If conda cannot solve the environment after attempting to install the dependencies create a new virtual environment with:
$ conda create -n autode_env $ conda activate autode_env
then install. The autodE environment will need to be activated each time a new shell is opened, with:
conda activate autode_env
.
If high and/or low level electronic structure methods cannot be found in your PATH
environment variable it is possible to set the paths manually in ade.Config
for e.g. XTB:
>>> from autode import methods, Config
>>> methods.get_lmethod()
autode.exceptions.MethodUnavailable
>>> Config.XTB.path = '/path/to/xtb/bin/xtb'
>>> methods.get_lmethod()
XTB(available = True)
alternatively, add to the PATH environment variable e.g. in bash:
$ export PATH=/path/to/xtb/bin/:$PATH
to set this permanently, add the above line to ~/.bash_profile or ~/.bashrc on Linux.
Automatically finding transition states (TS) with autodE can sometimes fail due to either a misidentification of the TS as incorrect or an insufficiently good TS guess geometry being found. If a TS hasn't been found first check the reaction_name_path.png visualisation of the initial path and proceed depending on whether a peak is or is not present.
Oh no! without an initial peak a TS won't be able to be found. Perhaps the reaction is electronically barrierless at this level of theory. If there are anions consider adding diffuse functions, if they are not already present e.g.:
>>> from autode import Config
>>> Config.ORCA.keywords.set_opt_basis_set('ma-def2-SVP')
If there is a peak then either (1) the TS guess geometry is not close enough to the TS for successful optimisation. To reduce the step size:
>>> from autode import Config
>>> Config.min_step_size = 0.02
Otherwise (2), it may be that the TS optimiser cannot locate the saddle point because of Hessian drift. Consider reducing the number of intermediate steps between Hessian updates (i.e. Config.ORCA.keywords.opt_ts for ORCA).
If you've found an interesting case where a TS cannot be found please do get in touch, we're always on the lookout for examples to improve the method!