We use an interferometric detection scheme to directly detect single gold nanoparticles with a diameter as small as 5 nm in an aqueous environment. We demonstrate both confocal and wide-field detection of nanoparticles and study signal strength as a function of particle size. Furthermore, we demonstrate a detection speed up to 2 μs. We also show that gold nanoparticles can be readily distinguished from background scatterers by exploiting the wavelength dependence of their plasmon resonances. Our studies pave the way for the application of this detection scheme for particle tracking in biological systems.
© 2006 Optical Society of America
Monitoring microscopic dynamical processes by tracking an optical label has a broad range of applications. In particular, particle tracking experiments have contributed a great deal to our understanding of biological phenomena such as transport in cells  and membranes  and motion of motor protein complexes . In most of these applications, the smaller the label, the greater the chance that its motion is unhindered. In principle, single fluorescent molecules satisfy this requirement in an ideal manner [4, 5]. However, photobleaching limits the observation time of single dye molecules to about one minute. Furthermore, integration times longer than several milliseconds are typically necessary and limit the time resolution in many experiments. Luminescent semiconductor nanocrystals, also known as quantum dots, offer a much better photostability, but they have two other key disadvantages. First, they show strong blinking, which results in missing segments in the particle trajectory. Secondly, these nanoparticles assume a diameter as large as 10–20 nm once their surfaces have been passivated to make them biocompatible .
A promising alternative approach is to use gold nanoparticles as optical labels [7, 8]. Gold nanoparticles are inert and have been used in biological electron microscopy applications for a long time . Furthermore, as opposed to fluorescent material, they do not degrade under laser illumination. Optical detection of single gold nanoparticles in the range of several tens of nanometers has been demonstrated via scattering in the dark-field or total internal reflection illumination. These techniques minimize the background scattering from the surrounding medium [8, 10]. However, as one decreases the particle diameter d, detection of the nanoparticles becomes increasingly difficult because the scattered intensity drops as d 6 . Recently, efforts have been undertaken to address this issue. An effective strategy has been to make use of the particle absorption, which scales only as d 3 rather than d 6 . One can either directly record the absorption  or measure the refractive index change in the heated vicinity of an absorbing particle [13, 14]. Detection of very small particles, down to 5 nm in the former case and 1.4 nm in the latter, has been achieved.
In a recent publication we reported on an alternative method for direct detection of gold nanoparticles via scattering, exploiting the interference between the background reflection and the scattered field . Let us consider Ei to be the incident electric field at the location of the sample, Er the electric field of the light reflected from the sample (i. e. background light) and Es the electric field of the light scattered from the particle. The measured intensity Im can be then written as
where r is the field reflectivity, and s = |s|eiψ is the complex scattering amplitude with phase ψ. The first term on the right denotes the background intensity whereas the second term represents the scattered intensity, which becomes smaller than the noise of the first term for very small particles. The third term, however, is proportional to d 3 and overwhelms the purely scattered light for very small particles. As in a homodyne or heterodyne detection scheme, the signal is amplified by a reference beam, in this case the background reflection. We point out that, because the cross term is proportional to r, it scales linearly with the shot noise of the background intensity, so that for shot noise limited illumination the detection sensitivity remains independent of the reflectivity of the interface.
In our previous work, we presented the principle of this detection scheme for single particles down to a diameter of about 5 nm in a confocal arrangement . By using a quasi-continuum focusable white light source, we recorded plasmon spectra of individual nanoparticles. The substrate supporting the gold particles was index-matched using immersion oil, minimizing the contribution of the background reflection. In the present article, we demonstrate that our scheme can be extended to detection of particles, 1) in aqueous media, 2) under illumination by monochromatic laser light, 3) in a wide-field arrangement, and 4) with an unprecedented time resolution. Moreover, we demonstrate that by taking advantage of the dispersion of the particle plasmon resonance, gold nanoparticles can be efficiently distinguished from background scattering.
The experimental setup is sketched in Fig. 1. The samples were preprepared by spin coating a dilute solution of gold nanoparticles (British Biocell) onto a microscope cover glass. A droplet of water was placed on the sample and contained by a small teflon cell. For illumination we used an intensity stabilized laser (Nd:YAG, CrystaLaser) with a wavelength λ = 532 nm near the peak of the plasmon resonance of a typical gold nanoparticle. The power delivered to the sample was on the order of a few milliwatts. Two plano-convex lenses were used in a telescope configuration to match the beam diameter to the aperture of the microscope objective (NA= 1.4, Zeiss). The experiments were performed on a commercial inverted microscope (Zeiss Axiovert 100), which had been modified to house a piezoceramic motion stage (Tritor 38 3D, Piezosystem Jena) to scan the sample in the focus of the objective. To avoid total internal reflection at the glass-water interface, the microscope objective was slightly underfilled to reduce the numerical aperture. The optical signal was collected in reflection, passed through a 50% beam splitter and focused by the microscope tube lens onto a variable diameter confocal pinhole. A photomultiplier served to detect the signal. For fast measurements we used a fast photomultiplier (Hamamatsu, H5783) in combination with a low-noise current-voltage amplifier (Stanford Research, band width limited to 1MHz). For wide-field experiments a power of up to 10mW was focused at the back focal plane of the objective, leading to collimated illumination of the sample. In these experiments the pinhole was opened to a range where it just blocked the stray light and the beam was re-collimated before being imaged with an infinity-corrected camera lens (80–200mm, Nikon) onto a charge-coupled device camera (PixelFly, PCO).
3. Confocal detection
Figure 2 shows scanning confocal microscope images from Au particles of 5, 10 and 20 nm diameter at the water-glass interface. Cross-sections of the respective images represent a normalized intensity σ(d) which we define as:
where Im(d) is the intensity measured from a particle of diameter d at the center of the focus, and Ir is the intensity of the reflected light without the particle, as determined from the average background in the same scan. For very small particles the pure scattering term in eq. (1) can be neglected so that σ(d) is proportional to the scattering amplitude s. Particles appear dark against the background due to the destructive interference between the scattered and reflected fields caused by the Gouy phase in the reflection of the focused incident beam.
The solid lines show theoretical fits using a dipole scattering model that includes radiation damping. The full symbols in Fig. 3 display the normalized intensity as a function of particle diameter at the water-glass interface. The solid line shows a theoretical fit to a dipole scattering model that includes radiation damping . As pointed out earlier, the signal contrast is negative due to the destructive interference between the scattered and reflected fields. This behavior is similar to the one observed at an oil-glass interface previously , but the particle diameter for which the contrast reversal occurs is shifted toward larger values due to a higher reflectivity at the interface (see Eq. (1)).
When studying single nanoparticles, it is important to verify that one does not detect aggregates. Since the resolution of the optical microscope is limited to about 200 nm in our case, we examine the size of the optical signal instead of its spatial distribution. The bottom row in Fig. 2 shows histograms of the signal σ(d) for three different particle sizes. The widths of the histograms are less than twice the value of the peak signal, showing that we indeed detect single particles. There are, however, also some particles that deliver a considerably larger signal at the tail of the histogram distribution. Here we note that an uncertainty of only 1–2 nm in the particle diameter leads to a variation in s of a factor of 2. Our previous electron microscopy measurements revealed an actual size distribution between 5 and 8 nm for particles with a nominal diameter of 5 nm .
As seen in the histogram of Fig. 2(c), a 5 nm particle causes a normalized intensity change of around 0.3%. To detect this small signal, a correspondingly low noise illumination and detection scheme is required. The intensity of our laser source was stable to about 0.1%. However, after coupling into an optical fiber active stabilization was required to reduce additional intensity fluctuations. Of course, the signal-to-noise ratio can be improved by reducing the speed of the confocal scan and thereby increasing the integration time at each pixel. The images shown in Fig. 2 were acquired using a scan speed of 0.2 s/line. We point out that the fluctuations in the backgrounds of images in Figs. 2(b) and (c) are due to light scattering by irregularities in the glass cover-slip. Repeated scans of the same sample produced exactly the same features, indicating they are not caused by laser intensity or detection noise. We will return to this issue later.
In order for the signal to stay well above the detector noise level, we typically used laser powers of the order of several milliwatts. To verify that photothermal effects [13, 14] do not influence our measurements, we have varied the illumination intensity. Figure 4 shows confocal images of the same nanoparticles imaged with 10mW and 0.2mW incident intensity. The line profiles in (c) display the same normalized intensity in both cases. We therefore infer that our contrast mechanism does not stem from thermal effects.
4. Fast detection
In tracking experiments, it is generally desirable to study the motion of a particle with as high of a time resolution as possible to ensure that one does not oversee processes and interactions that take place at very short time scales. The difficulty is that increasing the time resolution results in the reduction of the detected signal per unit time. In principle, this can be overcome by increasing the excitation intensity. In case of single fluorescent molecules, however, the integration times cannot be much shorter than a few milliseconds because of fluorescence saturation. In case of gold nanoparticles, on the other hand, there is no fundamental limitation for the detection speed because the scattered power remains proportional to the incident intensity. Indeed, this feature has been exploited in the recent work on protein diffusion in biological membranes where fast detection of gold nanoparticles with diameter of 40 nm has been achieved down to a time resolution of about 25 μs, allowing scientists to visualize short lived nano-scale confinements [18, 19]. Here we report first results on the detection of gold particles as small as 10 nm in a time window of about one microsecond.
To mimic the fast motion of a nanoparticle, we scanned the focused laser beam across the particle at a speed of 0.2 μm/μs, using a galvo-driven mirror. The curves in Figs. 5(a) and (b) show the signals obtained from single 30 nm and 20 nm particles, respectively, under laser power of about 2mW. The particles are clearly seen within a passage time of about 2 μs. We also tried this experiment for particles of diameter 10 nm, but the signal was not sufficient to detect single particles. Thus, we increased the illumination power to about 30mW, in which case even a 10 nm gold particle could be easily detected (see Figs. 5(c) and (d)). These results show the promise of gold nanoparticle for ultrafast tracking applications. We point out that time resolutions even below a microsecond should be easily achievable using higher illumination powers, but depending on the applications and susceptibility fo the sample to thermal damage, one might have to apply pulsed excitation to reduce heating.
5. Wide-field detection
In the preceding sections we have demonstrated the detection of gold nanoparticles using confocal microscopy. However, many applications prefer wide-field detection because it allows simultaneous observation of several objects within large areas. Here we demonstrate that our interference-based detection scheme can also be implemented in wide-field microscopy.
The image quality and measurement sensitivity in wide-field detection are governed by different factors than in the confocal case. In confocal scanning, temporal noise in the illumination light translates into spatial noise in the image because pixels are acquired sequentially in time. In wide-field studies the entire image is acquired simultaneously so that laser noise leads to fluctuations of the overall image intensity. On the other hand, spatial inhomogeneities in the reflected light result in unwanted features in the image. To overcome this kind of spurious contrast, we used a piezo to modulate the sample position by 500 nm between two subsequent images. Since the background contrast is independent of the lateral translation of the sample, the two images can be decomposed into a stationary background image and a displaced image containing the particles. To do this, we used an iterative numerical algorithm that started with the average of the two images as an initial guess for the stationary background. The difference between the first image and the background was decomposed into a component that correlated with the piezo displacement and one that did not. The former was taken as the current approximation for the signal and the latter was used to correct the background image for the next iteration step. The corrected background image went toward zero as the algorithm converged.
Figure 6 shows images and line profiles of gold nanoparticles with diameters of 15 and 20 nm. The signal-to-noise ratio is comparable to that of the confocal detection. Some of the small residual features that persist in the background were due to diffraction of the illumination by the objective edges in this preliminary experiment. More careful control of the illumination field and wavefronts should minimize these effects.
6. Distinction from background scatterers
A major issue in particle tracking is to distinguish the signal of the gold nanoparticle from that of other scatterers in the sample. Indeed, as we pointed out earlier, even very small corrugations or local modulation of the refractive index in the glass substrate are picked up in Figs. 2 and 4. Observation of these small variations in the optical contrast is, on the one hand, a sign of the high sensitivity of our detection scheme but, on the other hand, complicate identification of the tiny gold nanoparticles. The plasmon resonance in gold nanoparticles [11, 17] provides a convenient way of discriminating against the background scattering, which is typically wavelength independent. As shown in Fig. 7(a) for a 20 nm gold particle in water, the scattering cross section is about 2.5 times larger at λ = 532 nm as compared to λ = 488 nm. The scattering properties of most dielectrics and biological media, however, do not vary over this small spectral range.
As a second laser beam, we used light from an argon-ion laser (Innova 300, Coherent) at a wavelength of λ = 488 nm. In the following we will refer to the frequency-doubled Nd:YAG beam as the “green” beam and the argon-ion light as the “blue” beam. For most of the two-color experiments, both beams were coupled into a single-mode fiber in order to overlap the two modes. A drawback of the fiber was that it introduced intensity variations caused by thermal fluctuations, leading to variations of its polarization and transmission properties.We accounted for these intensity changes by taking a small fraction of the light from each laser beam behind the fiber as a reference.
The empty symbols in Fig. 3 display the normalized signal at λ = 488 nm for different particle sizes whereas the solid blue line shows a fit. We find that the blue signal is about 1.5× smaller than the green signal. Remembering that the signal is proportional to the scattering amplitude, this is in agreement with the prediction of Fig. 7(a) that the scattering intensities at the two wavelengths differ by about a factor of 2.5.
To demonstrate the discrimination of unwanted scattering in a two-color detection scheme, we used microtubules (about 25 nm in diameter) that glide on a kinesin-coated surface as described by Hess et al. . In our experiments, microtubules are assembled from biotinfunctionalized tubulin (Cytoskeleton, Denver, USA) and move with an average speed of ~ 100 nm/s (0.01mM ATP) on glass surfaces functionalized with kinesin motors. Anti-biotin coated gold nanoparticles (40 nm diameter, British Biocell, UK) added to the assay bind via antibody-biotin interaction to the biotinylated microtubules. Before measurement, the system was rinsed with buffer containing microtubule stabilizing additives (taxol, oxygen scavengers) as reported in more detail previously . Figures 7(b) and (c) show simultaneously acquired images of a microtubule (appearing as a strand in the center) with four gold particles attached to it, under illumination at wavelengths of λ = 532 nm and 488 nm, respectively. Note that both the gold nanoparticles and the microtubule can be seen directly in both images, but the particle contrast at λ = 532 nm exceeds the one at λ = 488 nm. To isolate the gold nanoparticles from other scatterers in the sample, we subtract the normalized signal at both wavelengths and plot the result in Fig. 7(d). As expected, the contrast of the microtubule disappears while the gold particles remain visible. Furthermore, we note that other residual features of the sample also disappear in the spectral difference image, meaning that these are not unbound gold particles but rather dielectric material such as protein agglomerations.
As pointed out previously, the background fluctuations in the images of 5 nm gold particles shown in Fig. 2 are dominated by scattering from inhomogeneities in the surface of the glass cover slip. Since the two-color experiment described above allows us to distinguish gold from dielectric scatterers, we can also employ this technique to improve the detection sensitivity for 5 nm gold particles, which was partly limited by the background inhomogeneities in the surrounding in Fig. 2(c). Figures 8(a) and (b) show the normalized intensities measured at λ = 488 nm and λ = 532 nm, respectively. Both images have been Fourier low-pass filtered to remove pixel noise and Fourier high-pass filtered to remove intensity variations due to slow drifts in laser intensity or sample position. Furthermore, due to the low signal-to-noise ratio in the detection of 5 nm particles, here we stabilized the intensity of the blue beam separately and overlapped it with the green beam without the use of a fiber to avoid unnecessary laser intensity noise. There is a clear resemblance between the background structures of the two images. In addition, Fig. 8(b) displays features that are not identifiable in Fig. 8(a). However, as shown by the white arrows, many of these are too faint to be distinguished from the background modulations. Figure 8(c) displays the difference in normalized intensity between (a) and (b). The signal-to-background ratio is significantly improved, allowing us to detect particles that are slightly smaller than those directly visible in Fig. 8(b) (see e. g. arrows). We point out that compared to the example in Fig. 7(c), in this case the background was not fully eliminated. We believe this is due to a slight lateral displacement between the blue and green laser spot.
We have demonstrated that the optical detection scheme described by Lindfors et al. , which was initially demonstrated in immersion oil covered samples using white-light illumination, can also be implemented using a single wavelength light source and allows detection of particles as small as 5 nm even at a water-glass interface. The detection sensitivity is high enough that even tiny variations of the optical contrast in the cover glass could be detected. In this light, our method could be also extended to the detection of dielectric nano-objects such as biological macromolecules in controlled environments. Moreover, we have demonstrated that single gold nanoparticles can be detected in very short times of the order of a microsecond. Furthermore, we showed that the detection can be done not only in a scanning confocal arrangement but also in a wide-field microscope, enabling simultaneous observation of many particles in large systems. Finally, the wavelength dependence of the plasmon resonance allows us to distinguish between gold nanoparticle labels and background scatterers. Our experiments address all issues that are necessary for robust and ultrafast optical detection and tracking of gold nanoparticles in biophysical systems.
This work was supported by the ETH Zurich and the Swiss Ministry of Education and Science (EU IP-Molecular Imaging).
References and links
4. G. Seisenberger, M. U. Ried, T. Endress, H. Buning, M. Hallek, and C. Bráuchle, “Real-time single-molecule imaging of the infection pathway of an adeno-associated virus,” Science 294, 1929–1932 (2001). [CrossRef]
5. A. Yildiz, J. N. Forkey, S. A. McKinney, T. Ha, Y. E. Goldman, and P. R. Selvin, “Myosin V walks hand-overhand: Single fluorophore imaging with 1.5-nm localization,” Science 300, 2061–2065 (2003). [CrossRef]
6. X. Michalet, F. F. Pinaud, L.A. Bentolila, J.M. Tsay, S. Doose, J.J. Li, G. Sundaresan, A. M. Wu, S. S. Gambhir, and S. Weiss, “Quantum dots for live cells, in vivo imaging, and diagnostics,” Science 307, 538–544 (2005). [CrossRef]
7. C. J. Cogswell, D. K. Hamilton, and C. J. R. Sheppard, “Color confocal reflection microscopy using red, green and blue lasers,” J. Microsc. 165, 103–117 (1992). [CrossRef]
8. S. Schultz, D. R. Smith, J. J. Mock, and D. A. Schultz, “Single-target molecule detection with nonbleaching multicolor optical immunolabels,” Proc. Natl. Acad. Sci. U.S.A. 97, 996–1001 (2000). [CrossRef]
10. C. Sönnichsen, S. Geier, N. E. Hecker, G. von Plessen, J. Feldmann, H. Ditlbacher, B. Lamprecht, J. R. Krenn, F. R. Aussenegg, V. Z.-H. Chan, J. P. Spatz, and M. Möller, “Spectroscopy of single metallic nanoparticles using total internal reflection microscopy,” Appl. Phys. Lett. 77, 2949–2951 (2000). [CrossRef]
11. C. F. Bohren and D. R. Huffman, Absorption and Scattering of Light by Small Particles (Wiley Interscience, New York, 1983).
12. A. Arbouet, D. Christofilos, N. Del Fatti, F. Vallée, J. R. Huntzinger, L. Arnaud, P. Billaud, and M. Broyer, “Direct measurement of the single-metal-cluster optical absorption,” Phys. Rev. Lett. 93, 127401 (2004). [CrossRef]
14. S. Berciaud, L. Cognet, G. A. Blab, and B. Lounis, “Photothermal heterodyne imaging of individual nonfluores-cent nanoclusters and nanocrystals,” Phys. Rev. Lett. 93, 257402 (2004). [CrossRef]
15. K. Lindfors, T. Kalkbrenner, P. Stoller, and V. Sandoghdar, “Detection and spectroscopy of gold nanoparticles using supercontinuum white light confocal microscopy,” Phys. Rev. Lett. 93, 037401 (2004). [CrossRef]
16. T. Kalkbrenner, U. Håkanson, and V. Sandoghdar, “Tomographic plasmon spectroscopy of a single gold nanopar-ticle,” Nano Lett. 4, 2309–2314 (2004). [CrossRef]
17. U. Kreibig and M. Vollmer, Optical Properties of Metal Clusters (Springer Series in Material Science, Berlin, 1995).
18. T. Fujiwara, K. Ritchie, H. Murakoshi, K. Jacobson, and A. Kusumi, “Phospholipids undergo hop diffusion in compartmentalized cell membrane,” J. Cell Biol. 157, 1071–1081 (2002). [CrossRef]
19. K. Ritchie, X.-Y. Shan, J. Kondo, K. Iwasawa, T. Fujiwara, and A. Kusumi, “Detection of non-Brownian diffusion in the cell membrane in single molecule tracking,” Biophys. J. 88, 2266–2277 (2005). [CrossRef]
21. C. Brunner, K.-H. Ernst, H. Hess, and V. Vogel, “Lifetime of biomolecules in polymer-based hybrid nanodevices,” Nanotechnology 15, S540–S548 (2004). [CrossRef]