Microsphere-assisted digital holographic microscopy integrated with on-line thermal lens spectrometry for 3D imaging and enhanced photothermal detection

Acquiring morphological and photothermal information is essential for characterization of a variety of samples in biochemistry, material science and process technologies. In this paper, we report on the integration of digital holographic microscopy (DHM) and thermal lens (TL) spectrometry techniques enabling simultaneous measurement of both class of properties: morphological images and photothermal parameters. DHM in transmission mode is an effective technique for label-free, non-contact, real-time, and non-invasive imaging and measurement of phase objects, such as biomaterials. On the other hand, TL is a highly sensitive methodology to detect low concentrations of analytes in low absorbing samples. We demonstrate the usefulness of the DHM-TL integrated system by applying it to obtain morphometrical information of samples as well as for acquisition of their photothermal maps. Further, by extending the concept of “microsphere-assisted (MS) imaging” into TL, we show that MS-TL enables detection and counting of particles, which otherwise their photothermal signals cannot be detected. The system has the potential to be applied for a variety of objects including bio-samples and may be proposed as a bench-top characterization device.


Introduction
In numerous applications in science and technology, it is proved that optical microscopies and nanoscopies are essential for observing, detecting and measurement of interesting phenomena and processes 1,2 . On the other hand, spectroscopy techniques through the light-matter interaction process, provides an exploratory analytical tool in the fields of physics, chemistry, biology and corresponding interdisciplinary branches of science and technology. They use a range of principles to reveal the physicochemical characteristics of materials, such as chemical composition, elemental concentrations, crystal structure and photoelectric properties 3 .
It is obvious that in several cases obtaining comprehensive information about the sample requires the utilization of different classes of methodologies at the same time. More importantly, simultaneous utilization of at least two complementary methods often is a crucial requirement, when the two combining methods are indispensable by their distinguishing characteristics in the application. Due to the nature of the optical characterization techniques, their combination seems straightforward. This has led to development of a number of integrated optical characterization systems, which in turn, enhances the applicability and usefulness of the individual techniques that are combined. Optical microscopies can be combined with spectroscopy techniques to deliver surface characteristics along with chemical composition and electronical information of materials simultaneously 4 . In recent years combination of various spectroscopy and microscopy techniques has been developed and reported 5 . For example, integrating an optical microscope with UV-visible spectroscopy in transmission and reflectance modes enables spectroscopy of microscopic samples, which is useful in forensic labs and in materials science and biological researches 6 , or the tip-enhanced Raman spectroscopy which is the atomic force microscopy coupled with Raman spectrometer enables surface information, such as topography, hardness, adhesion, and conductivity, along with molecular chemistry.
In this paper, we report on the integration of digital holographic microscopy (DHM) and thermal lens (TL) spectroscopy in a single device. Furthermore, by extending the concept of microsphere (MS)-assisted imaging into TL arrangement and its inclusion in the DHM-TL integrated setup we suggest a highly sensitive TL platform to be used for particle counting and depth profiling based on the TL effect, while being monitored in 3D by DHM. In the following we discuss the two combining methodologies and refer to the predated related combined systems.
Since conventional optical imaging techniques, such as bright field microscopy, works on the detection of the intensity and color of the light coming from the sample, the phase information, i.e. 3D information will be lost. To overcome this shortcoming, DHM is an elegant candidate. DHM possesses several advantages; it is non-contact, non-invasive, label-free, easy-to-implement, and inexpensive method, and having single exposure nature it enables "3D live monitoring" or equivalently "quantitative phase imaging" of time-varying phenomena 7 . Transmission DHM has been widely applied for 3D imaging of biological organelles, cells and membranes 8 . DHM in reflective mode is a suitable method for non-destructive and non-contact surface profile measurement 9 . The integration of DHM with other technologies, which also use lasers, is generally possible and straightforward, however, it needs to use proper optical elements in a smart configuration to achieve an effective integrated system. For example, the integrated arrangement of conventional optical microscopy, fluorescence microscopy and quantitative phase contrast microscopy has been successfully demonstrated 10 . These imaging techniques are complementary to each other; while the fluorescence image gives the high-contrast image of the nucleus in the cell, for instance, the phase contrast image provides the morphological structure of cell membrane in a transparent medium. Here, the other combining method is based on the TL effect and provides complementary to DHM results. TL effect is caused by the deposited heat through a non-radiative decay processes when the energy of a laser beam, so-called excitation beam, passing through the sample is absorbed 11 . The resulting local temperature changes and, hence, the refractive index gradient leads to production of a divergent lens. The TL effect depends on the absorption coefficient, concentration, power of the excitation beam, and thermo-optical properties of the medium. The effect is detected by tracking the induced changes in the wavefront of a second laser beam, so-called probe beam, which is passing through the sample under study. Therefore, TL signal may be considered as an effective structural representative of the sample. Due to the high sensitivity of TL measurement method, it is considered as a suitable optical characterization for a variety of samples, especially for detection of low concentrations of analytes in low absorbing samples as well as a photothermal detector integrated with gel electrophoresis [12][13][14] . It is also remarkable that, since DHM can extract the optical phase difference function of the sample, it may be considered as an alternative to measure the absorption coefficient using the TL effect 15,16 .
A very simple yet highly advantageous method for super-resolution microscopy has been utilizing a transparent MS in the working distance of a microscope objective toward synthetically increasing the effective NA of the imaging system 17,18 . Considering the versatility and capability of MS-assisted imaging it has been integrated in several imaging modalities to surpass the optical resolution limit. For example, it has been used in combination of dark field microscopy for the inspection of quasi-transparent sub-diffraction-limited structures 19 , polarimetric microscopy to reveal some important features in Mueller matrix images, which are not visible in a conventional Mueller matrix microscopy system 20 , and confocal microscopy to direct optical imaging with a resolution down to 25 nm by laser scanning 21 . It has been also used for Raman spectroscopy for obtaining enhancement in both resolution and signal strength of Raman mapping 22 . Recently, we showed that the MS-assisted imaging can be integrated with digital holographic microscopy (DHM) as an effective real-time and label-free method for quantitative phase imaging and 3D microscopy 8,23 . Therefore, in MS-DHM, 3D images with enhanced lateral resolution can be easily achieved through experimental recording and numerical reconstructing of digital holograms. It is also remarkable that MS-assisted microscopy can be used along with other super-resolution schemes such as structured illumination and oblique illumination for even further enhancement of the imaging resolution 24,25 . It seems straightforward to extend the MS idea into the presented DHM-TL setup. The TL signal generated in a liquid solution or PDMS sample is detected while the utilized MS is embedded inside the PDMS. MS-TL enables to acquire enhanced TL signal from a semitransparent sample using very low (15 mW) excitation power, in comparison with Ref 15 , in which a high power is used and longer path length sample is considered. Further the MS-DHM-TL arrangement is extended into biological specimens, for which as an example, red blood cells (RBCs) are used as focusing media in order to enhance the TL signal and detect the cells positions and depths 26 . It is remarkable that since the TL probe beam is used also as the DHM laser source, the combined DHM-TL system provides simultaneous 3D imaging and photothermal detection.

Results and Discussion
To demonstrate the capability of the combined DHM-TL system we designed and performed two experiments. In the first sets of experiments we consider the flow of a laminar fluid containing micro-objects inside a fluidic channel. The flow rate is controlled via adjusting the relative hight of the reservoir and the fluidic chamber. First, we use polystyrene particles of 40 µm diameter. The particles can move in different depths, i.e. different distances to the camera. In conventional microscopy, in order to acquire sharp images of the particles at different depths, one needs to adjust the objective-sample distance for each particle. This, in turn, makes it impossible to image and track the flowing of all the particles together. The incorporation of the numerical focusing feature of DHM, however, results in obtaining sharp images of all the particles at once.  particle 2 can be brought into a plane that provides a sharp intensity image. The distances to focus on particle 1 (d 1 ) or particle 2 (d 2 ) is precisely determined in the numerical process. The procedure can be applied during the experiment, regardless of that how many particles are present in the field of view. One the other hand, crossing the particles in different depths changes the thermal lensing signal differently. The different behavior of particles 1 and 2 in thermal lensing measurement is shown in Figs. 1(h-i); the maximum value of the visible intensity peaks corresponds to the particle's depth. Figure 1(g) is the reference thermal lensing signal where acquired when the excitation beam is switched off.
Furthermore, in order to demonstrate the capability of the present technique for characterization of biological objects we conduct similar experiment on flowing RBCs. The results are summarized in Fig. 2. Figure 2(a) shows the hologram of a couple of RBCs flowing in various depths. DHM is proved to be an effective 3D quantitative imaging and monitoring of RBCs dynamics 23,[27][28][29][30] , and for the flowing RBCs it is straightforward to reconstruct the 3D image at any desired time of the experiment. In Fig. 2(b) we show the reconstructed 3D image of the RBC 1 and its 2D map in the inset. Such images include detailed variations of the RBC thickness throughout its surface. In Fig. 2(c), as an example, a cross-sectional thickness profile along the AB line depicted in the inset of Fig. 2(c). The associated TL signals of the two crossing RBCs, taken by the TL part of the integrated setup, are shown in Fig. 2(d). It is remarkable that TL signal detection can be independently used for cell measurements [31][32][33] . For example, S. Vasudevan et. al. have used TL response for quantitatively monitoring the death process of individual RBCs 34 . Indeed, the enhanced TL detection is due to the focusing capability of RBCs 26,35 . Nevertheless, along with the TL information, the possibility for acquisition of detailed, high-resolution, quantitative 3D image highly enhances the cell measurement process. This is the main feature of the present DHM-TL system.
In the second experiment a single polystyrene particle which is embedded in a 300 µm layer of Polydimethylsiloxane (PDMS) is considered. In order to fabricate the microsphere-embedded PDMS films, the monodispersed polystyrene microspheres (MSs, diameter of 40 µm) are distributed in liquid PDMS elastomer and poured on a microscope coverslip as the substrates. The liquid is extended over the substrate by spin-coating. The compound is baked at 65 • C for one hour to dry the mixed compound and form a removable MS-embedded PDMS film. The results of the experiment are demonstrated in Fig. 3. Initially a hologram of the sample is recorded. Then by the use of a linear precision translation stage (M-126.DG1, Physik Instrumente), which is controlled by a computer, the sample is scanned and in each step the corresponding TL signal is registered to record a TL map of the sample. Figure 3(a) shows the recorded hologram on which the trajectory of the scanning procedure is also overlaid. The digital hologram is processed by the ASP method and its 3D reconstructed image is shown in Fig.  3(b). In Fig. 3(c) we show the TL signals of seven typical lines of the scanned sample. The resolution of the TL map, i.e., the photothermal signal, of the particle depends on the x-y scanning resolution. The symmetric structure of the signals is due to the reciprocating motion of the scanning device, i.e., as schematically shown in panel (a), the probe beam passes two times through the sample. The scanning nature imposes a limitation on the samples, and the static samples are suited more for DHM-TL experiments. However, by the use of high-speed scanners it will be also possible to acquire, simultaneously, 3D morphometric image from DHM and full thermo-optical microscopy map of a dynamic sample. The resolution of the scanning device and the TL probe beam size govern the final resolution of the TL map. With the current ultra-precise positioning technologies it seems the conclusive limitation should be attributed to the TL probe beam spot size. Nevertheless, for several samples commonly used TL arrangements provide sufficient performance. In all of the experiments, a control experiment is conducted initially in which, the excitation laser is switched off to attribute the observation solely to the TL effect. Additionally, the use of an embedded MS in PDMS, indeed, extends the MS-assisted super-resolved microscopy 36, 37 into TL imaging for an improved sensitivity. Figure  3(d) shows an out of focus MS in air, which does not generate thermal signal. However, if the MS crosses the probe beam in the PDMS layer the signal enhances, which is also confirmed by the symmetry of the signals in Fig. 3(c). Indeed, the TL signal is produced because of the absorption of the excitation light in the PDMS layer, but previously focused by the MS. The absorption peaks due to the particles in the air are negligible in comparison to the peaks in PDMS as the adsorption coefficient of air is much lower and the enhancement factor is also smaller. It is noticeable that there is not TL signal where the particles are absent (Fig. 3(c)) in the liquid or PDMS, thus demonstrating the enhancement effect produced by the MS. The fact that the TL signals are zero both for the particles in the air as well for PDMS without particles confirms the enhancement effect of MS on the TL signal of the absorbing medium. The results suggest a high sensitivity particle counting and depth profiling approach based on the TL effect with the present arrangement, which is also demonstrated in the first experiment ( Figs. 1 and 2). Therefore, the methodology can be used for sorting colloids based on their photothermal properties. For example, the TL part of the setup can distinguish dyed particles in a mixture from normal ones as well as their deepness, while DHM part images the whole sample. It should be noted that, similar to different MS-assisted microscopies 20,23,38,39 or MS-assisted Raman spectroscopy 40 , the enhancement depends on the MS size and refractive index, and adjustment of the objective-MS and MS-sample relative distances. The goal of this work is to extend the idea of MS-assisted signal enhancement in thermal lens spectrometry. For this purpose we image the depth profile of micro-particles and 3D image of RBCs. However, as a future application it is possible also to measure the absorption of the fluid or thin transparent sample because of the enhanced TL signal. Note that this has been possible using very low excitation power (15 mW).

Conclusion
In conclusion, we integrated DHM and TL as complementary techniques in material characterization. The integrated system which acquires simultaneous DHM and TL measurements can be used for a variety of samples in biochemistry, material science and process technologies. We presented the proof-of-concept experimental results on fluidic, solid and biological samples. The results demonstrated the imaging of micro-particles using DHM and simultaneously depth profiling of multiple particles through TL spectrometry. In our arrangement the induced TL effect due to MS utilization is strong, which suggests the MS-TL platform as a high sensitivity particle counting and depth profiling approach based on the TL effect. The DHM-TL system has the potential to be developed as a bench-top characterization device for measuring photothermal properties of thin transparent samples.

Experimental procedure.
By taking the DHM laser beam also as the probe beam in TL, it seams the integration of DHM with TL is possible and straightforward. The integration may be performed in a number of configurations. Figure 4 shows schematically the integration of the TL setup with the most common DHM configuration for transparent samples, i.e., Mach-Zehnder arrangement. The integrated system may be considered in three parts: (I) DHM, (II) Excitation, and (III) Probing. The outcome of a DHM setup includes the interferometric pattern of "reference" and "object" beams, both originated from the same coherent source and the latter passing through the sample under study. In DHM part of the integrated setup (part I), a He-Ne (λ =632.8 nm, 3 mW, 05-UR-111, Melles Griot) is steered by mirrors M 1 and M 2 toward the sample (S) which is positioned on a x-y-z stage. Lenses L 1 (LB1027-A, f=40 mm, Thorlabs) and L 2 (LB1676-A, f=100 mm, Thorlabs) are arranged in a telescope scheme to collimated and adjust the beam diameter to about 3 mm. The beam is split into two beams by the 50%-50% beam-splitter BS 1 (BS016, Thorlabs). The transmitted part, i.e. the object beam, is sent onto the sample (S) by mirror M 3 (DMSP605, Thorlabs). The laser light passing through the sample carries the phase information of the sample and is collected by the microscope objective MO 1 (LMH-10X-532, Thorlabs), and through a dichroic mirror (DM, DMSP605, Thorlabs) is sent onto a recombining beam-splitter (BS 2 , BS016, Thorlabs) to finally reach the DHM CMOS camera (acA 2440-20gc, Basler). L3 (LB1676-A, f=100 mm, Thorlabs) is used to complete the imaging train, M 4 redirects the beam, and F 1 is an interferometric filter (center wavelength 632.8 nm, Melles Griot) to allow only the He-Ne laser beam to hit the camera and block the other lights including the back scattering of the excitation laser beam. The reflected part, i.e., reference beam, is sent directly onto the DHM camera through BS 2 . A proper holographic fringe pattern has to be linear, uniform, high contrast, and should have proper fringe density. To achieve fringe patterns with the aforementioned properties, MO 2 , identical to MO 1 , is used to match the curvature of the reference beam with the object beam, a neutral density filter (NDF 1 ) is used to match the average intensity of the interfering beams, and the angle and position of BS 2 is carefully adjusted.
In the excitation part of the setup (part II), a 532 nm diode-pumped solid state laser (DPSS) (MGL-III-532 nm-100, UltraLasers), shown as Ex laser, is used to excite the sample. Its beam is modulated at 136 Hz using a function generator (FG, Rigol DG 2041A, Batronix), and using a rotating neutral density filter (NDF 2 , NDC 50S-3, Thorlabs) the excitation power is calibrated and is tuned during the experiment to 15 mW. Lenses L 4 (LB1027-A, f=40 mm, Thorlabs) and L 5 (LB1676-A, f=100 mm, Thorlabs) are used to collimate the excitation beam, mirrors M 5 and M 6 are used to redirect the laser beam, and the iris I truncates the beam to match its diameter with the back-aperture of MO 1 . MO 1 focuses the laser beam to the sample plane, where it is imaged through DHM part, including the same objective, on the camera. The focused light excites the sample and induces TL. In the present setup, as mentioned earlier, we use the same He-Ne laser of DHM to probe the TL effect. The probe laser (Pb), therefore, travels the same path of the DHM to reach the sample and it is carefully aligned counter-propagating with the Ex laser 11 . This particular configuration enables the use of the second laser as a probe beam for TL signal detection as well as DHM laser beam for acquisition of digital holograms. In fact, the probe beam diameter is larger than the excitation beam, testing slightly larger temperature profile, thus, higher sensitivity can be achieved 11 . However, after the second beam-splitter (BS 2 ) of the DHM part it is collimated by L 6 (LB1676-A, f=100 mm, Thorlabs) to reach the detector. The probe beam intensity changes resulting from the TL effect is detected by a Si photodiode detector (PDA 36A-EC, Thorlabs) equipped with a 0.5 mm diameter pinhole (Ph 1 ) and the interferometric filter F 2 to allow only the Pb beam to reach the detector. The TL signal is amplified by a lock-in amplifier (SR5 10, Stanford Research System) and processed by the data acquisition system. The data acquisition system mainly consists of a microcontroller-based digitization board and a graphical user interface (GUI) running on a PC. Arduino Mega 2560 board is used to digitize the analog input signal at 1k samples/s sampling rate and pre-process the data. The GUI is built using a LabVIEW graphical programming software. The main tasks of the GUI are to visualize the incoming data, capture the peaks and save the data to the local hard drive.
Even if the integration of DHM and TL setup is demonstrated on a Mach-Zehnder arrangement basis, however, it can be performed on the other arrangements. Recently, in order to minimize the influence of environmental vibrations and 6/10 noises on the digital hologram acquisition, smart DHM systems, in transmission and reflection modes, based on the use of interferometric microscope objectives, such as Mirau objective, and self-referencing arrangements have been developed 23,41,42 .
In self-referencing DHM configurations a part of the object beam is folded on the rest of the object beam or sheared and overlapped with itself, acting as the reference beam. Therefore, possible noises on object and reference beams are correlated, which leads to a robust and vibration-immune configuration. The TL measurement system has the capability to be integrated with a self-referencing DHM as well.

DHM reconstruction process.
The procedure of DHM consists of two steps: recording and reconstruction. The recorded digital holograms are subjected to numerical reconstruction to extract the full field distribution of the holographic image of the sample and further 3D and morphometric information therefrom. The registered TL signals peaks which are recorded simultaneously contain information about the spatial distribution of the samples as well as their deepness inside the liquid or PDMS. Based on the fact that the photodiode signal and the digital holograms are recorded simultaneously, these post-processing procedures enable us to collect morphometric and photothermal information of the sample at the same time. The reconstruction process of DHM is performed by simulating the diffraction from the digital hologram when illuminated by the reference wave and a numerical propagation into the desired image plane. It leads to the complex amplitude, from which the amplitude and the phase information can be extracted. The numerical reconstruction of digital holograms can be done by Fresnel transform, Huygens convolution, or angular spectrum propagation approach (ASP) 43 . In this research, we use the ASP approach.
In the reconstruction process, once the recorded digital hologram is illuminated by the reference beam, upon diffraction, it forms a virtual image, a real image, and zero-order terms. Dealing with Fourier space in ASP enables applying proper filtering to separate out the unwanted portions or to smooth the reconstructed information. In 28,44 we have explain the details of the reconstruction approach and its possibility to provide 3D imaging and tracking information of multiple micro-objects from the recorded holograms. Once the complex amplitude at the image plane, U(x, y; d) is extracted, the intensity of the image (I), as well as its phase (φ ), can be computed:  Figure 5. Flowchart of the numerical reconstruction process of the recorded digital holograms. FFT denotes fast Fourier transformation.
The intensity image (Eq. 1) is similar to a conventional microscope image, and 3D information of the object is obtained from the phase (Eq. 2). The phase obtained from Eq. 2 takes values between − π 2 and π 2 , which creates a discontinuity throughout the phase map. We convert the discontinued phases to continuous phase maps by the Goldstein's branch-cut unwrapping algorithm 45 . The phase is proportional to the optical path length, φ (x, y) = 2π λ n(x, y)L(x, y), where n is the refractive index and L is the physical length that the light beam passes through.
An important advantage of DHM over conventional microscopy is numerical refocusing. It means that in the reconstruction process, by varying the parameter d, it is possible to change the image formation plane. This is the key feature for 3D tracking of multiple objects based on DHM. By playing with d and formation of the focused images of any of the objects, indeed, their relative distances will be obtained. Acquisition and reconstruction of successive digital holograms lead to obtaining the 3D trajectories of multiple arbitrary shape objects. This feature of DHM, along with the possibility to acquire phase images, makes it a unique methodology to apply, for example, in several phenomena in microfluidics, as it provides complete space and time information of the dynamic process 46 . To remove the effects of the background contaminations and the aberrations of optical elements from the final data, before the onset of the experiment, a reference hologram is recorded. To this end, the chamber filled up with distilled water is used. The reconstructed phase of the reference hologram is then subtracted from the other recorded holograms in the reconstruction process. Figure 5 shows the the flowchart of the numerical reconstruction process of the recorded digital holograms.