Publication-Quality Formatting

From Notebook Sketch to Journal Figure

A figure that looks good in Jupyter often fails at the journal stage: fonts are too small, linewidths vanish at print scale, and file format is wrong. This section teaches the formatting rules for IEEE, Springer, and ACM venues — covering fonts, DPI, figure sizes, and export formats.

Definition:
rcParams — Global Style Configuration

matplotlib.rcParams is a dict-like object controlling every default:

import matplotlib.pyplot as plt

plt.rcParams.update({
    'font.family': 'serif',
    'font.serif': ['Times New Roman'],
    'font.size': 10,
    'axes.labelsize': 11,
    'axes.titlesize': 12,
    'xtick.labelsize': 9,
    'ytick.labelsize': 9,
    'legend.fontsize': 9,
    'figure.figsize': (3.5, 2.5),
    'figure.dpi': 150,
    'savefig.dpi': 300,
    'lines.linewidth': 1.5,
})

Use plt.rcdefaults() to reset to defaults.

Define rcParams at the top of your script, not scattered through plotting code. Consider a project-wide matplotlibrc file.

Definition:
IEEE Transactions Figure Specifications

IEEE requires:

Single column: 3.5 inches wide; double column: 7.16 inches
Font size: minimum 8 pt in the printed figure
Format: EPS or PDF (vector), or TIFF at 600 DPI (raster)
Font: preferably Times New Roman or Computer Modern

IEEE_SINGLE = {
    'figure.figsize': (3.5, 2.625),   # 4:3 aspect
    'font.family': 'serif',
    'font.size': 8,
    'savefig.dpi': 600,
    'savefig.format': 'pdf',
}

Theorem: DPI-Size Relationship

The pixel dimensions of a saved figure are:

$\text{pixels}_w = \text{figsize}_w \times \text{DPI}, \quad \text{pixels}_h = \text{figsize}_h \times \text{DPI}$

For a 3.5 in $\times$ 2.5 in figure at 300 DPI, the output is $1050 \times 750$ pixels. At 600 DPI, it is $2100 \times 1500$ pixels.

Set figsize to the physical print size (in inches) and DPI to the journal requirement. Never resize the figure after export.

Example: Complete IEEE-Ready BER Figure

Create a publication-quality BER curve with IEEE formatting that passes journal submission requirements.

Solution

Style setup

import matplotlib.pyplot as plt
import numpy as np
from scipy.special import erfc

plt.rcParams.update({
    'font.family': 'serif',
    'font.size': 8,
    'axes.labelsize': 9,
    'legend.fontsize': 7,
    'lines.linewidth': 1.2,
    'lines.markersize': 4,
    'figure.figsize': (3.5, 2.625),
    'savefig.dpi': 600,
    'axes.grid': True,
    'grid.alpha': 0.3,
})

Plotting

snr_db = np.arange(0, 21)
snr = 10**(snr_db / 10)

fig, ax = plt.subplots()
ax.semilogy(snr_db, 0.5*erfc(np.sqrt(snr)), 'o-',
            label='BPSK')
ax.semilogy(snr_db, 0.5*erfc(np.sqrt(snr/2)), 's--',
            label='QPSK')
ax.set(xlabel=r' $E_b/N_0$  (dB)', ylabel='BER')
ax.set_ylim(1e-6, 1)
ax.legend(loc='lower left')
fig.savefig('ber_ieee.pdf', bbox_inches='tight')

Example: Using LaTeX in Labels and Annotations

Add properly typeset mathematical labels using Matplotlib's built-in mathtext renderer (no LaTeX installation required).

Solution

Implementation

fig, ax = plt.subplots(figsize=(6, 4))
ax.set_xlabel(r' $E_b/N_0$  (dB)')
ax.set_ylabel(r' $P_b = \frac{1}{2}\mathrm{erfc}\!\left(\sqrt{E_b/N_0}\right)$ ')
ax.set_title(r'BER for BPSK in AWGN Channel')
ax.annotate(r' $P_b \approx 10^{-5}$ ',
            xy=(10, 1e-5), xytext=(6, 1e-3),
            arrowprops=dict(arrowstyle='-|>', color='red'),
            fontsize=10, color='red')

Key rules

Always use raw strings (r'...') to avoid backslash issues.
Use \mathrm{} for upright text inside math mode.
Use \! for negative thin space and \, for thin space.

Common rcParams Cheatsheet — The most commonly used rcParams grouped by category: fonts, lines, axes, figure, and saving.

Figure Export Formats

Format	Type	Quality	File Size	Best For
PDF	Vector	Lossless	Small	Journal papers, LaTeX
SVG	Vector	Lossless	Small	Web, presentations
PNG	Raster	DPI-dependent	Medium	Slides, reports
TIFF	Raster	DPI-dependent	Large	Some journals
EPS	Vector	Lossless	Small	Legacy LaTeX workflows

Common Mistake: Font Size Depends on Figure Size

Mistake:

Setting font.size=10 but figsize=(12, 8). When the figure is shrunk to column width in the paper, the 10pt font becomes ~4pt — unreadable.

Correction:

Set figsize to the final print size (e.g., 3.5 in for IEEE single column). Then 8pt font in the script stays 8pt in print.

Key Takeaway

Set figsize to the final print dimensions first, then adjust font sizes. If the figure will be 3.5 inches wide in the paper, use figsize=(3.5, 2.5) and choose fonts that look right at that size. Never rely on scaling after export.

Quick Check

A figure is saved with figsize=(3.5, 2.5) and dpi=300. What are the pixel dimensions?

350 x 250

1050 x 750

3500 x 2500

700 x 500

Correction:

1050 x 750

pixels = figsize x DPI: 3.5300=1050, 2.5300=750.

Historical Note: The Viridis Revolution

2015

In 2015, Matplotlib switched its default colormap from jet to viridis, designed by Stéfan van der Walt and Nathaniel Smith. Viridis was optimized for perceptual uniformity, colorblind-friendliness, and grayscale printability. The change sparked a broader movement across scientific software toward perceptually uniform colormaps.

DPI (Dots Per Inch)

The resolution of a raster image. Higher DPI means more pixels per physical inch, resulting in sharper prints.

rcParams

Matplotlib's global configuration dictionary controlling all default visual properties.

Subplots and Multi-Panel Figures Complex-Valued and Phase Plots