Analytica Chimica Acta 546 (2005) 37–45 Spectral filtering of light-emitting diodes for fluorescence detection Ebbing P. de Jong, Charles A. Lucy ∗ Department of Chemistry, Gunning/Lemieux Chemistry Centre, University of Alberta, Edmonton, AB, Canada T6G 2G2 Received 16 April 2005; received in revised form 29 April 2005; accepted 3 May 2005 Available online 8 June 2005 Abstract The use of light-emitting diodes (LEDs) for fluorescence detection has recently gained much interest. The broad wavelength emission of LEDs requires spectral filtering that is not necessary when using a laser. For instance, filtering the LED light using a bandpass filter improves the signal-to-background ratio for riboflavin by a factor of 70. The bandwidth of the necessary bandpass filters affects both the signal and the background in these measurements. Fluorescence signal can be maximized with wider-bandpass high-transmittance filters. Background is governed by scattering of the LED emission light transmitted by two bandpass filters. When there is large crosstalk between the filters, the LED intensity is linearly related to the background. By estimating and optimizing the crosstalk between excitation and emission filters with a method presented here, the signal-to-background can be optimized. Bandpass filters should be selected with sharp on–off transition, strong blocking outside their transmitting region and the widest bandwidth with minimal crosstalk. Using optimized spectral filtering and capillary electrophoresis analysis, LODs of 50, 3 and 20 nM are obtained for riboflavin, fluorescein and eosin Y, respectively. These results are superior to those reported in the literature for 5 mW LEDs. © 2005 Elsevier B.V. All rights reserved. Keywords: Light-emitting diode; Spectral filtering; Fluorescence detection; Bandpass filters 1. Introduction Laser-induced fluorescence can achieve detection lim- its that approach single-molecule levels. However, for applications such as sensors or miniaturized instrumenta- tion, commonly used lasers, such as the argon-ion [1], helium–cadmium [2], and to a lesser extent the helium–neon [3] lasers are too bulky, expensive, short lived and consume too much power. As such, there has been a drive towards using diode lasers which alleviate these problems [4]. Use of diode lasers was first introduced in the red spectral region [5], when these were the only available visible diode lasers. As semi- conductor technology developed shorter wavelength (laser) diodes, fluorescence detection was subsequently performed in the violet region [6–8]. Recently, the trend towards cheaper, more stable light sources for fluorescence detection has extended to light- emitting diodes (LEDs). Relative to lasers, commercially ∗ Corresponding author. Tel.: +1 780 492 0315; fax: +1 780 492 8231. E-mail address: charles.lucy@ualberta.ca (C.A. Lucy). available LEDs emit broad-band emissions from 350 to 1550 nm, with 291 nm LED recently being reported [9]. Stan- dard LEDs produce up to 1mW in the UV, between 2 and 5 mW in the visible, and up to 45 mW in the IR [10]. LEDs have been used extensively for fluorescence mea- surements of atmospheric gases using liquid-core waveguides [11–20], as the light source in fluorometers [21–23], for gas sensing [24–30] and in capillary electrophoresis [31–39]. An advantage of LEDs as a light source for fluorescence is that they are commercially available at wavelengths span- ning the entire visible spectrum, as well as the near-UV and near-IR. In contrast, diode lasers are available for a limited number of wavelengths in the visible and IR. How- ever, LEDs also possess a broader spectral bandwidth than diode lasers—typically 30 nm full width at half maximum (FWHM) versus 1 pm for diode lasers. This polychromaticity can be detrimental in fluorescence measurements, as scat- tered light from the LED that coincides with the detection wavelength will produce a high background signal. Fluo- rescence detection limits can be severely degraded if the longer-wavelength LED light reaches the detector [29,30]. 0003-2670/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2005.05.005