Highly Charged Ion Bombardment of Silicon Surfaces Jason E. Sanabia, Scott N. Goldie, Laura P. Ratliff, Lori S. Goldner, and John D. Gillaspy Physics Laboratory, National Institute of Standards and Technology, Gaithersburg, MD 20899 Abstract. Visible photoluminescence from Si(100) surfaces irradiated by highly charged ions has recently been reported [1]. In an attempt to reproduce these results, highly charged ion-irradiated silicon samples were prepared at the Electron Beam Ion Trap at the National Institute of Standards and Technology. Two highly sensitive fluorescence detection schemes were employed, both using ultraviolet light from an argonion laser for excitation. In the first detection scheme, the Xe44+-Si(100) samples were excited by the ultraviolet light while a spectrograph equipped with a liquid nitrogen-cooled charge-coupled device camera detected the fluorescence. The second detection scheme was a high throughput laser-scanning confocal microscope equipped with a photon-counting photomultiplier tube. We characterized the sensitivities in each detection scheme, allowing the assessment of the photoluminescence efficiency of Xe44+-Si(100). No photoluminescence was detected in either setup. describe visible light emission from porous silicon is the quantum confinement model [7,8]. In this model, confinement of electron and hole wave functions inside the silicon nanocrystals widens the silicon bandgap into the visible part of the light spectrum and makes a direct bandgap. Ion accelerators can also treat silicon to give it favorable optoelectronic properties by implanting impurities or introducing defects in silicon to make visible and infrared light emitting materials [9,10]. INTRODUCTION Highly charged ions (HCIs) are being explored for possible applications in the fabrication of nanostructures. Due to a rapid reneutralization that begins as the HCI approaches the surface, the ion’s potential energy (50keV for Xe44+) is released within 10 nm of the surface [2]. A single HCI on an atomically flat mica surface results in nanometer-sized damage to the surface [3]. The damage volume scales with HCI potential energy (charge state) from 10 nm3 for Xe30+ to 150 nm3 for Xe50+. On the other hand, HCI kinetic energy plays a small role in the damage scale: two orders of magnitude variation in HCI kinetic energy results in very little change in the size of the damage site [4]. Hamza, Newman, Thielen, Lee, Schenkel, McDonald, and Schneider reported that areas of silicon exposed to highly charged ions (Xe44+) emitted visible light when exposed to ultraviolet light [1]. The emission contained a number of sharp peaks between 540 nm and 570 nm. They hypothesize that highly charged ions form nanometer sized impact sites on the silicon surface and that quantum confinement occurs. Hence, highly charged ions have directly fabricated light emitting nanostructures on a silicon surface. Bulk silicon has an indirect bandgap in the infrared, and is therefore a poor emitter of visible and infrared light. Visible light emitting devices must therefore employ exotic semiconductors, like GaN, that can be expensive and toxic. Pure silicon can be treated to give it favorable optoelectronic properties. For example, photoluminescent porous silicon can be fabricated by wet chemical etching [5] or spark processing of silicon [6]. One of the models used to Here we seek to reproduce and expand upon the work of Hamza, et al. We describe experiments designed to measure the quantum yield and spectra from individual ion impact sites. The narrowness of the emission spectrum observed in Ref. [1] suggests CP680, Application of Accelerators in Research and Industry: 17th Int'l. Conference, edited by J. L. Duggan and I. L. Morgan 2003 American Institute of Physics 0-7354-0149-7/03/$20.00 568 that photoluminescent sites have the same emission spectrum, however, it may be the case that the emission spectrum is the sum of spectrally-distinct impact sites. One way to determine this is by measuring the photoluminescence from individual impact sites [11]. Pa. The surface is then allowed to reach room temperature. The prepared Si(100) surface is exposed to a continuous beam of Xe44+ ions from the EBIT [13]. The target chamber base pressure is 3×10-7 Pa. A beam of isotopically pure 136Xe44+ ions is extracted from the EBIT. Before exposure, a Faraday cup measures the ion beam current (typically 10 pA) through a 3 mm aperture. The Si(100) is translated into the ion beam path between the aperture and the Faraday cup. The kinetic energy is 350 keV. Exposures of several hours with a stable ion beam flux are possible. The two samples reported here have total doses of 1.3×1011 ions⋅cm-2 and 2.6×1011 ions⋅cm-2. We have constructed two highly sensitive fluorescence detection schemes to investigate silicon samples exposed to highly charged ions at the Electron Beam Ion Trap (EBIT) at the National Institute of Standards and Technology. In the first scheme, a fluorescence spectrometer samples a relatively large area of the sample. The second scheme is a confocal microscope capable of single fluorophore detection. The confocal microscope samples a relatively small area of the sample and was developed to measure fluorescence from individual ion impact sites. We bombarded Si(100) surfaces with Xe44+ using the same procedures as Hamza, et al. Unfortunately, we have not detected fluorescence from our samples with either optical setup, but our procedure places a limit on the absorption cross section of an individual ion impact site in Xe44+-Si(100). After exposure, the Xe44+-Si(100) samples are removed from the EBIT target chamber and exposed to atmosphere for approximately 1 hour during transfer to the vacuum chambers in which the fluorescence measurements are performed. Fluorescence Spectrometer The fluorescence spectrometer is schematically depicted in Figure 1. The Xe44+-Si(100) sample is held at 1 Pa in a vacuum chamber behind a fused quartz viewport at room temperature. The vacuum chamber is mounted on a platform that translates in the x, y, and z directions and rotates about the z-axis parallel to the sample plane. An argon ion laser supplies a continuous 364 nm (UV) excitation beam. This laser line is chosen because it is close to the wavelength of the pulsed 379 nm excitation that Hamza, et al. used. The UV excitation is focused to an ellipse on the surface of the sample with a minor axis diameter of 20 µm (diffraction limited). The UV excitation was polarized linearly in the plane of incidence and was incident at 56o (Brewster’s angle) to minimize ghost reflections of the excitation beam on the vacuum viewport. EXPERIMENTAL Xe44+-Si(100) Sample Preparation A 13 mm × 3 mm × 1 mm Si(100) sample is cut from a float zone Si(100) wafer that was sent to us by Hamza, et al. The sample is mounted in a molybdenum fixture and inserted into a heating assembly in an ultrahigh vacuum system. The base pressure is 3×10-8 Pa. The sample, fixture, and heating assembly are outgassed overnight by passing current through a tungsten filament heater such that the sample temperature is held at 870 K. The sample temperature is measured with an optical pyrometer. The silicon annealing procedure that we use has been shown to drive off adsorbates and the native oxide from the silicon surface and leave a clean, wellordered Si(100)-(2×1) reconstructed surface [12]. The sample is heated by passing a current directly through the sample. In the annealing procedure, the sample is heated rapidly to 1520 K until either the pressure rises above 3×10-7 Pa or 60 seconds has elapsed. Next, the sample is rapidly cooled to 1170 K and held there for 60 seconds. Finally, the sample is cooled to below 870 K at a rate of 2 K⋅s-1. This annealing cycle is repeated until the sample can remain at 1520 K for 60 seconds at a vacuum chamber pressure less than 3×10-7 The fluorescence from the sample is collected by an off-axis parabolic mirror with a 300 mm focal length and an estimated numerical aperture of 0.25. The solid angle of the parabolic mirror collects 1.7 % of the fluorescence from the sample, assuming isotropic emission. The fluorescence then passes through an uncoated pellicle beam splitter and reflects from two metallic steering mirrors. A 300 mm lens focuses the fluorescence onto the entrance slit of the liquid nitrogen cooled charge coupled detector (LNCCD) spectrograph. A 400 nm long pass filter is placed in front of the entrance slit of the LNCCD spectrograph to filter out the scattered UV excitation. 569 The beamsplitting mirror above the objective lens transmits 1 % of the UV reflected from the sample. SAMPLE IN VACUUM LIQUID NITROGEN-COOLED CCD SPECTROGRAPH PELLICLE BS MIRROR The solid angle of the objective collects 10 % of the fluorescence from the sample, assuming isotropic emission. The fluorescence then passes back through the dichroic mirror, which blocks most of the UV beam and transmits the longer wavelength fluorescence. A holographic notch filter blocks the remaining UV and transmits the fluorescent light (80 % at 500 nm). A lens focuses the fluorescence through a 150 µm pinhole to filter out background that does not originate from the focal spot on the sample. Behind the pinhole, a photon-counting photomultiplier tube (34 counts·s-1 dark noise) detects the fluorescence. This photomultiplier has a quantum efficiency of 25 % at 400 nm and 5 % at 550 nm. We estimate that the confocal microscope has a detection efficiency of 0.1 % at 550 nm. This estimate is based on estimated transmission/reflection losses for each of the mirrors, lenses, the dichroic, and the pinhole between the sample and the detector. 400 nm LONG PASS FILTER 633 nm ALIGNMENT LENS MIRROR LENS PARABOLIC LENS 364 nm EXCITATION FIGURE 1. Schematic of the fluorescence spectrometer. The LNCCD spectrograph has a 0.3 m focal length. A 300 groves·mm-1, 500 nm blazed grating with an average efficiency of 78 % at 550 nm disperses the light onto the LNCCD detector. The LNCCD detector is a 1340 × 100 pixel array with 90 % quantum efficiency at 550 nm and greater than 60 % quantum efficiency between 400 nm and 800 nm. The LNCCD has very low thermal noise allowing spectra to be recorded for at least 1000 seconds with no introduction of thermal noise (above readout noise) to the signal. We estimate that the parabolic setup has an overall detection efficiency of 0.4 % at 550 nm. The detection efficiency is defined as the fraction of the fluorescence emitted from the sample that is detected by the LNCCD. This estimate is based on assumed losses for each transmission/reflection from optics between the sample and the detector. The three mirrors inside the spectrograph and have been included in this loss estimate. CCD VIDEO CAMERA 364 nm EXCITATION EMISSION SCANNING MIRRORS AXIS LENS BEAMSPLITTING MIRROR DICHROIC BEAMSPLITTER HOLOGRAPHIC NOTCH FILTER Confocal Microscope SAMPLE HOLDER LENS The laser-scanning confocal microscope, similar in design to one described elsewhere [14], is schematically depicted in Figure 2. This microscope can acquire the fluorescence image of a sample and is sensitive to a single fluorescent molecule (see below). The Xe44+-Si(100) sample is held at 1 Pa behind a 0.5 mm fused quartz window in a vacuum chamber at room temperature. The vacuum chamber is mounted on a platform that translates in the x, y, and z directions. An argon ion laser supplies the 364 nm (UV) excitation beam, which reflects from a dichroic mirror and passes two rotating mirrors. A 0.6 NA objective lens with high transmission in the UV focuses the 364 nm light through the quartz window to a spot with a diameter of 730 nm (diffraction limited). The lens is corrected for the 0.5 mm window thickness. The spot can be scanned across the surface plane by rotating the scanning mirrors. For focusing, a video camera and lens mounted above the objective lens gives an image of the UV spot on the surface. OBJECTIVE LENS LENSES LENSES UV LASER PINHOLE PHOTOMULTIPLIER FIGURE 2. Schematic of the confocal microscope. RESULTS Using the fluorescence spectrometer, we did not detect any fluorescence for 10 W⋅cm-2 (60 µW into an area of 5.6×10-6 cm2) of 364 nm excitation on the ionexposed region of the Xe44+-Si(100) sample. The Xe44+-Si(100) sample had a dose of 1.3×1011 ions⋅cm-2. The LNCCD exposure time was 100 s. There was no background in the spectrum, only readout noise averaging 1.4 counts. We can impose an upper limit on the product of the absorption cross section at 364 nm, σ, and the fluorescence quantum yield, φ, of a single ion impact site in Xe44+-Si(100). In general, the signal-to-noise ratio, s/∆sn, is given by 570 s ∆s n = nDφσPτ ( Ahν ) nDφσPτ ( Ahν ) + CPτ + N d τ + s r2 As in the case of the fluorescence spectrometer measurement, we can impose an upper limit on φσ using Equation 1. The term sr2 = 0, since we are not using the LNCCD detector. The term CPτ + Ndτ = 220 counts·s-1 × 40 s = 8800 counts. The number of ion impact sites sampled in the A = 4.2×10-9 cm2 is n = 1100 ions. The detection efficiency is D = 0.001. This information gives the upper limit φσ = 9×10-22 cm2. The quantum yield is less than or equal to 1. Therefore, the upper limit for the absorption cross section is of order 10-21 cm2. Again, we conclude that the product of the absorption cross section at 364 nm and the fluorescence quantum yield for an individual impact site on Xe44+-Si(100) is extremely low. The value is 4 orders of magnitude higher for the confocal microscope for three reasons: First, the fluorescence spectrometer samples 730000 ion impact sites, whereas the confocal microscope samples 1100. Second, the photomultiplier is much noisier than the LNCCD. Finally, the optics in the confocal microscope autofluoresce and contribute to the detected background. (1) The numerator is the signal, s, in counts. n is the number of fluorophores in the area A. D is the detection efficiency. P is the incident beam power. τ is the collection time. A is the incident beam area on the sample. hν is the incident photon energy. The denominator is the total noise uncertainty, ∆sn, which is a propagation of the errors from the following noise sources: the statistical noise in the signal, s ; the statistical noise in the background, CPτ , where C is the background count rate per incident power; the statistical noise in the dark counts, N d τ , where Nd is the dark count rate; and the readout noise of the LNCCD camera, sr. Setting s/∆sn = 1, we can solve for φσ. The terms CPτ (background) and Ndτ (dark counts) are both effectively zero for the LNCCD. The number of ion impact sites in the sampled area, A = 5.6×10-6 cm2, is 7.3×105 sites. Assuming each impact site results in a fluorescent center, n = 7.3×105. The detection efficiency is D = 0.003. This information gives the upper limit φσ = 5×10-25 cm2. The quantum yield is less than or equal to 1, therefore, the upper limit for the absorption cross section is σ = 10-25 cm2. For comparison, the cross section for an organic fluorescent molecule is of order 10-16 cm2. We conclude that the product absorption cross section at 364 nm and the fluorescence quantum yield for a single ion-impact site on Xe44+-Si(100) is extremely low. The confocal microscope can detect single fluorescent molecules. To demonstrate this, we prepared a sample of Stilbene 3 fluorophores dispersed onto a clean quartz coverslip. We used spin-casting so that each molecule was separated by distances greater than the UV spot size. In this procedure, we dropped a few 10 µL drops of a dilute solution of Stilbene 3 in methanol onto a quartz coverslip spinning at 3000 revolutions per minute. During imaging, we used a 1.3 NA oil immersion objective lens to achieve a tighter focus and to collect more emitted light. In fluorescence images showing single Stilbene 3 molecules, the signal for each molecule was approximately 5000 counts·s-1 over a background of 300 counts·s-1. The excitation intensity was 7700 W⋅cm-2 (7.0 µW into an area of 9.1×10-10 cm2) The diameter of each molecule in the image was the diameter of the UV spot, roughly 340 nm. To further investigate these samples, confocal microscope images were taken of the Xe44+-Si(100) sample with a slightly higher dose of 2.6×1011 ions⋅cm-2. The sample was exposed to atmosphere for up to 24 hours before the measurement. We did not detect any fluorescence for 1200 W⋅cm-2 (5.1 µW into an area of 4.2×10-9 cm2) of 364 nm excitation on the ion-exposed region of the Xe44+-Si(100) sample. We collected 100 µm × 100 µm fluorescence images over a 13 mm length of the sample one image, translating the Xe44+-Si(100) sample 1 mm between images. Each image was collected in 40 seconds and showed no features. Summing over the pixels in each image and dividing by 40 seconds, the background in each image was 220 counts·s-1; 190 counts·s-1 is due to autofluorescence of the microscope optics and 30 counts·s-1 is due to dark noise in the photomultiplier. Stilbene 3 has an absorption cross-section of 1.5×10-16 cm2 at 364 nm, an emission maximum at 430 nm, and a quantum yield of 15 % [15]. Multiplying the excitation intensity by the absorption cross section and converting to counts·s-1 using the energy of a 364 nm photon, we find that a single Stilbene 3 is excited at a rate of 2.3×106 counts·s-1. A single Stilbene 3 molecule is therefore emitting photons at a rate of 3.5×105 counts·s-1. We estimate that 1.5 % of this fluorescence is detected, given that 24 % of the emission is collected by the oil immersion objective (assuming isotropic emission) and that the photomultiplier has a quantum efficiency of 25 % at 571 400 nm. Hence, we estimate a 5300 counts·s-1 signal from a single Stilbene 3 molecule, in good agreement with the measured 5000 counts·s-1. silicon atoms with a metal atom at the center can have a HOMO-LUMO transition at 520 nm [19]. ACKNOWLEDGEMENTS DISCUSSION We would like to thank Alex Hamza and Dieter Schneider for openly discussing their results and for providing the silicon wafer. We also acknowledge Garnett Bryant and Kimberly Briggman for their insightful discussions. JES and SNG are grateful for the support of NRC/NIST Postdoctoral Research Associateship. We did not detect photoluminescence from our Xe44+-Si(100) samples using two highly sensitive detection schemes. Our measurements place an upper limit of 10-25 cm2 for the 364 nm absorption crosssection of a single Xe44+ impact site on Si(100). On the other hand, Hamza, et al. detected appreciable fluorescence signal from their Xe44+-Si(100). We followed as closely as possible the reported sample preparation techniques and used similar ion beam exposures. We are confident in the exposure of our samples, based on our experience with scanning probe microscopy imaging of mica and graphite surfaces, where the distinct features at the ion impact sites make the ion irradiated regions of the surfaces easily recognizable [3,16]. We can only speculate that there are differences between our optical detection schemes and sample preparation techniques and those used by Hamza, et al. that would explain the discrepancy. REFERENCES 1. 2. 3. 4. 5. 6. One notable difference is that Hamza, et al. used a pulsed laser beam to excite Xe44+-Si(100), whereas our measurements were done with a continuous wave laser. When silicon is excited at high intensity, the simultaneous recombination of two electrons and two holes in silicon causes the emission of a photon with nearly twice the silicon bandgap energy, corresponding to approximately 550 nm [17,18]. If the pulse energy used by Hamza, et al. was high enough, it is conceivable that they observed a similar effect. 7. 8. 9. 10. 11. 12. 13. 14. Differences in sample handling might introduce impurities or defects on the silicon surface. Impurities such as transition metals and rare-earth metals, or defects such as dislocations, affect photoluminescence around the silicon band gap of 1.1 eV (corresponding to wavelengths near 1100 nm) [18], but there are no reports to our knowledge of impurities and defects causing visible photoluminescence in silicon (without the presence of silicon nanocrystals). However, one study suggests that a small cluster of 15. 16. 17. 18. 19. 572 Hamza, A. V., et al., Appl. Phys. Lett. 79, 2973 (2001). Hattass, M., et al., Phys. Rev. Lett. 82, 4795 (1999). Parks, D. C., et al., Nucl. Instrum. Methods Phys. Res., Sect. B 134, 46 ( 1998). Parks, D. C., et al., Journal of Vacuum Science & Technology B 13, 941 (1995). Canham, L. T., Appl. Phys. Lett. 57, 1046 (1990). Stora, M. E. and Hummel, R. E., J. Phys. Chem. Solids 63, 1655 (2002). Esaki, L. and Tsu, R., IBM J. Res. Develop. 14, 61 (1970). Dingle, R., Wiegmann, W., and Henry, C. H., Phys. Rev. Lett. 33, 827 (1974). Ng, W. L., et al., Nature 410, 192 (2001). Rebohle, L., et al., Appl. Phys. Lett. 77, 969 (2000). Empedocles, S. A., Norris, D. J., and Bawendi, M. G., Phys. Rev. Lett. 77, 3873 (1996). Yates, J. T., Experimental Innovations in Surface Science, New York: Springer-Verlag, 1997, p.791. Ratliff, L. P., et al., Rev. Sci. Instrum. 68, 1998 (1997). Stimson, M. J., Haralampus-Grynaviski, N., and Simon, J. D., Rev. Sci. Instrum. 70, 3351 (1999). Brackmann, U., Lambdachrome Laser Dyes, Third ed., Goettingen: Lambda Physik AG, 2000, pp. 82-83. Minniti, R., Ratliff, L. P., Gillaspy, J. D., Phys. Scr. T92, 22 (2001). Betzler, K. and Conradt, R., Phys. Rev. Lett. 28, 1562 (1972). Kimerling, L. C., et al., Solid State Physics, 50, 333 (1997). Kumar, V., and Kawazoe, Y., Phys. Rev. Lett. 87, art. no. 045503 (2001).
© Copyright 2025 Paperzz