Monte Carlo Configuration

The montecarlo section describes the parameters for the Monte Carlo radiation transport and convergence criteria:

type	object
properties
nthreads	The number of Numba threads for parallelisation. Must be between 1 and the environment variable NUMBA_NUM_THREADS (by default NUMBA_NUM_THREADS is equal to the number of CPU cores on the local system).
	type	number
	multipleOf	1.0
	default	1
seed	Seed for the random number generator
	type	number
	multipleOf	1.0
	default	23111963
no_of_packets	Seed for the random number generator
	type	number
	multipleOf	1.0
iterations	Number of maximum iterations
	type	number
	multipleOf	1.0
last_no_of_packets	This can set the number of packets for the last run. If set negative it will remain the same as all other runs.
	type	number
	multipleOf	1.0
	default	-1
no_of_virtual_packets	Setting the number of virtual packets for the last iteration.
	type	number
	multipleOf	1.0
	default	0
virtual_spectrum_spawn_range	Limits of virtual packet spawn spectrum (giving maximum and minimum packet frequency)
	type	object
	default
	properties
	start	type	quantity
		default	0 angstrom
	end	type	quantity
		default	inf angstrom
enable_reflective_inner_boundary	experimental feature to enable a reflective boundary.
	type	boolean
	default	False
inner_boundary_albedo	albedo of the reflective boundary
	type	number
	default	0.0
convergence_strategy	default	type	damped
	oneOf	Damped Convergence Strategy
		Custom Convergence Strategy
enable_full_relativity	Enables a more complete treatment of relativitic effects. This includes angle aberration as well as use of the fully general Doppler formula.
	type	boolean
	default	False
debug_packets	Decide whether to go into debugging mode. [EXPERIMENTAL FEATURE DO NOT RELY ON IT]
	type	boolean
	default	False
logger_buffer	Provides option to not log every line.
	type	number
	default	1
single_packet_seed	If debug_packets is true, this is the seed for the only packet.
	type	number
	default	-1
additionalProperties	False

The seed parameter seeds the random number generator first for the creation of the packets (\(\nu\) and \(\mu\)) and then the interactions in the actual Monte Carlo process. The no_of_packets parameter can take a float number for input convenience and gives the number of packets normally used in each Monte Carlo loop. The parameters last_no_of_packets and no_of_virtual_packets influence the last run of the Monte Carlo loop (which calculates the final spectrum!) when the radiation field should have converged. last_no_of_packets is normally higher than no_of_packets to create a less noisy output spectrum. no_of_virtual_packets can also be set to greater than 0 (a useful number is 3) to use the Virtual Packet formalism. Increasing this number drastically increases computational costs (and memory requirements if they are logged). The iterations parameter describes the maximum number of Monte Carlo loops executed in a simulation before it ends. Convergence criteria can be used to make the simulation stop sooner when the convergence threshold has been reached (see Convergence).

Convergence Strategy

The convergence_criteria section has a type keyword. Currently, one type is allowed: damped. All convergence criteria can be specified separately for the three variables for which convergence can be checked (t_inner, t_rad, ws) by specifying subsections in the convergence_criteria of the same name. These then override the defaults. Two more schemas are presented that further explain the damped and custom convergence strategies:

Damped Convergence Strategy

type	object
properties
type	enum	damped
default	damped
stop_if_converged	stop plasma iterations before number of specified iterations are reached if the simulation is plasma and inner boundary state is converged
	type	boolean
	default	False
fraction	the fraction of shells that have to converge to the given convergence threshold. For example, 0.8 means that 80% of shells have to converge to the threshold that convergence is established
	type	number
	minimum	0
	default	0.8
hold_iterations	the number of iterations that the convergence criteria need to be fulfilled before TARDIS accepts the simulation as converged
	type	number
	multipleOf	1.0
	default	3
damping_constant	damping constant
	type	number
	minimum	0
	default	1.0
threshold	specifies the threshold that is taken as convergence (i.e. 0.05 means that the value does not change more than 5%)
	type	number
	minimum	0
	default	0.05
t_inner	type	object
	properties
	damping_constant	damping constant
		type	number
		minimum	0
		default	0.5
	threshold	specifies the threshold that is taken as convergence (i.e. 0.05 means that the value does not change more than 5%)
		type	number
		minimum	0
	additionalProperties	False
t_rad	type	object
	properties
	damping_constant	damping constant
		type	number
		minimum	0
		default	0.5
	threshold	specifies the threshold that is taken as convergence (i.e. 0.05 means that the value does not change more than 5%)
		type	number
		minimum	0
	additionalProperties	False
w	type	object
	properties
	damping_constant	damping constant
		type	number
		minimum	0
		default	0.5
	threshold	specifies the threshold that is taken as convergence (i.e. 0.05 means that the value does not change more than 5%)
		type	number
		minimum	0
	additionalProperties	False
lock_t_inner_cycles	The number of cycles to lock the update of the inner boundary temperature. This process helps with convergence. The default is to switch it off (1 cycle)
	type	number
	multipleOf	1.0
	default	1
t_inner_update_exponent	L=4*pi*r**2*T^y
	type	number
	default	-0.5
additionalProperties	False

damped only has one parameter damping-constant and does not check for convergence. This can be used to fix the temperature of the inner boundary.

Custom Convergence Strategy

type	object
properties
type	enum	custom
description	Use this convergence_strategy for your specific needs. You need to change the codebase accordingly
additionalProperties	False

Parallel Execution with Numba

The montecarlo section of the Configuration file accepts the parameter nthreads which sets the number of threads to be used for parallelisation. Setting the value of the parameter between 1 and the environment variable NUMBA_NUM_THREADS (which is, by default, the number of CPU cores on your system) will automatically invoke Numba to parallelise the code. (See Configuration (Required Input) section).