Surface Science 502–503 (2002) 224–231 www.elsevier.com/locate/susc
Scanning probe energy loss spectroscopy R.E. Palmer *, B.J. Eves, F. Festy, K. Svensson Nanoscale Physics Research Laboratory, School of Physics and Astronomy, The University of Birmingham, Edgbaston, Birmingham B15 2TT, UK
Abstract This article gives a status report on the new technique of scanning probe energy loss spectroscopy (SPELS), which can be viewed as a hybrid between scanning tunnelling microscopy and electron energy loss spectroscopy. The technique provides local energy loss spectra from solid surfaces (at room temperature) with an expected spatial resolution of 1–10 nm and a current best energy resolution of 0.6 eV. Spectra are presented from the graphite and Si(1 1 1)-7 7 surfaces as well as metallic (silver and gold) thin films, which enables the chemical analysis capabilities of the technique to be demonstrated. New and future developments are also discussed, including the demonstration of SPELS with screened tips and the prospects for spectral measurements with energy resolution in the vibrational regime. Ó 2002 Elsevier Science B.V. All rights reserved. Keywords: Scanning tunneling microscopy; Electron energy loss spectroscopy (EELS); Graphite; Silicon; Silver; Gold
Most of the experimental techniques which are currently employed to obtain vibrational spectra from solid surfaces, such as high resolution electron energy loss spectroscopy (EELS) and reflection absorption infra-red spectroscopy, collect data from an extended region of the sample surface. By contrast, the family of scanning probe microscopes (SPMs), generated from the original invention of the scanning tunnelling microscope (STM) in the early 1980s, offers excellent lateral spatial resolution, typically down to the nanometre-scale regime and, in the case of the STM itself, to the level of individual atoms. Thus it is natural for a spectroscopist to ask the question ‘‘what kind of local spectroscopic information can be obtained with the help of SPM technology?’’.
*
Corresponding author. Fax: +44-121-414-7327. E-mail address:
[email protected] (R.E. Palmer).
An outstanding example of innovation in this area is provided by the work of Ho and colleagues in their development of inelastic electron tunnelling spectroscopy with the STM tip (STM-IETS) [1], so that vibrational spectra can be obtained with atomic resolution. However, even this technique is restricted––to cryogenic (liquid helium) temperatures, since the energy resolution scales with the thermal energy (kT)––which imposes some limits on the range of problems which can be investigated as well as the accessibility of the technique. In this paper we review the recent development of a new technique which is applicable at room temperature, and which we call ‘‘scanning probe energy loss spectroscopy’’ (SPELS) [2]. The concept of SPELS is illustrated schematically in Fig. 1; the technique is a hybrid of STM and EELS. The STM tip is retracted from the surface by a distance of, say, 100 nm and operated in the field emission mode, so that the bias voltage applied between tip
0039-6028/02/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 9 - 6 0 2 8 ( 0 1 ) 0 1 9 4 9 - 5
R.E. Palmer et al. / Surface Science 502–503 (2002) 224–231
225
Fig. 1. Schematic diagram showing the experimental configuration for SPELS, in which electrons field emitted from an STM tip are backscattered from the sample surface and collected with an angle-resolved electron energy analyser.
and surface is, for example, 100 V. Under these conditions a proportion of the electrons incident upon the surface will be scattered back from the surface and can be collected with an electron energy analyser. Thus a local EELS measurement can be conducted. One can also envisage measurements in which the analyser is set to collect a particular energy loss, or to scan repeatedly over a specific region of the EELS spectrum, while the tip is scanning over the surface. Of course, the tip can also be employed to acquire traditional STM images in the tunnelling regime before or after the SPELS measurements. What are the likely limits of the spatial and energy resolution of the SPELS technique? An idea of the spatial resolution is provided by Fig. 2, which shows some results of theoretical simulations which calculate the trajectories of (elastically scattered) electrons in a SPELS configuration; 1 details of the simulations will be given elsewhere. Fig. 2(a) shows a sample of electron trajectories incident upon the sample after field emission from the tip; the field emission process is described by the Fowler–Nordheim equation [4], which pre-
1 Laplace’s equation was solved by the boundary element method [3a], and the electron trajectories were obtained by solving the equations of motion with a Runge–Kutta 5(4) integration scheme [3b].
Fig. 2. Simulations of (elastic) electron trajectories in the SPELS configuration, with tip radius 28 nm, tip–surface distance 100 nm, tip bias voltage 100 V and emission current 1.16 nA. (a) Sample of electron trajectories incident on the sample after Fowler–Nordheim field emission from the tip. (b) Corresponding current density profile on the sample––the ‘‘illuminated spot’’. (c) Current density profile for (only) those electrons subsequently detected, after backscattering, in an angle-integrated fashion. (d) Current density profile for (only) those electrons subsequently detected in an angle-resolved measurement, with the analyser angle 85 0:5° from the surface normal.
sumes a free electron density of states in the tip. Note that the upper regions of the tip are screened by a co-axial grounded shield in the simulation, which damps out the long range effect of the tip field (see below). Fig. 2(b) plots the current density arriving at the surface; note that the width of the current density profile (the FWHM is 54 nm) is of
226
R.E. Palmer et al. / Surface Science 502–503 (2002) 224–231
the same order as the tip–sample separation (100 nm). At first sight, one might think that this width (i.e. the size of the illuminated spot) defines the likely spatial resolution of the technique, but this is not the case, as demonstrated in Fig. 2(c). Fig. 2(c) shows the current density profile on the surface of (only) those electrons which are subsequently reflected from the surface and collected by the analyser (integrating over the collection angle). The reason why the current density profile in this case is so strongly suppressed underneath the axis of the tip is that electrons reflected from the surface in this region are strongly influenced by the same field which accelerated them to the surface. In particular, the reflected electrons are bent back to the surface itself, as noted previously by V€ olkel et al. [5] in the context of electron beam writing with such tips. If we make the assumption (for purposes of simplicity) that the reflectivity for all electron scattering events from the surface is 1%, then we obtain the ‘‘black hole’’ underneath the tip in Fig. 2(c). What we now see is an annulus around the centre of the tip from which we can collect reflected electrons; the width (FWHM) of the annulus in Fig. 2(c) is 10 nm, and is a more likely indication of the spatial resolution achievable in SPELS. Fig. 2(d) illustrates a further interesting effect, since it shows a plot of the sample current density corresponding only to those electrons subsequently reflected and collected by an analyser positioned at a particular polar angle (85 0:5°) with respect to the surface. The effect of the selected polar angle (differential measurement) is to filter further the set of electrons incident upon the sample, leading to an annulus of width only 1.3 nm. While these simulations makes a number of assumptions about both tip and surface, e.g. the tip is perfectly symmetrical and the surface smooth, they do suggest that a spatial resolution in the range 1–10 nm may be achievable in SPELS measurements. Such a resolution suggests possible applications relevant not only to the field of microchip fabrication, where feature sizes are currently in the range 100–200 nm and predicted to reach 50 nm or below in the next 10 years, but also in the general area of ‘‘nanostructured surfaces’’, i.e. surfaces with controlled lateral structures on the length scale 1–100 nm [6–8].
The likely energy resolution in SPELS is determined by the process of field emission from the tip, which replaces the monochromator in a conventional EELS spectrometer. The width of the energy distribution from the tip is about 300 meV [9], and arises from the energy dependent probability for tunnelling from the occupied states of the tip out through the field emission barrier. The width is weakly dependent on temperature (because of Fermi level smearing), falling to about 200 meV at 5 K. With an energy resolution of this order, one would be well placed to make local measurements of electronic excitations in the range from, say, 0.5 to 50 eV, corresponding, in effect, to a local optical absorption spectrum from the IR through the visible to the UV and VUV, a region rich in plasmon excitations, interband transitions and molecular electronic excitations. By contrast, with this kind of resolution vibrational measurements are probably not feasible, except perhaps in special cases. However, if one could produce a tip with narrow energy bands (rather than a free electron density of states) arising, say, from spatially localised electronic states, it should be possible to extract from the tip a much narrower distribution of electron energies. The literature already proves the viability of this principle, since Binh and Purcell [10] have demonstrated an energy width in field emission from platinum nanotips of only 65 meV. The exploitation of such a tip in SPELS would thus appear to open the way to local vibrational measurements at room temperature. Having given some consideration to the likely limits of performance of the SPELS technique, we now turn to an experimental demonstration of SPELS measurements, with emphasis on the information content of the resulting local spectra; much of the data presented is new with respect to Ref. [2]. The spectra were collected at an emission angle of 90° from the surface normal, i.e. parallel to the surface plane. We begin with Fig. 3, which shows a series of SPELS spectra obtained with an electrochemically etched, polycrystalline tungsten STM tip and a hemispherical electron energy analyser, a distance of 4 cm from the surface; details of the experimental set-up will be given elsewhere [11]. Fig. 3(a) shows a SPELS spectrum from graphite (HOPG); the energy loss peak ob-
R.E. Palmer et al. / Surface Science 502–503 (2002) 224–231
227
Fig. 3. Energy loss spectra obtained in SPELS from (a) graphite and (b,c) different regions of an evaporated silver film on graphite. The incident electron energy (i.e. tip bias voltage) and sample current are given in each case. The assignment of the energy loss peaks is discussed in the text.
served at 8 eV can be assigned to the graphite p band plasmon, also reported in our first SPELS paper. 1 Figs. 3(b) and (c) show SPELS spectra from an evaporated film of silver on graphite; specifically, half the surface was masked during the metal evaporation. These two spectra, obtained at arbitrary points in this system, clearly show new loss features which can be assigned naturally to silver plasmon modes. The energy loss peak at 3.4 eV in Fig. 3(b) is not far from the surface plasmon frequency of a single crystal silver surface, 3.66 eV [12], or a film of silver nanoparticles on graphite, 3.85 eV [8], while the peak at 2.2 eV in Fig. 3(c) might be associated with non-spherical silver par-
ticles [13]. We do not at this stage attempt to correlate the local spectra obtained with microscopic (STM or scanning SPELS) measurements, but rather to emphasise the information content of the spectra and the obvious dependence of the spectra on surface site sampled. Note that the absolute count rates in the spectra lie typically in the range 1–10 kHz for the elastic peak (though closer to 100 kHz has been obtained on occasion), and 100 Hz to 1 kHz for the energy loss peaks. The energy resolution in the spectra of Figs. 3–6 typically lies in the range 1–2 eV, because the analyser pass energy was usually chosen to ensure good signal levels with a modest incident current (down
228
R.E. Palmer et al. / Surface Science 502–503 (2002) 224–231
Fig. 4. Energy loss spectra obtained in SPELS from (a) an evaporated gold film on mica and (b) a composite nanoparticle film consisting of gold and silver clusters. The incident electron energy (i.e. tip bias voltage) and sample current are given in each case. (c) compares the spectra from the Au/Ag film and the Au film, highlighting the silver plasmon feature between 3 and 4 eV.
to 10 nA) rather than to optimise the resolution. However, the FWHM of the elastic peak has been reduced to 0.6 eV, as illustrated by Fig. 3(a), where the elastic peak count rates is still 1 kHz. The chemical analysis capabilities of SPELS are illustrated further in Fig. 4, which shows spectra acquired from an evaporated film of gold on mica (Fig. 4(a)) and a patterned surface consisting of a mixture of gold and silver particles (Fig. 4(b)). The latter was prepared by the ‘‘template stripping’’ method [14]. The spectrum from the gold film (Fig. 4(a)) does not show very pronounced energy loss features, but broad bands centred at 5.5 and 15 eV are apparent. The SPELS spectrum from the mixed Au/Ag surface also shows these features, but in addition a well defined loss peak is visible at
3.6 eV, similar to the feature observed at 3.4 eV in the case of the silver particles on graphite (Fig. 3(b)). This suggests the notion that the Au/Ag spectrum can be replicated by the addition of individual Au and Ag spectra, and is supported by Fig. 4(c), which shows how a simple multiple of the Au spectrum from Fig. 4(a) reproduces the features of the Au/Ag spectrum of Fig. 4(b) except for the Ag peak at 3.6 eV. Fig. 5 shows that SPELS is not only sensitive to plasmon modes. Fig. 5(a) shows the SPELS spectrum of the Si(1 1 1)-7 7 surface and explores what energy loss features lie within the broad band of excitations visible in the spectrum. To do this, we exploit the method employed in some of the early measurements with conventional EELS, in
R.E. Palmer et al. / Surface Science 502–503 (2002) 224–231
229
Fig. 5. (a) SPELS spectrum from a Si(1 1 1)-7 7 surface; incident energy 200 V, sample current 10 nA. (b) Second derivative of the spectrum in (a), for comparison with (c), which shows the second derivative of a conventional EELS spectrum from a Si(1 1 1)-7 7 surface (from Ref. [15]). The spectral feature assignments are discussed in the text.
which the second derivative of the signal (N) with respect to energy (E) was plotted to unravel the contributions to the spectrum [15]. Technically, peaks in the second derivative––which highlights the curvature of the original spectrum––are believed to match (underlying) peaks in the original spectrum; note also that the intensities in the second derivative spectrum are not representations of real spectral feature intensities. Fig. 5(b) is a plot of d2 N =dE2 obtained from the Si SPELS spectrum of Fig. 5(a), while Fig. 5(c) is a plot of d2 N =dE2 obtained from an EELS study of a Si(1 1 1)-7 7 surface [15]. The degree of correspondence between the ‘‘decomposed’’ spectra is quite striking. In particular, comparison with the EELS assignments indicated that the second derivative SPELS spectrum contains not only the surface and bulk plasmon modes (at 10 and 17 eV, respectively) but also surface state interband transitions (S2 and S3 at 7.5 and 15 eV) and (at least) one bulk interband transition (E2) at 5 eV. Thus SPELS measurements are expected to reflect the whole gamut of
Fig. 6. Results obtained from the Si(1 1 1)-7 7 surface showing the effect of a screened tip (for details see text); incident energy 200 V, sample current 20 nA (screened tip) or 10 nA (regular tip). (a) Elastic peak intensity in the SPELS spectrum versus angle (here measured with respect to the surface plane) for the two tips, normalised at 0°. (b) SPELS spectra for the two tips, normalised to the elastic peak intensity.
electronic excitations at surfaces, while the second derivative analysis method may be helpful in drawing attention to individual contributions to the energy loss spectrum. As noted in our previous paper [2], the SPELS instrument is similar in concept to the ‘‘field emission scanning Auger microscope’’ reported in the 1980s, which was (briefly) employed to obtain Auger and energy loss spectra at high incident energies (547 eV) and relatively large tip–sample distances (a few hundred lms) [16]. At least two other groups have also explored this idea [17,18]. Important features of the present program include (i) the ability to make angle-resolved measurements, (ii) the recognition, based on theoretical
230
R.E. Palmer et al. / Surface Science 502–503 (2002) 224–231
simulations, that the trajectories of the backscattered electrons will be bent by the tip field, even at large distances from the STM-surface junction, such that the spectral intensity is peaked very close to the surface plane, and (iii) the control of the tip shape, which, together with (i) and (ii), allows the detection of significant backscattered signal even when the tip is only 100 nm from the surface. A further development in the tip technology is illustrated in Fig. 6, which compares angular measurements from the Si(1 1 1)-7 7 surface obtained using a regular tungsten tip with those obtained using a (prototype) ‘‘screened’’ tip. The screened (or ‘‘co-axial’’) tip was created by painting the tip shaft with, first, a layer of electrically insulating glue and, then, a layer of colloidal silver, down to within about 1 mm from the tip apex. Even the modest degree of long range electrostatic screening expected to arise from such a tip appears to have an impact on the experimental results. Fig. 6(a) indicates that the FWHM of the angular distribution of the backscattered elastic signal increases from 1.5° to 3° and also tails off with increasing angle from the surface much more slowly, while Fig. 6(b) suggests that the intensity of the energy loss features in the SPELS spectrum (obtained parallel to the surface) relative to the elastic peak is enhanced by the screened tip (consistent with the idea that the trajectories of electrons with lower kinetic energy are more susceptible to the influence of the long range tip field). These first results with a rather simple kind of screened tip suggest that in future the deployment of more sophisticated, microfabricated screened tips may provide valuable benefits to SPELS measurements. Enhanced count rates may be anticipated, as well as the prospect of collecting spectra further away from the sample plane, which may be especially helpful in the case of rough surfaces; the introduction of a microlens stack between tip and surface would be expected to have similar effects. Major enhancements in analyser detection efficiency also appear to be within reach; the addition of, say, a 128 channel multichannel detector to a standard dispersive electron energy analyser may lead to a decrease of two orders of magnitude in the time needed to collect a spectrum, likely to be particularly valuable in scanning measurements,
while some kind of time-of-flight analyser may give rise to even greater detection efficiencies. The application of SPELS to molecular or biomolecular systems would also benefit from improvements in detection efficiency, since in this case possible sample damage is an important consideration; detection efficiency can be traded against incident sample current. Note, however, that the sample current density employed at present (down to 10 nA in Fig. 3(b) and (c) over a simulated area of order 100 100 nm2 – while we have also collected spectra with a sample current down to 1 nA) is very much lower than in traditional STM measurements (although the ‘‘impact energy’’ is much higher). In summary, we have considered the ultimate limits of performance of the new SPELS technique, including the prospects for vibrational measurements at room temperature and spatial resolution approaching 1 nm. Angle-resolved measurements, especially close to the surface plane, theoretical simulations of the electron trajectories and good control of the field emitting tip are crucial components in this work. Experimental SPELS spectra reported from graphite, gold and Si(1 1 1)-7 7 surfaces as well as silver films on graphite and mixed Ag/Au nanoparticle films reveal the chemical identity of the substrates investigated, and also the additive nature of the spectral contributions in the case of mixed systems. The microfabrication of field emitting tips and the adoption of efficient electron detection schemes, e.g., multichannel detectors or time-of-flight analysers, offer significant opportunities for future instrument developments, raising at least the possibility of collecting broad band, spatially resolved energy loss spectra while scanning at standard STM rates as well as single molecule vibrational spectra in dilute surface systems at room temperature.
Acknowledgements We thank W. Frey for providing the Au/Ag samples prepared by the template stripping method, and acknowledge helpful discussions with P.R. Preece, as well as the contribution of P. Laitenberger to the early stages of this program from 1992 to 1997. B.J.E. is grateful to the ORS Award
R.E. Palmer et al. / Surface Science 502–503 (2002) 224–231
Scheme and the School of Physics and Astronomy, University of Birmingham for financial support.
[7]
References [8] [1] B.C. Stipe, M.A. Rezaei, W. Ho, Science 280 (1998) 1732; Rev. Sci. Instrum. 70 (1999) 137. [2] B.J. Eves, F. Festy, K. Svensson, R.E. Palmer, Appl. Phys. Lett. 77 (2000) 4223; F. Festy, K. Svensson, P. Laitenberger, R.E. Palmer, J. Phys. D: Appl. Phys. 34 (2001) 1849. [3] (a) R.L. Hartman, W.A. Mackie, P.R. Davis, J. Vac. Sci. Technol. B 12 (1994) 754; (b) W.H. Press, S.A. Teukolsky, W.T. Vetterling, B.P. Flannery, Numerical Recipes in C, second ed., Cambridge, Cambridge, 1992. [4] R. Gomer, Field Emission and Field Ionization, Harvard University Press, Cambridge, 1961. [5] B. V€ olkel, A. G€ olzh€auser, H.U. M€ uller, C. David, M. Grunze, J. Vac. Sci. Technol. B 15 (1997) 2877. [6] S.J. Carroll, S. Pratontep, M. Streun, R.E. Palmer, S. Hobday, R. Smith, J. Chem. Phys. 113 (2000) 7723; S.J. Carroll, P.D. Nellist, R.E. Palmer, S. Hobday, R. Smith, Phys. Rev. Lett. 84 (2000) 2654; J. Liu, J.C. Barnard, K. Seeger, R.E. Palmer, Appl. Phys. Lett. 73 (1998) 2030;
[9] [10] [11] [12] [13]
[14] [15] [16] [17] [18]
231
I.M. Goldby, L. Kuipers, B. von Issendorff, R.E. Palmer, Appl. Phys. Lett. 69 (1996) 2819. H. Osman, J. Schmidt, K. Svensson, R.E. Palmer, Y. Shigeta, J.P. Wilcoxon, Chem. Phys. Lett. 330 (2000) 1; P.J. Durston, J. Schmidt, R.E. Palmer, J.P. Wilcoxon, Appl. Phys. Lett. 71 (1997) 2940. C.-M. Grimaud, L. Siller, M. Andersson, R.E. Palmer, Phys. Rev. B 59 (1999) 9874. J.W. Gadzuk, E.W. Plummer, Rev. Mod. Phys. 45 (1973) 487. V.T. Binh, S.T. Purcell, Appl. Surf. Sci. 111 (1997) 157. B.J. Eves, Ph.D. Thesis, University of Birmingham (2002). M. Rocca, Surf. Sci. Rep. 22 (1995) 1. T.R. Jensen, M.L. Duval, K.L. Kelly, A.L. Lazarides, G.C. Schatz, R.P. Van Duyne, J. Phys. Chem. B 103 (1999) 9846; J.R. Krenn, G. Schider, W. Rechberger, B. Lamprecht, A. Leitner, F.R. Aussennegg, J.C. Weeber, Appl. Phys. Lett. 77 (2000) 3379. W. Frey, C.K. Woods, A. Chilkati, Adv. Mater. 12 (2000) 1515. J.E. Rowe, H. Ibach, Phys. Rev. Lett. 31 (1973) 102. B. Reihl, J.K. Gimzewski, Surf. Sci. 189 (1987) 36. M. Tomitori, H. Terai, T. Arai, Appl. Surf. Sci. 144 (1999) 123. H. Ibach, private communication.
