Advanced computing in electron microscopy

Chap5 Thin specimens

Phase grating approximation

  • The specimen potential is very small compare to the accelerating potential : where is the wave number of the accelerating potential and the interaction parameter.
  • The specimen is thin enough so that the effect of the potential can be summarized as an accumulated projected potential along the beam direction .

weak phase object

Since the interaction parameter the phase shift due to the projected potential of light atoms is usually quite small for Carbone at . As a result they may be considered as weak phase objects. However the potential exhibits a singularity in real space due to the atomic nucleus so the phase shift may be large if the incident beam passes very close to the nucleus.

electron density 3D potential projected potential

The atomic scattering factor can be calculated within the Born approximation which will be obtained from a simple Fourier transform. The projected potential can therefore conveniently be deduced using the Fourier projection theorem.

However, a proper calculation based on Moliere shows that the atomic scattering factor within the Born approximation makes significant errors in the phase of the scattered electrons especially at large scattering angles where the effect of the nucleus dominates which is actually the region where the neglected effect of bonding is most relevant. Applying the Born approximation for the atomic scattering amplitudes assumes that the atoms are weak phase objects.

Born approximation for C,Si,Cu,Au,U Moliere vs Born approximation for Gold

Image of thin specimens

Projected atomic potentials Image intensities

Chap6 Theory

Electron scattering is accounted for by solving Schrodinger's equation :

which is only an approximation of Dirac's equation the electron spin is neglected and relativistic effects not considered (accurate up to electron beam if the energy/wavelength of the electron already takes into account relativistic effects). The magnetic field due to the lenses can be ignored since it affects the electron path over a distance despite being .

ED modelling history

Date Name Advance
1928 Davisson,Bethe Electron diffraction exp,theory
1957 Cowley,Moodie Multislice approach
1959,1962 Fujimoto,Sturkey Scattering matrix Bloch
1961 Howie,Whelan Dynamical theory non crystal based
1972,1979 O,Keefe Multislice:First computer implementation SHRLI
1978 Glaeser,van Dyck Path integral formulation
1978,1987 Cowley,Kirkland Multi-slice CBED, ADF-STEM
1983 Kilaas Real space multislice

Notes

Prop Multislice Bloch
Number of beams Fourier components at the FFT stage Fourier components of the potential
Storage scales as
Computer time as
Bloch(up) vs mutlislice(down) CPU time

Bloch waves

see Bloch wave.

Fast electron wave equation

Using constant high energy collimated electrons, i.e. and slow varying envelope, i.e. assumptions, Schrodinger's equation reduces to:

where .

Howie Whelan

Assuming period potential and looking for solutions reduces to the system of differential equations :

Multislice

See multislice

Chap7 Applications

see multislice

Chap8 The programs

atompot+mulslice autoslic Si[100]

atompot and mulslic

These codes are essentially described in temsim

autoslic

Used for non periodic specimen :

  • The slices are computed automatically from the input file .xyz.
  • The projected potential is directly computed in real space.