We evaluated the optimal detection angle for maximizing the signal to noise ratio (SNR) in sub-diffraction resolution photothermal microscopy. The angular dependent photothermal signal was calculated based on scattering theory using the temporally modulated Yukawa potential, and its detection angle and modulation frequency dependencies were analyzed. We verified the theoretical findings by imaging gold nanoparticles using laser diode based photothermal microscopy with balanced detection scheme. High-sensitivity (SNR ~40) photothermal biological imaging of a mouse brain was also demonstrated.
© 2014 Optical Society of America
Photothermal microscopy is an efficient method for observing single molecules or nanomaterials using only their optical absorption [1–3]. One of the advantages of photothermal microscopy is that it is less affected by photobleaching, which is the major drawback in fluorescent measurement. By virtue of this fact, it has been used in several biological applications in recent years, such as imaging of live cells using 5nm gold nanoparticles for arbitrarily long periods , evaluation of the hydrodynamic properties of cytosol by means of photothermal absorption correlation spectroscopy [5–8], label-free imaging of heme proteins with two-photon excitation , and the visualization or tracking of membrane proteins moving at the surface of live cells [10, 11].
Typically in photothermal microscopy, two laser beams with different wavelengths (pump and probe beams) are incident on the sample through a focusing lens. The pump beam increases the temperature around the focal point of the optical absorbing sample. This results in local refractive index variations and induces deflection of the probe beam. The variation of the refractive index with temperature is typically in the order of −10−4 K−1, which is readily detected by a lock-in detection scheme as a change in probe beam transmissivity. The pump beam is modulated at 0.1 to 1 MHz and an integration time (the time constant of the lock-in amplifier) of 1-10 ms is used to attain a sufficient signal to noise ratio (SNR) . Thus, the image acquisition time is still slower than that achieved by conventional laser scanning fluorescence microscopy, with an integration time of down to ~1 μs per pixel.
Several technical improvements in photothermal microscopy have increased the SNR. As the relative change in the probe beam transmissivity is detected in a typical experimental setup, it is important to reduce the intensity noise of the beam. To reduce laser noise, a high frequency lock-in detection scheme of over 1 MHz is used, as laser noise is dominant at lower frequencies . Alternatively, an auto-balanced detection scheme can be used for common mode noise reduction at a lower lock-in frequency (up to ~100 kHz) . Using these techniques it is possible to reach shot noise limited sensitivity. It can also be useful to employ an embedding medium with large , or media exhibiting a phase transition with large variations of refractive index around the critical temperature [15, 16]. Furthermore, it is important to select the optimal detection angle for the probe beam (the numerical aperture of the condenser lens NAc) to maximize SNR. According to the literature on photothermal microscopy, NAc is empirically set to be smaller than that of the focusing lens [5, 10, 12, 17, 18]. However, no systematic studies have been conducted to investigate this practice.
In this study, theoretical and experimental analyses were carried out to provide clear understanding of the relationship between the detection angle of the probe beam and the SNR. High-sensitivity photothermal imaging of a mouse brain was also demonstrated with our sub-diffraction resolution photothermal microscopy, using intensity modulated laser diodes with a balanced detection scheme.
Several models have been proposed to describe the nanoscopic photothermal signal. Selmke and Cichos et al. calculated the difference in transmissivity of the focusing probe beam between two steady states, in the presence and absence of the pump beam, using generalized Lorenz-Mie theory [18–20]. They also suggested that the photothermal signal was analogous to Rutherford scattering [21, 22]. However, their model is based on a steady-state calculation and not applicable to the case when thermal diffusion length during one period of the modulating pump beam is comparable to or smaller than the focusing spot size of the probe beam. This condition is considered to be important to attain sub-diffraction resolution as detailed in Sec. 4. Berciaud and Lounis et al. provided an analytical description of the photothermal signal, based on the theory of light scattering from a temporally modulating medium . They introduced the time dependent refractive index profile induced by modulation of pump beam intensity, based on the thermal diffusion equation, and derived the photothermal signal intensity from the interference term between the scattering field and an isotropic local oscillator. On the basis of this model, the dependence of the modulation frequency on the photothermal signal was discussed for both forward and backward configurations. However, their plane-wave approximation of the incident probe field did not include the effect of focusing (or transmitted diverging field) on the photothermal signal. As the transmitted field also interferes with the scattered field, it is important to take into account the effect of the focusing field on the photothermal signal.
In this study, we extend these models by considering both the effects of the focusing field and modulating heat sources on the photothermal signal. We consider the case when pump and probe beams are incident on the sample and only the pump beam is absorbed by the point absorber, which then causes an increase in temperature in a surrounding medium. The temperature rise induces an increase or decrease in the refractive index depending on the sign of the thermooptic index and thus the probe beam is deflected. As modulation frequency of the pump beam intensity is much lower than the frequency of the electromagnetic field, we consider the quasi-static Helmholtz equation 11, 17]Eq. (2) into Eq. (1) readsEquation (3) is equivalent to the quasi-static Schrödinger equations, where corresponds to the potential field. The first term on the r.h.s. of Eq. (4) decays with the inverse of radial distance and corresponds to the Coulomb potential in quantum mechanical treatment. The Coulomb potential is effective over a greater distance. The second term is sinusoidally modulated and the amplitude (1/r)exp(r/rc) approaches zero at rc. This factor is equivalent to the Yukawa potential (the screened Coulomb potential), which was introduced to describe strong nuclear force between nucleons in particle physics . The Yukawa potential coincides with the Coulomb potential in the limit rc . In photothermal microscopy, the first term can be regarded as the interaction with the temporally averaged (static) point heat source, whereas the second term is the interaction with the modulated point heat sources and thus the interaction length should be smaller than rc.
The asymptotic solution of Eq. (3) in the far-field region for an incident plane wave readsFig. 1.The scattering amplitude is calculated under the first-order Born approximation asEq. (6) corresponds to the scattering amplitudes due to the Coulomb potential. The second and the third terms correspond to the scattering amplitude due to the temporally modulated Yukawa potential and their sum can be written as , where = tan−1(1/u) is the phase shift with . The scattering cross section, defined as the ratio of angularly integrated scattered field intensity to that of the incident field intensity, is given by24]. When the incident and deflected (scattered) photons are collected through a condenser lens with NAc = n0sinφc in the far field, as depicted in Fig. 1, the fraction of photons deflected outside the condenser lens (the residual scattering cross section) is expressed asEq. (8) is the fraction of photons deflected in the angle lower than φc. We note that the residual scattering cross section is zero when photons are collected in all angles, i.e., .
In photothermal microscopy, the focusing beam is incident on the sample through an objective lens and thus we consider a super-position of Eq. (5) at multiple incident angles to examine the effect of the focusing field on the photothermal signalEq. (15) refers the fraction of photons that have incident angle lower than φc and undergo scattering, while the second term is the fraction of the scattered photons in the angle lower than φc. Here, we assumed that the incident field is focused at the position of the point light absorber. The scattering amplitude is given by the sum of the three terms in Eq. (6), so Eq. (15) can be rewritten separately as
Figure 2(a) shows Yin and Yout as a function of NAc. Their magnitude and the phase of the output are also shown for the set of |Δk| = 2, n0 = 1, and NAf = 0.95. It is important to note that the magnitude has a peak at NAc = 0.73 and is zero for ΝΑc > ΝΑf. Thus, ΝΑc should be smaller than that of the focusing lens to achieve optimal SNR. We also note that the sign of Yin is negative for a negative , which suggests that transmissivity increases in the presence of the pump beam. This is because the wide-angled incident field component, with φ’ > ΝΑ c, is deflected to the angle lower than ΝΑ c and thus contributes to the positive signal gain. We found that the peak position of Yin is smaller than that of Yout, and so Φ has a tendency to increase with φ c. Figure 2(b) shows angular dependence of M for several values of rc. We found that the peak value decreases and the peak position slightly shifts to the lower angular side as rc decreases.
Detection angle dependence of the photothermal signal was examined experimentally in our pump-probe microscopy system, which uses intensity modulated laser diodes (LDs) and a balance detector. The details have been described previously . We used a 488 nm LD for the pump, and a 640 nm LD for the probe beam. Each beam was collimated through a spatial filter and two beams were combined using dichroic mirrors. The probe and pump-beam intensities were typically modulated at frequencies of ω1 and ω2, respectively. A beat signal |ω1-ω2| is generated by the bilinear interaction of the pump field and the probe field in the sample. We used a lock-in technique to detect the signal, and the lock-in amplifier was referenced to the beat frequency. By changing the modulation frequencies (ω1 and ω2) while holding the difference frequency |ω1-ω2| constant, this dual modulation scheme has the potential to acquire time-resolved information without a high-speed detector. To counteract the intensity fluctuation of the probe laser, we split the probe beam into two and directed the two resulting beams to an auto-balanced photodetector (New Focus, Nirvana). This auto-balanced detector serves to cancel both the laser noise and the intensity imbalance due to refractive index variations during the sample scanning. We confirmed that when the probe beam power incident on the photodiode was above ~10 μW, the signal to noise ratio was mainly limited by shot noise. The pump and probe beams were linearly polarized. A polarizing beam splitter (PBS) was used to direct the combined beam to the objective lens (UPLSAPO 40X2, Olympus) with numerical aperture NAf of 0.95. The beam size was adjusted to fill the back aperture of the objective lens. The sample position was raster scanned using a three-axis positioning stage driven by piezo actuators (MAX311D, Thorlabs or P-622.2, PI). A condenser lens with variable NAc (U-AAC, Olympus) was used to collect the transmitted light. This lens is equipped with a diaphragm for varying NAc. The maximum value of NAc was 1.4 with immersion oil, while it was ~0.95 without immersion oil. In this study, immersion oil was not employed since a dry objective was used for focusing the probe beam. The distance between the glass substrate and the front side of the condenser lens was ~0.5 mm. A narrowband filter matching the laser line was placed in front of the detector to filter out the pump beam, so that only the probe beam was detected. The pump and probe beams were collimated so that they focused at the same position.
We measured the photothermal signal from 20 nm diameter gold nanoparticles dispersed in a polyvinyl alcohol (PVA) film on a glass slide with varying NAc (Fig. 3). The wavelength of the pump beam overlapped the absorption spectrum of the gold nanoparticles as the plasmon resonance peaked at 524 nm. The pump and probe powers incident on the sample were 0.7 mW and 0.3 mW, respectively and were modulated at 1.015 MHz and 1.000 MHz, respectively. The time constant of the lock-in amplifier was 0.5 ms and the pixel dwell time was 1 ms. We found that the full width at half maximum (FWHM) value of a single particle is ~240 nm, which is 19% smaller than the diffraction limited spot size of the pump beam and 32% smaller than that of the probe beam . Furthermore, the axial spot size was 590 nm, which is 33% smaller than the diffraction-limited axial size of the probe beam. Upper and lower panels of Fig. 3 show magnitude, M, and phase shift, Φ, images, respectively, with several values of NAc. Since the thickness of the PVA film is about 20 μm, gold nanoparticles are distributed in both lateral and axial direction. This leads to the signal intensity distribution of gold nanoparticles in the xy-image. To characterize the images, we calculated the integrated intensity of M and the average Φ for each image, summarized in Fig. 4(a).It is important to note that with increasing NAc, integrated intensity increases and then decreases when NAc is larger than ~0.4, while Φ monotonically increases with NAc. These tendencies are in accordance with the theoretical calculation as shown in Fig. 2(a). Furthermore, we evaluate the effect of NAc on the signal to noise ratio (SNR) to ascertain the optimal NAc. To this end, the maximum value Smax and the root mean square (r.m.s.) of background level Sb are calculated for each image. Figure 4(b) shows Smax, Sb, and their ratio Smax/Sb as a function of NAc. We found that Smax has a peak at NAc = 0.45, while Sb increases with increase in NAc, as an increase in the beam power incident on the detector is accompanied by an increase in noise power. Consequently, Smax/Sb has a maximum value of 60 at NAc ~0.4 and significantly decreases to 12 at NAc = 0.85. We calculated maximum intensity Smax instead of the mean intensity Smean because it is difficult to determine the latter due to the background noise that is prominent for a large NAc. However, we found that Smean gives similar trend as Smax. We also checked the modulation frequency dependence on the signal intensity and found that the signal intensity decreases by approximately half when the modulation frequency reaches 2 MHz. This agrees with the experimental results in the literature [1, 2]. We also measured the incident beam power dependence and found that the signal is proportional to the product of the pump and probe power. These facts verify that the measured signal is to be attributed to the photothermal effect.
Biological photothermal imaging was demonstrated using mouse brain labeled with weakly fluorescent quantum dots (Fig. 5). The brain of a two month old C57BL/6 mouse was embedded in O.C.T. compound (SAKURA Seiki Co. Ltd., Tokyo Japan), frozen in liquid nitrogen, sliced at 14 μm, and air-dried. Sections were stained with anti-α-tubulin antibody labeled with eFluor 650NC (eBioscience, CA, San Diego), mounted on coverslips, and imaged. The samples were placed under atmospheric conditions for several months to induce surface oxidization of quantum dots, where rapid non-radiative exciton decay due to charge recombination dominates the relaxation pathway and photo excitation energy is mainly converted to heat. The pump and probe beam power were 4 mW and 0.6 mW, respectively. The time constant of the lock-in amplifier was 0.5 ms and the pixel dwell time was 1 ms. Pump and probe beams were modulated at 1.1 MHz and 1.0 MHz, respectively. Figure 5(a) shows an overview image with the image size of 200 x 200 μm. Figures 5(b)–5(d) show a set of images with different values of NAc. It can be clearly seen in Figs. 5(b)–5(d) that the largest SNR is attained for NAc = 0.36 [Fig. 5(b)]. The calculated values of Smax/Sb were (b) 40, (c) 17, and (d) 6. We confirmed that there was minimal damage to the sample after measurements. It is possible that stimulate emission contributed to the lock-in signal, as the laser wavelength overlapped the fluorescence of the quantum dot peak at 650 nm . However, we consider the photothermal signal is a majority component because the detected signal significantly decreases for a large NAc.
4. Discussion and conclusion
In this study, theoretical and experimental analyses were carried out to ascertain the optimal detection angle for the probe beam to maximize SNR. We found that signal intensity significantly depends on NAc. The magnitude of the photothermal signal has a maximum value when NAc = 0.4, which is less than half that of the focusing lens of NAf = 0.95. We also confirmed that the phase shift, Φ, increases as NAc increases. These trends qualitatively agree with the theoretical calculation in which the photothermal signal has a maximum value in the region NAc < NAf, while it is zero for NAc > NAf. However there are several discrepancies between the experimental data and theoretical calculations. In the experimental data, the photothermal signal remains even when NAc ~NAf, although the SNR is small. This is probably attributable to the static scattering field, such as Rayleigh scattering by a gold nanoparticle or light scattering at the surface of the glass plate, which serves as a local oscillator and causes interference with the temporary modulated scattering filed . In the present study, signal was detected in the forward direction, and thus we approximated that the photothermal signal mainly comes from the interference between temporally modulated scattering field and the diverging transmitted probe field. This makes the signal to vanish for NAc > NAf in the theoretical calculation. On the other hand, for the backward detection, the photothermal signal may be attributable to the interference between the static scattering field and the temporally modulated scattering field. In that case, signal intensity should be proportional to the square of NA .
The peak position of the photothermal signal also differs between the experimental data and the theoretical calculations. In this regard, we examined how the theoretical curve of M depends on the parameters, and found that the optimal value of NAc (position of the maximum M) decreases as λ or the focal spot size increases, while it increases with NAf. Furthermore, we assume uniform probe beam at the pupil of the focusing lens in theoretical calculations. However, experimentally, the incident beam is a Gaussian beam truncated by the limiting aperture of the objective pupil. For a Gaussian beam focusing at the sample position A(k) = exp(-φ’2/2σ2) with σ = 0.1π, the optimal value appears at NAc ~0.3. However, Φ decreases as NAc increases. Further study is needed in regard to this point including the effect of the focal position on the signal.
Theoretical analysis and experiments were carried out for a particle whose diameter is much smaller than wavelength of visible light (20-nm Au nanoparticle and 5-nm quantum dot), since a smaller particle is usually more preferred for a probe molecule in biological imaging. For a large particle, it is expected that forward static scattering field also interferes with the modulated scattering field, which gives a similar result as in the case of the transmitted probe field. In this regard, in images of the slice of mouse brain, we observed several spots with the sizes from several hundred nm to ~1 μm due to the aggregates of quantum dots. We found that signal intensities from these spots also decrease for a large NAc. This suggests that the optimal NAc is smaller than NAf regardless of the size of nanoparticles.
In a conventional thermal lens model, change in probe beam transmissivity is explained by considering the deflection by a concave lens induced by the temperature rise. This picture may be valid when rc is larger than the focal spot size. In such a case, the temporally modulated Yukawa potential in Eq. (4) is approximated by the Coulomb potential, and consequently the photothermal signal is described in a similar manner to Rutherford scattering [21, 22] or by a lens-like model [18, 20]. In the present experiment, by assuming the thermal diffusivity of 0.1 mm2/s and the modulation frequency of the pump beam of 1 MHz, rc is estimated to be ~0.1 μm, which is about four times smaller than the calculated focal spot size of the probe beam (0.61λ/NAf = 410 nm). In this regard, it is interesting to examine the spatial resolution with changing modulation frequencies, because when rc is much smaller than the focal spot size, the spatial resolution (the point spread function) should be determined by the product of the pump and probe beam intensity at the focal point. In this case, spatial resolution improves [14, 25]. The improvement of the spatial resolution is attributed to the fact that the photothermal signal is based on the nonlinear interaction between the two laser beams and the sample. In contrast, it is expected that the spatial resolution is determined by a pump beam intensity distribution when rc is much larger than the focal spot size. It is also important to examine the z-dependence of the photothermal signal with varying rc. In a thermal lens model, signal intensity exhibits two peaks that correspond to focusing or defocusing when the sample position is moved in the axial direction . However, in the present study, the amplitude of the lock-in signal exhibited only a single peak and the phase was nearly flat and did not exhibit the π shift at the origin. Further investigation is needed to clarify the modulation frequency dependence on the photothermal signal.
This study was financially supported by a Grant-in-Aid for Scientific Research (No. 24740261) received from the Japan Society for the Promotion of Science and a joint research project at the Institute of Laser Engineering, Osaka University, under contract number B1-27.
References and links
2. S. Berciaud, L. Cognet, G. A. Blab, and B. Lounis, “Photothermal heterodyne imaging of individual nonfluorescent nanoclusters and nanocrystals,” Phys. Rev. Lett. 93(25), 257402 (2004). [CrossRef] [PubMed]
3. A. Gaiduk, M. Yorulmaz, P. V. Ruijgrok, and M. Orrit, “Room-temperature detection of a single molecule’s absorption by photothermal contrast,” Science 330(6002), 353–356 (2010). [CrossRef] [PubMed]
4. C. Leduc, S. Si, J. Gautier, M. Soto-Ribeiro, B. Wehrle-Haller, A. Gautreau, G. Giannone, L. Cognet, and B. Lounis, “A highly specific gold nanoprobe for live-cell single-molecule imaging,” Nano Lett. 13(4), 1489–1494 (2013). [PubMed]
6. P. M. R. Paulo, A. Gaiduk, F. Kulzer, S. F. G. Krens, H. P. Spaink, T. Schmidt, and M. Orrit, “Photothermal correlation spectroscopy of gold nanoparticles in solution,” J. Phys. Chem. C 113(27), 11451–11457 (2009). [CrossRef]
8. C. Leduc, J. M. Jung, R. P. Carney, F. Stellacci, and B. Lounis, “Direct investigation of intracellular presence of gold nanoparticles via photothermal heterodyne imaging,” ACS Nano 5(4), 2587–2592 (2011). [CrossRef] [PubMed]
9. S. Lu, W. Min, S. Chong, G. R. Holtom, and X. S. Xie, “Label-free imaging of heme proteins with two-photon excited photothermal lens microscopy,” Appl. Phys. Lett. 96(11), 113701 (2010). [CrossRef]
10. D. Lasne, G. A. Blab, S. Berciaud, M. Heine, L. Groc, D. Choquet, L. Cognet, and B. Lounis, “Single nanoparticle photothermal tracking (SNaPT) of 5-nm gold beads in live cells,” Biophys. J. 91(12), 4598–4604 (2006). [CrossRef] [PubMed]
11. L. Cognet, C. Tardin, D. Boyer, D. Choquet, P. Tamarat, and B. Lounis, “Single metallic nanoparticle imaging for protein detection in cells,” Proc. Natl. Acad. Sci. U.S.A. 100(20), 11350–11355 (2003). [CrossRef] [PubMed]
14. J. Miyazaki, H. Tsurui, A. Hayashi-Takagi, H. Kasai, and T. Kobayashi, “Sub-diffraction resolution pump-probe microscopy with shot-noise limited sensitivity using laser diodes,” Opt. Express 22(8), 9024–9032 (2014). [CrossRef] [PubMed]
15. W.-S. Chang and S. Link, “Enhancing the sensitivity of single-particle photothermal imaging with thermotropic liquid crystals,” J. Phys. Chem. Lett. 3(10), 1393–1399 (2012). [CrossRef]
16. A. N. G. Parra-Vasquez, L. Oudjedi, L. Cognet, and B. Lounis, “Nanoscale thermotropic phase transitions enhancing photothermal microscopy signals,” J. Phys. Chem. Lett. 3(10), 1400–1403 (2012). [CrossRef]
17. S. Berciaud, D. Lasne, G. A. Blab, L. Cognet, and B. Lounis, “Photothermal heterodyne imaging of individual metallic nanoparticles: Theory versus experiment,” Phys. Rev. B 73(4), 045424 (2006). [CrossRef]
22. M. Selmke and F. Cichos, “Photonic Rutherford scattering: A classical and quantum mechanical analogy in ray and wave optics,” Am. J. Phys. 81(6), 405 (2013). [CrossRef]
23. H. Yukawa, “On the interaction of elementary particles. I,” Proc. Phys.-Math. Soc. Jpn. 17, 48–57 (1935).
24. K. Gottfried and T.-M. Yan, Quantum Mechanics: Fundamentals, Second Edition (Springer, 2003).
25. J. Miyazaki, K. Kawasumi, and T. Kobayashi, “Resolution improvement in laser diodes-based pump-probe microscopy with an annular pupil filter,” Opt. Lett. 39(14), 4219–4222 (2014). [CrossRef]
26. K. Uchiyama, A. Hibara, H. Kimura, T. Sawada, and T. Kitamori, “Thermal lens microscope,” Jpn. J. Appl. Phys. 39(9A), 5316–5322 (2000). [CrossRef]