1. Introduction

Recently, demands to measure the IR spectra of molecular species in microscopic targets have been growing in various fields, such as cellular biology and nano-industry. IR microscopy can directly measure the IR absorption by microscopic samples, and it plays an indispensable role to answer these demands. However, the application of IR spectroscopy in microscopes has not been as successful as another vibrational spectroscopy, Raman spectroscopy [1–3]. The largest problem of IR microscopy is its low spatial resolution. The resolution of microscopes in linear optics is determined by the diffraction limit (δ) of the illumination, and is expressed by δ=0.61λ/NA, where λ and NA are the wavelength of the illumination and the numerical aperture of objective lens, respectively [4]. Since the diffraction-limited spotsize of a microscope is proportional to the wavelength of the monitoring light, conventional IR microscopes cannot have better than several micro-meter resolution [5,6]. Although scanning near-field infrared microscopy (SNIM) has excellent resolution less than 30 nm [7,8], it can obtain the information on the structures near the cellular surface.

To overcome these difficulties of IR microscopy, we have recently developed a novel type of far-field IR super-resolution microscope using transient fluorescence detected IR (TFD-IR) spectroscopy which is a type of two-color laser spectroscopy [9,10]. Our microscope system can image IR absorption of fluorescent samples at the resolution of visible light which is about ten-times smaller than the IR diffraction limit, and we succeeded in obtaining IR super-resolution images of fluorescent beads and fluorescent dye labeled plant cells [9,10]. In this paper, we report an improvement in the resolution of this microscope to sub-micron level by using a long working distance objective. Furthermore, we present two-point resolution result for fluorescent beads located within 1.4 µm each other which is significantly smaller than the IR diffraction limit (~3.4 µm). Many biological molecules in cells are fluorescent (for example, flavins show visible fluorescence, aromatic residues of proteins show UV fluorescence and so on). The vibrational modes of such fluorescent species are difficult to be studied by typical IR and Raman microscopes. The result of this report suggests our microscope can obtain the vibrational information and the structural information of such molecules in vivo.

2. Basic concept and principle of the TFD-IR super-resolution microscope

The principle of TFD-IR spectroscopy is shown in Fig. 1(a). The molecule vibrationally excited by an IR photon is capable to be pumped to the first electronically excited state by a second visible photon whose energy is slightly smaller than the S₁←S₀ absorption of the molecule. Therefore, if the molecule is fluorescent, only the molecular species excited by IR can emit fluorescence (transient fluorescence). In this way we can convert the IR absorption to visible emission, and hence IR absorption can be imaged at visible resolution by monitoring the transient fluorescence.

The optical layout of TFD-IR microscope is schematically shown in Fig. 1(b). In this setup, we moderately focused both the visible and the IR beams on the whole sample area. This enabled us to acquire the fluorescent image of the whole area, which has a dimension of several tens of micrometers, simultaneously without scanning. This technique can considerably shorten the measurement time, which is very useful when applying the microscope to biological samples.

 

Fig. 1. (a). Energy diagram of transient fluorescence detected IR (TFD-IR) spectroscopy. (b) Optical layout of an IR super-resolution microscope using TFD-IR.

Download Full Size | PDF

3. Experimental methods

The laser setup for our two-color IR super-resolution microscope was described previously [9]. It can be summarized as follows: the output of a cw mode-locked Ti:sapphire laser (Tsunami, Spectra-Physics Lasers division of Newport Corp, Mountain View, CA) pumped by a 5 W diode laser (Millennia, Spectra-Physics Lasers division of Newport Corp, Mountain View, CA) was used to seed a pump laser equipped regenerative Ti:sapphire amplifier (Legend-Elite-Pico (Evolution-30), Coherent Inc., Santa Clara, CA). The intensity of the 800 nm output pulse was ~3.4 mJ/pulse at 1 kHz repetition rate, and its pulse width was 2 ps. This output pulse was divided into three parts and used to generate visible and IR beams. One of them was frequency doubled by a BBO crystal and then brought into a traveling-wave optical parametric amplifier system (TOPAS 400, Light conversion, Vilnius, LA) to obtain a tunable visible pulse. The second divided pulse of the output from the regenerative amplifier was brought into another OPA system (TOPAS 800, Light conversion, Vilnius, LA) and a frequency doubled idler pulse was combined with the remaining output of the regenerative amplifier in a difference frequency generating unit (employing a KTB crystal) to generate the IR beam (2.5–4 µm). In this study, we set the wavelength of the visible and the IR beams to 610 nm and 3.3 µm (3030 cm^-1), respectively.

The visible and IR beams were collinearly aligned by a beam-combiner and introduced into a commercial upright microscope (BX51, Olympus, Tokyo, Japan). Both beams were moderately focused by a CaF₂ lens (f=100 mm) onto a sample preparation. The focal spot sizes of the IR and visible light beams were adjusted to be about 100 µm at the sample position. In this study, 1 µm fluorescent beads were used as the sample. The beads absorb green light and emit fluorescence peaked at 560 nm. The beads were placed between two cover slips after being washed with fresh water for a few times, and were sealed by nail varnish.

If the NA of the objective which collects the transient fluorescence becomes large, its working distance is limited typically to ~1 mm. The IR pulse would cause optical damage on such an objective, because part of the elongated focal spot of the IR light penetrates the collective objective, in the optical layout shown in Fig. 1(b). In the experiments reported in this paper, we used a long working distance (=0.7 cm) semi-apochromat objective (x50, NA=0.5, LMPLFLN50x, Olympus, Tokyo, Japan) to avoid this problem. The transient fluorescence collected from the sample was projected onto an intensified charge-coupled device (ICCD) camera (PI-MAX-1K-HBf, Princeton Instruments, Trenton, NJ, USA) by projection and relaying (x3.5) lenses. To remove the background signal of the excitation lasers, notch and IR-cut (HA15, HOYA, Tokyo, Japan) filters were placed behind the objective. Furthermore, we used bandpass (centered at 555 nm and band-width=100 nm), and long cut (passing ≤590 nm) filters for optimum extraction of the fluorescence signal. The power of the visible and IR lasers were ~10 nJ/pulse and 10 µJ/pulse, respectively.

4. Results and Discussions

First we briefly investigate the expected IR image obtained by conventional IR microscopy. We numerically estimated the IR image of two 1 µm diameter beads separated by 3 µm from each other (Figs. 2(a)-2(c)). The FWHM of the diffraction limit of IR light was assumed to be 3.4 µm, in agreement with our experimental parameters of λ=3.3 µm and NA=0.5. In the calculation we made the following assumptions: the microscope has a much worse axial resolution than transverse resolution; molecular density is uniform inside the beads; and the IR absorption cross-section depends linearly on the local thickness of the beads. Accordingly, the estimated effective absorption cross section of the beads has a spherical profile (strongest absorption at the center of the beads where they have the largest thickness and the absorption cross section gradually going to zero at the edges), as indicated in Fig. 2(a) and the dotted curves in Fig. 2(c). The IR image was estimated by convolving the Gaussian diffraction-limited IR point-spread-function with the effective IR absorption cross-section of Fig. 2(a). As shown in Figs. 2(b) and 2(c), the two beads cannot be resolved using a conventional IR microscope setup.

 

Fig. 2. Numerical investigation of the resolution performance of ordinary IR microscopy; (a) schematic picture of two 1 µm microbeads separated by 3 µm; (b) image of (a) using an illumination wavelength of 3.3 µm and a focusing optics with NA=0.5; (the IR absorbing area is shown by white color) (c) intensity cross-sections of (a) [dotted line] and (b) [continuous line] at y=0 µm. As shown, two objects separated by 3 µm are not resolved by an ordinary IR system.

Download Full Size | PDF

Next, we discuss the results of our IR super-resolution microscope. Figure 3(a) shows the fluorescence image of a 1 µm fluorescent bead excited by a λ=532 nm beam. At this wavelength, the dye in the bead is directly pumped to the electronically excited state, so a clear fluorescence image can be observed. When we changed the excitation wavelength to 610 nm, which corresponds to photon energy smaller than the absorption band of the dye contained in the bead, the fluorescence disappeared completely (Fig. 3(b)). Similarly, an IR beam (λ=3.3 µm (3030 cm^-1)) did not independently cause any emission from the bead either (Fig. 3(c)). However, when the two beams used for Figs. 3(b) and 3(c) were introduced simultaneously, the fluorescent bead emitted a strong signal (Fig. 3(d)), and its image was almost identical to that obtained by resonant excitation by visible light (Fig. 3(a)). The cross sectional profile of this image is shown in Fig. 3(e) (gray solid line). We numerically fitted this profile by convolving the effective fluorescence cross-section of a 1 µm diameter bead (estimated in a way similar to the effective IR absorption cross-section of Fig. 2) with a Gaussian point spread function (dashed line in Fig. 3(e)). The FWHM of the point spread function which gave the best fitted curve was 880 nm. The IR absorption of the fluorescent bead was thus imaged at sub-micron resolution.

We found the 1-color fluorescence image (Fig. 3(a)) was somewhat different from the TFD-IR image (Fig. 3(d)). We believe that this may have been caused by the fact that the microbead was excited in different regions in the two cases. That is, in the 1-color experiment, most of the excitation beam (wavelength=532 nm) was absorbed at the surface of the microbead which has a strong absorption at this wavelength (Fig. 3(a)). In contrast, in the 2-color TFD-IR experiment, the microbead was excited homogeneously over its entire volume (because the absorptions of the visible (wavelength=610 nm) and the IR (wavelength=3.3 µm nm) beams were much weaker compared to the 1-color experiment) (Fig. 3(d)). Those differences in the excited region of the microbead should make the fluorescence distribution along the Z-axis different in the two experiments. As a result, because the focal point was optimized to 2-color excitation (Fig. 3(d)), this might have slightly lowered the resolution of the 1-color image (Fig. 3(a)). Similar phenomena were observed in the experiments with microbeads which have various diameters (data not shown).

When we changed the IR wavelength the TFD-IR signal disappeared where is no IR absorption by the dye. Therefore, we can conclude that the TFD-IR signal in Fig. 3(d) represents IR absorption of the dye in the microbead. The IR wavelength dependence of the TFD-IR signal was reported in a previous paper [9].

 

Fig. 3. (a). Fluorescence image of a 1 µm fluorescent bead (excitation wavelength=532 nm). Images of the same bead excited by (b) 610 nm visible beam (off resonant), (c) 3.3 µm IR beam and (d) 610 nm visible +3.3 µm IR beam. (e) The cross sectional profile (gray solid line) along the white line in (d). The black dashed line is the theoretical fluorescence intensity calculated as the convolution of the 1 µm bead and a Gaussian point spread function with FWHM=880 nm.

Download Full Size | PDF

To estimate the two-point resolution performance of our super-resolution IR microscope, we measured a fluorescence image of two adjacent 1 µm beads (Fig. 4(a)). As seen, we can distinguish adjacent beads whose diameters (1 µm) and distance (~1.4 µm) are significantly smaller than the IR diffraction limit (~3.4 µm calculated from the IR wavelength (=3.3 µm) and the NA of the objective (=0.5)). The cross sectional profile along the white line in (Fig. 4(a)) is shown in Fig. 4(b). It shows that the two-point-separation of our microscope is ≈1.4 µm which is considerably better than that of conventional IR microscopy (see Figs. 2(b) and 2(c)).

 

Fig. 4. (a). IR super-resolution image of two adjacent beads. (b) Cross sectional profile along the white line in (a).

Download Full Size | PDF

The above results obtained by our novel IR super-resolution microscope indicate its ability to measure the IR spectra of molecular species in microscopic samples at sub-micron resolution. For example, many cellular structures in biological cells have sub-micron dimensions. Our microscope is applicable even to such targets, and is expected to provide new insights on cellular biology.

5. Conclusions

Conventional IR microscopes cannot be applied to study microscopic objects whose dimension is ~1 µm or less. In this paper, we reported the performance of our recently developed TFD-IR super-resolution microscope which was designed to solve this problem of conventional IR microscopes. The resolution (FWHM of the point spread function) of our system was 880 nm which is smaller by a factor of ~3.8 than the diffraction limit of IR light (3.4 µm). Furthermore, this microscope is able to resolve 1 µm fluorescent beads separated by 1.4 µm from each other. These results indicate that TFD-IR super-resolution microscopy can be applied to microscopic targets such as sub-cellular structures in biological cells or nano-scale industrial materials.