Individual Detection of Single-Nanometer-Sized Particles in Liquid by

Anal. Chem. 1998, 70, 5037-5041

Individual Detection of Single-Nanometer-Sized Particles in Liquid by Photothermal Microscope Kazuma Mawatari, Takehiko Kitamori, and Tsuguo Sawada*

Department of Applied Chemistry, School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo 113-8656, Japan

We have developed a thermal lens microscope for liquidphase and surface microanalyses. By applying the thermal lens microscope to particle detection, we succeeded in detecting a pulsed photothermal signal from singlenanometer-sized particles in liquid and counting them individually. The samples were polystyrene latex particles (190 and 80 nm in diameter) and colloidal Ag particles (10 nm in diameter). To verify that the detected pulsed signals corresponded to the single-particle photothermal effects, we confirmed the items as described below using 190-nm polystyrene particles. First, no pulsed signal was generated under irradiation by either the excitation beam or the probe beam. Second, the pulse counts were proportional to the expectation value of the particles in the detection volume and zero for ultrapure water blank. Third, the pulse counts’ distribution in a series of unit times had a Poisson distribution when the expectation value of the sample was much less than 1. Then, we demonstrated counting 80-nm polystyrene particles and 10-nm Ag particles in water. The pulsed signals were clearly distinguished from noise, and the signal-to-noise ratio was as large as 5. Finally, we discussed differences between the conventional thermal lens effect and the single-particle photothermal effect. Individual nanometersized particle detection by photothermal effect was the first demonstration. Organic, metallic, and semiconductor particles, colloidal solutions and emulsions, and their chemical behaviors in liquids have become increasingly important in various fields such as environmental science,1,2 biological research,3 nano-scale materials development,4 and catalysts development.5,6 In particular, submicrometer- to nanometer-sized-particles (nanoparticles) are now attracting attention because they change chemical and physical properties such as catalytic activity and surface-enhanced Raman scattering (SERS) activity due to size and surface effects. It is desirable to know these chemical and physical properties, not as (1) Power, J. F.; Langford, C. H. Anal. Chem. 1988, 60, 842. (2) Zafriou, O. C.; Zepp, R. G.; Zika, R. D.; Joussot-Dubien, J. Environ. Sci. Technol. 1984, 18, 358A. (3) Alexander, F. Can. J. Physiol. Pharmacol. 1982, 60, 556. (4) Alivisatos, A. P. J. Phys. Chem. 1996, 100, 13226. (5) Murphy, J. C., Maclachlan Spicer, J. W., Aamodt, L. C., Royce, B. S. H., Eds. Photoacoustic and Photothermal Phenomena II; Springer Verlag: Berlin, 1990. (6) Colombo, D. P., Jr.; Bowman, R. M. J. Phys. Chem. 1995, 99, 11752. 10.1021/ac980250m CCC: $15.00 Published on Web 10/31/1998

© 1998 American Chemical Society

bulk material properties of a group of many particles but as properties of individual nanoparticles since physical properties, such as SERS activity, are quite different among particles generally.7 Averaged values of many particles are not useful to clarify the phenomena which are characteristic of nanoparticles. Therefore, spectrochemical detection methods are needed to characterize these properties in situ or in vivo. Besides these requirements, detection should be highly sensitive to allow single-particle detection in liquids. So far, single-nanoparticle spectrochemical detection methods have mainly used SERS measurements7 or a laser-induced fluorescence (LIF) method.8 However, these emission spectrochemical analyses are restricted to only the SERS-active particles or fluorescent ones. Many kinds of particles can be subjected to light absorption-based measurements of the bulk phase. Several groups have reported absorption measurements at a single-particle or emulsion level, but the minimum detectable size remains at micrometer scale because of poor sensitivity, and there are no measurements of nanometer order in liquids. As a more popular method, light scattering spectrometry has often been used for ultrafine particle counting in liquids. However, this method gives information only on whether the particle is present or not in the detection volume, and obtaining the spectroscopical, chemical, and physical information is difficult. Theoretically, the minimum detectable size is predicted to be about 80 nm or so, based on the Rayreigh ratio of the light scattering cross section of a particle to the solvent background. Photothermal techniques seem to be promising ways to overcome these problems because they are highly sensitive methods for monitoring the energy flow of heat resulting from absorption of optical irradiation, and photothermal phenomena are based on the most common relaxation process following absorption of light. For example, these techniques include photoacoustic spectroscopy (PAS), optical beam deflection (OBD), and thermal lens spectroscopy (TLS). Their advantages of high sensitivity and wide applicability have been utilized in the analyses of nonfluorescent molecules or particles in liquid and solid.9,10 Several unusual characteristics of photothermal signals were reported; in particular, as the diameter of the particle decreased from 300 to (7) Nie, S.; Emory, S. R. Science 1997, 275, 1102. (8) Kim, H.-B.; Hayashi, M.; Nakatani, K.; Kitamura, N.; Sasaki, K.; Hotta, J.; Masuhara, H. Anal. Chem. 1996, 68, 409. (9) Waldron, K. C.; Dovichi, N. J. Anal. Chem. 1992, 64, 1396. (10) For the latest review, see: Bialkowski, S. E. Photothermal Spectroscopy Method for Chemical Analysis; John Wiley & Sons: New York, 1996.

Analytical Chemistry, Vol. 70, No. 23, December 1, 1998 5037

25 µm, the sensitivity of the OBD signal increased due to the larger temperature gradient around the particle in air.11 Also, in liquid, 40-nm polystyrene particle counting was conducted using laser-induced breakdown acoustic spectroscopy.12,13 However, the OBD technique is still lacking in sensitivity for nanoparticle counting in liquid, and laser-induced breakdown acoustic spectroscopy is not a nondestructive method. Therefore, at present, no researchers have succeeded in counting single nanoparticles nondestructively in liquid. For the purpose of ultratrace and ultramicro analyses in liquids and on solid surfaces, we have developed a thermal lens microscope (TLM).14 In our TLM, coaxial excitation and probe beams are introduced to an optical microscope, and they are tightly focused in/on a sample. Irradiation of the excitation beam induces the thermal lens effect in the sample under an objective lens. From the confined optical configuration using chromatic aberration of the objective lens, the probe beam passing through the sample is converged by the concave thermal lens effect, and the thermal lens signal is detected by measuring the probe beam intensity change when passed through a pinhole. The signal amplitude is proportional to the absorbance of the sample and excitation beam power, and it also characterizes the thermal properties of the sample. In thermal lens theory, tight focusing of the excitation beam enhances the signal amplitude15 because the temperature gradient becomes large. Therefore, the TLM should be a very sensitive tool. In fact, we applied the TLM to ultrasensitive detection in liquids and demonstrated that it was possible to detect subzeptomole (10-21 mol) dye molecules in water. In this paper, we applied the TLM to particle detection, and, for the first time, we demonstrated counting of single nanoparticles in liquid using the photothermal effect. We verified that the detected pulsed signals were generated by the photothermal effect of individual single particles. Polystyrene particles were used for the basic experiments, and we also successfully counted 10 nm Ag colloidal particles. When we apply the TLM to the counting of nanoparticles, some essential differences in the signal generation mechanism can be anticipated in comparison with the conventional thermal lens effect. We discuss these and the signal generation mechanism. EXPERIMENTAL SECTION Reagents and Samples. Monodispersed polystyrene latex particles (190 and 80 nm, Japan Synthetic Rubber Co., Ltd.) were used as standard samples. These polystyrene particles include dye molecules uniformly for absorbing the excitation beam in the visible laser radiation line used. They have a carboxyl type surface and are well known to be highly monodispersed. The coefficient of variation (CV) of the diameter was less than 5%, which was determined by transmission electron microscopy. Stock solutions at several number densities were prepared by stepwise dilution with ultrapure water. These were filtered to remove particulate impurities larger than 0.2 µm before the measurement. The trace (11) Wu, J.; Kitamori, T.; Sawada, T. J. Appl. Phys. 1991, 69, 7015. (12) Kitamori, T.; Yokose, K.; Sakagami, M.; Sawada, T. Jpn. J. Appl. Phys. 1989, 28, 1195. (13) Kitamori, T.; Matsui, T.; Sakagami, M.; Sawada, T. Chem. Lett. 1989, 2205. (14) Harada, M.; Shibata, M.; Kitamori, T.; Sawada, T. Anal. Chim. Acta 1995, 299, 33. (15) Weimer, W. A.; Dovichi, N. J. J. Appl. Phys. 1986, 59, 225.

5038 Analytical Chemistry, Vol. 70, No. 23, December 1, 1998

Figure 1. Schematic illustration of the photothermal measurement of a nanoparticle in liquid microspace, including a sectional view around the focal point of the excitation beam and probe beam in the microchannel. The chromatic aberration between the focal point of the excitation beam and that of the probe beam was 4 µm.

turbid solutions were sufficiently degassed by an aspirator. The number density of the stock samples was determined from a calculation based on specific gravity of 1.1 g/mL for the particle and 1.0 g/mL for the solution. The number density was 2.5 × 1016 L-1 for the 190-nm particle and 3.6 × 1017 L-1 for the 80-nm ones from the ratio of total weight to single-particle mass. The Ag colloidal solutions were prepared according to the published method.16 The number density of their stock sample was calculated as 4.8 × 1015 L-1. Apparatus. The TLM system used was the same as described previously.13 In brief, the excitation beam was the 488-nm lasing line of an Ar laser, and the beam power was 2 mW at the sample. The excitation beam was modulated at 4 kHz with a light chopper. The probe beam was the 633-nm lasing line of a He-Ne laser, and the beam power was 0.1 mW. The numerical aperture (N.A.) of the objective lens (Olympus Optical Co., Ltd., Japan) was 0.85, and therefore the beam waist of the excitation beam was calculated as about 1 µm. The excitation beam which passed through the pinhole was spatially separated from the probe beam using a diffraction grating before being introduced to the detection part. The detector of the probe beam was a photodiode, and the electrical signal was fed into a low-noise preamplifier (×100). In all measurements, a lock-in amplifier (LI-574, NF Corp., Japan) was used as an ultranarrow band-pass filter and noise reduction tool, which differed from the usual harnessing methods for lockin amplifiers. The time constant of the lock-in amplifier was set at 10 ms which determined the temporal resolution power of the pulsed signal and influenced the signal amplitude. This is described later. All the results were recorded on a chart recorder. The sample cell used was a microchannel fabricated in a glass microchip.17 An expanded sectional view around the detection area is schematically illustrated in Figure 1. The cross section of the microchannel was 150 µm wide and 100 µm high. The sample was introduced to the microchannel as follows. First, the microchannel was filled with ultrapure water, and a certain portion of the sample was dropped on the sample reservoir of the glass (16) Roberti, T. W.; Smith, B. A.; Zhang, J. Z. J. Chem. Phys. 1995, 102, 3860. (17) Sato, K.; Kitamori, T.; Sawada, T. Anal. Chem. submitted.

Figure 3. Dependence of the pulse counts during a 2-min measurement on the expectation value of the particle number. The sample was 190-nm polystyrene trace turbid solution. Figure 2. Typical monitored signals for the trace turbid solution of 190-nm polystyrene particles with irradiation of (a) both excitation beam and probe beam, (b) probe beam only, and (c) excitation beam only. The expectation value in the detection volume was 7.5 × 10-4.

microchip. Then, more water was sucked up from another reservoir, and the microchannel was completely filled with the sample. After the glass chip was left for a few minutes to equilibrate, the photothermal effect was measured at the center of the microchannel both vertically and horizontally. The detection volume of the thermal lens signal generation area was confirmed to be 3 fL on the basis of thermal lens measurement of a carbon thin film scanned in the z-direction (normal to the sample plane). The expectation value of the particle number was determined from the product of the detection volume and the number density of the sample. A chromatic aberration occurred at 4 µm between the focal point of the excitation beam and that of the probe beam, which is well known to enhance the thermal lens signal strongly.18 RESULTS AND DISCUSSION First of all, the photothermal signal of the 190-nm polystyrene turbid solution was measured. The results are shown in Figure 2a. The expectation value of the sample was 7.5 × 10-4, much smaller than 1. Measurement period was 60 s. Pulsed signals were observed when both the excitation beam and the probe beam irradiated the sample. To confirm that the pulsed signals were surely generated by the photothermal effect, we measured the same sample under the irradiation of the probe beam and the excitation beam only, and these results are shown in Figure 2b,c. Pulsed signals might also be generated by light scattering of the particles because photothermal microscopy monitors the intensity change of the probe beam when it is passed through the pinhole. However, no pulsed signal was observed for either of these cases. Therefore, the pulsed signals of Figure 2a were caused by neither the probe beam scattered by the particles nor the stray light of the scattered excitation beam. Furthermore, the average pulse height and the excitation beam power were confirmed to be proportional from 0.5 to 2 mW. Therefore, we concluded that the observed pulsed signals were caused by the photothermal effect of the particles. Second, we needed to confirm whether the pulsed signal corresponded to an individual particle, that is, whether the pulsed (18) Berthoud, T.; Delorme, N.; Mauchien, P. Anal. Chem. 1985, 57, 1216.

signals were generated from a single particle. Then we investigated the dependence of the pulse counts during the measurement period on the expectation value of the sample. The results are shown in Figure 3. The number density of the sample solution was converted to the expectation value of the particle number in the detection region, which is the confocal region according to the thermal lens theory. The expectation value ranged from 4.7 × 10-5 to 1.5 × 10-3. The measurement period was 2 min, which was long enough to obtain statistically reliable values because particles go into and out of the detection volume (3 fL) in a few hundred milliseconds according to the Stokes-Einstein equation. As is shown in Figure 3, the measured pulse counts exhibited a relationship proportional to the expectation values of the sample. Also, no pulsed signal was obtained for the ultrapure water blank. These results also strongly supported the earlier conclusion that the detected pulsed signals represent the photothermal effect of the individual polystyrene particles. Furthermore, when the expectation value of the particle number is much less than 1, it is predicted from statistical theory that the pulse counts’ distribution should be a Poisson distribution. Next, the number of pulses in each unit time was counted, and the distribution of the pulse counts in a series of unit times was compared with the Poisson distribution. The sample was 190-nm polystyrene particles, and the expectation value was 7.5 × 10-4. The procedures for the experiment and data analysis were as follows. First, the results of a 2-min measurement were divided into consecutive 1.6-s periods; that is, the unit time was set at 1.6 s. Next, the number of pulses was counted for each period. Then, the frequency of the time unit with the same pulse counts was obtained. Finally, the frequency versus the pulse counts was plotted as in Figure 4. For comparison, the theoretical Poisson distribution is plotted as the solid line. The theoretical expectation value was normalized by the total pulse counts from the experimental data. The frequency distribution of the pulse counts in unit time was well fitted with the Poisson distribution. From the results of Figures 3 and 4 concerning the pulse counts, we concluded that the pulsed signals were generated from individual particles, and we anticipated that the pulsed signals were caused by momentarily generated photothermal effects when the individual particles happened to pass through the detection volume due to Brownian motion. The pulse heights were quite different, and this might be due to variation of the pathway in the detection volume. As for the optical trapping force, 190-nm polystyrene particles are very small, and the maximum power of the excitation Analytical Chemistry, Vol. 70, No. 23, December 1, 1998

5039

Figure 4. Comparison of the pulse counts’ distribution in a series of unit times with the Poisson distribution. The sample was 190-nm polystyrene trace turbid solution with an expectation value of 7.5 × 10-4. The measurement period was 2 min, divided into consecutive 1.6-s periods (unit time). The frequency of the time units with the same pulse counts was plotted against the pulse counts. The expectation values of the theoretical Poisson distribution (solid line) was normalized by the total pulse counts in the experimental data.

Figure 5. Monitored signals for the trace turbid solution of 80-nm polystyrene particles. The expectation values were (a) 8.7 × 10-4, (b) 1.8 × 10-4, and (c) 3.6 × 10-5. (d) Ultrapure water (blank).

beam (2 mW) is not large enough to trap submicrometer-sized polystyrene particles in water, so the optical trapping force can be ignored in these experiments. Furthermore, we tried to measure smaller particles in liquid. The results of monitoring the 80-nm polystyrene solution are shown in Figure 5. The detection conditions were the same as those in the experiment for 190-nm particles. The pulsed signals were identified from the noise level, and the pulse counts had the same characteristics as discussed above for the 190-nm particles. The Ag particles (10 nm average size) were also successfully counted, and the results are shown in Figure 6. The pulsed signals could be identified clearly. The excitation wavelength was set at 488 nm, which differed from the absorption maximum wavelength of 390 nm of Ag particles. However, Ag particles exhibit broad light absorption due to the surface plasmon band16 and the absorption cross section was about 1 order larger than that of the 80-nm polystyrene particle, even at 488 nm. From 5040 Analytical Chemistry, Vol. 70, No. 23, December 1, 1998

Figure 6. Result from monitoring the average diameter in 10-nm Ag colloid solution. The expectation value in the detection volume was 1 × 10-4 (a). (b) Ultrapure water (blank).

these results, we could demonstrate counting of individual nanoparticles in liquid by the photothermal effect. Next, we discuss the influence of the system response on the average pulse height. In the experiments shown above, a lock-in amplifier was used as a narrow band-pass filter around 4 kHz (modulation frequency) and noise reduction tool. As a result, the system response was governed by the time constant of the lockin amplifier, and the value was set as small as 10 ms. As in conventional thermal lens measurements, the time constant was set large in order to improve the signal-to-noise ratio. In the case of single-nanoparticle detection, however, the mean period for which particles stay in the detection volume, namely the staying time, is a few tens to hundreds of milliseconds order, which differs with the diameter of the particle. Therefore, the smaller value of the time constant is more suitable to detecting pulsed signals with high signal-to-noise ratio by the lock-in amplifier. The average pulse height of the 10-nm Ag colloid solution in Figure 6 was comparable to that of the 80-nm polystyrene turbid solution in Figure 5, though the absorption cross section of the Ag particles was 10 times larger at 488 nm. This might be due to the difference of the staying time in the detection volume as determined from the Stokes-Einstein equation and insufficient response time of the lock-in amplifier for the 10-nm Ag colloid solution. Details of the theoretical analysis will be discussed later in another report. Finally, we discuss the differences between the conventional thermal lens effect in liquid with the observed single-particle photothermal effect. The conventional thermal lens effect assumes that the sample is homogeneously distributed in the solution and the intensity distribution of the excitation beam induces a temperature distribution, resulting in a refractive index distribution in the heating area which acts as a lensing effect. However, for a single nanoparticle, the particle was much smaller than the beam waist, and the particles should work as a moving point heat source. Therefore, the refractive index distribution is independent of the intensity distribution of the excitation beam. This situation is considered to a momentary photothermal deflection by the moving point heat source in the submicrometer space rather than the thermal lens effect. As evidence of this, the signal generation area of our experiments was somewhat larger than that expected from thermal lens theory. From the random walk theory with the Stokes-Einstein equation, the staying time can be calculated as about 200 ms for 190-nm particles. Then, the

total staying time during the measurement period can be calculated from the production of the total pulse counts and the staying time of 200 ms. The ratio of this value to the measurement period, 120 s, must be smaller than 1, but the results are 2 orders larger than the detection volume determined from thermal lens measurement of a thin carbon film. This might be attributed to the detection area being 2 orders larger for the volume than 3 fL for the thermal lens measurement. This result shows that the singleparticle photothermal effect did not obey the theory of the thermal lens effect. The deviation from the theoretical Poisson distribution line seen in Figure 4 also suggests this. The particle acts as a moving point heat source of nanometer scale, and the temperature gradient around the particle enhanced the photothermal effect, as is seen in localization of the temperature field on the particle and enhancing the photothermal deflection effect.11 When the discussion is valid, the photothermal signal can be generated outside the confocal volume, and the detection volume can become

larger, which is consistent with the experimental results. This may be a unique and interesting photothermal effect characteristic of single nanoparticles in liquid. In conclusion, we observed the new photothermal effect of individual nanoparticles in liquid for the first time, and we successfully counted nanometer-sized polystyrene particles. This technique is expected to become a powerful and widely used analytical tool for nanoparticles in liquid. ACKNOWLEDGMENT We would thank Dr. Takeda of Japan Synthetic Rubber Co., Ltd. for providing the dye containing polystyrene particles. This work is supported by a Grant-in-Aid for Specially Promoted Research (No. 07102004) from the Ministry of Education, Science and Culture of Japan. Received for review March 5, 1998. Accepted September 8, 1998. AC980250M

Analytical Chemistry, Vol. 70, No. 23, December 1, 1998

5041