--- title: Proudman–Johnson Equation url: https://www.emergentmind.com/topics/proudman-johnson-equation type: topic --- # Proudman–Johnson Equation The Proudman–Johnson equation denotes a family of nonlinear one-dimensional evolution equations that appear in several closely related normal forms, and whose modern analysis centers on transport–stretching dynamics, flow-map degeneration, and finite-time singularity formation. In contemporary PDE literature, the term usually refers to the generalized inviscid family \[ u_{txx}+u\,u_{xxx}-a\,u_xu_{xx}=0, \] or equivalent parameterizations such as \[ u_{txx}+(1+2\lambda)u_xu_{xx}+u\,u_{xxx}=0, \] together with periodic, Dirichlet, or non-periodic boundary conditions; the name also persists in the older reversed-stagnation-point literature for a distinct similarity-reduced equation of boundary-layer type [1902.03787] [2101.03601] [1302.2391]. ## 1. Nomenclature and normal forms A standard modern formulation is the generalized Proudman–Johnson equation \[ u_{txx}+u\,u_{xxx}-a\,u_xu_{xx}=0, \] studied on either \([0,1]\) with Dirichlet boundary conditions or on the periodic domain with mean-free periodic boundary conditions. In this convention, the classical Proudman–Johnson equation is the case \(a=1\); the same paper records \(a=-3\) as Burgers and \(a=-2\) as Hunter–Saxton [1902.03787]. A second common convention writes the generalized inviscid family on \(\mathbb R\) or \(S^1\) as \[ u_{txx}+(1+2\lambda)u_xu_{xx}+u\,u_{xxx}=0. \] In the non-periodic theory, this is paired with the identification \(\lambda=\frac1r\in(0,1)\), which links the equation to the \(r\)-Hunter–Saxton equation [2101.03601]. A third convention, used in an information-geometric formulation, is \[ u_{txx}+(2-\alpha)u_xu_{xx}+u\,u_{xxx}=0, \] where \(\alpha=3\) gives the original Proudman–Johnson equation, \(\alpha=0\) gives the Hunter–Saxton equation, and \(\alpha=-1\) gives the \(\mu\)-Burgers equation [2508.00371]. For periodic problems, an integrated form is often more convenient: \[ u_{xt}+u\,u_{xx}-\lambda u_x^2=I(t),\qquad I(t)=-(\lambda+1)\int_0^1 u_x^2\,dx, \] with periodic boundary conditions on \(u\) and \(u_x\) [1306.4437]. In this formulation, the nonlocal term \(I(t)\) enforces periodic compatibility. The older aerodynamic usage is different. In the reversed stagnation-point setting, the similarity-reduced equation associated with Proudman and Johnson’s 1962 analysis is \[ f_{\eta\tau}-(f_\eta)^2+f f_{\eta\eta}-f_{\eta\eta\eta}=-1, \] with \[ f(0,\tau)=f_\eta(0,\tau)=0,\qquad f_\eta(\infty,\tau)=1. \] That equation arises from a stream-function ansatz for two-dimensional reversed stagnation-point flow near a flat wall [1302.2391]. This suggests that “Proudman–Johnson equation” functions partly as a historical label spanning more than one reduction framework. ## 2. Characteristic and flow-map formulations The modern theory is largely driven by Lagrangian reduction. For periodic generalized inviscid Proudman–Johnson, \[ \dot\gamma(\alpha,t)=u(\gamma(\alpha,t),t),\qquad \gamma(\alpha,0)=\alpha, \] and \[ \dot\gamma_\alpha=u_x(\gamma(\alpha,t),t)\,\gamma_\alpha. \] Introducing \[ \mathcal J(\alpha,t)=1-\lambda \eta(t)u_0'(\alpha), \] together with \[ \mathcal K_i(\alpha,t)=\frac{1}{\mathcal J(\alpha,t)^{i+\frac1\lambda}}, \qquad \overline{\mathcal K}_i(t)=\int_0^1\frac{d\alpha}{\mathcal J(\alpha,t)^{i+\frac1\lambda}}, \] one obtains \[ \gamma_\alpha(\alpha,t)=\frac{\mathcal K_0(\alpha,t)}{\overline{\mathcal K}_0(t)} \] and the representation formula \[ u_x(\gamma(\alpha,t),t)= \frac{1}{\lambda \eta(t)\,\overline{\mathcal K}_0(t)^{2\lambda}} \left( \frac{1}{\mathcal J(\alpha,t)}-\frac{\overline{\mathcal K}_1(t)}{\overline{\mathcal K}_0(t)} \right). \] The same framework gives \[ u_{xx}(\gamma(\alpha,t),t)=u_0''(\alpha)\,\gamma_\alpha(\alpha,t)^{2\lambda-1}, \] so the sign pattern of \(u_{xx}\) is transported by the flow [1306.4437]. A complementary flow-map formulation on \([0,1]\) or the periodic domain starts from \[ F_t(\xi,t)=u(F(\xi,t),t),\qquad F(\xi,0)=\xi, \] with \(F_\xi>0\). Along trajectories, \[ \hat u_{xx}(\xi,t)=u_{xx}^0(\xi)\,F_\xi(\xi,t)^a. \] For \(a\neq -1\), setting \(f_a=F_\xi^{a+1}\) yields the inhomogeneous Liouville equation \[ \frac{\partial^2}{\partial t\,\partial \xi}\log f_a(\xi,t) =(a+1)u_{xx}^0(\xi)\,f_a(\xi,t). \] This can be solved explicitly up to a scalar auxiliary function \(\eta(t)\), leading to \[ F_\xi(\xi,t)= \big(\eta'(t)\big)^{\frac{1}{a+1}} \left[ 1-\eta(t)\frac{a+1}{2}u_x^0(\xi) \right]^{-\frac{2}{a+1}}. \] In this formulation, continuation is equivalent to preserving the diffeomorphism property \(F_\xi>0\); breakdown occurs when \[ 1-\eta(t)\frac{a+1}{2}u_x^0(\xi)=0 \] for some \((\xi,t)\) [1902.03787]. ## 3. Blow-up, global existence, and the role of initial curvature For smooth periodic data with quadratic behavior near extrema, the generalized inviscid Proudman–Johnson equation admits a sharp parameter-dependent classification. In the \(\lambda\)-formulation, solutions are global for \(\lambda\in[0,1]\). For \(\lambda>1\), finite-time blow-up is two-sided and everywhere in the sense that \[ M(t)\to+\infty,\qquad m(t)\to-\infty, \] and for every non-extremal label \(\alpha\), \(u_x(\gamma(\alpha,t),t)\to-\infty\). For \(\lambda\in(-2,0)\), blow-up is one-sided and discrete: only the minimum diverges. For \(\lambda\le -2\), blow-up is again two-sided and everywhere, but now non-minimizing points diverge to \(+\infty\) while minimizing points diverge to \(-\infty\) [1306.4437]. Norm growth is more delicate than \(L^\infty\)-blow-up. For \(p>1\), \(\|u_x\|_p\to+\infty\) for \(\lambda\in\mathbb R\setminus(-2,1]\), and also for \(\lambda\in(-2,-2/p]\). The energy \[ E(t)=\|u_x(\cdot,t)\|_2^2 \] remains finite for \(\lambda\in(-2/3,1]\) and diverges for \(\lambda\in\mathbb R\setminus(-2/3,1]\), so one-sided blow-up on \((-2/3,0)\) can remain compatible with finite \(L^2\)-energy up to blow-up time [1306.4437]. The curvature of the initial slope near its extrema refines these thresholds. If, near each maximizing point, \[ u_0'(\alpha)\sim M_0+C_1|\alpha-\overline\alpha_i|^q,\qquad C_1<0,\quad q>0, \] then the decisive positive-\(\lambda\) threshold is \[ \lambda=\frac q2. \] Solutions are global for \(0\le \lambda\le q/2\); if \(0<\lambda1\) and \[ \lambda<\frac{q}{1-q}, \] the blow-up becomes two-sided and everywhere [1306.4449]. For smooth data with \(u_0''\) vanishing to order \(k\) at the relevant extremum, \(q=k+1\), so the positive threshold becomes \(\lambda=(k+1)/2\) [1306.4449]. An alternative interval/periodic flow-map theory in the \(a\)-parameterization gives global existence for \(a=-1\), global existence for \(-1-1\) under the smallness condition \[ u_{\max}<\frac{1}{\sqrt{1+a}\,\|u_x^0\|_{L^2(0,1)}}. \] The same theory gives a blow-up criterion for \(a\ge1\) in terms of square-integrability of \[ \psi(\eta)=\int_0^1\frac{ds}{1-\frac{a+1}{2}\eta u_x^0(s)} \] on \[ \left[0,\frac{2}{(a+1)u_{\max}}\right] \] [1902.03787]. ## 4. Damping and nonstandard boundary conditions A damped version of the generalized inviscid Proudman–Johnson equation has been studied on \([0,1]\) under the homogeneous three-point boundary condition \[ u(1,t)=u_x(0,t)=u_x(1,t)=0. \] The equation is \[ u_{xt}+u\,u_{xxx}+\beta\,u_xu_{xx}+a(t)u_{xx}=0, \] with \[ \beta=\frac{2-n}{n},\qquad n\in\mathbb Z_+,\quad n\ge2. \] Writing \(H(t)=u(0,t)=-\int_0^1u_x(x,t)\,dx\), one obtains the identity \[ H'(t)+a(t)H(t)+\frac{n-1}{n}\int_0^1|u_x(x,t)|^2\,dx=0, \] and therefore the Riccati-type inequality \[ H'(t)+a(t)H(t)+\frac{n-1}{n}H(t)^2\le0. \] This scalar reduction is the core of the blow-up mechanism [2106.00068]. For bounded smooth damping, if \[ M=\sup_{t\ge0}a(t),\qquad H_0=u_0(0)<\frac{M(1-n)}{n}, \] then the solution blows up in finite time at \(x=0\), \[ \lim_{t\uparrow t^*}u(0,t)=-\infty, \] with explicit upper bound \[ t^*=-\frac{1}{M}\ln\!\left(1+\frac{M(1-n)}{nH_0}\right). \] The result shows that bounded damping shifts the threshold quantitatively but does not prevent blow-up for sufficiently negative boundary data [2106.00068]. The same paper proves that even unbounded smooth damping may fail to arrest singularity formation. For the explicit choice \[ a(t)=e^{ct},\qquad c>0, \] if \[ u_0(0)<-\frac{c(n-1)}{n e^{1/c}E_1(1/c)}, \] where \(E_1\) is the exponential integral, then there exists a finite \(t^*>0\) such that \[ \lim_{t\uparrow t^*}u(0,t)=-\infty. \] The proved singularity is boundary-value blow-up of \(u(0,t)\) itself, rather than a derivative-only blow-up scenario [2106.00068]. ## 5. Geometric and variational interpretations On the real line, the non-periodic generalized inviscid Proudman–Johnson equation is formally equivalent to the \(r\)-Hunter–Saxton equation under \[ \lambda=\frac1r,\qquad r\in(1,\infty). \] In this regime, it is the Eulerian geodesic equation of a right-invariant homogeneous \(W^{1,r}\)-Finsler metric on \[ \mathrm{Diff}_{-\infty}(\mathbb R) = \left\{\varphi=\mathrm{id}+f:\ f'\in W^{\infty,1}(\mathbb R),\ f'>-1,\ \lim_{x\to-\infty}f(x)=0\right\}. \] The decisive linearizing map is \[ \Phi(\varphi)=r\big(\varphi_x^{1/r}-1\big), \] which is an isometric embedding into an open convex subset of \(W^{\infty,1}(\mathbb R)\). Geodesics therefore become straight lines, and the explicit solution formula is \[ \varphi(t,x)=x+\int_{-\infty}^x\left(\left(1+\frac{t\,u_0'(y)}{r}\right)^r-1\right)\,dy. \] The maximal positive existence time is \[ T^*(u_0)=\inf_{x\in\mathbb R}\frac{-r}{u_0'(x)} \] whenever some \(u_0'(x)<0\); otherwise the solution exists for all \(t>0\). The same work emphasizes that the equivalence with \(r\)-Hunter–Saxton fails on the circle [2101.03601]. A different geometric interpretation uses information geometry. On the space of positive densities, the Amari–Čencov \(\alpha\)-connections \(\nabla^{(\alpha)}\) are realized as Levi-Civita connections of explicit metrics \(G^\alpha\). Pulling \(G^\alpha\) back by \[ \Theta:\mathcal D(\mathbb R)\to\mathcal M,\qquad \varphi\mapsto\varphi_*dx \] gives the metric \[ G^{\alpha\dot H^1}_{\varphi}(u\circ\varphi,v\circ\varphi) = \int \varphi_x^{-\alpha-1}(u\circ\varphi)_x(v\circ\varphi)_x\,dx, \] and the corresponding Euler–Arnold equation is \[ u_{txx}+(2-\alpha)u_xu_{xx}+u\,u_{xxx}=0. \] Here the metric is right-invariant iff \(\alpha=0\); for \(\alpha\neq0\), the generalized Proudman–Johnson equation is the Euler–Arnold equation of a non-right-invariant metric. The same theory states that these generalized Proudman–Johnson equations are globally well-posed on \(\mathcal D(\mathbb R)\) iff \(\alpha=1\) [2508.00371]. ## 6. Self-similar regimes and recent developments Recent periodic work near the classical threshold \(a=1\) studies \[ u_{txx}+u\,u_{xxx}=a\,u_xu_{xx}+\nu u_{xxxx}, \qquad x\in\mathbb T=[-\pi,\pi], \] under odd symmetry. In this convention, \(a=1\) is the classical Proudman–Johnson equation. There exists \(\delta_1>0\) such that for some \(C^\infty\) initial data the inviscid equation \((\nu=0)\) blows up in finite time when \(a\in(1,1+\delta_1)\). The same paper constructs self-similar solutions \[ \omega(x,t)=\frac{1}{1+c_{\omega,a}t}\,\omega_a(x), \qquad \omega=u_{xx}, \] for \(a\in(1-\delta_1,1+\delta_1)\), with \(c_{\omega,a}<0\) in the supercritical range \(a>1\), \(c_{\omega,a}=0\) at \(a=1\), and \(c_{\omega,a}>0\) for \(a<1\), yielding \(O(t^{-1})\) decay in the subcritical regime. The same analysis also proves finite-time self-similar blow-up for some \(C^\alpha\) data at the critical case \(a=1\), and finite-time blow-up in the viscous case \(\nu>0\) for \(a>1\) sufficiently close to \(1\) [2511.15166]. A separate recent development uses the periodic mean-free equation \[ \partial_t a+ \left(\int_{-\pi}^x a(t,\bar x)\,d\bar x\right)\partial_x a -a^2 +\frac{1}{\pi}\int_{-\pi}^{\pi}a^2\,dx =0, \qquad \int_{-\pi}^{\pi}a(t,x)\,dx=0, \] as the renormalized form of a blow-up problem for the incompressible porous medium equation. Under the corresponding change of variables, the explicit blow-up profile becomes the steady state \[ a(t,x)=\mu\cos x. \] The stationary \(C^2\) solutions are completely classified: they are exactly \[ a=\mu\cos(kx) \quad\text{or}\quad a=\mu\sin\!\left(\frac{2k+1}{2}x\right). \] For every \(\mu>0\), small \(C^3\) perturbations of \(\mu\cos x\) are globally stable and satisfy \[ \|a(t,\cdot)-\mu_*\cos(\cdot)\|_{C^1([-\pi,\pi])} \lesssim \sigma e^{-\mu t/2}, \] where \[ \mu_*=-\partial_x^2 a_0(x_0^*) \] is selected by the initial curvature at the point of maximum. By contrast, for every \(\epsilon,\sigma>0\) there exist perturbations arbitrarily small in \(C^{2-\epsilon}\) for which the solution cannot converge in \(L^\infty\) to any \(\mu\cos x\). This identifies a sharp regularity threshold between \(C^{2+\epsilon}\)-type stability and \(C^{2-\epsilon}\)-instability, and it simultaneously yields corresponding stability statements for associated special classes of solutions of 2D Euler and inviscid primitive or hydrostatic Euler [2507.17381]. Source: https://www.emergentmind.com/topics/proudman-johnson-equation