How to Make Subplots in Matplotlib (Python Multi-Panel Figures Guide)

Most real scientific figures are not single plots — they are multi-panel layouts with an (a), (b), (c) grid of related views. In matplotlib, the tool for this is plt.subplots(), with subplot_mosaic() for irregular layouts. This guide covers grids, shared axes, unequal panel sizes, named mosaic layouts, and panel labeling — everything you need to assemble a journal-ready composite figure.

If you are still learning matplotlib basics, read the complete matplotlib tutorial first, then use this guide to combine plots into one figure.

1. A grid of subplots

Call plt.subplots(nrows, ncols). With more than one row and column you get a 2D array of axes, indexed as axes[row, col]. Always set the figure size up front — resizing later rescales fonts and breaks journal sizing.

import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(0, 2 * np.pi, 200)

# 2 rows x 2 columns of axes
fig, axes = plt.subplots(2, 2, figsize=(8, 6))

axes[0, 0].plot(x, np.sin(x), color="#7c3aed")
axes[0, 0].set_title("sin")
axes[0, 1].plot(x, np.cos(x), color="#ec4899")
axes[0, 1].set_title("cos")
axes[1, 0].plot(x, np.tan(x), color="#10b981")
axes[1, 0].set_ylim(-5, 5)
axes[1, 0].set_title("tan")
axes[1, 1].plot(x, np.exp(-x), color="#f59e0b")
axes[1, 1].set_title("decay")

fig.tight_layout()

2. Loop over panels cleanly

When every panel does the same thing with different data, do not hardcode indices. Use axes.flat to iterate over all axes in row-major order in a single loop — this scales from a 2×2 to a 5×5 grid without rewriting the code.

import matplotlib.pyplot as plt
import numpy as np

rng = np.random.default_rng(42)
fig, axes = plt.subplots(2, 3, figsize=(10, 6))

# axes is a 2x3 array; flatten it to loop in one line
for i, ax in enumerate(axes.flat):
    ax.hist(rng.normal(i, 1, 300), bins=20, color="#7c3aed", edgecolor="white")
    ax.set_title(f"Sample {i + 1}")

fig.tight_layout()

3. Share axes to align comparisons

Stacked time series or spectra should share an x-axis so features line up vertically. Pass sharex=True, tighten the vertical spacing with gridspec_kw={"hspace": ...}, and call fig.align_ylabels() so the y-labels sit at a consistent x position.

import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(0, 10, 200)
fig, axes = plt.subplots(3, 1, figsize=(6, 7),
                         sharex=True,            # one shared x-axis
                         gridspec_kw={"hspace": 0.08})

axes[0].plot(x, np.sin(x), color="#7c3aed")
axes[1].plot(x, np.sin(2 * x), color="#ec4899")
axes[2].plot(x, np.sin(3 * x), color="#10b981")
axes[2].set_xlabel("Time (s)")

for ax in axes:
    ax.set_ylabel("Amp")

fig.align_ylabels(axes)

4. Unequal panel sizes with height ratios

A common scientific layout is a tall main panel with a short residual strip beneath it. Use gridspec_kw={"height_ratios": [3, 1]} to size the rows proportionally, and sharex=True so the two panels stay aligned.

import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(0, 10, 200)

# Top panel 3x taller than the bottom residual strip
fig, (ax_main, ax_resid) = plt.subplots(
    2, 1, figsize=(6, 5), sharex=True,
    gridspec_kw={"height_ratios": [3, 1], "hspace": 0.05},
)

ax_main.plot(x, np.sin(x), color="#7c3aed", label="model")
ax_main.legend(frameon=False)
ax_resid.axhline(0, color="gray", lw=0.8)
ax_resid.scatter(x, np.random.normal(0, 0.1, x.size), s=5, color="#ec4899")
ax_resid.set_xlabel("x")
ax_resid.set_ylabel("resid")

5. Irregular layouts with subplot_mosaic

For anything that is not a clean grid — a wide overview spanning the top and two detail panels below — plt.subplot_mosaic() is far cleaner than manual gridspec. You describe the layout as a list of rows of names, then access each axis by name. Repeating a name makes that panel span those cells.

import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(0, 10, 200)

# Named layout: "wide" spans the top row, two panels share the bottom
fig, axd = plt.subplot_mosaic(
    [["wide", "wide"],
     ["lower_left", "lower_right"]],
    figsize=(8, 5.5),
)

axd["wide"].plot(x, np.sin(x), color="#7c3aed")
axd["wide"].set_title("Overview")
axd["lower_left"].scatter(x, np.cos(x), s=6, color="#ec4899")
axd["lower_right"].hist(np.cos(x), bins=20, color="#10b981", edgecolor="white")

fig.tight_layout()

6. Label the panels (a, b, c)

Journals expect each panel to carry a bold letter. Place it in axes-fraction coordinates with transform=ax.transAxes so the position is consistent regardless of the data range.

import matplotlib.pyplot as plt
import string

fig, axes = plt.subplots(1, 3, figsize=(10, 3.2))

for ax, letter in zip(axes.flat, string.ascii_uppercase):
    ax.text(-0.1, 1.05, letter, transform=ax.transAxes,
            fontsize=13, fontweight="bold", va="bottom", ha="right")

fig.tight_layout()

Quick reference: subplot arguments

Common mistakes and fixes

• Indexing a 1D axes array as 2D — with a single row or column, axes is 1D (axes[i]), not axes[i, j]. Pass squeeze=False if you want it always 2D.
• Overlapping labels — call fig.tight_layout() or build with constrained_layout=True.
• Repeated axis ticks — use sharex/sharey so inner panels drop redundant tick labels.
• Resizing after plotting — set figsize in the subplots() call; resizing later rescales fonts unpredictably.
• One shared legend repeated per panel — collect handles and call fig.legend() once, outside the axes.

Frequently asked questions