Realistic NanoARPES

Here are the facts and some guidelines of what can be realistically done using nanoARPES.

Samples with insulating or marginally insulating character need careful consideration. This is discussed in the NanoARPES Samples page.

NanoARPES spatial resolution:

a) with Fresnel Zone Plate: 120 nm routinely achievable. You should expect the flux in nm-scale nanoARPES to be about 100 times less than what you are used to in conventional microARPES. Thus, count times are much longer, although this can be alleviated by going to higher pass energy or angle mode (thus reducing resolution). You cannot relax the beamline energy resolution without also sacrificing spatial resolution at the sample.

There is very little space charge limitation due to the low photon count rate (10⁹ to 10¹⁰ ph/sec), however beam damage effects can be considerable due to high flux density. Due to low count rates, measurements typically are done at large pass energy and angular mode.

b) with elliptical capillary: we have recently commissioned an elliptical capillary with very high flux (similar to microARPES or conventional ARPES) and sub micron spot size. This can clearly help with fast identification of samples, and also allows better energy/angle resolution. Beam damage will obviously be worse, but experiments have been successfully conducted on graphene and WS2 without too much problem. Space charge limitations will certainly arise. You can see the paper by Rotenberg and Bostwick (and references therein) for estimates of space charge broadening vs. spot size.

Currently, the focus quality (which we think is on the order of 1/2 micron) is still being determined with respect to beam tails and beam shape issues that can limit spectroscopy (although images look great). We are working closely with the vendor (SigRay) to improve these optics and expect to install a next-generation optic sometimes this term.

c) Can I go to larger spot size? Yes, the best way is to just defocus the optic (move it up or downstream). But, the beam profile is not gaussian, it is a doughnut shape, so you should understand that when interpreting images.

d) Prospects for improved spatial resolution. The limitations on this are

i) sufficient numeric aperture. This requires large diameter zone plates in combination with close distance to sample. Compared to a STXM both of these limitations act to reduce the accessible electron solid angle, and also to limit whether normal electron emission can be accessed. We could go to the same spot size as a STXM (25nm) now in principle but it would sacrifice electron access so we prefer to invest our energy in ever larger zone plates. These zone plates require state-of-the art manufacturing which is delaying their implementation.

ii) diffraction limit of photons, which can be overcome by going to higher photon energy. We like ARPES at lower photon energy (we run typically at 94 eV) where coherent flux, energy resolution, ARPES cross sections are favorable.

Automated / User driven operation (or lack thereof)

Our biggest limitation right now is manpower. The sample transfer to nanoARPES is both more dangerous and less automated than the other chambers. Although we make incremental progress in this area, do not expect to be handed the reins very soon. This is especially true for sample loading/locating, and selecting/focussing the optics all of which must be done by the BLS (i.e. only Aaron). If once all this setup is done, we will consider letting users "point and shoot" a particular sample.

Given the low count rate and the need for BLS, nanoARPES is typically run in a 24 hour mode. The first shift is in the daytime when the BLS is present, and is used for setup. Then longer scans (as discussed below) are programmed for overnight operation.

Unrealistic NanoARPES Experiments

- Fermi Surface at 120 nm for more than a few sample locations- . (i.e. X-Y-E-kx-ky) scans are not feasible with the zone plate and feasible with the capillary but only for a very course scan.

- "I think my sample is somewhere near the corner" -- We can find samples with such poor descriptors but it takes hours and we would rather look at better-located samples.

- "I don't know if the sample is clean enough to show good bands" -- Sorry, you should try in microarpes first.

- Samples that didn't follow our guidelines, see "NanoARPES Samples" in the sidebar.

- Variable temperature. We routinely run at RT and we think we can run at LT (20K) but don't expect sample tracking to be sufficiently precise for VT operation

Coming Soon or Conceivably Feasible

- In operando. We have recently modified the stage for up to 8 contacts for in operandi geometries but this will require commissioning experiments. We are looking for good commissioning experiments to develop this technique in partnership with a user group.

- Variable Temperature. It requires an upgrade of our laser-interferometry, which is currently budget- and manpower-limited.

- In situ growth (K atoms). We installed it but did not test. We have not developed the protocols for safe operation (i.e. not to deposit all over the optics. K deposition also requires LT in most cases which is not routine yet.

- NEXAFS or RESPES. Sample tracking during optic motion would be required to maintain focus. Currently not calibrated.

Realistic NanoARPES Experiments

Here are some examples of what can be done realistically using the Fresnel Zone Plate. The capillary performance is much better, similar to microARPES and is not discussed here.

1) Single band map. It takes a few minutes to get a basic spectrum at high pass energy. Typically the real time display shows a recognizable band structure especially for good quality sharp bands like graphene, but the counts are sparsely spread across the detector. There is no fundamental energy (small pass energy) or angle resolution (small angle mode) limit, except the time it takes to accumulate the signal.

2) Finding samples. Unless samples are space filling the substrate (e.g. polycrystalline macroscopic film) it is essential to provide fiducial marks to help locate the sample. These are typically natural (sample edges, scratches, etc) but it helps if these are visible in our telescope. It is also very useful to acquire accurate distance of sample of interest to fiducial. Lastly, some idea of crystalline axes is useful, although we can compensate random oriented crystals by rotating the analyzer about its lens, it is better if sample axes are at least roughly aligned to the horizontal or vertical planes.

3) XY map of sample using core level or integrated (angle+energy signature) valence band. Such maps can be done pretty quickly, like 0.5 to 1.0 hours

4) XY Map, acquire single band structure image at each pixel. We typically do this in 5 to 10 hours

5) Fermi surface at a single spot: can be done in under a half hour, but presumably you want good quality so 1-2 hours minimum is needed.

Some typical count rate times for good quality graphene are indicated in the illustration below. The data in the rightmost panel is not (as implied by the label) from a single 120 nm region, but averaged over 3x3 pixel² (as suggested by the size of the red dot in the middle panel).