Flow Lab

Syllabus / Module 15: Applications: Biology, Chemistry & Society

Lesson 15.5

Reaction–Diffusion & Turing Patterns

How diffusion can destabilize: Turing instability, Gray–Scott and FitzHugh–Nagumo patterns, spots, stripes and spirals.

Written byClaude Sonnet

🌊
Why this matters for simulation: The same solver structure as a fluid sim: diffuse + react on a grid.

By the end of this lesson you can…

  • Derive the Fisher-KPP front speed $2\sqrt{Dr}$ from the linearised leading edge and check it, with its slow approach, on a grid.
  • Write the growth rate of a Fourier mode as an eigenvalue of $J - q^2 D$ and say when diffusion destabilises.
  • Derive the Turing conditions, the unstable band, $q_c$ and the Brusselator threshold in $b$.
  • Compare the fastest-growing linear wavelength with the wavelength of a nonlinear 1D pattern and explain the gap.
  • Find and measure the explicit stability limit of a reaction-diffusion scheme, and say what IMEX changes.
  • Explain why Gray-Scott patterns need a finite seed and what Turing patterns do not show about animals.
On this page
·
  1. Reaction plus diffusion
  2. Fisher-KPP: a front with a speed
  3. Linear stability with diffusion
  4. Turing instability of the Brusselator
  5. A pattern from noise: linear against nonlinear
  6. Computing it: the explicit limit and IMEX
  7. Gray-Scott in 2D, and why the Brusselator does the analysis
  8. What this does and does not show
  9. Where this goes next
  10. The exercise: Laplacian, Fisher speed, dispersion relation and step limit
  11. Concept checks
  12. Problems
  13. Key Takeaways

15.4 Chemical Kinetics & Stiff Systems ended with a determinant going negative. The Brusselator with a=1a = 1, b=1.9b = 1.9 is a stable spiral alone, yet once the two species diffuse at different speeds, J−q2 diag(Dx,Dy)J - q^2\,\mathrm{diag}(D_x, D_y) has determinant −0.6-0.6 at q2=0.4q^2 = 0.4 and a mode grows. That was a promise. This lesson cashes it: where the unstable band is, which wavelength wins, what the nonlinear pattern looks like and what it costs to compute. On the way we settle an IOU from 15.3 Epidemic Models: the front speed 2Dr2\sqrt{Dr} quoted there from memory is now measured.

▶ Video A reaction term plus diffusion, the Fisher-KPP front and its speed, the dispersion relation and the Turing band, a pattern growing from noise, the explicit step limit, and Gray-Scott spotscaptions only
Transcript
  1. Last lesson promised a pattern. The Brusselator at b equal 1.9 is stable alone, yet with D x equal 1 and D y equal 10 the determinant of J minus q squared D is minus 0.6 at q squared equal 0.4, and a mode grows. One diffusing species only smooths; a pair with unequal diffusivities can make structure. First, an IOU: the front speed 2 root D r from the epidemics lesson, now measured.
  2. Take logistic growth with diffusion, u t equals D u x x plus r u times one minus u. Far ahead of a front u is tiny, so a tail e to the minus lambda times x minus c t gives c of lambda equals D lambda plus r over lambda. Steep tails are limited by diffusion, shallow ones by growth, with the minimum at lambda star equal root r over D. The speed is c star equals 2 root D r: for D and r equal 1, 2. Doubling D multiplies it by root 2, not 2.
  3. Explicit FTCS with dx equal 0.1, over t from 20 to 40, gives 1.946 against 2.000. For D equal 2 it gives 2.752 against 2.828, and for r equal 0.5, 1.330 against 1.414: 3 percent to 6 percent low, since pulled fronts approach c star slowly, reaching 1.983 for t from 80 to 100. The tail matters: e to the minus 0.5 x runs at 2.499 and e to the minus 0.7 x at 2.127, but a steeper e to the minus 2 x runs at 1.947, near 2, not 2.5. That selection rule is from memory.
  4. Perturb a uniform state by cos of q x. Then sigma v equals J minus q squared D, times v, and the top eigenvalue is the dispersion relation. One species can only be stabilised. For two, stable without diffusion, only the determinant can go negative: D 1 D 2 q to the fourth minus B q squared plus det J. That needs B above 2 root D 1 D 2 det J: a self-activating species and a faster inhibitor, local activation with long range inhibition. For the Brusselator, b must exceed 1 plus a root D x over D y, squared.
  5. At a equal 1, D y equal 10 and b equal 1.9, J has eigenvalues minus 0.05 plus or minus 0.9987 i, a stable spiral. The determinant vanishes at q squared equal 0.1551 and 0.6449: the band is q from 0.394 to 0.803, wavelengths 7.82 to 15.96, with threshold b equal 1.7325. The fastest mode has q squared 0.3216, growth 0.1425, wavelength 11.08. From b equal 1.75 to 1.95 peak growth rises about twelvefold, while the wavelength stays near 11. Chemistry and diffusion set it, not the domain.
  6. Linear theory says which modes grow, not where growth stops. The nonlinear Brusselator on 0 to 100, with 201 nodes and noise of size 0.01, settles into stationary peaks by t equal 400; seed 1 spaces them 11.773 apart. Over 12 seeds, 7 gave 11.7, 3 gave 11.1 and 2 gave 10.5: all in the band, within 6.3 percent of 11.08. Modes 17, 18 and 19 grow within 3 percent of each other, so the nonlinear terms decide, and the linear winner came out only 3 times in 12.
  7. Explicit stepping needs the checkerboard mode inside the stability interval; the faster diffusivity binds. With the reaction the most negative eigenvalue is minus 160.99, so h must stay below 0.012423. The checkerboard factor is 0.9962 at h equal 0.0124 and 1.0284 at 0.0126, which blows up within about 800 steps. At h equal 0.01 the run takes 40000 steps. IMEX treats diffusion implicitly: at h equal 0.2 it needs 2000 steps, 20 times fewer, with spacing 11.776 against 11.773. At h equal 0.3 the pattern is lost, for a reason I did not isolate.
  8. Gray-Scott is a poor model to analyse: its empty state is stable for every F and k, so no noise grows and patterns need a finite seed, like a Fisher front. With D u 0.16, D v 0.08 and a periodic 128 by 128 grid, F 0.030 and k 0.062 give 89 blobs at t equal 5000, and 110 at t equal 15000: still changing. F 0.040 and k 0.060 give one connected labyrinth covering 52.9 percent. The explicit limit is 1.5267: the checkerboard decays at 0.965 for h equal 1.5 and grows at 1.030 for 1.55.
  9. Turing proposed in 1952 that morphogens could pattern an embryo. The mathematics is exact; whether a given animal is patterned this way is a hypothesis. The diffusivity ratio must exceed 6.98, the wavelength depends on diffusivities only through a fourth root, and noise changes the outcome. Nothing here was fitted to an animal. Next, a neuron rests stably, but a big kick sends a pulse down the axon, then it recovers: an excitable medium, with a slow recovery variable that need not diffuse. The next lesson, on neuron models, supplies the reaction terms. Whether 2 root D r survives is the first question to ask.

Reaction plus diffusion

A reaction-diffusion system adds a Laplacian to each rate equation of a reaction network:

∂tu=D ∇2u+f(u),D=diag(D1,D2,… ).\partial_t \mathbf u = D\,\nabla^2 \mathbf u + \mathbf f(\mathbf u), \qquad D = \mathrm{diag}(D_1, D_2, \dots).

Without the Laplacian this is the ODE of the last three lessons at every point; without f\mathbf f it is the heat equation of 9.2 The Heat (Diffusion) Equation. One species with diffusion can only spread and smooth. A pair with unequal diffusivities can do the opposite and make structure from uniformity.

Fisher-KPP: a front with a speed

Take one species with logistic growth, as in 15.1 Population Models:

ut=D uxx+r u (1−u).u_t = D\,u_{xx} + r\,u\,(1 - u).

With u=1u = 1 on the left and u=0u = 0 on the right, a front U(x−ct)U(x - ct) invades the unstable state u=0u = 0. How fast?

Derivation. Far ahead of the front uu is tiny, u(1−u)≈uu(1-u) \approx u, and the equation is linear: ut=Duxx+ruu_t = Du_{xx} + ru. Try a tail u=e−λ(x−ct)u = e^{-\lambda(x - ct)}. Then λc=Dλ2+r\lambda c = D\lambda^2 + r, so

c(λ)=Dλ+rλ,λ∗=r/D,c∗=2D r.c(\lambda) = D\lambda + \frac{r}{\lambda}, \qquad \lambda^* = \sqrt{r/D}, \qquad c^* = 2\sqrt{D\,r}.

c(λ)c(\lambda) is large for steep tails (diffusion-limited) and shallow ones (growth-limited), with its minimum at λ∗\lambda^*. The Kolmogorov-Petrovsky-Piskunov result (1937; the selection rule is from memory) is that compactly supported data run at exactly the minimum. It is a pulled front: the linear leading edge sets the speed and the saturated bulk follows. For D=r=1D = r = 1, c(λ)=λ+1/λc(\lambda) = \lambda + 1/\lambda is minimal at λ=1\lambda = 1 with c∗=2c^* = 2; doubling DD multiplies the speed by 2\sqrt2, not 22.

Measured. The sim runs explicit FTCS (10.2 Explicit Methods for the Heat Equation) with Δx=0.1\Delta x = 0.1, h=Δx2/(10D)h = \Delta x^2/(10D), tracks u=1/2u = 1/2, and takes the speed over t∈[20,40]t \in [20, 40]. Scipy’s RK45 on the same semi-discrete system is the independent check.

DDrr2Dr2\sqrt{Dr}simscipy
11112.0002.0001.9461.9461.9481.948
22112.8282.8282.7522.7522.7542.754
110.50.51.4141.4141.3301.3301.3311.331

The measured speeds sit below c∗c^*, by 3%3\% to 6%6\%. Pulled fronts approach c∗c^* slowly. Bramson’s correction (from memory) puts the front at c∗t−32λ∗ln⁡tc^*t - \tfrac{3}{2\lambda^*}\ln t, giving window averages 1.9481.948, 2.7552.755 and 1.3411.341: within 0.0010.001 of scipy for the first two rows and 0.010.01 for the third. On a longer run, t∈[80,100]t \in [80, 100] gives 1.9831.983 (scipy 1.9851.985, correction 1.9831.983). The time step matters: h=0.4 Δx2/Dh = 0.4\,\Delta x^2/D reads 1.9401.940 in the first row.

The tail picks the speed. If the initial data decay slower than e−λ∗xe^{-\lambda^* x}, that tail sets the speed c(λ)c(\lambda) (from memory). Test with D=r=1D = r = 1:

initial tailpredictedsimscipy
e−0.5xe^{-0.5x}2.5002.5002.4992.4992.5002.500
e−0.7xe^{-0.7x}2.1292.1292.1272.1272.1292.129
e−2xe^{-2x}22 (not c(2)=2.5c(2) = 2.5)1.9471.9471.9491.949
◆ Interactive Fisher-KPP: change D and r, watch the front, and compare its measured speed with 2 sqrt(D r)

Loading simulation…

Linear stability with diffusion

Let u∗\mathbf u^* be a uniform steady state with Jacobian JJ. Perturb it by ε v eσtcos⁡(qx)\varepsilon\,\mathbf v\,e^{\sigma t}\cos(qx). The Laplacian turns cos⁡(qx)\cos(qx) into −q2cos⁡(qx)-q^2\cos(qx), so

σ v=(J−q2D) v.\sigma\,\mathbf v = (J - q^2 D)\,\mathbf v.

The growth rate σ(q)\sigma(q) is the largest real part of the eigenvalues of J−q2DJ - q^2D: the dispersion relation. For one species σ=f′−Dq2\sigma = f' - Dq^2, so diffusion only stabilises. For two it need not.

Take tr J<0\mathrm{tr}\,J \lt 0 and det⁡J>0\det J \gt 0 (stable without diffusion). Then tr(J−q2D)\mathrm{tr}(J - q^2D) is negative for every qq, and only a negative determinant can destabilise:

det⁡(J−q2D)=D1D2 q4−B q2+det⁡J,B=D2J11+D1J22.\det(J - q^2D) = D_1D_2\,q^4 - B\,q^2 + \det J, \qquad B = D_2J_{11} + D_1J_{22}.

An upward parabola in q2q^2 with vertex at B/(2D1D2)B/(2D_1D_2) dips below zero iff

B>2D1D2det⁡J,qc2=det⁡J/(D1D2) at threshold.B \gt 2\sqrt{D_1D_2\det J}, \qquad q_c^2 = \sqrt{\det J/(D_1D_2)} \text{ at threshold.}

This forces a sign pattern. Since B>0B \gt 0 and J11+J22<0J_{11} + J_{22} \lt 0, one diagonal entry is positive, say J11>0J_{11} \gt 0: species 1 activates itself. Then D2/D1>−J22/J11>1D_2/D_1 \gt -J_{22}/J_{11} \gt 1: the inhibitor must diffuse faster. The picture is local activation with long-range inhibition.

Turing instability of the Brusselator

Here J=(b−1a2−b−a2)J = \begin{pmatrix} b - 1 & a^2 \\ -b & -a^2 \end{pmatrix}, xx activates when b>1b \gt 1, det⁡J=a2\det J = a^2, and the threshold becomes

Dy(b−1)−Dxa2>2aDxDy  ⟺  b>(1+aDx/Dy)2,qc2=aDxDy.D_y(b - 1) - D_xa^2 \gt 2a\sqrt{D_xD_y} \;\Longleftrightarrow\; b \gt \left(1 + a\sqrt{D_x/D_y}\right)^2, \qquad q_c^2 = \frac{a}{\sqrt{D_xD_y}}.

At fixed bb, Dy/Dx>a2/(b−1)2D_y/D_x \gt a^2/(\sqrt b - 1)^2. Take a=1a = 1, Dx=1D_x = 1, Dy=10D_y = 10, b=1.9b = 1.9. The eigenvalues of JJ are −0.05±0.9987i-0.05 \pm 0.9987i, a stable spiral, and

det⁡(J−q2D)=10q4−8q2+1=0  ⇒  q2=0.1551 or 0.6449.\det(J - q^2D) = 10q^4 - 8q^2 + 1 = 0 \;\Rightarrow\; q^2 = 0.1551 \text{ or } 0.6449.

The unstable band is q∈(0.394,0.803)q \in (0.394, 0.803), wavelengths 7.827.82 to 15.9615.96. The threshold is b=1.7325b = 1.7325, qc2=0.3162q_c^2 = 0.3162, and the critical ratio at b=1.9b = 1.9 is 6.986.98. The fastest mode (found numerically) has q2=0.3216q^2 = 0.3216, σ=0.1425\sigma = 0.1425, wavelength 11.0811.08. Sim and numpy agree on all of these.

bbband in q2q^2fastest q2q^2wavelengthσmax⁡\sigma_{\max}
1.751.750.250.25 to 0.400.400.31680.316811.1611.160.01480.0148
1.801.800.200.20 to 0.500.500.31850.318511.1311.130.05720.0572
1.901.900.15510.1551 to 0.64490.64490.32160.321611.0811.080.14250.1425
1.951.950.14110.1411 to 0.70890.70890.32310.323111.0511.050.18540.1854

The band widens (about 0.150.15, 0.300.30, 0.490.49 in q2q^2, roughly like b−bc\sqrt{b - b_c}) around a nearly fixed centre, so the fastest wavelength hardly moves while σmax⁡\sigma_{\max} grows about twelvefold. And the wavelength is set by chemistry and diffusion, not by the domain: 2π/qc=2π(DxDy)1/4/a2\pi/q_c = 2\pi(D_xD_y)^{1/4}/\sqrt a, so quadrupling DyD_y stretches the pattern only by 2\sqrt2.

◆ Interactive The dispersion relation: slide b and Dy and watch the unstable band open and close; the right panel grows the pattern

Loading simulation…

A pattern from noise: linear against nonlinear

Linear theory says which modes grow, not where growth stops. The sim runs the nonlinear Brusselator on [0,100][0, 100], Δx=0.5\Delta x = 0.5 (201201 nodes), no-flux ends, b=1.9b = 1.9, from the uniform state plus noise of size 0.010.01, explicit h=0.01h = 0.01 to t=400t = 400. Each run settles into regular stationary peaks of xx, between about 0.600.60 and 1.791.79 for seed 1. I measured the mean spacing of the interior peaks for 1212 seeds and recomputed each with scipy’s LSODA from identical noise (the generator, mulberry32, is reproduced in Python): the spacings agree to three decimals.

spacingseedsallowed mode nn, wavelength 200/n200/n
≈11.7\approx 11.777 of 12121717: 11.7611.76
≈11.1\approx 11.133 of 12121818: 11.1111.11
≈10.5\approx 10.522 of 12121919: 10.5310.53

All twelve land inside the band and within 6.3%6.3\% of the linear optimum 11.0811.08; the mean is 11.3811.38. They do not coincide with it. The no-flux ends allow only q=nπ/Lq = n\pi/L, but the nearest allowed wavelength (n=18n = 18, 11.1111.11) is close, so that alone does not explain the spread. The competition is close: the linear growth rates are 0.13870.1387 (n=17n = 17), 0.14250.1425 (n=18n = 18) and 0.13970.1397 (n=19n = 19), within 3%3\%, and the nonlinear terms decide. The linear winner, n=18n = 18, was the outcome only 33 times in 1212. Linear theory gives the scale of the pattern, not an oracle for its wavelength.

Sensitivity is the flip side. At Dy=5D_y = 5, σ(q)≤−0.05\sigma(q) \le -0.05 for every qq: no pattern from any noise. At Dy=8D_y = 8 the band exists but σmax⁡\sigma_{\max} is 0.05750.0575. A pattern-former lives within a factor of two of a boundary where it forms nothing.

Computing it: the explicit limit and IMEX

Explicit FTCS needs the highest grid mode, the checkerboard (−1)i(-1)^i with q2=4/Δx2q^2 = 4/\Delta x^2, inside the Euler stability interval (10.2 Explicit Methods for the Heat Equation). For heat alone that is h≤Δx2/(2D)h \le \Delta x^2/(2D); with two diffusivities the faster binds: 0.52/(2⋅10)=0.01250.5^2/(2\cdot10) = 0.0125. Including the reaction Jacobian, the checkerboard’s most negative eigenvalue is −160.99-160.99, so h<2/160.99=0.012423h \lt 2/160.99 = 0.012423. Three checks on the full scheme:

  • Scipy’s eigenvalues of the full 402×402402\times402 discrete Jacobian give the same 0.0124230.012423.
  • The sim measures the checkerboard growth factor per step by running the scheme: 0.99620.9962 at h=0.0124h = 0.0124, 1.02841.0284 at h=0.0126h = 0.0126.
  • To t=400t = 400, h=0.0124h = 0.0124 gives the normal pattern (88 interior peaks, spacing 11.7411.74); h=0.0126h = 0.0126 blows up within about 800800 steps.

At h=0.01h = 0.01 the run takes 4000040000 steps. Halving Δx\Delta x quarters the limit and multiplies the work about eightfold.

IMEX. Treat the stiff diffusion implicitly (the tridiagonal solve of 10.4 Implicit Methods & Crank–Nicolson) and the mild reaction explicitly: u♯=un+h f(un)\mathbf u^\sharp = \mathbf u^n + h\,\mathbf f(\mathbf u^n), then (I−hD ∂xx)un+1=u♯(I - hD\,\partial_{xx})\mathbf u^{n+1} = \mathbf u^\sharp. It is the operator splitting of 15.4 Chemical Kinetics & Stiff Systems in miniature, first order in hh. At h=0.2h = 0.2 it takes 20002000 steps, 2020 times fewer, with spacing 11.77611.776 against 11.77311.773 explicit and xx range [0.6026,1.7933][0.6026, 1.7933] against [0.6026,1.7929][0.6026, 1.7929] (a dense-solve numpy version agrees). At h=0.3h = 0.3 the pattern is lost: no peaks, xx confined to [0.708,0.748][0.708, 0.748]. The splitting’s linear growth rate for the fastest mode is still 0.13910.1391 there against the exact 0.14250.1425, so mode analysis does not explain it and I did not isolate the cause. IMEX removes the Δx2\Delta x^2 limit; another limit, found by experiment, takes over.

Gray-Scott in 2D, and why the Brusselator does the analysis

The Gray-Scott model,

ut=Du∇2u−uv2+F(1−u),vt=Dv∇2v+uv2−(F+k)v,u_t = D_u\nabla^2u - uv^2 + F(1 - u), \qquad v_t = D_v\nabla^2v + uv^2 - (F + k)v,

is the standard demonstration: a feed FF and a kill rate kk give spots, labyrinths and splitting blobs. It is a poor model to analyse here. The empty state (1,0)(1, 0) has Jacobian diag(−F,−(F+k))\mathrm{diag}(-F, -(F + k)), stable for every F,k>0F, k \gt 0 and every diffusivity, so no noise grows: patterns start only when a finite seed of vv invades, like a Fisher front, and then splits. The Brusselator is the opposite: last lesson’s Jacobian, a closed-form threshold, growth from noise. So we analyse the Brusselator in 1D and play with Gray-Scott in 2D.

The sim uses Du=0.16D_u = 0.16, Dv=0.08D_v = 0.08, a periodic 128×128128\times128 grid of spacing 11, a central square seed, and h=1h = 1. In 2D the checkerboard has q2=8/Δx2q^2 = 8/\Delta x^2, so the limit is about Δx2/(4Dmax⁡)=1.5625\Delta x^2/(4D_{\max}) = 1.5625, and with the reaction 2/(8Du+F)=1.52672/(8D_u + F) = 1.5267 at F=0.03F = 0.03. Measured by running the scheme (and in numpy): checkerboard growth 0.9650.965 at h=1.5h = 1.5, 1.0301.030 at h=1.55h = 1.55, and h=1.7h = 1.7 blows up within 9898 steps. The default h=1h = 1 has margin.

FF, kkafter t=5000t = 5000after t=15000t = 15000
0.0300.030, 0.0620.0628989 blobs of v>0.2v \gt 0.2, 22.3%22.3\% of area110110 blobs, 23.2%23.2\%
0.0400.040, 0.0600.06011 connected structure, 52.9%52.9\%11 structure, 52.3%52.3\%

(Blobs are 4-connected with periodic wrap; sim and numpy/scipy agree exactly.) The first row is spots whose count is still changing at t=15000t = 15000, so it is a snapshot; the second is one connected labyrinth. Spirals and other regimes exist in the FF-kk plane (from memory, not shown here).

◆ Interactive Gray-Scott: choose a preset, change F and k, and paint v onto the field with the pointer

Loading simulation…

What this does and does not show

Turing proposed in 1952 that morphogens obeying such equations could pattern an embryo. The mathematics is exact: the band, threshold and wavelength are facts about a PDE. Whether a given animal is patterned this way is a hypothesis to be tested against biology, not against the PDE. Support is mixed (from memory): some systems have been argued to fit and others, leopard spots among them, remain interpretations, and real tissue has growth, curvature and more than two species. Our own numbers show three limits. The diffusivity ratio must exceed a threshold (6.986.98 here), hard to arrange when small molecules diffuse at similar speeds (from memory). The wavelength depends on diffusivities only through a fourth root. And noise and geometry change the outcome, as the twelve seeds showed. Nothing here was fitted to an animal.

Where this goes next

So far we have a front (Fisher) and a stationary pattern (Turing). A neuron is different: its membrane rests stably, but a large enough kick sends a stereotyped pulse down the axon, after which it recovers. That is an excitable medium, again ∂tu=D∇2u+f(u)\partial_t\mathbf u = D\nabla^2\mathbf u + \mathbf f(\mathbf u), with a fast voltage variable and a slow recovery variable that, unlike the Turing inhibitor, need not diffuse at all. The Hodgkin-Huxley and FitzHugh-Nagumo models of 15.6 Neuron Models supply f\mathbf f, and the front becomes a travelling pulse. Whether anything like 2Dr2\sqrt{Dr} survives there is the first question to ask; it should not be assumed.

The exercise: Laplacian, Fisher speed, dispersion relation and step limit

The playground first grows the 1D Brusselator pattern with explicit steps and tests the step limit.

🐍 PythonA Turing pattern in 1D, and the step limit
⌘/Ctrl + Enter

Now build the pieces yourself: the no-flux Laplacian, the Fisher front speed, the dispersion relation, the unstable band and the explicit limit.

🧪 Coding exerciseExercise: Laplacian, Fisher speed, sigma(q), Turing band and step limit
⌘/Ctrl + Enter

Concept checks

✅ Concept check

A Fisher-KPP front has $D = 4$ and $r = 0.25$. What is its asymptotic speed $2\sqrt{Dr}$?

✅ Concept check

Which combination is necessary for a Turing instability of a two-species system whose uniform state is stable without diffusion?

✅ Concept check

For the Brusselator with $a = 1$, $D_x = 1$ and $D_y = 4$, at what value of $b$ does the Turing threshold $(1 + a\sqrt{D_x/D_y})^2$ sit?

✅ Concept check

A 1D explicit reaction-diffusion run has its step limit set by the diffusion term. By what factor does the allowed step change when $\Delta x$ is halved?

✅ Concept checkselect all that apply

Select all true statements about the Gray-Scott runs in this lesson.

Problems

✏️ Pen & paperSpeed of an invasion●○○

A species spreads with D=0.09D = 0.09 and r=0.4r = 0.4 (consistent units). Find the asymptotic front speed and the time to cover 3030 length units. Is the real time shorter or longer?

Show worked solution

c∗=20.036=0.3795c^* = 2\sqrt{0.036} = 0.3795, and 30/0.3795=79.130/0.3795 = 79.1. The real time is longer: the measured speed approaches c∗c^* from below, as in the table above.

✏️ Pen & paperTuring conditions for a given Jacobian●●○

A reaction has J=(1−23−4)J = \begin{pmatrix} 1 & -2 \\ 3 & -4 \end{pmatrix} at its uniform state, with Dx=1D_x = 1. (a) Confirm it is stable without diffusion. (b) Find the smallest DyD_y for a Turing instability. (c) Find qc2q_c^2 there.

Show worked solution

(a) tr=−3<0\mathrm{tr} = -3 \lt 0, det⁡=−4+6=2>0\det = -4 + 6 = 2 \gt 0. (b) The condition is Dy−4>22DyD_y - 4 \gt 2\sqrt{2D_y}. With s=Dys = \sqrt{D_y}: s2−22 s−4>0s^2 - 2\sqrt2\,s - 4 \gt 0, so s>2+6s \gt \sqrt2 + \sqrt6 and Dy>8+43=14.93D_y \gt 8 + 4\sqrt3 = 14.93. (A numerical scan finds an unstable mode at Dy=15D_y = 15 and none at 14.914.9.) (c) qc2=2/14.93=0.3660=(3−1)/2q_c^2 = \sqrt{2/14.93} = 0.3660 = (\sqrt3 - 1)/2.

✏️ Pen & paperThe cost of resolving a pattern●●○

For the Brusselator with a=1a = 1, b=1.9b = 1.9, Dx=1D_x = 1, Dy=10D_y = 10 on [0,100][0, 100], find the explicit step limit and the steps to T=400T = 400 at Δx=0.5\Delta x = 0.5 and 0.250.25, and the ratio of work (steps times nodes).

Show worked solution

The limit is 2/∣λmin⁡∣2/|\lambda_{\min}| for J−(4/Δx2)diag(1,10)J - (4/\Delta x^2)\mathrm{diag}(1, 10): 0.0124230.012423 at Δx=0.5\Delta x = 0.5 (3219832198 steps, 201201 nodes) and 0.00312010.0031201 at 0.250.25 (128200128200 steps, 401401 nodes). The work is 6.47×1066.47\times10^6 against 5.14×1075.14\times10^7, a ratio of 7.947.94, close to 232^3: steps ×4\times4, nodes ×2\times2.

✏️ Pen & paperThe tail sets the speed●●●

For ut=uxx+u(1−u)u_t = u_{xx} + u(1 - u) the tail e−λxe^{-\lambda x} gives c(λ)=λ+1/λc(\lambda) = \lambda + 1/\lambda. (a) Show c(λ)≥2c(\lambda) \ge 2, with equality only at λ=1\lambda = 1. (b) Predict the speed for data decaying like e−0.5xe^{-0.5x} and like e−2xe^{-2x}. (c) The sim gave 2.4992.499 and 1.9471.947. Do these support (b)?

Show worked solution

(a) c(λ)−2=(λ−1)2/λ≥0c(\lambda) - 2 = (\lambda - 1)^2/\lambda \ge 0. (b) For λ=0.5<1\lambda = 0.5 \lt 1 the tail carries the front at c(0.5)=2.5c(0.5) = 2.5; for λ=2>1\lambda = 2 \gt 1 the front selects the minimum, 22, not c(2)=2.5c(2) = 2.5 (both from memory). (c) Yes: 2.4992.499 matches 2.52.5, and 1.9471.947 is near 22 (below it, as in the table) and far from 2.52.5. One window is evidence, not proof.

Key Takeaways

  • Adding D∇2D\nabla^2 to a reaction gives a PDE; one diffusing species only smooths, but two with unequal diffusivities can create structure.
  • A Fisher-KPP front is pulled by its linear tail: c(λ)=Dλ+r/λc(\lambda) = D\lambda + r/\lambda is minimal at λ∗=r/D\lambda^* = \sqrt{r/D}, giving 2Dr2\sqrt{Dr}; measured speeds are 3%3\% to 6%6\% below it at t=20t = 20 to 4040 and reach 1.9831.983 near t=90t = 90.
  • A mode cos⁡(qx)\cos(qx) grows at the top eigenvalue of J−q2DJ - q^2D; only det⁡(J−q2D)<0\det(J - q^2D) \lt 0 can destabilise a state stable without diffusion.
  • Turing needs an activator and a faster inhibitor: b>(1+aDx/Dy)2b \gt (1 + a\sqrt{D_x/D_y})^2 for the Brusselator, band q2∈(0.155,0.645)q^2 \in (0.155, 0.645) and fastest wavelength 11.0811.08 at Dy=10D_y = 10.
  • The nonlinear wavelength lands in the band and near, not at, the fastest mode: twelve seeds gave 10.510.5, 11.111.1 or 11.711.7, at most 6.3%6.3\% away.
  • Explicit schemes need h<2/∣λmin⁡∣h \lt 2/|\lambda_{\min}| of the grid checkerboard, about Δx2/(2Dmax⁡)\Delta x^2/(2D_{\max}): 0.0124230.012423 here, measured as growth factors 0.99620.9962 and 1.02841.0284.
  • IMEX took h=0.2h = 0.2 with 20×20\times fewer steps and the same pattern, but failed at h=0.3h = 0.3 for a reason I did not isolate.
  • Gray-Scott patterns need a finite seed because the empty state is always stable, and have long transients (8989 then 110110 spots).
  • Turing’s mechanism as the origin of real animal markings is a hypothesis; the mathematics is exact, the biology unsettled.
Finished this lesson?

Mark it complete to track your progress (saved in this browser). Reading every section does this for you.

Where this leads

These later lessons build directly on this one: