Diverging palettes

library(circadia)
library(ggplot2)

Diverging palettes are designed for continuous data with a meaningful midpoint — typically zero, or a reference value such as a group mean. The colour transitions outward from a neutral centre in two directions, making positive and negative departures immediately distinguishable.

circadia provides two diverging palettes:

Palette Hue strategy Best for
diverging Multi-hue — blue → teal → antique white → amber → coral Figures where maximum contrast across the full range matters
diverging_simple Single-axis interpolation — deep blue → antique white → coral Smaller figures, print, or when chromatic noise should be minimised

diverging — complex

Nine stops, multi-hue. High perceptual contrast across the full range.

#014370
#1B6799
#4A9BBF
#9BDFE2
#FFECD4
#FFC99A
#FFA75D
#FC7060
#FC544A 
# Correlation matrix
vars    <- c("mpg", "cyl", "disp", "hp", "drat", "wt", "qsec")
cor_mat <- cor(mtcars[, vars])
cor_df  <- data.frame(
  Var1  = rep(vars, each  = length(vars)),
  Var2  = rep(vars, times = length(vars)),
  value = as.vector(cor_mat)
)

ggplot(cor_df, aes(Var1, Var2, fill = value)) +
  geom_tile(colour = "white", linewidth = 0.5) +
  geom_text(aes(label = round(value, 2)), size = 3, colour = "white") +
  scale_fill_circadia_c("diverging", limits = c(-1, 1)) +
  labs(title = "Motor Trend — variable correlations",
       fill = "r", x = NULL, y = NULL) +
  theme_circadia(grid = "none") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

Correlation heatmap of mtcars variables using the diverging palette

diverging_simple — simple

Seven stops, direct blue → antique white → coral interpolation. Lower chromatic complexity — cleaner on small multiples or in print.

#014370
#567B91
#AAB4B3
#FFECD4
#FEB9A6
#FD8778
#FC544A 
ggplot(cor_df, aes(Var1, Var2, fill = value)) +
  geom_tile(colour = "white", linewidth = 0.5) +
  geom_text(aes(label = round(value, 2)), size = 3, colour = "#333333") +
  scale_fill_circadia_c("diverging_simple", limits = c(-1, 1)) +
  labs(title = "Motor Trend — variable correlations (simple palette)",
       fill = "r", x = NULL, y = NULL) +
  theme_circadia(grid = "none") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

Correlation heatmap using the diverging_simple palette

Side-by-side comparison

library(patchwork)

# Standardised sleep-like data: z-scores across a small grid
set.seed(42)
grid_df <- expand.grid(x = 1:12, y = 1:8)
grid_df$z <- as.vector(scale(rnorm(96, mean = grid_df$x - 6, sd = 2)))

make_tile <- function(pal) {
  ggplot(grid_df, aes(x, y, fill = z)) +
    geom_tile() +
    scale_fill_circadia_c(pal, limits = c(-3, 3)) +
    labs(title = paste0('palette = "', pal, '"'), x = NULL, y = NULL, fill = "z") +
    theme_circadia(grid = "none") +
    theme(axis.text = element_blank(), axis.ticks = element_blank())
}

make_tile("diverging") | make_tile("diverging_simple")

Side by side comparison of diverging and diverging_simple on a z-score heatmap

Controlling limits and midpoints

Use the limits argument (passed to scale_fill_gradientn()) to pin the palette midpoint to a specific value:

# Centre the palette at 0 with symmetric limits
scale_fill_circadia_c("diverging", limits = c(-1, 1))

# Asymmetric data — still centre at 0
scale_fill_circadia_c("diverging", limits = c(-3, 3))