Research guide
Scientific Color Palettes for Papers and Figures (2026)
Colorblind-safe scientific color palettes with hex codes, plus rules for choosing colors in research figures, charts, and posters.
The best scientific color palettes for papers are perceptually uniform and colorblind-safe: use viridis for continuous data, the Okabe–Ito set for categorical colors, and a ColorBrewer diverging scheme when values run above and below a midpoint. Good color is not decoration — it encodes meaning, survives grayscale printing, and stays legible to the roughly 1 in 12 men and 1 in 200 women with color-vision deficiency. This guide gives ready-to-use palettes with hex codes and the rules for applying them.
Ready-to-use scientific color palettes
| Palette | Type | Colorblind-safe | Best for |
|---|---|---|---|
| Viridis | Sequential | Yes | Heatmaps, continuous scales, density plots |
| Cividis | Sequential | Yes (optimized) | Continuous data printed in grayscale |
| Okabe–Ito | Categorical | Yes | Up to 8 discrete groups or series |
| ColorBrewer Set2 | Categorical | Yes | 3–6 soft categorical groups |
| ColorBrewer RdBu | Diverging | Yes | Correlations, log-fold change, ± values |
| Grayscale | Sequential | Always | Single-variable figures, print-only venues |
The Okabe–Ito categorical palette (with hex)
This eight-color set was designed to stay distinguishable across the common types of color blindness, which makes it a safe default for bar charts, scatter series, and schematic labels:
- Black
#000000 - Orange
#E69F00 - Sky blue
#56B4E9 - Bluish green
#009E73 - Yellow
#F0E442 - Blue
#0072B2 - Vermillion
#D55E00 - Reddish purple
#CC79A7
Use them in order and stop at the number of categories you actually have. More than six or seven colors becomes hard to tell apart no matter how the palette was designed.
Rules for color in research figures
Encode data, not decoration. Every distinct color should mean something. If two bars are the same category, give them the same color; if color is just for visual variety, use a single hue instead.
Design for grayscale first. Many readers still print in black and white, and some journals charge for color figures. A figure that relies only on hue fails in grayscale — add differences in lightness, line style, or texture so the meaning survives.
Avoid the rainbow (jet) colormap. The classic red-to-blue rainbow is not perceptually uniform: it creates false boundaries where the data is smooth and hides real gradients. Viridis and cividis were built specifically to fix this and are now the defaults in most plotting libraries.
Limit your palette. Two to four hues is usually enough for a single figure. Consistency across a figure set matters more than variety — reuse the same color for the same variable throughout the paper.
Never encode meaning by red/green alone. It is the most common confusion axis. Pair color with shape, position, or a direct label so a red/green-blind reader loses nothing.
Matching color to figure type
A conceptual or mechanism diagram uses color to group related parts — keep the palette to a few muted hues so labels stay readable. When you draft one with the Scientific Illustration Generator or a Graphical Abstract Maker, recolor the elements from a colorblind-safe set rather than the default output. A scientific poster needs higher contrast because it is read from a distance: dark text on a light background, with one accent color for emphasis rather than a different color per section.
Tools and where to get palettes
Most scientific plotting libraries ship these palettes: matplotlib and seaborn include viridis and cividis by default, R's viridis and RColorBrewer packages cover the rest, and the ColorBrewer website lets you preview and copy hex codes for any scheme. For hand-built figures, keep a saved swatch of your chosen palette so every figure in the paper draws from the same set.
FAQ
What is the best colorblind-safe palette? For categorical data, the Okabe–Ito eight-color set; for continuous data, viridis or cividis. Both remain distinguishable across the common types of color-vision deficiency.
Why avoid the rainbow colormap? It is not perceptually uniform, so it invents boundaries where data is smooth and hides real gradients. Viridis and cividis were designed to correct this.
How many colors should a scientific figure use? Usually two to four. Beyond six or seven, categories become hard to distinguish regardless of palette.
Do my figures need to work in grayscale? Yes. Many readers print in black and white and some journals charge for color, so encode meaning with lightness or pattern as well as hue.
Ready to build a figure with a clean, colorblind-safe palette? Generate an editable scientific figure free.