subwavelength spatial resolution with an atomic force microscope tip used as a photothermal sensor. A. Dazzi, R. Prazeres, F. Glotin, and J. M. Ortega.
2388
OPTICS LETTERS / Vol. 30, No. 18 / September 15, 2005
Local infrared microspectroscopy with subwavelength spatial resolution with an atomic force microscope tip used as a photothermal sensor A. Dazzi, R. Prazeres, F. Glotin, and J. M. Ortega Centre Laser Infrarouge Orsay, CLIO/LURE, Bâtiment 209D, Université Paris-Sud, BP34, 91898 Orsay Cedex, France Received March 28, 2005; revised manuscript received May 20, 2005; accepted May 24, 2005 We describe a new method of infrared microspectroscopy. It is intended for performing chemical mapping of various objects with subwavelength lateral resolution by using the infrared vibrational signature characterizing different molecular species. We use the photothermal expansion effect, detected by an atomic force microscope tip, probing the local transient deformation induced by an infrared pulsed laser tuned at a sample absorbing wavelength. We show that this new tool opens the way for measuring and identifying spectroscopic contrasts not accessible by far-field or near-field optical methods and with a subwavelength lateral resolution. © 2005 Optical Society of America OCIS codes: 110.0180, 120.6200.
The combination of infrared spectroscopy and microscopy is a technique of long-standing interest, as the spectral distribution of vibrational frequencies gives each molecular species their own infrared signature (chemical imaging). Use of the high brightness of synchrotron radiation instead of blackbody sources has already permitted the ultimate far-field resolution to be reached.1 However, lowering the lateral resolution beyond the diffraction limit is of paramount importance for numerous studies involving optical spectroscopy. For example, the size of many living cells is comparable with the wavelength in the spectral region of interest for chemical mapping 共3–20 m兲. Various infrared near-field techniques have been developed. Among them, the apertureless configuration2,3 seemed promising; most of the apetureless techniques use CO2 lasers. However, the lack of tunability of cw lasers makes the interpretation of the images difficult and greatly restricts its applicability. Several pulsed infrared lasers, particularly free-electron lasers (FELs) offer tunability. The apertureless technique is then not operating as it does with cw lasers, and various near-field techniques are employed.4–7 With the FEL CLIO,8 nearfield spectra were recorded9 by using the photon scanning tunneling microscope configuration9 but with poor spatial resolution. Moreover, in all optical methods, the characterization of very small volumes requires measurement of very tiny absorption, i.e., transmitted light, changes so small that it becomes practically impossible. In fact, theoretical studies10 show that the intensity distribution of the evanescent field differs from the spatial distribution of the imaginary part of the index of refraction. Therefore mapping at a single wavelength generally provides no chemical information, since the influence of real (topography) and imaginary (absorption) parts of the index of refraction are mixed. Indeed, complete spectra have to be recorded at each point, as is usually done in infrared far-field microscopy. 0146-9592/05/182388-3/$15.00
In order to succeed in local spectroscopy, other methods have to be envisioned. Methods measuring absorption directly rather than transmission, such as photoacoustical and photothermal techniques, seem more adapted. Recently we explored11 the possibility of working with the photothermal deflection beam (PTDB) effect. The PTDB microscope measures the deflection of a visible laser induced by irradiation of the sample. However, resolution is limited by the visible laser spot size, and a reflecting surface is needed. The new solution that we propose is to use an atomic force microscope (AFM) tip sensor. Infrared irradiation of the sample produces a local transient heating that can be detected through the corresponding surface deformation (a method that we call “AFMIR”, AFM used in the infrared). Spectroscopy with an AFM has been studied elsewhere,12 but without either spectral or spatial resolution, and studies have also been performed with a thermal sensor13 associated with an AFM: this method is sensitive to the temperature increase rather than to the surface deformation, as in our case, and does not provide subwavelength resolution. Indeed, we have to use short excitation duration and fast measurements because natural dilution of the heating would prevent any measurement or, at least, reduce the spatial resolution. Therefore we use pulsed lasers, a 10 ns CO2 laser and 10 s long FEL pulses. The lateral resolution is expected to depend on the propagation of the local deformation, i.e., on the thermal propagation on the substrate during the measurement. It should also depend on the structure of the sample and its thermal contact with the substrate. The sample is deposited on the surface of a ZnSe prism [Fig. 1(a)]. The infrared pulsed laser is sent through it in order to reach the surface supporting the sample at the total reflection angle. However, in parts covered by the sample, the laser is allowed to propagate through it and experiences total reflection only at the interface with air. On the other side of the © 2005 Optical Society of America
September 15, 2005 / Vol. 30, No. 18 / OPTICS LETTERS
Fig. 1. (a) Experimental setup. (b) Signal on the resin at wavelengths = 9.55 m and = 10.5 m of the CO2 laser. Note that the first oscillation of the pulse train, representing a time duration of about 20 s, is a parasitic signal that is due to the high-voltage discharge of the CO2 laser.
sample is the tip of the AFM. To avoid residual absorption by the tips, these are covered with a thin gold layer (60 nm). We employ a commercially available AFM (Veeco Explorer Model) possessing a fast detection: a visible laser is focused on the standard cantilever, and a four-quadrant detector measures its deflection. Infrared laser pulses, heating the sample, create a vibration of the cantilever as a result of sample deformation. The feedback of the AFM tends to counteract the signal produced by the four-quadrant detector, but the time response of the feedback loop is too slow to compensate for during the measurement for the vibration induced by the laser burst. Therefore the output signal represents the actual vibration of the cantilever, which is resonant at a frequency near 60 kHz. By using frequency analysis, we measure this oscillation amplitude and reject noise and parasitic signals. The AFM tip is used in contact mode. The force applied on the tip is close to 10 nN. Here this operating point (set point) of the AFM is adjusted to the middle of the linear part of the vertical displacement of the tip, corresponding to a deflection of 50 nm. The first sample is a layer of Epon-SU8 resin, deposited on a ZnSe prism. The resin absorption peaks at 9.7 m, and the CO2 laser is tuned either at 9.55 or 10.5 m. Imaginary parts of the refractive indices are respectively 4.9⫻ 10−2 and 2.6⫻ 10−2. The laser produces a series of pulses of up to 10 mJ during 50 ns, and its average power is made identical at both wavelengths (52 mW at 10 Hz repetition rate). It is focused on a large spot, ⬵1 mm, in order to spread over the entire region of interest. The large difference between the resulting signals (Fig. 1) shows that the amplitude of the cantilever vibration readily depends on the resin absorption value. Interestingly, these vibration curves are symmetrical with respect to the initial equilibrium position. Two conclusions are
2389
drawn from this. First, the induced deformation of the surface decreases rapidly. Otherwise, if the deformation time duration was larger than, typically, 10 s, the average value of the vibration signal would follow the deformation amplitude, or at least the oscillation would be unsymmetrical. Second, the symmetry of the vibration shows that the AFM still remains in a linear part of the operating point, even for large oscillations (30 nm), as expected. Therefore the signal amplitude is proportional to the local thermal expansion. Here the ratio of the two signals is larger by 40% than the absorbance ratio, which may be due to nonuniform heating (see below), to the remaining absorption of the metallized silicon tip at 9.5 m, or to a nonlinearity of the process. Figure 2 displays two curves as a function of the tip’s lateral position: (1) the amplitude of vibration of the cantilever, corresponding to the AFMIR absorption measurement, and (2) the topographic signal from the AFM. The absorption profile is well fitted by the topographic measurement. The contrast is very good between the resin sample and the ZnSe substrate area, with a signal-to-noise ratio of about 10. The spatial width of the resin being 1.1 m, the resolution of the measurement is of the order of 0.1 m, about 100 times smaller than the infrared wavelength 共10 m兲, well below the Rayleigh criterion for optical far-field measurements. Also, the left-hand border exhibits a larger (reproducible) deflection than the right-hand border of the object. This effect is due to nonsymmetrical illumination by the laser (incidence of 30° from the right-hand side), producing nonhomogeneous heating of the sample. The second sample consists in bacteria of Escherichia coli deposited on the ZnSe substrate. Their distribution is recorded by the AFM working in its normal mode. It can be seen (Fig. 3) that bacteria are distributed randomly on the substrate. Their lateral size is about 0.7 m and thickness about 450 nm. After a scan in contact mode, the AFM tip is easily positioned above a selected position (point A of Fig. 3),
Fig. 2. Lateral position scan of the AFM tip. The AFM topographic signal (without irradiation) and the tip vibration amplitude (under irradiation) are measured successively.
2390
OPTICS LETTERS / Vol. 30, No. 18 / September 15, 2005
Fig. 3. Contact AFM mapping of the bacteria deposited on a ZnSe substrate.
of the surface. This explains why the nonabsorbing region practically does not contribute to the signal, as the heat flows slowly compared with the rise time of the pulsed laser (10 ns for the CO2 laser, less than 1 s for the FEL). Therefore the heat diffusion has little influence on the spatial resolution, and its ultimate value should be of the order of the tip end size, i.e. a few nanometers. This is in contrast with the above-mentioned PTDB method, where the beam deflection is instead proportional to the derivative of the surface displacement versus lateral position (and not elapsed time). Therefore the PTDB is more sensitive to heat diffusion. These results demonstrate that the AFMIR can perform local spectroscopy on extremely small objects with a subwavelength spatial resolution of the order of a few tens of nanometers. This opens the way for microspectroscopy and chemical mapping of various objects, such as biological cells. The authors are grateful to Antoine Boivin (Centre de Génétique Moléculaire—CNRS-UPR 2167—Paris XI) for preparing and conditioning the bacteria for the experiments. References
Fig. 4. AFMIR spectrum of a single bacterium (solid curve) recorded at point A (Fig. 3) compared with the farfield spectrum of a stockpile of bacteria (dotted curve).
and an AFMIR spectrum of a single bacterium is recorded. The pulsed laser in this case is the abovementioned CLIO FEL, continuously tunable between 3 and 100 m, with a bandwidth of about 0.5%. The average power (for 8 s long pulses at 25 Hz) is reduced to about 50 mW to avoid damaging the sample and is also focused on a large spot (1 mm). The recorded spectrum (Fig. 4) is in complete agreement with a far-field measurement of a much larger ensemble of bacteria deposited on a substrate and with the data recorded in the literature.14 This agreement indicates that, at least in this case, the recorded signal is linear with respect to the optical absorption. It seems that the excellent lateral resolution of the AFMIR is due to the fact that the cantilever highfrequency vibration amplitude is sensitive to the speed of the surface displacement when it is hit by the laser, rather than by the displacement itself, i.e., to the derivative versus time of the vertical position
1. P. Dumas, G. L. Carr, and G. P. Williams, Analusis 1, 68 (2000). 2. R. Bachelot, P. Gleyzes, and C. Boccara, Opt. Lett. 20, 1924 (1995). 3. B. Knoll and F. Keilmann, Nature 399, 134 (1999). 4. A. Piednoir, F. Creuzet, C. Licoppe, and J. M. Ortega, Ultramicroscopy 57, 282 (1995). 5. M. Hong, A. Jeung, T. I. Smith, H. A. Schwettman, P. Huie, and S. Erramilli, Nucl. Instrum. Methods Phys. Res. B 144, 246 (1998). 6. A. Cricenti, R. Generosi, P. Perfetti, J. M. Gilligan, H. Tolk, C. Coluzza, and G. Margaritondo, Appl. Phys. Lett. 73, 151 (1998). 7. D. Palanker, G. Knippels, T. I. Smith, and H. A. Schwettman, Opt. Commun. 148, 215 (1998). 8. R. Prazeres, F. Glotin, C. Insa, D. A. Jaroszynski, and J. M. Ortega, Eur. Phys. J. D 3, 87 (1998). 9. N. Gross, A. Dazzi, J. M. Ortega, R. Andouart, R. Prazeres, C. Chicanne, J.-P. Goudonnet, Y. Lacroute, C. Boussard, G. Fonteneau, and S. Hocdé, Eur. Phys. J. Appl. Phys. 16, 91 (2001). 10. A. Dazzi, S. Goumri-Said, and L. Salomon, Opt. Commun. 235, 351 (2004). 11. W. Seidel, H. Foerstendorf, K. H. Heise, R. Nicolai, A. Schamlott, J. M. Ortega, F. Glotin, and R. Prazeres, Eur. Phys. J. Appl. Phys. 25, 39 (2004). 12. M. S. Anderson, J. Appl. Spectrosc. 54, 349 (2000). 13. A. Hammiche, L. Bozec, H. M. Pollock, M. German, and M. Reading, J. Microsc. 213-2, 129 (2004). 14. D. Naumann, in Encyclopedia of Analytical Chemistry, R. A. Meyers, ed. (Wiley, 2000), p. 102.
