BaseTrajectoryModule

class BaseTrajectoryModule(name=None)

Base implementation for Trajectory modules compatible with the Bloch simulation module as well as containing a guaranteed method for position pre-calculation on call.

All derived classes must implement the abstract method “increment_particles” which must be compatible with a tf.function decoration. The __call__ function must also be implemented but is not meant to be called inside a tf.function

Methods:

__call__(initial_positions, timing, **kwargs)	Evaluates the positions for particles at given initial positions for all times specified in the timing argument.
increment_particles(particle_positions, dt, ...)	Evaluates the new position of particles at given locations r for a temporal step width dt.

abstract __call__(initial_positions, timing, **kwargs)

Evaluates the positions for particles at given initial positions for all times specified in the timing argument. For implementations using the increment_particles function in a loop the maximal time-delta must be specified.

Parameters:

• initial_positions (Tensor) – (N, 3)

• timing (Tensor) – (T, )

• kwargs – Can vary in concrete implementation

Return type:

(Tensor, dict)

Returns:

• r_new [tf.Tensor, (T, N, 3)]

• additional_fields [dict] containing the lookup values for each step

abstract increment_particles(particle_positions, dt, **kwargs)

Evaluates the new position of particles at given locations r for a temporal step width dt. If the concrete implementation involves a look up (e.g. velocity fields) the values at the old location is also returned as dictionary.

Note

concrete implementations must be compatible with tf.function decoration

Parameters:

• particle_positions (Tensor) – (N, 3) Current particle positions.

• dt (Tensor) – (,) Temporal step width in milliseconds to evaluate the next positions

• kwargs – Can vary in concrete implementation

Return type:

(Tensor, dict)

Returns:

r_new [tf.Tensor, (N, 3)], additional_fields [dict] containing the lookup values