This data deposit contains the main files needed to reproduce the results in Atomic resolution imaging of the strength and 3D direction of atomic displacements in functional oxide thin films by MacLaren et al.

To reproduce figure 2, the data is calculated from the 4DSTEM dataset inside the file:
binned_diff_20210421_160712.hdf5
The calculations are performed by the jupyter notebook:
Figure2 PolarSpeedup BinnedDif160712.ipynb
This runs with slightly older versions of hyperspy (1.7.1) and py4DSTEM (0.13.6).  There is no known reason why it shouldnt run with more recent versions after some minor modifications (e.g. py4DSTEM.io.datastructure.DataCube will need to be updated to py4DSTEM.DataCube).  Also file path specifications will have to be adjusted to your own computer storage architecture.

Figure 3 is calculated by the notebook:
Figure3_PatternAverage160712v2.ipynb
This uses no special python libraries, just standard ones present in any normal python environment
The HAADF image is:
binned_diff_20210421_160712_fast_HAADFfull.npy
which is calculated by a version of the Figure 2 notebook
similarly, the parameter maps come from:
binned_diff_20210421_160712_fast_full_fitparams.npy
which is calculated by the same notebook
The A site positions are in the file:
binned_diff_20210421_160712_fast_HAADFfull_zoom_Alist.p
which is generated by atomap.  A notebook for doing this is:
FindColumns_v2_for template average.ipynb
The simulation parameters are in:
simparams.npy
which is calculated from a set of simulations arranged in a 4DSTEM like structure in a numpy file (because it is relatively small  just covering one supercell):
L2CM_111_scan_a20_z+0_4DSTEM.npy
The polar transformed version of this is in:
L2CM_111_scan_a20_z+0_4DSTEM_PT.npy
Finally, this is fitted using a notebook much like that for Figure 2 to give the fit parameters in the file:
L2CM_111_scan_a20_z+0_4DSTEM_params.npy
At the same time, this makes the HAADF image in:
L2CM_111_scan_a20_z+0_HAADF.npy

Figure 4 is calculated by the notebook:
L2CM_111_scan_a20_z+0_4DSTEM_PT.npy
This was all calculated from the dataset in 
binned_diff_20210421_154716.hdf5
the parameter maps are in:
binned_diff_20210421_154716_fast_full_fitparams.npy
and the HAADF in:
binned_diff_20210421_154716_HAADF.npy
A lot of the mapping is driven from the detection of columns in parameter maps using atomap with the following notebook:
FindColumns_v2_Figure4.ipynb
Which resulted in the following for A-site positions:
binned_diff_20210421_154716_HAADF_Aposlist.p
and this for A1 peaks:
binned_diff_20210421_154716_fast_full_fitparams_A1aposlist.p

Full files for every simulation presented in the Supplemental Information are not provided as this would be large and complex, but can be provided individually on reasonable request.
