ukaea · chris-ashe · Jun 3, 2026 · Jun 3, 2026 · Jun 3, 2026 · Jun 3, 2026
@@ -4,7 +4,7 @@ Two cost models are available, determined by the switch `cost_model`.
 
 ## 1990 cost model (`cost_model = 0`)
 
-This combines methods[^1] used in the TETRA code [^2] and the Generomak[^3] scheme. The costs are split into accounting categories[^4]. The best references for the algorithms used are[^5], and source file `costs.f90` in the code itself. The majority of the costed items have a unit cost associated with them. These values scale with (for example) power output, volume, component mass etc., and many are available to be changed via the input file. All costs and their algorithms correspond to 1990 dollars.
+This combines methods[^1] used in the TETRA code [^2] and the Generomak[^3] scheme. The costs are split into accounting categories[^4]. The best references for the algorithms used are[^5], and source file `costs.py` in the code itself. The majority of the costed items have a unit cost associated with them. These values scale with (for example) power output, volume, component mass etc., and many are available to be changed via the input file. All costs and their algorithms correspond to 1990 dollars.
 
 The unit costs of the components of the fusion power core are relevant to "first-of-a-kind" items. That is to say, the items are assumed to be relatively expensive to build as they are effectively prototypes and specialised tools and machines have perhaps been made specially to create them. However, if a "production line" has been set up, and R & D progress has allowed more experience to be gained in constructing the power core components, the cost will be reduced as a result. Variable `fkind` may be used to multiply the raw unit costs of the fusion power core items (by a factor less than one) to simulate this cost reduction for an *N<sup>th</sup>*-of-a-kind device. In other systems studies of fusion power plants[^6], values for this multiplier have ranged from 0.5 to 0.8.
 

@@ -5,17 +5,14 @@ optimisation figure of merit and constraints to the `PROCESS` code.
 
   **At all times the [`PROCESS` style guide](../development/standards.md) must be used.**
 
-!!! note
-
-    As the code is quickly converging towards a wholly Python codebase the respective files may change in type from `.f90` to `.py`.
 
 -----------------
 
 ## Add an input
 
 To add a `PROCESS` input, please follow below:
 
-1. Choose the most relevant module `XX` and add the variable in the `XX_variables` defined in `XX_variables.f90`.
+1. Choose the most relevant module `XX` and add the variable in the `XX_variables` defined in `XX_variables.py`.
 
 2. Add a description of the input variable below the declaration, using the FORD      formatting described in the standards section specifying the units.
 
@@ -26,11 +23,10 @@ To add a `PROCESS` input, please follow below:
 Here is an example of the code to add:
 
 
-Variable definition example in `tfcoil_variables.f90`:
-```fortran
-  real(dp) :: rho_tf_joints
-  !! TF joints surfacic resistivity [ohm.m]
-  !! Feldmetal joints assumed.
+Variable definition example in `tfcoil_variables.py`:
+```python
+  e_tf_coil_magnetic_stored: float = 0.0
+  """Stored magnetic energy in a single TF coil (J)"""
 ```
 
 Variable initialization example in `tf_coil.py`:
@@ -55,10 +51,10 @@ Code example in the `input.py` file:
 To add a `PROCESS` iteration variable please follow the steps below, in addition to the instructions for adding an input variable:
 
 
-1. The parameter `ipnvars` in module `numerics` of `numerics.f90` will normally be greater than the actual number of iteration variables, and does not need to be changed.
+1. The parameter `ipnvars` in module `numerics` of `numerics.py` will normally be greater than the actual number of iteration variables, and does not need to be changed.
 2. Append a new iteration number key to the end of the `ITERATION_VARIABLES` dictionary  in `iteration_variables.py`. The associated variable is the corresponding key value.
 3. Set the variable origin file and then the associated lower and upper bounds
-4. Update the `lablxc` description in `numerics.f90`.
+4. Update the `lablxc` description in `numerics.py`.
 
 It should be noted that iteration variables must not be reset elsewhere in the
 code. That is, they may only be assigned new values when originally
@@ -115,15 +111,13 @@ After following the instruction to add an input variable, you can make the varia
 
 
 `nsweep` comment example:
-```fortran
-
-  integer :: nsweep = 1
-  !! nsweep /1/ : switch denoting quantity to scan:<UL>
-  !!         <LI> 1  aspect
-  !!         <LI> 2  pflux_div_heat_load_max_mw
-  ...
-  !!         <LI> 54 GL_nbti upper critical field at 0 Kelvin
-  !!         <LI> 55 `dr_shld_inboard` : Inboard neutron shield thickness </UL>
+```python
+  nsweep: int = None
+  """Switch denoting quantity to scan:<UL>
+  <LI> 1  aspect
+  <LI> 2  pflux_div_heat_load_max_mw
+  <LI> 3  p_plant_electric_net_required_mw
+  <LI> 4  hfact
 ```
 
 `SCAN_VARIABLES` case example:

@@ -37,7 +37,7 @@ Switches should start with the `i_` prefix in their name, be of integer type and
 
 Use an uppercase single letter, word, or words. Separate words with underscores to improve readability.
 
-Refrain from declaring or typing known numerical constants directly in the code. Instead call the value from `constants.f90`
+Refrain from declaring or typing known numerical constants directly in the code. Instead call the value from `constants.py`
 If the constants doesn't exist then add it with a source link and uncertainty value.
 
 ---------------------
@@ -524,13 +524,13 @@ Try to keep names to a sensible length while also keeping the name explicit and
 
 The physical type of the variable should form the first part of the variable name, e.g. for plasma resistance the variable should be named:
 
-```fortran
+```python
 res_plasma = 1.0
 ```
 
 Another example would be pulse length
 
-```fortran
+```python
 time_pulse_length = 7200.0
 ```
 
@@ -540,17 +540,17 @@ time_pulse_length = 7200.0
 
 Inside PROCESS all variables should be in SI units unless otherwise stated. For example:
 
-```fortran
-! Fusion power [W]
-p_fusion_total = 1000.0d6
+```python
+p_fusion_total = 1e6
+"""Fusion power [W]"""
 ```
 
 If a variable is not in SI units then its units should be put at the end of of the variable name.
 Example:
 
-```fortran
-! Fusion power [MW]
-p_fusion_total_mw = 1000.0d0
+```python
+p_fusion_total_mw = 1000.0
+"""Fusion power [W]"""
 ```
 
 ---------------------
@@ -559,22 +559,21 @@ p_fusion_total_mw = 1000.0d0
 
 Coordinates should be defined as
 
-```fortran
-r_plasma_centre = 9.0d0
+```python
+r_plasma_centre = 9.0
 
-z_plasma_centre = 0.0d0
+z_plasma_centre = 0.0
 
 theta_ = 
 ```
 
 For dimensions
 
-```fortran
+```python
 dr_cs = 
 
 dz_cs = 
 
-dtheta_description =
 ```
 
 ---------------------

@@ -82,7 +82,7 @@ ixc = 169 * Achievable Temperature of the ECRH at the ignition point
 
 ## Code specifics
 
-The stellarator model is largely contained within source file `stellarator.f90`.
+The stellarator model is largely contained within source file `stellarator.py`.
 
 The model call is in the following order
 

@@ -125,8 +125,8 @@ There are 4 key functions for calculating the fusion reaction for the plasma. Th
 #### Detailed Steps
 1. **Initialize Bosch-Hale Constants**: Initializes the Bosch-Hale constants for the required reaction using predefined reaction constants stored in the BoschHaleConstants dataclass.
 2. **Calculate Fusion Reaction Rate**: Uses Simpson's rule to integrate the fusion reaction rate over the plasma profile.
-3. **Calculate Fusion Power Density**: Compute the fusion power density produced by the given reaction. Using the reaction energy calculated and stored in `constants.f90`. The reactant density is given by $\mathtt{f\_deuterium, f\_tritium}$ or $\mathtt{f\_helium3}$ multiplied by the volume averaged ion density. For the D-D reactions the fusion reaction rate is scaled with the output of [`deuterium_branching()`](#deuterium-branching-fraction--deuterium_branching) to simulate the different branching ratios.
-4. **Calculate Specific Fusion Power Densities**: Compute the fusion power density for alpha particles, neutrons and other charged particles, depending on the reaction. Energy branching fractions used are calculated and called from `constants.f90`.
+3. **Calculate Fusion Power Density**: Compute the fusion power density produced by the given reaction. Using the reaction energy calculated and stored in `constants.py`. The reactant density is given by $\mathtt{f\_deuterium, f\_tritium}$ or $\mathtt{f\_helium3}$ multiplied by the volume averaged ion density. For the D-D reactions the fusion reaction rate is scaled with the output of [`deuterium_branching()`](#deuterium-branching-fraction--deuterium_branching) to simulate the different branching ratios.
+4. **Calculate Specific Fusion Power Densities**: Compute the fusion power density for alpha particles, neutrons and other charged particles, depending on the reaction. Energy branching fractions used are calculated and called from `constants.py`.
 5. **Calculate Fusion Rate Densities**: Compute the total fusion rate density and fusion rates just for the alpha particles, neutrons and other charged particles, depending on the reaction.
 6. **Update Reaction Power Density**: Updates the object attribute for the specific reaction power density.
 7. **Sum Fusion Rates**: Call the [`sum_fusion_rates()`](#sum-the-fusion-rates--sum_fusion_rates) function to add the reaction to the global plasma power balance.

@@ -94,7 +94,7 @@ variable values that maximises or minimises a certain function of them,
 known as the figure of merit. 
 
 Several possible figures of merit are available, all of which are in the 
-source file `evaluators.f90`. 
+source file `evaluators.py`. 
 
 Switch `minmax` is used to control which figure of merit is to be used. If the 
 figure of merit is to be minimised, `minmax` should be **positive**, and if a 

@@ -249,8 +249,7 @@ def _call_models_once(self, xc: np.ndarray):
         """Call the physics and engineering models.
 
         This method is the principal caller of all the physics and
-        engineering models. Some are Fortran subroutines within modules, others
-        will be methods on Python model objects.
+        engineering models.
 
         Parameters
         ----------

@@ -47,7 +47,7 @@ class InputVariable:
     """A variable to be parsed from the input file."""
 
     module: Any
-    """The Fortran module that this variable should be set on."""
+    """The module that this variable should be set on."""
     type: type
     """The expected type of the variable"""
     range: tuple[NumberType, NumberType] | None = None
@@ -1156,7 +1156,6 @@ def parse_input_file(data_structure_obj: DataStructure):
             continue
 
         # matches (variable name, array index, value)
-        # NOTE: array index is Fortran-based hence starts at 1.
         line_match = re.match(
             r"([a-zA-Z0-9_]+)(?:\(([0-9]+)\))?[ ]*=[ ]*([ +\-a-zA-Z0-9.,]+).*",
             stripped_line,
@@ -1342,12 +1341,8 @@ def set_scalar_variable(name: str, value: ValidInputTypes, config: InputVariable
     """
     current_value = getattr(config.module, name, ...)
 
-    # use ... sentinel because None is probably a valid return from Fortran
-    # and definately will be when moving to a Python data structure
     if current_value is ...:
-        error_msg = (
-            f"Fortran module '{config.module}' does not have a variable '{name}'."
-        )
+        error_msg = f"Module '{config.module}' does not have a variable '{name}'."
         raise ProcessValueError(error_msg)
 
     setattr(config.module, name, value)
@@ -1356,9 +1351,6 @@ def set_scalar_variable(name: str, value: ValidInputTypes, config: InputVariable
 def set_array_variable(name: str, value: str, array_index: int, config: InputVariable):
     """Set an array variable in the `config.module`.
 
-    The way PROCESS input files are structured, each element of the array is provided on one line
-    so this function just needs to set the `value` at `array_index` (-1) because of Fortran-based indexing.
-
     Parameters
     ----------
     name :

@@ -59,7 +59,6 @@ def publish(self):
 
 
 class SourceDictionary(Dictionary):
-    # Dictionary created from Fortran source
     def __init__(self, name, dict_creator_func):
         Dictionary.__init__(self, name)
         # Function that creates the dict
@@ -221,7 +220,7 @@ def remove_comments(line):
 
 def dict_var_type():
     """Function to return a dictionary mapping variable name to variable type
-    eg. 'real_variable' or 'int_array'. Looks in input.f90 at the process
+    eg. 'real_variable' or 'int_array'. Looks in input.py at the process
     functions that read in variables from IN.DAT.
 
     Example of line we are looking for:
@@ -245,7 +244,7 @@ def dict_var_type():
 def dict_input_bounds():
     """Returns a dictionary matching variable names to dictionary containing
     upper and lower bounds that PROCESS checks variable lies between when
-    reading IN.DAT. Looks in input.f90 for parse_real_variable and
+    reading IN.DAT. Looks in input.py for parse_real_variable and
     parse_int_variable.
 
     Example of a line we are looking for:

@@ -1152,11 +1152,6 @@ def process_line(self, line_type, line):
             line_commentless = line.split("*")[0]
             array_name = line_commentless.split("(")[0]
             empty_array = dicts["DICT_DEFAULT"][array_name]
-            # empty_array is what the array is initialised to when it is
-            # declared in the Fortran. If the array is declared but not
-            # initialised until later in a separate "init" subroutine, then
-            # empty_array will be None. This is a side-effect of the need to
-            # re-initialise Fortran arrays in a separate subroutine.
 
             if empty_array is None:
                 # Array is declared but not initialised at the same time;
@@ -1405,61 +1400,25 @@ def process_array(self, line, empty_array):
         empty_array:
             Default array for this array name
         """
-        # TODO This is a mess after the Fortran module variable
-        # re-initialisation work. It is hard to see how this can be improved
-        # as the regex method of Fortran inspection (i.e. Python-Fortran
-        # dictionaries) is increasingly untenable. An attempt is made here,
-        # but with a view to the dictionaries method being dropped in future
-        # in light of increasing Python f2py conversion.
-
         line_commentless = line.split("*")[0] if "*" in line else line
 
         name = line_commentless.split("(")[0]
         index = int(line_commentless.split("(")[1].split(")")[0]) - 1
-        # The -1 assumes all Fortran arrays begin at 1: they don't in Process!
-        # This bug would cause a Fortran index of 0 to be interpreted as a
-        # Python index of -1 (last element). This didn't cause any exceptions,
-        # but would obviously set the wrong list index. This now throws
-        # exceptions when trying to access [-1] of an empty list []. Hence it
-        # must be guarded against with the following:
+
         if index == -1:
             index = 0
-        # This will cause Fortran index 0 and 1 to overwrite the same Python
-        # index of 0, which is clearly awful. However, it is equally bad to the
-        # previous case above, where Fortran [0] would overwrite Python [-1].
-        # There isn't a way of reconciling this without knowing whether the
-        # Fortran array begins at 0 or 1.
-        # TODO Either enforce Fortran arrays that start at 1 throughout, or
-        # find a way of determining the starting index of the array
 
         value = line_commentless.split("=")[-1].replace(",", "")
 
-        # Many arrays are now declared but not initialised until the "init"
-        # subroutine is run in each Fortran module, to allow re-initialisation
-        # on demand for a new run etc.
-        # In Fortran, we have a declared but uninitialised array of a given
-        # length. This results in an empty list in the value attribute here.
-        # However, we need to set the value for a given index. Therefore
-        # make a list of Nones, so that we can set the value for a given
-        # index. We don't know the length of the Fortran array (its value is
-        # []), so we have to make the Python list as long as this Fortran index.
-        # This way, at least the list is long enough for this Fortran index.
-        # TODO Again, this requires a more robust solution
-
         list_len = len(self.data[name].value)
         # Length of Python list in self.data
         max_list_index = list_len - 1
         # The greatest index in that Python list
         array_len = index + 1
-        # The Fortran array must be at least this long
 
-        # If the default array is an empty list, make a list of Nones of
-        # matching length to the Fortran array
         if len(empty_array) == 0:
             empty_array = [None] * array_len
 
-        # If the Pyhton list is an empty list, make a list of Nones of
-        # matching length to the Fortran array
         if list_len == 0:
             self.data[name].value = [None] * array_len
         # Check the Python list is long enough to store the new index

@@ -350,7 +350,6 @@ def set_variable_in_indat(in_dat, varname, value):
 
     elif "(" in varname:
         name = varname.split("(")[0]
-        # Fortran numbering converted to Python numbering!
         identity = int((varname.split("("))[1].split(")")[0]) - 1
         in_dat.change_array(name, identity, float(value))
 

@@ -16,7 +16,6 @@ def write(models, data, _outfile):
     models : process.main.Models
         physics and engineering model objects
     _outfile : int
-        Fortran output unit identifier
 
     """
     # ensure we are capturing warnings that occur in the 'output' stage as these are warnings

@@ -173,7 +173,6 @@ def ovarre(file, descr: str, varnam: str, value, output_flag: str = ""):
     if isinstance(value, np.ndarray):
         value = value.item()
     if isinstance(value, bytes):
-        # TODO: remove when Fortran is gone
         value = value.decode().strip()
     if isinstance(value, str):
         # try and convert the value to a float

@@ -189,7 +189,7 @@ def _missing_(cls, var):
 
 
 class Scan:
-    """Perform a parameter scan using the Fortran scan module."""
+    """Perform a parameter scan using the scan module."""
 
     def __init__(self, models: Model, solver: str, data: DataStructure):
         """Immediately run the run_scan() method.

@@ -15,7 +15,7 @@ class IterationVariable:
     name: str
     """The name of the variable"""
     module: str | Any
-    """The Fortran module that this variable should be set on."""
+    """The module that this variable should be set on."""
     lower_bound: float
     """The default lower bound of the iteration variable"""
     upper_bound: float
@@ -28,7 +28,6 @@ class IterationVariable:
     """If `module.name` is an array, the iteration variable can only modify
     `array_index` of that array.
 
-    NOTE: The indexes start at 0 (despite indexing Fortran arrays).
     """
 
 
@@ -263,8 +262,6 @@ def load_iteration_variables(data):
         variable_index = data_structure.numerics.ixc[i]
         iteration_variable = ITERATION_VARIABLES[variable_index]
 
-        # use ... as the default return value because None might be a valid return from Fortran?
-
         module = (
             getattr(data, iteration_variable.module)
             if isinstance(iteration_variable.module, str)