED was found to also show two weak emission
LED-detector. This feature must be due to differences inherent
bands at 370 nm and at 480 nm which are beyond the spectral
in the two detectors. The conventional detector has a specified
range of the figure. The band at 480 nm can be visibly observed.
spectral bandwidth of 6.5 nm, as determined by the built in
monochromator, and that of the UV-LED was found to be
approximately 6 nm (see above). Thus the bandwidths of the
two systems are almost identical and this factor alone therefore
cannot be responsible for the observation. It is presumed that the
cause of the variation are minor differences in the distribution
of the spectral sensitivity within these spectral windows. In any
case, this is not a problem for the application of the device.

Fig. 3 Absorbance spectra of the 4 compounds tested (left hand scale)

and emission spectrum of the UV-LED (right hand scale).

Also shown in Fig. 3 are the absorbance spectra of the 4

compounds used for characterization of the detector. One of the
species, nitrobenzene shows indeed a strong peak with a maxi-
mum close to the emission maximum of the LED, while benzoic
acid and and methyl benzoate have their maxima at lower
wavelengths and show an absorbance which is approximately
10 times less pronounced at 255 nm than that of nitrobenzene.
The fourth compound, benzyl acetate, shows only a very weak
absorbance at this wavelength. Clearly, highest sensitivity for the
UV-LED based detector can be expected for nitrobenzene.
Also considered has to be the spectral sensitivity of the
photodiode used for detection. A silicon photodiode shows a
strong reduction in sensitivity with decreasing wavelength even
if it is specified for use in the UV-range. Typically, the response
at 250 nm is about one order of magnitude lower than at
300 nm and the sensitivity at longer wavelengths can be expected
to be even greater. It is therefore important to consider the
additional emission peaks of the UV-LED at longer wavelengths,
despite their lower intensity, as these could lead to a significant
background signal from the photodiode detector. Such a signal
has a similar effect as stray light and will lead to a reduced Fig. 4 Chromatograms obtained with the LED-detector (A) and
sensitivity for the analyte peak. One possibility to exclude these the commercial detector (255 nm) (B) on injection of a test mixture
unwanted wavelengths would be the use of an interference filter consisting of nitrobenzene (50 lg·mL−1 ), benzoic acid (250 lg·mL−1 ),
in front of either the LED or the photodiode. The chosen methyl benzoate (250 lg·mL−1 ) and benzyl acetate (2500 lg·mL−1 ).

This journal is © The Royal Society of Chemistry 2008 Analyst, 2008, 133, 465–469 | 467

A baseline disturbance is apparent after about 1.5 min in
the chromatogram acquired with the LED-detector. Presumably
this is an artifact due to refractive index changes caused by
the solvent of the injected solution. On the other hand, no
discernible baseline drift could be observed for the duration of
the chromatogram, which would be the case in gradient elution
if such an effect was significant. Also apparent is a higher level
of baseline noise for the LED-detector, which, of course, has
a bearing on the signal-to-noise ratio and the detection limits.
These aspects will be discussed in the following section.
Published on 26 February 2008. Downloaded by Aston University on 14/01/2014 15:36:34.
Signal and noise, calibration
The photocurrents obtained are in the nA-range and thus orders
of magnitude lower than what would be achieved using an LED
in the visible range and a regular photodiode. This must be due to
both, low emission intensity of the UV-LED and low sensitivity
of the photodiode at this wavelength. It can be assumed that the
low intensity is the main reason for the baseline noise evident.
This was quantified by determining the maximum and minimum
baseline values within a period of 1.5 min, which corresponds
approximately to about 5 times the width of a peak, and a typical
value of 1.6 mV was found. At a detector output voltage of about
0.12 V this translates into a signal-to-noise ratio (S/N) of about
75 for the raw output. It was also observed that over the duration
of the project, the intensity of the LED was gradually decreasing.
The initial baseline signal was at approximately 0.16 V, so that
for the approximately 100 hours that the detector was in use a
loss of about 25% of intensity occurred.
Calibration curves were acquired for nitrobenzene, benzoic
acid and methyl benzoate. For benzyl acetate calibration was
not attempted as the sensitivity for this compound at 255 nm is
too low (compare Fig. 3). Eight standard solutions over almost
3 orders of magnitude were prepared for the 3 compounds. For
nitrobenzene these ranged from 1.6 lg·mL−1 to 400 lg·mL−1 and
for the other two compounds from 8 lg·mL−1 to 2.5 mg·mL−1 .
Higher concentrations were not tested as such levels would be
of very limited practical relevance. For the commercial detector
the calibration curves for peak areas were found to be perfectly
linear, but as anticipated, this was not the case for the LED-
detector. The reason for the expected non-linearity is the fact
that by design the output signal of the detector is proportional
to transmittance, T, and not to absorbance, the parameter which
scales linearly with concentration. Linear curve fitting was thus
not carried out, but quadratic equations could be fitted well
to the data. The coefficients of determination, r2 , obtained for
the regression analysis are given in Table 1. Also given are the
Table 1 Quantification
Coefficient of Determination (r2 )a Sensitivityb
Nitrobenzene 0.9945 1.31
Benzoic acid 0.9972 0.104
Methyl benzoate 0.9996 0.104
a
For quadratic regression equations for peak areas for 7 concentrations between 1.6 lg·mL−1 and 400 lg·mL−1 (nitrobenzene) and 8 lg·mL−1
and 2.5 mg·mL−1 (benzoic acid, methyl benzoate). b Slope for linear regression equations for concentrations up to 40 lg·mL−1 (nitrobenzene) and
200 lg·mL−1 (benzoic acid, methyl benzoate). c n = 7, for 80 lg·mL−1 (nitrobenzene), 400 lg·mL−1 (benzoic acid) and 200 lg·mL−1 (methylbenzoate).
Values in brackets are for the commercial detector (255 nm). d Concentrations giving peak heights 3 times the baseline noise. Values in brackets are
for the commercial detector (255 nm).
468 | Analyst, 2008, 133, 465–469
View Article Online
sensitivities, which reflect the different absorptivities as evident
from the spectral fits of the compounds to the wavelength band
of the detector (compare Figs. 3 and 4).
The reproducibilities for the peak areas were determined by
repeated injections for some of the standards. Typical values
of the standard deviations obtained are given in Table 1 as
well, and were found to be below 1%. A comparison with
the values obtained for the detector built into the instrument
shows that the performance of the new detector in this regard is
almost as good. Also given in the table are the lower limits of
detection, determined as the concentrations giving peak heights
corresponding to 3 times the baseline noise (measured as the
maximum baseline deviations for a time period of 5 times the
peak width). The detection limits correlate directly with
the sensitivities (the initial slope values as given in the table)
as the baseline noise is constant for all compounds but their
absorption coefficients are different. The same correlation is
found for the detection limits obtained with the commercial de-
tector built into the instrument. Clearly, the new detector cannot
match the commercial detector in this performance parameter.
It is interesting to note that detection limits for nitrobenzene
obtained in preliminary work using a different HPLC instrument
yielded values of 300 ng·mL−1 and 200 lg·mL−1 for the UV-
LED based and conventional detectors respectively. The fact
that the value for the LED-detector was better initially is due
to the somewhat higher signal level in earlier work (as discussed
above), and the fact that the value for the commercial detector
was poorer is due to a variation in performance of different
models of instruments.
Conclusions
The simple UV-absorbance detector based on solid-state opto-
electronic components was found to give a performance which
is close to that of a commercial unit based on a deuterium lamp,
except for the detection limits. To our knowledge this is the
first report on the use of such a light source for detection in
HPLC in the deep-UV. Where flexibility in wavelength is not
essential, the new device is therefore an attractive alternative
which is significantly less expensive. Its compactness and low
power consumption is also of interest for the design of portable
and miniaturized instrumentation or for lab-on-chip devices. By
incorporating a logarithmic convertor in the electronic circuitry,
direct absorbance, rather than transmission measurements are
possible, and thus a linear detector output could be obtained
if desired (see for example ref. 15). Furthermore, it can be
expected that UV-LEDs with higher output intensity will
Relative Standard Deviationc Limit of Detectiond
0.35% (0.19%) 750 ng·mL−1 (77 ng·mL−1 )
0.27% (0.22%) 5.8 lg·mL−1 (0.55 lg·mL−1 )
0.83% (0.37%) 12 lg·mL−1 (1.2 lg·mL−1 )
This journal is © The Royal Society of Chemistry 2008

become available. This will lead to improved signal-to-noise
ratios and hence lower limits of detection, which may ultimately
exceed that of detectors based on conventional light sources.
Acknowledgements
Partial financial support from the Swiss National Science
Foundation (Grant No. 200020-113335/1) is gratefully acknowl-
edged.
Published on 26 February 2008. Downloaded by Aston University on 14/01/2014 15:36:34.
References
1 H. Flaschka, C. McKeithan and R. Barnes, Anal. Lett., 1973, 6,
585–594.
2 D. Betteridge, E. L. Dagless, B. Fields and N. F. Graves, Analyst,
1978, 103, 897–908.
3 M. Trojanowicz, P. J. Worsfold and J. R. Clinch, TrAC, 1988, 7,
301–305.
4 M. Trojanowicz and J. Szpunar-Łobińska, Anal. Chim. Acta, 1990,
230, 125–130.
5 P. C. Hauser and S. S. S. Tan, Analyst, 1993, 118, 991–995.
6 G. Neurauter, I. Klimant and O. S. Wolfbeis, Anal. Chim. Acta, 1999,
382, 67–75.
7 M. Macka, P. Andersson and P. R. Haddad, Electrophoresis, 1996,
17, 1898–1905.
This journal is © The Royal Society of Chemistry 2008
View Article Online
8 C. B. Boring and P. K. Dasgupta, Anal. Chim. Acta, 1997, 342, 123–
132.
9 P. A. G. Butler, B. Mills and P. C. Hauser, Analyst, 1997, 122, 949–
953.
10 M. King, B. Paull, P. R. Haddad and M. Macka, Analyst, 2002, 127,
1564–1567.
11 C. Johns, M. Macka and P. R. Haddad, Electrophoresis, 2004, 25,
3145–3152.
12 S. Hapuarachchi, G. A. Janaway and C. A. Aspinwall, Electrophore-
sis, 2006, 27, 4052–4059.
13 M. C. Breadmore, R. D. Henderson, A. R. Fakhari, M. Macka and
P. R. Haddad, Electrophoresis, 2007, 28, 1252–1258.
14 P. K. Dasgupta, I.-Y. Eom, K. J. Morris and J. Li, Anal. Chim. Acta,
2003, 500, 337–364.
15 P. K. Dasgupta, H. S. Bellamy, H. Liu, J. L. Lopez, E. L. Loree, K.
Morris, K. Petersen and K. A. Mir, Talanta, 1993, 40, 53–74.
16 G. J. Schmidt and R. P. W. Scott, Analyst, 1984, 109, 997–1002.
17 M. O’Toole, K. T. Lau, B. Shazmann, R. Shepherd, P. N. Nesterenko,
B. Paull and D. Diamond, Analyst, 2006, 131, 938–943.
18 L. Barron, P. N. Nesterenko, D. Diamond, M. O’Toole, K. T. Lau
and B. Paull, Anal. Chim. Acta, 2006, 577, 32–37.
19 A. Berthod, M. Click and J. D. Winefordner, J. Chromatogr., 1990,
502, 305–315.
20 S. Wiesufer, A. Boddenberg, A. P. Ligon, G. Dallmann, W. V. Turner
and S. Gab, Environ Sci Pollut R, 2002, 41–47.
21 R. P. W. Scott, Chromatographic Detectors, Design, Function, and
Operation, Marcel Dekker Inc., New York, 1996.
Analyst, 2008, 133, 465–469 | 469