X-ray photoelectron spectroscopy with a laser-plasma source

Transcription

1 Proc. SPIE Vol.3157 (1997) pp X-ray photoelectron spectroscopy with a laser-plasma source Toshihisa TOMIE a, Hiroyuki KONDO b, Hideaki SHIMIZU a, and Peixiang Lu a a Electrotechnical Laboratory, Tsukuba, Ibaraki, Japan b Nikon Tsukuba Research Laboratory, Tsukuba, Ibaraki, Japan ABSTRACT For the study of future nm devices, electronic and chemical states must be observed with nm resolution, and a photoelectron microscope is strongly desired to be developed. A compact system for in-house analysis can be realized with a laser-plasma x-ray source. Experimental results confirm that an x-ray photoelectron microscope with a laser-plasma can become a powerful tool with potential spectrum acquisition time comparable to that with a synchrotron source. Reported are high energy resolution photoelectron spectra for Si revealing the existence of surface states, and 2-dimensional mapping with resolution of 6 µm demonstrating that simultaneous detection of multi-elements by the time-of-flight energy analysis is powerful in obtaining a reliable mapping. High electron detection efficiency of the system with a pulse source is discussed to be essentially important for nm resolution. 1. INTRODUCTION The size of semiconductor market has been growing exponentially with annual growth rate of 17%, and is expected to reach 2 billion dollars in 2. This market growth is technologically supported by continuous reduction of feature size of integrated circuits which is expected to reach.1 µm between years of 22 and 26. In fabricating fine resolution features in semiconductor devices, electron beam etching or reactive ion etching are known to induce electronic damages on the surface. The thickness of the damaged layer is of the order of 1 nm and increases with etching duration. Although this etching-induced damage can be neglected when the feature size is large, it will become serious when the feature size reduces to around.1 µm. Further in nm devices, extremely low concentration impurities could affect the performance, and the role of surface and interfaces become important. Therefore, detection of electronical defects or low concentration impurities and analysis of surfaces and interfaces are very important in developing future semiconductor devices. As for atomic scale spatial resolution, various kinds of microscopes, such as scanning (SEM) or transmission electron microscopes (TEM), scanning tunneling microscopes (STM), atomic force microscopes (AFM), and scanning near field optical microscope (SNOM) are available. However, all of them give information just on physical structures, but not on electronical states. For example, SEM, STM, AFM, and SNOM give images of surface relief, and TEM and x-ray microscope give information on mass density. For the analysis of electronic states, photoelectron spectroscopy is the best method. In photoelectron spectroscopy, highly monochromatic light irradiates a sample, and energy distribution of electrons ejected to vacuum is analyzed to get distribution of electronic states in the sample. X-ray photoelectron spectroscopy (XPS) in which x-ray is employed to excite electrons in inner shells of atoms, is especially useful because we can identify element surrounding which the electronic states are observed. However, the best resolution of commercial in-house system with an x-ray tube is only 1 µm 1) which is far from the requirement in the study of.1µm semiconductor devices. A microscope based on XPS should be the one to be developed for the study of nm devices. Extremely brilliant x-ray source is required in XPS microscope, and all studies have been performed at synchrotron facilities. In this paper, we describe how a low average-power Fig.1 Photoelectron spectroscopy gives information on electronic states of materials

2 laser-plasma source can serve as an x-ray source for XPS, report the experimental demonstration of observing spin-orbit splitting of Si 2p electrons and high precision 2D mapping of a gold dot on a Si wafer with 6 µm resolution. Finally, we discuss that a laser-plasma source is the best x-ray source for nm resolution. 2. ISSUES IN XPS MICROSCOPE In XPS microscope, spatial mapping of electronic states is obtained by scanning x-ray micro-beam on sample. Micro-beam of tens nm diameter can be produced with use of a zone-plate or multi layer-coated Schwarzschild optics which work in the photon energy range from several tens to several hundreds ev. Therefore, XPS microscope of tens nm resolution should be possible, in principle. However, practical resolution limit is determined by other factors as discussed later in detail in 5.DISCUSSIONS. In an XPS microscope, a monochromatic x-ray source and a high resolution electron energy analyzer are more important, in some sense, than x-ray focusing optics, because the best advantage of XPS is in its capability of observing chemical states of elements. In order to distinguish chemical shifts of elements, 1 ev is required for the energy resolution, and very often.1 to.3 ev resolution is required to get more detailed information as will be shown in section 4.2. To get.2 ev resolution for 2 ev photons, monochromaticity λ/ λ of the source should be better than 1, which is larger than that in a commonly used grating spectrometer. At high energy resolution, utilizing efficiency of an x-ray source and detection efficiency of photoelectrons are very low. Therefore, a highly brilliant x-ray source is required for reasonably large detection rate of photoelectrons. Because of this extremely low efficiency of the method, XPS microscopy is being studied so far only at huge synchrotron facilities 2). Even bending synchrotron radiation (SR) is not bright enough and extremely brilliant undulator source is required for sub-µm resolution. Therefore, general belief is that any in-house system using a compact x- ray source will be practically of no importance in sub-µm resolution. However, in-house inspection is crucially important for the study of nm devices, because taking samples out of a fabrication chamber for the inspection could introduce non-negligible impurities for nm devices. Therefore, a compact inhouse XPS microscope is strongly desired to be developed. Moreover, presently developed systems with an undulator source will encounter a serious radiation damage problem in tens nm resolution, as is discussed later in 5.DISCUSSIONS. These problems will be solved by employing a laser-plasma as an x-ray source, as explained in next section. 3. MEANS FOR MAKING XPS WITH A LASER-PLAMA SOURCE A PRACTICAL SYSTEM Laser-plasma x-ray source (LPX) is well known as a compact and high brilliance x-ray source. As shown in Fig.2, the peak brilliance of LPX is higher than that of a bending synchrotron radiation (SR) source and comparable to that of an undulator SR source. In fact, tens nm resolution x-ray images of living cells in solution can be taken only with LPX, because extremely short exposure time is indispensable to avoid the effect of instant vaporization of the specimen within a few ns 3) under the dose for high resolution imaging. However, emitting duration of LPX is extremely short as a few ns, and hence, the duty ratio is as low as 2x1-6 even at 1 khz rep-rate operation of the laser, and the time-averaged photon flux is not high. Therefore, mere replacement of a SR source with LPX cannot make a XPS microscope with LPX a practical tool although the system is compact. Low average-power of LPX was considered fatal in XPS and there was no try of applying LPX to XPS before the proposal by Tomie 4). In 1992, Tomie 4) proposed that an XPS system employing a laser-plasma x-ray source (LPX) can become a practical system with the aid of two magic. Things appeared impossible is changed to reality by magic, but very often the secret of magic is simple after Fig.2 Peak brilliance of various x-ray sources.

3 it is made open. The first magic is the monochromatization of LPX without employing a grating. Transmission efficiency of a grating spectrometer is very low partly due to low diffraction efficiency and mainly due to small acceptance solid angle. Especially for a point source, acceptance solid angle should be as large as possible for efficient use of the source. A LPX emits discrete spectral lines of very narrow bandwidth if the plasma generation is well controlled. Then, a single line can be selected out with use of a filter of a few %. The filter monochromatization is very efficient mainly because of large acceptance solid angle. The second magic is adoption of the time-of-flight (TOF) method for the electron energy analysis. In TOF, a sample is excited by a pulse source, and velocity distribution of ejected electrons is observed by a current detector placed at some distance from the sample. In principle, electrons of all energies can be detected in one shot, and electrons emitted to all directions can be collected with use of some electron lens without sacrificing energy resolution. Due to these two advantages, the detection efficiency of TOF can be several orders higher than methods for a continuously emitting x-ray source. The high detection efficiency is crucially important for high spatial resolution as discussed later. Thus, the pulse nature of LPX which was considered to be disadvantage causing low average power, is changed to a great advantage. These years we have been studying XPS with LPX 5). To get a confidence that an XPS microscope with LPX can become a really powerful tool, many things are to be experimentally confirmed, and many technologies are to be developed. We have to experimentally confirm that; filtering a LPX produces a single line of narrow line width, photon flux in one shot on sample is large enough, energy resolution in TOF can be high enough, and the effect of space charge in pulse excitation is negligible. High rep-rate generation of LPX, high rep-rate acquisition of TOF spectra, increase of solid angle for electron detection, and relaxing debris problem from LPX are to be implemented for the system to have the best performance. As is reported below, our study for these years demonstrated that an XPS microscope with LPX is feasible. 4. EXPERIMENTAL RESULTS 4.1 Single line from a LPX First, we show that a single emission line can be obtained from a LPX with use of a simple filter. Experimental set-up is shown in Fig.3. A boron nitride (BN) plate was irradiated by the second harmonic of a Nd:YAG laser. Pulse duration of the laser was a few ns, and the pulse energy was about one hundred mj, and the focal diameter was 5 µm. Hydrogen-like and helium-like boron and nitrogen ions emitted several lines of wavelength between 2.5 nm and 6 nm, as shown in Fig.4. By filtering a LPX with a mylar foil of a few µm thickness, which has the absorption edge at 4.4 nm, a nearly monochromatic spectrum was obtained as shown in Fig.4 b), in which Lyα line emission of hydrogen-like boron ions at 4.86 nm dominated. The energy resolution of the irradiating x-ray is determined by the line width of LPX emission itself. Therefore, plasma generation must be well controlled to get narrow bandwidth lines. As discussed later in subsection 4.2, the bandwidth of 4.86 nm (255 ev) was better than.5 ev in our case. Fig.3: Experimental setup of time-of-flight XPS with a laser-plasma x-ray source. Optical Density Optical Density (a) Target : BN B Lyman a Laser intensity : W/cm 2 (2ω) N Lyman a N He-like a B Lyman b B He-like b Wavelength [nm] (a) spectrum from a BN plasma (b) Filter : Mylar 2.4 mm x 1 B Lyman α B He-like a Β Ηε λικε α Wavelength [nm] (b) spectrum through a mylar filter Fig.4. Experimentally observed x-ray spectra

4 4.2 Observation of chemical shifts and photon flux on sample Next, we show high energy resolution is achieved with the time-of-flight (TOF) method. In TOF, photoelectrons travel through a magnetically shielded flight tube, and the electron current is detected by a MCP detector at the end of the flight tube. The electron current is recorded with a digitizing oscilloscope, and the current waveform is converted to a photoelectron energy spectrum. The energy resolution can be limited by the temporal response time of the source and the detecting system, which were about a few ns in our case. Because the energy resolution E limited by temporal resolution T scales as T /T 3, where T is the flight time in the flight tube, E can be improved by applying a retarding potential to slow electrons. Figure 5 shows the photoelectron spectra from two samples One sample was a Si Fig.5: Chemical shifts of Si 2p electrons observed by TOF with a laser-plasma x-ray source wafer half covered with SiO 2 stripes, and the other with SiN stripes. By applying a retarding field of 135V, the energy resolution was improved so that chemical shifts of Si 2p electron in SiO 2, SiN, and Si crystal were clearly resolved 6). The laser pulse duration was 3 ns and the combined response time of the MCP detector and the oscilloscope was also 3 ns, and the energy resolution limited by these temporal response time was.4 ev. The observed resolution of photoelectron spectra was estimated to be.5 ev. From these, we found the spectral width λ/λ of the line emission from LPX was around 1/1. The ordinate in Fig.5 shows the number of photoelectrons detected by the MCP. From this number, we can estimate the photon number on the sample. On the other hand, we know the brilliance of the plasma is about 1 18 photons/sec/mm 2 /mrad 2 /.1%BW from the line intensity of the spectrum shown in Fig. 4, which is plotted as a dot B in Fig.2. By knowing the brilliance of the x-ray source, we can calculate the photon number reaching the sample. The calculated photon number agreed well 5) with the photon number observed in the XPS experiment. source energy range (ev) energy resolution (ev) Table 1 space resolution (mm) flux on sample (photons/s) ETL & Nikon LPX <3 (.3) (.5) (1 5 /shot) time-average brightness (relative) (1 4 at 1 Hz) MAXIMUM at Wisconsin undulator < (2x1 9 ) (8x1 7 ) [7] X1-SPEM at NSLS undulator x x1 5 [8] HASYLAB at Hamburg undulator x1 9 3x1 4 [9] MAX I at Lund Hitachi at KEK undulator [1] bending [11] ref.

5 In our future system, we plan to focus x-rays to around.3 µm with a Schwarzschild optic of NA number of.2. As stated above, we can estimate the photon flux when the x-ray beam is focused. Table 1 shows the expected performance of our future system and published performances of systems at synchrotron facilities. In the estimation of photon number of our future system, conservative numbers are used; the peak brilliance of LPX is assumed 1 17 photons/sec /mm 2 /mrad 2 /.1%BW, the pulse duration is 1 ns, the numerical aperture of Schwarzschild optics is.1, and the reflectivity of the multilayer is 3 % per surface. The bandwidth λ/λ of line emission from LPX is assumed 1-3, and no grating is used. People will be surprised to see that one-shot photon-number expected from a micro-beam of LPX is equal to the experimentally observed photon number for 1 seconds accumulation of a bending magnet SR source. The average brilliance of a bending magnet SR source is about 1 13 photons/sec/mm 2 /mrad 2 /.1%BW. Therefore, one might simply expect 1 ms accumulation would give the same photon number obtained from one shot LPX. Quantitative estimation of flux reduction is discussed in ref.9 based on their experimental measurements. The flux reduction of SR source arises from the situation that λ/λ of 1-3 is attained only by the use of a very low efficiency monochrometer. Because of low diffraction efficiency of gratings and low reflectivities of mirrors, the monochromatized flux suffers nearly two order attenuation. More seriously, only radiation emitted in a quite small solid angle is allowed to pass through a monochromator. This very small acceptance angle of a high resolution monochromator reduces the intensity by another few orders. 4.3 Observation of surface states of Si crystal The role of surface and interface of thin layers is very important in nm devices. discussions on surface structures of crystal silicon several 2 years ago. Theoretical calculations predicted laser-plasma reconstruction of surface Si atoms. Surface reconstruction was predicted to produce shifts of the Si 2p binding energy 15 Si (111) of the order of.1 ev. In fact, complicated XPS spectral H-terminated profiles 12) recorded with photons of around 15 ev were decomposed and interpreted with those of surface atoms. 1 In an x-ray tube source, x-rays of 1.5 kev to 2 kev photon energy are generated, and the energy of Si 2p 5 photoelectrons is larger than 1 kev. The escape depth of 1 kev electrons is larger than 2 nm. Thickness of 2 nm is very thin in usual sense, but when detection of a few atomic layers is discussed, layer of 2 nm thickness is considered as bulk. Therefore, the obtained information with these For example, there were hot photons is mostly that of bulk. To get high sensitivity to a few atomic layers, contribution from bulk states must be Fig.6 Detailed spectral profile of Si 2p suppressed. Observation of electrons at very shallow photoelectrons. Tail on the low binding energy angle near the target surface is a technique to reduce the side is produced by the contribution from a few effective depth of detected layer. However, this hampers atomic layers of reconstructed surface. lateral spatial resolution and can not be applied to a high resolution microscope. Escape depth of electrons has the minimum of about.3 nm at the electron energy of around 5 ev. Therefore, excitation by x-rays of 15-2eV photon energy gives information sensitive to surface layers without tilting a sample, as was done when surface states were observed with a SR source. In our laser-plasma XPS system, generated x-rays are 1-3 ev, and our system should be sensitive to surface states. Figure 6 shows the detailed photoelectron spectral profile of 2p electrons from a hydrogen-terminated Si <111> crystal obtained in our experiment. The profile had a tail on the low binding-energy side, which is quite similar to that obtained using a SR source 12). As is demonstrated, a LPX-XPS is very good at observing chemical states of surfaces. Photoelectrons(/shot/eV) 1/ Energy of photoelectrons (ev) 3/2 4.4 Simultaneous detection of multi-elements in 2-dimensional mapping The most important advantage of TOF analysis is simultaneous detection of multi-elements and multi electronic states. In TOF, electrons of all energies are detected simultaneously and under the exactly same condition, and then, the obtained spectral profile is highly reliable. With highly precise profiles, least square fitting can analyze very faint changes, enabling detection of low concentration impurities and electronic defects. In conventional methods, electrons in narrow energy slit are detected in one measurement, and photoelectron spectra are obtained through many measurements by scanning the energy slit. Hence, fluctuations in x-ray intensity or target circumstances could produce artificial shapes in spectral profile. The above advantage of TOF is demonstrated in 2-dimensional mapping described below.

6 Experimental set-up for 2-dimensional mapping is shown in Fig.7. A grazing incidence Wolter type mirror was employed to form an x-ray micro-beam of 6 µm diameter. In this experiment, the target for x-ray generation was carbon, and pulse duration of the laser was 8 ns. The sample was a gold dot deposited on a Si wafer. Photoelectron spectra recorded in the regions of pure gold and pure Si are shown in Fig.7 (a) and (b), respectively. Mapping obtained by the conventional method is shown in Fig.9. Figure 9 (a) is a simple mapping of intensities at 25 ev where 4d electron of gold has the peak, and the mapping of intensity at 1 ev where 2p electron of Si has the peak is shown in Fig.9 (b). As seen in Fig.9 (b), simple mapping of intensity at the Si 2p peak does not clarify distribution of Si. This is simply understood by looking at Fig.8; not only Si but also gold have peaks at 1 ev. Even in mapping of signal intensities at 25 ev, Fig. 9 (a), distribution of gold is not reproduced satisfactorily; lower half of a circle is stronger although real pattern was a uniform circle. This was caused by the fluctuation of x-ray intensity during the scanning of micro-beam. 12 Nd:YAG laser (2ω) plasma tape target micro-channel plate (MCP) photoelectron filter x-ray Wolter mirror (1/4) sample to oscilloscope Fig.7 Experimental configuration for 2-D mapping. retarding field Electron Number (/ev/shot) Au Electron Energy (ev) (a) spectrum in the pure gold region Electron Number (/ev/shot) Si Electron Energy (ev) (b) spectrum in the pure silicon region Fig.8. Photoelectron spectra taken in two regions Fig.9 Mapping by conventional method of a gold dot deposited on a silicon wafer.

7 As shown in Fig.9, in conventional mapping, coincidence of peaks for different elements prohibit producing distribution of some elements, and fluctuation of x-ray intensity or other changes can generate artifacts in mapping. The above problems can be solved by least square fitting of whole spectral profiles. Using spectral profiles for pure gold and pure Si, spectral profile at any position can be decomposed to give fraction of gold and Si with high accuracy. In this method, fluctuation of x-ray intensity is easily normalized. The resultant mapping is shown in Fig.1. Now the true distributions of gold and Si are obtained. Thus, TOF is demonstrated to be quite advantageous for obtaining a highly reliable mapping of multi-elements through least square fitting analysis. 5. Discussions 5.1 Space charge effect One concern in TOF is the space charge. Because short pulse excitation produces many electrons simultaneously, repulsion among electrons might modify electron velocities. This space charge effect caused by pulse excitation was observed by Munakata et al. 13). They reported that photoelectron spectra broaden by several tens mev when the ejected total photoelectron number was larger than 5x1 5 /pulse. In our experiment, estimated electron number in one shot was 1 5 at most, and space charge effect was negligible, which is demonstrated by the observation of surface states as reported in 4.3. In XPS microscopy, the available photon number will be smaller than 1 5 /shot, as shown in Table 1, and then the number of ejected electron will be less than 1. Because electrons of 1 ev energy travel 6 mm in 1 ns, electron density does not increase by focusing x-rays below.1 mm. Hence, space charge caused by pulse excitation will not be a serious problem in XPS microscopy, while this must be confirmed experimentally. 5.2 Spatial resolution limited by radiation damage Very often, people believe that spatial resolution of XPS microscope is determined by focusing performance of x-ray optics. It is true when the resolution is larger than 1 µm. However, for the resolution of sub µm and better, it will be determined by radiation damage. One big advantage of XPS is non-destructive phortoelectrons observation. However, photoelectrons excitation photons excitation photons when the spatial resolution becomes higher and higher, XPS can introduce significant radiation damage. This is understood with the aid of Fig.11. When pixel size is 1 µm, 6x1 9 atoms share the burden of ejecting electrons. On the other hand, when the resolution is 3nm, one pixel has only 5x1 4 atoms. If 1 6 electrons are to be detected to draw a decent photoelectron spectrum, each atom should emit 2 electrons even when all electrons are detected. If the detection efficiency is lower than.1 %, as is usual,. Fig.1 Mapping of a gold dot on a silicon wafer. Least square fitting using whole spectral profiles gives quantitative concentration of multi-elements with high accuracy. low resolution high resolution (1µm x 1µm x.5nm)/(.2nm) 3 (3nm x 3nm x.5nm)/(.2nm) 3 = 6x1 9 = 5 x 1 4 Fig.11 Even in XPS which is known as non-destructive analyzing method, radiation damage becomes serious for high resolution, and practical spatial resolution is limited by radiation damage.

8 2x1 4 electrons must be extracted from one atom. In an experiment with e-beam 14), the radiation damage was observed for the dose corresponding to 2 electrons per atom. We expect that more electrons per atom will be allowed in XPS, but the limit will be lower than 1 4. As is easily understood, for resolution better than.1 µm, practical spatial resolution will be limited not by focusing optics, but by radiation damage. Therefore, increase of detection efficiency is crucially important for improving the spatial resolution. From this point of view, TOF should be adopted as electron energy analysis, and a pulse source is required. 6. SUMMARY We discussed that development of XPS microscope is very important in the study of future nm devices, and that LPX can serve as an x-ray source for it. In our experiment, selection of highly monochromatic line emission, sufficiently large number of photons on sample, and high energy resolution in TOF method were demonstrated. Surface atomic layers of hydrogen terminated Si crystal were also successfully observed. 2-D mapping experiment with 6 µm resolution demonstrated that simultaneous detection of multi-elements by TOF is greatly advantageous in getting precise mapping. We also discussed that an XPS microscope with LPX is not a substitute of that with SR. The reason is that spatial resolution in extremely high resolution is not determined by focusing optics but by radiation damage, and from this point of view, an XPS microscope of tens nm resolution will be realized only with LPX. REFERENCES 1) 2) Ade H 1992 Nucl. Instrum. Methods Phys. Research. A ) Tomie T et al. H 1996 Int. Conf. X-ray Microscopy and Spectromicroscopy XRM`96, O41 4) T.Tomie; applied to Japanese Patent in July 1992, , US pat. No ) Kondo H, Tomie T, and Shimizu H 1996 Appl. Phys. Lett ) Kondo H, Tomie T, and Shimizu H 1996 Int.Conf.X-ray Microscopy and Spectromicroscopy XRM`96, O57 7) Ray-Chaudhuri A K et al J.Vac.Sci.Technol. A , 8) Ade H et al. 199 Appl. Phys. Lett ) Voss J et al J.X-ray Sci.Technol. 3 85, 1) Nyholm R et al Rev. Sci.Instrum , and Johansson U et al Rev. Sci.Instrum , 11) M.Hasegawa and K.Ninomiya 1995 Rev. Sci.Instrum , 12) Wachs A L et al Phys. Rev. B , 13) Munakata T et al Rev. Sci. Instrum , 14) Strecker C L et al J. Appl. Phys Further author information -- T.T (correspondence) : E mail: tomie@etl.go.jp; Telephone ; Fax: H.K.: E mail: kondo@tsukubagw.nikon.co.jp; H.S.: E mail: h.shimiz@etlrips.etl.go.jp; P.L.: E mail:peixiang@etl.go.jp