We report the first chemical identification of a cluster of atoms from identifiable sites on an extended crystal surface. The instrument used combines the functions of a scanning tunneling microscope (STM) with a time-of-flight atom probe. Atoms are transferred from regions of interest identified in STM images to the tip, from which they are ejected into a time-of-flight spectrometer. Preliminary results are shown in which a cluster of silicon atoms taken from the Si (111) 7X7 surface is identified by TOF spectra. Applications of these "atomic tweezers" for microanalysis are discussed.

Recently developed techniques for atomic manipulation and the creation of artificial handmade atomic structures by scanning probe microscopy (SPM) has stimulated work in many fields, from organic chemistry to low temperature physics and semiconductor lithography on the nanometer scale1. Likewise, SPM imaging has opened up entirely new possibilities for the study of the structure and dynamics of individual molecules in solution, of biomolecules, of catalytic reactions on metal surfaces, and of semiconductor crystal growth, amongst other phenomena. In all this work, one crucial piece of evidence is often missing - the chemical identification of the atomic species seen in the images. In some surface science applications this is not a serious drawback, since surfaces of known structure are prepared with atomic cleanliness (confirmed by Auger spectroscopy), and interest focuses on changes due to the addition of one known additional species. In most cases, however, it is believed that impurities control properties, yet these cannot be identified on extended surfaces by any atomic resolution microscopy. (The atom probe provides a powerful method for samples which can be formed into tips for controlled field-evaporation). We report here a new scientific instrument, based on the scanning tunneling microscope (STM) which makes this possible.

The interpretation of STM images is always difficult at the atomic level2. Despite the existence of both accurate ab-initio and approximate computational schemes for image simulation for given structures, no general solution to the inversion problem of finding unique atomic coordinates (and atomic types) from a given image of an unknown structure exists. Additional information on the species present and their crystallographic sites would greatly assist with image interpretation, particularly for multicomponent systems. In the study of crystal growth, for example, the ability to identify foreign atoms at special sites such as kinks and surface steps would be invaluable. Additional uses for chemical mapping at atomic resolution are listed below.

A number of proposals have appeared for chemical identification of atoms in STM images, based on optical excitation, elastic and inelastic tunneling spectroscopy and the excitation of Auger emission from an STM tip at large distances from the sample3. Measurement of the characteristic ionization energy of inner-shell electrons is most commonly used to identify atoms, and presently the techniques which offer the highest spatial resolution (apart from the atom probe) are transmission energy loss spectroscopy (ELS) and scanning auger microscopy (SAM). Both use the subnanometer probe of a transmission electron microscope: the first is capable of identifying a few atoms within a thin film from their ELS spectra in correlation with an atomic resolution image of the projected structure4, the second has very recently achieved a resolution of about 2 nm for surface imaging of bulk samples5. The atom probe provides the ultimate in single-atom microanalysis, however the range of materials is limited, and crystal growth on extended flat surfaces cannot be studied. What is needed is a method of identifying atoms at particular sites selected from an STM or AFM image.

Our instrument achieves this as follows: by applying a short voltage pulse to the tip at a region of interest in an STM image, atoms are transferred from this region to the tip. The sample is then removed, and a much larger voltage pulse applied, causing field evaporation of these atoms into a time-of-flight spectrometer for identification, as in the atom probe (For an excellent review of atom probe principles see Miller and Smith6. Instruments related to ours, but intended for different purposes, have been described Nichikawa and Sakurai amongst others 7 ). Our instrument arose from previous work studying the form of material transferred between tip and sample. This was observed by transmission electron microscopy, using an STM within a TEM, operated in the reflection mode during STM tunneling8.

Figure 1 shows the general arrangement of the instrument, which is housed in a UHV chamber. A tube scanner is used to support the sample holder, which sits on three balls and may be removed to a heating station. The scanner provides all fine motion in X,Y and Z.Tunnel current is obtained from the sample via a conducting support ball connected to the input of a current preamplifier. The non-scanning tip is fixed to the end of a Burleigh inchworm which is used for z-motion. This arrangement simplifies high voltage (<15kV) connections to the tip for field-evaporation. The scan tube is glued directly to the inchworm body with its inner, grounded electrode supporting up to 15 kV electrical isolation from the tip. The tip sits in a removable kinematic mount, allowing the tip to be removed for cleaning by heating, field evaporation or sputtering.

Figure 1. General arrangement of the instrument, not to scale.

The bias lead to the tip passes to a ceramic UHV high voltage capacitor (which supplies the field evaporation pulse, < 2 kV), through a 5 kOhm UHV resistor to a vacuum feedthrough, to which either the bias (STM mode) or the <10 kV d.c. field evaporation voltage is supplied. The resistor prevents pulse reflections from the DC supply and does not interfere with STM operation. The 50 ohm pulse line supplies a < 2kV, 10 ns wide pulse to the tip. The shape of this pulse controls the energy resolution of the time of flight (TOF) spectrum. Multiple reflections are minimized by placing our 50 ohm symmetrical terminator as close to the tip as possible. Ions are detected in TOF mode by a two-stage chevron channel plate, capable of single ion detection (gain 107). A biased fluorescent screen allows observation of either the electron field emission image from the tip or the field ion image of the tip. Capacitive coupling to the screen leads the ion signal to a digital oscilloscope, which is triggered by the 10 ns pulse to the tip. The scope transfers spectra to a small computer using Labview software, which initiates pulsing, provides a graphic user interface, and accumulates a histogram of ion flight times, labelled with m/n values. A leak valve provides a supply of neon imaging gas, which may also be used for sputter sharpening the tip. This form of atom probe, which does not contain a probe hole, resembles the imaging atom probe design of Panitz9.

For a total effective potential Veff at the tip, the mass to charge ratio m/n is obtained in the usual way from the energy conservation equation

where t is the flight time, m the atomic mass, n the ionic charge and L = 200mm (figure 1) on the assumption that the ions are accelerated to their final velocity within a few microns of the tip, due to the rapid falloff in potential near the tip. Ion trajectories in electrostatic fields are independent of m/n. The effective potential is calibrated using spectra from tungsten tips containing both W3+ and W4+ peaks, as shown in figure 2.

Figure 2. Calibration time-of-flight spectrum from a bare tungsten STM tip with sample removed. The oscillations near the time origin are due to the electromagnetic field pulse from the tip, since surpressed by improved shielding. The peak at the extreme right has not been identified.

The mass resolution of the TOF spectrometer is influenced by several factors including the variation in time between axial and marginal rays (2% error) and the shape of the evaporation pulse (ions leaving the tip at the beginning of the pulse are accelerated for longer than those emitted near the end). We expect about Dm/m = 2Dt/t = 1/50 (i.e. 2%) due to the variation in flight time alone - the resolution measured in figure 5 (about 4%) may be due to additional energy deficit effects from the tip pulse shape. (The Si28 isotope accounts for 92% abundance, so that an isotopic energy spread is not expected). The choice of L is a compromise between improved mass resolution (larger L ) and the need to enlarge the angular view of the tip (about 20o full width in this design). This limits the size of the region on the tip from which atoms can reach the detector.

Figure 3 demonstrates the stability of the STM, showing the Si (111) 7X7 structure we have used to test the instrument. The doped sample was resistively heated to 1350oC and cooled through the phase transition at 830oC.

Figure 3. Si (111) 7X7 reconstruction imaged in constant current mode with the STAP.

Figure 4 shows the effects of a 5 volt , 20 ms pulse applied to the sample (tip positive). After pulsing, the right hand image was obtained in a second scan (with the transferred material still on the tip). The images are identical except for the pit where a cluster of atoms have been transferred to the tip. A time of flight spectrum from this region is shown in figure 5 - two successive pulses produced the Si+ and Si++ peaks shown. Further experiments have confirmed that STM scanning with a clean tip does not produce Si peaks in the spectrum without low voltage pulsing for transfer of material.

Figure 4. Identical Si 7X7 regions before and after applying a 10 msec positive 5 volt pulse to the tip near the center of the image, showing modification to the surface.

Figure 5. Successive TOF spectra showing the Si+ and Si++ peaks indicated. Oscillations near time origin are RF noise, since suppressed by improved shielding.

The question arises as to whether the same region of sample can be found for STM imaging after spectroscopy. The original sample holder included a straight, vee-shaped groove, into which two of the three supporting balls fit. This constrains the stage motion to one dimensional motion. Since the stage can easily be positioned to within 100 microns (0.1mm) along this track using the wobble-stick, and since the lateral range of the scan tube is 4 microns, only 25 frames need be searched to find the same area again. However, we have not demonstrated this experimentally yet.

A considerable literature exists on experiments and theory for atom transfer and manipulation by STM10,11. Two regimes have been distinguished - large and small gaps. For microanalysis we require atomic transport from sample to tip only. For small gaps, this depends on tip polarity and the relative field evaporation thresholds Ev for the tip and sample. For small gaps with W tips on Si where chemical forces act, a negative tip produces most efficient transfer. There is evidence that, for large gaps (> 1nm) material flows from sample to tip independent of polarity, since ions of both types may form in response to the field12. In general, refractory metal tips should be used for microanalysis of softer materials, for silicon and tungsten, Ev(W) = Ev(Si) + 2 eV. For large tunneling gaps, some field enhanced diffusion of adsorbed species may occur13, degrading spatial resolution. This effect will depend on the pulse length, temperature, gap size and diffusion coefficients. Following standard practice with atom probes, it may be possible to determine the number of atoms transferred to the tip from the height of the peaks in the spectrum if detection inefficiencies are allowed for. In the time-gated mode, an elemental map of the distribution of sample atoms adsorbed on the tip might be obtained14. A comparison of these images with the STM pit images may be instructive.

Applications of this method include the measurement of diffusion profiles at interfaces (quantum wells, grain boundaries, ceramic-metal), the study of non-stoichiometry during crystal growth, the identification of foreign species at special sites (such as kinks and steps) during crystal growth and the study of their effect on growth kinetics, segregation at defects, such as Cotrell atmospheres around emerging dislocation cores, the study of the atomic mechanisms involved in surfactants during crystal growth, and the possibility of picking up fragments of large molecules lying on surfaces for analysis (in correlation with their atomic resolution images). Other possibilities also arise such as continuous erosion for three-dimensional compositional analysis, research in catalyst poisoning, and the study of changes in surface composition due to surface diffusion at elevated temperatures.

In conclusion, we have demonstrated that a small cluster of atoms may be transferred from a surface onto an STM tip, then identified using time-of-flight spectrometery in a new type of instrument (the Scanning Tunneling Atom Probe). Due, for example, to the limited open area of the channel-plate, ion detection is known to be less than 100% efficient. We therefore do not expect that this method will always be capable of single-atom detection. The statistical sensitivity remains to be evaluated. An atomic force microscope (AFM) variant of the instrument is under consideration.

Acknowledgments

This work was supported by NSF awards DMR9112550 and DMR9526100. We are particularly grateful to Prof G. Smith for his interest and encouragement. Thanks also to M. Scheinfein, D. Seidman, P. McClernon, H. Rohrer and S. Mo.

References.

1. P. Avouris Ed., Atomic scale modifications of material. Nato ASI Series Vol E239. (Plenum, New York, 1992). See also J.Wiesendanger, Scanning Probe Microscopy and Spectroscopy (C.U.P., Cambridge, UK., 1994) and J. Vac Sci. B12 (3) (1994)

2. C.J. Chen, Scanning Microscopy, Supplement 7, p.281-299 (1993)

3. R. Allenspach and A. Bischof, Appl. Phys. Lett. 60, 1908 (1989)

4. P. E. Batson, Nature 366, 727 (1993).

5. G.Hembree, J.Drucker, F.Luo, M.Krishnamurty, J.Venables,Appl. Phys .Lets . 58, 1 (1991)

6. M.Miller and G.Smith,Atom Probe Microanalysis (Materials Research Society, Pittsburgh 1989)

7. T. Sakurai. Prog. Surf. Sci. 33, 1 (1990); O. Nishikawa et al.J.Vac.Sci.Tech. B13 599 (1995)

8. W. Lo and J.C.H. Spence, Ultramic. 48, 433 (1993),

9. J.A. Panitz, Rev. Sci. Instr. 44 1034 (1973).

10. R. Becker, J. Golovchenko, B. Schwartzentruber. Nature 325 419 (1987).

11. P. Avouris and L. Lyo. Applied Surface Science 60 426 (1992); W.Mizutani et al Appl. Surf. Sci. 87/88 399 (1995)

12. K. Kobayashi et al. in ref 1. (Avouris, Ed.).

13. L. Whitman, J. Strocio, R. Dragoset, R. Celotta p 25. in ref 1., Avouris Ed.

14. G. Kellogg, Prog. Surf. Sci. 21 1 (1994).