Keywords

1 Introduction

Microfluidic chips have the potential to integrate all experimental processes in a research laboratory (sample preparation, reaction, separation, and detection) onto a microscale device. Ever since the micro total analysis system (μ-TAS) was firstly proposed by Manz et al. in the early 1990s [1], it has achieved great advances in many scientific research fields, such as medical diagnostics [2], drug screening [3], proteomics [4], and environmental and food analysis [5, 6]. To be able to read out the analytical results captured by a microchip, the detection system is the core part and is responsible for signal acquisition. To date, various detection modules including optical detection, electrochemical detection, mass spectrometry and others, have been assembled or partly assembled onto the microchip. The ultimate goal of having the detection system on the microfluidic chip is to incorporate it as an inseparable part of a portable microdevice with high sensitivity and fast response.

In this review, we mainly focus on the recent developments in optical systems assembled onto a chip because optical detection is easily implemented in chemical/biochemical laboratories, has acceptable sensitivity in regular analysis, and simplicity in coupling with other functional units. Such topics have been touched on in previous reviews. A general overview of earlier detection methods has been published elsewhere [7, 8]. The specialized subjects of optical systems such as detectors for separation and unconventional detection can be found in the literature [9–11]. The current optical detection systems on microfluidic chips can be summarized from two perspectives. The first is the traditional optics configuration strategy also called “off-chip” or “free space,” in which light propagates in air before and after interaction with the sample (see Sect. 2). In this mode, the detection units (e.g., light source, mirrors, and detectors) are separated from the chip platform. We present the detection methods of absorption, luminescence and scattering in Sects. 2.1–2.5. The second strategy is the “on-chip” mode, in which the optics are integrated with or fabricated together with the fluidic functional units (see Sect. 3). When the optical property of the integrated optics can be tuned by fluidic technology, the microfluidic is also called “optofluidic.” This mode, however, is in the development stage, and a highly integrated optofluidic is not fully realized yet. Some related optical components such as waveguides, microlens, detectors and microscopes will be discussed in Sects. 3.1–3.4. The nonoptical detection methods (e.g., electrochemical detection, mass spectrometers, and nuclear magnetic resonance) are not covered in this review but can be found in previous reviews [7, 8, 12].

2 Conventional Free-Space Optical Detection

2.1 Absorbance Detection

UV/Vis absorption spectroscopy is the most straightforward and popular detection method in separation and microseparation science due to its generality, sensitivity, and adequate detection limit. Absorption spectra are related to the structure and concentration of the analyte and are based on the capability of samples to attenuate the incident electromagnetic radiation at various wavelengths.

The depth of the channels in microchips is relatively shallow compared to their width due to the character of the fabrication method. If light is designed to pass through the channel perpendicular to the chip plane, the optical length through the sample is limited by the depth of the channel, which is shortened from millimeters in conventional analysis to tens of micrometers in microanalysis, adversely impacting the sensitivity. To address the issue of the low sensitivity caused by a short optical length, various approaches were developed to increase the optical path length.

With free-space optics, light first propagates in the air and then enters the detection cell. The optical length is enlarged simply by changing the geometry of the channel at the detection point, such as incorporating a short, wide bubble cell within the channel [13], or sandwiching an additional glass substrate with a through-hole at the detection point to increase the optical length in the vertical direction of the chip plane [14]. Despite that, these methods suffer from the limited improvement of optical length because the changes of channel geometry are detrimental to maintaining separation efficiency and to blocking stray light. Introducing the light along the channel axially is a practical alternative for enhancing absorbance. The light can be reflected between two mirrors by coating 200-μm aluminum films above and below the channel. Effective optical lengths ranging from 50 to 272 μm were achieved by adjusting the angle of laser incidence (Fig. 1) [15]. Billot et al. implemented a similar idea on a glass–PDMS–glass microchip by coating the mirrors on the inside surfaces of the channel and exciting the sample with a normal illumination configuration [16]. Besides making light reflect multiple times inside the channel, passing light along the channel axially in a straight line is another way of increasing the optical length. This idea is implemented by fabricating two parallel 45° mirrors at each end of the microchannel so that light was reflected into the channel and out to the detector by the mirrors, making the 5-mm channel the optical path (Fig. 2) [17]. Within this category, other efforts were also made to improve the sensitivity of UV absorption and to maintain the high separation performance of microscale electrophoresis. For examples, Lin et al. fabricated a hybrid poly(dimethylsiloxane) (PDMS)/quartz chip with a 100-μm thick PDMS membrane on the detection window for UV absorbance detection [18]; Ou et al. developed a PDMS/SU-8/quartz hybrid chip incorporating UV absorbance-based whole-channel imaging detection [19].

Fig. 1
figure 1_144

Diagram showing a planar multireflection absorbance cell. Aluminum mirrors (thick lines) were coated above and below a channel. The narrowest region inside the cell is the fluid flow path (Reprinted from [15] with permission of Wiley-VCH). i (1,2,3,4) refers to different medium; θi is the angle within the medium i with respect to the surface normal; hi is the thickness of the medium i; li is the longitudinal distance in each of medium and li = hi tan (θi); Ln is the total longitudinal travel for once reflection; n the total number of reflections

Full size image
Fig. 2
figure 2_144

Cross-sectional diagram of a microchip. Two parallel 45° mirrors were fabricated at each end of the microchannel to increase the optical length (Reprinted from [17] with permission of Elsevier)

Full size image

Embedding waveguide optics into the microchannel is another way to enlarge the optical length. Integrated waveguides on-chip are similar to electrical circuits on electronic chips, carrying information about the analytes. This has good potential for design of a lab-on-a-chip because it can realize the practical miniaturization of the optical elements. Actually, one of the initial motivations for developing waveguides was to extend the optical length by coupling the light effectively into the axial direction of the channel. The channel length then becomes the light path length. Integrated optical waveguides on microchips offer the capability to deliver and capture light at nearly arbitrary positions and angles on the chip. With the help of an in-plane waveguide, various absorbance detection geometries have been developed, such as a 750-μm U-cell [20], a 1,000-μm U-cell on an electrochromatography chip [21] and a hollow Abbe prism on a PDMS chip [22]. However, the enhancement of optical losses with increasing optical path length was observed because of stray light and light dispersion, leading to a reduction in the signal-to-noise ratio [23, 24]. Accordingly, a light collimating system was integrated in the front of the extended optical path to block stray light and increase the sensitivity [25]. Llobera et al. integrated optical fibers, air mirrors, and collimation microlenses on a chip to realize a multireflection absorbance detection system in which the detection limit of 41 nM for fluorescein was achieved [26]. Fang et al. provided an example of miniaturized absorbance detection. They employed a U-shape bent capillary to couple light into the capillary and out to the photodetector with less light loss and less dead volume, resulting in a 15 mm effective optical path length, which significantly simplified the system for use as a hand-held photometer [27].

A technique based on absorbance of evanescent waves has been developed very recently, in which the evanescent field was coupled into a microchannel by a U-bend SU-8 waveguide. When the analytes flow through the evanescent field, they interact with the light and induce evanescent wave absorption. This technique is expected to also suppress the scattered light when light passes directly through the detection region, thereby enhancing the sensitivity [28].

Waveguide-based cavity ring-down absorption spectroscopy [29] can increase the optical path length by reflecting a pulse of light multiple times through the sample without changing the channel geometry, and is also a promising way for absorption detection on the microchip. In the case of the tartrazine detection, microfluidic-based fiber-loop ring-down absorption spectroscopy has shown a satisfactory detection limit of 5 μM [30, 31].

Despite the fact that absorbance is more limited in sensitivity than fluorescence, the potential of an inexpensive miniaturized optical setup still provides an advantage in certain applications, such as when changes in optical density or color are sufficient for clinical diagnosis.

2.2 Fluorescence Detection

Today’s lab-on-a-chip devices take advantage of various optical detection strategies to achieve microvolume detection. Of these, fluorescence detection is still a preferred choice to be incorporated onto microdevices primarily due to its high sensitivity and high selectivity for cellular and molecular sensing. Fluorescence is the emission of light of fluorphores where the cycle of excitation and emission will continue until the excitation source is turned off or the fluorophore is photodecomposed. The fluorescence of biomacromolecules originates either from the native chromophore inside the molecules or from chemically linked fluorescent labels. Recently, the fluorescent labels for microanalysis have been extended from synthetic dyes, fluorescent proteins, and quantum dots (QDs) to nanoparticles. Development and exploitation of new types of fluorescent probes with high quantum yield and high specificity on lab-on-a-chip will be helpful to improve the detection sensitivity and minimize the size of optical components in the sensing system.

The complete optical setup of induced fluorescence involves an excitation part and an emission part. The intersection point of the two parts is the detection window in the microdevice. The excitation part starts from the light source and ends at the microdevice, while the emission part originates from the microdevice and stops at the detector. In free-space fluorescence detection, the common excitation sources include the laser, light-emitting diode (LED), and mercury or xenon arc lamp, and each light source has its distinctive spectroscopic property and practical benefits. Laser-induced fluorescence (LIF) is a highly sensitive optical detection method and is able to perform even single molecule detection. LIF has been introduced into the microchip at the early stage of lab-on-a-chip and remains to be a popular approach, especially for analyzing low-abundance molecules after combining with point detectors like photomultiplier tubes (PMTs). The high pressure vapor discharge lamp is the standard light source for modern fluorescence microscopy due to its broad spectrum, ranging from UV to visible regions, and because it is less expensive and more flexible by offering different wavelengths with the assistance of proper optical filter sets. The lamp-based fluorescence microscope is usually coupled with an imaging detector like a charge-coupled device (CCD) to analyze cellular fluorescence. An LED is a small semiconductor light source and can efficiently produce light. These merits indicate that LED is suitable for integration into microfluidic chips. Yao et al. succeeded in integrating a green organic LED and thin film interference filter into a microfluidic device through layer-by-layer stacking [32], and also designed a similar LED-induced fluorescence-based microdevice for the imaging of whole-column isoelectric focusing [33] and for high throughput capillary array electrophoresis on the chip [34].

Fluorophores emit photons with 4 pi steradians once they are excited. Therefore, fluorescence can be collected at any solid angle by an objective or a lens. One prevalent optical configuration for induced fluorescence detection is that the collection objective in the emission part is the same as the focus components in the excitation part. Fluorescence collected by the same objective passes through the dichroic and emission bandpass filter and reaches the detector. Such configuration is called epi-illumination in fluorescence microscopy and is also the basis of confocal LIF as long as a pinhole is placed in the conjugate plane to the sample plane to block the out-of-focus light and a laser diffraction-limited spot is used to excite samples. A typical confocal LIF mode on-chip is depicted in Fig. 3 and is almost the same as for the confocal microscope except that the microchip is used to replace the microscope stage. Additionally, the emission part of another setup for fluorescence detection does not share the same objective with the excitation part, and each part has its own focusing systems. In this configuration, the collection angle seriously affects the sensitivity and has been optimized to 45° [35] by Fang’s group. Subsequently, Whiting et al. determined under- and oversampling of individual particle distributions in microfluidic electrophoresis with orthogonal LIF [36]. Using this orthogonal LIF detector, a detection limit of 1,761 molecules of fluorophore was obtained under a constant flow of fluorescein. Compared to the common LIF detection scheme used with microfluidic devices, this is a significant improvement.

Fig. 3
figure 3_144

Confocal fluorescence detection on a microchip. A laser was used as the excitation source. The laser light passed through a biconvex lens, was focused into an illumination pinhole, and was subsequently reflected by a dichroic mirror and focused in the channel on the capillary electrophoresis (CE) chip by a microscope objective. The fluorescence signal from the sample, collected by the microscope objective, was passed through the dichroic mirror, focused by the tube lens into a confocal pinhole, and then detected by a photomultiplier tube (PMT). To improve the signal-to-noise ratio, band-pass filter and notch filter were inserted above PMT for spectral filtering

Full size image

From the viewpoint of instrumentation development, the aim of optical detection systems of lab-on-a-chip is miniaturization, but under the premise that the detection sensitivity is maintained at a practical level. Therefore, we shall review the fluorescence detection on-chip from the points of view of improving sensitivity and miniaturization.

LIF is appropriate to deal with the minute amounts of sample in the microchannel and to demonstrate the feasibility of the concept of μ-TAS in the early stage of development [37, 38]. Optimization of detection parameters to improve the sensitivity of confocal LIF on-chip was accomplished [39] and the authors detected 300 fM fluorescein with a signal-to-noise ratio of 6.1 in continuous-flow mode and 1 pM fluorescein in capillary zone electrophoresis. Consequently, the confocal LIF was successfully used to detect single molecules of a 100–1,000 bp DNA ladder in a microchannel by electrodynamic focusing [40]. High-quality DNA sequencing separations have also been performed in microfabricated electrophoresis channels [41]. Early results of detection of single molecules labeled with organic dyes in microchannels were reviewed by Manz in 2005 [42]. The point detection mode suffers from low detection efficiency because the diffraction-limited focused laser spot is so small that the whole channel cannot be covered and most molecules do not pass through the laser beam. Chiu et al. designed a line-confocal detection geometry comprising an elliptically shaped probe volume to enlarge the fraction of molecules that flow through the laser beam without severely decreasing the signal-to-noise ratio [43]. Another way to improve the single molecule detection efficiency in microchannels is to employ an imaging detector, either an intensified charged coupled device (ICCD) or an electron multiplied CCD (EMCCD), by imaging a larger detection area than covered by the point detector. Single moving lambda DNA molecules forced by pressure or electrokinetic flow in the microchannel are easily observed with an imaging detector [44]. When the total internal reflection excitation mode is coupled to the microchannel, the velocity of molecules near the channel wall can also be measured. Therefore, the velocity distribution in the microchannel can be profiled by adjusting the excitation depth of the evanescent wave [45–47]. In addition to organic dyes that are used as the fluorescent label, recently developed fluorescence nanoparticles are also used to label molecules. Our group has improved the fluorescence characteristic of core–shell QDs by adding thiol-group chemicals and labeled QDs to the oligonucleic acids and demonstrated the advantages of QDs for microchips [48–50].

The microdevice is also an ideal platform for single-cell analysis when the functional units of cell sampling, cell sorting, cell breakage, chemical separation, and detection are integrated onto a single chip. Meanwhile, single-cell chemical analysis on the microchip is a typical application of LIF detection on-chip. Yu developed a high-throughput single-cell analysis method on PDMS microfluidic chips to simultaneously analyze glutathione and rhodamine 123 in single K562 cells. They could analyze about ten cells each minute in one channel [51]. For nucleic acid molecules, the labeling process can be implemented in the channel by intercalating dyes into the base pairs. Kleparnik [52] assessed the effects of doxorubicin, a cytostatic, on the apoptosis of single cardiomyocyte cells by separating DNA fragments labeled with ethidium bromide. For protein molecules, label-free LIF detection is a suitable method due to the UV fluorescence of the three amino acids tryptophan, tyrosine, and phenylalanine. Hellmich reported LIF detection of native protein lysed from single cells and excited with a 266-nm laser [53]. The authors further demonstrated the CE separation of two proteins from a single cell and improved the separation efficiency and detection sensitivity by shortening the electrical lysis time and improving the injection plug size [54].

LIF detection is the most sensitive optical method so far, but is hard to miniaturize in order to satisfy the ultimate goal of a microfluidic chip that assembles all analytical processes within one micrometerscale microstructure. Therefore, how to achieve the miniaturization of fluorescence detection on microdevices is becoming an active field for lab-on-a-chip research. Several examples demonstrate recent advances in miniaturized LIF detection on the microchip. In 2005, Renzi et al. designed a hand-held microchip-based analytical instrument that combines fluidic, optics, electrical power, and interface modules and integrates the functions of fluidics, microseparation, lasers, power supplies etc., into an 11.5 × 11.5 × 19.0 cm system (Fig. 4a) [55]. This device allows picomolar (10−11 M) detection sensitivity for fluorescent dyes and nanomolar sensitivity (10−9 M) for fluorescamine-labeled proteins. Novak et al. built a low-cost mini-fluorescence detection system of 30 × 30 × 11 mm (Fig. 4b), in which LED, photodiode, mirrors, lenses, and electronic circuit are housed together, with a sensitivity in the low nanomolar range [56]. In 2010, Backhouse et al. presented a hand-held isotachophoresis microchip device [240 g, (Fig. 4c)] with LIF detection, which was powered by a universal serial bus link connected to a laptop computer. A limit of detection of 100 pM for Alexa Fluor 647 was demonstrated [57].

Fig. 4
figure 4_144

Examples of portable microfluidic instruments. (a) The system integrates fluidics, microseparation chips, lasers, optics, high-voltage power supplies, electronic controls, data algorithms, and a user interface into a hand-portable instrument (Reprinted from [55] with permission of The American Chemistry Society). (b) A miniature LED-induced fluorescence microdevice (Reprinted from [56] with permission of The Royal Society of Chemistry). (c) Hand-held isotachophoresis instrument (dimensions: 7.6 × 5.7 × 3.8 cm) (Reprinted from [57] with permission of The Royal Society of Chemistry)

Full size image

Making the size of each optical component for LIF as small as possible and assembling them onto microfluidic systems is a straightforward way to develop a portable analyzer as described above. Another idea to minimize the instruments is to design and fabricate all optics directly on-chip, which means that the optics such as light source [58–60], lens [61, 62], and detector [63] are part of the microchip rather than minimizing them and interfacing them with microchip. This self-contained optical microfluidic, also called optofluidic, is reviewed in the Sect. 3 of this article.

2.3 Chemiluminescence

Chemiluminescence (CL) is a molecular luminescence spectrometry, originating from the release of energy in a chemical reaction: a metastable reaction intermediate or product in a singlet excited state relaxes to the ground state with the emission of a photon. By measuring the as-generated luminous intensity, one can conveniently determine the analyte concentration. Compared to other spectroscopic techniques, there is an absence of any excitation light source is the inherent advantage of CL detection, which not only eliminates the problematic background emission from interfering compounds or from the microfluidic substrate itself. CL is also more appropriate for miniaturization of the on-line detection system and to build a portable microfluidic instrument [64, 65]. Thus, CL is a promising alternative detection technique to be coupled to microfluidic analysis for a miniaturized device. However, CL is limited to some special reaction reagents, such as luminol and its derivatives [66], peroxyalate and oxalate [64], and ruthenium complexes and to direct oxidations [65]. Till now, CL detection coupled to capillary electrophoresis has been well reviewed [67–70]. Herein, we briefly introduce recent developments in the use of CL on microchips.

In 1998, CL detection coupled to microfluidic was first described by Mangru and Harrison [66]. Based on horseradish peroxidase (HRP)-catalyzed reaction of luminol with peroxide, they designed a postseparation detection scheme on a microchip by transporting peroxide into the separation channel from a bypass channel. A 7 nM detection limit for the fluorescein conjugate of HRP was achieved. To optimize the microchip layout and reaction conditions for different CL reagents, Liu et al. [71] compared two chip layouts (cross-combining with Y, and cross-combining with V) in PDMS and evaluated their performance with respect to two model CL systems that included the metal-ion-catalyzed luminol–peroxide reaction and the dansyl species conjugated peroxalate–peroxide reaction. A comparison between these chip layouts showed that the chip pattern of cross-combining with Y may be more promising for the luminol–peroxide CL system, while the chip pattern of cross-combining with V was preferred for the peroxalate–peroxide system.

Hashimoto et al. [64] simplified the chip layout without etching the manifolds. The detection point was set at the interface of separation channel and reaction reservoir. They successfully separated and detected dansyl amino acids (glycine and glycine) using peroxyoxalate CL reaction using CL detection-based microchip capillary electrophoresis. They further incorporated an immune reaction with CL detection to determine human serum albumin or immunosuppressive acidic protein as a cancer marker in human serum [72].

After CL detection was successfully introduced to the microchip and optimal channel design and reaction conditions were obtained, it was used for high-sensitivity detection, especially for single cells analysis. Zhao et al. [73] first reported that CL detection with microchip electrophoresis was used to determine the intracellular content of glutathione in single human red blood cells. CL detection showed about 100 times more sensitivity than LIF detection for single cell analysis. The same group subsequently quantified Gly and Ala in single rat hepatocytes cells [74], taurine and amino acids in mice single fibrosarcoma cells [75], and sulfydryl compounds in single red blood cells [76]. They further developed a CL resonance energy transfer method to detect the amino acids in single human red blood cells [77]. Besides intracellular content analysis, the counting of intact cells by CL detection on microfluidic chip was designed by Wang et al. [78]. They counted CD4-positive (CD4+) T-cells from whole blood by capturing CD4+ T-cells with anti-CD4 antibody. The trapped cells were then bound to CD3 antibodies conjugated with HRP and incubated with chemiluminescent substrate. The emitted photons were proportional to the number of captured CD4+ T-cells. The results can be provided automatically and digitally within seconds of running the test. Chemiluminescence detection on the microfluidic chip provides the highest potential to reduce the size of instruments compared with other optical detection systems due to the lack of an external light source, which makes it possible to produce low-cost point-of-care diagnosis microdevices. However, the crucial drawback of CL detection in microfluidics is its poor reproducibility because CL measurements are strongly affected by many factors, including temperature, pH, solvent, solution composition, etc. Furthermore, the number of CL reagents is limited.

2.4 Surface Plasmon Resonance and Localized Surface Plasmon Resonance

Surface plasmon resonance (SPR) is an optical method for probing the refractive index changes of medium (sample) in contact with the surface of a noble metal film in which coherent electron oscillations result in resonance. SPR is a particularly suitable technique for measuring rates and equilibrium constants of processes that involve molecular interactions and adsorption of biomacromolecules on surfaces. A light wave is coupled to the metal film to excite the surface plasmons. If analytes bind to the metal film surface, a change in the refractive index of the medium alters the characteristics of the light wave coupled to the surface plasmon. Measuring the alteration of the excitation light reflects the species and concentrations of the analytes. The basic principles, couplers and applications of SPR have been well documented by Homola [79]. An SPR spectroscope usually incorporates a flow cell to confine and control the sample flowing through the sensing surface. Microfluidic chips provide an easy manner for sensing very small amounts of samples without an optical tag when SPR or localized surface plasmon resonance (LSPR) is integrated onto microchips.

Coupling SPR to microfluidic systems is usually implemented by prism, waveguide, and metal grating. In the prism coupling mode, the walls of microfluidic channels are deposited with a layer of gold film for sensing the change of analytes flowing through the film [80–82]. The excited light is totally internally reflected by the prism, which is in contact with the metal film, and is introduced to the metal film through evanescent wave. Figure 5a shows a schematic of prism-coupled SPR for a microfluidic chip. This traditional coupling approach is simple, sensitive, and robust, but it is less advantageous in miniaturization and integration. Optical fibers and waveguide have been developed as alternatives to the prism to generate evanescent waves and to excite the surface plasmon. Kurihara etched a fiber to form a gold-coated microtip at the end of the fiber and inserted it into a microchannel [83]. The coated microtip has features like a microprism (Fig. 5b). Fiber-based SPR on-chip has the problem of alignment and assembly. To overcome such issues, it was proposed that the planar optical waveguide coupling mode should be integrated on the microfluidic chip (Fig. 5c) [84, 85]. The waveguide-based SPR on-chip has the potential of high throughput analysis of multiple samples in parallel because multiple waveguide structures can be fabricated into one single platform. Another promising alternative method for the miniaturization is SPR-based grating, which is low-cost and can be mass-produced by disposable polymer materials instead of the glass chip used for prism or waveguide SPR [86, 87]. In this coupling mode, a metal grating profile coated on the channel diffracts the incident light wave, which is capable of matching the momentum of the plasmon and the wave and generating a surface plasmon (Fig. 5d). In this approach, noncontact optical excitation obviates the need for a prism and index-matching fluids and greatly simplifies instrument configuration.

Fig. 5
figure 5_144

Coupling schemes of incident light to surface plasmons: (a) prism coupling, (b) experimental setup for the fiber-optic SPR microsensor, (c) waveguide coupling (Reprinted from [83] with permission of Elsevier), and (d) grating coupling

Full size image

The excitation of surface plasmons that occurs at the planar surfaces is denoted as SPR; that which occurs at the surface of nanometer-sized metallic structures is called LSPR. SPR-based systems require coating of metal film on the walls of the microchannel and finding an appropriate way to couple the light to the surface plasmon. However, LSPR merely requires the immobilization of metal nanoparticles in the channel [88], which substantially reduces the fabrication cost and the optical setup complexity. LSPR seems to be sensitive enough on the microfluidic chip because Cao et al. have measured the changes in the LSPR signal of a single Au nanoparticle by dark field microscopy [89].

2.5 Surface Enhanced Raman Spectroscopy/Imaging

Surface enhanced Raman spectroscopy (SERS) occurring on the surfaces of noble metal substrates or nanoparticles and is another highly sensitive detection method based on surface plasmon, without requiring labeling. In contrast to conventional Raman spectroscopy, the detection sensitivity of SERS is enhanced up to 6–10 orders of magnitude. The advantages of SERS detection on microfluidic chips are expected to yield reliable and reproducible results and to obtain precise quantitative measurement because the experimental conditions in the microchannel are easier to control than that of bulk SERS, which has been discussed in a review of chip-based SERS published in 2008 [90]. Besides probing the fingerprint of molecules, a SERS-coupled microfluidic device has been used for characterization of single Chinese hamster ovary cells, into which gold nanoparticles were introduced for Raman enhancement [91]. Herein, we only focus on progress published after 2007 and in which metal nanoparticles were exploited as the SERS enhancer.

The targets of present SERS on-chip are (1) to solve the problem of poor detection sensitivity caused by the small sample volume and (2) to develop robust methods to obtain a reliable and reproducible SERS signal. Aggregation of metal nanoparticles and mixing of enhancer with analytes are the straightforward ways to increase the SERS signal intensity and the uniformity of the SERS active site.

In general, SERS signal intensity is related to the number of molecules, Raman cross-section and the enhancement factors. Among these parameters, molecular quantity is the only controllable variable for the signal enhancement at given experimental conditions. Therefore, increasing the amount of absorbed molecules in the detection area can amplify the SERS signal. Various ways have been proposed for accumulating nanoparticles in microdevices and three are listed below:

  1. 1.

    Hampering nanoparticles by a dam or restricting them in reservoir: Choi et al. [92] concentrated rhodamine 6G-adsorbed gold nanoparticles by repeating a “filling–drying” cycle on a (compact disk) CD-based platform. They readily achieved a SERS signal of 1 nM rhodamine 6G. Wang et al. [93] developed a pinched and step microfluidic–nanochannel junction like an inverted dam, where nanoparticles and target molecules were stopped and assembled, and obtained electromagnetic enhancement factor of ~108. This technique was also used for monitoring the conformational transition of beta-amyloid peptide [94].

  2. 2.

    Use of electrokinetic force: Huh et al. [95] embedded microwells in a microchannel, the top and bottom of which were fabricated with a pair of electrodes. When electrically actuated, gold nanoparticles bound with DNA molecules were attracted into the microwells by the electric field, and the sensitivity of SERS was significantly increased. The authors reported the detection limit of 30 pM for Dengue sequences. Similarly, Cho et al. [96] reported accumulation of electrokinetically driven charge molecules on an Au surface and amplified the SERS signal by almost eight orders of magnitude.

  3. 3.

    Coupling a laser tweezer to a microdevice for aggregating metal nanoparticles: When nanoparticles flowed through focused red laser beam, they were trapped by the laser beam and assembled together. Another green laser was used for SERS excitation. Although the instrument was complex, it provided potential for controllable aggregation of nanoparticles and sequential analysis of samples in one run [97]. Based on above advances, we will not be surprised if magnetic force will be used for trapping metal enhancer in the near future.

SERS detection is restricted in quantitative analysis due to its poor reproducible results. A consistent and homogeneous mixing between the analytes and nanoparticles before they arrive the detection point is helpful in improving reproducibility. Therefore, various shapes of microchannels have been designed to increase the mixing effect, such as alligator-teeth, zigzag, 3D channel, etc [90]. Recently, Quang et al. [98] fabricated a pillar array in a PDMS microfluidic channel to efficiently mix sample and enhancer. Two confluent streams of analytes and silver nanocolloids were introduced into the pillar array channel and then the SERS spectra were recorded at the downstream end of the channel. Using this pillar array channel-based SERS, they achieved the quantitative analysis of dipicolinic acid and malachite green, and the detection limits were estimated to be 200 and 500 ppb, respectively.

3 Detection Inside the Chip

Optical detection inside the chip means that optical detection relevant units, such as light sources, filters, mirrors, waveguides, lenses, and detectors, are integrated on the chip. Those microfluidic chips integrated with optical components that are merged with or made of fluids are also called “optofluidic.” It is obvious that the more functional units, including optical components, are integrated on the chip, the degree of miniaturization of the microdevice is higher. Compared with free-space optics systems, the integrated optical microfluidic system is alignment-free and technician-independent [99]. Since such integrated microsystem is compact and firm, it is insensitive to shock and vibration, and can be used under extreme environments. In addition, if the light is well guided inside the chip, it can minimize light loss and the emergence of stray light as much as possible, increase the utilization efficiency of probing light, and decrease the background. Most current progress on detection inside a chip is concentrated on developing various strategies for assembly of optical elements on the chip. For example, previous reviews have discussed the structure and fabrication of the optofluidic waveguide [100], optofluidic microlens [101], microscope on-chip [102] and so on. With the advance of manufacturing technologies, integrated optical devices in the near future would be less expensive, with response times much faster than ever before. The full integration of optical elements into microfluidic chip is becoming possible.

3.1 Planar Optical Waveguide

Optical waveguide is a physical structure that transmits light along its axis, which is generally composed of “core/cladding.” A planar optical waveguide is a waveguide fabricated in a flat platform, and herein refers particularly to microfluidic chip. A waveguide is a key optical element for integrated optofluidics, and has the same importance as the fluidic channel in directing fluids. When the optical waveguide and the fluidic guide are merged together, the degree of integration is maximum. In microfluidic systems, planar waveguides have been used to enhance the detection sensitivity, such as focusing the light to the channel to increase the excitation power for fluorescence measurements [103], extending the optical path length in absorbance detection [104], and splitting the laser beam for interferometry sensing [105, 106]. The current status is that two types of waveguides have been assembled on microchips. One type is based on the refractive index, whereby the light is confined in the core material by total internal reflection because the core has a higher refractive index than that of the cladding. There is an electric field in the cladding layer, which typically penetrates 100–200 nm depths and decays exponentially outside the core. This field is called the evanescent wave. The other type of waveguide is the interference-based waveguide, in which the light is reflected back to the core by interference rather than internal reflection; therefore, the refractive index of the core need not be larger than that of the cladding.

3.1.1 Evanescent Wave-Based Waveguide

Depending on the materials used for core and cladding, waveguides based on total internal reflection are divided into three types: solid-state waveguides, liquid-core waveguides (LCWs), and liquid-core/liquid-cladding (L2) waveguides.

Solid-state waveguides are usually buried in the chip and intersect the fluidic channel to excite the analytes and collect the response signal. Solid-state waveguides are fabricated with the same materials as the chip, such as silica, glass and polymer. Silica and glass have the advantages over polymer materials of being chemically inert, biocompatible, and optically clear, and are usually used for improving sensitivity in absorption and fluorescence detection. Petersen et al. buried a doped silica waveguide on microfluidic separation devices for UV/Vis absorbance measurements [20]. A ninefold higher sensitivity was achieved with this approach compared to a conventional CE instrument with a 75-μm inner diameter capillary. The authors further fabricated 128 waveguides on-chip for multipoint detection to measure particle velocities [107]. Cleary et al. fabricated a silica waveguide on a chip and realized time-correlated single-photon counting of nile blue with concentrations as low as 1.5 nM [108].

However, glass and silica materials suffer from fabrication and mass production difficulties due to their complex processing technology and high cost. A variety of recent developments in assembling waveguides on microdevices have concentrated on polymer materials such as PDMS [109, 110] and poly(methyl methacrylate) (PMMA) [111] for large-scale production and cost reduction. Very recently, Okagbare et al. [112] developed a cyclic olefin copolymer waveguide embedded in a PMMA fluidic chip. The evanescent wave outside the cyclic olefin copolymer waveguide was used to simultaneously excite fluorphores in 11 fluidic channels for high-throughput analyses. The limit of detection for an Alexa Fluor 647 dye solution was 7.1 × 10−20 mol at a signal-to-noise ratio of 2.

The solid state waveguide is assembled into the microchip but separately fabricated with the microchannels. The connection point of fluidic channel and solid waveguide is generally the interrogation area. However, LCWs share the same physical volume with the fluidic channel. The microchannel not only transports the sample, but also transmits the light, which greatly simplifies the design complexity and reduces light loss through the channel walls. If liquid (particularly water in a microfluidic chip, with refractive index of 1.33) is used as the core of waveguide, the refractive indices of the cladding should be smaller than that of water, based on the principle of total internal reflection. Teflon amorphous fluoropolymers (Teflon AF) with an indices of 1.29–1.31 is a suitable cladding material for a LCW in a microfluidic chip. Datta et al. [113, 114] coated Teflon AF on silicon and glass substrate as cladding layer, and fabricated the coated substrates into microchannels that were also LCWs to propagate light. Detection of nanomolar concentrations of fluorescent dye was achieved on the liquid-core waveguide integrated chips. In order to reduce the time-consuming and high-cost fabrication processes using glass and silica substrates, Wu et al. [115] coated Teflon on the PDMS substrate and assembled a liquid-core waveguide on a PDMS/glass hybrid microfluidic chip to sense nitrite. Cho et al. [116] developed a method to construct LCWs in PDMS microchannels by flowing Teflon AF solution through the channels. This coating approach avoids the difficulties of bonding between two Teflon-coated PDMS substrates [115]. A micro fluorescence-activated cell sorter in the Teflon-coated LCW was further demonstrated. The laser was introduced into the microchannel, i.e., the LCW, from its one end and the prestained cells in microchannel were all excited. Therefore, the cells can be sorted into a split channel at the downstream side [117]. Replacing a commercialized Teflon tube with a fabricated liquid core in channel is an alternative method for improving sensitivity in a microdevice. For example, Du et al. [118] coupled a Teflon AF capillary to the chip channel as the detection cell for microchip absorption detection and achieved an effective optical path length of 17.4 mm with a detection volume of only 90 nL. In 2010, Pan et al. [27] integrated Teflon LCWs, photometer, LEDs and other components into a hand-held instrument. Teflon has long been the only option for low refractive index cladding for LCWs until nanoporous dielectrics were developed. Nanoporous materials are formed with a high-index solid material by adding air pores, therefore the refractive index can be adjusted by varying the air fraction. At the same time, the wall of nanoporous cladding can be chemically functionalized, overcoming some of the adhesion problems of Teflon [119]. Korampally et al. fabricated nanoporous film of low refractive index on the silicon microchannel and developed a miniaturized LCW [120].

L2 waveguides are formed by two or more different laminar liquids of different refractive index flowing inside a fluidic channel. The index of the cladding liquid is smaller than that of the core liquid so that the light is guided in the channel by the total internal reflection that occurred at the interface of the liquids. By changing the flow rates of the liquids, the size and the direction of the waveguides can be reconstructed (Fig. 6) [121, 122]. The advantages of L2 waveguides over solid-state waveguide are: (1) the optical performance is tunable by controlling the fluidic properties; and (2) the smooth interfaces between the core and cladding streams minimize optical loss. To couple an L2 waveguide with a biosensor, Rosenauer et al. [123] fabricated a liquid-core/liquid-twin-cladding waveguide for on-chip fluorescence spectroscopy. The design included an inner and an outer cladding fluid outside of the core fluid; analyte was suspended in the inner cladding fluid and excited by the evanescent wave at the fluid interface of core and inner cladding. By tuning the flow rate ratios of the three different liquids, considerable reduction in sample volume and a low background noise were achieved. The detection limit of rhodamine B dye was 20 μM. The L2 waveguide in a 2D plane suffers from optical loss at the top and bottom interfaces of core and channel walls. Lee et al. [124] therefore developed a 3D microfluidic L2 waveguide by using microfluidic drifting technology [125], where the liquid core cannot come in contact with the channel walls because it is hydrodynamically focused in the center of the microchannel by the cladding fluids. In addition, the laminar flow-based L2 configuration has been widely applied to construct other optical components, such as reconfigurable lenses [126], optical switches [127], and tunable micro-optofluidic prisms [128]. Although current L2 waveguides may be restricted by the sensitivity of the stream interface to vibrations and diffusion effects, and limited on finding liquids with specific refractive indexes, the tunability of L2 waveguides makes it possible to provide functionality and integration in future optofluidic devices.

Fig. 6
figure 6_144

Three parallel waveguides formed with liquid core and cladding in laminar flow systems. The direction of the light propagation can be altered by different flow rates of the adjacent fluids. hv refers to the energy of light; h is the Plank's constant; v is the frequency of the light. (Reprinted from [121] with permission of The American Chemistry Society)

Full size image

3.1.2 Interference-Based Waveguide

The interference-based waveguide is a conceptually different approach for realizing LCWs in microfluidic chips. In contrast to the evanescent wave-based waveguides, interference-based waveguides do not require a cladding material with a lower index than that of the core material. Light is multiply reflected from a periodic dielectric cladding layer by wave interference. There are two types of interference-based waveguides: photonic crystal (PC) waveguides and antiresonant reflecting optical waveguides (ARROWs).

The integration of PC fibers on glass chip has been demonstrated by means of femtosecond laser exposure and chemical etching [129]. However, so far no microfabricated PC-based LCW has been reported. ARROWs have attracted more attention from researchers in the microfluidic chip field due to their relatively easy fabrication from accessible materials using current micro-electromechanical techniques. The main feature of ARROWs is that light is confined by Fabry–Perot antiresonant reflectors, rather than total internal reflection. Therefore, ARROWs can be constructed in a low refractive index core, such as an aqueous solution, surrounded by high refractive index cladding layer such as a silicon substrate. Planar integrated networks of intersecting solid and LCWs on a chip were used for single dye molecule detection [130] and single liposome particle detection [131]. Further information about integrated waveguiding can be found in reviews by Schmidt et al. [100, 132].

3.2 Microlens

The microlens is an important and key optical component for focusing and collimating light in the microfluidic optical detection system, and can be made up of a hard solid surface or interface (soft solid/liquid, liquid/liquid/…). Hard solid state lenses on-chip are usually fixed focal length lenses similar to the miniaturized traditional lenses used in the free-space detection system. They can be fabricated on-chip with the optical axial vertical to the chip plane, focusing light to the channel from top or bottom surface [133], as well as with optical axis parallel to the plane, introducing light from the side wall of the chip [25, 134]. Although solid lenses suffer from the complexity of fabrication and alignment, they are robust and vibration resistant. Recently, such microlenses have been fabricated on a SU-8/PDMS hybrid microfluidic chip for optical manipulation of microparticles [135]. Other promising and popular microlenses on-chip utilize the curvature formed by the interface between two liquids, between liquid and soft solid, or between liquid and air. Such microlenses are also called optofluidic microlens, and they are dynamic optical elements and their focal lengths can be easily tuned by changing fluid characteristics such as species, concentration, and flow rate. Next we will focus the discussion on optofluidic microlenses.

The major advantages of these interface-based microlenses are that they are reconfigurable and optically smooth meniscus surfaces. Light can be modulated by simple adjustment of the flow conditions, thus eliminating the need for mechanical or electrical parts for manipulating the lens on the microchip.

Elastic PDMS is a commonly used polymer material for fabricating microfluidic chip and because it is readily curved by air pressure or liquid hydrodynamic, it has been used to shape a curvature and form a microlens by deforming with pneumatic force. For example, a PDMS wall with a thickness of 50 μm between chambers and microchannel can be distorted by applying liquid pressures flowing through chambers. The deformation of the PDMS wall and the index-matching fluid act as a double convex lens to focus both the excitation light source and the fluorescent emission signal. The focal length can be adjusted through different liquid pressures [136]. Changing the curvature of the PDMS surface can also be realized out-of-plane by exerting electrical force [137] and electromagnetic force [138]. However, the distortion of the surface is limited, and a large curvature of the PDMS surface is hard to form by external force. To address this issue, Shi et al. [139] fabricated a tunable in-plane optofluidic microlens set involving a static PDMS lens and an air–water interface that diverged the light and was reshaped by adjusting the flow rate of the injected water. The tunable divergence in the incident light beam, in turn tuning the overall focal length of the microlens, was converged by the static PDMS lens. The microlens achieved tunable focal distance of 11 mm and a ninefold light intensity enhancement at the focal point. In addition, the liquid–liquid interface promises another way of forming a large curvature microlens.

Liquid–liquid microlenses were generated for the first time on-chip by Mao et al. [140]. Subsequently, Rosenauer et al. designed a hydrodynamically adjustable convex lens with 3D light focusing ability instead of line focusing [141] (Fig. 7). The authors co-injected miscible fluids of different refractive indices and formed the interface in the channel. When the liquids flow through a 90° curve in a microchannel, a centrifugal effect also curves the fluidic interface and yields a reliable cylindrical microlens. The curvature of the fluidic interface can be altered by simply changing the flow rate. Higher flow rates generate a microlens with larger curvature and, hence shorter focal length. Tang et al. [142] reported another L2 lens formed by laminar flow of three streams of fluids (Fig. 8). The refractive index of the central (core) stream was higher than that of the sandwiching (cladding) streams. When the streams entered an expansion chamber, the lens was formed. The curvature of the interface between the liquid core and the liquid cladding can be modulated by adjusting the relative flow rates, resulting in variable focal distance. As a result of the change of the expansion chamber to a circular chamber, interfaces with a perfect arc shape and a short focal length can be obtained [143, 144]. However, diffusion of solute across the fluid–fluid interface deteriorates the smoothness of the interface. Therefore, Huang’s group developed in-plane [126] and out-of-plane [145] lenses on the microchip that were dependent on the liquid gradient refractive index (L-GRIN), and focus light due to the gradient of the refractive index within the liquid medium, rather than due to curved refractive interfaces (Fig. 9).

Fig. 7
figure 7_144

A 3D light-focusing optofluidic microlens. The displayed cross-sections show the shaping of the lens and the refractive index step defined by the choice of core and cladding fluids (Reprinted from [141] with permission of The Royal Society of Chemistry)

Full size image
Fig. 8
figure 8_144

(a) Experimental setup for focusing light exiting an optical fiber through the liquid-core/liquid-cladding (L2) lens. (b) Bright-field image of the L2 lens. The green laser beam from the fiber is visible in front of the aperture because PDMS contains nanoparticles of silica that scatter light. The focused beam in the beam-tracing chamber is orange due to the fluorescence of the dye. (Reprinted from [142] with permission of The Royal Society of Chemistry)

Full size image
Fig. 9
figure 9_144

Principle and design of the L-GRIN lens. (a) Comparison between the classic refractive lens (a1) and GRIN lens (a2). Change of the refractive index contrast in GRIN lens can result in change of focal distance (a2, a3), and shift of optical axis can result in change of output light direction (a4). (b) Schematic of the L-GRIN lens design (b1), microscopic image of the L-GRIN lens in operation (b2, left), refractive index distribution at two locations (b2, right) (c) Drawings showing two operation modes of the L-GRIN lens: the translation mode with variable focal length including no-focusing (c1), a large focal distance (c2), and a small focal distance (c3); and the swing mode with variable output light direction (c3–c5) (Reprinted from [126] with permission of The Royal Society of Chemistry)

Full size image

3.3 Optofluidic Laser

To achieve the ultimate goal of integration of many optical components onto a compact chip, laser integration on-chip is a key challenge, although conventional bulky lasers have been widely applied in microfluidic systems through free-space coupling [7, 8]. The miniaturized laser had been independently developed before the microfluidic chip was put forth. However, only when the microfluidic device and microphotonic system are synergetically combined, can the optofluidic chip be successful. Therefore, we will not attempt to discuss light sources that are miniaturized and assembled on the microchip but have nothing to do with fluids, although they have made contributions to the development of laser miniaturization. For example, the laser diode, photodetector, and waveguide have all been integrated on a silicon chip [146], as well as an organic semiconductor film laser assembly on a PMMA substrate [147]. The dye laser is currently the optofluidic laser most usually integrated on microfluidic chips [148]. The advantages of optofluidic dye lasers are that they can be fabricated on the same materials as the microfluidic system and readily integrated with other fluidic functionalities, and that the characteristics of dye lasers are amenable to tuning by changing the dye solution. Since the first optofluidic dye laser was reported in 2003 [149], rapid and significant progress has been obtained in terms of various configurations and materials. Lasers consists of three principal elements: gain medium, pump source, and resonator cavity. Since the pump sources for the present optofluidic dye laser are all external light sources such as lasers, and the gain mediums are various dyes in different solvents, both of which are less related to chip structure, we will summarize the development of optofluidic lasers with respect to the resonator, which is an optical feedback volume that repeatedly transmits a resonant light through the gain medium to create a very intense, coherent laser. This optical feedback system is conventionally called the optical resonator or optical cavity.

The simplest optical cavity comprises only two facing plane mirrors, denoted as a Fabry–Perot cavity. Helbo et al. [149] first deposited a pair of opposing metal mirrors on the top and bottom of a microchannel acting as a resonator in which rhodamine 6G solution flew through as the gain medium (Fig. 10). Song and Psaltis have recently developed a pneumatically tunable optofluidic dye laser [150], in which a LCW filled with dye solution is the resonator at the ends of which two air chambers act as the mirrors. The wavelength is chosen by inflating the air-gap etalon with compressed air.

Fig. 10
figure 10_144

Cross-sectional outline of the microcavity fluidic dye laser. The laser dye is pumped through a microfluidic channel. The laser cavity is formed inside the microfluidic channel by metal mirrors, deposited on the top and bottom glass wafers. The microcavity fluidic dye laser is pumped optically by an external laser, and output is emitted through the semitransparent top mirror (Reprinted from [149] with permission of IOP Publishing)

Full size image

A ring resonator is the second simple configuration for a cavity consisting of three or more mirrors to loop the light in the resonator. Galas et al. fabricated a ring resonator using four mirrors (air chamber) to generate a tunable dye laser [151]. Suter et al. [152] presented side-coupled ring geometry to generate an optofluidic ring resonator laser, which can be produced in PDMS substrates with contact molding (Fig. 11). A kind of variation of the ring cavity structures is the droplet-based whispering gallery mode. A droplet can be treated as an optical microcavity when the index of refraction of enclosed dye solution is higher than that of the carrier liquid. Light emitting from the dyes is confined and amplified in the drops by total internal reflection. This method could provide a multicolor laser in a train of alternating dye drops and is capable of switching the wavelength at high frequency [153].

Fig. 11
figure 11_144

(a) Cross-sectional outline of the microcavity fluidic dye laser; WGM whispering gallery mode. (b) Photograph of a glass/SU-8 chip with a microcavity fluidic dye laser (Reprinted from [152] with permission of The Optical Society of America)

Full size image

Besides the mirror-based optical cavity, there is another type of optical feedback structure involving diffraction grating, called distributed feedback (DFB). The group of Kristensen demonstrated a 130th order Bragg grating-based optofluidic dye laser [154] to achieve a single mode laser. However, lower-order DFB gratings are expected to minimize the intrinsic scattering losses and decrease the threshold of pump energies. Therefore, the15th order [155] and third order [156] Bragg gratings were successively fabricated on a microfluidic chip. Song et al. [157] have systematically studied third, second, and first order DFB lasers, where the second order lasers exhibited the lowest pump threshold. Vannahme et al. [158] embedded a first order DFB optofluidic dye laser in a foil for mass production and cost reduction.

At least two issues should be improved to fulfill the promise of optofluidic dye lasers as on-chip coherent light sources. Firstly, new strategies should be developed to widen the laser wavelength range and to reduce the tuning time. Secondly, the optical pump source should be integrated onto the same laser chip, which should demonstrate the lowest laser threshold.

3.4 Optical Detector

Optical detectors have been integrated on microfluidic chips, such as silicon photodiode, organic photodiodes (OPDs), and complementary metal–oxide–semiconductor (CMOS) chips. The former two are point detectors without spatial resolution, whereas CMOS detectors can capture an image. In the early stage, integrated detectors on microfluidic chips were mainly silicon detectors due to the mature fabrication technique of semiconductors, e.g., a microavalanche photodiode on PDMS [159], a photodiode built on a silicon substrate [160], and a hydrogenated amorphous silicon photodiode assembly on glass [161]. Generally speaking, the fabrication procedure for a detector on-chip becomes less and less complex, and the cost of detector trends to become lower and lower. For example, if the substrates are changed from silicon to glass and then to polymer, the fabrication temperature is decreased from a high temperature to ambient temperature. Recently, Pereira et al. presented an example by integrating a PIN photodiode on a microfluidic chip to perform chemiluminescent detection, and the detection results were comparable with laboratory analysis for detection of HRP [162].

Silicon photodiodes could render relatively high sensitivity for detection of low analyte concentrations but are too expensive and complicated to fabricate as an integral part of a disposable microfluidic chip. In contrast to the silicon counterpart, OPDs are a promising potential for future point-of-care technology because they are inexpensive and easily fabricated, and are attracting more and more attention from researchers in the field of lab-on-chip.

Hofmann et al. [163] showed the feasibility of integration of thin-film OPDs on microchips for microscale chemiluminescence. The detector was fabricated by a layer-by-layer vacuum deposition method. The quantitation of hydrogen peroxide yielded a detection limit of 1 mM. They further implemented the integration of OPD on PDMS instead of silicon as the substrate material [164] and applied it to analyze antioxidant capacity [165]. To reduce the background and increase the detection limit of the OPD detector on chips, Pais [166] placed a second polarizer layer before fluorescence entered into the OPD detector by preventing the leaking of excitation light into the detector. The detectable concentrations of rhodamine 6 G and fluorescein dyes were as low as 100 nM and 10 μM, respectively. Wojciechowski et al. [167] developed a chemiluminescence immunoassay-based miniaturized biosensor, which was composed of a disposable microscope slide and a hand-held reader. The OPD substrate was printed on the glass slide. This OPD system is approaching the final requirements of point-of-care, i.e., inexpensive, easy to use, highly sensitive, and robust.

Besides the various photodiodes mentioned above integrated on microfluidic chips, the CMOS is another type of detector that can be assembled on microchips and offers the great advantage of the ability to observe the image of sample object [63]. One of the most important applications of an integrated CMOS detector on-chip is the lensless optofluidic microscopy (OFM). In 2005, Lange et al. [168] attached a CMOS chip at the bottom of a microfluidic chip to record the projection image of Caenorhabditis elegans. But, the quality of this imaging method needs to be further improved by choosing appropriate apertures. Yang et al. [169, 170] were inspired to develop a high-resolution optofluidic microscope by covering a nanohole array layer on the imaging chip. They measured a resolution limit of 490 ± 40 nm, comparable resolution to a conventional microscope. The principle of lensless microscopy is shown in Fig. 12. The authors extended this method to color OFM by integrating a RGB CMOS color sensor chip substrate [171]. Laurberg et al. then exploited OFM to 3D spatial resolution and on-chip particle image velocimetry [172].

Fig. 12
figure 12_144

Principles of nanohole microscopy. (a) Direct projection imaging scheme. (b) By placing the specimen on a grid of apertures, we can obtain a sparse image. (c) Raster-scanning the specimen over the aperture grid. In this case, the image resolution is limited by the aperture size. (d) The scanning scheme can be simplified into a single-pass flow of the specimen across the grid. (e) The aperture grid can be simplified by substitution with a long linear aperture array (Reprinted from [170] with permission of The National Academy of Science, USA)

Full size image

Yang’s group further presented a non-nanohole method for reconstruction of a high-resolution image from a low-resolution image processed with a pixel super-resolution algorithm [173]. This method further compacts the OFM by eliminating the nanoaperture array, and follows the consistent requirement of microfluidic chip in miniaturization as small as possible. The rapid development of OFM is following the advances in optofluidics but also contributes to them.

4 Conclusions

Optical detection still maintains its predominant application in microfluidics due to its characteristics of being noninvasive, easily coupled, quick to respond, and highly sensitive. As discussed, although conventional free-space detection strategy is less amenable to being miniaturized and compacted in a portable microdevice, it is highly familiar to analytical chemists, and is still helpful when a proof-of-principle needs to be quickly demonstrated on microfluidic chips and extreme sensitivity is a particular concern. As the most promising alternative to free-space detection, optofluidics is becoming an active field in lab-on-a-chip, merging fluidics with optics in the same microchip, in which optics are generated by fluids and fluids are controlled by optics. Current optofluidics is developing and optimizing single optics, but has potential to minimize and integrate all components onto the same microscale device, which includes analytical units and optical pieces. In optofluidics, various optical components such as light source, filters, microlens, detectors, and waveguides can be made of fluids; therefore, the space in the chip is greatly saved and the complexity of fabrication is simplified. The microchannel both handles the analytes and transmits light to probe the sample and collect analytical results. Future optofluidic chips undoubtedly can integrate more components and be mass-produced to meet the requirements of point-of-care diagnostics.