pychnosz 1.1.1__cp310-cp310-win_amd64.whl

pychnosz/core/diagram.py

@@ -0,0 +1,3336 @@
+"""
+Diagram module for plotting chemical activity and predominance diagrams.
+
+This module provides Python equivalents of the R functions in diagram.R:
+- diagram(): Plot equilibrium chemical activity and predominance diagrams
+- Supporting utilities for 1D line plots and 2D predominance diagrams
+
+Author: CHNOSZ Python port
+"""
+
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from typing import Union, List, Optional, Dict, Any, Tuple
+import warnings
+import copy
+from ..utils.expression import _format_species_latex
+
+
+def copy_plot(diagram_result: Dict[str, Any]) -> Dict[str, Any]:
+    """
+    Create a deep copy of a diagram result, allowing independent modification.
+
+    This function addresses a fundamental limitation in Python plotting libraries:
+    matplotlib figure and axes objects are mutable, so passing them between
+    functions causes modifications to affect all references. This function
+    creates a true deep copy that can be modified independently.
+
+    Parameters
+    ----------
+    diagram_result : dict
+        Result dictionary from diagram(), which may contain 'fig' and 'ax' keys
+
+    Returns
+    -------
+    dict
+        A deep copy of the diagram result with independent figure and axes objects
+
+    Examples
+    --------
+    Manual copying workflow (advanced usage - normally use add_to parameter instead):
+
+    >>> import pychnosz
+    >>> # Create base plot (Plot A)
+    >>> basis(['SiO2', 'Ca+2', 'Mg+2', 'CO2', 'H2O', 'O2', 'H+'])
+    >>> species(['quartz', 'talc', 'chrysotile', 'forsterite'])
+    >>> a = affinity(**{'Mg+2': [4, 10, 500], 'Ca+2': [5, 15, 500]})
+    >>> plot_a = diagram(a, fill='terrain')
+    >>>
+    >>> # Manual approach: create copies first, then modify the axes directly
+    >>> plot_a1 = copy_plot(plot_a)  # For modification 1
+    >>> plot_a2 = copy_plot(plot_a)  # For modification 2
+    >>> # ... then modify plot_a1['ax'] and plot_a2['ax'] directly
+    >>>
+    >>> # Recommended approach: use add_to parameter instead
+    >>> # This automatically handles copying internally
+    >>> basis('CO2', -1)
+    >>> species(['calcite', 'dolomide'])
+    >>> a2 = affinity(**{'Mg+2': [4, 10, 500], 'Ca+2': [5, 15, 500]})
+    >>> plot_a1 = diagram(a2, type='saturation', add_to=plot_a, col='blue')
+    >>> plot_a2 = diagram(a2, type='saturation', add_to=plot_a, col='red')
+    >>> # Now you have three independent plots: plot_a, plot_a1, plot_a2
+
+    Notes
+    -----
+    - This function uses copy.deepcopy() which works well for matplotlib figures
+    - For very large plots, copying may be memory-intensive
+    - Interactive plots (plotly) may not copy perfectly - test before relying on this
+    - The copied plot is fully independent and can be saved, displayed, or modified
+      without affecting the original
+
+    Limitations
+    -----------
+    Python's matplotlib (unlike R's base graphics) uses mutable objects for plots.
+    Without explicit copying, all references point to the same plot. This is a
+    known limitation of matplotlib that this function works around.
+
+    See Also
+    --------
+    diagram : Create plots that can be copied with this function
+    """
+    return copy.deepcopy(diagram_result)
+
+
+def diagram(eout: Dict[str, Any],
+            type: str = "auto",
+            alpha: bool = False,
+            balance: Optional[Union[str, float, List[float]]] = None,
+            names: Optional[List[str]] = None,
+            format_names: bool = True,
+            xlab: Optional[str] = None,
+            ylab: Optional[str] = None,
+            xlim: Optional[List[float]] = None,
+            ylim: Optional[List[float]] = None,
+            col: Optional[Union[str, List[str]]] = None,
+            col_names: Optional[Union[str, List[str]]] = None,
+            lty: Optional[Union[str, int, List]] = None,
+            lwd: Union[float, List[float]] = 1,
+            cex: Union[float, List[float]] = 1.0,
+            main: Optional[str] = None,
+            fill: Optional[str] = None,
+            fill_NA: str = "0.8",
+            limit_water: Optional[bool] = None,
+            plot_it: bool = True,
+            add_to: Optional[Dict[str, Any]] = None,
+            contour_method: Optional[Union[str, List[str]]] = "edge",
+            messages: bool = True,
+            interactive: bool = False,
+            annotation: Optional[str] = None,
+            annotation_coords: List[float] = [0, 0],
+            width: int = 600,
+            height: int = 520,
+            save_as: Optional[str] = None,
+            save_format: Optional[str] = None,
+            save_scale: float = 1,
+            normalize: Union[bool, List[bool]] = False,
+            as_residue: bool = False,
+            **kwargs) -> Dict[str, Any]:
+    """
+    Plot equilibrium chemical activity and predominance diagrams.
+
+    This function creates plots from the output of affinity() or equilibrate().
+    For 1D diagrams, it produces line plots showing how affinity or activity
+    varies with a single variable. For 2D diagrams, it creates predominance
+    field diagrams.
+
+    Parameters
+    ----------
+    eout : dict
+        Output from affinity() or equilibrate()
+    type : str, default "auto"
+        Type of diagram:
+        - "auto" (default): Plot affinity values (A/2.303RT)
+        - "loga.equil": Plot equilibrium activities from equilibrate()
+        - "saturation": Draw affinity=0 contour lines (mineral saturation)
+        - Basis species name (e.g., "O2", "H2O", "CO2"): Plot equilibrium
+          log activity/fugacity of the specified basis species where affinity=0
+          for each formed species. Useful for Eh-pH diagrams and showing
+          oxygen/water fugacities at equilibrium.
+    alpha : bool or str, default False
+        Plot degree of formation instead of activities?
+        If "balance", scale by balancing coefficients
+    balance : str, float, or list of float, optional
+        Balancing coefficients or method for balancing reactions
+    names : list of str, optional
+        Custom names for species (for labels)
+    format_names : bool, default True
+        Apply formatting to chemical formulas?
+    xlab : str, optional
+        Custom x-axis label
+    ylab : str, optional
+        Custom y-axis label
+    xlim : list of float, optional
+        X-axis limits [min, max]
+    ylim : list of float, optional
+        Y-axis limits [min, max]
+    col : str or list of str, optional
+        Line colors for 1-D plots and boundary lines in 2-D plots (matplotlib color specs)
+    col_names : str or list of str, optional
+        Text colors for field labels in 2-D plots (matplotlib color specs)
+    lty : str, int, or list, optional
+        Line styles (matplotlib linestyle specs)
+    lwd : float or list of float, default 1
+        Line widths for 1-D plots and boundary lines in 2-D predominance
+        diagrams. Set to 0 to disable borders in 2-D diagrams. If fill is
+        None and lwd > 0, uses white fill with black borders (R CHNOSZ default).
+    cex : float or list of float, default 1.0
+        Character expansion factor for text labels. Values > 1 make text larger,
+        values < 1 make text smaller. Can be a single value or a list (one per species).
+        Used for contour labels in type="saturation" plots.
+    main : str, optional
+        Plot title
+    fill : str, optional
+        Color palette for 2-D predominance diagrams. Can be any matplotlib
+        colormap name (e.g., 'viridis', 'plasma', 'terrain', 'rainbow',
+        'Set1', 'tab10', 'Pastel1'). If None, uses discrete colors from
+        the default color cycle. Ignored for 1-D diagrams.
+    fill_NA : str, default "0.8"
+        Color for regions outside water stability limits (water instability regions).
+        Matplotlib color specification (e.g., "0.8" for gray, "#CCCCCC").
+        Set to "transparent" to disable shading. Default "0.8" matches R's "gray80".
+    limit_water : bool, optional
+        Whether to show water stability limits as shaded regions (default True for
+        2-D diagrams). If True, also clips the diagram to the water stability region.
+        Set to False to disable water stability shading.
+    plot_it : bool, default True
+        Display the plot?
+    add_to : dict, optional
+        A diagram result dictionary from a previous diagram() call. When provided,
+        this plot will be AUTOMATICALLY COPIED and the new diagram will be added to
+        the copy. This preserves the original plot while creating a modified version.
+        The axes object is extracted from add_to['ax'].
+
+        This parameter eliminates the need for a separate 'add' boolean - when
+        add_to is provided, the function automatically operates in "add" mode.
+
+        Example workflow:
+        >>> plot_a = diagram(affinity1, fill='terrain')  # Create base plot
+        >>> plot_a1 = diagram(affinity2, add_to=plot_a, col='blue')  # Copy and add
+        >>> plot_a2 = diagram(affinity3, add_to=plot_a, col='red')   # Copy and add again
+        >>> # plot_a remains unchanged, plot_a1 and plot_a2 are independent modifications
+    contour_method : str or list of str, optional
+        Method for labeling contour lines. Default "edge" labels at plot edges.
+        Can be a single value (applied to all species) or a list (one per species).
+        Set to None, NA, or "" to disable labels (only for type="saturation").
+        In R CHNOSZ, different methods like "edge", "flattest", "simple" control
+        label placement; in Python, this mainly controls whether labels are shown.
+    interactive : bool, default False
+        Create an interactive plot using Plotly instead of matplotlib?
+        If True, calls diagram_interactive() with the appropriate parameters.
+    annotation : str, optional
+        For interactive plots only. Annotation text to add to the plot.
+    annotation_coords : list of float, default [0, 0]
+        For interactive plots only. Coordinates of annotation, where [0, 0] is
+        bottom left and [1, 1] is top right.
+    width : int, default 600
+        For interactive plots only. Width of the plot in pixels.
+    height : int, default 520
+        For interactive plots only. Height of the plot in pixels.
+    save_as : str, optional
+        For interactive plots only. Provide a filename to save this figure.
+        Filetype is determined by `save_format`.
+    save_format : str, optional
+        For interactive plots only. Desired format of saved or downloaded figure.
+        Can be 'png', 'jpg', 'jpeg', 'webp', 'svg', 'pdf', 'eps', 'json', or 'html'.
+        If 'html', an interactive plot will be saved.
+    save_scale : float, default 1
+        For interactive plots only. Multiply title/legend/axis/canvas sizes by
+        this factor when saving the figure.
+    **kwargs
+        Additional arguments passed to matplotlib plotting functions
+
+    Returns
+    -------
+    dict
+        Dictionary containing:
+        - plotvar : str, Variable that was plotted
+        - plotvals : dict, Values that were plotted
+        - names : list, Names used for labels
+        - predominant : array or NA, Predominance matrix (for 2D)
+        - balance : str or list, Balancing coefficients used
+        - n.balance : list, Numerical balancing coefficients
+        - ax : matplotlib.axes.Axes, The axes object used for plotting (if plot_it=True)
+        - fig : matplotlib.figure.Figure, The figure object used for plotting (if plot_it=True)
+        - All original eout contents
+
+    Examples
+    --------
+    >>> import pychnosz
+    >>> pychnosz.basis(["Fe2O3", "CO2", "H2O", "NH3", "H2S", "oxygen", "H+"],
+    ...              [0, -3, 0, -4, -7, -80, -7])
+    >>> pychnosz.species(["pyrite", "goethite"])
+    >>> a = pychnosz.affinity(H2S=[-60, 20, 5], T=25, P=1)
+    >>> d = diagram(a)
+
+    Notes
+    -----
+    This implementation is based on R CHNOSZ diagram() function but adapted
+    for Python's matplotlib plotting instead of R's base graphics. The key
+    differences from diagram_from_WORM.py are:
+    - Works directly with Python dict output from affinity() (no rpy2)
+    - Uses matplotlib for 1D plots by default
+    - Can optionally use plotly if requested
+    """
+
+    # Handle add_to parameter: automatically copy the provided plot
+    # This extracts the axes object and creates an independent copy
+    # When add_to is provided, we're in "add" mode
+    ax = None
+    add = add_to is not None
+    plot_was_provided = add
+
+    if add_to is not None:
+        # Make a deep copy of the provided plot to preserve the original
+        plot_copy = copy_plot(add_to)
+        # Extract the axes from the copied plot
+        if 'ax' in plot_copy:
+            ax = plot_copy['ax']
+        else:
+            raise ValueError("The 'add_to' parameter must contain an 'ax' key (a diagram result dictionary)")
+
+    # If interactive mode is requested, delegate to diagram_interactive
+    if interactive:
+        df, fig = diagram_interactive(
+            eout=eout,
+            type=type,
+            main=main,
+            borders=lwd,
+            names=names,
+            format_names=format_names,
+            annotation=annotation,
+            annotation_coords=annotation_coords,
+            balance=balance,
+            xlab=xlab,
+            ylab=ylab,
+            fill=fill,
+            width=width,
+            height=height,
+            alpha=alpha,
+            plot_it=plot_it,
+            add=add,
+            ax=ax,
+            col=col,
+            lty=lty,
+            lwd=lwd,
+            cex=cex,
+            contour_method=contour_method,
+            save_as=save_as,
+            save_format=save_format,
+            save_scale=save_scale,
+            messages=messages
+        )
+        # Return in a format compatible with diagram's normal output
+        # diagram_interactive returns (df, fig), wrap in a dict for consistency
+        # Include eout data so water_lines() can access vars, vals, basis, etc.
+        result = {
+            **eout,  # Include all original eout data
+            'df': df,
+            'fig': fig,
+            'ax': fig  # For compatibility, store fig in ax key for add=True workflow
+        }
+        return result
+
+    # Check that eout is valid
+    efun = eout.get('fun', '')
+    if efun not in ['affinity', 'equilibrate', 'solubility']:
+        raise ValueError("'eout' is not the output from affinity(), equilibrate(), or solubility()")
+
+    # Determine if eout is from affinity() (as opposed to equilibrate())
+    # Check for both Python naming (loga_equil) and R naming (loga.equil)
+    eout_is_aout = 'loga_equil' not in eout and 'loga.equil' not in eout
+
+    # Check if type is a basis species name
+    plot_loga_basis = False
+    if type not in ["auto", "saturation", "loga.equil", "loga_equil", "loga.balance", "loga_balance"]:
+        # Check if type matches a basis species name
+        if 'basis' in eout:
+            basis_species = list(eout['basis'].index) if hasattr(eout['basis'], 'index') else []
+            if type in basis_species:
+                plot_loga_basis = True
+                if alpha:
+                    raise ValueError("equilibrium activities of basis species not available with alpha = TRUE")
+
+    # Handle type="saturation" - requires affinity output
+    if type == "saturation":
+        if not eout_is_aout:
+            raise ValueError("type='saturation' requires output from affinity(), not equilibrate()")
+        # Set eout_is_aout flag
+        eout_is_aout = True
+
+    # Get number of dimensions
+    # Handle both dict (affinity) and list (equilibrate) values structures
+    if isinstance(eout['values'], dict):
+        first_values = list(eout['values'].values())[0]
+    elif isinstance(eout['values'], list):
+        first_values = eout['values'][0]
+    else:
+        first_values = eout['values']
+
+    if hasattr(first_values, 'shape'):
+        nd = len(first_values.shape)
+    elif hasattr(first_values, '__len__'):
+        nd = 1
+    else:
+        nd = 0  # Single value
+
+    # For affinity output, get balancing coefficients
+    if eout_is_aout and type == "auto":
+        n_balance, balance = _get_balance(eout, balance, messages)
+    elif eout_is_aout and type == "saturation":
+        # For saturation diagrams, use n_balance = 1 for all species (don't normalize by stoichiometry)
+        if isinstance(eout['values'], dict):
+            n_balance = [1] * len(eout['values'])
+        elif isinstance(eout['values'], list):
+            n_balance = [1] * len(eout['values'])
+        else:
+            n_balance = [1]
+        if balance is None:
+            balance = 1
+    else:
+        # For equilibrate output, use n_balance from equilibrate if available
+        if 'n_balance' in eout:
+            n_balance = eout['n_balance']
+            balance = eout.get('balance', 1)
+        else:
+            if isinstance(eout['values'], dict):
+                n_balance = [1] * len(eout['values'])
+            elif isinstance(eout['values'], list):
+                n_balance = [1] * len(eout['values'])
+            else:
+                n_balance = [1]
+            if balance is None:
+                balance = 1
+
+    # Determine what to plot
+    plotvals = {}
+    plotvar = eout.get('property', 'A')
+
+    # Calculate equilibrium log activity/fugacity of basis species
+    if plot_loga_basis:
+        # Find the index of the basis species
+        basis_df = eout['basis']
+        ibasis = list(basis_df.index).index(type)
+
+        # Get the logarithm of activity used in the affinity calculation
+        logact = basis_df.iloc[ibasis]['logact']
+
+        # Check if logact is numeric
+        try:
+            loga_basis = float(logact)
+        except (ValueError, TypeError):
+            raise ValueError(f"the logarithm of activity for basis species {type} is not numeric - was a buffer selected?")
+
+        # Get the reaction coefficients for this basis species
+        # eout['species'] is a DataFrame with basis species as columns
+        nu_basis = eout['species'].iloc[:, ibasis].values
+
+        # Calculate the logarithm of activity where affinity = 0
+        # loga_equilibrium = loga_basis - affinity / nu_basis
+        plotvals = {}
+        for i, (sp_idx, affinity_vals) in enumerate(eout['values'].items()):
+            plotvals[sp_idx] = loga_basis - affinity_vals / nu_basis[i]
+
+        plotvar = type
+
+        # Set n_balance (not used for basis species plots, but needed for compatibility)
+        n_balance = [1] * len(plotvals)
+        if balance is None:
+            balance = 1
+    elif eout_is_aout:
+        # Plot affinity values divided by balancing coefficients
+        # DEBUG: Check balance application
+        if False:  # Set to True for debugging
+            print(f"\nDEBUG: Applying balance to affinity values")
+            print(f"  n_balance: {n_balance}")
+
+        # Handle dict-based values (from affinity)
+        if isinstance(eout['values'], dict):
+            for i, (species_idx, values) in enumerate(eout['values'].items()):
+                if False:  # Set to True for debugging
+                    print(f"  Species {i} (ispecies {species_idx}): values/n_balance[{i}]={n_balance[i]}")
+                plotvals[species_idx] = values / n_balance[i]
+        # Handle list-based values
+        elif isinstance(eout['values'], list):
+            for i, values in enumerate(eout['values']):
+                species_idx = eout['species']['ispecies'].iloc[i]
+                plotvals[species_idx] = values / n_balance[i]
+
+        if plotvar == 'A':
+            plotvar = 'A/(2.303RT)'
+            if nd == 1:
+                if messages:
+                    print(f"diagram: plotting {plotvar} / n.balance")
+    else:
+        # Plot equilibrated activities
+        # Check for both Python naming (loga_equil) and R naming (loga.equil)
+        loga_equil_key = 'loga_equil' if 'loga_equil' in eout else 'loga.equil'
+        loga_equil_list = eout[loga_equil_key]
+
+        # For equilibrate output, keep plotvals as a dict with INTEGER indices as keys
+        # This preserves the 1:1 correspondence with the species list, including duplicates
+        # Do NOT use ispecies as keys because duplicates would overwrite each other
+        if isinstance(loga_equil_list, list):
+            for i, loga_val in enumerate(loga_equil_list):
+                plotvals[i] = loga_val  # Use integer index, not ispecies
+        else:
+            # Already a dict
+            plotvals = loga_equil_list
+
+        plotvar = 'loga.equil'
+
+    # Handle alpha (degree of formation)
+    if alpha:
+        # Convert to activities (remove logarithms)
+        # Use numpy arrays for proper element-wise operations
+        act_vals = {}
+        for k, v in plotvals.items():
+            if isinstance(v, np.ndarray):
+                act_vals[k] = 10**v
+            else:
+                act_vals[k] = np.power(10, v)
+
+        # Scale by balance if requested
+        if alpha == "balance":
+            species_keys = list(act_vals.keys())
+            for i, k in enumerate(species_keys):
+                act_vals[k] = act_vals[k] * n_balance[i]
+
+        # Calculate sum of activities (element-wise for arrays)
+        # Get the first value to determine shape
+        first_val = list(act_vals.values())[0]
+        if isinstance(first_val, np.ndarray):
+            # Multi-dimensional case
+            sum_act = np.zeros_like(first_val)
+            for v in act_vals.values():
+                sum_act = sum_act + v
+        else:
+            # Single value case
+            sum_act = sum(act_vals.values())
+
+        # Calculate alpha (fraction) - element-wise division
+        plotvals = {k: v / sum_act for k, v in act_vals.items()}
+        plotvar = "alpha"
+
+    # Get species information for labels
+    species_df = eout['species']
+    if names is None:
+        names = species_df['name'].tolist()
+
+    # Format chemical names if requested
+    if format_names and not alpha:
+        names = [_format_chemname(name) for name in names]
+
+    # Prepare for plotting
+    if nd == 0:
+        # 0-D: Bar plot (not implemented yet)
+        raise NotImplementedError("0-D bar plots not yet implemented")
+
+    elif nd == 1:
+        # 1-D: Line plot
+        result = _plot_1d(eout, plotvals, plotvar, names, n_balance, balance,
+                       xlab, ylab, xlim, ylim, col, lty, lwd, main, add, plot_it, ax, width, height, plot_was_provided, **kwargs)
+
+    elif nd == 2:
+        # 2-D: Predominance diagram or saturation lines
+        # Pass lty and cex through kwargs for saturation plots
+        result = _plot_2d(eout, plotvals, plotvar, names, n_balance, balance,
+                       xlab, ylab, xlim, ylim, col, col_names, fill, fill_NA, limit_water, lwd, main, add, plot_it, ax,
+                       type, contour_method, messages, width, height, plot_was_provided, lty=lty, cex=cex, **kwargs)
+
+    else:
+        raise ValueError(f"Cannot create diagram with {nd} dimensions")
+
+    # Handle Jupyter display behavior
+    # When plot_it=True, we want the figure to display
+    # When plot_it=False, we want to suppress display and close the figure
+    if not plot_it and result is not None and 'fig' in result:
+        # Close the figure to prevent auto-display in Jupyter
+        # The figure is still in the result dict, so users can access it via result['fig']
+        # but it won't be displayed automatically
+        plt.close(result['fig'])
+    elif plot_it and result is not None and 'fig' in result:
+        # Try to use IPython display if available (for Jupyter notebooks)
+        try:
+            from IPython.display import display
+            display(result['fig'])
+        except (ImportError, NameError):
+            # Not in IPython/Jupyter, regular matplotlib display
+            pass
+
+    return result
+
+
+def _get_balance(eout: Dict[str, Any], balance: Optional[Union[str, float, List[float]]], messages: bool = True) -> Tuple[List[float], Union[str, int, List[float]]]:
+    """
+    Get balancing coefficients for formation reactions.
+
+    This implements the R CHNOSZ balance() function logic for determining
+    how to balance formation reactions when calculating diagrams.
+
+    Parameters
+    ----------
+    eout : dict
+        Output from affinity()
+    balance : str, float, list of float, or None
+        Balancing specification
+
+    Returns
+    -------
+    tuple of (list of float, balance_name)
+        - Balancing coefficients for each species
+        - The balance identifier used
+    """
+    species_df = eout['species']
+    basis_df = eout['basis']
+    n_species = len(species_df)
+
+    # Get basis species column names (exclude metadata columns)
+    basis_cols = [col for col in species_df.columns
+                 if col not in ['ispecies', 'name', 'state', 'logact']]
+
+    if balance is None:
+        # Auto-select using which_balance logic
+        ibalance = _which_balance(species_df, basis_cols)
+        if len(ibalance) == 0:
+            raise ValueError("no basis species is present in all formation reactions")
+        balance_col = basis_cols[ibalance[0]]
+        n_balance = species_df[balance_col].tolist()
+        if messages:
+            print(f"balance: on moles of {balance_col} in formation reactions")
+        balance = balance_col
+    elif balance == 1 or balance == "1":
+        # Balance on one mole of species (formula units)
+        n_balance = [1] * n_species
+        if messages:
+            print("balance: on supplied numeric argument (1) [1 means balance on formula units]")
+        balance = 1
+    elif isinstance(balance, (int, float)):
+        # Use a specific basis species by index
+        if 0 < balance <= len(basis_cols):
+            balance_col = basis_cols[int(balance) - 1]
+            n_balance = species_df[balance_col].tolist()
+            if messages:
+                print(f"balance: on moles of {balance_col} in formation reactions")
+            balance = balance_col
+        else:
+            warnings.warn(f"Balance index {balance} out of range, using 1")
+            n_balance = [1] * n_species
+            balance = 1
+    elif isinstance(balance, str):
+        # Use named basis species
+        if balance in species_df.columns:
+            n_balance = species_df[balance].tolist()
+            if messages:
+                print(f"balance: on moles of {balance} in formation reactions")
+        else:
+            warnings.warn(f"Balance species '{balance}' not found, using 1")
+            n_balance = [1] * n_species
+            balance = 1
+    elif isinstance(balance, list):
+        # Use provided list
+        if len(balance) == n_species:
+            n_balance = balance
+            if messages:
+                print(f"balance: on supplied numeric argument ({','.join(map(str, balance))})")
+        else:
+            warnings.warn(f"Balance list length ({len(balance)}) doesn't match species count ({n_species}), using 1")
+            n_balance = [1] * n_species
+            balance = 1
+    else:
+        n_balance = [1] * n_species
+        balance = 1
+
+    # Handle negative coefficients (make all positive if all negative)
+    if all(x < 0 for x in n_balance):
+        n_balance = [-x for x in n_balance]
+
+    return n_balance, balance
+
+
+def _which_balance(species_df: pd.DataFrame, basis_cols: List[str]) -> List[int]:
+    """
+    Find basis species present in all formation reactions.
+
+    This implements R CHNOSZ which.balance() function.
+
+    Parameters
+    ----------
+    species_df : pd.DataFrame
+        Species dataframe with stoichiometric coefficients
+    basis_cols : list of str
+        Names of basis species columns
+
+    Returns
+    -------
+    list of int
+        Indices of basis species present in all reactions (0-indexed)
+    """
+    ib = []
+    for i, col in enumerate(basis_cols):
+        coeffs = species_df[col].values
+        # Check if all coefficients are non-zero
+        if np.all(coeffs != 0):
+            ib.append(i)
+    return ib
+
+
+def _plot_1d(eout: Dict[str, Any],
+             plotvals: Dict,
+             plotvar: str,
+             names: List[str],
+             n_balance: List[float],
+             balance: Optional[Union[str, float, List[float]]],
+             xlab: Optional[str],
+             ylab: Optional[str],
+             xlim: Optional[List[float]],
+             ylim: Optional[List[float]],
+             col: Optional[Union[str, List[str]]],
+             lty: Optional[Union[str, int, List]],
+             lwd: Union[float, List[float]],
+             main: Optional[str],
+             add: bool,
+             plot_it: bool,
+             ax: Optional[Any],
+             width: int = 600,
+             height: int = 520,
+             plot_was_provided: bool = False,
+             **kwargs) -> Dict[str, Any]:
+    """
+    Create a 1-D line plot.
+
+    Parameters
+    ----------
+    (See diagram() for parameter descriptions)
+
+    Returns
+    -------
+    dict
+        Output dictionary with plot data and metadata
+    """
+
+    # Get x-axis values
+    xvar = eout['vars'][0]
+    xvals = eout['vals'][xvar]
+
+    # Convert to numpy array if needed
+    if not isinstance(xvals, np.ndarray):
+        xvals = np.array(xvals)
+
+    # Set up axis labels
+    if xlab is None:
+        xlab = _axis_label(xvar, eout)
+
+    if ylab is None:
+        ylab = _axis_label(plotvar, eout)
+
+    # Set up x limits
+    if xlim is None:
+        xlim = [xvals[0], xvals[-1]]
+
+    # Set up colors and line styles
+    n_species = len(plotvals)
+
+    if col is None:
+        # Use matplotlib default color cycle
+        prop_cycle = plt.rcParams['axes.prop_cycle']
+        colors = prop_cycle.by_key()['color']
+        col = [colors[i % len(colors)] for i in range(n_species)]
+    elif isinstance(col, str):
+        col = [col] * n_species
+    else:
+        col = list(col) * (n_species // len(col) + 1)
+        col = col[:n_species]
+
+    if lty is None:
+        lty = ['-'] * n_species
+    elif isinstance(lty, (str, int)):
+        lty = [lty] * n_species
+    else:
+        lty = list(lty) * (n_species // len(lty) + 1)
+        lty = lty[:n_species]
+
+    if isinstance(lwd, (int, float)):
+        lwd = [lwd] * n_species
+    else:
+        lwd = list(lwd) * (n_species // len(lwd) + 1)
+        lwd = lwd[:n_species]
+
+    # Convert numeric line styles to matplotlib styles
+    lty_map = {1: '-', 2: '--', 3: '-.', 4: ':', 5: '-', 6: '--'}
+    lty = [lty_map.get(lt, lt) if isinstance(lt, int) else lt for lt in lty]
+
+    # Temporarily disable interactive mode if plot_it=False
+    # This prevents Jupyter from auto-displaying the figure
+    was_interactive = plt.isinteractive()
+    if not plot_it:
+        plt.ioff()
+
+    # Convert width and height from pixels to inches for matplotlib
+    # Use standard 96 DPI for consistency with web/screen displays
+    dpi = 96
+    figsize_inches = (width / dpi, height / dpi)
+
+    # Create figure and axes (always, even if plot_it=False)
+    # This allows the plot to be used with add_to parameter later
+    fig = None
+    ax_was_provided = ax is not None  # Track if ax was passed as parameter
+
+    if ax is not None:
+        # Use provided axes
+        fig = ax.get_figure()
+    elif not add:
+        # Create new figure and axes with specified size
+        fig, ax = plt.subplots(figsize=figsize_inches, dpi=dpi)
+    else:
+        # Try to get current axes, create new if none exists
+        try:
+            ax = plt.gca()
+            fig = ax.get_figure()
+        except:
+            fig, ax = plt.subplots(figsize=figsize_inches, dpi=dpi)
+
+    # Plot each species (always draw, regardless of plot_it)
+    # plot_it only controls display, not drawing
+    for i, (species_idx, yvals) in enumerate(plotvals.items()):
+        # Convert to numpy array if needed
+        if not isinstance(yvals, np.ndarray):
+            yvals = np.array([yvals] * len(xvals))
+
+        ax.plot(xvals, yvals,
+               color=col[i],
+               linestyle=lty[i],
+               linewidth=lwd[i],
+               label=names[i],
+               **kwargs)
+
+    # Set labels and limits
+    ax.set_xlabel(xlab)
+    ax.set_ylabel(ylab)
+
+    if xlim is not None:
+        ax.set_xlim(xlim)
+
+    if ylim is not None:
+        ax.set_ylim(ylim)
+
+    # Add legend
+    ax.legend()
+
+    # Add title
+    if main is not None:
+        ax.set_title(main)
+
+    # Add grid
+    ax.grid(True, alpha=0.3)
+
+    if not add:
+        plt.tight_layout()
+
+    # Build output dictionary
+    result = {
+        **eout,
+        'plotvar': plotvar,
+        'plotvals': plotvals,
+        'names': names,
+        'predominant': np.nan,
+        'balance': balance,
+        'n.balance': n_balance
+    }
+
+    # Add figure and axes to output if they were created
+    if fig is not None:
+        if not ax_was_provided or plot_was_provided:
+            result['ax'] = ax
+            result['fig'] = fig
+
+    # Always restore interactive mode to its original state
+    if was_interactive and not plt.isinteractive():
+        plt.ion()
+    elif not was_interactive and plt.isinteractive():
+        plt.ioff()
+
+    return result
+
+
+def _plot_2d(eout: Dict[str, Any],
+            plotvals: Dict[int, np.ndarray],
+            plotvar: str,
+            names: List[str],
+            n_balance: List[float],
+            balance: Union[str, int, List[float]],
+            xlab: Optional[str],
+            ylab: Optional[str],
+            xlim: Optional[List[float]],
+            ylim: Optional[List[float]],
+            col: Optional[Union[str, List[str]]],
+            col_names: Optional[Union[str, List[str]]],
+            fill: Optional[str],
+            fill_NA: str,
+            limit_water: Optional[bool],
+            lwd: Union[float, List[float]],
+            main: Optional[str],
+            add: bool,
+            plot_it: bool,
+            ax: Optional[Any],
+            type: str = "auto",
+            contour_method: Optional[str] = "edge",
+            messages: bool = True,
+            width: int = 600,
+            height: int = 520,
+            plot_was_provided: bool = False,
+            **kwargs) -> Dict[str, Any]:
+    """
+    Create a 2-D predominance diagram (internal function).
+
+    This function determines which species has the maximum affinity at each
+    point on a 2D grid and creates a predominance field diagram.
+
+    Parameters
+    ----------
+    eout : dict
+        Output from affinity()
+    plotvals : dict
+        Values to plot (affinity or equilibrium values)
+    plotvar : str
+        Variable being plotted
+    names : list of str
+        Species names for labels
+    n_balance : list of float
+        Balancing coefficients
+    balance : str, int, or list
+        Balance identifier
+    xlab, ylab : str or None
+        Axis labels
+    xlim, ylim : list or None
+        Axis limits
+    col : str, list, or None
+        Colors for boundary lines in 2D plots
+    col_names : str, list, or None
+        Colors for field labels (text) in 2D plots
+    fill : str or None
+        Matplotlib colormap name for coloring predominance fields
+        (e.g., 'viridis', 'plasma', 'terrain', 'rainbow', 'Set1', 'tab10')
+    lwd : float or list of float
+        Line width for drawing boundaries between predominance fields.
+        Set to 0 to disable borders.
+    main : str or None
+        Plot title
+    add : bool
+        Add to existing plot?
+    plot_it : bool
+        Display the plot?
+    **kwargs
+        Additional matplotlib arguments
+
+    Returns
+    -------
+    dict
+        Result dictionary with predominance information
+    """
+
+    # Get the two variables
+    vars_list = eout['vars']
+    if len(vars_list) != 2:
+        raise ValueError(f"Expected 2 variables for 2-D plot, got {len(vars_list)}")
+
+    # R CHNOSZ convention: first variable in affinity() → x-axis, second → y-axis
+    # In the array: rows correspond to first var, columns to second var
+    xvar = vars_list[0]  # First variable (rows in array) → x-axis
+    yvar = vars_list[1]  # Second variable (cols in array) → y-axis
+
+    # Get the values for each variable
+    xvals = eout['vals'][xvar]
+    yvals = eout['vals'][yvar]
+
+    # Convert to numpy arrays if needed
+    xvals = np.asarray(xvals)
+    yvals = np.asarray(yvals)
+
+    # Get axis labels
+    if xlab is None:
+        xlab = _axis_label(xvar, eout)
+    if ylab is None:
+        ylab = _axis_label(yvar, eout)
+
+    # Handle saturation lines separately from predominance diagrams
+    if type == "saturation":
+        # Extract lty and cex from kwargs (they're in kwargs from diagram() call)
+        lty_param = kwargs.pop('lty', None)
+        cex_param = kwargs.pop('cex', 1.0)
+
+        return _plot_saturation_2d(eout, plotvals, plotvar, names, n_balance, balance,
+                                   xlab, ylab, xlim, ylim, col, lwd, lty_param, cex_param,
+                                   main, add, plot_it, ax, contour_method, messages, width, height, plot_was_provided, **kwargs)
+
+    # For non-saturation plots, remove lty and cex from kwargs to avoid passing to matplotlib
+    kwargs.pop('lty', None)
+    kwargs.pop('cex', None)
+
+    # Print message about diagram method
+    if messages:
+        print(f"diagram: using maximum affinity method for 2-D diagram")
+
+    # Stack all species values into a 3D array (species, rows, cols)
+    # The plotvals dict can have two types of keys:
+    # 1. Integer indices (0, 1, 2, ...) for equilibrate output - preserves duplicates
+    # 2. ispecies values (1844, 1876, ...) for affinity output - unique species only
+    species_keys = list(plotvals.keys())
+    n_species = len(species_keys)
+
+    # Check if species_keys are integer indices (equilibrate) or ispecies values (affinity)
+    # Integer indices will be 0, 1, 2, ..., n-1 and directly map to names list
+    # ispecies values are typically > 100 and need mapping
+    species_df = eout['species']
+    if all(isinstance(k, int) and k < len(names) for k in species_keys):
+        # Integer indices: direct mapping to names
+        predominant_to_names_idx = {i: species_keys[i] for i in range(n_species)}
+    else:
+        # ispecies values: need to find matching rows
+        # Use first matching row (for consistency with affinity output)
+        predominant_to_names_idx = {}
+        for i, sp_idx in enumerate(species_keys):
+            matching_rows = species_df[species_df['ispecies'] == sp_idx].index.tolist()
+            if len(matching_rows) > 0:
+                predominant_to_names_idx[i] = matching_rows[0]
+            else:
+                predominant_to_names_idx[i] = i  # Fallback
+
+    # DEBUG: Print species_keys order
+    # print(f"DEBUG: species_keys = {species_keys}")
+
+    # Get dimensions from first species
+    first_vals = plotvals[species_keys[0]]
+    if len(first_vals.shape) != 2:
+        raise ValueError(f"Expected 2-D array for each species, got shape {first_vals.shape}")
+
+    # Array shape: (n_xvar, n_yvar) since first var → x-axis, second → y-axis
+    # For our example: (n_T, n_P) = (3, 10)
+    n_xvar, n_yvar = first_vals.shape
+
+    # Stack all species affinities into a 3D array
+    affinity_stack = np.zeros((n_species, n_xvar, n_yvar))
+    for i, sp_idx in enumerate(species_keys):
+        affinity_stack[i, :, :] = plotvals[sp_idx]
+
+    # DEBUG: Check species order
+    if False:  # Set to True for debugging
+        print(f"\nDEBUG affinity_stack:")
+        print(f"  n_species: {n_species}")
+        print(f"  species_keys: {species_keys}")
+        for i, sp_idx in enumerate(species_keys):
+            print(f"  Stack position {i}: species {sp_idx}")
+
+    # Find the species with maximum affinity at each point
+    # predominant will have indices 0, 1, 2, ... for species
+    predominant_indices = np.argmax(affinity_stack, axis=0)
+
+    # Get the affinity values at predominant points
+    predominant_values = np.max(affinity_stack, axis=0)
+
+    # Convert indices to species indices (1-based for R compatibility)
+    # In R, predominant contains 1, 2, 3, etc.
+    predominant = predominant_indices + 1
+
+    # Calculate water stability limits if requested (default True for 2-D diagrams)
+    H2O_predominant = None
+    if limit_water is None:
+        limit_water = not add  # True unless adding to existing plot
+
+    if limit_water:
+        # Call water_lines with plot_it=False to get boundaries
+        wl = water_lines(eout, plot_it=False, messages=messages)
+        # Check if water_lines produced valid results
+        if not (wl['y_oxidation'] is None or wl['y_reduction'] is None):
+            # Create a copy of predominant matrix for water stability masking
+            # Convert to float to allow NaN values
+            H2O_predominant = predominant.astype(float).copy()
+
+            # For each x-point, find y-values outside water stability limits
+            for i in range(len(wl['xpoints'])):
+                ymin = min(wl['y_oxidation'][i], wl['y_reduction'][i])
+                ymax = max(wl['y_oxidation'][i], wl['y_reduction'][i])
+
+                if not wl['swapped']:
+                    # Normal orientation: x is first var, y is second var
+                    # eout['vals'][yvar] contains the y-axis values
+                    yvals = np.asarray(eout['vals'][yvar])
+                    # Find indices where y is outside stability range
+                    iNA = (yvals < ymin) | (yvals > ymax)
+                    # Mark those regions as NA (using nan)
+                    H2O_predominant[i, iNA] = np.nan
+                else:
+                    # Swapped: first var is y-axis
+                    xvals = np.asarray(eout['vals'][xvar])
+                    iNA = (xvals < ymin) | (xvals > ymax)
+                    H2O_predominant[iNA, i] = np.nan
+
+    # For plotting: transpose arrays so that x-axis is horizontal and y-axis is vertical
+    # imshow expects (n_rows, n_cols) = (n_yaxis, n_xaxis)
+    # Current shape is (n_xvar, n_yvar), so transpose to (n_yvar, n_xvar)
+    predominant_indices_T = predominant_indices.T
+    if H2O_predominant is not None:
+        H2O_predominant_T = H2O_predominant.T
+    else:
+        H2O_predominant_T = None
+
+    # Temporarily disable interactive mode if plot_it=False
+    # This prevents Jupyter from auto-displaying the figure
+    was_interactive = plt.isinteractive()
+    if not plot_it:
+        plt.ioff()
+
+    # Convert width and height from pixels to inches for matplotlib
+    # Use standard 96 DPI for consistency with web/screen displays
+    dpi = 96
+    figsize_inches = (width / dpi, height / dpi)
+
+    # Create figure and axes (always, even if plot_it=False)
+    # This allows the plot to be used with add_to parameter later
+    fig = None
+    ax_was_provided = ax is not None  # Track if ax was passed as parameter
+
+    if ax is not None:
+        # Use provided axes
+        fig = ax.get_figure()
+    elif not add:
+        # Create new figure and axes with specified size
+        fig, ax = plt.subplots(figsize=figsize_inches, dpi=dpi)
+    else:
+        # Try to get current axes, create new if none exists
+        try:
+            ax = plt.gca()
+            fig = ax.get_figure()
+        except:
+            fig, ax = plt.subplots(figsize=figsize_inches, dpi=dpi)
+
+    # When add=True, don't draw fill by default (to overlay on existing plot)
+    # User can explicitly provide fill parameter to override this
+    draw_fill = True
+    if add and fill is None:
+        draw_fill = False
+
+    # Draw the plot content (always, regardless of plot_it)
+    # plot_it only controls display, not drawing
+    # Determine fill colors for predominance fields
+    # Priority: fill parameter > default
+    # R CHNOSZ default: white fill with black borders
+    if fill is not None:
+        # Use a matplotlib colormap
+        try:
+            import matplotlib.cm as cm
+            cmap = cm.get_cmap(fill)
+            # Sample the colormap evenly across species
+            # Use range 0.1 to 0.9 to avoid very light/dark ends
+            color_indices = np.linspace(0.1, 0.9, n_species)
+            fill_colors = [cmap(idx) for idx in color_indices]
+        except (ValueError, KeyError):
+            warnings.warn(f"Colormap '{fill}' not found, using default colors")
+            prop_cycle = plt.rcParams['axes.prop_cycle']
+            colors = prop_cycle.by_key()['color']
+            fill_colors = [colors[i % len(colors)] for i in range(n_species)]
+    elif fill is None and lwd > 0:
+        # R CHNOSZ default behavior: white fill with black borders
+        fill_colors = ['white'] * n_species
+    else:
+        # Use default matplotlib colors
+        prop_cycle = plt.rcParams['axes.prop_cycle']
+        colors = prop_cycle.by_key()['color']
+        fill_colors = [colors[i % len(colors)] for i in range(n_species)]
+
+    # Determine boundary line colors (col parameter)
+    if col is None:
+        # Default to black for boundary lines
+        boundary_colors = ['black'] * n_species
+    elif isinstance(col, str):
+        boundary_colors = [col] * n_species
+    else:
+        boundary_colors = list(col)
+        if len(boundary_colors) < n_species:
+            # Repeat colors if not enough provided
+            boundary_colors = boundary_colors * (n_species // len(boundary_colors) + 1)
+        boundary_colors = boundary_colors[:n_species]
+
+    # Determine text label colors (col_names parameter)
+    if col_names is None:
+        # Default to black for text labels
+        text_colors = ['black'] * n_species
+    elif isinstance(col_names, str):
+        text_colors = [col_names] * n_species
+    else:
+        text_colors = list(col_names)
+        if len(text_colors) < n_species:
+            # Repeat colors if not enough provided
+            text_colors = text_colors * (n_species // len(text_colors) + 1)
+        text_colors = text_colors[:n_species]
+
+    # NOTE: Water instability shading is NOT drawn here in diagram()
+    # It is only drawn when water_lines() is explicitly called
+    # We keep H2O_predominant for the limit_water feature, but don't show gray shading
+
+    # Draw filled predominance fields only if not adding to existing plot
+    # (or if user explicitly provided fill parameter)
+    if draw_fill:
+        # Create a colored map showing predominance fields
+        # Map each predominance index to its color
+        # Shape is (n_yvar, n_xvar) after transpose
+        colored_predominant = np.zeros((n_yvar, n_xvar, 3))  # RGB image
+        from matplotlib.colors import to_rgb
+
+        for i in range(n_species):
+            mask = predominant_indices_T == i
+            rgb = to_rgb(fill_colors[i])
+            colored_predominant[mask] = rgb
+
+        # Display the predominance map
+        # imshow plots: rows → y-axis (vertical), cols → x-axis (horizontal)
+        # extent sets the data coordinates: [x_start, x_end, y_start, y_end]
+        # IMPORTANT: Use original order (xvals[0] to xvals[-1]), not min/max
+        # This preserves axes with decreasing values (e.g., F- from -2 to -9)
+        extent = [xvals[0], xvals[-1], yvals[0], yvals[-1]]
+
+        # Expand extent if any dimension has identical limits to avoid matplotlib warning
+        if extent[0] == extent[1]:
+            x_range = abs(extent[0]) * 0.1 if extent[0] != 0 else 0.1
+            extent[0] -= x_range
+            extent[1] += x_range
+        if extent[2] == extent[3]:
+            y_range = abs(extent[2]) * 0.1 if extent[2] != 0 else 0.1
+            extent[2] -= y_range
+            extent[3] += y_range
+
+        im = ax.imshow(colored_predominant, aspect='auto', origin='lower',
+                      extent=extent, interpolation='nearest', **kwargs)
+
+    # Add species labels at the center of their predominance regions
+    for i in range(n_species):
+        # Find all points where this species predominates (in transposed array)
+        mask = predominant_indices_T == i
+        if np.any(mask):
+            # Get row and column indices from transposed array
+            # rows correspond to y-axis values, cols to x-axis values
+            rows, cols = np.where(mask)
+            # Calculate mean position
+            mean_row = rows.mean()
+            mean_col = cols.mean()
+            # Convert to data coordinates
+            # cols index into xvals, rows index into yvals
+            x_pos = xvals[int(mean_col)]
+            y_pos = yvals[int(mean_row)]
+
+            # Map predominant index to correct names index
+            # This handles cases where duplicate species exist in the species list
+            names_idx = predominant_to_names_idx[i]
+
+            # Add text label with color from col_names
+            ax.text(x_pos, y_pos, names[names_idx],
+                   ha='center', va='center',
+                   color=text_colors[i],
+                   bbox=dict(boxstyle='round', facecolor='white', edgecolor='none', alpha=0.7),
+                   fontsize=10, fontweight='bold')
+
+    # Set labels and limits
+    ax.set_xlabel(xlab)
+    ax.set_ylabel(ylab)
+
+    if xlim is not None:
+        ax.set_xlim(xlim)
+    else:
+        # Preserve original axis direction (important for decreasing axes)
+        x_min, x_max = xvals[0], xvals[-1]
+        # Expand range if limits are identical to avoid matplotlib warning
+        if x_min == x_max:
+            x_range = abs(x_max) * 0.1 if x_max != 0 else 0.1
+            x_min -= x_range
+            x_max += x_range
+        ax.set_xlim([x_min, x_max])
+
+    if ylim is not None:
+        ax.set_ylim(ylim)
+    else:
+        # Preserve original axis direction (important for decreasing axes)
+        y_min, y_max = yvals[0], yvals[-1]
+        # Expand range if limits are identical to avoid matplotlib warning
+        if y_min == y_max:
+            y_range = abs(y_max) * 0.1 if y_max != 0 else 0.1
+            y_min -= y_range
+            y_max += y_range
+        ax.set_ylim([y_min, y_max])
+
+    # Add title
+    if main is not None:
+        ax.set_title(main)
+
+    # Draw borders between predominance fields
+    if lwd > 0:
+        # Use matplotlib's contour function to draw boundaries between species
+        # This matches R CHNOSZ's use of contourLines()
+        # Following R CHNOSZ approach: loop over species and draw contours at level 0.5
+
+        # Get unique species values (excluding any that don't appear)
+        unique_species = np.unique(predominant_indices_T)
+        unique_species = unique_species[~np.isnan(unique_species)]
+
+        # Create meshgrid for contour (matches actual data coordinates)
+        X, Y = np.meshgrid(xvals, yvals)
+
+        # Loop over species (except the last one to avoid double-plotting)
+        for i in range(len(unique_species) - 1):
+            species_idx = int(unique_species[i])
+
+            # Create a binary mask: 1 where this species predominates, 0 elsewhere
+            z = (predominant_indices_T == species_idx).astype(float)
+
+            # Draw contour at level 0.5 (boundary between 0 and 1)
+            # This creates smooth boundaries following the actual grid
+            try:
+                line_color = boundary_colors[species_idx]
+                cs = ax.contour(X, Y, z, levels=[0.5], colors=[line_color],
+                               linewidths=lwd, zorder=10)
+            except:
+                pass  # Skip if contour can't be drawn (e.g., species doesn't appear)
+
+    if not add:
+        plt.tight_layout()
+
+    # Don't close the figure when plot_it=False
+    # The plt.ioff() above already prevents auto-display in Jupyter
+    # This keeps the figure available for adding titles, legends, and later display
+    # Users can display with: d['fig'].show() or display(d['fig']) in Jupyter
+
+    # Build output dictionary (matching R CHNOSZ structure)
+    result = {
+        **eout,
+        'plotvar': plotvar,
+        'plotvals': plotvals,
+        'names': names,
+        'predominant': predominant,
+        'predominant.values': predominant_values,
+        'balance': balance,
+        'n.balance': n_balance
+    }
+
+    # Add figure and axes to output if they were created
+    if fig is not None:
+        if not ax_was_provided or plot_was_provided:
+            result['ax'] = ax
+            result['fig'] = fig
+
+    # Always restore interactive mode to its original state
+    if was_interactive and not plt.isinteractive():
+        plt.ion()
+    elif not was_interactive and plt.isinteractive():
+        plt.ioff()
+
+    return result
+
+
+def _plot_saturation_2d(eout: Dict[str, Any],
+                        plotvals: Dict[int, np.ndarray],
+                        plotvar: str,
+                        names: List[str],
+                        n_balance: List[float],
+                        balance: Union[str, int, List[float]],
+                        xlab: str,
+                        ylab: str,
+                        xlim: Optional[List[float]],
+                        ylim: Optional[List[float]],
+                        col: Optional[Union[str, List[str]]],
+                        lwd: Union[float, List[float]],
+                        lty: Optional[Union[str, int, List]],
+                        cex: Union[float, List[float]],
+                        main: Optional[str],
+                        add: bool,
+                        plot_it: bool,
+                        ax: Optional[Any],
+                        contour_method: Optional[Union[str, List[str]]],
+                        messages: bool = True,
+                        width: int = 600,
+                        height: int = 520,
+                        plot_was_provided: bool = False,
+                        **kwargs) -> Dict[str, Any]:
+    """
+    Plot saturation lines (affinity=0 contours) for 2-D diagrams.
+
+    This function draws contour lines where affinity = 0 for each species,
+    indicating saturation boundaries (e.g., mineral precipitation thresholds).
+
+    Parameters
+    ----------
+    (Most parameters are the same as _plot_2d)
+    contour_method : str, list of str, or None
+        Method for labeling contour lines. Can be a single value (applied to all species)
+        or a list (one per species). If None, NA, or "", disable labels.
+        Matplotlib doesn't support the same contour methods as R, so this mainly
+        controls whether labels are drawn (any non-None/non-empty value enables labels).
+
+    Returns
+    -------
+    dict
+        Result dictionary
+    """
+
+    # Get the two variables
+    vars_list = eout['vars']
+    xvar = vars_list[0]
+    yvar = vars_list[1]
+
+    # Get the values for each variable
+    xvals = np.asarray(eout['vals'][xvar])
+    yvals = np.asarray(eout['vals'][yvar])
+
+    species_keys = list(plotvals.keys())
+    n_species = len(species_keys)
+
+    if messages:
+        print(f"diagram: plotting saturation lines for 2-D diagram")
+
+    # Set up colors and line styles
+    if col is None:
+        # Use matplotlib default color cycle
+        prop_cycle = plt.rcParams['axes.prop_cycle']
+        colors = prop_cycle.by_key()['color']
+        col = [colors[i % len(colors)] for i in range(n_species)]
+    elif isinstance(col, str):
+        col = [col] * n_species
+    else:
+        col = list(col)
+        if len(col) < n_species:
+            col = col * (n_species // len(col) + 1)
+        col = col[:n_species]
+
+    if isinstance(lwd, (int, float)):
+        lwd = [lwd] * n_species
+    else:
+        lwd = list(lwd)
+        if len(lwd) < n_species:
+            lwd = lwd * (n_species // len(lwd) + 1)
+        lwd = lwd[:n_species]
+
+    # Handle line styles (lty)
+    if lty is None:
+        lty = ['-'] * n_species
+    elif isinstance(lty, (str, int)):
+        lty = [lty] * n_species
+    else:
+        lty = list(lty)
+        if len(lty) < n_species:
+            lty = lty * (n_species // len(lty) + 1)
+        lty = lty[:n_species]
+
+    # Convert numeric line styles to matplotlib styles
+    lty_map = {1: '-', 2: '--', 3: '-.', 4: ':', 5: '-', 6: '--'}
+    lty = [lty_map.get(lt, lt) if isinstance(lt, int) else lt for lt in lty]
+
+    # Handle text size (cex) for contour labels
+    if isinstance(cex, (int, float)):
+        cex_list = [cex] * n_species
+    else:
+        cex_list = list(cex)
+        if len(cex_list) < n_species:
+            cex_list = cex_list * (n_species // len(cex_list) + 1)
+        cex_list = cex_list[:n_species]
+
+    # Determine if labels should be drawn (per species)
+    # Convert contour_method to a list (one per species)
+    if contour_method is None or contour_method == "" or (isinstance(contour_method, str) and contour_method.upper() == "NA"):
+        # No labels for any species
+        drawlabels = [False] * n_species
+    elif isinstance(contour_method, str):
+        # Same method for all species
+        drawlabels = [True] * n_species
+    elif isinstance(contour_method, list):
+        # Per-species methods
+        if len(contour_method) != n_species:
+            # Repeat/extend to match number of species
+            contour_method_extended = list(contour_method) * (n_species // len(contour_method) + 1)
+            contour_method_extended = contour_method_extended[:n_species]
+        else:
+            contour_method_extended = contour_method
+
+        # Check each method to determine if labels should be drawn
+        drawlabels = []
+        for method in contour_method_extended:
+            if method is None or method == "" or (isinstance(method, str) and method.upper() == "NA"):
+                drawlabels.append(False)
+            else:
+                drawlabels.append(True)
+    else:
+        drawlabels = [True] * n_species
+
+    # Temporarily disable interactive mode if plot_it=False
+    # This prevents Jupyter from auto-displaying the figure
+    was_interactive = plt.isinteractive()
+    if not plot_it:
+        plt.ioff()
+
+    # Convert width and height from pixels to inches for matplotlib
+    # Use standard 96 DPI for consistency with web/screen displays
+    dpi = 96
+    figsize_inches = (width / dpi, height / dpi)
+
+    # Create figure and axes (always, even if plot_it=False)
+    # This allows the plot to be used with add_to parameter later
+    fig = None
+    ax_was_provided = ax is not None  # Track if ax was passed as parameter
+
+    if ax is not None:
+        fig = ax.get_figure()
+    elif not add:
+        fig, ax = plt.subplots(figsize=figsize_inches, dpi=dpi)
+    else:
+        try:
+            ax = plt.gca()
+            fig = ax.get_figure()
+        except:
+            fig, ax = plt.subplots(figsize=figsize_inches, dpi=dpi)
+
+    # Only do the actual plotting if plot_it=True
+    # Draw the plot content (always, regardless of plot_it)
+    # plot_it only controls display, not drawing
+    if not add:
+        ax.set_xlabel(xlab)
+        ax.set_ylabel(ylab)
+
+        if xlim is not None:
+            ax.set_xlim(xlim)
+        else:
+            ax.set_xlim([xvals.min(), xvals.max()])
+
+        if ylim is not None:
+            ax.set_ylim(ylim)
+        else:
+            ax.set_ylim([yvals.min(), yvals.max()])
+
+        if main is not None:
+            ax.set_title(main)
+
+    # Plot saturation lines (affinity = 0 contours) for each species
+    for i, sp_idx in enumerate(species_keys):
+        zs = plotvals[sp_idx]
+
+        # Skip plotting if this species has no possible saturation line
+        if len(np.unique(zs)) == 1:
+            if messages:
+                print(f"diagram: no saturation line possible for {names[i]}")
+            continue
+
+        # Skip if line is outside the plot range
+        if np.all(zs < 0) or np.all(zs > 0):
+            if messages:
+                print(f"diagram: beyond range for saturation line of {names[i]}")
+            continue
+
+        # Draw the contour line at affinity = 0
+        # matplotlib's contour needs (X, Y, Z) where X and Y are meshgrids
+        X, Y = np.meshgrid(xvals, yvals)
+        # Transpose zs to match meshgrid orientation
+        # plotvals has shape (n_xvar, n_yvar), but contour expects (n_yvar, n_xvar)
+        zs_T = zs.T
+
+        try:
+            # Calculate font size from cex (matplotlib default is ~10pt)
+            fontsize = 9 * cex_list[i]
+
+            if drawlabels[i]:
+                CS = ax.contour(X, Y, zs_T, levels=[0], colors=col[i],
+                               linewidths=lwd[i], linestyles=lty[i], **kwargs)
+                ax.clabel(CS, inline=True, fontsize=fontsize, fmt=names[i])
+            else:
+                ax.contour(X, Y, zs_T, levels=[0], colors=col[i],
+                          linewidths=lwd[i], linestyles=lty[i], **kwargs)
+        except Exception as e:
+            warnings.warn(f"Could not plot contour for {names[i]}: {e}")
+
+    if not add:
+        plt.tight_layout()
+
+    # Build output dictionary
+    result = {
+        **eout,
+        'plotvar': plotvar,
+        'plotvals': plotvals,
+        'names': names,
+        'predominant': np.nan,
+        'balance': balance,
+        'n.balance': n_balance
+    }
+
+    # Add figure and axes to output if they were created
+    if fig is not None:
+        if not ax_was_provided or plot_was_provided:
+            result['ax'] = ax
+            result['fig'] = fig
+
+    # Always restore interactive mode to its original state
+    if was_interactive and not plt.isinteractive():
+        plt.ion()
+    elif not was_interactive and plt.isinteractive():
+        plt.ioff()
+
+    return result
+
+
+def _axis_label(var: str, eout: Dict[str, Any]) -> str:
+    """
+    Generate axis label for a variable.
+
+    Parameters
+    ----------
+    var : str
+        Variable name
+    eout : dict
+        Output from affinity()
+
+    Returns
+    -------
+    str
+        Formatted axis label
+    """
+
+    # Special cases
+    if var == 'A/(2.303RT)':
+        return r'A/(2.303RT)'
+    elif var == 'alpha':
+        return r'$\alpha$'
+    elif var == 'loga.equil':
+        return r'log activity'
+    elif var == 'pH':
+        return 'pH'
+    elif var == 'pe':
+        return 'pe'
+    elif var == 'Eh':
+        return 'Eh (V)'
+    elif var == 'T':
+        return 'Temperature (°C)'
+    elif var == 'P':
+        return 'Pressure (bar)'
+    elif var == 'IS':
+        return 'Ionic strength'
+    else:
+        # Check if it's a basis species
+        basis_df = eout.get('basis')
+        if basis_df is not None and var in basis_df.index:
+            state = basis_df.loc[var, 'state']
+            # Format the chemical formula with proper LaTeX subscripts and superscripts
+            var_formatted = _format_species_latex(var)
+            if state in ['aq', 'liq', 'cr']:
+                return f'$\\log\\ a_{{{var_formatted}}}$'
+            else:
+                return f'$\\log\\ f_{{{var_formatted}}}$'
+        return var
+
+
+def _format_chemname(name: str) -> str:
+    """
+    Format a chemical formula for display in matplotlib.
+
+    Uses LaTeX formatting for proper subscripts and superscripts.
+    Delegates to _format_species_latex from expression.py for consistency.
+
+    Parameters
+    ----------
+    name : str
+        Chemical formula
+
+    Returns
+    -------
+    str
+        Formatted formula (using LaTeX for matplotlib)
+    """
+    # Use the centralized formatting function and wrap in math mode for matplotlib
+    latex_formula = _format_species_latex(name)
+    return f'${latex_formula}$'
+
+
+def water_lines(eout: Dict[str, Any],
+                which: Union[str, List[str]] = ['oxidation', 'reduction'],
+                lty: Union[int, str] = 2,
+                lwd: float = 1,
+                col: Optional[str] = None,
+                plot_it: bool = True,
+                messages: bool = True) -> Dict[str, Any]:
+    """
+    Draw water stability limits for Eh-pH, logfO2-pH, logfO2-T or Eh-T diagrams.
+
+    This function adds lines showing the oxidation and reduction stability limits
+    of water to diagrams. Above the oxidation line, water breaks down to O2.
+    Below the reduction line, water breaks down to H2.
+
+    Parameters
+    ----------
+    eout : dict
+        Output from affinity(), equilibrate(), or diagram()
+    which : str or list of str, default ['oxidation', 'reduction']
+        Which line(s) to draw: 'oxidation', 'reduction', or both
+    lty : int or str, default 2
+        Line style (matplotlib linestyle or numeric code)
+    lwd : float, default 1
+        Line width
+    col : str, optional
+        Line color (matplotlib color spec). If None, uses current foreground color
+    plot_it : bool, default True
+        Whether to plot the lines and display the figure. When True, the lines
+        are added to the diagram and the figure is displayed (useful when the
+        original diagram was created with plot_it=False). When False, only
+        calculates and returns the water line coordinates without plotting.
+
+    Returns
+    -------
+    dict
+        Dictionary containing all keys from the input diagram (including 'fig', 'ax',
+        'plotvar', 'plotvals', 'names', 'predominant', etc. if present) plus the
+        following water line specific keys:
+        - xpoints: x-axis values
+        - y_oxidation: y values for oxidation line (or None)
+        - y_reduction: y values for reduction line (or None)
+        - swapped: whether axes were swapped
+
+    Examples
+    --------
+    >>> # Add water lines to an existing displayed diagram
+    >>> basis(["Fe+2", "SO4-2", "H2O", "H+", "e-"], [0, math.log10(3), math.log10(0.75), 999, 999])
+    >>> species(["rhomboclase", "ferricopiapite", "hydronium jarosite", "goethite", "melanterite", "pyrite"])
+    >>> a = affinity(pH=[-1, 4, 256], pe=[-5, 23, 256])
+    >>> d = diagram(a, main="Fe-S-O-H, after Majzlan et al., 2006")
+    >>> water_lines(d, lwd=2)
+
+    >>> # Add water lines and display when diagram was created with plot_it=False
+    >>> d = diagram(a, main="Fe-S-O-H", plot_it=False)
+    >>> water_lines(d, lwd=2)  # This will display the figure with water lines
+
+    Notes
+    -----
+    This function only works on diagrams with a redox variable (Eh, pe, O2, or H2)
+    on one axis and pH, T, P, or another non-redox variable on the other axis.
+    For 1-D diagrams, vertical lines are drawn.
+    """
+
+    # Import here to avoid circular imports
+    from ..utils.units import convert, envert
+    from ..core.subcrt import subcrt
+
+    # Create a deep copy of the input to preserve all diagram information
+    # This allows us to return all the original keys plus water line data
+    result = copy_plot(eout)
+
+    # Detect if this is a Plotly figure (interactive diagram)
+    is_plotly = False
+    if 'fig' in result and result['fig'] is not None:
+        is_plotly = hasattr(result['fig'], 'add_trace') and hasattr(result['fig'], 'update_layout')
+
+    # Ensure which is a list
+    if isinstance(which, str):
+        which = [which]
+
+    # Get number of variables used in affinity()
+    nvar1 = len(result['vars'])
+
+    # Determine actual number of variables from array dimensions
+    # Check both loga.equil (equilibrate output) and values (affinity output)
+    if 'loga_equil' in result or 'loga.equil' in result:
+        loga_key = 'loga_equil' if 'loga_equil' in result else 'loga.equil'
+        first_val = result[loga_key][0] if isinstance(result[loga_key], list) else list(result[loga_key].values())[0]
+    else:
+        first_val = list(result['values'].values())[0] if isinstance(result['values'], dict) else result['values'][0]
+
+    if hasattr(first_val, 'shape'):
+        dim = first_val.shape
+    elif hasattr(first_val, '__len__'):
+        dim = (len(first_val),)
+    else:
+        dim = ()
+
+    nvar2 = len(dim)
+
+    # We only work on diagrams with 1 or 2 variables
+    if nvar1 not in [1, 2] or nvar2 not in [1, 2]:
+        result.update({'xpoints': None, 'y_oxidation': None, 'y_reduction': None, 'swapped': False})
+        return result
+
+    # Get variables from result
+    vars_list = result['vars'].copy()
+
+    # If needed, swap axes so redox variable is on y-axis
+    # Also do this for 1-D diagrams
+    if len(vars_list) == 1:
+        vars_list.append('nothing')
+
+    swapped = False
+    if vars_list[1] in ['T', 'P', 'nothing']:
+        vars_list = list(reversed(vars_list))
+        vals_dict = {vars_list[0]: result['vals'][vars_list[0]]} if vars_list[0] != 'nothing' else {}
+        if len(result['vars']) > 1:
+            vals_dict[vars_list[1]] = result['vals'][vars_list[1]]
+        swapped = True
+    else:
+        vals_dict = result['vals']
+
+    xaxis = vars_list[0]
+    yaxis = vars_list[1]
+    xpoints = np.asarray(vals_dict[xaxis]) if xaxis in vals_dict else np.array([0])
+
+    # Make xaxis "nothing" if it is not pH, T, or P
+    # (so that horizontal water lines can be drawn for any non-redox variable on the x-axis)
+    if xaxis not in ['pH', 'T', 'P']:
+        xaxis = 'nothing'
+
+    # T and P are constants unless they are plotted on one of the axes
+    T = result['T']
+    if vars_list[0] == 'T':
+        T = envert(xpoints, 'K')
+    P = result['P']
+    if vars_list[0] == 'P':
+        P = envert(xpoints, 'bar')
+
+    # Handle the case where P is "Psat" - keep it as is for subcrt
+    # (subcrt knows how to handle "Psat")
+
+    # logaH2O is 0 unless given in result['basis']
+    basis_df = result['basis']
+    if 'H2O' in basis_df.index:
+        logaH2O = float(basis_df.loc['H2O', 'logact'])
+    else:
+        logaH2O = 0
+
+    # pH is 7 unless given in eout['basis'] or plotted on one of the axes
+    if vars_list[0] == 'pH':
+        pH = xpoints
+    elif 'H+' in basis_df.index:
+        minuspH = basis_df.loc['H+', 'logact']
+        # Special treatment for non-numeric value (happens when a buffer is used)
+        try:
+            pH = -float(minuspH)
+        except (ValueError, TypeError):
+            pH = np.nan
+    else:
+        pH = 7
+
+    # O2 state is gas unless given in eout['basis']
+    O2state = 'gas'
+    if 'O2' in basis_df.index:
+        O2state = basis_df.loc['O2', 'state']
+
+    # H2 state is gas unless given in eout['basis']
+    H2state = 'gas'
+    if 'H2' in basis_df.index:
+        H2state = basis_df.loc['H2', 'state']
+
+    # Where the calculated values will go
+    y_oxidation = None
+    y_reduction = None
+
+    if xaxis in ['pH', 'T', 'P', 'nothing'] and yaxis in ['Eh', 'pe', 'O2', 'H2']:
+        # Eh/pe/logfO2/logaO2/logfH2/logaH2 vs pH/T/P
+
+        # Reduction line (H2O + e- = 1/2 H2 + OH-)
+        if 'reduction' in which:
+            logfH2 = logaH2O  # usually 0
+
+            if yaxis == 'H2':
+                # Calculate equilibrium constant for gas-aqueous conversion if needed
+                logK = subcrt(['H2', 'H2'], [-1, 1], ['gas', H2state], T=T, P=P, convert=False, messages=messages, show=False).out['logK']
+                # This is logfH2 if H2state == "gas", or logaH2 if H2state == "aq"
+                logfH2 = logfH2 + logK
+                # Broadcast to match xpoints length
+                if isinstance(logfH2, (int, float)):
+                    y_reduction = np.full_like(xpoints, logfH2)
+                else:
+                    logfH2_val = float(logfH2.iloc[0]) if hasattr(logfH2, 'iloc') else float(logfH2[0])
+                    y_reduction = np.full_like(xpoints, logfH2_val)
+            else:
+                # Calculate logfO2 from H2O = 1/2 O2 + H2
+                logK = subcrt(['H2O', 'O2', 'H2'], [-1, 0.5, 1], ['liq', O2state, 'gas'], T=T, P=P, convert=False, messages=messages, show=False).out['logK']
+                # This is logfO2 if O2state == "gas", or logaO2 if O2state == "aq"
+                logfO2 = 2 * (logK - logfH2 + logaH2O)
+
+                if yaxis == 'O2':
+                    # Broadcast to match xpoints length
+                    if isinstance(logfO2, (int, float)):
+                        y_reduction = np.full_like(xpoints, logfO2)
+                    else:
+                        logfO2_val = float(logfO2.iloc[0]) if hasattr(logfO2, 'iloc') else float(logfO2[0])
+                        y_reduction = np.full_like(xpoints, logfO2_val)
+                elif yaxis == 'Eh':
+                    y_reduction = convert(logfO2, 'E0', T=T, P=P, pH=pH, logaH2O=logaH2O, messages=messages)
+                elif yaxis == 'pe':
+                    Eh_val = convert(logfO2, 'E0', T=T, P=P, pH=pH, logaH2O=logaH2O, messages=messages)
+                    y_reduction = convert(Eh_val, 'pe', T=T, messages=messages)
+
+        # Oxidation line (H2O = 1/2 O2 + 2H+ + 2e-)
+        if 'oxidation' in which:
+            logfO2 = logaH2O  # usually 0
+
+            if yaxis == 'H2':
+                # Calculate logfH2 from H2O = 1/2 O2 + H2
+                logK = subcrt(['H2O', 'O2', 'H2'], [-1, 0.5, 1], ['liq', 'gas', H2state], T=T, P=P, convert=False, messages=messages, show=False).out['logK']
+                # This is logfH2 if H2state == "gas", or logaH2 if H2state == "aq"
+                logfH2 = logK - 0.5*logfO2 + logaH2O
+                # Broadcast to match xpoints length
+                if isinstance(logfH2, (int, float)):
+                    y_oxidation = np.full_like(xpoints, logfH2)
+                else:
+                    logfH2_val = float(logfH2.iloc[0]) if hasattr(logfH2, 'iloc') else float(logfH2[0])
+                    y_oxidation = np.full_like(xpoints, logfH2_val)
+            else:
+                # Calculate equilibrium constant for gas-aqueous conversion if needed
+                logK = subcrt(['O2', 'O2'], [-1, 1], ['gas', O2state], T=T, P=P, convert=False, messages=messages, show=False).out['logK']
+                # This is logfO2 if O2state == "gas", or logaO2 if O2state == "aq"
+                logfO2 = logfO2 + logK
+
+                if yaxis == 'O2':
+                    # Broadcast to match xpoints length
+                    if isinstance(logfO2, (int, float)):
+                        y_oxidation = np.full_like(xpoints, logfO2)
+                    else:
+                        logfO2_val = float(logfO2.iloc[0]) if hasattr(logfO2, 'iloc') else float(logfO2[0])
+                        y_oxidation = np.full_like(xpoints, logfO2_val)
+                elif yaxis == 'Eh':
+                    y_oxidation = convert(logfO2, 'E0', T=T, P=P, pH=pH, logaH2O=logaH2O, messages=messages)
+                elif yaxis == 'pe':
+                    Eh_val = convert(logfO2, 'E0', T=T, P=P, pH=pH, logaH2O=logaH2O, messages=messages)
+                    y_oxidation = convert(Eh_val, 'pe', T=T, messages=messages)
+
+    else:
+        # Invalid axis combination
+        result.update({'xpoints': xpoints, 'y_oxidation': None, 'y_reduction': None, 'swapped': swapped})
+        return result
+
+    # Route to Plotly or matplotlib implementation
+    if is_plotly:
+        return _water_lines_plotly(result, xpoints, y_oxidation, y_reduction, swapped,
+                                  lty, lwd, col, plot_it)
+
+    # Matplotlib implementation
+    # Only draw water lines if eout already has an axes (meaning it's from a diagram)
+    # If no axes, this is being called just for calculation (e.g., from within diagram())
+    if 'ax' not in eout or eout['ax'] is None:
+        # No axes to plot on - just return the calculated values
+        result.update({'xpoints': xpoints, 'y_oxidation': y_oxidation, 'y_reduction': y_reduction, 'swapped': swapped})
+        return result
+
+    # Use the axes from result
+    ax = result['ax']
+
+    # First, shade the water-unstable regions with gray
+    # This creates the same effect as R's fill.NA for H2O.predominant
+    if y_oxidation is not None and y_reduction is not None:
+        from matplotlib.colors import ListedColormap
+
+        # Get current axis limits to create shading
+        xlim = ax.get_xlim()
+        ylim = ax.get_ylim()
+
+        # Create a high-resolution mesh for smooth shading
+        n_points = 500
+        if swapped:
+            # When swapped, xpoints is on the y-axis
+            y_mesh = np.linspace(ylim[0], ylim[1], n_points)
+            x_mesh = np.linspace(xlim[0], xlim[1], n_points)
+            X, Y = np.meshgrid(x_mesh, y_mesh)
+
+            # For each y-value, determine if it's in water-unstable region
+            # Interpolate oxidation and reduction values to the mesh
+            y_ox_interp = np.interp(y_mesh, xpoints, y_oxidation)
+            y_red_interp = np.interp(y_mesh, xpoints, y_reduction)
+
+            # Create mask: unstable where x < min or x > max
+            unstable = np.zeros_like(X, dtype=bool)
+            for i in range(n_points):
+                ymin = min(y_ox_interp[i], y_red_interp[i])
+                ymax = max(y_ox_interp[i], y_red_interp[i])
+                unstable[i, :] = (X[i, :] < ymin) | (X[i, :] > ymax)
+        else:
+            # Normal: xpoints on x-axis, y values on y-axis
+            x_mesh = np.linspace(xlim[0], xlim[1], n_points)
+            y_mesh = np.linspace(ylim[0], ylim[1], n_points)
+            X, Y = np.meshgrid(x_mesh, y_mesh)
+
+            # Interpolate oxidation and reduction values to the mesh
+            y_ox_interp = np.interp(x_mesh, xpoints, y_oxidation)
+            y_red_interp = np.interp(x_mesh, xpoints, y_reduction)
+
+            # Create mask: unstable where y < min or y > max
+            unstable = np.zeros_like(Y, dtype=bool)
+            for i in range(n_points):
+                ymin = min(y_ox_interp[i], y_red_interp[i])
+                ymax = max(y_ox_interp[i], y_red_interp[i])
+                unstable[:, i] = (Y[:, i] < ymin) | (Y[:, i] > ymax)
+
+        # Create masked array for unstable regions
+        import numpy.ma as ma
+        unstable_mask = ma.masked_where(~unstable, np.ones_like(X))
+
+        # Draw the shading with gray (matching R's gray80 = 0.8)
+        fill_na_cmap = ListedColormap(['0.8'])
+        extent = [xlim[0], xlim[1], ylim[0], ylim[1]]
+        ax.imshow(unstable_mask, aspect='auto', origin='lower',
+                 extent=extent, interpolation='nearest',
+                 cmap=fill_na_cmap, vmin=0, vmax=1, zorder=1)
+
+    # Set line color
+    if col is None:
+        col = 'black'
+
+    # Convert numeric line style to matplotlib style
+    lty_map = {1: '-', 2: '--', 3: '-.', 4: ':', 5: '-', 6: '--'}
+    if isinstance(lty, int):
+        lty = lty_map.get(lty, '--')
+
+    if swapped:
+        if nvar1 == 1 or nvar2 == 2:
+            # Add vertical lines on 1-D diagram
+            if y_oxidation is not None and len(y_oxidation) > 0:
+                ax.axvline(x=y_oxidation[0], linestyle=lty, linewidth=lwd, color=col)
+            if y_reduction is not None and len(y_reduction) > 0:
+                ax.axvline(x=y_reduction[0], linestyle=lty, linewidth=lwd, color=col)
+        else:
+            # xpoints above is really the ypoints
+            if y_oxidation is not None:
+                ax.plot(y_oxidation, xpoints, linestyle=lty, linewidth=lwd, color=col)
+            if y_reduction is not None:
+                ax.plot(y_reduction, xpoints, linestyle=lty, linewidth=lwd, color=col)
+    else:
+        if y_oxidation is not None:
+            ax.plot(xpoints, y_oxidation, linestyle=lty, linewidth=lwd, color=col)
+        if y_reduction is not None:
+            ax.plot(xpoints, y_reduction, linestyle=lty, linewidth=lwd, color=col)
+
+    # Update the figure and axes references in result to reflect the water lines
+    fig = ax.get_figure()
+    result['fig'] = fig
+    result['ax'] = ax
+
+    # Display the figure if plot_it=True
+    # This allows water_lines() to display a figure that was created with plot_it=False
+    if plot_it and fig is not None:
+        try:
+            from IPython.display import display
+            display(fig)
+        except (ImportError, NameError):
+            # Not in IPython/Jupyter, matplotlib will handle display
+            pass
+
+    # Update result with water line data and return
+    result.update({'xpoints': xpoints, 'y_oxidation': y_oxidation, 'y_reduction': y_reduction, 'swapped': swapped})
+    return result
+
+
+def _water_lines_plotly(eout: Dict[str, Any],
+                        xpoints: np.ndarray,
+                        y_oxidation: Optional[np.ndarray],
+                        y_reduction: Optional[np.ndarray],
+                        swapped: bool,
+                        lty: Union[int, str],
+                        lwd: float,
+                        col: Optional[str],
+                        plot_it: bool) -> Dict[str, Any]:
+    """
+    Add water stability lines to a Plotly interactive diagram.
+
+    This helper function adds dashed lines showing water oxidation and reduction
+    stability limits, plus gray shading for water-unstable regions, to an
+    interactive Plotly diagram.
+    """
+    import plotly.graph_objects as go
+    import numpy as np
+
+    # Get the Plotly figure from eout
+    fig = eout['fig']
+
+    # Set line color (default to black)
+    if col is None:
+        col = 'black'
+
+    # Convert numeric/matplotlib line styles to Plotly dash types
+    lty_map = {
+        1: 'solid', '-': 'solid',
+        2: 'dash', '--': 'dash',
+        3: 'dashdot', '-.': 'dashdot',
+        4: 'dot', ':': 'dot',
+        5: 'solid', 6: 'dash'
+    }
+    dash_style = lty_map.get(lty, 'dash') if (isinstance(lty, int) or lty in lty_map) else 'dash'
+
+    # Get axis limits to determine shading extent
+    # We need to extract axis ranges from the existing figure
+    if fig.layout.xaxis.range:
+        xlim = fig.layout.xaxis.range
+    else:
+        # If not set, estimate from data
+        xlim = [xpoints.min(), xpoints.max()]
+
+    if fig.layout.yaxis.range:
+        ylim = fig.layout.yaxis.range
+    else:
+        # If not set, estimate from y_oxidation and y_reduction
+        if y_oxidation is not None and y_reduction is not None:
+            ylim = [min(y_oxidation.min(), y_reduction.min()),
+                   max(y_oxidation.max(), y_reduction.max())]
+        else:
+            ylim = [0, 1]  # Fallback
+
+    # Add gray shading for water-unstable regions
+    if y_oxidation is not None and y_reduction is not None:
+        # Create high-resolution mesh for smooth shading
+        n_points = 200
+
+        if swapped:
+            # When swapped, xpoints is on y-axis, y values on x-axis
+            # Create shading shapes for regions outside water stability
+
+            # Upper unstable region (above oxidation line)
+            y_mesh = np.linspace(ylim[0], ylim[1], n_points)
+            x_ox_interp = np.interp(y_mesh, xpoints, y_oxidation)
+
+            # Fill from oxidation line to right edge
+            fig.add_trace(go.Scatter(
+                x=np.concatenate([x_ox_interp, [xlim[1]] * len(y_mesh), x_ox_interp[::-1]]),
+                y=np.concatenate([y_mesh, y_mesh[::-1], y_mesh[::-1]]),
+                fill='toself',
+                fillcolor='rgba(128, 128, 128, 0.5)',  # Gray with transparency
+                line=dict(width=0),
+                showlegend=False,
+                hoverinfo='skip'
+            ))
+
+            # Lower unstable region (below reduction line)
+            x_red_interp = np.interp(y_mesh, xpoints, y_reduction)
+
+            # Fill from left edge to reduction line
+            fig.add_trace(go.Scatter(
+                x=np.concatenate([[xlim[0]] * len(y_mesh), x_red_interp[::-1], [xlim[0]] * len(y_mesh)]),
+                y=np.concatenate([y_mesh, y_mesh[::-1], y_mesh[::-1]]),
+                fill='toself',
+                fillcolor='rgba(128, 128, 128, 0.5)',
+                line=dict(width=0),
+                showlegend=False,
+                hoverinfo='skip'
+            ))
+
+        else:
+            # Normal: xpoints on x-axis, y values on y-axis
+            # Interpolate to create smooth shading boundaries
+            x_mesh = np.linspace(xlim[0], xlim[1], n_points)
+            y_ox_interp = np.interp(x_mesh, xpoints, y_oxidation)
+            y_red_interp = np.interp(x_mesh, xpoints, y_reduction)
+
+            # Upper unstable region (above oxidation line)
+            fig.add_trace(go.Scatter(
+                x=np.concatenate([x_mesh, x_mesh[::-1]]),
+                y=np.concatenate([y_ox_interp, [ylim[1]] * len(x_mesh)]),
+                fill='toself',
+                fillcolor='rgba(128, 128, 128, 0.5)',
+                line=dict(width=0),
+                showlegend=False,
+                hoverinfo='skip'
+            ))
+
+            # Lower unstable region (below reduction line)
+            fig.add_trace(go.Scatter(
+                x=np.concatenate([x_mesh, x_mesh[::-1]]),
+                y=np.concatenate([[ylim[0]] * len(x_mesh), y_red_interp[::-1]]),
+                fill='toself',
+                fillcolor='rgba(128, 128, 128, 0.5)',
+                line=dict(width=0),
+                showlegend=False,
+                hoverinfo='skip'
+            ))
+
+    # Add water stability lines
+    if swapped:
+        # When swapped: xpoints is on y-axis, y values on x-axis
+        if y_oxidation is not None:
+            fig.add_trace(go.Scatter(
+                x=y_oxidation,
+                y=xpoints,
+                mode='lines',
+                line=dict(color=col, width=lwd, dash=dash_style),
+                name='H₂O oxidation limit',
+                showlegend=False,
+                hoverinfo='skip'
+            ))
+        if y_reduction is not None:
+            fig.add_trace(go.Scatter(
+                x=y_reduction,
+                y=xpoints,
+                mode='lines',
+                line=dict(color=col, width=lwd, dash=dash_style),
+                name='H₂O reduction limit',
+                showlegend=False,
+                hoverinfo='skip'
+            ))
+    else:
+        # Normal orientation
+        if y_oxidation is not None:
+            fig.add_trace(go.Scatter(
+                x=xpoints,
+                y=y_oxidation,
+                mode='lines',
+                line=dict(color=col, width=lwd, dash=dash_style),
+                name='H₂O oxidation limit',
+                showlegend=False,
+                hoverinfo='skip'
+            ))
+        if y_reduction is not None:
+            fig.add_trace(go.Scatter(
+                x=xpoints,
+                y=y_reduction,
+                mode='lines',
+                line=dict(color=col, width=lwd, dash=dash_style),
+                name='H₂O reduction limit',
+                showlegend=False,
+                hoverinfo='skip'
+            ))
+
+    # Update the figure reference in eout to reflect the water lines
+    eout['fig'] = fig
+
+    # Display the figure if plot_it=True
+    if plot_it:
+        try:
+            from IPython.display import display
+            display(fig)
+        except (ImportError, NameError):
+            # Not in IPython/Jupyter, use fig.show()
+            fig.show()
+
+    # Update eout with water line data and return
+    eout.update({'xpoints': xpoints, 'y_oxidation': y_oxidation, 'y_reduction': y_reduction, 'swapped': swapped})
+    return eout
+
+
+def find_tp(predominant: np.ndarray) -> np.ndarray:
+    """
+    Find triple points in a predominance diagram.
+
+    This function identifies the approximate positions of triple points
+    (where three phases meet) in a 2-D predominance diagram by locating
+    cells with the greatest number of different neighboring values.
+
+    Parameters
+    ----------
+    predominant : np.ndarray
+        Matrix of integers from diagram() output indicating which species
+        predominates at each point. Should be a 2-D array where each value
+        represents a different species/phase.
+
+    Returns
+    -------
+    np.ndarray
+        Array of shape (n, 2) where n is the number of triple points found.
+        Each row contains [row_index, col_index] of a triple point location.
+        Indices are 1-based to match R behavior.
+
+    Examples
+    --------
+    >>> from pychnosz import *
+    >>> reset()
+    >>> basis(["corundum", "quartz", "oxygen"])
+    >>> species(["kyanite", "sillimanite", "andalusite"])
+    >>> a = affinity(T=[200, 900, 99], P=[0, 9000, 101], exceed_Ttr=True)
+    >>> d = diagram(a)
+    >>> tp = find_tp(d['predominant'])
+    >>> # Get T and P at the triple point
+    >>> Ttp = a['vals'][0][tp[0, 1] - 1]  # -1 for 0-based indexing
+    >>> Ptp = a['vals'][1][::-1][tp[0, 0] - 1]  # reversed and -1
+
+    Notes
+    -----
+    This is a Python translation of the R function find.tp() from CHNOSZ.
+    The R version returns 1-based indices, and this Python version does too
+    for consistency. When using these indices to access Python arrays,
+    remember to subtract 1.
+
+    The function works by:
+    1. Rearranging the matrix as done by diagram() for plotting
+    2. For each position, examining a 3x3 neighborhood
+    3. Counting the number of unique values in that neighborhood
+    4. Returning positions with the maximum count (typically 3 or more)
+    """
+    # Rearrange the matrix in the same way that diagram() does for 2-D predominance diagrams
+    # R code: x <- t(x[, ncol(x):1])
+    # This means: first reverse columns, then transpose
+    x = np.transpose(predominant[:, ::-1])
+
+    # Get all positions with valid values (> 0)
+    valid_positions = np.argwhere(x > 0)
+
+    if len(valid_positions) == 0:
+        return np.array([])
+
+    # For each position, count unique values in 3x3 neighborhood
+    counts = []
+    for pos in valid_positions:
+        row, col = pos
+
+        # Define the range to look at (3x3 except at edges)
+        r1 = max(row - 1, 0)
+        r2 = min(row + 1, x.shape[0] - 1)
+        c1 = max(col - 1, 0)
+        c2 = min(col + 1, x.shape[1] - 1)
+
+        # Extract the neighborhood
+        neighborhood = x[r1:r2+1, c1:c2+1]
+
+        # Count unique values
+        n_unique = len(np.unique(neighborhood))
+        counts.append(n_unique)
+
+    counts = np.array(counts)
+
+    # Find positions with the maximum count
+    max_count = np.max(counts)
+    max_positions = valid_positions[counts == max_count]
+
+    # Convert to 1-based indexing (to match R)
+    # Return as [row, col] with 1-based indices
+    result = max_positions + 1
+
+    return result
+
+
+def diagram_interactive(eout: Dict[str, Any],
+                        type: str = "auto",
+                        main: Optional[str] = None,
+                        borders: Union[float, str] = 0,
+                        names: Optional[List[str]] = None,
+                        format_names: bool = True,
+                        annotation: Optional[str] = None,
+                        annotation_coords: List[float] = [0, 0],
+                        balance: Optional[Union[str, float, List[float]]] = None,
+                        xlab: Optional[str] = None,
+                        ylab: Optional[str] = None,
+                        fill: Optional[Union[str, List[str]]] = "viridis",
+                        width: int = 600,
+                        height: int = 520,
+                        alpha: Union[bool, str] = False,
+                        add: bool = False,
+                        ax: Optional[Any] = None,
+                        col: Optional[Union[str, List[str]]] = None,
+                        lty: Optional[Union[str, int, List]] = None,
+                        lwd: Union[float, List[float]] = 1,
+                        cex: Union[float, List[float]] = 1.0,
+                        contour_method: Optional[Union[str, List[str]]] = "edge",
+                        messages: bool = True,
+                        plot_it: bool = True,
+                        save_as: Optional[str] = None,
+                        save_format: Optional[str] = None,
+                        save_scale: float = 1) -> Tuple[pd.DataFrame, Any]:
+    """
+    Create an interactive diagram using Plotly.
+
+    This function produces interactive versions of the diagrams created by diagram(),
+    using Plotly for interactivity. It accepts output from affinity() or equilibrate()
+    and creates either 1D line plots or 2D predominance diagrams.
+
+    Parameters
+    ----------
+    eout : dict
+        Output from affinity() or equilibrate().
+    main : str, optional
+        Title of the plot.
+    borders : float or str, default 0
+        Controls boundary lines between regions in 2D predominance diagrams.
+        - If numeric > 0: draws grid-aligned borders with specified thickness (pixels)
+        - If "contour": draws smooth contour-based boundaries (like diagram())
+        - If 0 or None: no borders drawn
+    names : list of str, optional
+        Names of species for activity lines or predominance fields.
+    format_names : bool, default True
+        Apply formatting to chemical formulas?
+    annotation : str, optional
+        Annotation to add to the plot.
+    annotation_coords : list of float, default [0, 0]
+        Coordinates of annotation, where 0,0 is bottom left and 1,1 is top right.
+    balance : str or numeric, optional
+        How to balance the transformations.
+    xlab : str, optional
+        Custom x-axis label.
+    ylab : str, optional
+        Custom y-axis label.
+    fill : str or list of str, default "viridis"
+        For 2D diagrams: colormap name (e.g., "viridis", "hot") or list of colors.
+        For 1D diagrams: list of line colors.
+    width : int, default 600
+        Width of the plot in pixels.
+    height : int, default 520
+        Height of the plot in pixels.
+    alpha : bool or str, default False
+        For speciation diagrams, plot degree of formation instead of activities?
+        If True, plots mole fractions. If "balance", scales by stoichiometry.
+    messages : bool, default True
+        Display messages?
+    plot_it : bool, default True
+        Show the plot?
+    save_as : str, optional
+        Provide a filename to save this figure. Filetype of saved figure is
+        determined by save_format.
+    save_format : str, default "png"
+        Desired format of saved or downloaded figure. Can be 'png', 'jpg', 'jpeg',
+        'webp', 'svg', 'pdf', 'eps', 'json', or 'html'. If 'html', an interactive
+        plot will be saved.
+    save_scale : float, default 1
+        Multiply title/legend/axis/canvas sizes by this factor when saving.
+
+    Returns
+    -------
+    tuple
+        (df, fig) where df is a pandas DataFrame with the data and fig is the
+        Plotly figure object.
+
+    Examples
+    --------
+    1D diagram:
+    >>> basis("CHNOS+")
+    >>> species(info(["glycinium", "glycine", "glycinate"]))
+    >>> a = affinity(pH=[0, 14])
+    >>> e = equilibrate(a)
+    >>> diagram_interactive(e, alpha=True)
+
+    2D diagram:
+    >>> basis(["Fe", "oxygen", "S2"])
+    >>> species(["iron", "ferrous-oxide", "magnetite", "hematite", "pyrite", "pyrrhotite"])
+    >>> a = affinity(S2=[-50, 0], O2=[-90, -10], T=200)
+    >>> diagram_interactive(a, fill="hot")
+
+    Notes
+    -----
+    This function requires plotly to be installed. Install with:
+        pip install plotly
+
+    The function adapts the pyCHNOSZ diagram_interactive() implementation
+    to work with Python CHNOSZ's native data structures.
+    """
+
+    # Import plotly (lazy import to avoid dependency issues)
+    try:
+        import plotly.express as px
+        import plotly.graph_objects as go
+        import plotly.io as pio
+    except ImportError:
+        raise ImportError("diagram_interactive() requires plotly. Install with: pip install plotly")
+
+    # Check that eout is valid
+    efun = eout.get('fun', '')
+    if efun not in ['affinity', 'equilibrate', 'solubility']:
+        raise ValueError("'eout' is not the output from affinity(), equilibrate(), or solubility()")
+
+    # Determine if this is affinity or equilibrate output
+    calc_type = "a" if ('loga_equil' not in eout and 'loga.equil' not in eout) else "e"
+
+    # Get basis species and their states
+    basis_df = eout['basis']
+    basis_sp = list(basis_df.index)
+    basis_state = list(basis_df['state'])
+
+    # Get variable names and values
+    xyvars = eout['vars']
+    xyvals_dict = eout['vals']
+    # Convert vals dict to list format for easier access
+    xyvals = [xyvals_dict[var] for var in xyvars]
+
+    # Determine balance if not provided
+    if balance is None or balance == "":
+        # For saturation diagrams, use balance=1 (formula units) to match R behavior
+        # This avoids issues when minerals don't have a common basis element
+        if type == "saturation":
+            balance = 1
+            n_balance = [1] * len(eout['values'])
+        else:
+            # Call diagram with plot_it=False to get balance
+            # Need to import matplotlib to close the figure afterward
+            import matplotlib.pyplot as plt_temp
+            temp_result = diagram(eout, messages=False, plot_it=False)
+            balance = temp_result.get('balance', 1)
+            n_balance = temp_result.get('n_balance', [1])
+            # Close the matplotlib figure created by diagram() since we don't need it
+            if 'fig' in temp_result and temp_result['fig'] is not None:
+                plt_temp.close(temp_result['fig'])
+    else:
+        # Calculate n_balance from balance
+        try:
+            balance_float = float(balance)
+            n_balance = [balance_float] * len(eout['values'])
+        except (ValueError, TypeError):
+            # balance is a string (element name)
+            # Get species from eout instead of global state
+            if 'species' in eout and eout['species'] is not None:
+                sp_df = eout['species']
+            else:
+                # Fallback to global species if not in eout
+                from .species import species as species_func
+                sp_df = species_func()
+
+            # Check if balance is a list (user-provided values) or a string (column name)
+            if isinstance(balance, list):
+                n_balance = balance
+            elif balance in sp_df.columns:
+                n_balance = list(sp_df[balance])
+            else:
+                n_balance = [1] * len(eout['values'])
+
+    # Get output values
+    if calc_type == "a":
+        # handling output of affinity()
+        out_vals = eout['values']
+        out_units = "A/(2.303RT)"
+    else:
+        # handling output of equilibrate()
+        loga_equil_key = 'loga_equil' if 'loga_equil' in eout else 'loga.equil'
+        out_vals = eout[loga_equil_key]
+        out_units = "log a"
+
+    # Convert values to a list format
+    if isinstance(out_vals, dict):
+        nsp = len(out_vals)
+        values_list = list(out_vals.values())
+        species_indices = list(out_vals.keys())
+    else:
+        nsp = len(out_vals)
+        values_list = out_vals
+        species_indices = eout['species']['ispecies'].tolist()
+
+    # Get species names
+    from .info import info as info_func
+    # Convert numpy types to Python types
+    species_indices_py = [int(idx) for idx in species_indices]
+    sp_info = info_func(species_indices_py, messages=False)
+    sp_names = sp_info['name'].tolist()
+
+    # Use custom names if provided
+    if isinstance(names, list) and len(names) == len(sp_names):
+        sp_names = names
+
+    # Determine dimensions
+    first_val = values_list[0]
+    if hasattr(first_val, 'shape'):
+        nd = len(first_val.shape)
+    else:
+        nd = 1 if hasattr(first_val, '__len__') else 0
+
+    # Handle type="saturation" - plot contour lines where affinity=0
+    if type == "saturation":
+        if nd != 2:
+            raise ValueError("type='saturation' requires 2-D diagram")
+        if calc_type != "a":
+            raise ValueError("type='saturation' requires output from affinity(), not equilibrate()")
+
+        # Delegate to saturation plotting function
+        return _plot_saturation_interactive(
+            eout, values_list, sp_names, xyvars, xyvals,
+            xlab, ylab, col, lwd, lty, cex, contour_method,
+            main, add, ax, width, height, plot_it,
+            save_as, save_format, save_scale, messages
+        )
+
+    # Build DataFrame
+    if nd == 2:
+        # 2D case - flatten the data
+        xvals = xyvals[0]
+        yvals = xyvals[1]
+        xvar = xyvars[0]
+        yvar = xyvars[1]
+
+        # Flatten the data - transpose first so coordinates match
+        # Original shape is (nx, ny) where nx=len(xvals), ny=len(yvals)
+        # After transpose, shape is (ny, nx)
+        # Flattening with C-order then gives: [row0, row1, ...] = [x-values at y[0], x-values at y[1], ...]
+        flat_out_vals = []
+        for v in values_list:
+            # Transpose then flatten so coordinates align correctly
+            flat_out_vals.append(v.T.flatten())
+        df = pd.DataFrame(flat_out_vals, index=sp_names).T
+
+        # Apply balance if needed
+        if calc_type == "a":
+            if isinstance(balance, str):
+                # Get balance from species dataframe
+                # Get species from eout instead of global state
+                if 'species' in eout and eout['species'] is not None:
+                    sp_df = eout['species']
+                else:
+                    # Fallback to global species if not in eout
+                    from .species import species as species_func
+                    sp_df = species_func()
+
+                # Check if balance is a list (user-provided values) or a string (column name)
+                if isinstance(balance, list):
+                    n_balance = balance
+                elif balance in sp_df.columns:
+                    n_balance = list(sp_df[balance])
+            # Divide by balance
+            for i, sp in enumerate(sp_names):
+                df[sp] = df[sp] / n_balance[i]
+
+        # Find predominant species
+        df["pred"] = df.idxmax(axis=1, skipna=True)
+        df["prednames"] = df["pred"]
+
+        # Add x and y coordinates
+        # After transpose and flatten, data is ordered as:
+        # [x0,y0], [x1,y0], ..., [xn,y0], [x0,y1], [x1,y1], ...
+        xvals_full = list(xvals) * len(yvals)
+        yvals_full = []
+        for y in yvals:
+            yvals_full.extend([y] * len(xvals))
+        df[xvar] = xvals_full
+        df[yvar] = yvals_full
+
+    else:
+        # 1D case
+        xvar = xyvars[0]
+        xvals = xyvals[0]
+
+        flat_out_vals = []
+        for v in values_list:
+            flat_out_vals.append(v)
+        df = pd.DataFrame(flat_out_vals, index=sp_names).T
+
+        # Apply balance if needed
+        if calc_type == "a":
+            if isinstance(balance, str):
+                # Get species from eout instead of global state
+                if 'species' in eout and eout['species'] is not None:
+                    sp_df = eout['species']
+                else:
+                    # Fallback to global species if not in eout
+                    from .species import species as species_func
+                    sp_df = species_func()
+
+                # Check if balance is a list (user-provided values) or a string (column name)
+                if isinstance(balance, list):
+                    n_balance = balance
+                elif balance in sp_df.columns:
+                    n_balance = list(sp_df[balance])
+            # Divide by balance
+            for i, sp in enumerate(sp_names):
+                df[sp] = df[sp] / n_balance[i]
+
+        # Handle alpha (degree of formation)
+        if alpha:
+            df = df.apply(lambda x: 10**x)
+            df = df[sp_names].div(df[sp_names].sum(axis=1), axis=0)
+
+        df[xvar] = xvals
+
+    # Create axis labels
+    unit_dict = {"P": "bar", "T": "°C", "pH": "", "Eh": "volts", "IS": "mol/kg"}
+
+    for i, s in enumerate(basis_sp):
+        if basis_state[i] in ["aq", "liq", "cr"]:
+            if format_names:
+                unit_dict[s] = f"log <i>a</i><sub>{_format_html_species(s)}</sub>"
+            else:
+                unit_dict[s] = f"log <i>a</i><sub>{s}</sub>"
+        else:
+            if format_names:
+                unit_dict[s] = f"log <i>f</i><sub>{_format_html_species(s)}</sub>"
+            else:
+                unit_dict[s] = f"log <i>f</i><sub>{s}</sub>"
+
+    # Set x-axis label
+    if not isinstance(xlab, str):
+        xlab = xvar + ", " + unit_dict.get(xvar, "")
+        if xvar == "pH":
+            xlab = "pH"
+        if xvar in basis_sp:
+            xlab = unit_dict[xvar]
+
+    # Create the plot
+    if nd == 1:
+        # 1D plot
+        # Melt the dataframe for plotting
+        df_melted = pd.melt(df, id_vars=[xvar], value_vars=sp_names, var_name='variable', value_name='value')
+
+        # Format species names if requested
+        if format_names:
+            df_melted['variable'] = df_melted['variable'].apply(_format_html_species)
+
+        # Set y-axis label
+        if not isinstance(ylab, str):
+            if alpha:
+                ylab = "alpha"
+            else:
+                ylab = out_units
+
+        fig = px.line(df_melted, x=xvar, y="value", color='variable',
+                      template="simple_white", width=width, height=height,
+                      labels={'value': ylab, xvar: xlab},
+                      render_mode='svg')
+
+        # Apply custom colors if provided
+        if isinstance(fill, list):
+            for i, color in enumerate(fill):
+                if i < len(fig.data):
+                    fig.data[i].line.color = color
+
+        # Check for LaTeX format in axis labels
+        if xlab and _detect_latex_format(xlab):
+            warnings.warn(
+                "LaTeX formatting detected in 'xlab' parameter. "
+                "Plotly requires HTML format (<sub>, <sup>) instead of LaTeX ($, _, ^). "
+                "For activity ratios, use ratlab_html() instead of ratlab().",
+                UserWarning
+            )
+        if ylab and _detect_latex_format(ylab):
+            warnings.warn(
+                "LaTeX formatting detected in 'ylab' parameter. "
+                "Plotly requires HTML format (<sub>, <sup>) instead of LaTeX ($, _, ^). "
+                "For activity ratios, use ratlab_html() instead of ratlab().",
+                UserWarning
+            )
+
+        fig.update_layout(xaxis_title=xlab,
+                          yaxis_title=ylab,
+                          legend_title=None)
+
+        if isinstance(main, str):
+            fig.update_layout(title={'text': main, 'x': 0.5, 'xanchor': 'center'})
+
+        if isinstance(annotation, str):
+            # Check for LaTeX format and warn user
+            if _detect_latex_format(annotation):
+                warnings.warn(
+                    "LaTeX formatting detected in 'annotation' parameter. "
+                    "Plotly requires HTML format (<sub>, <sup>) instead of LaTeX ($, _, ^). "
+                    "For activity ratios, use ratlab_html() instead of ratlab().",
+                    UserWarning
+                )
+
+            fig.add_annotation(
+                x=annotation_coords[0],
+                y=annotation_coords[1],
+                text=annotation,
+                showarrow=False,
+                xref="paper",
+                yref="paper",
+                align='left',
+                bgcolor="rgba(255, 255, 255, 0.5)")
+
+        # Configure download button
+        save_as_name, save_format_final = _save_figure(fig, save_as, save_format, save_scale,
+                                                        plot_width=width, plot_height=height, ppi=1)
+
+        config = {'displaylogo': False,
+                  'modeBarButtonsToRemove': ['resetScale2d', 'toggleSpikelines'],
+                  'toImageButtonOptions': {
+                      'format': save_format_final,
+                      'filename': save_as_name,
+                      'height': height,
+                      'width': width,
+                      'scale': save_scale,
+                  }}
+
+        # Store config on figure so it persists when fig.show() is called later
+        fig._config = fig._config | config
+
+    else:
+        # 2D plot
+        # Map species names to numeric values
+        mappings = {s: lab for s, lab in zip(sp_names, range(len(sp_names)))}
+        df['pred'] = df['pred'].map(mappings).astype(int)
+
+        # Reshape data
+        # Data is flattened as [x0,y0], [x1,y0], ..., [xn,y0], [x0,y1], ...
+        # Reshape to (ny, nx) for proper orientation in Plotly
+        # Plotly expects data[i,j] to correspond to x[j], y[i]
+        data = np.array(df['pred'])
+        shape = (len(yvals), len(xvals))
+        dmap = data.reshape(shape)
+
+        data_names = np.array(df['prednames'])
+        dmap_names = data_names.reshape(shape)
+
+        # Set y-axis label
+        if not isinstance(ylab, str):
+            ylab = yvar + ", " + unit_dict.get(yvar, "")
+            if yvar in basis_sp:
+                ylab = unit_dict[yvar]
+            if yvar == "pH":
+                ylab = "pH"
+
+        # Check for LaTeX format in axis labels (2D plot)
+        if xlab and _detect_latex_format(xlab):
+            warnings.warn(
+                "LaTeX formatting detected in 'xlab' parameter. "
+                "Plotly requires HTML format (<sub>, <sup>) instead of LaTeX ($, _, ^). "
+                "For activity ratios, use ratlab_html() instead of ratlab().",
+                UserWarning
+            )
+        if ylab and _detect_latex_format(ylab):
+            warnings.warn(
+                "LaTeX formatting detected in 'ylab' parameter. "
+                "Plotly requires HTML format (<sub>, <sup>) instead of LaTeX ($, _, ^). "
+                "For activity ratios, use ratlab_html() instead of ratlab().",
+                UserWarning
+            )
+
+        # Create heatmap
+        fig = px.imshow(dmap, width=width, height=height, aspect="auto",
+                        labels={'x': xlab, 'y': ylab, 'color': "region"},
+                        x=xvals, y=yvals, template="simple_white")
+
+        fig.update(data=[{'customdata': dmap_names,
+                          'hovertemplate': xlab + ': %{x}<br>' + ylab + ': %{y}<br>Region: %{customdata}<extra></extra>'}])
+
+        # Set colormap
+        if fill == 'none':
+            colormap = [[0, 'white'], [1, 'white']]
+        elif isinstance(fill, list):
+            colmap_temp = []
+            for i, v in enumerate(fill):
+                colmap_temp.append([i / (len(fill) - 1) if len(fill) > 1 else 0, v])
+            colormap = colmap_temp
+        else:
+            colormap = fill
+
+        fig.update_traces(dict(showscale=False,
+                               coloraxis=None,
+                               colorscale=colormap),
+                          selector={'type': 'heatmap'})
+
+        fig.update_yaxes(autorange=True)
+
+        if isinstance(main, str):
+            fig.update_layout(title={'text': main, 'x': 0.5, 'xanchor': 'center'})
+
+        # Add species labels
+        for s in sp_names:
+            if s in set(df["prednames"]):
+                df_s = df.loc[df["prednames"] == s]
+                namex = df_s[xvar].mean()
+                namey = df_s[yvar].mean()
+
+                if format_names:
+                    annot_text = _format_html_species(s)
+                else:
+                    annot_text = str(s)
+
+                fig.add_annotation(x=namex, y=namey,
+                                   text=annot_text,
+                                   bgcolor="rgba(255, 255, 255, 0.5)",
+                                   showarrow=False)
+
+        if isinstance(annotation, str):
+            # Check for LaTeX format and warn user
+            if _detect_latex_format(annotation):
+                warnings.warn(
+                    "LaTeX formatting detected in 'annotation' parameter. "
+                    "Plotly requires HTML format (<sub>, <sup>) instead of LaTeX ($, _, ^). "
+                    "For activity ratios, use ratlab_html() instead of ratlab().",
+                    UserWarning
+                )
+
+            fig.add_annotation(
+                x=annotation_coords[0],
+                y=annotation_coords[1],
+                text=annotation,
+                showarrow=False,
+                xref="paper",
+                yref="paper",
+                align='left',
+                bgcolor="rgba(255, 255, 255, 0.5)")
+
+        # Add borders if requested
+        if borders == "contour":
+            # Use contour-based boundaries (smooth, like diagram())
+            # Draw boundaries using matplotlib contour extraction without filling
+
+            # Get unique species (excluding any that don't appear)
+            unique_species_names = sorted(df["prednames"].unique())
+
+            # Create a temporary matplotlib figure to extract contour paths
+            # We won't display it, just use it to calculate contours
+            temp_fig, temp_ax = plt.subplots()
+
+            # For each species, create a binary mask and extract contours
+            for i, sp_name in enumerate(unique_species_names):
+                # Create binary mask: 1 where this species predominates, 0 elsewhere
+                z = (dmap_names == sp_name).astype(float)
+
+                # Create meshgrid for contour
+                X, Y = np.meshgrid(xvals, yvals)
+
+                # Find contours at level 0.5 using matplotlib
+                try:
+                    cs = temp_ax.contour(X, Y, z, levels=[0.5])
+
+                    # Extract the contour segments
+                    # cs.allsegs is a list of lists: [level][segment]
+                    for level_segs in cs.allsegs:
+                        for segment in level_segs:
+                            # segment is an (N, 2) array of (x, y) coordinates
+                            # Add as a scatter trace with lines
+                            fig.add_trace(
+                                go.Scatter(
+                                    x=segment[:, 0],
+                                    y=segment[:, 1],
+                                    mode='lines',
+                                    line=dict(color='black', width=2),
+                                    hoverinfo='skip',
+                                    showlegend=False
+                                )
+                            )
+
+                    # Clear the temp axes for next species
+                    temp_ax.clear()
+                except Exception as e:
+                    if messages:
+                        warnings.warn(f"Could not draw contour for {sp_name}: {e}")
+                    pass  # Skip if contour can't be drawn
+
+            # Close the temporary figure
+            plt.close(temp_fig)
+
+        elif isinstance(borders, (int, float)) and borders > 0:
+            unique_x_vals = sorted(list(set(df[xvar])))
+            unique_y_vals = sorted(list(set(df[yvar])))
+
+            # Skip border drawing if there are fewer than 2 unique values
+            # (single point or single line - no borders to draw between regions)
+            if len(unique_x_vals) < 2 or len(unique_y_vals) < 2:
+                if messages:
+                    warnings.warn("Skipping border drawing: need at least 2 unique values in each dimension")
+            else:
+                def mov_mean(numbers, window_size=2):
+                    moving_averages = []
+                    for i in range(len(numbers) - window_size + 1):
+                        window_average = sum(numbers[i:i + window_size]) / window_size
+                        moving_averages.append(window_average)
+                    return moving_averages
+
+                x_mov_mean = mov_mean(unique_x_vals)
+                y_mov_mean = mov_mean(unique_y_vals)
+
+                x_plot_min = x_mov_mean[0] - (x_mov_mean[1] - x_mov_mean[0])
+                y_plot_min = y_mov_mean[0] - (y_mov_mean[1] - y_mov_mean[0])
+
+                x_plot_max = x_mov_mean[-1] + (x_mov_mean[1] - x_mov_mean[0])
+                y_plot_max = y_mov_mean[-1] + (y_mov_mean[1] - y_mov_mean[0])
+
+                x_vals_border = [x_plot_min] + x_mov_mean + [x_plot_max]
+                y_vals_border = [y_plot_min] + y_mov_mean + [y_plot_max]
+
+                # Find border lines
+                def find_line(dmap, row_index):
+                    return [i for i in range(len(dmap[row_index]) - 1) if dmap[row_index][i] != dmap[row_index][i + 1]]
+
+                nrows, ncols = dmap.shape
+                vlines = [find_line(dmap, row_i) for row_i in range(nrows)]
+
+                dmap_transposed = dmap.transpose()
+                nrows_t, ncols_t = dmap_transposed.shape
+                hlines = [find_line(dmap_transposed, row_i) for row_i in range(nrows_t)]
+
+                y_coord_list_vertical = []
+                x_coord_list_vertical = []
+                for i, row in enumerate(vlines):
+                    for line in row:
+                        x_coord_list_vertical += [x_vals_border[line + 1], x_vals_border[line + 1], np.nan]
+                        y_coord_list_vertical += [y_vals_border[i], y_vals_border[i + 1], np.nan]
+
+                y_coord_list_horizontal = []
+                x_coord_list_horizontal = []
+                for i, col in enumerate(hlines):
+                    for line in col:
+                        y_coord_list_horizontal += [y_vals_border[line + 1], y_vals_border[line + 1], np.nan]
+                        x_coord_list_horizontal += [x_vals_border[i], x_vals_border[i + 1], np.nan]
+
+                fig.add_trace(
+                    go.Scatter(
+                        mode='lines',
+                        x=x_coord_list_horizontal,
+                        y=y_coord_list_horizontal,
+                        line={'width': borders, 'color': 'black'},
+                        hoverinfo='skip',
+                        showlegend=False))
+
+                fig.add_trace(
+                    go.Scatter(
+                        mode='lines',
+                        x=x_coord_list_vertical,
+                        y=y_coord_list_vertical,
+                        line={'width': borders, 'color': 'black'},
+                        hoverinfo='skip',
+                        showlegend=False))
+
+                fig.update_yaxes(range=[min(yvals), max(yvals)], autorange=False, mirror=True)
+                fig.update_xaxes(range=[min(xvals), max(xvals)], autorange=False, mirror=True)
+
+        # Configure download button
+        save_as_name, save_format_final = _save_figure(fig, save_as, save_format, save_scale,
+                                                        plot_width=width, plot_height=height, ppi=1)
+
+        config = {'displaylogo': False,
+                  'modeBarButtonsToRemove': ['zoom2d', 'pan2d', 'zoomIn2d', 'zoomOut2d',
+                                             'autoScale2d', 'resetScale2d', 'toggleSpikelines',
+                                             'hoverClosestCartesian', 'hoverCompareCartesian'],
+                  'toImageButtonOptions': {
+                      'format': save_format_final,
+                      'filename': save_as_name,
+                      'height': height,
+                      'width': width,
+                      'scale': save_scale,
+                  }}
+
+        # Store config on figure so it persists when fig.show() is called later
+        fig._config = fig._config | config
+
+    if plot_it:
+        fig.show(config=config)
+
+    return df, fig
+
+
+def _detect_latex_format(text: str) -> bool:
+    """
+    Detect if a string contains LaTeX formatting (incompatible with Plotly).
+
+    Parameters
+    ----------
+    text : str
+        Text to check for LaTeX formatting
+
+    Returns
+    -------
+    bool
+        True if LaTeX formatting is detected (e.g., $...$, _{...}, ^{...})
+    """
+    import re
+    # Check for common LaTeX patterns:
+    # - Text wrapped in $ $
+    # - LaTeX subscripts _{...}
+    # - LaTeX superscripts ^{...}
+    latex_patterns = [
+        r'\$[^$]+\$',           # $...$
+        r'_\{[^}]+\}',          # _{...}
+        r'\^\{[^}]+\}'          # ^{...}
+    ]
+
+    for pattern in latex_patterns:
+        if re.search(pattern, text):
+            return True
+    return False
+
+
+def _format_html_species(formula: str) -> str:
+    """
+    Format a chemical formula for HTML rendering in Plotly.
+
+    Converts chemical formulas like "H2O" to "H<sub>2</sub>O" and
+    "Ca+2" to "Ca<sup>2+</sup>".
+
+    Parameters
+    ----------
+    formula : str
+        Chemical formula to format
+
+    Returns
+    -------
+    str
+        HTML-formatted formula
+    """
+    import re
+
+    # Handle charge notation (e.g., +2, -1)
+    # Match patterns like +2, -2, +, -
+    charge_pattern = r'([+-])(\d*)'
+
+    def format_charge(match):
+        sign = match.group(1)
+        num = match.group(2)
+        if num == '' or num == '1':
+            return f"<sup>{sign}</sup>"
+        else:
+            return f"<sup>{num}{sign}</sup>"
+
+    # First handle charges at the end
+    formula = re.sub(charge_pattern + r'$', format_charge, formula)
+
+    # Handle subscript numbers (digits that aren't part of the charge)
+    # Match digits that come after letters and aren't preceded by < or >
+    def format_subscript(match):
+        return f"<sub>{match.group(0)}</sub>"
+
+    # Find all digits that should be subscripts
+    # This matches digits that come after letters
+    result = []
+    i = 0
+    while i < len(formula):
+        if formula[i].isdigit() and i > 0 and formula[i-1].isalpha():
+            # Start of a number sequence
+            num_start = i
+            while i < len(formula) and formula[i].isdigit():
+                i += 1
+            result.append(f"<sub>{formula[num_start:i]}</sub>")
+        else:
+            result.append(formula[i])
+            i += 1
+
+    return ''.join(result)
+
+
+def _save_figure(fig, save_as, save_format, save_scale, plot_width, plot_height, ppi):
+    """
+    Save a Plotly figure to a file.
+
+    Parameters
+    ----------
+    fig : plotly figure
+        The figure to save
+    save_as : str or None
+        Filename (without extension) to save as
+    save_format : str or None
+        Format to save ('png', 'jpg', 'jpeg', 'webp', 'svg', 'pdf', 'eps', 'json', 'html')
+    save_scale : float
+        Scale factor for saving
+    plot_width : int
+        Width of the plot
+    plot_height : int
+        Height of the plot
+    ppi : int
+        Pixels per inch
+
+    Returns
+    -------
+    tuple
+        (save_as, save_format) - processed values for use in config
+    """
+    import plotly.io as pio
+
+    valid_formats = ['png', 'jpg', 'jpeg', 'webp', 'svg', 'pdf', 'eps', 'json', 'html']
+
+    if isinstance(save_format, str) and save_format not in valid_formats:
+        raise ValueError(f"{save_format} is an unrecognized save format. "
+                         f"Supported formats include: {', '.join(valid_formats)}")
+
+    if isinstance(save_format, str) and save_as is not None:
+        if not isinstance(save_as, str):
+            save_as = "newplot"
+
+        if save_format == "html":
+            fig.write_html(save_as + ".html")
+            print(f"Saved figure as {save_as}.html")
+            save_format = 'png'
+        elif save_format in ['pdf', 'eps', 'json']:
+            pio.write_image(fig, save_as + "." + save_format, format=save_format,
+                            scale=save_scale, width=plot_width * ppi, height=plot_height * ppi)
+            print(f"Saved figure as {save_as}.{save_format}")
+            save_format = "png"
+        else:
+            pio.write_image(fig, save_as + "." + save_format, format=save_format,
+                            scale=save_scale, width=plot_width * ppi, height=plot_height * ppi)
+            print(f"Saved figure as {save_as}.{save_format}")
+    else:
+        save_format = "png"
+
+    return save_as, save_format
+
+
+def _plot_saturation_interactive(eout, values_list, sp_names, xyvars, xyvals,
+                                  xlab, ylab, col, lwd, lty, cex, contour_method,
+                                  main, add, ax, width, height, plot_it,
+                                  save_as, save_format, save_scale, messages):
+    """
+    Plot saturation lines (affinity=0 contours) for interactive 2-D diagrams using Plotly.
+
+    This function draws contour lines where affinity = 0 for each species,
+    indicating saturation boundaries (e.g., mineral precipitation thresholds).
+    """
+    import plotly.graph_objects as go
+    import plotly.io as pio
+
+    # Get x and y values
+    xvals = xyvals[0]
+    yvals = xyvals[1]
+    xvar = xyvars[0]
+    yvar = xyvars[1]
+
+    n_species = len(sp_names)
+
+    if messages:
+        print(f"diagram: plotting saturation lines for interactive 2-D diagram")
+
+    # Set up colors
+    if col is None:
+        # Use default Plotly colors
+        default_colors = ['#636EFA', '#EF553B', '#00CC96', '#AB63FA', '#FFA15A',
+                         '#19D3F3', '#FF6692', '#B6E880', '#FF97FF', '#FECB52']
+        col = [default_colors[i % len(default_colors)] for i in range(n_species)]
+    elif isinstance(col, str):
+        col = [col] * n_species
+    else:
+        col = list(col)
+        if len(col) < n_species:
+            col = col * (n_species // len(col) + 1)
+        col = col[:n_species]
+
+    # Set up line widths
+    if isinstance(lwd, (int, float)):
+        lwd = [lwd] * n_species
+    else:
+        lwd = list(lwd)
+        if len(lwd) < n_species:
+            lwd = lwd * (n_species // len(lwd) + 1)
+        lwd = lwd[:n_species]
+
+    # Handle line styles (lty)
+    if lty is None:
+        lty = ['solid'] * n_species
+    elif isinstance(lty, (str, int)):
+        lty = [lty] * n_species
+    else:
+        lty = list(lty)
+        if len(lty) < n_species:
+            lty = lty * (n_species // len(lty) + 1)
+        lty = lty[:n_species]
+
+    # Convert numeric/matplotlib line styles to Plotly dash types
+    lty_map = {
+        1: 'solid', '-': 'solid',
+        2: 'dash', '--': 'dash',
+        3: 'dashdot', '-.': 'dashdot',
+        4: 'dot', ':': 'dot',
+        5: 'solid', 6: 'dash'
+    }
+    lty = [lty_map.get(lt, 'solid') if (isinstance(lt, int) or lt in lty_map) else 'solid' for lt in lty]
+
+    # Handle contour label control
+    if contour_method is None or contour_method == "" or (isinstance(contour_method, str) and contour_method.upper() == "NA"):
+        show_labels = [False] * n_species
+    elif isinstance(contour_method, str):
+        show_labels = [True] * n_species
+    elif isinstance(contour_method, list):
+        if len(contour_method) != n_species:
+            contour_method_extended = list(contour_method) * (n_species // len(contour_method) + 1)
+            contour_method_extended = contour_method_extended[:n_species]
+        else:
+            contour_method_extended = contour_method
+
+        show_labels = []
+        for method in contour_method_extended:
+            if method is None or method == "" or (isinstance(method, str) and method.upper() == "NA"):
+                show_labels.append(False)
+            else:
+                show_labels.append(True)
+    else:
+        show_labels = [True] * n_species
+
+    # Handle text size (cex) for contour labels
+    if isinstance(cex, (int, float)):
+        cex_list = [cex] * n_species
+    else:
+        cex_list = list(cex)
+        if len(cex_list) < n_species:
+            cex_list = cex_list * (n_species // len(cex_list) + 1)
+        cex_list = cex_list[:n_species]
+
+    # Base font size for labels (Plotly default ~12)
+    base_font_size = 12
+    font_sizes = [base_font_size * c for c in cex_list]
+
+    # Create or get figure
+    if add and ax is not None:
+        # ax is actually the Plotly figure from previous call
+        fig = ax
+    else:
+        # Create new figure
+        fig = go.Figure()
+
+    # Add contour lines for each species
+    for i, sp_name in enumerate(sp_names):
+        # Get affinity values for this species
+        # values_list[i] has shape (nx, ny)
+        affinity_2d = values_list[i]
+
+        # Create contour trace for affinity=0 only
+        # Use ncontours=1 with start=0 and end=0 to force a single contour at zero
+        # Note: Plotly doesn't support custom text on contour lines, so we rely on the legend
+        contour = go.Contour(
+            x=xvals,
+            y=yvals,
+            z=affinity_2d.T,  # Transpose to match Plotly's expected orientation
+            ncontours=1,  # Only generate one contour level
+            contours=dict(
+                coloring='lines',  # Draw only contour lines, not filled regions
+                start=0,  # Start at 0
+                end=0,    # End at 0
+                showlabels=False  # Don't show "0" labels on contour lines
+            ),
+            line=dict(
+                color=col[i],
+                width=lwd[i],
+                dash=lty[i]
+            ),
+            colorscale=[[0, col[i]], [1, col[i]]],  # Force uniform color
+            showscale=False,
+            hoverinfo='skip',
+            name=sp_name,
+            legendgroup=sp_name,
+            showlegend=True
+        )
+
+        fig.add_trace(contour)
+
+    # Set axis labels if not adding to existing plot
+    if not add:
+        # Create axis labels with proper units
+        from .thermo import thermo as thermo_func
+        basis_df = eout['basis']
+        basis_sp = list(basis_df.index)
+        basis_state = list(basis_df['state'])
+
+        unit_dict = {"P": "bar", "T": "°C", "pH": "", "Eh": "volts", "IS": "mol/kg"}
+
+        for i, s in enumerate(basis_sp):
+            if basis_state[i] in ["aq", "liq", "cr"]:
+                unit_dict[s] = f"log <i>a</i><sub>{s}</sub>"
+            else:
+                unit_dict[s] = f"log <i>f</i><sub>{s}</sub>"
+
+        if not isinstance(xlab, str):
+            xlab = xvar + ", " + unit_dict.get(xvar, "")
+            if xvar == "pH":
+                xlab = "pH"
+            if xvar in basis_sp:
+                xlab = unit_dict[xvar]
+
+        if not isinstance(ylab, str):
+            ylab = yvar + ", " + unit_dict.get(yvar, "")
+            if yvar in basis_sp:
+                ylab = unit_dict[yvar]
+            if yvar == "pH":
+                ylab = "pH"
+
+        fig.update_xaxes(title_text=xlab)
+        fig.update_yaxes(title_text=ylab)
+
+        fig.update_layout(
+            template="simple_white",
+            width=width,
+            height=height,
+            showlegend=True
+        )
+
+        if isinstance(main, str):
+            fig.update_layout(title={'text': main, 'x': 0.5, 'xanchor': 'center'})
+
+    # Configure download button
+    save_as_name, save_format_final = _save_figure(fig, save_as, save_format, save_scale,
+                                                    plot_width=width, plot_height=height, ppi=1)
+
+    config = {'displaylogo': False,
+              'modeBarButtonsToRemove': ['resetScale2d', 'toggleSpikelines'],
+              'toImageButtonOptions': {
+                  'format': save_format_final,
+                  'filename': save_as_name,
+                  'height': height,
+                  'width': width,
+                  'scale': save_scale,
+              }}
+
+    fig._config = fig._config | config if hasattr(fig, '_config') else config
+
+    # Show plot if requested
+    if plot_it:
+        fig.show(config=config)
+
+    # Return empty DataFrame (saturation doesn't produce tabular data like predominance diagrams)
+    df = pd.DataFrame()
+
+    return df, fig
+
+
+# Export main functions
+__all__ = ['diagram', 'diagram_interactive', 'water_lines', 'find_tp']