A Rapidly Modulated Multifocal Detection Scheme for Parallel

Jun 3, 2014 - Fast Confocal Raman Imaging Using a 2-D Multifocal Array for Parallel Hyperspectral Detection. Lingbo Kong , Maria Navas-Moreno , and ...
0 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/ac

A Rapidly Modulated Multifocal Detection Scheme for Parallel Acquisition of Raman Spectra from a 2‑D Focal Array Lingbo Kong† and James Chan*,†,‡ †

Center for Biophotonics, University of California, Davis, Sacramento, California 95817, United States Department of Pathology and Laboratory Medicine, University of California, Davis, Sacramento, California 95817, United States

‡

S Supporting Information *

ABSTRACT: We report the development of a rapidly modulated multifocal detection scheme that enables full Raman spectra (∼500−2000 cm−1) from a 2-D focal array to be acquired simultaneously. A spatial light modulator splits a laser beam to generate an m × n multifocal array. Raman signals generated within each focus are projected simultaneously into a spectrometer and imaged onto a TE-cooled CCD camera. A shuttering system using different masks is constructed to collect the superimposed Raman spectra of different multifocal patterns. The individual Raman spectrum from each focus is then retrieved from the superimposed spectra with no crosstalk using a postacquisition data processing algorithm. This system is expected to significantly improve the speed of current Raman-based instruments such as laser tweezers Raman spectroscopy and hyperspectral Raman imaging.

■

INTRODUCTION Laser tweezers Raman spectroscopy (LTRS)1 is a method of using a single laser beam for both optical trapping and performing Raman spectroscopy. The laser tweezers technique allows microparticles suspended in an aqueous medium to be analyzed independently and continuously by immobilizing a single particle within the laser focus. The Raman spectrum acquired from the trapped single particle provides direct chemical information without requiring exogenous labels or extensive sample preparation and perturbation. LTRS has proven to be a powerful tool for the molecular analysis of single cells 2−4 and has been applied for cancer detection, 5 heterogeneity analysis,6 and cell dynamics monitoring.7,8 However, because of the intrinsically weak Raman scattering signal, Raman-based instruments have the disadvantage of requiring long signal acquisition times (from seconds to minutes). For LTRS, this limits the number of live cells that can be analyzed in a reasonable amount of time and makes it challenging and impractical to measure long-term dynamics of a large number of cells simultaneously and under the same exact environmental conditions. For hyperspectral Raman microscopy, live cell imaging at high resolutions can be impractical or impossible. Recently, in order to increase the analytical throughput of single-trap LTRS and extend it to more practical applications that require a large number of individual microparticles to be measured, multifocal LTRS9,10 was demonstrated. A time-sharing trapping scheme was created in which a single laser beam was rapidly scanned by galvanometer mirrors to generate multiple foci that each trap an individual particle. Parallel detection of the Raman signals of these trapped particles was achieved by projecting all of the signals © 2014 American Chemical Society

simultaneously into a spectrometer and onto a charge-coupled device (CCD), with each spectrum positioned onto different pixel rows of the camera to avoid overlap and cross-talk between nearby spectral channels. The time multiplexing technique based multifocal LTRS has been demonstrated to achieve parallel analysis of ∼10 individual particles. However, two main limitations of this design prohibit further improvement of the analytical throughput of multifocal LTRS. First, the time-sharing technique that uses only one laser focus makes it difficult to trap a larger number of microparticles (>50). Second, the vertical dimension of the spectrometer’s CCD detector ultimately limits the number of spectra that can be detected simultaneously. The first limitation can be avoided by using a multiplexing technique, such as a spatial light modulator (SLM) or a microlens array, to generate multiple optical traps.11,12 For the second limitation, a new technique is needed to further increase the parallel acquisition of spectral signals beyond the 1-D (i.e., spectra separated along the vertical axis of the CCD detector) detection schemes of current multifocal LTRS systems.9,10 It should be noted that parallel Raman microspectroscopy from multiple points using a SLM and widefield Raman imaging has recently been reported to improve the speed of Raman imaging.13,14 In this article, we report on the development of a modulated multifocal detection scheme that enables the parallel acquisition of full Raman spectra (∼500−2000 cm−1) from a 2-D m × n array of optically trapped particles. This design overcomes the Received: April 3, 2014 Accepted: June 3, 2014 Published: June 3, 2014 6604

dx.doi.org/10.1021/ac5012188 | Anal. Chem. 2014, 86, 6604−6609

Analytical Chemistry

Article

Figure 2A shows a brightfield image of a 2 × 3 array of polystyrene beads trapped by the laser tweezers array. The

two limitations discussed above by allowing for parallel detection of spectra along both the vertical and horizontal dimensions of the CCD chip and providing more stable particle trapping. A spatial light modulator (SLM) was used to generate holographic laser tweezers in a 2-D array. Raman signals from trapped beads were detected simultaneously by a spectrometer equipped with a wide area CCD camera. A shutter system consisting of different mask patterns was placed in the detection path to modulate the 2-D array pattern of Raman signals that was allowed to pass into the spectrometer. A data processing algorithm was developed to reconstruct the individual Raman spectra of each laser focus based on the different superimposed Raman spectra that were collected from the different patterns. By allowing the individual Raman spectra from a 2-D multifocal array to be acquired in parallel, this novel system is expected to significantly improve the analytical throughput of LTRS as well as increase the imaging speed of hyperspectral Raman microscopy.

■

EXPERIMENTAL SECTION Samples. Two types of polymer beads were used in this study, polystyrene and poly methyl methacrylate (PMMA). Ten microliters of polystyrene beads (∼106 beads per mL, mean diameter: 3.005 μm, Duke Scientific Corporation) and 10 μL of PMMA beads (∼106 beads per mL, mean diameter: 1.14 μm, Bangs Laboratories, Inc.) were added to 500 μL of distilled water. The solution was placed into a cell chamber (Attofluor, Molecular Probes) that holds a fused silica coverslip with no. 1 thickness. Instrumentation. The schematic of the experimental setup is shown in Figure 1. A laser beam at 785 nm from a diode laser

Figure 2. (A) Bright field image of laser trapped 3 μm polystyrene beads in a 2 × 3 array. (B) Raman spectral image of laser trapped 2 × 3 polystyrene beads recorded by the CCD camera of the spectrometer. (C) Raman spectra of the trapped 2 × 3 polystyrene beads generated by binning five vertical pixels (y direction) for each row in B.

backward propagating Raman scattering light from the trapped particles was collected by the same objective. Instead of a conventional single slit, a custom designed multislit array (HTA Photomask) was placed at the entrance of the spectrometer (Princeton Instruments, LS785). The slit array consists of five 100 μm wide slits spaced 350 μm (center to center) apart to match the dimensions of the array of Raman signals entering the spectrometer. This multislit array improves both the reduction of background noise and the confocal detection of the Raman signals. The Raman spectral image from all trapped particles was detected by a CCD camera (PIXIS 100BR, Princeton Instruments) mounted onto the spectrometer. Figure 2B is a representative image of Raman spectra from an array of trapped particles (2 × 3 in this case) that is captured by the CCD camera. Figure 2C is a plot of the intensity as a function of pixel position (x direction) for the image in Figure 2B. Modulated Multifocal Detection. The dimensions of the trap array can be adjusted, and the vertical pixels of the CCD chip can be binned accordingly in order to detect, with no cross talk, the individual Raman spectra displaced vertically on the camera. However, the overlap and superposition of the Raman spectra in the horizontal x direction is unavoidable if high spectral resolution (5 cm−1/pixel) and broad spectral coverage (200−2300 cm−1) are desired, which our current Raman detection system provides. In Figure 2B and C, the three Raman spectra in each row overlap each other, with a specific pixel shift between the spectra. It is possible to avoid overlap of adjacent spectra horizontally by using a grating with a smaller number of grooves per millimeter and increasing the distance between neighboring trapped particles. However, the trade-off is a decrease in spectral resolution, and additionally, the horizontal dimensions of the CCD chip will become a factor in limiting the number of spectra that can be measured simultaneously. In order to decompose the superimposed

Figure 1. Schematic of the 2-D laser tweezers Raman spectroscopy system with a modulated multifocal detection scheme. Abbreviations: SLM, Spatial Light Modulator; O, Objective; D, Dichroic Mirror; L, Lens; M, Mirror; S, Spectrometer; C, Camera.

with a maximum power of 1 W (Sacher Lasertechnik) was illuminated onto a SLM (Boulder Nonlinear Systems). The SLM was programmed by a computer to modulate the phase of the incident laser beam, which generated a holographic optical tweezers array. In this work, a TEM00 Gaussian beam was used for the optical traps. The iterative Gerchberg−Saxton (GS) algorithm was used for the hologram calculation.15,16 After passing through a dichroic mirror, the phase modulated laser beam from the SLM was directed into an objective (Olympus, 60×/1.2 W) to form the 2-D optical trapping array. The total laser power before the SLM was 170 mW, and the average power of each focus in the laser tweezers array was 5.5 mW. 6605

dx.doi.org/10.1021/ac5012188 | Anal. Chem. 2014, 86, 6604−6609

Analytical Chemistry

Article

Figure 3. Modulated multifocal detection for reconstructing individual Raman spectra. (A) Bright-field image of trapped 3 μm polystyrene beads and 1 μm PMMA beads in a 4 × 5 laser tweezers array. (B) Reference Raman spectra of a 3 μm polystyrene bead and a 1 μm PMMA bead using 5 mW laser power. (C, D, E, and F) Reconstructed individual Raman spectra of rows 1, 2, 3, and 4, respectively. For all the spectra, intensities of the Raman band at 1017 and 813 cm−1 for polystyrene and PMMA, respectively, were normalized to the same level.

Raman spectra of the mth row of the ith designed Raman measurement pattern. The individual Raman spectra of the mth row can be reconstructed by the following linear equations:

spectra in the horizontal direction and retrieve the individual spectra, we have invented a modulated multifocal detection method for spectral collection and reconstruction. Consider a 2-D 4 × 5 laser tweezers array that traps 20 particles. The entire 4 × 5 array of overlapped Raman spectra is defined as the initial pattern and can be expressed mathematically as a matrix M0:

⎛1 ⎜ 1 M0 = ⎜ ⎜1 ⎜ ⎝1

1 1 1 1

1 1 1 1

1 1 1 1

1⎞ ⎟ 1⎟ 1⎟ ⎟ 1⎠

⎛0 ⎜ ⎜1 ⎜1 ⎜ ⎜1 ⎝1

(1.1)

in which each element of the array represents the Raman spectrum from each individual focus. To resolve the 20 individual Raman spectra, five different Raman spectral array patterns need to be measured. For example, the following five Raman measurement patterns are designed and defined as Mi (i = 1, ... 5): ⎛0 ⎜ 0 M1 = ⎜ ⎜0 ⎜ ⎝0

1 1 1 1 ⎛1 ⎜ 1 M3 = ⎜ ⎜1 ⎜ ⎝1 ⎛1 ⎜ 1 M5 = ⎜ ⎜1 ⎜ ⎝1

1 1 1 1 1 1 1 1 1 1 1 1

1 1 1 1 0 0 0 0 1 1 1 1

⎛1 0 1⎞ ⎟ ⎜ 1 0 1⎟ , M2 = ⎜ ⎟ ⎜ 1 0 1 ⎟ ⎜ ⎝1 0 1⎠ ⎛1 1 1⎞ ⎟ ⎜ 1 1 1⎟ , M4 = ⎜ ⎜1 1 1⎟ ⎟ ⎜ ⎠ ⎝1 1 1 1 0⎞ ⎟ 1 0⎟ 1 0⎟ ⎟ 1 0⎠

1 1 1 1 1 1 1 1

1 1 1 1 1 1 1 1

1 1 0 1 1

1 1 1 0 1

m ⎛ I ⎞ ⎛ I1 ⎞ 1 ⎞⎜ m1 ⎟ ⎜ ⎟ ⎟ ⎜ m⎟ 1 ⎟⎜ Im 2 ⎟ ⎜ I 2 ⎟ ⎜ ⎟ m 1 ⎟⎜ Im3 ⎟ = ⎜ I3 ⎟ ⎟ ⎜ ⎟ 1 ⎟⎜ Im 4 ⎟ ⎜ I m ⎟ 4 ⎜ ⎟ 0 ⎠⎜ I ⎟ ⎜ m ⎟ ⎝ m5 ⎠ ⎝ I5 ⎠

(1.3)

The solutions for these equations are 4(Im1) = I2m + I3m + I4m + I5m − 3(I1m); 4(Im2) = I1m + I3m + I4m + I5m − 3(I2m); 4(Im3) = I1m + I2m + I4m + I5m − 3(I3m);

1⎞ ⎟ 1⎟ , 1⎟ ⎟ 1⎠ 0 0 0 0

1 0 1 1 1

4(Im4) = I1m + I2m + I3m + I5m − 3(I4m); 4(Im5) = I1m + I2m + I3m + I4m − 3(I5m)

(1.4)

By processing the data using the above equations, the 20 individual Raman spectra of each trapped particle in the 2-D array can be reconstructed. In general, for a 2-D m × n array, n different array patterns (i.e., i = n) need to be measured to retrieve all individual Raman spectra. In the present setup, the modulated multifocal detection scheme was realized by using custom designed mask patterns placed in the detection arm of the instrument to block the Raman light. Five different mask patterns were fabricated using a 3D printer (UP! Start Plus v1.1s, TierTime) and mounted in a motorized filter wheel (FW102W, Thorlabs) and placed in the focal planes of two lenses, as shown in Figure 1. A custom written MatLab program was used to synchronize the switching of the filter wheel position to change block patterns and the collection of the Raman image by the CCD camera. In our

1⎞ ⎟ 1⎟ , 1⎟ ⎟ 1⎠

(1.2)

where 0 in the matrix means no Raman signal is measured for the corresponding trapped particle in the laser trap array. We define Imn as the individual Raman spectrum of each trapped particle (m = 1...4; n = 1...5), and Imi=1...5 as the overlapped 6606

dx.doi.org/10.1021/ac5012188 | Anal. Chem. 2014, 86, 6604−6609

Analytical Chemistry

Article

Figure 4. (A) Bright-field image of laser trapped 3 μm polystyrene beads in a 1 × 5 line array along the horizontal x direction. (B) Overlapped Raman spectra of the laser trapped 1 × 5 polystyrene beads. (C) Calibrated Raman spectrum of the central polystyrene bead # 3. (D) Dependence of Raman peak positions on the laser trap positions of each particle. (E) Relative positions of the Raman peaks of beads #1, 2, 4, and 5 compared to the peak positions of central polystyrene bead #3.

current instrument configuration. The biggest power loss in our system occurs with the SLM, which has a zero-order diffraction efficiency of ∼65% to realize this tweezers array. While the SLM gave the greatest flexibility in the initial design and testing of this approach, future iterations of this instrument can use a microlens array instead of the SLM to generate the laser tweezers array, which is also more cost-effective. Although the exposure time for each Raman pattern in matrices 1.2 was 2 s, the reconstructed spectra of each bead shown in Figure 3 actually reflects an 8 s acquisition time. The reason is that after the five modulated patterns are acquired, the Raman signal of each particle has been collected four times. This can also be understood from the factor of 4 in eqs 1.4. Therefore, the added advantage of this scheme is that in addition to parallel acquisition of spectra, the signals that are acquired for each pattern are accumulated and lead to a final spectrum with improved total signal intensity. The total signal intensity of select Raman peaks (1001 and 813 cm−1) and the signal-to-noise ratio (SNR) of the reconstructed Raman spectra for the polystyrene and PMMA beads in rows 1 and 2 (Figure 3a) were calculated. These values were compared to those obtained from the measured 2 s spectra (i.e., rows 1 and 2 in the M0 pattern in eq 1.1) from which the reconstructed spectra were derived. These results are shown in Figures S5 and S6 and Table S1 in the Supporting Information. The reconstructed spectra have a total signal intensity that is approximately 4 times greater than the original 2 s spectra, as expected. In addition, the reconstructed spectra also has a factor of ∼1.5 improvement in the SNR compared to the 2 s spectra (Table S1, Supporting Information). Therefore, a significant advantage of this modulated multifocal detection scheme is that the reconstructed spectra have both higher total signal intensity and better SNR compared to individual spectra that would be acquired in only 2 s from each of the five patterns in eq 1.2. The primary objective of this work was to demonstrate the feasibility of the modulated, multifocal detection scheme for

experiments, the exposure time for each Raman measurement pattern in matrices 1.2 was 2 s.

■

RESULTS AND DISCUSSION Figure 3A shows a bright field image of 3 μm polystyrene and 1 μm PMMA beads trapped in the 4 × 5 array. Rows 1 and 2 are polystyrene and PMMA beads, respectively, while rows 3 and 4 consist of a random mix of beads. Typical Raman spectra of a single polystyrene and PMMA bead are shown in Figure 3B. In our Raman system, the measured Raman intensity of the 3 μm polystyrene bead was ∼10 times higher than that of the 1 μm PMMA bead with the same laser power and acquisition times. By switching the filter wheel position to change the measured Raman pattern array as described in the matrices in eq 1.2, five different Raman signal patterns were collected (see Supporting Information, Figure S1−S4). Following the data processing using eqs 1.4, individual Raman spectra of the 20 trapped beads were reconstructed. Figure 3C−F show the reconstructed spectra of each bead for all four rows. The signal intensities are plotted as a function of the pixel position on the camera to better visualize the offset of the spectra on the CCD detector in the horizontal dimension. The spectral profile for all beads was accurately reconstructed and matched the standard spectra in Figure 3B. Because of the index and size differences between the polystyrene and PMMA beads, we were able to visually identify the beads in each position in the array, which allowed us to then confirm the accuracy of the reconstructed spectra. Also, the results indicate that there is no spectral crosstalk between the bead samples following the reconstruction algorithm. This is particularly evident in the reconstructed PMMA spectra, which show no spectral contributions from the polystyrene beads, which have a ∼10 times stronger signal that would be clearly noticeable if crosstalk was present. Given that our laser source has a maximum power of 1 W, we anticipate that there should be enough power to be able to generate at least 100 foci (5.5 mW for each focus) under the 6607

dx.doi.org/10.1021/ac5012188 | Anal. Chem. 2014, 86, 6604−6609

Analytical Chemistry

Article

retrieving individual Raman spectra from a multifocal array of overlapping, superposed spectra. The results presented here were obtained using just one example of five simple mask patterns (matrices in eq 1.2) where each pattern blocks only one column of Raman spectra. There are, however, a number of different patterns that can be used to reconstruct individual Raman spectra. An example of a different set of five detection pattern matrices for the 4 × 5 laser tweezers array is shown in the Supporting Information. For different laser tweezers array dimensions, a specific combination of different mask patterns can be chosen such that the signal intensities and SNR of the reconstructed Raman spectra are optimized. Moreover, because this detection scheme allows for a large number of spectra to be acquired and accumulated from the same laser focus, noise reduction algorithms can be applied during postdata processing to further improve the SNR of the final reconstructed individual Raman spectra. These optimization studies will be pursued in the future. A six-position motorized filter wheel was used to switch between the block patterns. This device, though simple, was functional and allowed us to demonstrate the basic concept of the technique. For future designs of this instrument in which detection of a 2-D m × n array is needed where n is large and where faster switching of the blocking patterns is desired, alternative devices such as a digital micromirror device (DMD) can be used to rapidly modulate the detection pattern. The spectra that are acquired with this multifocal detection scheme are offset along the horizontal axis of the CCD detector. Therefore, the development of a robust and accurate method for calibrating the Raman wavenumbers of these spectra needs to be considered. Figure 4A and B show a brightfield image of trapped beads in a row, and their corresponding overlapped Raman spectra, which is composed of the five individual spectra of each bead that passes through the five slits of the slit array. We defined the central slit that passes the Raman signal of the center bead labeled #3 in Figure 4A as the standard slit for the wavenumber calibration of the spectrometer. Figure 4C shows the calibrated Raman spectra of polystyrene bead #3. The Raman spectra of the other four beads are shifted in position along the wave dispersion direction (horizontal x direction) of the spectrometer’s CCD chip relative to the spectrum of bead # 3. We have examined the dependence of the pixel positions of the major polystyrene Raman peaks on the laser trap position (or slit position) of each bead. Figure 4D shows a linear relationship between each Raman peak position in the CCD chip with the laser trap position, which is consistent with previous reported literature.14 However, the slopes of the fitted lines in Figure 4D are different. For example, the 621, 1001, and 1602 cm−1 lines had slopes of 27.2, 28.7, and 31.9, respectively, which indicates that the different Raman peaks are shifted by a different number of pixels. We have also examined the relative position of the Raman peaks of bead 1, 2, 4, and 5 compared to the peak positions of the center bead #3. The highly linear relationship that is shown in Figure 4E indicates that the parameters obtained from a linear fit of the data in Figure 4E allows us to accurately shift and calibrate the Raman spectra of all the beads adjacent to the center bead #3 and align it to the center position of the spectrometer. Figure 5A and B are plots of the individual Raman spectra from Figures 3C−F after the calibration procedure, which shows that the Raman peaks for both the polystyrene and the PMMA spectra are well aligned.

Figure 5. Relative wavenumber calibrated individual Raman spectra for (A) polystyrene and (B) PMMA beads from Figure 3C−F.

■

CONCLUSIONS We have developed a rapidly modulated multifocal detection scheme that enables Raman spectra from a 2-D focal array to be acquired simultaneously. Key components of the system include a spatial light modulator (SLM) and a Raman detection modulation method that collects spectra from different combinations of multifocal patterns. While the current work shows that spectra from 20 individual foci can be acquired simultaneously, we expect to be able to increase this to at least 100 foci (a 10 × 10 array). Ultimately, the maximum number of foci will be limited by the maximum power of the laser source. We modulated the detection arm using different block patterns. In theory, the laser illumination pattern could be modulated as well by programming different patterns with the SLM. However, there are two drawbacks with modulating the laser excitation pattern. First, changing the patterns with the SLM can be inefficient. More importantly, changing the illumination pattern can adversely affect the stable trapping of a large number of particles. For nontrapping applications, this obviously would not be an issue. This present work was motivated by the need to improve the analytical capabilities of LTRS by enabling parallel detection of spectra from a 2-D array of trapped cells and to maintain stable trapping of the particles by keeping all of the optical traps “on” simultaneously. However, in addition to its application in LTRS, this new multifocal detection scheme can also be used to improve the imaging speed of hyperspectral spontaneous Raman and broadband coherent anti-Stokes Raman scattering (CARS) microscopes. In fact, this technique is not solely limited to the detection of Raman signals but rather is broadly applicable to the hyperspectral parallel detection of any type of optical signal. 6608

dx.doi.org/10.1021/ac5012188 | Anal. Chem. 2014, 86, 6604−6609

Analytical Chemistry

■

Article

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work has been supported by funding from the Keaton Raphael Memorial Foundation and the National Science Foundation, Creation of an Ecosystem for Biophotonics Innovation IIP 1127888 and Ecosystem for Biophotonic Innovation Building Sustainability IIP 1343479.

■

REFERENCES

(1) Xie, C. G.; Dinno, M. A.; Li, Y. Q. Opt. Lett. 2002, 27, 249−251. (2) Petrov, D. V. J. Opt. A: Pure Appl. Opt. 2007, 9, S139−S156. (3) Kong, L. B.; Zhang, P. F.; Wang, G. W.; Yu, J.; Setlow, P.; Li, Y. Q. Nat. Protoc. 2011, 6, 625−639. (4) Chan, J. W. J. Biophotonics 2013, 6, 36−48. (5) Chan, J. W.; Taylor, D. S.; Lane, S. M.; Zwerdling, T.; Tuscano, J.; Huser, T. Anal. Chem. 2008, 80, 2180−2187. (6) Chen, D.; Huang, S. S.; Li, Y. Q. Anal. Chem. 2006, 78, 6936− 6941. (7) Singh, G. P.; Creely, C. M.; Volpe, G.; Grotsch, H.; Petrov, D. Anal. Chem. 2005, 77, 2564−2568. (8) Kong, L. B.; Zhang, P. F.; Setlow, P.; Li, Y. Q. Anal. Chem. 2010, 82, 3840−3847. (9) Liu, R.; Taylor, D. S.; Matthews, D. L.; Chan, J. W. Appl. Spectrosc. 2010, 64, 1308−1310. (10) Zhang, P. F.; Kong, L. B.; Setlow, P.; Li, Y. Q. Opt. Lett. 2010, 35, 3321−3323. (11) Dufresne, E. R.; Spalding, G. C.; Dearing, M. T.; Sheets, S. A.; Grier, D. G. Rev. Sci. Instrum. 2001, 72, 1810−1816. (12) Sow, C. H.; Bettiol, A. A.; Lee, Y. Y. G.; Cheong, F. C.; Lim, C. T.; Watt, F. Appl. Phys. B: Lasers Opt. 2004, 78, 705−709. (13) Qi, J.; Shih, W. C. Opt. Lett. 2012, 37, 1289−1291. (14) Qi, J.; Li, J. T.; Shih, W. C. Biomed. Opt. Express 2013, 4, 2376− 2382. (15) Polin, M.; Ladavac, K.; Lee, S. H.; Roichman, Y.; Grier, D. G. Opt. Express 2005, 13, 5831−5845. (16) Di Leonardo, R.; Ianni, F.; Ruocco, G. Opt. Express 2006, 15, 1913−1922.

6609

dx.doi.org/10.1021/ac5012188 | Anal. Chem. 2014, 86, 6604−6609