Vous êtes sur la page 1sur 4

NANO LETTERS

Electromagnetic Quantum-Size Effects in Directional Near-Field EELS of Nanocrystals


H. Cohen, B. I. Lembrikov,*,, M. A. Itskovsky, and T. Maniv
Weizmann Institute of Science, RehoVot 76100, Israel, Chemistry Department, Technion-Israel Institute of Technology, Haifa 32000, Israel, and Holon Academic Institute of Technolodgy, Golomb str. 52, Holon 58102, Israel
Received September 19, 2002; Revised Manuscript Received December 19, 2002

2003 Vol. 3, No. 2 203-206

ABSTRACT
Pronounced electromagnetic quantum-size effects are observed in directional near-field electron energy loss spectroscopy of rectangular nanocrystals. A theoretical analysis, exploiting a unique coincidence of all relevant length parameters in this spectroscopy and the momentum space filtering imposed by the exponentially decaying electronplasmon interaction in the vacuum, reveals strong sensitivity of the EEL signal to the size and shape of the nanocrystal. The breakdown of momentum conservation along the beam direction, associated with the finite size of the nanocrystal, drastically alters the EEL signal pattern.

Electron energy loss spectroscopy (EELS) provides a powerful tool for the study of nanoparticles, combining the subnm lateral resolution of scanning transmission electron microscopy (STEM) with broad spectral capabilities at typical energy resolution of 0.5 eV.1 Exploiting STEM focused beams at highly controlled distances2 in the nearfield (NF) regime of a nanoparticle,3 a new spectroscopy, which emphasizes collective aspects of the nanoparticle, is established. Potential applications of this spectroscopy to soft matter and biological nanoobjects are especially promising, proposing a key advantage in elimination of beam-induced damage. In this paper we study, both theoretically and experimentally, the NF spectroscopy of a uniaxially anisotropic, rectangular nanocrystal, applying two perpendicular 1D scans. The evolution of EEL intensities along these scans exhibits strong electromagnetic (EM) quantum size effects and distinct directionality, the latter proposed as an effective geometrical means for controlling EM interactions between nanoparticles. Experimental spectra were recorded on a VG STEM HB 501 instrument (100 keV), fitted with a Gatan 666 parallelEELS spectrometer.4 Beam diameter was ca. 5 , and the energy width (fwhm of the zero loss peak) 0.5 eV. A collection angle of 6 mrad was determined by the forward direction aperture. For the study of directional effects, 2HMoS2 platelets were chosen, exhibiting both high geometrical
* Corresponding author. Weizmann Institute of Science. Technion-Israel Institute of Technology. Holon Academic Institute of Technolodgy. 10.1021/nl0258022 CCC: $25.00 Published on Web 01/14/2003 2003 American Chemical Society

Figure 1. STEM image of MoS2 platelets distributed at different orientations and suspended into the vacuum, held by electrostatic forces only. Vertically standing platelets, like the ones labled 4 and 5, are studied here spectroscopically.

aspect ratios and considerable anisotropy of the dielectric function in the energy range below 10 eV.5 The platelets were prepared in a gas-phase reaction between MoO3-x and H2S in reducing atmosphere of H2(5%)/N2(95%) at 950 C.6 Using a holey carbon grid, isolated particles suspended by electrostatic forces into the vacuum could be selected, free of support and grid contributions. Figure 1 shows a typical distribution of MoS2 platelets, part of them lay horizontally, with their narrow dimension parallel to the e-beam, while two platelets (labled 4 and 5) are oriented with their narrow face (10 nm in size) perpendicular to the e-beam. The experimental results below correspond to a Vertically standing platelet, ca. 100 100 10 nm in size, and an e-beam propagating along the x axis and progressively

Figure 2. Near-field EEL spectra of MoS2 nanoplatelet in z-scan (b) and y-scan (c) configurations. Values of the impact parameter b are indicated. All calculated spectra (dashed curves) are scaled by the same incoming beam flux parameter. Experimental dielectric function of MoS6 5 is used in the calculations. (d) Loss intensity at p ) 18 eV, plotted for different scans on a log scale vs b: full symbols stand for experimental data, and empty symbols for calculated data. The upper-right panel (a) illustrates the geometry of the beam-platelet system for both the z- and y- scan configurations.

advancing along a z or y line scans, as illustrated in Figure 2a. Spectra of representative pixels along the two line scans are shown in Figure 2b,c, the z-scan advancing along the z-axis toward the large platelet face and the y-scan approaching the platelet narrow face along the y-axis. Comparing yand z-scan spectra, one may observe some spectral differences, yet the most striking observation concerns the spectral intensities, which show significant differences in decay rates. The high energy mode, for example, attenuates toward the y-direction on a length scale about 2 times shorter than the corresponding z-scan attenuation. This result is demonstrated in Figure 2d. A coherent interpretation of EEL spectra may become quite complicated for anisotropic media, as realized in the low energy region in MoS2. Developing the theory of the proposed spectroscopy, we exploit a simplified, classical model of the e-beam, commonly used in the literature.7 In this framework the loss function is calculated from the total power loss W of the electron moving along an unperturbed classical trajectory (selected to be along the x-axis). The corresponding differential scattering cross section 2P/x is related to the power loss per unit path length by dW/dx ) pd(2P/x), where is the excitation frequency. 0 This can be computed from the electric field component along the beam direction, induced in the free space by the current density associated with the e-beam and by the charge oscillations in the nanoparticle. The desired field is obtained from the solution of Maxwells equations, both in the nanoparticle and in the vacuum, with the appropriate boundary conditions. We consider an e-beam propagating with velocity V along the x-axis over the x - y or x - z face of the nanoparticle (corresponding to current density jx(r,t)
204

) - eV(x1 - b)(x2)(x - Vt), with either x1 ) z - c*, x2 ) y describing z scanning, or x1 ) y - b*, x2 ) z describing y scanning). To make things analytically tractable, we assume the beam coordinate x2 ) 0 to be located sufficiently far from the particle edges and its impact parameter b sufficiently small compared to the length of these edges, such that the distortion of the field lines at the edges may be neglected. This assumption simplifies considerably the application of boundary conditions by allowing separation of variables. Inside an anisotropic uniaxial medium with a dielectric tensor: xx ) yy ) (); zz ) | (), the EM waves can be separated into ordinary (o) and extraordinary (e) modes.8 Since the e-mode contains an electric field component parallel to the symmetry axis z, it is a transverse magnetic (TM) mode. The o-mode is polarized in the plane perpendicular to the symmetry axis, which is therefore a transverse electric (TE) mode. Excitation of these modes by an external e-beam in the two scan configurations studied here is essentially different: In a z-scan configuration, consisting of an electric field component perpendicular to the surface, only the e-mode can be resonantly excited, an essential requirement by a relativistic e-beam.9 In the y-scan configuration the situation is more complicated since inside the nanoparticle (|y| e b*, |z| e c* ) both o and e-modes have electric field components perpendicular to the surface. A detailed analysis, which will be presented elsewhere,10 yields for the z-scan differential excitation probability: d2P e2 ) 2 2 ddx 2 V p
0

(Rz2 - qo2) (2 - 1) z + Rz [2qoRz coth(2qoc*) + (Rz2 + qo2)] z z z Rz[( Rz)2 - qe2] z k2[2 Rzqe coth(2qec* ) + ( Rz)2 + qe2] z z z

0 dky exp(-2Rzb) Im

{[

Rz k2

(1)

where ) V/c ) 0.54, Rz ) k2-2/c2 is the spatial attenuation parameter in the free space above the x - y face, qo ) k2- 2/c2 and qe ) ( / |)(k2- |2/c2) are the z z wavenumbers for the propagation of the o- and e-mode, respectively, along the z-axis inside the medium, and k2 ) kx2 + ky2. The value of kx in the above expression is fixed by the conditions of conservation of energy and momentum along the beam direction, kx ) /V (see below, however). For the y-scan configuration the exact solution is very involved due to the mixture of the o- and e-modes mentioned above. A relatively simple, closed analytical expression can be derived10 for a semi-infinite medium, which is a good approximation when a large aspect ratio b* . c* is realized. Calculated loss functions, using an experimentally derived dielectric function,6 are presented in Figure 2. The reader should note the fairly good reproduction of the corresponding experimental EEL intensities, particularly in the z-scan configuration, and the significantly better agreement with the intensity decay rates at both scans. The linear fits in Figure 2d yield a directional ratio, Ry/Rz, of about 2. The
Nano Lett., Vol. 3, No. 2, 2003

Figure 3. Surface plasmon-polariton dispersion lines, calculated for isotropic dielectric film. Three oscillators are used for the dielectric tensor. Damping parameters are neglected. The normalization parameters are pn ) 10 eV, kn ) n/c. Bound frequencies (1,2,3) are indicated by horizontal dashed lines, the light line (LL) ) ck by the dashed oblique line, and the e-beam line (eL) ) Vk by the dotted line (corresponding to V/c ) 0.54 or 100 keV beam energy). Symmetric (antisymmetric) SPP branches are indicated by s* (s). Note the appearance of waveguide modes (indicated by w*) within the energy gaps of the SPP bands along the LL. The inset shows an enlarged portion of the dispersion lines above 1.

Figure 4. z-Scan configuration EEL spectra, calculated for a MoS2 platelet of thickness 2c* ) 40 with a three-oscillator dielectric tensor at various impact parameters: b ) 10, 30, 50 . Note the increasing energy splitting in the high-energy band, due to the momentum space filtering effect. The inset shows a similar spectrum for a very thin platelet, 2c* ) 12 , at b ) 36 , calculated with experimentally derived dielectric tensor.5

origin of this effect is in the smaller size of the particle along the z-axis and the strong quantization of the corresponding wavenumber kz. The minimal attenuation parameter along the z-scan is Rmin ) (/V)(1 - 2)1/2, while along the y-scan z 2 Rmin ) (/V)[V2kz,min/2 + (1 - 2)]1/2. Here a cutoff kz,min y 1/2 (c*/V) (1/c*) is determined by the boundary conditions at z ) ( c* (ref 10). For a platelet 100 thick we find the theoretical estimate Ry/Rz 1. 95, in a very good agreement with the experimentally derived value. In the y-scan configuration the agreement between theory and experiment is significantly poorer than in the z-scan, and it quickly deteriorates as the impact parameter b increases. This may arise from the complete neglect of the platelet corners in our calculation, a serious shortcoming in the y -scan configuration when b becomes comparable to the narrow size c*. The salient features of the loss function can be derived from the dispersion relations of the various collective modes, defined by the poles of the integrand in eq 1. As shown in Figure 3, the high-energy region above 12 eV is dominated by a single, strongly dispersed surface-plasmon-polariton (SPP) mode. In this spectral region the dielectric tensor is Drude-like, i.e., exhibiting weak dispersion and anisotropy. The low-energy region below 10 eV in MoS2 represents a qualitatively different situation: Re[ ()] is strongly dispersive and highly anisotropic, generally smaller in magnitude than Im[ ()]. Also, the electronic interband transitions determine in this region a set of spectrally close singleelectron oscillators, which are coupled by the common EM field to yield several polariton branches with limited spatial dispersion. A simple phenomenological model of the threeoscillator dielectric function is used in Figure 3, roughly
Nano Lett., Vol. 3, No. 2, 2003

reproducing the measured dielectric tensor.5 Note that fine structures which arise from the dielectric function anisotropy10 are not resolved in the above experimental spectra due to smearing by a relatively large damping of the dielectric function. A remarkable property of the STEM-EEL spectroscopy concerns the close proximity of the electron beam line (eL), defined by ) Vk, to the light line (LL), ) ck. The former line sets the scale of the momentum transfer in the scattering event, while in the close vicinity of the latter the dispersion curve undergoes a major variation. At least two additional length parameters are involved in the loss process: The impact parameter, b, and the object dimensions, e.g., the platelet thickness, 2c*. The impact parameter imposes a further constraint on the momentum transfer, k j kb ) 1/b, as larger momenta are effectively suppressed by the exponential factors in the loss function. Hence it may be used as a variable low-pass filter in momentum space. The object size parameter determines a region in k-space where the antisymmetric branch of the dispersion curve undergoes a major deviation from the symmetric branch (see Figure 3). The unique feature of the NF EEL spectroscopy is that all these parameters happen to vary within the same length scale, their interplay can lead to striking effects. Thus, by varying the impact parameter near the characteristic e-L wavelength, k-1 V/, the loss function exhibits strong frequency dependence, which reflects the strong k-dispersion in the e-L region, and so results in significant sensitivity of the EEL signal pattern to the finite size of the nanoparticle. A nice demonstration of this effect is shown in Figure 4, where a well-known finite size splitting, due to the interaction between the two parallel slab faces,9 is detected by varying b. This splitting starts to be visible at b 2c*, when the loss function is dominated by wavelengths comparable to the particle size. The growing energy splitting with increasing
205

Figure 5. Calculated loss functions for a 2H-MoS2 nanoplatelet in a z-scan configuration, including a quantum mechanical correction for the finite platelet size: 2a* ) 200, 600, 1000 , and . The other parameters used are 2c* ) 100 , V/c ) 0.54, and b ) 30 . Experimental dielectric tensor of MoS26 is used.

b, shown in Figure 4, is associated with a shift of kb toward smaller wavenumbers, where the effective energy gap between the symmetric and the antisymmetric branches is highly pronounced. A calculated spectrum for c* ) 6 , shown in the inset to Figure 4, agrees remarkably well with the experimental data presented in ref 11, taken from a two-shell WS2 nanotube. In view of the specific differences between the dielectric functions of the two materials, the similarity of the overall shape of the spectra reflects the collective EM nature of the proposed spectroscopy. The close proximity between the e-L and the LL leads to another striking effect. In Figure 5, the shape of the calculated loss function undergoes drastic changes by varying a*, the platelet size along the e-beam direction (x-axis). The driving force for this effect is the breakdown of the strict requirement of momentum conservation, kx ) /V. This constraint is inherent to the rigid classical trajectory model used above for the beam electron, which ignores momentum uncertainty originating from the finite size of the target object. We account for this uncertainty by integrating over a continuous distribution function of (kxV - ), its form determined by comparing the classical expression, eq 1, with the one derived in a full quantum mechanical approach, ref 3. The reduction in power loss, seen in Figure 5, results from momentum transfer to the entire platelet rather than to its internal degrees of freedom. It becomes particularly strong in the energy region near V/a*, where the momentum smearing kx 1/a* approaches the e-L momentum /V. Thus, by varying the platelet size a*, the shape of the entire

electron loss spectrum changes drastically as compared to the classical one. One should note that for sufficiently small values of a* there should be significant contributions to the integral over kx for which the values of Rz are purely imaginary, and the SPP dispersion lines could penetrate beyond the SPP lightline ) ck. The e-beam couples then to radiative modes of the platelet, and hence the signal dependence on b is expected to change into smoothly decaying (nonexponential) behavior. Electron-beam coupling to radiative modes is known in the context of patterned surfaces, for example.9 This striking situation in nanoparticles requires a further study. In conclusion, two different quantum size effects have been presented using NFEELS. The first effect arises from sample finite dimensions perpendicular to the e-beam path, imposing stronger attenuation in corresponding (near field) directions. The second effect, associated with the finite object size along the beam direction, introduces uncertainty in kx, which alters significantly the overall EEL spectrum. We have shown that the experiment provides an effective filter in momentum space, varied via the impact parameter, thus applicable to the study of size effects. A theoretical model developed for uniaxial, strongly dissipative media, accounts well for the quantitative variations in signal intensity near the nanoplatelet and manifests the essentially collective EM nature of the probed excitations. Acknowledgment. We thank David Muller for helpful discussions. O. Stephan and C. Colliex are acknowledged for kindly providing us with experimental data. This research was supported by the BIKURA foundation of the Israel Academy of Science and Humanities, by the Association Franco-Israelienne pour la Recherche Scientifique et Technologique, and by the KAMEA foundation of the Center for Absorption in Science, Ministry of Immigrant Absorption, Israel. References
(1) Colliex, C. in Transmission EELS in Material Sciences; Disco, M. M., Ahn, C. C., Fultz, B., Eds.; The TMS Monographs Ser. 2 85; TMS: Warrendale, 1992; p 185. (2) Howie, A.; Milne, R. H. Ultramicroscopy 1985, 18, 427. (3) Cohen, H.; Maniv, T.; Tenne, R.; Rosenfeld Hacohen, Y.; Stephan, O.; Colliex, C. Phys. ReV. Lett. 1998, 80, 782. (4) Tence, M.; Quartuccio, M.; Colliex, C. Ultramicroscopy 1995, 58, 42. (5) Zeppenfeld, K. Optics Commun. 1970, 1, 377. (6) Feldman, Y.; Wasserman, E.; Srolovitz, D. J.; Tenne, R. Science 1995, 267, 222. (7) Wang, Z. I. MICRON 1996, 27, 265-299. (8) Landau, L. D.; Lifshitz, E. M. Electrodynamics of continuous media; Pergamon Press: New York, 1960. (9) Heinz Raether; Surface Plasmons on Smooth and Rough Surfaces and on Gratings; Springer: New York, 1988. (10) Lembrikov, B. I.; Itskovsky, M. A.; Cohen, H.; Maniv, T., to be published. (11) Kociak, M. et al., Phys ReV. Lett. 2001, 87, 075501.

NL0258022

206

Nano Lett., Vol. 3, No. 2, 2003

Vous aimerez peut-être aussi