Gray–Scott model

A reaction–diffusion system heavily studied for its complex dynamics is the Gray–Scott system, given by

\[\begin{aligned}\pd{u}{t}&=\nabla^2 u+u^2v - (a+b)u,\\ \pd{v}{t}&=D\nabla^2v-u^2v + a(1 - v),\end{aligned}\]

where we take $D=2$ and only vary $a,b>0$. This model has a wide range of behaviours, shown in another WebGL simulator that partially inspired VisualPDE.

A famous 1993 paper on this model explored a range of the parameters $a$ and $b$ to classify different behaviours, and many people have hence made these parameters depend linearly on $x$ and $t$ to see all of this behaviour at once.

Building from the previous simulation, we can rescale the heterogeneity to still be monotonic, but to use up more of the domain to see different dynamical regimes.

Interestingly, the value of $D=2$ used gives a very rich parameter space, whereas making $D$ smaller reduces the regions of patterned behaviour, and taking $D$ larger increases it at the cost of making things more stationary and more spot-like for most of the parameter domain.

Furthermore, when $D=1$ the system no longer supports stationary patterns, but does exhibit waves similar to the spiral waves in the equal-diffusion case of the cyclic competition models.

Below we’ve listed some parameter combinations that give rise to different and interesting behaviours.

One of our favourites is the moving spots simulation, which exhibits spots bobbing around.

$a$ $b$ Description
0.037 0.06 Labyrinthine
0.03 0.062 Spots
0.025 0.06 Pulsating spots
0.078 0.061 Worms
0.039 0.058 Holes
0.026 0.051 Spatiotemporal chaos
0.034 0.056 Intermittent chaos/holes
0.014 0.054 Moving spots (glider-like)
0.018 0.051 Small waves
0.014 0.045 Big waves
0.062 0.061 U-skate world