pychnosz 1.1.12__cp310-cp310-macosx_15_0_x86_64.whl

pychnosz/core/affinity.py

@@ -0,0 +1,1256 @@
+"""
+Affinity calculation module.
+
+This module provides Python equivalents of the R functions in affinity.R:
+- affinity(): Calculate chemical affinities of formation reactions
+- Energy calculation utilities and argument processing
+- Variable expansion and multi-dimensional calculations
+
+Author: CHNOSZ Python port
+"""
+
+import numpy as np
+import pandas as pd
+from typing import Union, List, Optional, Dict, Any, Tuple
+import warnings
+
+from .thermo import thermo
+from .basis import get_basis, is_basis_defined
+from .species import get_species, is_species_defined
+from .subcrt import subcrt
+
+
+class AffinityError(Exception):
+    """Exception raised for affinity-related errors."""
+    pass
+
+
+def affinity(messages: bool = True, basis: Optional[pd.DataFrame] = None,
+             species: Optional[pd.DataFrame] = None, iprotein: Optional[Union[int, List[int], np.ndarray]] = None,
+             loga_protein: Union[float, List[float]] = 0.0, **kwargs) -> Dict[str, Any]:
+    """
+    Calculate affinities of formation reactions.
+
+    This function calculates chemical affinities for the formation reactions of
+    species of interest from user-selected basis species. The affinities are
+    calculated as A/2.303RT where A is the chemical affinity.
+
+    Parameters
+    ----------
+    messages : bool, default True
+        Whether to print informational messages
+    basis : pd.DataFrame, optional
+        Basis species definition to use (if not using global basis)
+    species : pd.DataFrame, optional
+        Species definition to use (if not using global species)
+    iprotein : int, list of int, or array, optional
+        Build proteins from residues (row numbers in thermo().protein)
+    loga_protein : float or list of float, default 0.0
+        Activity of proteins (log scale)
+    **kwargs : dict
+        Variable arguments defining calculation conditions:
+        - Basis species names (e.g., CO2=[-60, 20, 5]): Variable basis species activities
+        - T : float or list, Temperature in °C
+        - P : float or list, Pressure in bar
+        - property : str, Property to calculate ("A", "logK", "G", etc.)
+        - exceed_Ttr : bool, Allow extrapolation beyond transition temperatures
+        - exceed_rhomin : bool, Allow calculations below minimum water density
+        - return_buffer : bool, Return buffer activities
+        - balance : str, Balance method for protein buffers
+
+    Returns
+    -------
+    dict
+        Dictionary containing:
+        - fun : str, Function name ("affinity")
+        - args : dict, Arguments used in calculation
+        - sout : dict, Subcrt calculation results
+        - property : str, Property calculated
+        - basis : pd.DataFrame, Basis species definition
+        - species : pd.DataFrame, Species of interest definition
+        - T : float or array, Temperature(s) in Kelvin
+        - P : float or array, Pressure(s) in bar
+        - vars : list, Variable names
+        - vals : dict, Variable values
+        - values : dict, Calculated affinity values by species
+
+    Examples
+    --------
+    >>> import pychnosz
+    >>> pychnosz.reset()
+    >>> pychnosz.basis(["CO2", "H2O", "NH3", "H2S", "H+", "O2"])
+    >>> pychnosz.species(["glycine", "tyrosine", "serine", "methionine"])
+    >>> result = pychnosz.affinity(CO2=[-60, 20, 5], T=350, P=2000)
+    >>> print(result['values'][1566])  # Glycine affinities
+
+    >>> # With proteins
+    >>> import pandas as pd
+    >>> aa = pd.read_csv("POLG.csv")
+    >>> iprotein = pychnosz.add_protein(aa)
+    >>> pychnosz.basis("CHNOSe")
+    >>> a = pychnosz.affinity(iprotein=iprotein, pH=[2, 14], Eh=[-1, 1])
+
+    Notes
+    -----
+    This implementation maintains complete fidelity to R CHNOSZ affinity():
+    - Identical argument processing including dynamic basis species parameters
+    - Same variable expansion and multi-dimensional calculations
+    - Exact energy() function behavior for property calculations
+    - Identical output structure and formatting
+    - Support for protein calculations via iprotein parameter
+    """
+
+    # Get thermo object for protein handling
+    thermo_obj = thermo()
+
+    # Handle iprotein parameter
+    ires = None
+    original_species = None
+    if iprotein is not None:
+        # Convert to array
+        if isinstance(iprotein, (int, np.integer)):
+            iprotein = np.array([iprotein])
+        elif isinstance(iprotein, list):
+            iprotein = np.array(iprotein)
+
+        # Check all proteins are available
+        if np.any(np.isnan(iprotein)):
+            raise AffinityError("`iprotein` has some NA values")
+        if thermo_obj.protein is None or not np.all(iprotein < len(thermo_obj.protein)):
+            raise AffinityError("some value(s) of `iprotein` are not rownumbers of thermo().protein")
+
+        # Add protein residues to the species list
+        # Amino acids in 3-letter code
+        aminoacids_3 = ["Ala", "Cys", "Asp", "Glu", "Phe", "Gly", "His", "Ile", "Lys", "Leu",
+                        "Met", "Asn", "Pro", "Gln", "Arg", "Ser", "Thr", "Val", "Trp", "Tyr"]
+
+        # Use _RESIDUE notation (matches R CHNOSZ affinity.R line 84)
+        resnames_residue = ["H2O_RESIDUE"] + [f"{aa}_RESIDUE" for aa in aminoacids_3]
+
+        # Save original species
+        from .species import species as species_func
+        original_species = get_species() if is_species_defined() else None
+
+        # Add residue species with activity 0 (all in "aq" state)
+        species_func(resnames_residue, state="aq", add=True, messages=messages)
+
+        # Get indices of residues in species list
+        species_df_temp = get_species()
+        ires = []
+        for name in resnames_residue:
+            idx = np.where(species_df_temp['name'] == name)[0]
+            if len(idx) > 0:
+                ires.append(idx[0])
+        ires = np.array(ires)
+
+    # Check if basis and species are defined (use provided or global)
+    if basis is None:
+        if not is_basis_defined():
+            raise AffinityError("basis species are not defined")
+        basis_df = get_basis()
+    else:
+        basis_df = basis
+
+    if species is None:
+        if not is_species_defined():
+            raise AffinityError("species are not defined")
+        species_df = get_species()
+    else:
+        species_df = species
+
+    # Process arguments
+    args_orig = dict(kwargs)
+
+    # Handle argument recall (if first argument is previous affinity result)
+    if len(args_orig) > 0:
+        first_key = list(args_orig.keys())[0]
+        first_value = args_orig[first_key]
+        if (isinstance(first_value, dict) and
+            first_value.get('fun') == 'affinity'):
+            # Update arguments from previous result
+            aargs = first_value.get('args', {})
+            # Update with new arguments (skip the first one)
+            new_args = dict(list(args_orig.items())[1:])
+            aargs.update(new_args)
+            return affinity(**aargs)
+
+    # Process energy arguments
+    args = energy_args(args_orig, messages, basis_df=basis_df)
+
+    # Get property to calculate
+    property_name = args.get('what', 'A')
+
+    # Get thermo data
+    thermo_obj = thermo()
+    # basis_df and species_df are already set above
+
+    # Determine if we need specific property calculation
+    if property_name and property_name != 'A':
+        # Calculate specific property using energy function
+        energy_result = energy(
+            what=property_name,
+            vars=args['vars'],
+            vals=args['vals'],
+            lims=args['lims'],
+            T=args['T'],
+            P=args['P'],
+            IS=args.get('IS', 0),
+            exceed_Ttr=kwargs.get('exceed_Ttr', True),
+            exceed_rhomin=kwargs.get('exceed_rhomin', False),
+            basis_df=basis_df,
+            species_df=species_df,
+            messages=messages
+        )
+        affinity_values = energy_result['a']
+        energy_sout = energy_result['sout']
+    else:
+        # Calculate affinities (A/2.303RT)
+        energy_result = energy(
+            what='A',
+            vars=args['vars'],
+            vals=args['vals'],
+            lims=args['lims'],
+            T=args['T'],
+            P=args['P'],
+            IS=args.get('IS', 0),
+            exceed_Ttr=kwargs.get('exceed_Ttr', True),
+            exceed_rhomin=kwargs.get('exceed_rhomin', False),
+            basis_df=basis_df,
+            species_df=species_df,
+            messages=messages
+        )
+        affinity_values = energy_result['a']
+        energy_sout = energy_result['sout']
+
+    # Handle protein affinity calculations if iprotein was provided
+    if iprotein is not None and ires is not None:
+        # Calculate protein affinities from residue affinities using group additivity
+        # Normalize loga_protein to match number of proteins
+        if isinstance(loga_protein, (int, float)):
+            loga_protein_arr = np.full(len(iprotein), loga_protein)
+        else:
+            loga_protein_arr = np.array(loga_protein)
+            if len(loga_protein_arr) < len(iprotein):
+                loga_protein_arr = np.resize(loga_protein_arr, len(iprotein))
+
+        # Calculate affinity for each protein
+        protein_affinities = {}
+
+        for ip, iprot in enumerate(iprotein):
+            # Get protein amino acid composition from thermo().protein
+            # Columns 4:24 contain chains and amino acid counts (0-indexed: columns 4-23)
+            protein_row = thermo_obj.protein.iloc[iprot]
+            aa_counts = protein_row.iloc[4:24].values.astype(float)
+
+            # Calculate protein affinity by summing residue affinities weighted by composition
+            # affinity_values keys are ispecies indices
+            # Get the ispecies for each residue
+            species_df_current = get_species()
+            residue_ispecies = species_df_current.iloc[ires]['ispecies'].values
+
+            # Initialize protein affinity with same shape as residue affinities
+            first_residue_key = residue_ispecies[0]
+            if first_residue_key in affinity_values:
+                template_affinity = affinity_values[first_residue_key]
+                protein_affinity = np.zeros_like(template_affinity)
+
+                # Sum up contributions from all residues
+                for i, res_ispecies in enumerate(residue_ispecies):
+                    if res_ispecies in affinity_values:
+                        residue_contrib = affinity_values[res_ispecies] * aa_counts[i]
+                        protein_affinity = protein_affinity + residue_contrib
+
+                # Subtract protein activity
+                protein_affinity = protein_affinity - loga_protein_arr[ip]
+
+                # Use negative index to denote protein (matches R CHNOSZ convention)
+                protein_key = -(iprot + 1)  # Negative of (row number + 1)
+                protein_affinities[protein_key] = protein_affinity
+
+        # Add ionization affinity if H+ is in basis (matching R CHNOSZ behavior)
+        if 'H+' in basis_df.index:
+            if messages:
+                print("affinity: ionizing proteins ...")
+
+            # Get protein amino acid compositions
+            from ..biomolecules.proteins import pinfo
+            from ..biomolecules.ionize_aa import ionize_aa
+
+            # Get aa compositions for these proteins
+            aa = pinfo(iprotein)
+
+            # Determine pH values from vars/vals or basis
+            # Check if H+ is a variable
+            if 'H+' in args['vars']:
+                # H+ is a variable - get pH from vals
+                iHplus = args['vars'].index('H+')
+                pH_vals = -np.array(args['vals'][iHplus])  # pH = -log(a_H+)
+            else:
+                # H+ is constant - get from basis
+                pH_val = -basis_df.loc['H+', 'logact']  # pH = -log(a_H+)
+                pH_vals = np.array([pH_val])
+
+            # Get T values (already processed earlier)
+            T_vals = args['T']
+            if isinstance(T_vals, (int, float)):
+                T_celsius = T_vals - 273.15
+            else:
+                T_celsius = T_vals - 273.15
+
+            # Get P values
+            P_vals = args['P']
+
+            # Calculate ionization affinity
+            # ionize_aa expects arrays, so ensure T, P, pH are properly shaped
+            # For grid calculations, we need to expand T, P, pH into a grid matching the affinity grid
+            if len(args['vars']) >= 2:
+                # Multi-dimensional case - create grid
+                # Figure out which vars are T, P, H+
+                var_names = args['vars']
+                has_T_var = 'T' in var_names
+                has_P_var = 'P' in var_names
+                has_Hplus_var = 'H+' in var_names
+
+                # Build T, P, pH grids matching the affinity calculation grid
+                if has_T_var and has_Hplus_var:
+                    # Both T and pH vary - create meshgrid
+                    T_grid, pH_grid = np.meshgrid(T_celsius, pH_vals, indexing='ij')
+                    T_flat = T_grid.flatten()
+                    pH_flat = pH_grid.flatten()
+                    if isinstance(P_vals, str):
+                        P_flat = np.array([P_vals] * len(T_flat))
+                    else:
+                        P_flat = np.full(len(T_flat), P_vals if isinstance(P_vals, (int, float)) else P_vals[0])
+                elif has_T_var:
+                    # Only T varies
+                    T_flat = T_celsius if isinstance(T_celsius, np.ndarray) else np.array([T_celsius])
+                    pH_flat = np.full(len(T_flat), pH_vals[0])
+                    P_flat = np.array([P_vals] * len(T_flat)) if isinstance(P_vals, str) else np.full(len(T_flat), P_vals if isinstance(P_vals, (int, float)) else P_vals[0])
+                elif has_Hplus_var:
+                    # Only pH varies
+                    pH_flat = pH_vals
+                    T_flat = np.full(len(pH_flat), T_celsius if isinstance(T_celsius, (int, float)) else T_celsius[0])
+                    P_flat = np.array([P_vals] * len(pH_flat)) if isinstance(P_vals, str) else np.full(len(pH_flat), P_vals if isinstance(P_vals, (int, float)) else P_vals[0])
+                else:
+                    # No T or pH variables
+                    T_flat = np.array([T_celsius if isinstance(T_celsius, (int, float)) else T_celsius[0]])
+                    pH_flat = pH_vals
+                    P_flat = np.array([P_vals] if isinstance(P_vals, str) else [P_vals if isinstance(P_vals, (int, float)) else P_vals[0]])
+            else:
+                # Single or no variable case
+                T_flat = np.array([T_celsius if isinstance(T_celsius, (int, float)) else T_celsius[0]])
+                pH_flat = pH_vals if isinstance(pH_vals, np.ndarray) else np.array([pH_vals[0] if hasattr(pH_vals, '__getitem__') else pH_vals])
+                P_flat = np.array([P_vals] if isinstance(P_vals, str) else [P_vals if isinstance(P_vals, (int, float)) else P_vals[0]])
+
+            # Call ionize_aa to get ionization affinity
+            ionization_result = ionize_aa(aa, property="A", T=T_flat, P=P_flat, pH=pH_flat)
+
+            # Add ionization affinity to formation affinity for each protein
+            for ip, iprot in enumerate(iprotein):
+                protein_key = -(iprot + 1)
+                ionization_affinity = ionization_result.iloc[:, ip].values
+
+                # Reshape to match formation affinity dimensions if needed
+                formation_affinity = protein_affinities[protein_key]
+                if isinstance(formation_affinity, np.ndarray):
+                    if formation_affinity.shape != ionization_affinity.shape:
+                        # Reshape ionization affinity to match formation affinity
+                        ionization_affinity = ionization_affinity.reshape(formation_affinity.shape)
+
+                # Add ionization to formation affinity
+                protein_affinities[protein_key] = formation_affinity + ionization_affinity
+
+        # Replace affinity_values with protein affinities
+        affinity_values = protein_affinities
+
+        # Calculate stoichiometric coefficients for proteins using matrix multiplication
+        # This matches R CHNOSZ: protbasis <- t(t((resspecies[ires, 1:nrow(thermo$basis)])) %*% t((thermo$protein[iprotein, 5:25])))
+        # IMPORTANT: Get the species list BEFORE deletion
+        species_df_with_residues = get_species()
+
+        # Extract basis species coefficients from residue species (rows = residues, cols = basis species)
+        # ires contains indices of residues in the species list
+        # We need the columns corresponding to basis species
+        basis_cols = list(basis_df.index)  # e.g., ['CO2', 'H2O', 'NH3', 'H2S', 'e-', 'H+']
+
+        # Create residue coefficient matrix (n_residues x n_basis)
+        # resspecies[ires, 1:nrow(thermo$basis)] in R
+        res_coeffs = species_df_with_residues.iloc[ires][basis_cols].values.astype(float)
+
+        # Get amino acid composition matrix (n_proteins x n_residues)
+        # thermo$protein[iprotein, 5:25] in R (columns 5-25 contain chains and 20 amino acids)
+        # In Python (0-indexed): columns 4:24 contain chains and 20 amino acids
+        aa_composition = []
+        for iprot in iprotein:
+            protein_row = thermo_obj.protein.iloc[iprot]
+            # Columns 4:24 contain: chains, Ala, Cys, Asp, Glu, Phe, Gly, His, Ile, Lys, Leu,
+            #                       Met, Asn, Pro, Gln, Arg, Ser, Thr, Val, Trp, Tyr
+            aa_counts = protein_row.iloc[4:24].values.astype(float)
+            aa_composition.append(aa_counts)
+        aa_composition = np.array(aa_composition)  # Shape: (n_proteins, 21)
+
+        # Matrix multiplication: (n_proteins x 21) @ (21 x n_basis) = (n_proteins x n_basis)
+        # Note: res_coeffs has shape (21, n_basis) - first row is H2O, next 20 are amino acids
+        # R code: t(t(resspecies) %*% t(protein)) means: (n_basis x n_residues) @ (n_residues x n_proteins) = (n_basis x n_proteins)
+        # Then transpose to get (n_proteins x n_basis)
+        # In Python: (n_proteins x n_residues) @ (n_residues x n_basis) = (n_proteins x n_basis)
+        protein_coeffs = aa_composition @ res_coeffs  # Shape: (n_proteins, n_basis)
+
+        # Delete residue species from species list now that we have the coefficients
+        from .species import species as species_func
+        species_func(ires.tolist(), delete=True, messages=False)
+
+        if original_species is not None:
+            # Restore original species (but we've already calculated, so just update species_df)
+            pass
+
+        # Create DataFrame for proteins with basis species coefficients
+        species_data = {}
+
+        # Add basis species columns
+        for j, basis_sp in enumerate(basis_cols):
+            species_data[basis_sp] = protein_coeffs[:, j]
+
+        # Add metadata columns
+        protein_names = []
+        protein_ispecies = []
+
+        for iprot in iprotein:
+            prot_row = thermo_obj.protein.iloc[iprot]
+            # Escape underscores for LaTeX compatibility in diagram labels
+            protein_name = f"{prot_row['protein']}_{prot_row['organism']}"
+            # Replace underscores with escaped version for matplotlib/LaTeX
+            protein_name_escaped = protein_name.replace('_', r'\_')
+            protein_names.append(protein_name_escaped)
+            protein_ispecies.append(-(iprot + 1))  # Negative index
+
+        species_data['ispecies'] = protein_ispecies
+        species_data['logact'] = loga_protein_arr[:len(iprotein)]
+        species_data['state'] = ['aq'] * len(iprotein)
+        species_data['name'] = protein_names
+
+        species_df = pd.DataFrame(species_data)
+
+    # Process temperature and pressure for output
+    T_out = args['T']
+    P_out = args['P']
+    vars_list = args['vars']
+    vals_dict = {}
+
+    # Convert variable names and values for output
+    # Important: Keep vars_list with actual basis species names (H+, e-) for internal use
+    # but create display versions in vals_dict with user-friendly names (pH, pe, Eh)
+    vars_list_display = vars_list.copy()
+    for i, var in enumerate(vars_list):
+        # Handle pH, pe, Eh conversions for output
+        if var == 'H+' and 'pH' in args_orig:
+            vars_list_display[i] = 'pH'
+            vals_dict['pH'] = [-val for val in args['vals'][i]]
+        elif var == 'e-' and 'pe' in args_orig:
+            vars_list_display[i] = 'pe'
+            vals_dict['pe'] = [-val for val in args['vals'][i]]
+        elif var == 'e-' and 'Eh' in args_orig:
+            vars_list_display[i] = 'Eh'
+            # Convert from log(a_e-) back to Eh using temperature-dependent formula
+            # log(a_e-) = -pe, so pe = -log(a_e-)
+            # Eh = pe * (ln(10) * R * T) / F = -log(a_e-) * T / 5039.76
+            T_kelvin = args['T'] if isinstance(args['T'], (int, float)) else args['T'][0] if hasattr(args['T'], '__len__') else 298.15
+            conversion_factor = T_kelvin / 5039.76  # volts per pe unit
+            vals_dict['Eh'] = [-val * conversion_factor for val in args['vals'][i]]
+        else:
+            vals_dict[var] = args['vals'][i]
+
+    # Keep vars_list as-is (with basis species names) for internal calculations
+    # vars_list_display will be used for output only
+
+    # Check if T or P are variables
+    if 'T' in vars_list:
+        T_out = []  # Variable T
+        # Convert back to Celsius for output
+        T_vals = vals_dict['T']
+        vals_dict['T'] = [T - 273.15 for T in T_vals]
+    else:
+        # Convert to Kelvin for output (matching R)
+        T_out = args['T']
+
+    if 'P' in vars_list:
+        P_out = []  # Variable P
+    else:
+        P_out = args['P']
+
+    # Build output dictionary matching R CHNOSZ structure
+    result = {
+        'fun': 'affinity',
+        'args': {
+            **args_orig,
+            'property': property_name,
+            'exceed_Ttr': kwargs.get('exceed_Ttr', False),
+            'exceed_rhomin': kwargs.get('exceed_rhomin', False),
+            'return_buffer': kwargs.get('return_buffer', False),
+            'balance': kwargs.get('balance', 'PBB')
+        },
+        'sout': energy_sout,
+        'property': property_name,
+        'basis': basis_df,
+        'species': species_df,
+        'T': T_out,
+        'P': P_out,
+        'vars': vars_list_display,  # Use display version with 'Eh', 'pH', 'pe' for output
+        'vals': vals_dict,
+        'values': affinity_values
+    }
+
+    return result
+
+
+def energy_args(args: Dict[str, Any], messages: bool = True, basis_df: Optional[pd.DataFrame] = None) -> Dict[str, Any]:
+    """
+    Process arguments for energy calculations.
+
+    Converts variable arguments into consistent format for multi-dimensional
+    calculations, handling T, P, IS and basis species variables.
+
+    Parameters
+    ----------
+    args : dict
+        Raw arguments from affinity() call
+
+    Returns
+    -------
+    dict
+        Processed arguments with consistent variable structure
+    """
+
+    thermo_obj = thermo()
+    if basis_df is None:
+        basis_df = get_basis()
+
+    # Default values
+    T = 298.15
+    P = "Psat"
+    IS = 0
+    T_is_var = P_is_var = IS_is_var = False
+
+    # Process T, P, IS arguments
+    if 'T' in args:
+        T = args['T']
+        if hasattr(T, '__len__') and len(T) > 1:
+            T_is_var = True
+        # Convert to Kelvin if needed (assuming Celsius input)
+        if T_is_var:
+            if isinstance(T, (list, tuple)):
+                # Handle [T1, T2, npoints] format or [T1, T2] (default to 256 points)
+                if len(T) == 3:
+                    T = np.linspace(T[0] + 273.15, T[1] + 273.15, int(T[2]))
+                elif len(T) == 2:
+                    # Default resolution: 256 points (R CHNOSZ standard)
+                    T = np.linspace(T[0] + 273.15, T[1] + 273.15, 256)
+                else:
+                    T = np.array(T) + 273.15
+            else:
+                T = T + 273.15
+        else:
+            T = T + 273.15
+
+    if 'P' in args:
+        P = args['P']
+        if hasattr(P, '__len__') and len(P) > 1:
+            P_is_var = True
+        if P_is_var and P != "Psat":
+            if isinstance(P, (list, tuple)):
+                if len(P) == 3:
+                    P = np.linspace(P[0], P[1], int(P[2]))
+                elif len(P) == 2:
+                    # Default resolution: 256 points (R CHNOSZ standard)
+                    P = np.linspace(P[0], P[1], 256)
+
+    if 'IS' in args:
+        IS = args['IS']
+        if hasattr(IS, '__len__') and len(IS) > 1:
+            IS_is_var = True
+            if isinstance(IS, (list, tuple)):
+                if len(IS) == 3:
+                    IS = np.linspace(IS[0], IS[1], int(IS[2]))
+                elif len(IS) == 2:
+                    # Default resolution: 256 points (R CHNOSZ standard)
+                    IS = np.linspace(IS[0], IS[1], 256)
+
+    # Print status messages
+    if messages:
+        if not T_is_var:
+            T_celsius = T - 273.15 if isinstance(T, (int, float)) else T[0] - 273.15
+            print(f'affinity: temperature is {T_celsius:.0f} ºC')
+
+        if not P_is_var:
+            if P == "Psat":
+                print("affinity: pressure is Psat")
+            else:
+                print(f'affinity: pressure is {P} bar')
+
+        if not IS_is_var and IS != 0:
+            print(f'affinity: ionic strength is {IS}')
+
+    # Default property
+    what = 'A'
+    if 'what' in args:
+        what = args['what']
+
+    # Process variable arguments
+    # Preserve the order in which variables were specified (R CHNOSZ compatibility)
+    vars_list = []
+    vals_list = []
+    lims_list = []
+
+    # Track which T/P/IS are variables and process them in the order they appear in args
+    tps_vars = {'T': (T_is_var, T), 'P': (P_is_var, P), 'IS': (IS_is_var, IS)}
+
+    # Add T, P, IS in the order they appear in args (preserves user's specification order)
+    for arg_name in args.keys():
+        if arg_name in ['T', 'P', 'IS'] and tps_vars[arg_name][0]:
+            var_name = arg_name
+            var_value = tps_vars[arg_name][1]
+
+            vars_list.append(var_name)
+            vals_list.append(var_value)
+
+            if isinstance(args[arg_name], (list, tuple)):
+                if len(args[arg_name]) == 3:
+                    # User specified [min, max, npoints]
+                    if arg_name == 'T':
+                        lims_list.append([args[arg_name][0] + 273.15, args[arg_name][1] + 273.15, args[arg_name][2]])
+                    else:
+                        lims_list.append([args[arg_name][0], args[arg_name][1], args[arg_name][2]])
+                elif len(args[arg_name]) == 2:
+                    # User specified [min, max], default to 256 points
+                    if arg_name == 'T':
+                        lims_list.append([args[arg_name][0] + 273.15, args[arg_name][1] + 273.15, 256])
+                    else:
+                        lims_list.append([args[arg_name][0], args[arg_name][1], 256])
+                else:
+                    # User provided explicit array of values
+                    lims_list.append([var_value.min(), var_value.max(), len(var_value)])
+            else:
+                lims_list.append([var_value.min(), var_value.max(), len(var_value)])
+
+    # Process basis species variables
+    basis_names = basis_df.index.tolist()
+
+    for arg_name, arg_value in args.items():
+        # Skip T, P, IS, and non-basis arguments
+        if arg_name in ['T', 'P', 'IS', 'what', 'property', 'exceed_Ttr', 'exceed_rhomin', 'return_buffer', 'balance']:
+            continue
+
+        # Handle pH -> H+, pe -> e-, Eh -> e-
+        var_name = arg_name
+        var_values = arg_value
+
+        if arg_name == 'pH':
+            var_name = 'H+'
+            if hasattr(var_values, '__len__'):
+                if len(var_values) >= 3:
+                    # [pH1, pH2, npoints] -> [-pH1, -pH2, npoints] for H+ (logact)
+                    # pH and log(a_H+) are related by: pH = -log(a_H+), so log(a_H+) = -pH
+                    var_values = np.linspace(-var_values[0], -var_values[1], int(var_values[2]))
+                elif len(var_values) >= 2:
+                    var_values = [-v for v in var_values]
+                else:
+                    # Single value in a list [pH]
+                    var_values = np.array([-var_values[0]])
+            else:
+                # Scalar value
+                var_values = np.array([-var_values])
+        elif arg_name == 'pe':
+            var_name = 'e-'
+            if hasattr(var_values, '__len__'):
+                if len(var_values) >= 3:
+                    # pe = -log(a_e-), so log(a_e-) = -pe
+                    # For pe range [pe1, pe2], log(a_e-) range is [-pe1, -pe2]
+                    var_values = np.linspace(-var_values[0], -var_values[1], int(var_values[2]))
+                elif len(var_values) >= 2:
+                    var_values = [-v for v in var_values]
+                else:
+                    # Single value in a list [pe]
+                    var_values = np.array([-var_values[0]])
+            else:
+                # Scalar value
+                var_values = np.array([-var_values])
+        elif arg_name == 'Eh':
+            var_name = 'e-'
+            # Convert Eh (volts) to log(a_e-) using temperature-dependent formula
+            # pe = Eh * F / (ln(10) * R * T) where pe = -log(a_e-)
+            # Therefore: log(a_e-) = -pe = -Eh * F / (ln(10) * R * T)
+            # where R = 0.00831470 kJ/(mol·K), F = 96.4935 kJ/(V·mol), T in Kelvin
+            # This gives: log(a_e-) = -Eh * 96.4935 / (2.303 * 0.00831470 * T)
+            #           = -Eh * 96.4935 / (0.019145 * T)
+            #           = -Eh * 5039.76 / T
+
+            # Get temperature for conversion (default to 25°C if not specified)
+            T_kelvin = T if isinstance(T, (int, float)) else T[0] if hasattr(T, '__len__') else 298.15
+            conversion_factor = 5039.76 / T_kelvin  # pe per volt (need to negate for log(a_e-))
+
+            if hasattr(var_values, '__len__') and len(var_values) >= 2:
+                if len(var_values) == 3:
+                    # [Eh1, Eh2, npoints] format
+                    # Convert to log(a_e-) = -pe = -Eh * conversion_factor
+                    logact_start = -var_values[0] * conversion_factor
+                    logact_end = -var_values[1] * conversion_factor
+                    var_values = np.linspace(logact_start, logact_end, int(var_values[2]))
+                elif len(var_values) == 2:
+                    # [Eh1, Eh2] format - default to 256 points like R
+                    logact_start = -var_values[0] * conversion_factor
+                    logact_end = -var_values[1] * conversion_factor
+                    var_values = np.linspace(logact_start, logact_end, 256)
+                else:
+                    # List of explicit Eh values
+                    var_values = [-v * conversion_factor for v in var_values]
+            else:
+                # Single value
+                var_values = -var_values * conversion_factor
+
+        # Check if this is a basis species
+        if var_name in basis_names:
+            vars_list.append(var_name)
+
+            # Process values
+            if isinstance(var_values, (list, tuple)):
+                if len(var_values) == 3:
+                    # [min, max, npoints] format
+                    vals_array = np.linspace(var_values[0], var_values[1], int(var_values[2]))
+                    vals_list.append(vals_array)
+                    lims_list.append(var_values)
+
+                    # Print variable info
+                    if messages:
+                        n_vals = int(var_values[2])
+                        print(f'affinity: variable {len(vars_list)} is log10(a_{var_name}) at {n_vals} values from {var_values[0]} to {var_values[1]}')
+
+                elif len(var_values) == 2:
+                    # [min, max] format - default to 256 points (R CHNOSZ behavior)
+                    vals_array = np.linspace(var_values[0], var_values[1], 256)
+                    vals_list.append(vals_array)
+                    lims_list.append([var_values[0], var_values[1], 256])
+
+                    # Print variable info
+                    if messages:
+                        print(f'affinity: variable {len(vars_list)} is log10(a_{var_name}) at 256 values from {var_values[0]} to {var_values[1]}')
+
+                else:
+                    # Explicit array of values
+                    vals_list.append(np.array(var_values))
+                    lims_list.append([min(var_values), max(var_values), len(var_values)])
+            else:
+                # Single value
+                if not hasattr(var_values, '__len__'):
+                    var_values = [var_values]
+                vals_list.append(np.array(var_values))
+                lims_list.append([var_values[0], var_values[-1], len(var_values)])
+        else:
+            # Not a recognized basis species or variable
+            raise AffinityError(f"{arg_name} is not one of T, P, or IS, and does not match any basis species")
+
+    return {
+        'what': what,
+        'vars': vars_list,
+        'vals': vals_list,
+        'lims': lims_list,
+        'T': T,
+        'P': P,
+        'IS': IS
+    }
+
+
+def energy(what: str, vars: List[str], vals: List, lims: List,
+           T: Union[float, np.ndarray] = 298.15,
+           P: Union[float, str] = "Psat",
+           IS: float = 0,
+           sout: Optional[Dict] = None,
+           exceed_Ttr: bool = True,
+           exceed_rhomin: bool = False,
+           basis_df: Optional[pd.DataFrame] = None,
+           species_df: Optional[pd.DataFrame] = None,
+           messages: bool = True) -> Dict[str, Any]:
+    """
+    Calculate energy properties over multiple dimensions.
+
+    This is the core calculation function that handles multi-dimensional
+    property calculations for basis and formed species.
+
+    Parameters
+    ----------
+    what : str
+        Property to calculate ("A", "logK", "G", "H", etc.)
+    vars : list of str
+        Variable names
+    vals : list of arrays
+        Variable values
+    lims : list of limits
+        Variable limits [min, max, npoints]
+    T : float or array
+        Temperature(s) in Kelvin
+    P : float or str
+        Pressure(s) in bar or "Psat"
+    IS : float
+        Ionic strength
+    sout : dict, optional
+        Pre-calculated subcrt results
+    exceed_Ttr : bool
+        Allow extrapolation beyond transitions
+    exceed_rhomin : bool
+        Allow below minimum density
+
+    Returns
+    -------
+    dict
+        Dictionary with 'sout' (subcrt results) and 'a' (property values)
+    """
+
+    # Get system data
+    thermo_obj = thermo()
+    if basis_df is None:
+        basis_df = get_basis()
+    if species_df is None:
+        species_df = get_species()
+
+    n_basis = len(basis_df)
+    n_species = len(species_df)
+
+    # Determine array dimensions
+    if len(vars) == 0:
+        mydim = [1]
+    else:
+        mydim = [lim[2] for lim in lims]
+
+    # Prepare subcrt call
+    if what in ['G', 'H', 'S', 'Cp', 'V', 'E', 'kT', 'logK'] or what == 'A':
+        # Need to call subcrt for thermodynamic properties
+
+        # Prepare species list (basis + formed species)
+        all_species = basis_df['ispecies'].tolist() + species_df['ispecies'].tolist()
+
+        # Prepare T, P, IS for subcrt (convert T from Kelvin to Celsius)
+        subcrt_T = T - 273.15 if isinstance(T, (int, float)) else T - 273.15
+        subcrt_P = P
+        subcrt_IS = IS
+
+        # Handle variable T, P, IS
+        if 'T' in vars:
+            # T in vals is already in Kelvin, convert to Celsius for subcrt
+            T_vals = vals[vars.index('T')]
+            subcrt_T = T_vals - 273.15 if isinstance(T_vals, (int, float)) else T_vals - 273.15
+        if 'P' in vars:
+            subcrt_P = vals[vars.index('P')]
+        if 'IS' in vars:
+            subcrt_IS = vals[vars.index('IS')]
+
+        # Call subcrt
+        # Skip sout calculation for affinity (what=='A') since the affinity block
+        # has its own optimized batch subcrt call
+        if sout is None and what != 'A':
+            try:
+                # Determine grid parameter for subcrt
+                grid_param = None
+                if len(vars) > 1:
+                    # Multi-variable case - use appropriate grid
+                    subcrt_vars = [v for v in vars if v in ['T', 'P', 'IS']]
+                    if len(subcrt_vars) >= 2:
+                        grid_param = subcrt_vars[0]  # Use first subcrt variable
+
+                sout_result = subcrt(
+                    species=all_species,
+                    T=subcrt_T,
+                    P=subcrt_P,
+                    IS=subcrt_IS,
+                    property='logK',
+                    grid=grid_param,
+                    exceed_Ttr=exceed_Ttr,
+                    exceed_rhomin=exceed_rhomin,
+                    messages=messages,
+                    show=False
+                )
+                sout_data = sout_result.out
+
+            except Exception as e:
+                warnings.warn(f"subcrt calculation failed: {e}")
+                # Create dummy sout data
+                n_conditions = np.prod(mydim) if len(mydim) > 0 else 1
+                sout_data = pd.DataFrame({
+                    'T': np.full(n_conditions, T if isinstance(T, (int, float)) else T[0]) - 273.15,
+                    'P': np.full(n_conditions, 1.0 if P == "Psat" else (P if isinstance(P, (int, float)) else P[0])),
+                    'logK': np.full(n_conditions, np.nan)
+                })
+        else:
+            sout_data = sout
+
+    # Calculate the requested property
+    if what == 'A':
+        # Calculate affinities A/2.303RT following R CHNOSZ logic exactly
+        affinity_values = {}
+
+        # Get basis and species information
+        basis_names = basis_df.index.tolist()
+        n_conditions = np.prod(mydim) if len(mydim) > 0 else 1
+
+        # Create activity arrays for each basis species using multi-dimensional grid expansion
+        # This implements R's expand.grid functionality using numpy.meshgrid
+        logact_basis_arrays = {}
+
+        if len(vars) > 1:
+            # Multi-dimensional case: create meshgrid for all variables
+            var_arrays = []
+            var_names_ordered = []
+
+            # Collect variable arrays in order
+            for var_name in vars:
+                if var_name in basis_names:
+                    var_idx = vars.index(var_name)
+                    var_arrays.append(np.array(vals[var_idx]))
+                    var_names_ordered.append(var_name)
+
+            # Create meshgrid for basis species variables
+            if var_arrays:
+                # meshgrid creates N-D arrays where each variable varies along its own axis
+                # indexing='ij' gives matrix indexing (first index varies down rows)
+                meshgrids = np.meshgrid(*var_arrays, indexing='ij')
+
+                # Map meshgrid results back to basis species
+                for i, var_name in enumerate(var_names_ordered):
+                    logact_basis_arrays[var_name] = meshgrids[i]
+
+        # Handle all basis species (variables and fixed)
+        for j, basis_name in enumerate(basis_names):
+            if basis_name in vars and basis_name not in logact_basis_arrays:
+                # Single variable case
+                var_idx = vars.index(basis_name)
+                logact_basis_arrays[basis_name] = np.array(vals[var_idx])
+            elif basis_name not in logact_basis_arrays:
+                # Fixed activity from basis definition - broadcast to full grid
+                basis_logact = basis_df.iloc[j]['logact']
+                try:
+                    logact_val = float(basis_logact)
+                except (ValueError, TypeError):
+                    logact_val = 0.0
+
+                if len(mydim) > 1:
+                    # Multi-dimensional: broadcast scalar to full grid shape
+                    logact_basis_arrays[basis_name] = np.full(mydim, logact_val)
+                else:
+                    # Single dimension
+                    logact_basis_arrays[basis_name] = np.full(n_conditions, logact_val)
+
+        # For affinities, we need logK of balanced formation reactions
+        # Optimize by calling subcrt once for all basis + non-basis species
+        # to get logK of formation from elements, then calculate formation from basis
+        formation_logK = {}
+
+        # Convert T from Kelvin back to Celsius for subcrt (subcrt expects Celsius)
+        T_celsius = T - 273.15
+
+        # Get all unique species (basis + formed species) using ispecies indices
+        # to avoid redundant info_character lookups
+        basis_ispecies_list = basis_df['ispecies'].tolist()
+        species_ispecies_list = species_df['ispecies'].tolist()
+        all_species_indices = list(dict.fromkeys(basis_ispecies_list + species_ispecies_list))
+
+        # Create mapping from names to ispecies indices
+        # Note: multiple names (e.g., "Fe" and "iron") can map to the same ispecies
+        basis_names_list = basis_names  # Already defined at line 548
+        species_names_list = species_df['name'].tolist()
+
+        # Build a name->ispecies mapping
+        name_to_ispecies = {}
+        for name, ispec in zip(basis_names_list, basis_ispecies_list):
+            name_to_ispecies[name] = ispec
+        for name, ispec in zip(species_names_list, species_ispecies_list):
+            name_to_ispecies[name] = ispec
+
+        # Build ispecies->result_index mapping for batch result access
+        ispecies_to_result_idx = {ispec: idx for idx, ispec in enumerate(all_species_indices)}
+
+        # All unique names (may have duplicates that refer to same ispecies)
+        all_species_names = list(dict.fromkeys(basis_names_list + species_names_list))
+
+        # Single batch subcrt call to get logK of formation from elements for all species
+        # Use ispecies indices to avoid redundant lookups
+        try:
+            # Determine grid parameter for subcrt when we have multiple T/P variables
+            grid_param = None
+            if len(vars) >= 2:
+                # Check if we have T and/or P as variables
+                if 'T' in vars and 'P' in vars:
+                    # Both T and P vary - use T as grid variable (R CHNOSZ convention)
+                    grid_param = 'T'
+                elif 'T' in vars:
+                    grid_param = 'T'
+                elif 'P' in vars:
+                    grid_param = 'P'
+
+            batch_result = subcrt(all_species_indices, property="logK", T=T_celsius, P=P, grid=grid_param, messages=messages, show=False)
+
+            # Extract logK values from batch result
+            # batch_result.out is a dict with 'species_data' list
+            # When T/P are variable, each species_data DataFrame has multiple rows
+            species_logK_from_elements = {}
+            if isinstance(batch_result.out, dict) and 'species_data' in batch_result.out:
+                # Map each name to its data using the ispecies->result_idx mapping
+                for sp_name in all_species_names:
+                    ispec = name_to_ispecies[sp_name]
+                    result_idx = ispecies_to_result_idx[ispec]
+                    sp_data = batch_result.out['species_data'][result_idx]
+
+                    if 'logK' in sp_data.columns:
+                        # Get all logK values (may be array if T/P variable)
+                        logK_vals = sp_data['logK'].values
+                        # Handle NaN values by keeping them as nan (they will propagate to affinity)
+                        # DO NOT replace nan with 0.0 as this causes incorrect affinity calculations
+                        # logK_vals = np.where(np.isnan(logK_vals), 0.0, logK_vals)
+
+                        # Reshape if we have a 2-D grid
+                        if len(mydim) > 1 and len(logK_vals) == np.prod(mydim):
+                            # Reshape flattened array to match grid dimensions
+                            # mydim is [nT, nP] or similar, and grid='T' gives row-major order
+                            logK_vals = logK_vals.reshape(mydim)
+
+                        species_logK_from_elements[sp_name] = logK_vals
+                    else:
+                        # No logK column - use zeros
+                        n_rows = len(sp_data)
+                        if len(mydim) > 1 and n_rows == np.prod(mydim):
+                            species_logK_from_elements[sp_name] = np.zeros(mydim)
+                        else:
+                            species_logK_from_elements[sp_name] = np.zeros(n_rows)
+            elif isinstance(batch_result.out, pd.DataFrame):
+                # Single species case - result.out is a DataFrame directly
+                sp_data = batch_result.out
+                sp_name = all_species_names[0]
+                if 'logK' in sp_data.columns:
+                    logK_vals = sp_data['logK'].values
+                    # Handle NaN values by keeping them as nan (they will propagate to affinity)
+                    # DO NOT replace nan with 0.0 as this causes incorrect affinity calculations
+                    # logK_vals = np.where(np.isnan(logK_vals), 0.0, logK_vals)
+
+                    # Reshape if we have a 2-D grid
+                    if len(mydim) > 1 and len(logK_vals) == np.prod(mydim):
+                        logK_vals = logK_vals.reshape(mydim)
+
+                    species_logK_from_elements[sp_name] = logK_vals
+                else:
+                    n_rows = len(sp_data)
+                    if len(mydim) > 1 and n_rows == np.prod(mydim):
+                        species_logK_from_elements[sp_name] = np.zeros(mydim)
+                    else:
+                        species_logK_from_elements[sp_name] = np.zeros(n_rows)
+            else:
+                # Fallback if structure is different
+                for sp_name in all_species_names:
+                    if len(mydim) > 1:
+                        species_logK_from_elements[sp_name] = np.zeros(mydim)
+                    else:
+                        species_logK_from_elements[sp_name] = np.array([0.0])
+
+            # Now calculate formation logK from basis species for each formed species
+            for i in range(n_species):
+                species_idx = species_df.iloc[i]['ispecies']
+                species_name = species_df.iloc[i]['name']
+
+                # Check if this species is also a basis species
+                is_basis_species = species_idx in basis_df['ispecies'].values
+
+                if is_basis_species:
+                    # Species is in the basis - formation from basis is trivial
+                    formation_logK[species_idx] = 0.0
+                else:
+                    # Calculate formation logK from basis using stoichiometry
+                    # The species() coefficients represent: species = basis_products - basis_reactants
+                    # For logK from elements: logK_formation = logK_species - sum(coeff_i * logK_basis_i)
+                    logK_formation_val = species_logK_from_elements.get(species_name, 0.0)
+
+                    # Subtract contribution from basis species
+                    for basis_name in basis_names_list:
+                        coeff = species_df.iloc[i][basis_name]
+                        basis_logK = species_logK_from_elements.get(basis_name, 0.0)
+                        logK_formation_val -= coeff * basis_logK
+
+                    formation_logK[species_idx] = logK_formation_val
+
+        except Exception as e:
+            warnings.warn(f"Batch subcrt call failed, falling back to individual calls: {e}")
+            # Fallback to old method if batch call fails
+            for i in range(n_species):
+                species_idx = species_df.iloc[i]['ispecies']
+                is_basis_species = species_idx in basis_df['ispecies'].values
+
+                if is_basis_species:
+                    formation_logK[species_idx] = 0.0
+                else:
+                    try:
+                        species_name = species_df.iloc[i]['name']
+                        formation_result = subcrt([species_name], [1], T=T_celsius, P=P, messages=messages, show=False)
+
+                        # Handle both single DataFrame and dict of DataFrames
+                        if hasattr(formation_result, 'out'):
+                            if isinstance(formation_result.out, dict) and 'species_data' in formation_result.out:
+                                # Multiple conditions (T/P arrays) - result.out is a dict
+                                sp_data = formation_result.out['species_data'][0]
+                                if 'logK' in sp_data.columns:
+                                    logK_vals = sp_data['logK'].values
+                                    # Keep nan values as is
+                                    # logK_vals = np.where(np.isnan(logK_vals), 0.0, logK_vals)
+                                    logK_val = logK_vals
+                                else:
+                                    logK_val = np.zeros(len(sp_data))
+                            elif isinstance(formation_result.out, pd.DataFrame):
+                                # Single condition - result.out is a DataFrame
+                                if 'logK' in formation_result.out.columns:
+                                    logK_val = formation_result.out['logK'].values
+                                    # Keep nan values as is
+                                    # logK_val = np.where(np.isnan(logK_val), 0.0, logK_val)
+                                else:
+                                    logK_val = 0.0
+                            else:
+                                logK_val = 0.0
+                        else:
+                            logK_val = 0.0
+                        formation_logK[species_idx] = logK_val
+                    except Exception as e2:
+                        warnings.warn(f"Could not get formation logK for species {species_idx}: {e2}")
+                        formation_logK[species_idx] = 0.0
+
+        # Calculate affinities for each formed species
+        for i in range(n_species):
+            species_idx = species_df.iloc[i]['ispecies']
+
+            # Get the formation reaction logK (already balanced)
+            logK_formation = formation_logK[species_idx]
+
+            # Get formation reaction stoichiometry from species DataFrame
+            # These are the stoichiometric coefficients from the balanced reaction
+            formation_coeffs = {}
+            for basis_name in basis_names:
+                formation_coeffs[basis_name] = species_df.iloc[i][basis_name]
+
+            # Calculate logQ using R CHNOSZ logic:
+            # logQ = +1 * logact_species + sum(-coeff_i * logact_basis_i)
+            # Species gets +1 coefficient (product), all basis species get negative coefficients (reactants)
+
+            # Species activity (always +1 coefficient on product side)
+            species_logact = species_df.iloc[i]['logact']
+            try:
+                species_logact_val = float(species_logact)
+            except (ValueError, TypeError):
+                species_logact_val = 0.0
+
+            # Start with species contribution: +1 * logact_species
+            # Create array with proper dimensions to match the grid
+            if len(mydim) > 1:
+                logQ_arrays = np.full(mydim, species_logact_val)
+            else:
+                logQ_arrays = np.full(n_conditions, species_logact_val)
+
+            # Add contributions from all basis species: -coeff_i * logact_basis_i
+            for basis_name in formation_coeffs:
+                coeff = formation_coeffs[basis_name]
+                logact_array = logact_basis_arrays[basis_name]
+                # DEBUG
+                if False and species_idx == 763:  # ethanol
+                    print(f"  Basis {basis_name}: coeff={coeff}, logact_array[0]={logact_array[0] if hasattr(logact_array, '__getitem__') else logact_array}")
+                # All basis species contributions are negative (reactant side)
+                logQ_arrays += (-coeff) * logact_array
+
+            # Calculate affinity: A/2.303RT = logK - logQ
+            # Handle shape broadcasting when logK varies along fewer dimensions than logQ
+            # This happens when we have basis variables (e.g., H2S) and subcrt variables (e.g., T)
+            # logK only varies with subcrt variables (T, P, IS) but logQ varies with all variables
+            if isinstance(logK_formation, np.ndarray) and isinstance(logQ_arrays, np.ndarray):
+                if logK_formation.shape != logQ_arrays.shape:
+                    # Need to broadcast logK to match logQ dimensions
+                    if len(mydim) > 1 and logK_formation.ndim == 1:
+                        # logK is 1-D but should be broadcast to 2-D
+                        # Determine which dimension logK varies along
+                        # Check if logK length matches first dimension of mydim (typically T)
+                        if len(logK_formation) == mydim[0]:
+                            # logK varies along first dimension, broadcast to second
+                            logK_formation = np.broadcast_to(logK_formation[:, np.newaxis], mydim)
+                        elif len(logK_formation) == mydim[1]:
+                            # logK varies along second dimension, broadcast to first
+                            logK_formation = np.broadcast_to(logK_formation[np.newaxis, :], mydim)
+                        elif len(logK_formation) == np.prod(mydim):
+                            # logK is flattened, reshape it
+                            logK_formation = logK_formation.reshape(mydim)
+
+            affinity_array = logK_formation - logQ_arrays
+
+            # DEBUG: Check first value
+            if False:  # Set to True for debugging
+                if hasattr(affinity_array, '__getitem__'):
+                    print(f"\nDEBUG affinity for species {species_idx}:")
+                    print(f"  logK_formation[0] = {logK_formation[0] if hasattr(logK_formation, '__getitem__') else logK_formation}")
+                    print(f"  logQ_arrays[0] = {logQ_arrays[0] if hasattr(logQ_arrays, '__getitem__') else logQ_arrays}")
+                    print(f"  affinity_array[0] = {affinity_array[0]}")
+
+            # Store result with proper dimensions
+            # Keep array structure if we have multiple variables, even if n_conditions == 1
+            # This ensures diagram() can detect the correct dimensionality (matching R behavior)
+            if n_conditions == 1 and len(mydim) <= 1:
+                # True scalar case: no variables or single variable with 1 point
+                affinity_values[species_idx] = affinity_array.item() if hasattr(affinity_array, 'item') else affinity_array
+            else:
+                # Multi-dimensional case: preserve array structure
+                # Array already has correct shape from meshgrid
+                affinity_values[species_idx] = affinity_array
+
+        return {
+            'sout': sout_data,
+            'a': affinity_values
+        }
+
+    elif what == 'logK':
+        # Extract logK values from subcrt results
+        logK_values = {}
+
+        for i in range(n_species):
+            species_idx = species_df.iloc[i]['ispecies']
+
+            if hasattr(sout_data, 'iloc') and len(sout_data) > n_basis + i:
+                logK_val = sout_data.iloc[n_basis + i]['logK'] if 'logK' in sout_data.columns else np.nan
+            else:
+                logK_val = np.nan
+
+            # Expand to proper dimensions
+            if np.prod(mydim) > 1:
+                logK_values[species_idx] = np.full(mydim, logK_val)
+            else:
+                logK_values[species_idx] = logK_val
+
+        return {
+            'sout': sout_data,
+            'a': logK_values
+        }
+
+    else:
+        # Other thermodynamic properties
+        prop_values = {}
+
+        for i in range(n_species):
+            species_idx = species_df.iloc[i]['ispecies']
+
+            if hasattr(sout_data, 'iloc') and len(sout_data) > n_basis + i:
+                prop_val = sout_data.iloc[n_basis + i][what] if what in sout_data.columns else np.nan
+            else:
+                prop_val = np.nan
+
+            # Expand to proper dimensions
+            if np.prod(mydim) > 1:
+                prop_values[species_idx] = np.full(mydim, prop_val)
+            else:
+                prop_values[species_idx] = prop_val
+
+        return {
+            'sout': sout_data,
+            'a': prop_values
+        }
+
+
+# Export main functions
+__all__ = [
+    'affinity', 'energy_args', 'energy', 'AffinityError'
+]