Photothermal spectroscopy and micro/nanofluidics

As is well known, the field of microfluidics was initiated by on-chip electrophoresis. A few years later, pressure-driven microfluidics was developed independently. Our contribution to this field was to initiate pressure-driven microfluidics and also pioneer nanofluidics. Regarding microfluidics, we proposed basic concepts and methods to innovate pressure-driven techniques for microfluidics and made the device technology more widely applicable than electrophoresis. Furthermore, we demonstrated some typical applications that are still a main current of microfluidics today. Another contribution was to pioneer nanofluidics, which has a channel size 1000-fold smaller than that of microfluidics, while the concepts and methods are the same. Ultimately, nanofluidics has been initiating small-scale chemical and biomedical sciences evolving from the molar scale to a relatively small number of molecules. Nanofluidics will surely bring breakthroughs to other science and technology efforts, such as the analysis of an absolute number of molecules, solution and interface chemistry, and fluid mechanics in nanoscale liquid spaces. Figure 1 shows a chronograph of microfluidics and nanofluidics with our contribution of initiating pressure-driven microfluidics and pioneering nanofluidics, which makes the timeline of progress clear. Changes in published papers shown in Fig. 1 also clearly show the development of this field. Microfluidics started in 1990 as capillary electrophoresis in microchannels fabricated on a glass substrate, and it was called a micro-total analytical system.^1,2 At almost the same time, we initiated pressure-driven microfluidics. Our method introduced some solvents and solutions into microchannels by pressure, and many types of chemical operations, such as mixing, reaction, and separation, were realized in microchannel networks. We called this method micro-unit operation (MUO) at first and continuous flow chemical processing (CFCP) afterward.³ This method was quite successful, and we pioneered and demonstrated a variety of microfluidics applications, as shown in Fig. 1.

Regarding the detection of the results of chemical processing or the monitoring of chemical processes in the device, the readout method for the microchannel on the device is critical, and we applied our thermal lens microscope (TLM), which can detect even nonfluorescent molecules with high sensitivity, to a variety of analysis and synthesis methods.⁴ The sensitivity of the TLM was almost the same as the laser-induced fluorescence method, and therefore, it became a powerful tool for microfluidics. However, as shown in Fig. 1, the development of the TLM and microfluidics occurred at the same time and was published in 1993 because our first microfluidic device was fabricated for experiments used to develop the TLM. Under an optical microscope, there was no space for a liquid vessel, and we placed a droplet on a glass substrate and placed a cover glass over it. The cover glass floating on the droplet vibrated freely and disturbed precise measurements of TLM experiments. Furthermore, the thickness of the sample solution became on the order of millimeters and convection occurred, which also disturbed the TLM measurement. Then, we fabricated a shallow straight channel on the order of 100 μm and confined sample solutions in it by pushing and attaching the cover glass tightly to the straight channel to obtain a stable signal. After this experiment, we fabricated an inlet and an outlet to the channel, the channel shape became Y-shaped, and the number of inlets increased to six. This microchannel structure enabled reproducible measurements due to sample exchange without touching the device, and we succeeded in developing the TLM. After that, we used this Y-shaped device for mixing and reacting two solutions, forming aqueous/organic two-phase parallel flow and solvent extraction between the two-phase flows, and other kinds of chemical experiments. These experiments were the base of our MUO and CFCP methods. We summarized and systemized our methodology in 2002, and it became a generalized design method for our pressure-driven microfluidics. We pioneered the current applications of microfluidics, such as immunoassay, extraction, and other kinds of chemical and biological experiments, in the 1990s and implemented microfluidics for practical use in biomedicine and chemical synthesis, primarily chemical plant applications from the early 2000s up to now. The challenge of and transition from microfluidics to nanofluidics was a natural evolution in our group.^5,6 Though there were a lot of technical difficulties and unique solutions,⁷ the development of nanofluidics was similar to the R&D history of our microfluidics seen in Fig. 1. However, the TLM could not be applicable anymore because of the limitation of the geometrical optical configuration of the TLM principle in nanospace, and we decided to introduce wave optics. The details are given in a later section, but we also succeeded in developing a powerful photothermal detection tool for nanofluidics, which was the differential interference contrast (DIC)-TLM or photothermal optical phase shift (POPS) method.⁸ The detection method partially contributed to some recently developed nanofluidic devices.^9–13 

In this way, photothermal spectroscopy and micro/nanofluidics are intertwined. We would like to introduce this scientific background of microfluidics in this special issue celebrating Professor Andreas Mandelis. In this invited review, we summarize not only the TLM but other photothermal methods for sensitive detection in microfluidic and nanofluidic channels in Secs. II–IV. We intend to shed light on many challenges and uses of microfluidics and nanofluidics with photothermal spectroscopy.

For planar microfluidic chips, the collinear beam configuration is more favorable than the crossed-beam configuration used for flow cells and capillaries in conventional photothermal deflection spectrometry. The first TLM, which had a perfectly collinear beam configuration, was reported in 1993.⁴ In addition to traditional photothermal techniques, the principle of a TLM is based on the optical absorption of analyte molecules in the sample solution followed by nonradiative relaxation. When a laser beam is focused on the sample, the heat generated in the focal point diffuses into the solution. Because the temperature distribution reflects a Gaussian intensity distribution of the laser and the temperature coefficient of the refractive index (dn/dT) is negative for a liquid, a change in the refractive index that works as a concave lens is induced. This is called the thermal lens effect, and it can be detected by another laser beam. In a TLM, the excitation beam used for heating and the probe beam used for detection are focused using the same objective lens. Herein, a small shift of the two laser focal points leads to efficient refraction (diversion or conversion) of the probe beam by the thermal lens, as shown in Fig. 2. Theoretically, the optimum value of the shift is approximately the confocal length of the probe beam. For the detailed theory of the thermal lens and its detection, see other papers, books, and reviews.^14–17 

Figure 3 shows a schematic diagram of the experimental setup of a TLM. The shift of focal points is introduced by small divergences of collimated laser beams using laser beam expanders because chromatic aberration is compensated in modern objective lenses. The excitation beam is intensity-modulated using a mechanical chopper, and synchronous detection is performed using a lock-in amplifier for sensitive detection of the thermal lens effect. The modulation frequency f of the excitation beam is an important parameter to characterize the scale of thermal diffusion. The thermal diffusion length L is defined as $L= α / π f$ using the thermal diffusivity of solvent α. Herein, L is typically several micrometers when the solvent is water and f is 1 kHz, which allows sensitive detection even in microchannels.

The high sensitivity of the TLM and its marriage with microchips realized the detection of ultratrace amounts of molecules in a liquid. Tokeshi et al. reported the determination of a subyoctomole amount of nonfluorescent molecules.¹⁸ The limit of detection (LOD) described in the article reached 4 × 10⁻⁶ AU, and the estimated number of molecules in the focal volume was 0.34. Therefore, the analyte molecule concentration was predominantly zero, sometimes one, and rarely two considering the Poisson distribution. Then, the probability was time-averaged and detected through thermal phenomena. The temperature change induced in the microchannel was described as being 3.1 μK at the LOD. Although a TLM can detect nonfluorescent molecules, the molecule needs to have optical absorption at the excitation wavelength. In addition, a TLM cannot distinguish molecules that have absorption at the same wavelength. Therefore, coupled with separation technologies,¹⁹ molecular recognition processes such as chelate reaction²⁰ and immunoreaction²¹ are effective for selectivity. Since energy differences of the absorbed and emitted photons are released as heat in the case of fluorescent molecules, photothermal spectroscopy can be applied to measurements of fluorescence quantum yields.²² 

The rapid development of microfluidics accompanied the rise of nanofluidics, as indicated by studies of anomalous phenomena specific to the nanospace and proposals of functional devices using nanofluidics.²³ However, the size of the channel is also the limitation of a TLM because the spatial size of the thermal lens is several micrometers. The main three difficulties of photothermal detection in a nanochannel are summarized as follows. First, the amount of heat generated in the channel is proportional to the optical path length, that is, the depth of the channel. Second, the heat quickly dissipates into the channel wall because the material of the nanochannel is generally glass, which has a high thermal conductivity. Third, because the principle of the TLM is thermal lensing, detection of a nanoscale thermal lens is difficult. In other words, principles based on geometrical optics do not work in nanochannels that are smaller than the wavelength of light. Hence, new principles based on wave optics are required.

The DIC method is one of the modalities in microscopic imaging. As well as phase-contrast microscopy, DIC microscopy can visualize transparent objects by extracting differences in the refractive index using interference. By introducing the DIC method into a TLM, DIC-TLM, or a POPS detector, was developed.⁸ As illustrated in Fig. 2, the probe beam is separated into two beams and integrated again using a pair of DIC prisms. The excitation beam is not separated, which induces a photothermal effect only for one arm of probe beams. Then, a phase shift that is proportional to the change in refractive index is detected through interference.

A schematic diagram of the DIC-TLM is shown in Fig. 3. The key parameter for realizing this principle is the distance between the two arms of the probe beams, which is called the shear value. To obtain good phase contrast, the shear value was specifically designed to be 5.3 μm, which is approximately equal to the thermal diffusion length. In addition, to separate the probe beam but not the excitation beam, their polarization planes are controlled.

One of the advantages of a DIC-TLM over a normal TLM is background-free detection. In principle, the interference of probe beams is destructive when no heat is generated at the excitation focus. Therefore, the signal due to phase contrast is detected in a dark field, which improves the signal-to-background ratio and LOD. Moreover, this principle based on wave optics applies to nanochannels. Shimizu et al. reported a sensitive determination of the concentration of a nonfluorescent molecule in nanochannels.²⁴ Because the depth of the nanochannel was shorter than the confocal length, the detection volume was precisely determined to be 250 al for a 500-nm deep channel. Although the LOD for concentration was 2.4 μM, the number of detected molecules was 300 in the detection volume. Later, the same authors improved the LOD down to 30 molecules using UV excitation and nonlabeled bovine serum albumin.²⁵ As well as the nanochannel, the sensitive detection of absorbance was performed in femtoliter droplets by Maceiczyk et al. to investigate enzyme kinetics of β-galactosidase and monitor the activity of HL-60 cells at single cell level.²⁶ The limitation of the DIC-TLM is also the channel size. Because the dn/dT of water and glass are negative and positive, respectively, their phase shifts may cancel each other out, especially for a depth less than 100 nm. Le et al. solved the problem using another material that has a negative dn/dT and detected 800 molecules by sputtering TiO₂ on the bottom of a channel 50 nm deep.²⁷ The detection performance can be improved further by optimizing the thickness and modulation frequency.²⁸ Note that the optical path length change with temperature (ds/dT) should be considered for some materials with high thermal expansion coefficient because ds/dT can be positive by sample thickness change even when dn/dT is negative, which affects the focusing or defocusing behavior and the direction of the optical phase shift by the photothermal effect.^29,30

As mentioned above, even a DIC-TLM that uses interference suffers from a channel size limitation, and another detection principle is desirable for much smaller nanochannels. Optical diffraction is another well-known phenomenon based on wave optics. For transparent or nonabsorbent objects, the diffracted light intensity depends on the refractive index difference between the object and the surrounding medium. In the case of micro/nanofluidic devices, a fabricated channel serves as a diffractive object if the solutions in the channel and the surrounding glass substrate have different refractive indices. This means that optical diffraction can be used for measuring the refractive index of solutions in microchannels and nanochannels. Several papers have reported the use of optical diffraction by grating microchannels and nanochannels for refractive index measurement and monitoring of sample solutions. From the viewpoint of photothermal readout, optical diffraction by microchannels and nanochannels is useful due to its high sensitivity to refractive index change, that is, increased temperature due to the photothermal effect. Using optical diffraction in a single nanochannel, Tsuyama et al. developed photothermal optical diffraction (POD) for nanofluidic devices, as shown in Fig. 2.³¹ The detection principle of POD is as follows. When a laser beam is focused on a nanochannel, part of the light passes through the glass substrate because the typical width of the nanochannel is smaller than the focused beam spot. The difference in refractive index (Δn) between the solution and glass substrate gives rise to diffracted light in a far-field region, and the diffracted light intensity depends on Δn. The photothermal effect of analytes induced by coaxially focused excitation laser changes refractive indices of a solution and glass substrate by heat generation, thermal diffusion, and temperature change. Generally, dn/dT is negative for liquid solutions and positive for glass. For example, the dn/dT of water is −9.1 × 10⁻⁵ K⁻¹, while it is +9.8 × 10⁻⁶ K⁻¹ for fused silica. The opposite polarity enables Δn to increase even if the heat diffuses into the glass substrate, followed by an increase in the diffracted light intensity. Thus, the photothermal effect of analytes in the nanochannel can be measured as a diffracted light intensity change.

The experimental setup of a POD detector is shown in Fig. 3. The optical system is similar to a TLM and relatively simple. A coaxial probe laser and an excitation laser are focused onto a single nanochannel, and light diffracted by the nanochannel can be collected by a high-NA objective lens. Different from grating structures, which give strong diffracted light at specific angles, optical diffraction by a single nanochannel gives a broad diffraction pattern. Therefore, part of the diffracted light can be separated from the transmitted light by a slit. Although precise position alignment of the laser spot on the nanochannel is required, sensitive analyte detection can be realized without precise phase or polarization adjustment.

The characteristics and advantages of POD are as follows. First, the detection principle of the POD enables sensitive molecule detection in a nanochannel whose size is much smaller than the thermal diffusion length (on the order of micrometers for 1-kHz excitation modulation). Generally, thermal diffusion outside the channel leads to a sensitivity decrease by thermal loss and cancellation effect due to the difference in thermal and optical properties between solution and the glass substrate, which limits the application of photothermal detection methods to nanochannels. The POD method overcomes the above problems because thermal diffusion outside the channel also contributes to the signal as explained above. For example, the detection of sunset yellow FCF, a nonfluorescent dye, in an aqueous solution was realized in nanochannels 70 nm wide and deep. The LOD was 300 μm, which corresponds to 1200 molecules in a detection volume of 7 al (calculated from the focal spot diameter and channel dimensions). Experimental results showed that the sensitivity has 1.5th power dependence on the channel depth and 1.3th power dependence on the channel width, suggesting that thermal diffusion outside the channel had little effect.

Second, the sensitivity of diffraction-based photothermal detection depends not only on the thermal properties of solvents but also on the optical properties. It is well known that organic solvents with low thermal diffusivity and high dn/dT can improve the sensitivity of photothermal detection methods. In the case of diffraction-based photothermal detection, the optical property of the solvent also affects the sensitivity because changes in the initial diffracted light (background signal) and diffracted light intensity (photothermal signal) depend on the initial refractive index of the solvent. Theoretical calculations and experimental results show that the sensitivity of the POD was improved more than 30 times by organic solvents with suitable thermal and optical properties, and an LOD of 75 nM was realized for a Sudan IV hexane solution in nanochannels 400 nm wide and deep, which corresponds to an average of 10 molecules in a detection volume of 0.23 fl.³² 

Furthermore, the diffraction-based detection principle enables the detection of all analytes flowing in a nanochannel. Because the size of a nanochannel is smaller than the focused beam spot and the Rayleigh length, all analytes in the nanochannel pass through the detection region. Such an experimental setup also enables signal-based evaluation of measured analytes because all analytes experience almost same laser irradiation, which is useful for the nanoparticle counting analysis, as discussed later.³³ 

Although optical readout offers high sensitivity for photothermal detection, precise adjustment of excitation and probe beams is required to obtain the best performance. Recently, some other readout schemes have been proposed and applied to microfluidic systems. Pfeiffer and Nagl integrated a temperature-dependent luminescent sensor into a microfluidic channel for photothermal detection.³⁴ An approach that detects the temperature rise directly was also proposed by Kwon et al. using a nickel-resistive temperature detector.³⁵ Moreover, Fu et al. used a thermometer for photothermal immunoassay and even integrated the thermometer part as a bar-chart microfluidic chip.^36–40 These kinds of easy readout techniques may expand the possibilities of photothermal detection in terms of on-site detection and point-of-care (POC) diagnosis.

In the history of separation science, miniaturization has been highly significant for smaller-volume samples and higher separation efficiency. However, microscale separation is always alongside the difficulty of detection, which requires novel detection techniques with high sensitivity and versatility. Due to its nature of absorbance-based and label-free detection, photothermal detection has been used for chromatography since the 1980s.^41–44 Liquid chromatography combined with sensitive photothermal detection is still used in, for example, the food analysis.^45–48 Later, capillary electrophoresis was developed for separation.^49,50 Kitagawa et al. reported highly sensitive detection using a TLM coupled with capillary electrophoresis and an interface chip that realized an online sample preconcentration with a 3.9 × 10⁶-fold enhancement of sensitivity.¹⁹ Recently, the ultimate miniaturization of separation columns using the nanospace has been implemented using nanocapillary or nanofluidic channels fabricated on chips.^9,51,52 In such nanospaces smaller than the wavelength of visible light, fluorescence is almost the only detection method. However, the family of TLMs designed for nanospaces has been combined with nanoscale separation techniques. The DIC-TLM was first used to detect nonfluorescent dyes separated in a nanochannel.⁵³ The detection volume (21 fl) was precisely determined by the depth of the nanochannel (350 nm) and the diameter of the excitation spot. Then, the DIC-TLM was estimated to be able to detect 250 zmol injected into the nanochannel. The POD was also combined with nanofluidic chromatography in a nanochannel 800 nm wide and 300 nm deep.⁵⁴ The detection performance of the POD was 5.4 zmol (3300 molecules) for the same dye [Figs. 4(a)–4(c)]. Because reversed-phase separation and fabrication of a long nanochannel have been also achieved, nanofluidic chromatography has the potential to be a powerful tool for biological studies at the single-cell scale.^55,56 In this respect, introduction of UV excitation lasers is also noteworthy.^57,58 Specifically, a DIC-TLM succeeded in the detection of cytochrome c digested in a nanofluidic reactor.⁵⁹ Furthermore, the use of nanoscale open-tubular columns may contribute to studies of separation mechanisms.

Counting and evaluation of micro/nanoparticles, such as plasmonic metal particles, viruses, and exosomes, are important technologies in materials science, environmental monitoring, and chemical/biological analysis. The combination of micro/nanofluidics and photothermal detection methods is promising for measuring the absorption of such particles due to the high throughput and sensitivity at a single-particle level. In 1998, Mawatari et al. demonstrated the counting of 80-nm polystyrene particles and 10-nm Ag particles in a microchannel by a TLM, although not under flow conditions.⁶⁰ Later, Shimizu et al. realized detection of single gold nanoparticles with a diameter of 5 nm in a microchannel by POPS.⁸ By introducing a DIC, the signal-to-background ratio was improved by one order of magnitude compared with a TLM. Recently, Maceiczyk et al. demonstrated high-throughput detection of gold nanoparticles flowing in a microcapillary by differential detection photothermal interferometry.⁶¹ They realized the detection of 50-nm gold particles at a linear flow speed of 100 mm/s, corresponding to an analyzed volume of more than 1 nl s⁻¹ and counting rates of 400 particles per second.

One of the problems with nanoparticle detection by photothermal methods is low detection efficiency. Because photothermal detection requires tightly focused probe and excitation lasers with several micrometer-scale beam spots, only a few nanoparticles pass through the detection region in the channel, which is typically much larger than the beam spots. Furthermore, the signal intensity of each particle depends not only on the absorption cross section but also on the position and trajectory of particles passing through the detection region; signal-based evaluation of each nanoparticle is difficult. Several techniques have been proposed to mitigate this issue. For example, Yamaoka et al. developed a three-dimensional flow-focusing microfluidic device and measured focused particles by a TLM.⁶² They realized the measurement of 500-nm red polystyrene particles with a detection efficiency of 85 ± 6%. Another way to control the position and trajectory of nanoparticles is using narrow and shallow channels whose size is comparable to or smaller than a focused laser spot. Seta et al. used microchannels 1 μm wide and deep and measured 130-nm polystyrene nanoparticles with a TLM.⁶³ By confining the nanoparticles within a small microchannel, the detection efficiency improved almost 100%. Recently, Tsuyama et al. demonstrated gold nanoparticle detection in a nanochannel by POD. They also realized almost 100% detection efficiency with nanochannels 800 nm wide and 710 nm deep, and 20- and 40-nm gold nanoparticles were discriminated from their photothermal signals.³³ 

Recently, the combination of plasmonic nanoparticles and photothermal detection methods has been used for chemical/biological assays that measure the aggregation or changes in surface conditions of plasmonic nanoparticles by chemical/biological processing. For example, Heshi et al. developed fluorescence resonance energy transfer-thermal lens spectrometry (FRET-TLS) to identify the presence of methamphetamine in human blood plasma and urine.⁶⁴ In FRET-TLS, the energy absorbed by fluorescein transferred to gold nanoparticles causes enhancement of the thermal lens signal. Methamphetamine in the sample was quantified from a decrease in the thermal lens signal because the stronger adsorption of methamphetamine causes the detachment of fluorescein molecules from the surface of gold nanoparticles. FRET-TLS achieved an LOD of 1.5 nM. As another example, Shokoufi et al. demonstrated a DNA hybridization assay using gold nanoparticles and a TLM.⁶⁵ They used the difference in adsorption propensities of single- and double-stranded DNA on gold nanoparticles. Hybridization of target DNA leads to low binding affinity to citrate-protected gold nanoparticles, followed by the aggregation of gold nanoparticles. The aggregation of gold nanoparticles was quantified from the photothermal signal, and the smallest amount of target DNA was 2.6 zmol.

As shown above, particle counting and analysis by photothermal detection methods mainly target or use plasmonic nanoparticles due to their large absorption cross section for the visible light region and photostability. Introducing UV excitation will expand the application of the above methods to biological particles such as viruses and extracellular vesicles. Specifically, extracellular vesicles are known to have heterogeneous size and composition;^66,67 measuring their absorbance will provide information for understanding their characteristics and functions at a single-particle level.

Light absorption and the subsequent heat dissipation are common phenomena in nature. Continuous efforts have been made to measure light absorption of biological samples from tissue and cells to the level of single molecules and particles for its potential for label-free detection.^68–70 Although a TLM is nearly insensitive to scattering of the sample, the complex scattered field of heterogeneous biological samples of tissues and cells may complicate the quantitative interpretation of the photothermal signal and rapidly deteriorate phase-sensitive measurement. Several approaches have been developed to increase the sensitivity of the photothermal signal. One is to choose appropriate wavelengths of laser beams for the label-free detection of biological samples. A variety of endogenous cellular chromophores are good optical absorbers, such as cytochrome c in mitochondria,⁷¹ lipofuscins in lysosomes,⁷² hemoglobin in red blood cells,^73,74 and melanin in cancer cells.⁷⁵ Label-free dynamic photothermal imaging of mitochondria and lysosomes in living cells has been demonstrated using two excitation laser diodes [Fig. 5(a)].⁷² It has also been demonstrated to retain both spectral and spatial resolutions by employing mid-infrared excitation and visible probe beams, respectively.⁷⁶ Other more general approaches include combining a TLM with separation methods (e.g., microchip extraction^20,77–79 and liquid chromatography^58,80) and labeling techniques [e.g., the enzyme-linked immunosorbent assay, or ELISA,^21,81–86 and metallic nanoparticle-tagged systems, as in Fig. 5(b) ⁸⁷]. Labeling target molecules, although a complex process, increases the specificity, sensitivity, and applicability of a TLM, as the same enzymatic reaction or nanoparticle not only amplifies the photothermal signal but also can be used as the readout for the detection of various analytes by changing only the recognition molecule.

Micro/nanofluidic technology provides the capability necessary for efficient chemical processing. For instance, detection procedures can be automated in a microchip-based ELISA system.⁸⁸ In addition, excellent performance in sensitivity (10–1000 fg ml⁻¹), analysis time (10–20 min), and sample volume (1 μl) has been demonstrated using even clinical samples, which makes the microchip-based ELISA system, especially, advantageous for the practical rare cell analysis.⁸⁴ Furthermore, Nakao et al. demonstrated the detection of interleukin-6 cytokines secreted by a single B cell at a sensitivity of five molecules by performing an on-chip ELISA [Fig. 5(c)].¹³ A single-molecule ELISA has been realized by integrating nanofluidics and a DIC-TLM.¹² The photothermal signal may be greatly enhanced using gold or silver nanoparticles due to their large absorption cross section and photostability. Boyer et al. demonstrated the detection of 10-nm gold nanoparticles using a DIC-TLM, which is insensitive to the strongly scattering 300-nm latex beads.⁸⁹ It was found that for small gold particles (<40 nm), the photothermal signal is linearly proportional to absorbed energy. However, for large gold nanoparticles (40–100 nm), the strong scatter field of the particle interferes with the scattered field of the thermal lens, which can be considered as an enhancement of the photothermal signal, providing additional information that may be exploited for label-free detection in complex environments.⁹⁰ 

In addition to the development of ultrasensitive detection using a TLM, photothermal effects have also been exploited for sensitive POC biomarker quantitation in a cost-effective manner. Enzymes; in situ-generated agents; and a large variety of metallic nanoparticles, carbon nanomaterials, and transition metal oxide nanoparticles have been explored with photothermal sensing. Considerable progress has been made and described in several reviews.^91,92 Effort has also been devoted to developing simple signal readouts for POC testing. For instance, Zhou et al. reported a low-cost bar-chart microfluidic chip with an LOD of 2.1 ng ml⁻¹ for prostate-specific antigen in human serum samples.⁴⁰ The photothermal effect mediated by iron oxide (Fe₃O₄) nanoparticles in the ELISA assay was used to convert light into increased vapor pressure inside the device, driving the preloaded ink through microfluidic channels, which could be easily visualized as shown in Fig. 5(d).

Micro/nanofluidic devices coupled with photo-thermal/thermo-photo capabilities are now finding applications in the fields of environment and energy. Fu et al. developed a photothermal microfluidic paper-based analytical device (called μPAD), which consists of photothermally responsive hydrogel rods incorporating Prussian blue dye, a laser irradiation zone, and multiple microchannels on a cellulose chromatography paper substrate.³⁸ When the hydrogel rods are irradiated with a near-infrared laser, the photothermal phenomenon induces structural deformation in the gel, resulting in the release of liquid dye into microchannels. By measuring the thermal image and the dye flow distance into the microchannels, the detection of silver ions in environmental water was achieved at a concentration of 0.25 μM. Liu et al. developed a microfluidic analytical device combining the microfluidic flow-injection analysis (μFIA) with a TLM and demonstrated high-throughput and sensitive detection of Cr(VI) in sub-microliter solutions at an LOD of 0.6 ng ml⁻¹.⁹³ A key aspect of this μFIA-TLM device is the rapid and selective complexation of Cr(VI) with a diphenylcarbazone ligand in a microchannel on a chip and its matching with the wavelength of the excitation laser. An analytical approach using bundles of threads was developed for the sequential determination of Cu(II) and Zn(II) in environmental and biological samples by Yan et al.⁹⁴ In this system, sample and colorimetric solutions were moved along thread-based microfluidic channels by capillary and gravity actions. The colorimetric changes due to the complexation of Cu(II) and Zn(II) with a chelating ligand were detected by a near-field laser TLM, and an LOD of about 0.20 μg ml⁻¹ was obtained for both ions.

Such micro/nanofluidics have proved to be suitable even in the nuclear field. Highly efficient separation (extraction and stripping) of radionuclides, that is, lanthanides and actinides (including uranium, plutonium, and americium), from simulated nuclear waste solutions has been demonstrated by controlled microfluidics. Most of these devices are configured to inject the sample solution recovered from microchannels into the detection equipment, such as inductively coupled plasma mass spectrometry, gamma-ray detectors, and liquid scintillation counters.^95–99 To overcome the drawbacks of not being able to measure on-line, light/laser-assisted detection techniques on a chip have been devised. For example, Aileen et al. realized real-time determination and extraction kinetics analysis of trace Se(IV) by a combination of microextraction due to fluorogenic-chemosensor material and fluorescence microscopy.^100, In situ laser-induced fluorescent detections of not only trace U(VI) in capillary electrophoresis but also Eu(III) in micro-plug liquid-liquid extraction were demonstrated by Haraga et al. and Angeli et al., respectively.^101,102 TLM detection allows ultrasensitive and online detection of radionuclides in microchannels. In 2002, micro-Co(II) wet analysis composed of multiphase microextraction and TLM detection was successfully demonstrated.³ This technique has made a great contribution in ensuring the safety of radiation workers because the concentration monitoring of radioactive cobalt ions in primary coolant in nuclear reactor plants is essential. We realized real-time and in situ determination of U(VI) in both aqueous nitric acid and organic phases, including tributyl phosphate, during the liquid–liquid microextraction process by a TLM. The TLM system achieved an LOD of approximately 10⁻⁴M for U(VI), and the kinetic analysis was performed from U(VI) concentration changes at each detection point in a microchannel [Fig. 6(a)].¹⁰³ Also, rapid electrochemical valence control of U(IV)-U(VI) has been achieved in a microchannel equipped with microelectrodes and online monitoring of concentration changes of U(IV) and U(VI) in a microchannel by a TLM [Fig. 6(b)].¹⁰⁴ Moreover, it was found that the nanofluidic device made it possible to mutually separate U(IV) and U(VI) ions in acid solutions.¹⁰⁵ 

Miniaturized energy conversion devices are expected to offer superior features, such as faster mass transfer, higher energy density, ease of integration, and portability compared with conventional large-scale devices.^106,107 Various laminar flow-based micro-fuel cells (μFCs) have been developed.^108,109 They are typically constructed with carbon-metal composite electrodes and microflows such as fuel stream (formic acid, methanol), an aqueous stream containing an oxidant (O₂, H₂O₂), or electrolyte stream (sulfuric acid, potassium hydroxide). Because the interface formed by laminar flows in microchannels acts as a virtual membrane to prevent hydrodynamic mixing, effective electrochemical reactions and mass transfer can be generated. As a result, the μFCs have been capable of generating power densities in the range of 10–100 mW cm⁻². Different types of μFCs, which used a Nafion membrane as an electrolyte, nanoporous electrodes,¹¹⁰ and enzymes or microorganisms as biocatalysts,¹¹¹ have been investigated and their cell performances have been evaluated.

Optofluidic approaches composed of microfluidics and optics have become of greater interest because they easily convert naturally abundant solar energy into electrical energy due to their light energy harvesting abilities and mass/photon transfer efficiencies within the microchannels. Li et al. and Zhang et al. fabricated micro-photocatalytic fuel cells with embedded TiO₂ nanoparticles and a TiO₂ nanorod array immobilized on an fluorine-doped tin oxide glass as photocatalysts have succeeded in generating electricity.^112,113 Bhattacharjee et al. developed a novel microfluidic-based energy harvesting device.¹¹⁴ By flowing aqueous NaCl microdroplets loaded with gold nanoparticles into a microchannel integrated with an array of Schottky-junction electrodes under solar irradiation, a conversion efficiency of a few percent and power density at a level of 100 μW cm⁻² were achieved. Such unique properties result from the combined effects of the streaming potential due to the microflow and surface plasmon resonance due to the nanoparticles. Recently, we proposed an autonomous solar-light-driven μFC on a glass microchip that integrates WO₃/BiVO₄ nanorods as a photocatalyst and nanofluidic channels as ultra-fast proton conductors. This microdevice achieves a power density of ∼10 μW cm⁻² as a result of the internal photocatalytic hydrogen fuel generation by solar irradiation.¹¹⁵ Also of note, micro-thermophotovoltaic systems have been realized by Chan et al.¹¹⁶ In this microsystem, a catalytic micro-combustion reactor with metallic photonic crystals as emitters leads to efficient heat-to-thermal radiation conversion, and the radiated photons are converted into electricity via photovoltaic diodes made from GaInAsSb materials. The electric power generation reaches about 300 mW cm⁻² at around 800 °C.

Unique physicochemical properties differing from bulk water and surface-adsorbing water have been found in 10- to 100-nm-scale nanofluidic channels because the nanofluidic channel is a transitional space between a single molecule and a bulk condensed phase and comparable to the thickness of an electric double layer. Several researchers, including ourselves, have investigated the physicochemical properties of liquids in nanofluidic spaces. We pioneered flow visualization and time-resolved fluorescence measurements of water confined in 330- and 850-nm-sized nanochannels and found a reduction in water velocity due to capillary action and changes in the decay time of aqueous solutions containing rhodamine dyes compared with the case of a 250-μm-sized microchannel, respectively. The results showed that nano-confined water had high viscosity and a low dielectric constant in comparison to bulk water. Since then, other measurement techniques for nanofluidics have been developed, and various liquid properties in nanospaces, such as enhancement of liquid viscosity and ion conductivity,^117–119 changes of diffusion coefficients,^120,121 and reduction in pH,^122,123 have been clarified. The streaming current method enables us to nonintrusively measure zeta potential, conductivity, pH, and the dielectric constant of confined liquids without the addition of a probe ion [Fig. 7(a)].^124–126 

Unique microscopic structural and dynamical behavior of liquids in nanofluidic channels has also been characterized by means of nuclear magnetic resonance (NMR) spectroscopy.^127,128 The size dependence of ¹H-NMR chemical shifts and spin-lattice relaxation rates (¹H-1/T₁) of water confined in 4- to 5 000-nm-sized channels were examined, and a spectrum broadening without chemical shift changes and a drastic increase in ¹H-1/T₁ values in the vicinity of 800 and 200 nm were found. In addition, comparison to heavy water (²H₂O) results confirmed that translational motion of water molecules, rather than rotational motion, was modulated in nano-confining geometries [Fig. 7(b)]. These results suggest that an intermediate phase, which has a highly ordered structure consisting of loosely coupled water molecules with an efficient proton transfer process, exists between the adsorbed and bulk phases in nanofluidic spaces. Such a well-ordered water structure was supported by the results of nano-x-ray diffractometry and Raman spectroscopy.¹²⁹ A fluorescence microscopic observation of pH-controlled solutions containing a fluorescent probe molecule showed greater protonic diffusion in nanofluidic channels because fluorescence quenching progressed with the elapse of time, as shown in Fig. 7(c). From quenching profiles, an approximately fourfold higher proton diffusion coefficient in 180-nm channels compared with bulk could be determined according to the Einstein relation L = (2Dt)^1/2.⁷ Thus, specific nanoscience applications play an important role in producing advanced and innovative nanofluidic applications. Photothermal spectroscopy may also contribute to the investigation of thermophysical properties such as thermal diffusivity.¹³⁰ 

For easy operation and miniaturization of analytical systems, integrated photothermal detectors are required in many application areas. A thermal lens detection device that uses optical fibers and microlenses was developed and used for a microfluidic ELISA system.¹³¹ Other examples have used microfabricated mirrors and lenses¹³² and Young interferometers.¹³³ 

Single-cell analyses have been the most frequent targets of applied micro/nanofluidics.^134–137 A small fluid volume confined in a nanochannel greatly reduces the dilution of the secretome or molecular content of a small number of cells or even a single cell. Recently, a mechanical nanovalve has been reported, which makes nanofluidic operation much easier and more robust.¹³⁸ The integration of a TLM and micro/nanofluidics has shown great promise for single-cell multi-omics.

The authors acknowledge support from the Taiwan Ministry of Science and Technology under Grant No. MOST 110-2639-E-007-002-ASP.

The authors have no conflicts to disclose.

Hisashi Shimizu: Conceptualization (equal); Writing – original draft (equal); Writing – review and editing (equal). Chihchen Chen: Conceptualization (equal); Writing – original draft (equal); Writing – review and editing (supporting). Yoshiyuki Tsuyama: Conceptualization (equal); Writing – original draft (equal). Takehiko Tsukahara: Conceptualization (equal); Writing – original draft (equal). Takehiko Kitamori: Conceptualization (equal); Supervision (equal); Writing – original draft (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.