Methods
Methods are a core concept in the Digichem submission process. Methods represent a job (normally a calculation) to perform. Typically, each method corresponds to a single calculation type (a single point, a geometry optimisation etc.), but some compound calculations (such as a two part geometry optimisation and frequency calculation) can be represented by a single method. As a general rule, if the calculation would normally require more than one input file to submit manually, then it will require more than one method to submit with Digichem.
Each method contains three core components, which together describe how the calculation should be performed. These are:
The destination, which describes where to perform the calculation. Normally this corresponds to a specific server queue/partition.
The program, which describes the computational engine to run. This is where paths to the program binaries are set.
The calculation, which describes the calculation itself. Most of the configuration options you’ll need are set here.
Every part of the calculation is controlled by one of these three components, except for the starting geometry (the molecule). When you submit a calculation with Digichem, the molecules (starting geometries) and methods are specified separately. Each method acts like a template, and will apply to every molecule that is submitted. See the digichem submit command for more information.
Method components
The three components of a method are generalizable and interchangeable. This means, for example, that a program definition can be written only once and used with multiple different server queues and/or calculation types.
Many calculation options are generalized across different calculation programs too. For example, the setting to perform a geometry optimisation:
properties:
opt:
calc: True
Is the same for Gaussian, Turbomole, and Orca. This makes it easy to repeat calculations across different engines.
Specifying methods
Methods are normally specified in one of two ways:
Through the internal library.
In method files.
It is nearly always a good idea to specify the available server queues (destinations) and calculation engines (programs) in the library, because these settings rarely change. See setup for more information on configuring these files, if you have not done so already. Once setup, these destination and program definitions can then be reused in other method files to avoid having to write them again.
Digichem also comes pre-installed with a large number of calculation definitions which cover the most common calculation types.
Method files
Method files are written in YAML, which is a hierarchical, program independent text format. The normal file extension for Digichem method files is .sim (for in-silico method), and this is used by Digichem throughout. However, the file extension is entirely optional and can be omitted or changed if desired.
Each method file consists of three top-level options, corresponding to the three components of the method:
destination:
... # Destination/queue options go here.
program:
... # Program/engine options go here.
calculation:
... # Calculation specific options go here.
Each component can either manually specify the desired options, or refer back to a definition in the internal library.
To use a library definition, simply set the option to the appropriate definition name. These names are set during
setup, and can be changed in the relevant config files under ~/.config/digichem6/Destinations
and ~/.config/digichem6/Programs.
For example, if a Gaussian engine was setup with the following options:
link:
tag: Gaussian 16
... # Other options.
Then the corresponding method would simply be:
destination:
... # Destination/queue options go here.
program: Gaussian 16
calculation:
... # Calculation specific options go here.
The digichem config show command will display the available program and destination names:
$ digichem config show -b destination -d 1
Destination:
[1] 'SLURM'
[2] 'SLURM Array'
[3] 'PBS'
[4] 'PBS Array'
[5] 'Series'
[6] 'Parallel'
$ digichem config show -b program -d 1
Destination:
[1] 'Gaussian 16'
[2] 'Turbomole'
[3] 'Orca'
Each part of the method can either link to the library or be defined directly in the method file. Typically, this means that the destination and program options will use a library definition, while the calculation will be set manually:
destination: SLURM # Use an existing destination/queue called 'SLURM'.
program: Gaussian 16 # Use an existing progrm/engine called 'Gaussian 16'.
calculation: # Define a new calculation (a single-point at the HF level).
method:
hf:
calc: True
properties:
sp:
calc: True