Scanning near-field optical microscopy and spectroscopy as a tool for chemical analysis

Tags: SNOM, resolution, topographic image, SNOM tips, Ultramicroscopy, tip, ablation, D. W. Pohl, D. Zeisel, SERS, B. Dutoit, J. K. Trautman, Renato Zenobi, optical microscopy, Raman intensity, Raman spectrum, spectroscopic methods, diffraction limit, A. Lewis, laser pulses, H. Fuchs, microscopy, Raman scattering, hot spots, fluorescence detection, fluorescence, E. Betzig, R. Zenobi, V. Deckert, analytical methods, analytical method, J. Microsc, chemical analysis, A. Harootunian, SNOM probes, chemical information, nanometer scale, R. Zenobi V. Deckert Volker Deckert, Appl, optical transmission, spectroscopy, scanning electrochemical microscopy, M. Isaacson, Volker Deckert Research, optical spectroscopy, optical resolution, C. Gerber, Raman spectra, confocal fluorescence, H. Rohrer, fluorescence imaging, topographic images, surface-enhanced Raman, silver spheres, fluorescent molecules, SNOM-Raman, signal intensity, U. C. Fischer, intensity dependence, G. Binnig
Content: REVIEWS
Scanning Near-Field Optical Microscopy and Spectroscopy as a Tool for Chemical Analysis Renato Zenobi* and Volker Deckert
Research into the nanometer-scale region is currently of relevance in many branches of modern science and engineering, such as in microelectronics, supramolecular chemistry, and in a biological context. A great deal of attention is given to the design of molecular devices, usually towards understanding the function of existing Єmolecular machinesє. Central to this task are powerful diagnostic tools capable of recording chemical information with spatial resolution in the nanometer range. While elemental analysis of surfaces with a lateral resolution of a few dozen nanometers is almost routine, analysis of molecular species with a resolution of < 1 mm is very difficult. Scanning-tunneling and
atomic-force microscopies usually do not give any chemical information. By combining scanning near-field optical microscopy (SNOM)Рthe Єoptical memberє of the family of scanningprobe microscopiesРwith optical spectroscopy, it is possible to obtain molecular information from sample areas as little as 50 nm in diameter. In SNOM, a light source is scanned above the object of interest at a distance of a few nanometers. In the optical near field, the illuminated area is not subject to the AbbeВ diffraction limit, but merely by the size of the illuminating source. High quality SNOM probes can be reproducibly prepared by a chemical-etching method. These etched probes have an optical transmission
up to 1000-fold higher than commercial (pulled) SNOM tips and can withstand higher laser power. This last advantage allows not only high resolution optical imaging, but also localized spectroscopic investigations of surfaces and even optical Єnanosamplingє by pulsed-laser ablation. The ablated material can be transported over a considerable distance, which opens the possibility for its subsequent analysis with a complementary, highly sensitive analytical method, such as mass spectrometry. Keywords: analytical methods ґ mass spectrometry ґ optical near-field microscopy ґ optical spectroscopy ґ surface analysis
1. Introduction A great breakthrough in surface science was the development of the scanning-probe microscopy techniques, led by the development of the scanning-tunneling microscope (STM), which earned Heinrich Rohrer and Gerd Binnig the 1986 Nobel Prize in Physics.[1±3] Soon thereafter, a host of related methods were developed, all based on piezoelectric scanning of a very sharp probe relative to the sample, such as atomicforce microscopy (AFM), Scanning electrochemical microscopy, scanning magnetic microscopy, and many others. They are sometimes collectively referred to as the Єchildren of the STMє.[4] [*] Prof. R. Zenobi, Dr. V. Deckert Laboratorium fьr Organische Chemie Eidgenцssische Technische Hochschule (ETH) Zьrich Universitдtsstrasse 16, 8092 Zьrich (Switzerland) Fax: ( 41) 1-632-12-92 E-mail: [email protected]
Of particular interest to applications in molecular science is scanning near-field optical microscopy (SNOM),[5±8] the Єopticalє version among the children of the STM. SNOM delivers light to spots of less than 100 nm diameter and thereby holds great promise for molecular analysis on the nanometer scale. An optical resolution of as little as 12 nm has been claimed by Betzig and Trautman[9] but only after deconvolution of the size of the near-field optical probe. This spatial resolution has not been reproduced lately by other researchers. Delivering light to such small spots is not possible by conventional optical methods, such as by a focussed laser beam. It is also impossible to resolve features separated by less than l/2 by conventional optical microscopy, where l is the wavelength of the illuminating light. This limitation is imposed by the AbbeВ diffraction limit. At visible wavelengths, a resolution approaching 200 ± 250 nm should be theoretically possible but, except for confocal schemes employing oilimmersion lenses and therefore subject to other limitations, a resolution better than 0.5 ± 1 mm is rarely achieved. This diffraction limit is the reason that short-wavelength ultra-
Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756 WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000 1433-7851/00/3910-1747 $ 17.50+.50/0
1747
REVIEWS violet light or even X-rays, where l is smaller still, find increasing use in microelectronics and other applications of lithography but at the cost of more expensive optics and instrumentation. The possibility of illumination in the visible, in the near ultraviolet, or in the infrared range onto a spot smaller than the diffraction limit permits the application of a wide range of spectroscopic methods to problems in chemistry, biology, materials science, and other fields in which the molecular composition of nanoscale objects is of interest. Analyzing the elemental composition with a resolution less than 100 nm is now established and a routine procedure with scanning Auger microscopy, some secondary-ion mass spectrometers, and certain forms of electron-microprobe analysis.[10] These methods typically operate under ultrahigh vacuum conditions, often require special sample pretreatment, and rely on focussing an ion or electron beam to a spot < 100 nm in size, which is detrimental for most molecular species. The chemical analysis of the molecular composition with such a resolution was impossible in practise until now, and is just starting to be explored by a combination of SNOM technology with spectroscopic methods. A large advantage of the SNOM methods is that they can operate under atmospheric pressure. 2. Experimental Section Commercially available near-field scanning optical microscopes (models Lumina and Aurora, ThermoMicroscopes (formerly Topometrix), Santa Clara, CA (USA) form the basic units of our instrumentation (Figure 1). Laser light penetrating through the nanometer-sized aperture of the probe illuminates the surface under investigation. The sample stage is scanned relative to the tip in order to maintain a fixed
R. Zenobi and V. Deckert Figure 1. Equipment for scanning near-field microscopy. The sample is mounted on a piezoelectric x,y,z scanning stage. The SNOM tip is held either by a dither piezo (shown) or by a tuning fork (not shown). Shear force damping is used for controlling the tip ± sample distance through a feedback loop. Light from a continuous-wave (CW) laser is coupled onto the blunt end of the fiber, illuminating the sample through the SNOM tip. Scattered light is collected either in transmission (shown) or in reflection (not shown) using conventional far-field optics. Detection uses either an avalanche photodiode (APD) or by an optical spectrograph equipped with a charge-coupled device (CCD). The entrance slit of the spectrograph or a pinhole mounted in front of the APD further reduce stray light. optical path. The emitted or scattered light from the sample is collected with high numerical aperture microscope objectives either in transmission or in reflection (not shown in Figure 1).
Renato Zenobi is Professor of Analytical Chemistry at the Swiss Federal Institute of Technology (ETH) in Zьrich (Switzerland). He is also one of the founders and the current chairman of the Center of Excellence in Analytical Chemistry at the ETH Zьrich. He obtained his diploma from the ETH (1986) and his PhD degree from Stanford University (1990). After two postdoctoral appointments at the University of Pittsburgh and the University of Michigan (Ann Arbor), he returned to Switzerland in 1992. He held an Alfred Werner Fellowship at the Ecole Polytechnique FeВdeВrale in Lausanne until 1994. He joined the faculty of the ETH Zьrich in 1995. His research interests include laser-based analytical chemistry, laser-assisted mass spectrometry, and near-field optical microscopy and -spectroscopy.
R. Zenobi
V. Deckert
Volker Deckert is pursuing his Habilitation at the ETH Zьrich. He obtained both his diploma (1991) and his PhD degree (1994) from the University of Wьrzburg (Germany), working in the area of Raman spectroscopy. He was a post-doc at the University of Tokyo and at the Kanagawa Academy of science and technology (1994/95) and worked on nonlinear and time-resolved laser spectroscopy. Since 1993, he has also been a freelancing developer of spectroscopy software for Photometrics, Tucson, AZ (USA). He joined the ETH Zьrich in 1996 and is currently leading the Surface Analysis and Spectroscopy team in the group of Prof. R. Zenobi.
1748
Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756
Optical Near-Field Microscopy Collected light is then filtered and projected onto a sensitive detector for fluorescence detection or, alternatively, onto the entrance slit of an optical spectrograph. The latter is optimized for a greater amount of transmitted light. In our arrangement, a transmission holographic-grating spectrograph with a high numerical aperture was used (Holospec f/1.8i, Kaiser Optical Systems, Ann Arbor, MI (USA)). To maintain a constant tip-to-sample distance, an optically detected shear-force feedback loop is most frequently used,[11, 12] although other methods have been employed.[13±17] For the feedback loop, the tip dithers at its resonance frequency. The dithering amplitude decreases dramatically when the tip is within 5 ± 15 nm of the sample surface. The reasons for the shear force damping near the surface are still not fully understood; probably a combination of contact, frictional, and adhesive forces induced by a thin water film on the sample surface are responsible, at least under ambient conditions.[18, 19] In the Єconstant gapє mode, a feedback loop maintains a constant distance, normally < 10 nm, between the probe and surface. For flat samples, scans in Єconstant height modeє are advantageous to avoid topographic artifacts.[20] Such scans are performed in the absence of a distance feedback loop once the tip ± sample distance has been established. The probe is the most delicate component of any SNOM equipment. In this review, we will be dealing exclusively with Єaperture SNOM probesє, that is, tips with an apex smaller than the wavelength of the illuminating light, and made of a tapered glass fiber coated with metal except at the apex.[12, 21] Other probes are also used, including point-source emission probes[22] or apertureless probes.[23±28] The important properties of aperture SNOM probes are good transparency, a well defined circular aperture, no light leaking through defects in the metal coating, and a high optical-damage threshold. Several methods have been proposed to prepare the tapered glass core probes. Adiabatic pulling of optical fibers during heating with a CO2 laser,[29, 30] similar to preparing a pipette, is currently the most popular method. In another case, tiny splinters of glass as tetrahedral SNOM tips are used.[31, 32] Several groups are employing focused ion-beam milling to create very well defined apertures at the tip apex.[33, 34] Recently, microfabricated tips have been described in the literature[35±38] and such tips will become commercially available soon. A very practical method for the preparation of probes is based on etching glass fibers at the meniscus between hydrofluoric acid (etching fluid) and a protective organic overlayer.[39] Etched fiber probes usually provide a higher optical throughput due to larger cone angles than the pulled probes,[38±42] but a high yield of good tips produced by etching is often not obtained. A more reproducible and efficient method, called Єtube etchingє, was recently developed in our laboratory.[43] By this method, the protective polymer coating of the optical fiber is retained, and a tip forms due to a convective mechanism inside the polymer jacket. The method is tolerant to perturbations during preparation such as vibrations and temperature changes. The resulting tips are produced reproducibly and in high yield with large cone angles and smooth aluminum coating. It should also be noted Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756
REVIEWS that the composition of the glass is crucial for obtaining good etching results and also to avoid large background signals during SNOM experiments arising, for example, by Raman scattering of the fiber material itself (see Section 3.2). The left side of Figure 2 shows an electron micrograph of an aluminum-coated SNOM tip produced by the tube-etching method. The large opening angle is immediately apparent Figure 2. Left: Electron micrograph of a SNOM tip produced by the tubeetching method and a closeup of the apex region of the same tip. Right: Representation of the origin of differences between optical and topographic images. and, in the closeup, an aperture of well below 100 nm diameter is visible. On the right side, the difference between the Єtopographic tipє and the Єoptical tipє is illustrated. Sometimes a grain of the metal coating will protrude such that the center of the tapered optical fiber and that of the topographic tip can be separated by several tens of nanometers. If this is the case, the offset of the topographic and the optical images can lead to a misinterpretation of the combined SNOM and topographic data. 3. Results and Discussion 3.1. Fluorescence Imaging with Etched Optical-Fiber Tips Fluorescence is by far the most widespread and easily implemented mode for chemical imaging by near-field optical methods, although the resolution offered by modern confocal fluorescence-microscopy techniques is often sufficient.[44] A fluorescent tag must be attached to the molecules of interest, unless their natural fluorescence can be exploited. Due to high signal intensity, it is even possible to image single fluorescent molecules with SNOM technology. Fascinating applications, involving single molecule fluorescence detection, have been described by Betzig,[45] Xie,[46±48] and van Hulst.[49] A very promising area for the use of SNOM in combination with fluorescence detection is the study of biological nanostructures, in particular, parts of cells. An example from our laboratory is shown in Figure 3 A. In this application, dog kidney (Madine ± Darby canine kidney, MDCK) cells were grown on a flat silica slide using glutaraldehyde and lysed such that the interior of their cell walls became accessible to SNOM imaging.[50] The actin filaments of the cytoskeleton were stained selectively with a rhodamine-labeled phalloidin; they appear green in the image. Figure 3 shows immediately that much finer features are resolved with SNOM. Specialized laser-scanning confocal- 1749
REVIEWS
R. Zenobi and V. Deckert
Figure 4. A) SNOM/fluorescence image and B) topographic image of a 10 В 10 mm area of a PVB film containing fluorescent microspheres. The spheres had a diameter of 288 nm. The fluorescence was excited with the 488 nm line of an Ar-ion laser.
Figure 3. A) SNOM fluorescence image of the cytoskeleton of Madine ± Darby canine kidney cells stained with a fluorescent dye (see text for details). B) Confocal micrograph image of the same sample (different area).
tion. It would be, at best, very difficult to identify unknown molecules based solely on fluorescence. Other optical spectroscopies are better suited for this purpose.
microscopy techniques push the resolution below 200 nm by, for example, using two-photon excitation. However, the lateral resolution in Figure 3 B is about 300 nm, typical for confocal fluorescence microscopes. Actin filament bundles can be resolved by confocal microscopy but the resolution in the SNOM image (Figure 3 A) considerably exceeds that of the confocal-microscope image (Figure 3 B). Near-field optical detection is not strictly limited to surfaces. This is illustrated in Figure 4. Here, a sample consisting of fluorescent latex spheres embedded in a polyvinyl butyral (PVB) thin film was imaged by SNOM using fluorescence detection in transmission mode. The spheres had a diameter of 288 nm and were impregnated with a fluorescent dye that is easily excited with the 488 nm line of an Arion laser. Spheres that are located on the surface of the polymer matrix appear with sharp contours both in the topographic and in the SNOM/fluorescence image, albeit with a larger apparent diameter of about 375 nm. This allows determination of the tip diameter by deconvolution, in this case an optical tip diameter of 100 nm may be deduced. Other latex spheres appear blurred in the optical image at locations where no protrusions were apparent in the simultaneously recorded topographic image. This can be attributed to latex spheres that are not located directly at the surface but are embedded deeper in the PVB film. Detection to a depth of at least 150 nm is therefore possible. A disadvantage of fluorescence imaging is the need for a fluorescent moiety or tag to be employed, which is not always possible. A further shortcoming is that fluorescence spectra are broad and structureless, to reveal little chemical informa- 1750
3.2. Near-Field Vibrational Spectroscopy Vibrational spectroscopy is very useful to identify unknown molecules due to its high chemical-information content. It is, however, difficult to combine infrared spectroscopy with SNOM for two reasons: 1) Fiber materials in the mid-infrared region, where most molecular vibrations are detected, are not readily available and are difficult to handle; 2) The discrepancy between the desired resolution and the diffraction limit is accentuated due to the much longer excitation wavelength. Nevertheless, there are some very interesting, even heroic, attempts to use SNOM methods for imaging the mid-infrared range.[51] Keilmann et al.[28] and Boccara et al.[25, 26] use an apertureless approach: An AFM tip is used for field enhancement within the focus of a fixed-frequency infrared laser. A resolution of l/600, about 17 nm, was claimed with this approach.[25] However, true IR spectroscopy with a resolution on the order of a few nanometers has not been achieved yet. An alternative to IR spectroscopy is Raman spectroscopy. The difficulty lies here with the exceedingly low scattering cross-section, such that its combination with SNOM seems hopeless at first. However, there are certain advantages of Raman over IR spectroscopy: Suitable optical fibers are readily available, there is a large choice of very sensitive detectors, and it is possible to use resonance-excitation techniques. Finally, Raman cross-sections are increased by orders of magnitude if surface enhancement (SERS) is employed. It is relatively straightforward to combine surface-enhanced spectroscopies with SNOM to improve the overall photon yield. Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756
Optical Near-Field Microscopy Several research groups have begun to implement Raman microscopes where the excitation uses an aperture SNOM probe.[52±62] The alternative is again to use apertureless probes: A very sharp tip, usually a metallic STM or AFM tip, can be used for field enhancement inside the diffraction-limited focus of the illuminating laser beam. The position of the tip can be temporally modulated in order to employ lock-in detection schemes.[23, 24] Resolution to 10 has been claimed using this method, albeit not with Raman scattering. It is, however, still unclear where the optical contrast originates from. For example, it is possible that light reflected off the side of the probe contributes more to the overall signal than radiation from the apex of the probe. A shadowing effect can also occur, which may even cause a reversal of the optical contrast. Clever schemes, for example using two-photon excitation, can be used to circumvent these problems.[27] Figure 5 shows an image of a silver SERS substrate that was specially designed to contain isolated islands of silver without any topographic substructure.[63] The substrate was then coated with an adlayer of rhodamine 6G (R6G). First, the Figure 5. Shear-force topographic image of a substrate with a SERS coating. For near-field Raman measurements, a monolayer of rhodamine 6G is applied. At various locations, marked by the circles, near-field SERS measurements were made. The circles are 70 nm in diameter and represent the optical resolution obtained with the SNOM tip. The pole heights above the transparent plane denote the relative signal intensity of the Raman peak at 1650 cmА1. The signals are normalized against the average Raman intensity measured on unmetallized glass slides. Each segment of the poles corresponds to an 1000-fold signal enhancement. surface morphology was measured using the shear-force feedback-loop mechanism. The SNOM probe was then parked at selected positions of the depicted 1 В 1 mm area and the Raman spectrum of the R6G layer was measured. Each measured position in Figure 5 is depicted by a white circle, the diameter of which (about 70 nm) indicates the approximate optical resolution obtained by the SNOM tip. Each position was illuminated for one minute through the SNOM tip, to yield the characteristic Raman spectrum of R6G. The optical resolution of the tip was determined by inherent spectral variation in the Raman spectra of adjacent locations; scanning electron micrographs of the used SNOM Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756
REVIEWS
tips confirmed this estimate of 70 nm. Assuming a monolayer
coverage of the dye, a spot size of this diameter corresponds to
a sensitivity of less than 100 R6G molecules!
As a measure for the Raman enhancement factor (EF), the
normalized and background-corrected Raman peak intensi-
ties at 1650 cmА1 were selected, which correspond to an
aromatic CАC stretch vibration of R6G. The magnitude of this
band was directly compared with the average value obtained
for unmetallized, R6G-
coated substrates. The cal-
culated EFs are superim-
posed on the topographic
image of Figure 5. Each
segment of the poles corre-
sponds to an additional
1000-fold signal enhance-
ment. The EFs lay between
2000 to well above 6000,
with over a 50 % variation
observed in the Raman in-
tensity for two points sepa-
rated by less than 100 nm.
Interestingly, the overall
EF measured by far-field
optics was higher still,
around 11 000. We still can-
not fully explain the dis-
crepancy between the en-
hancement factors deter-
mined by near-field and
far-Field measurements.
Perhaps Єhot spotsє with
much higher EFs exist on
the substrate, which cannot
be found in the near field.
Alternatively, cumulative
effects of neighboring is-
lands may play a role. A
further possibility is that
the aluminum coating of
the SNOM tip deteriorates
the EF near the silver islands by some electrostatic interaction as soon as the SNOM tip is in proximity.
Figure 6. SNOM-SERS imaging of a BCB-labelled DNA sample deposited on a substrate consisting of monodisperse silver spheres for SERS
We have also demonstrated high resolution Raman imaging is possible with the near-field optics. Figure 6 A shows the shear-
enhancement. Lateral dependence of: A) shear-force topography; B) the Raman signal intensity at 800 cmА1 (glass); C) the near-field surface-enhanced Raman intensity at 1641 cmА1 (BCB) before normal-
force topograpic image and Figure 6 B ± D the optical images of a 2 В 2 mm area of a SERS-coated sub-
ization; and D) the near-field surface-enhanced Raman intensity at 1641 cmА1 after normalization. In images (B) ± (D), the pixel size was only 100 В 100 nm and, for each
strate, which consisted of silver-coated Teflon microspheres with an overall diameter of 300 nm.[61] The shear-force topography im-
graph, the intensity of one line of the complete Raman spectrum was plotted. In images (C) and (D), the contrast between the lowest and the highest Raman intensity was about a factor of three.
1751
REVIEWS age shows that aggregated silver spheres are clearly resolved and their spatial distribution is quite irregular. This substrate was coated with a monolayer of a 18-mer DNA fragment labelled with brilliant cresyl blue (BCB).[62] To achieve a reasonable acquisition time for the entire image, we measured the near-field Raman spectra as an array of only 20 В 20 data points. This acquisition time was twenty times longer compared to normal SNOM imaging times. Nevertheless, for an aperture smaller than 100 nm, it is the tip size and not the step size that limits the optical resolution. This represents a slight undersampling that we accept in order to obtain the entire spectral information within a reasonable measurement time. The image of the Raman spectrum feature at 800 cmА1 (Figure 6 B), which stems from the glass of the SNOM tip itself, is at least partially dependent upon the topography. From the comparison of Figures 6 A and B, it is obvious that the correlation of the topographic information and the reflected light from such a highly corrugated sample is by no means trivial. In particular, an interpretation of the image and an estimation of the optical resolution is difficult if topographic coupling or shadowing effects[20] occur. The image in Figure 6 C shows the lateral intensity dependence of the Raman band from the BCB label (at 1641 cmА1). By comparing Figures 6 B and 6 C, it is apparent that Raman signals from the label and from the glass tip show some correlation, such as the increased intensity in the lower left corner. However, the Raman signal stems from the BCBlabelled DNA, whereas the glass scattering signal originates entirely from the SNOM tip. The bands from the glass therefore should be proportional to the local reflectivity of the surface. The Raman scattered signal is subject to the same reflectivity changes as the glass signal. If, for instance, a particle between the illuminated spot and the collection optics decreases the reflectivity, the Raman signal is attenuated in exactly the same fashion and extent. In other words, the topography of our sample is encoded in the reflectivity signal, such that the topographic effects can be removed from the Raman signal by normalizing it against the intensity of the glass band. The effect of this normalization is demonstrated in Figure 6 D, in which the Raman intensity of the 1641 cmА1 BCB band at each position after dividing the data in Figure 6 C by the intensity of Figure 6 B. A comparison of Figures 6 D and 6 C shows that instead of two bright spots, the corrected image shows only one spot of high SERS intensity. At the position in the lower left, an enhanced sample reflectivity is responsible for the increased Raman signal, whereas the bright spot on the upper left remains after the normalization. In the corrected Raman images at all frequencies, we observe a small, highly SERS-active region of about 400 nm diameter on the center-left side of the image. If we assume a uniform distribution of the sample on the surface, this increase in intensity can only be attributed to a particularly strong enhancement by the substrate (Єhot spotsє), although concentration fluctuations leading to a variation of Raman signals cannot entirely be excluded. Interestingly, the maximum SERS enhancement does not coincide with the position of a silver particle but, in fact, the location of the maximum enhancement seems to be at a position where a gap between 1752
R. Zenobi and V. Deckert
the silver spheres is visible (Figure 6 A). We have to be aware
of the possibility that the topographic and the SNOM-Raman
image may not be in perfect congruence. It has been
postulated that Єhot spotsє may be located on SERS
substrates where gaps or cavities exist.[64] In previous work
by Emory and Nie,[56] there are hints for such Єhot spotsє but
they were unable to image these sites directly. In contrast, the
SNOM method allows directly visualization of Єhot spotsє
because of the simultaneous
recording of sample topography
and Raman intensity at each
data point; this is particularly
straightforward if low-density
silver-island films are used as
SERS substrates.[63]
It is interesting to compare
the relative magnitude of Ram-
an scattered light and fluores-
cence signals. The Cross Sections
found for SERS on Єhot spotsє
was of the same order of mag-
nitude as that of the fluores-
cence signals![56, 65] Upon ab-
sorption of a single photon, an
individual molecule will either
undergo a Raman scattering
event or fluoresce, but not both.
Photobleaching processes are
intimately linked with fluorescence and therefore should not affect the Raman emission
Figure 7. Raman lines superimposed on a fluorescence background in the optical spec-
process. This is demonstrated in Figure 7, where three consecutive spectra of R6G deposited on a silver-island SERS substrate are shown. The intensity
trum of a monolayer of rhodamine 6G deposited on a silver SERS substrate (excitation at 488 nm). The spectra were recorded consecutively (from top to bottom), using an integra-
of the fluorescence background decreases due to photobleaching, while that of the Raman bands remains constant, in
tion time of 5 min per spectrum. Slow photobleaching is observed for the fluorescence background, while the intensity of the Raman lines remains
agreement with the expecta- constant.
tion. The reason for this behav-
ior may be that Єhot spotsє on the SERS substrate coax the
adsorbed molecules into undergoing Raman scattering rather
than fluorescence, thereby protecting them from photo-
bleaching. The average coverage of the SERS substrate by
the dye was slightly more than one monolayer; apparently
only a fraction of the adsorbed R6G molecules are associated
with these protective sites but are responsible for quite strong
SERS signals. This again supports the idea that only a few
Єhot spotsє exhibit a very large Raman enhancement and are
responsible for the overall intensity observed.
Are SERS methods the only hope to obtain Raman spectra
excited by near-field optical probes? The answer to this
question is, as demonstrated in Figure 8, a clear Єnoє. The
near-field Raman spectrum of a chemical-vapor deposited
(CVD) diamond film shows a sharp feature at 1330 cmА1,
which is the well known Raman band of diamond. Raman
scattering from graphitic carbon is expected around
Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756
Optical Near-Field Microscopy
REVIEWS cules will not be randomly oriented and the structure of an interface can be studied through symmetry considerations. Figure 9 A and 9 B are far-field Raman spectra of liquid pxylol, while 9 C and 9 D are near-field Raman spectra of the same. Over the transition from parallel to crossed orientation of the polarizers in the excitation and detection beam path, the expected intensity decrease is observed for many Raman modes, for example, the strong aromatic-ring vibrations at 820 and 1200 cmА1. In contrast, the band at about 1600 cmА1 remains strong even when crossed polarizers are used, indicating large anisotropic components in its scattering tensor. Figure 9 shows impressively that the SNOM tip used for recording the near-field data preserves the polarization quite well.
Figure 8. Near-field Raman spectrum of a diamond film produced by CVD methods. The excitation wavelength was 488 nm and the laser line was filtered out in front of the spectrograph by a holographic notch filter. Strong bands due to Raman scattering from the SERS tip itself dominate the spectrum (glass). Inset: the sharp band at 1330 cmА1 is assigned to diamond, and the broad feature at % 1600 cmА1 is due partially to glass and partially to graphitic carbon.
1600 cmА1, but unfortunately a fairly strong residual signal
from the long progression of glass Raman bands, originating
from the tip itself, overlaps with the sp2 carbon band. This is
partially due to the low Raman intensity of the sample, but the
Raman emission from the glass of the SNOM tip depends
strongly on the glass fiber material used. Our latest generation
of SNOM tips exhibit much
lower Raman scattering in
this spectral region, such
that, in principle, the purity
of the film could be meas-
ured at any given location.
The exquisite resolution
in the z direction makes the
SNOM an attractive tool to
study the interface between
two liquids. In particular,
molecular orientation at
such interfaces may be stud-
ied by the anisotropies of
the Raman scattering ten-
sor using polar-filtered ex-
citing light and detection.
For the isotropic part of the
Figure 9. Raman spectra recorded in liquid p-xylol using polar-filtered exciting light and detection. A) Far-
scattering tensor of the fully symmetric vibrational modes, a strong decrease in
field spectrum with parallel polarizers; B) Far-field spectrum with perpendicular polarizers; C) Nearfield spectrum with parallel polarizers; D) Near-field spectrum with
signal intensity is expected if the polarizers are crossed. In contrast, for anisotropic modes, some intensity will
perpendicular polarizers. The absolute intensities for each spectrum are not directly comparable. The laser power was 10 mW (at l 488 nm) and the signal, averaged over 10 min, was used.
still be detected even for crossed polarizers. This is illustrated in Figure 9. At an interface, the orientation of the investigated mole-
Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756
3.3. Laser Ablation through SNOM Tips The high transmission SNOM tips that we use permit not only high resolution optical imaging and spectroscopic investigations of surfaces, but also optical Єnanosamplingє by pulsed-laser ablation through the tip.[66] The laser-ablated material can be transported over a considerable distance and analyzed in a second step by a highly sensitive analytical method such as mass spectrometry. A schematic of this concept is shown in Figure 10. Figure 10. Schematic representation of the coupling of laser ablation through the SNOM tip with mass-spectrometric analysis of the vaporized material. This localized ablation has already been described and a crude realization has appeared in the literature.[67] The authors guided a sharpened optical fiber over a transmission Electron Microscopy (TEM) grid inside the ion source of a mass spectrometer and performed a matrix-assisted laserdesorption/ionization (MALDI) mass-spectrometry experiment. The TEM grid was covered with an acetylcholine sample embedded in a 2,5-dihydroxybenzoic acid matrix. The advantage of this experiment over earlier MALDI studies, where the laser light was also brought in by an optical fiber,[68, 69] remains unclear. The spatial resolution was of several micrometers, far greater than what is possible with the SNOM method. Several mechanisms are possible for laser-induced ablation through SNOM tips.[70] The photochemical mechanism assumes that the energy is deposited in the sample via optical absorption and ablation follows as a consequence of breaking chemical bonds. The photothermal mechanism is similar but assumes the absorbed optical energy would be converted into 1753
REVIEWS
heat, which causes ablation from the locally heated spot. A
ballistic mechanism is also possible, in which the material
from the tip is sputtered onto the sample and material ablates
from the surface. A ballistic mechanism is the most likely
explanation for an experiment carried out on an anthracene
surface using a completely metallized SNOM tip,[41] since, in
this case, no photons reach the sample and therefore a sputter
process, either by metal atoms from the tip or by material
adsorbed on its surface, was believed to be the reason for the
ablation process. Fourth, one also has to take transient
thermal expansion[71±73] into account as a possible mechanism
for the creation of surface indentations, but it is likely to be
negligible.[70]
We now discuss an experiment that gave clear results about
the ablation mechanism of a rhodamine B film. The sample
was irradiated with laser pulses either at its optical adsorbtion
maximum (Єon resonanceє, 532 nm) or nearby this point (Єoff
resonanceє, 650 nm).
SNOM tips with aper-
tures of < 100 nm diam-
eter were used. No sig-
nificant difference in
the optical transmission
of the SNOM tip over
the 400 to 650 nm
range, although some
decrease is expected
with an increasing dif-
ference between the
wavelength and tip
aperture size. Figure 11
shows shear-force topo-
graphic images record-
ed after several laser
pulses had been fired
onto the sample. No
ablation was observed
for the off-resonance
Figure 11. Shear-force topographic im- wavelength
(Figure
ages of a rhodamine B film after transmitting laser pulses through the SNOM tip at A) 2.1 mJ at l 650 nm, far from the maximum absorption of rhodamine B, and B) 1.4 mJ at l 532 nm (B), a wavelength near the maximum absorp-
11 A;
l 650 nm,
2.1 mJ per pulse). In
contrast, well defined,
% 70 nm
diameter
(FWHM), 5 nm deep
tion. The total topographic contrast in the z direction is A) 10 nm and B) 30 nm.
holes were created when the laser operated at a wavelength that is
strongly absorbed by
the sample (Figure 11 B; l 532 nm, 1.4 mJ per pulse). This
clearly points to an optical ablation mechanism, either
photochemical or photothermal. Energy transfer by absorp-
tion of photons appears to be more efficient than by a ballistic
process. The photon energy used here is not sufficient for
inducing bond dissociation in rhodamine molecules, in con-
trast to ablation experiments with ultraviolet laser radiation,
where photochemical dissociation is often used to rationalize
the observed phenomena.[74] We therefore suggest that a
photothermal mechanism is responsible for the ablation of the
rhodamine film: The absorbed photon energy is converted to
1754
R. Zenobi and V. Deckert vibrational excitation, which causes a rapid temperature increase and leads to thermal desorption of the molecules. Another interesting observation in Figure 11 is that the ablated material is redeposited on the sample surface close to the ablation crater. The distribution of redeposited material was not symmetric in this experiment but was always to the left side of the crater and independent of the scan direction. In other SNOM ablation experiments, the distribution was more symmetric. The origin of the directionality appears to be the tip itself: The apex of a SNOM tip is frequently not symmetric and the ablated material is ejected in a specific direction. Transport over micrometer distances was observed in several cases. The result of such an experiment is shown in Figure 12.[70] Here, an anthracene crystal surface was continuously irradiated with a pulsing laser (6 ns, 532 nm, 50 mJ) operated at a 20 Hz repetition rate over a 1 В 2 mm area. A shear-force image of a larger area was then recorded without Figure 12. Transport of material over many micrometers: A topographic image of a 5 В 8 mm area of an anthracene film crystal surface, previously irradiated in a 1 В 2 mm trench in the upper left corner with ablating laser pulses (pulse: 6 ns pulse; wavelength: 532 nm; pulse energy: 50 mJ; repetition rate: 20 Hz). Total topographic contrast in the z direction is 100 nm. Adapted from ref. [70]. irradiation. An approximately 1 mm wide trench with well defined edges could be detected in the upper-left corner of the image, while a 2 mm wide, 60 nm high mound was found about 5 mm away running roughly parallel to the trench. We interpret this again as vaporization followed by redeposition of material in a preferred direction. The direction of deposition is most likely determined by an asymmetry of the tip. For collection of ablated molecules, for example by a vacuum interface to a mass spectrometer, this asymmetry could be an advantage if the SNOM tip could be arranged to direct the ablated material towards the collector. From Figure 12 we also estimate that only a few, larger chunks of material are ablated. Most of the anthracene is found in the form of a diffuse and structureless mound, which is expected if the molecules are redeposited from the gas phase. We have also shown that the laser-ablated molecules do not decompose.[70] We conclude that material transport does not take place by a mechanical dragging from the SNOM tip, but instead by vaporization and redeposition in a direction dictated by the assymetry of the tip. 4. Conclusions and Outlook Near-field imaging based on aperture probes can now provide Spectroscopic data with a lateral resolution on the Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756
Optical Near-Field Microscopy order of 50 nm. Some improvements in resolution may still be feasible but the aperture SNOM will probably never reach a resolution of 1 nm or less, which is possible by STM and AFM. However, as shown in this article, SNOM can provide a wealth of chemical information, which, in our opinion, more than compensates for the lack in spatial resolution. High quality, high transmission SNOM tips can now be produced by various methods. The chemical etching method described here is one of the most easily implemented and provides high quality SNOM tips in a good yield. The Єtube etchedє tips can also withstand higher laser energies[75] and are therefore brighter near-field light sources. We have shown examples of fluorescence and SERS imaging in the optical near-field with a lateral optical resolution of less than 100 nm, relatively short exposure times, and high signal-to-noise ratios. The different origins of the reflectivity and Raman signals can be used for an intensity correction of the Raman images by a simple normalization. This procedure corrects topographic artifacts present in standard SNOM images and can always be used if the detection of the interesting compound can be made at a different wavelength. Surface enhancement of optical transitions appears to be due to a small number of Єhot spotsє. With the combined optical and topographic information available in SNOM, we have a good chance to identify the nature of these sites. Surface enhancement, however, is not always a prerequisite for obtaining near-field Raman spectra, as shown on the example of a CVD diamond film. Finally, the tip itself could be used as a platform to carry a small particle[76, 77] that provides strong Raman enhancement. Such a Єtip SERSє approach would not require the sample to be deposited on a rough silver film and thereby makes the technique more generally usable. In the area of pulsed-laser ablation through SNOM tips, we have shown that it is possible to create highly localized (around 70 nm diameter) ablation craters on the surface of organic materials. The ablated molecules remain intact and often fly in a direction determined by an asymmetry of the SNOM tip. Material can be transported over several micrometers, collected, and analyzed by a complementary method. We have recently built an interface to a mass spectrometer for this purpose. In contrast to common laserablation ion sources for mass spectrometry, this interface allows the sample to be at ambient pressure. Mass spectral data of material ablated by single diffraction-limited laser pulses have been successfully recorded, and the signal-tonoise ratio was good enough to put an equivalent near-field experiment within reach. We wish to thank Dr. Dieter Zeisel for many contributions to this article, Dr. Yung Doug Suh, Raoul Stцckle, Christian Fokas, Anja Vinckier, Urs Ziegler, Dr. Roland Hauert, Dr. Jцrg Patscheider, and Dr. Tuan Vo-Dinh for their help, and the Swiss National Science Foundation for financial support (National Research Project ЄNanotechnologyє, NFP-36). Received: July 21, 1999 [A 355] Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756
REVIEWS [1] G. Binnig, H. Rohrer, Helv. Phys. Acta 1982, 55, 726 ± 735. [2] G. Binnig, H. Rohrer, C. Gerber, E. Weibel, Phys. Rev. Lett. 1982, 49, 57 ± 59. [3] G. Binnig, H. Fuchs, C. Gerber, H. Rohrer, E. Stoll, E. Tosatti, Europhys. Lett. 1986, 1, 31 ± 36. [4] R. Pool, Science 1990, 247, 634 ± 636. [5] D. W. Pohl, W. Denk, M. Lanz, Appl. Phys. Lett. 1984, 44, 651 ± 653. [6] A. Lewis, M. Isaacson, A. Harootunian, A. Muray, Ultramicroscopy 1984, 13, 227 ± 231. [7] D. W. Pohl, U. C. Fischer, U. T. Dьrig, Scanning Microscopy Tech- nologies and Applications, Washington, DC, 1988, pp. 84 ± 90. [8] E. Betzig, M. Isaacson, A. Lewis, K. Lin, Scanning Microsc. Technol. Appl. 1988, 897, 91 ± 99. [9] E. Betzig, J. K. Trautman, Science 1992, 257, 189 ± 195. [10] D. Brune, R. Hellborg, H. J. Whitlow, O. Hunderi, Surface Character- ization, WILEY-VCH, Weinheim, 1997. [11] R. Toledo-Crow, P. C. Yang, Y. Chen, M. Vaez-Iravani, Appl. Phys. Lett. 1992, 60, 2957 ± 2959. [12] E. Betzig, P. L. Finn, J. S. Weiner, Appl. Phys. Lett. 1992, 60, 2484 ± 2486. [13] M. Garcia-Parajo, E. Cambril, Y. Chen, Appl. Phys. Lett. 1994, 65, 1498 ± 1500. [14] K. Karrai, R. D. Grober, Ultramicroscopy 1995, 61, 197 ± 205. [15] J. W. P. Hsu, M. Lee, B. S. Deaver, Rev. Sci. Instrum. 1995, 66, 3177 ± 3181. [16] J.-K. Leong, C. C. Wiliams, Appl. Phys. Lett. 1995, 66, 1432 ± 1434. [17] J. Barenz, O. Hollricher, O. Marti, Rev. Sci. Instrum. 1996, 67, 1912 ± 1916. [18] H. Heinzelmann, D. W. Pohl, Appl. Phys. A 1994, 59, 89 ± 101. [19] C. Durkan, I. V. Shvets, J. Appl. Phys. 1996, 80, 5659 ± 5664. [20] B. Hecht, H. Bielefeldt, Y. Inouye, D. W. Pohl, L. Novotny, J. Appl. Phys. 1997, 81, 2492 ± 2498. [21] D. W. Pohl, Adv. Opt. Electron Microsc. 1991, 243 ± 312. [22] N. Kuck, K. Lieberman, A. Lewis, Appl. Phys. Lett. 1992, 61, 139 ± 141. [23] F. Zenhausern, M. P. OBoyle, H. K. Wickramasinghe, Appl. Phys. Lett. 1994, 65, 1623 ± 1625. [24] F. Zenhausern, Y. Martin, H. K. Wickramasinghe, Science 1994, 269, 1083 ± 1085. [25] A. Lahrech, R. Bachelot, P. Gleyzes, A. C. Boccara, Opt. Lett. 1996, 21, 1315 ± 1317. [26] A. Lahrech, R. Bachelot, P. Gleyzes, A. C. Boccara, Appl. Phys. Lett. 1997, 71, 575 ± 577. [27] E. J. SaВ nchez, L. Novotny, X. S. Xie, Phys. Rev. Lett. 1999, 82, 4014 ± 4017. [28] B. Knoll, F. Keilmann, Nature 1999, 399, 134 ± 137. [29] E. Betzig, J. K. Trautman, T. D. Harris, J. S. Weiner, R. L. Kostelak, Science 1991, 251, 1468 ± 1470. [30] G. A. Valaskovic, M. Holton, G. H. Morrison, Appl. Opt. 1995, 34, 1215 ± 1228. [31] U. C. Fischer, J. Koglin, H. Fuchs, J. Microsc. (Oxford) 1994, 176, 231 ± 237. [32] J. Koglin, U. C. Fischer, H. Fuchs, J. Biomed. Opt. 1996, 1, 75 ± 78. [33] J. A. Veerman, A. M. Otters, L. Kuipers, N. F. van Hulst, Appl. Phys. Lett. 1998, 72, 3115 ± 3118. [34] S. Pilevar, K. Edinger, W. Atia, I. Smolyaninov, C. Davis, Appl. Phys. Lett. 1998, 72, 3133 ± 3135. [35] S. Mьnster, S. Werner, C. Mihalcea, W. Scholz, E. Oesterschulze, J. Microsc. 1997, 186, 17 ± 22. [36] W. Noell, M. Abraham, K. Mayr, A. Ruf, J. Barentz, O. Hollricher, O. Marti, P. Gьnther, Appl. Phys. Lett. 1997, 70, 1236 ± 1238. [37] D. Drews, W. Noell, W. Ehrfeld, M. Lacher, K. Mayr, O. Marti, C. Serwatzy, M. Abraham, SPIE Conference on Materials and Device Characterization in Micromachining, Santa Clara, CA, 1998, pp. 76 ± 83. [38] T. Yatsuia, M. Kourogi, M. Ohtsu, Appl. Phys. Lett. 1998, 73, 2090 ± 2092. [39] P. Hoffmann, B. Dutoit, R.-P. SalatheВ, Ultramicroscopy 1995, 61, 165 ± 170. [40] L. Novotny, D. W. Pohl, B. Hecht, Opt. Lett. 1995, 20, 970 ± 972. [41] D. Zeisel, S. Nettesheim, B. Dutoit, R. Zenobi, Appl. Phys. Lett. 1996, 68, 2491 ± 2492. 1755
REVIEWS [42] M. N. Islam, X. K. Zhao, A. A. Said, S. S. Mickel, C. F. Vail, Appl. Phys. Lett. 1997, 71, 2886 ± 2888. [43] R. Stцckle, V. Deckert, C. Fokas, R. Zenobi, B. Hecht, B. Sick, U. P. Wild, Appl. Phys. Lett. 1999, 75, 160 ± 162. [44] J. Pawley, Handbook of Biological Confocal Microscopy, Plenum, New York, 1995. [45] J. K. Trautman, J. J. Macklin, L. E. Brus, E. Betzig, Nature 1994, 369, 40 ± 42. [46] X. S. Xie, R. C. Dunn, Science 1994, 265, 361 ± 364. [47] R. C. Dunn, G. R. Holtom, L. Mets, X. S. Xie, J. Phys. Chem. 1994, 98, 3094 ± 3098. [48] X. S. Xie, J. K. Trautman, Ann. Rev. Phys. Chem. 1998, 49, 441 ± 480. [49] A. G. T. Ruiter, J. A. Veerman, M. F. Garcia-Parajo, N. F. van Hulst, J. Phys. Chem. A 1997, 101, 7318 ± 7323. [50] U. Ziegler, A. Vinckier, P. Kernen, D. Zeisel, J. Biber, G. Semenza, H. Murer, P. Groscurth, Mol. Biol. Cell 1998, 9, 600. [51] M. K. Hong, S. Erramilli, P. Huie, G. James, A. Jeung, SPIE 1996, 2863, 54 ± 63. [52] D. P. Tsai, A. Othonos, M. Moskovits, Appl. Phys. Lett. 1994, 64, 1768 ± 1770. [53] C. L. Jahncke, M. A. Paesler, H. D. Hallen, Appl. Phys. Lett. 1995, 67, 2483 ± 2485. [54] C. L. Jahncke, H. D. Hallen, M. A. Paesler, J. Raman Spectrosc. 1996, 27, 579 ± 586. [55] D. A. Smith, S. Webster, M. Ayad, S. D. Evans, D. Fogherty, D. Batchelder, Ultramicroscopy 1995, 61, 247 ± 252. [56] S. R. Emory, S. Nie, Anal. Chem. 1997, 69, 2631 ± 2635. [57] Y. Narita, T. Tadokoro, T. Ikeda, T. Sakai, S. Mononobe, M. Ohtsu, Appl. Spectrosc. 1998, 52, 1141 ± 1144. [58] S. J. Stranick, L. J. Richter, R. R. Cavanagh, J. Vac. Sci. Technol. B 1998, 16, 1948 ± 1952. [59] D. Zeisel, B. Dutoit, V. Deckert, T. Roth, R. Zenobi, Anal. Chem. 1997, 69, 749 ± 754.
R. Zenobi and V. Deckert [60] D. Zeisel, V. Deckert, R. Zenobi, T. Vo-Dinh, Chem. Phys. Lett. 1998, 283, 381 ± 385. [61] T. Vo-Dinh, Surface-Enhanced Raman Scattering, Elsevier Science, 1995. [62] V. Deckert, D. Zeisel, R. Zenobi, T. Vo-Dinh, Anal. Chem. 1998, 70, 2646 ± 2650. [63] R. Stцckle, V. Deckert, C. Fokas, D. Zeisel, R. Zenobi, Vibr. Spectrosc. 2000, 22, 39 ± 48. [64] M. Moskovits, J. Mod. Phys. 1985, 57, 783 ± 826. [65] K. Kneipp, Y. Wang, H. Kneipp, I. Itzkan, R. R. Dasari, M. S. Feld, Phys. Rev. Lett. 1996, 76, 2444 ± 2446. [66] G. Krausch, J. Mlynek, Microelectron. Eng. 1996, 32, 219 ± 228. [67] D. Kossakovski, S. D. OConnor, M. Widmer, J. D. Baldeshwieler, J. L. Beauchamp, Ultramicroscopy 1998, 71, 111 ± 115. [68] J. D. Hogan, D. A. Laude, The 38th ASMS Conference on Mass Spectrometry and Allied Topics, Tucson, Arizona, 1990, pp. 848 ± 849. [69] K. Dreisewerd, M. Schьrenberg, M. Karas, F. Hillenkamp, Int. J. Mass Spectrom. Ion Processes 1995, 141, 127 ± 148. [70] B. Dutoit, D. Zeisel, V. Deckert, R. Zenobi, J. Phys. Chem. B 1997, 101, 6955 ± 6959. [71] M. Stдhelin, M. A. Bopp, G. Tarrach, A. J. Meixner, I. Zschokke- Grдnacher, Appl. Phys. Lett. 1996, 68, 2603 ± 2605. [72] H. J. Mamin, Appl. Phys. Lett. 1996, 69, 433 ± 435. [73] C. Lienau, A. Richter, T. Elsaesser, Appl. Phys. Lett. 1996, 69, 325 ± 327. [74] D. Singleton, G. Paraskevopoulos, R. S. Taylor, Chem. Phys. 1990, 144, 415 ± 423. [75] R. Stцckle, N. Schaller, V. Deckert, C. Fokas, R. Zenobi, J. Microsc. 1999, 194, 378 ± 382. [76] K. Sasaki, J. Vac. Sci. Technol. B 1997, 15, 2786 ± 2790. [77] R. Stцckle, Y.-D. Suh, V. Deckert, R. Zenobi, Chem. Phys. Lett. 2000, 318, 131 ± 136.
1756
Angew. Chem. Int. Ed. 2000, 39, 1746 ± 1756

File: scanning-near-field-optical-microscopy-and-spectroscopy-as-a.pdf
Title: Scanning Near-Field Optical Microscopy and Spectroscopy as a Tool for Chemical Analysis
Published: Fri May 5 06:10:10 2000
Pages: 11
File size: 0.47 Mb


, pages, 0 Mb
Copyright © 2018 doc.uments.com