Nano-scale energy filtered scanning confocal electron
microscopy (EFSCEM) using a double aberration-corrected
transmission electron microscope
Peng Wang1, Gavin Behan1, Masaki Takeguchi2, Ayako Hashimoto2, Kazutaka
Mitsuishi2, Masayuki Shimojo3, Angus I. Kirkland1, and Peter D. Nellist1*
Supporting Information (SI)
Establishment of the EFSCEM optical configuration
The EFSCEM configuration starts by establishing elastic confocal trajectories
following Ref [1], To regain the confocal point for inelastic electrons with a desired
energy loss, Eloss the accelerating voltage, E0 is increased by a factor k of Eloss
where
k  Cc2 (Cc1  Cc2 )
(S1)
and where, Cc1 and Cc2 are the chromatic aberration coefficients of the pre- and postspecimen optics respectively. The focal length of the pre-specimen optics thus
increases by
z1  kElossCc1 / Eo
(S2)
focusing the incident electrons at point c in Fig. 1 (b) (Main Text). At the increased
incident beam energy, electrons with the energy-loss of interest passing through the
post-specimen optics have energy, (k  1)Eloss relative to the beam energy at which
the BFSCEM trajectory was established, and therefore the focal length of the postspecimen optics is reduced
z 2  (k  1)ElossCc2 / E0 .
(S3)
Substituting (1) into (2) and (3) shows that z1  z 2 and therefore the changes in
focal length of the pre- and post-specimen optics are exactly compensated, regaining
the confocal trajectory for the desired energy loss. The final alignment step requires
simply that the sample is lowered by a distance z1 to the confocal point.
Establishing these conditions now takes a skilled operator approximately 30 minutes.
Experimental:
Instrument:
The experiments reported were performed using the Oxford-JEOL 2200MCO with
third order aberration correctors for both pre and post specimen optics and an in
column energy filter (an upgraded version of the instrument described in Reference
[2]).
Samples:
For both EFSTEM and the 1D EFSCEM experiments, the sample used was a holey
carbon film obtained from Agar Scientific. Independent measurements of the
thickness of the sample were carried out using the ratio of the strength of inelastic to
elastic scattering [3] to give an estimate for the thickness of 25±5nm. For the 2D
EFSCEM experiment, the specimen used was a perforated carbon film decorated with
Au nanoparticles. Using same methods ,the thickness of this sample at point 1 (Fig. 3
(a) Main Text) was estimated at 65±7nm and 32±3nm at point 2.
EFSTEM experimental configuration:
EFSTEM is a non-confocal configuration as shown in Fig. S1. The one-dimensional
EFSTEM scans in the z-direction were performed by changing the height of the
sample using a piezo-driven sample stage. The sample was illuminated with a beam
convergence semi-angle of 22 mrad. EFSTEM data was recorded using an energy
filter [4] with a 20 eV Energy-selecting window (ESW), centred at an energy loss of
290 eV in order to record the carbon K-edge, and a collection semi-angle of 10 mrad
into the spectrometer was used.
Figure S1 shows the overall optical configuration for EFSTEM.
EFSCEM experimental configuration:
For both 1D and 2D EFSCEM optical sectioning experiments, the microscope was
adjusted to the confocal mode for an energy loss of 290 eV to collect carbon K-edge
scattering.
1D EFSCEM
For 1D EFSCEM data acquisition, the sample was moved along the z-axis using a
piezo-stage. Semi-angles of 22 and 36 mrad were used for the pre- and post-specimen
lens apertures respectively, and data was recorded using ESWs of 1, 5 and 20 eV
centred at a 290eV energy-loss and also with no ESW present. Series of images of the
EF- confocal probe were recorded on a 4kx4k Gatan CCD camera with 4x pixel
binning and a 4s dwell time. A movie composed of a series of experimental probe
images with a 5eV ESW centred at 290 eV is shown as Movie S1. A virtual collector
aperture with a diameter of 0.4 nm was applied to the stack of CCD images by
integrating the intensities within a circular region centred at each probe image as
shown in the insert (a) in Fig.2 (Main Text). The insert (b) in Fig. 2 shows that the
signal is spread outside this aperture when the sample is located away from the
confocal point giving rise to a strongly depth dependent signal (Fig. 2 (Main Text))
that accurately measures the height of the carbon film.
2D EFSCEM
For the 2D experiments, a purpose designed piezo-driven stage-scanning system [5]
was used to scan the sample in either two-dimensional lateral (x-y) scans, or to slice in
depth, forming an (x-z) scan. For these experiments the CCD camera was replaced
with a physical collector aperture corresponding to a diameter of 0.32 nm positioned
in front of a fast electron detector, allowing two dimensional images to be recorded in
approximately 40 s for a 512 x 512 pixel array [5]. The post-specimen optics were
adjusted to give a numerical aperture with a 17.7 mrad semi-angle to exclude effects
due to high order aberrations whilst the pre-specimen optical numerical aperture was
maintained at 22 mrad. A 20 eV ESW was centered at an energy-loss of 290eV. 2D
scans with an array size of 256 x 256 pixels were recorded in 35s by averaging 50
measurements to improve the signal to noise ratio in the final data. The pixel size in
the images is 1.75 nm for both x and y directions and 21.09 nm for the z direction.
The images in shown Figure 3 (Main Text) were cropped from these raw images. The
depth line profiles in Fig. 3 c) and f) (Main Text) were subsequently generated using
an integration width of 10 pixels along the dotted lines L1 (—) and L2 (---) in
Figs. 3 b) and e) (Main Text), respectively.
Z-Response (or Plane spread function)
For a completely incoherent scattering object function
V (R )
located at a height, z=0
and in the confocal condition, the intensity of the three-dimensional incoherent SCEM
image[1] can be written as:
I (R 1 , z)   V (R) P1 (R  R 1 , z) P2 (R 1  R, z) dR (S4)
2
2
in which we have used the notation provided in Reference[1]. The point spread
function (PSF) for incoherent confocal imaging can thus be identified as:
2
PSF (R, z )  P1 (R, z ) P2 ( R, z )
2
(S5)
The z-response describes the depth resolution of a laterally extended object. For an
infinitely thin incoherent scattering plane, the z-response is given by integrating the
PSF in the R plane as:
 SCEM ( z)   P1 (R, z) P2 (R, z) dR (S6)
2
2
Using the same analysis, for the STEM geometry, the z-response can be written as:
 STEM ( z)   P1 (R, z) dR
2
(S7)
which is constant due to the total intensity of the beam propagating through the plane
for any value of z being unchanged. This lack of z-response for STEM is evident
experimentally in Fig. 2 (Main Text), which confirms that the STEM cannot provide
depth determination for an extended object.
[1] P. D. Nellist et al., Applied Physics Letters 89, 124105 (2006).
[2] J. L. Hutchison et al., Ultramicroscopy 103, 7 (2005).
[3] R. F. Egerton, Electron Energy-loss Spectroscopy in the Electron Microscope
(Springer, 1996).
[4] K. Tsuno et al., J Electron Microsc 46, 357 (1997).
[5] M. Takeguchi et al., J Electron Microsc 57, 123 (2008).