Increased resolution in neutral atom microscopy

Authors


P.J. Witham, Department of Physics, Portland State University, Portland, Oregon 97201, U.S.A. Tel: 001-503-725-3812; fax: 001-503-725-2815; e-mail: pjw@pdx.edu

Summary

The neutral atom microscope uses a beam of thermal noncharged atoms or molecules to probe an atomic surface with very low interaction energies (<70 meV). Continued optimization of the ‘pinhole’ neutral atom microscope has improved resolution to 0.35 μm. Recent images are presented demonstrating resolution and the contrast mechanisms identified so far. The future potential for sub-100 nm resolution is discussed.

At our lab we previously developed a neutral helium microscope which makes no attempt to focus the beam. This approach relies on locating the sample very close to a beam forming aperture, and mechanically scanning the sample, then detecting the scattered beam particles. This microscope produced the first images obtained in reflection mode from gas scattering, with resolution similar to the best published using focusing systems and transmission mode by Koch et al. (2008). It also demonstrated high beam intensity (1010–1011 atoms s−1) and superior signal-to-noise ratio. It has allowed one to begin exploring the unique imaging contrast mechanisms seen by gas scattering. Others working with molecular beam experiments have shown the potential for new science using neutral atom scattering, for example, Engel & Rieder (1981) through Jardine et al. (2009).

Improving spatial resolution

Improving resolution has been our main task since this design was published (Fig. 1.) Better resolution can be obtained by miniaturizing the critical components and dimensions of the system, and also by increasing the detector sensitivity or source intensity. This microscope and the method used for predicting resolution is described in Witham & Sánchez (2011a).

Figure 1.

Neutral atom microscope helium scattering image of a biological sample (uncoated Crocosmia pollen grains on adhesive.)

Important improvements were made on the miniaturization front (Fig. 2.) First, the aperture diameter was reduced. Then the working distance was reduced (from 25–100 μm to 10–30 μm) by using a more precise sample positioner. Working distance is currently limited by the stereo (optical) microscope used to view the sample and avoid contact with the aperture. A higher resolution optical microscope would allow a smaller working distance. The source nozzle inside diameter and flow rate were reduced. In addition, the distance from the source to the aperture was reduced, to 0.3–0.6 mm currently. Finally, the cone shaped ‘aperture holder’ component was formed to a sharper point. This modification has improved image contrast by providing more clearance from the sample for scattered beam particles to reach the detector nozzle inlet.

Figure 2.

Diagram of critical components for forming the beam and detecting scattered atoms.

Additional changes include better alignment of the source to the aperture along the desired beam axis, which has improved image contrast (misalignment partially scatters the beam). The sample and scan plane were tilted somewhat towards the detector, effectively raising the illumination angle in the images.

Vibration measurements done using the electromagnetic sample positioner/scanner indicate that vibration (due to vacuum pumps) is a significant component of the present resolution limitation. A primary factor in the vibration magnitude is the low resonant frequency associated with the XY scanner, which can be easily replaced. It was also found important to electrically ground the sample platform since a buildup of electrical charge can cause movement of the sample due to electrostatic attraction to the aperture.

Beam full width half maximum measurement is often used for resolution. The full width half maximum can be measured using a 12–88% line step transition measurement in cases were the beam has a Gaussian profile (Senoner et al. 2005). The edge 12–88% transition distance is 0.260 μm, averaged over the four line profiles of Figure 3, with the 10–90% figure being 0.274 μm. The image scale is calibrated using a 127 μm spacing Quantifoil™ Transmission Electron Microscope (TEM) grid. Our resolution estimate, 0.35 ± 0.05 μm, is more conservative than the full width half maximum to account for image noise (Senoner et al. 2010) and the scale calibration precision.

Figure 3.

(a) Neutral atom microscope image of the central area of a pollen grain, with the four thin white lines in the upper area marking the location of the line profiles of panel (b). These sites are chosen because the sample topography presents bright linear edges facing the detector, folded under and contrasting against a dark shadowed background. These nearly sharp edges provide a reasonable resolution measurement.

Detector sensitivity

The detector has not been improved as such, but optimization of the detector nozzle has increased sensitivity. The function of this component is to select and increase the pressure contrast sensed by the detector due to scattering in some chosen direction. A discussion of this as a network of vacuum conductances is found in Witham & Sánchez (2011b). The nozzle is a sheet metal cone shape with a narrow inlet area located close to the sample. Its large end connects to a tubular volume enclosing the detector's ionizer and quadrupole mass spectrometer head. The small inlet area facing the sample is the only area open to the vacuum system, and is thus also the outlet for any gas which enters the nozzle. In effect this has a large partial pressure ‘multiplying’ effect since the scattered beam particles radiate from the point of impact on the sample in free molecular flow. The detector nozzle inlet area must always cover an optimum included angle for the highest contrast (Witham & Sánchez 2011a). Using a small inlet area allows the inlet to be placed closer to the beam landing point where the pressure from scattered beam particles (impingement rate per unit area) is higher. Therefore a smaller inlet can sample a higher pressure. The trade-off is in pressure response time, essentially a first-order low-pass filter with a time constant τ=v/C, where v is the empty detector volume enclosed behind the nozzle, in L, and C is the vacuum conductance through the nozzle, in L s−1. For an estimate we can treat the nozzle opening as an aperture.

image(1)

where Mb is the beam particle mass in g mol−1, ai is the nozzle open area and k= 630 m s−1 at 300 K, which equals (R · T · mol/(2π · g))1/2.

Assuming a constant optimum included angle at the nozzle opening, the inflow of particles is consistent regardless of the open area. Also assuming the conductance of the vacuum chamber is much higher than the nozzle, the pressure contrast seen by the detector is inversely proportional to the area of the inlet. For our current microscope, the effective detector inlet area without the nozzle is ∼380 mm2. The current inlet area is 6 mm2. The increase in pressure is then roughly 63:1. A ‘beaming’ effect is ignored here which should increase the effectiveness of the nozzle, gas radiating into the nozzle from the beam landing point, on average, reaches deep within the nozzle before scattering since the nozzle area expands behind the inlet. Gas returning randomly to the chamber faces the conductance limitation (higher scattering probability) of a decreasing area cone.

The result is a trade-off which has been optimized for a response time similar to the present pixel sample time of 0.15 s. An unexpected result of this trade-off between sensitivity and speed is that the optimized performance of this detector system is inversely proportional to its unfilled internal volume.

Contrast mechanisms

The contrast mechanisms clearly identified to date are all topographic effects (Fig. 4.) One important difference from the behaviour of light is that He atoms do not undergo significant absorption. The detected partial pressure, converted to image intensity, can be considered as a mixed reflection to the detector (the source of ‘white’) and to the vacuum pumping system (the source of ‘black’). The angle of the surface to the detector is found to produce contrast as one would expect from a diffuse reflecting surface, with the highest brightness corresponding to the specular geometry. More specifically, the brightness at each point appears to roughly match the expected cosine distribution for diffuse scattering, corresponding to the visible included angular area of the beam spot as seen from the direction of the detector inlet. Shadowing of the visibility of the detector inlet area from the beam landing point also produces contrast as expected if you consider the detector inlet as the apparent source of ‘illumination’. Surfaces facing the lower right in Figures 1 and 4 and the bottom of Figure 3(a) demonstrate shadowing.

Figure 4.

Neutral atom microscope images of (a) the pollen grain on the right-hand side of Figure 1 and (b) debris cluster on the edge of a TEM grid. Points 1,2,3 and 4: areas of reduced contrast due to multiple scattering in the path between the beam landing point and the detector.

Generally, sample areas from which beam particles undergo multiple scattering events on the path to the detector have a mid-level grey intensity. Grey intensity indicates more balanced probabilities of scattered particles reaching the vacuum system or the detector inlet first (Fig. 5). In a similar situation for a light image, the multiple reflections would typically result in high absorption, these areas of the sample would be darker. This effect can especially be seen in narrow indented areas, such as point 1 within a grain in Figure 4(a), and other obscured areas such as points 2, 3 and 4 of Figure 4(b). On the upper right side of Figure 1, a simplified case of this effect can be seen where one pollen grain overlaps another. The reflection off of the closer grain produces a diffuse illumination of the grain below it. At the same time, contrast is reduced in that area. This varying contrast effect can be understood as a quality of the illumination of the area. A direct view of the detector nozzle inlet or the vacuum system produces direct, high contrast illumination, whereas areas where escape requires multiple scattering are diffusely illuminated. Note that significant diffuse illumination is reflected from the detector side of the conical aperture holder, just above the sample.

Figure 5.

Diagram illustrating a few of many possible scattering paths between the beam landing spot and the detector given a deep (aspect ratio) sample surface topography. Areas without a direct view from the detector inlet have reduced contrast due to multiple scattering.

Images of silicon crystal samples (Fig. 6) have not shown obvious nontopographic effects, such as especially bright specular reflections.

Figure 6.

Neutral atom microscope images of a silicon wafer sample: (a) corner, (b) chipped edge and (c) broken edge with debris. These are of lower resolution than the later images of Figures 1 and 3.

Conclusions and future work

For reflection mode neutral atom microscope, the realization that reducing the internal empty volume of the detector is as important to sensitivity as increasing ionization efficiency should have an important effect on future designs.

Unfortunately, if one desired to use a time-of-flight detection method, which required rapid sampling at the detector, the optimized detector nozzle method of increasing the pressure signal could not be used to the extent we did here. An example of that would be the use of a source beam chopper and time-of-flight-specific sampling to select a narrow range of particle energies. The slow pressure response would eliminate the time-of-flight information. Selecting a particular beam energy by some method is important to many molecular beam experiments. It might also be used to reduce chromatic aberration in a focused microscope. Chromatic aberration is not a concern for a pinhole neutral atom microscope, however.

Resolution has improved dramatically with modest efforts and the results presented give confidence in the method we previously published for predicting the resolution. Optimizing the detector nozzle as described has produced a large improvement in the detector sensitivity without changes to the detector itself. At this point, four paths remain to further improvement. Replacing the ionizer section of the Residual Gas Analyzer detector (RGA) with a more effective one would allow up to perhaps a 103× increase in sensitivity based on the best ionizer designs found so far (Alderwick et al., 2008). Better optics for monitoring the aperture to sample distance would allow the working distance to be reduced, which then allows a reduction in the source to aperture distance without excessive beam divergence and thus a reduction in the aperture diameter while maintaining constant beam intensity. Finally, reducing the mechanical vibration amplitude at the scanner and improving the source intensity offer some degree of improvement.

Replacing our present vibration sensitive scanner would improve resolution by at least 100 nm.

Spot size calculations indicate that reaching sub-100 nm resolution can be done without detector improvements by reducing working distance to 3 μm (from the present ∼20 μm), adjusting the source distance and drilling a smaller aperture diameter. Improving the detector sensitivity by ∼40× would produce the same resolution increase while allowing a greater working distance. We believe therefore that this nanoscale resolution goal is certainly reachable.

The most remarkable finding on contrast mechanisms so far is the paucity of nontopographic mechanisms. In this, it seems to be unique among microscopes that use a beam. Other contrast effects appear in small areas of a few images, but have not yet been identified. Various effects would be expected, and over large areas, from the body of helium scattering experiments mentioned previously. The vacuum condition of the present microscope is poor however, between 10−8 and 10−7 Torr with water and hydrocarbons at significant levels. Quite possibly, sample surfaces are masked by surface contamination and absorbates. If so, the scattering seen is off of the contamination layer and not the underlying sample.

Quantitative measurement of brightness versus surface angle has yet to be done. A full prediction of this would require numerical simulation of scattering from all surfaces nearby the beam landing point into the detector inlet, which would be useful for optimizing the inlet shape and position. Imaging of many crystal surface types would be interesting. Imaging of clean hydrophilic and hydrophobic surfaces on the same sample, before and after the introduction of water vapour to the vacuum would also be interesting, as would images taken at various stages of monolayer growth. Finally, methods for in situ surface cleaning, which do not have damaging particle energies or charges, need to be utilized.

Acknowledgements

Many thanks to Sean King, Everett Lapp, Rich Swinford, Mike DeArmond, Derek Nowak and Intel Corporation for help, suggestions and support.

Ancillary