Abstract
We will prove that solutions of the Allen–Cahn equations that satisfy the equipartition of the energy can be transformed into solutions of the Euler equations with constant pressure. As a consequence, we obtain De Giorgi type results, that is, the level sets of entire solutions are hyperplanes. Also, we will determine the structure of solutions of the Allen–Cahn system in two dimensions that satisfy the equipartition. In addition, we apply the Leray projection on the Allen–Cahn system and provide some explicit entire solutions. Finally, we obtain some examples of smooth entire solutions of the Euler equations. For specific type of initial conditions, some of these solutions can be extended to the Navier–Stokes equations. The motivation of this paper is to find a transformation that relates the solutions of the Allen–Cahn equations to solutions of the minimal surface equation of one dimension less. We prove this result for equipartitioned solutions in dimension three.
Similar content being viewed by others
1 Introduction
As it is well known, De Giorgi in 1978 [11] suggested a stricking analogy of the Allen Cahn equation \( \Delta u =f(u) \) with minimal surface theory that led to significant developments in Partial Differential equations and the Calculus of Variations, by stating the following conjecture about bounded solutions on \( {{\mathbb {R}}}^n\):
Conjecture: (De Giorgi) Let \( u \in C^2({{\mathbb {R}}}^n) \) be a solution to
such that: 1. \( |u|<1 \), 2. \( \dfrac{\partial u}{\partial x_n} >0 \;\, \forall x \in {{\mathbb {R}}}^n\).
Is it true that all the level sets of u are hyperplanes, at least for \( n \le 8\)?
De Giorgi’s conjecture refers to an analogy between diffused interfaces and minimal surfaces. The relationship with the Bernstein problem for minimal graphs is the reason why \( n \le 8 \) appears in the conjecture.
The first partial results on the De Giorgi conjecture was established by Modica and Mortola in [21] and [20].
In 1997 Ghoussoub and Gui in [16] proved the De Giorgi conjecture for \( n=2 \). Building on [16], Ambrosio and Cabre in [3] proved the conjecture for \( n=3 \). Also, Ghoussoub and Gui showed in [17] that the conjecture is true for \( n=4,5 \) for special class of solutions that satisfy an anti-symmetry condition.
In 2003 the conjecture was proved up to \( n=8 \) by Savin in [27], under the additional hypothesis: \( \lim _{x_n \rightarrow \pm \infty } u(x',x_n) = \pm 1 \), using entirely different methods.
Finally, Del Pino, Kowalczyk and Wei in [12] gave a counterexample to the De Giorgi’s conjecture for \( n \ge 9 \). This counterexample satisfies also the limiting assumption \( \lim _{x_n \rightarrow \pm \infty } u(x',x_n) = \pm 1. \) The construction is based on a careful perturbation argument building on the Bombieri, De Giorgi and Giusti [4] result for minimal surfaces.
The relation of the Allen–Cahn with minimal surfaces can be seen via the theory of \( \Gamma \)-convergence (see [21, 23] and [26] for further details). The family of functionals
\( \Gamma \)-converges as \( \varepsilon \rightarrow 0 \) to the perimeter functional and the Euler-Lagrange equations are
therefore one expects that the level sets of the minimizers will minimize the perimeter.
So, one question could be, whether there exists a transformation that transforms the Allen–Cahn equation \( \Delta u = f(u) \;\, (u:\Omega \subset {{\mathbb {R}}}^n \rightarrow {\mathbb {R}} \)) to the minimal surface equation of one dimension lower (i.e. \( (n-1) \)-dimensional minimal surface equation). The answer is positive for the class of solutions that satisfy the equipartition, at least in dimension 3, by Corollary 3.1 and then by applying a Bernstein-type theorem for the minimal surface equation (see [9, 13]) we obtain that the level sets of solutions are hyperlanes.
For bounded entire solutions of the Allen–Cahn equation that satisfy the equipartition holds a more general result (see Theorem 5.1 in [8]), that is, the level sets of entire solutions of the Allen Cahn equations that satisfy the equipartition are hyperplanes. This was already known by Modica and Mortola in 1980 (see final remark in [20]). In fact, any solution of the Allen–Cahn equation is smooth and satisfies the bound \( | u | \le 1 \) (see Proposition 1.9 in [14]). The point in Corollary 3.1 is that, we can obtain that the level set of solutions are hyperplanes in any open, convex domain with the appropriate boundary conditions, utilizing the result in [13].
As we can see in Appendix B, we propose a De Giorgi type property for the 2D Euler equations. The relations between different classes of equations, allow us to obtain some explicit smooth entire solutions for the 2D and 3D Isobaric Euler equations. Those solutions can be extended when the pressure is linear function in the space variables. Some of these solutions have linear dependent components. Thus, if we impose linear dependency of the components of the initial conditions, we can obtain some explicit entire solutions and can be extended to other type of equations. In Appendix C we give some examples of smooth entire solutions of the Navier–Stokes equations with linear dependent components of the initial conditions.
One of the observations in this paper, is to view the equipartition as the Eikonal equation. As stated in Proposition 2.1, the Eikonal equation can be transformed to the Euler equations with constant pressure (without the divergence free condition). Thus, solutions of the Allen Cahn equations that satisfy the equipartition can be transformed into the Euler equations with constant pressure, and we obtain the divergence free condition from the Allen Cahn equations. This observation plays a crucial role in the proof of Corollary 3.1, which was the initial motivation of this work.
Furthermore, we state this result to the equation \( a(u) \Delta u + b(u) | \nabla u|^2 = c(u) \), under the hypothesis that \( u = \Phi (v) \) for some v that is also in this class of equation. This hypothesis is quite reasonable since the equation \( a(u) \Delta u + b(u) | \nabla u|^2 = c(u) \) is invariant under such transformations, in the sence that if u is a solution then \( v= F(u) \) is also in this class of equations.
In the last section, we propose an analogue of a De Giorgi type result for the vector Allen–Cahn equations and we will prove that entire solutions of the Allen–Cahn system in dimension 2 that satisfy the equipartition have such a specific structure. Finally, we apply the Helmholtz-Leray decomposition in the Allen–Cahn system and obtain an equation, independent from the potential W. Then we apply the Leray projection (i.e. only the divergence free term from the decomposition) and we can determine explicit entire solutions. In Appendix A, we give some examples of such solutions and compare them to the structure we have obtained from Theorem 4.1. One such example, for a particular potential \( W \ge 0 \) with finite number of global minima has the property that \( \lim _{x \rightarrow \pm \infty } u(x,y)= a^{\pm } \), where \( a^{\pm } \in \lbrace W=0 \rbrace \) and \( \lim _{y \rightarrow \pm \infty } u(x,y) = U^{\pm } (x) \) where \( U^{\pm } \) are heteroclinic connections of the system (i.e. \( U^{\pm ''} = W_u(U^{\pm } ) \)). If fact, we can have infinitely many such solutions.
2 The Allen–Cahn equation and the equipartition
2.1 The equipartition of the energy and the Euler equations
We begin with a brief discussion on the equipartition of the energy. Let \( u: {\mathbb {R}} \rightarrow {\mathbb {R}}^m \) be a minimizer of the energy functional
where \( W: {\mathbb {R}}^m \rightarrow {\mathbb {R}} \) is a \( C^2 \) function (the potential energy) such that \( W >0 \) on \( {\mathbb {R}}^m {\setminus } \lbrace a^+,a^- \rbrace ,\; W_u (u) = ( \frac{\partial W}{\partial u_1},\ldots , \frac{\partial W}{\partial u_m} )^T \).
Then u will satisfy
We are interested in connecting the phases \( u_1 = a^-,\; u_2 =a^+,\, W(a^{\pm }) = 0 \). Consider also the length functional
which is invariant under the group of orientation preserving diffeomorphisms \( \psi : {\mathbb {R}} \rightarrow {\mathbb {R}},\, \psi ' > 0 \), that is, \( L( u \circ \psi ) = L(u). \)
So it holds that \( L (u) \le J(u) \) and equality holds when we have equipartition of the energy (or u is equipartitioned), that is,
In the case of heteroclinic connections (i.e. \( u: {\mathbb {R}} \rightarrow {\mathbb {R}}^m \) minimizer of J such that \( \lim _{x \rightarrow \pm \infty }u(x) = a^{\pm } \)) the equipartition of the energy holds (see also Theorem 2.1 in [2]).
More generally, let u be a solution of
the equipartition of the energy takes the form
However, when dealing with solutions \( u: {\mathbb {R}}^n \rightarrow {\mathbb {R}}^m \) of (2.5), even in the scalar case \( m=1 \), the equipartition of the energy does not hold for all solutions in general. In the scalar case, Modica in [22] proved a gradient bound \( \frac{1}{2} | \nabla u |^2 \le W(u) \) for entire solutions \( u: {\mathbb {R}}^n \rightarrow {\mathbb {R}} \). If equality holds even at a single point, then it holds for every \( x \in {\mathbb {R}}^n \) and the solutions will be one dimensional (see Theorem 5.1 in [8]). Therefore saddle-shaped solutions constructed in [5] or the counterexample for the De Giorgi’s conjecture in [12] do not satisfy the equipartition of the energy. For solutions of the Allen–Cahn equation that satisfy the equipartition in an arbitrary domain \( \Omega \subset {\mathbb {R}}^n \) there is no such a characterization in general.
In the vector case things are far more complicated, and there are examples that violate even the Modica inequality, see Sect. 2 in [28]. Nevertheless, in contrast to the scalar case, we can have a wide variety of solutions that satisfy the equipartition. In the last section we analyze the structure of solutions to the Allen–Cahn system that satisfy the equipartition and in Appendix A we provide some examples of solutions related to that structure.
We now illustrate a transformation that relates the Eikonal equation and the Euler equation with constant pressure and without the incompressibility condition. Note that \( x_n \) plays the role of the “time parameter” and \( x_n \in {\mathbb {R}} \) instead of \( x_n >0 \). We could choose any of \( x_i,\; i=1,\ldots ,n,\; n \ge 2 \) as a “time parameter”, supposing the monotonicity condition with respect to \( x_i \).
Proposition 2.1
Let \( v: \Omega \subset {{\mathbb {R}}}^n \rightarrow {\mathbb {R}} \) be a smooth solution of
where \( G: {\mathbb {R}}\rightarrow {\mathbb {R}} \) is a smooth function and suppose that \( v_{x_n} >0 \).
Then the vector field \( F=(F_1,\ldots ,F_{n-1}) \) where \( F_i = \dfrac{v_{x_i}}{v_{x_n}},\; i=1,\ldots ,n-1 \) satisfies the Euler equations
Proof
Differentiating (2.7) over \( x_i \) gives
Now we have
Thus, by (2.10) and (2.11) (for \( i=1,\ldots ,n-1 \)), we have
finally, by (2.9), the last equation becomes
\(\square \)
Remark 2.2
Note that since \( v_{x_n} >0 \) it holds that \( v( \Omega ) \cap \lbrace G = 0 \rbrace = \emptyset \). Indeed, if \( v(x_0) \in \lbrace G = 0 \rbrace \Rightarrow | \nabla v(x_0) |^2 = 0 \) which contradicts \( v_{x_n} >0 \). So, by setting \( {\tilde{v}} = P(v) \), where \( P'(v) = \frac{1}{\sqrt{G(v)}} \) we have \( \nabla {\tilde{v}} = P'(v) \nabla v \Rightarrow | \nabla {\tilde{v}}|^2 = (P'(v))^2 | \nabla v |^2 \Rightarrow | \nabla {\tilde{v}}|^2 = 1. \) Thus \( {\tilde{v}} \) satisfies \( | \nabla {\tilde{v}}|^2 =1 \) and \( F_i = \dfrac{v_{x_i}}{v_{x_n}} = \dfrac{{{\tilde{v}}}_{x_i}}{{{\tilde{v}}}_{x_n}}. \) So, at first, it seems that this transformation can be inverted: \( F_1^2 +... + F_{n-1}^2 = \dfrac{{{\tilde{v}}}_{x_1}^2 +... + {{\tilde{v}}}_{x_{n-1}}^2}{{{\tilde{v}}}_{x_n}^2} = \dfrac{1}{{{\tilde{v}}}_{x_n}^2} -1 \Rightarrow {{\tilde{v}}}_{x_n} = \dfrac{1}{\sqrt{F_1^2 +... + F_{n-1}^2 +1 }} \)
That is, if \( F_i,\; i=1,\ldots ,n-1 \) satisfies the Euler equations \( F_{x_n} +F {\nabla }_y F =0 \), then v defined by (2.15) will satisfy the Eikonal equation. This statement is true for \( n=2 \) (see [25]). But to generalize for \( n \ge 3 \) it appears that further assumptions are needed. So, the class of solutions of the Euler equations with constant pressure seem to be “richer” in some sense than the class of solutions of the Eikonal equation, that is, for every smooth solution of the Eikonal equation, we can obtain a solution of the Euler equation, but not vice versa.
Theorem 2.3
Let \( u,v: \Omega \subset {{\mathbb {R}}}^n \rightarrow {\mathbb {R}} \) such that \( u_{x_n} >0 \) satisfy the equations
and suppose that \( u = \Phi (v) \) for some \( \Phi : {\mathbb {R}} \rightarrow {\mathbb {R}} \;( \Phi ' \ne 0) \) and \( p(t) \ne 0,\; a(t) \ne 0 \), where \(p(t):= k(t) a( \Phi (t)) \Phi ''(t) + k(t) b( \Phi (t)) ( \Phi '(t))^2 - l(t) a( \Phi (t) ) \Phi '(t) \, \).
Then the vector field \( F=(F_1,\ldots ,F_{n-1}) \) defined as \( F_i = \dfrac{u_{x_i}}{u_{x_n}}, i=1,..,n-1 \), will satisfy the Euler equations
Also, \(div_y F = 0 \) if and only if \( \Phi \) is a solution of the ODE
where \( G(t):= \dfrac{k(t) f( \Phi (t)) - g(t) a( \Phi (t)) \Phi '(t)}{p(t)} \;\;\; (p \) as defined above)
Proof
We have \( u= \Phi (v) \) and \( \nabla u = \Phi ' (v) \nabla v \), therefore
since u is a solution of \( a(u)\Delta u + b(u) | \nabla u|^2 = f(u) \).
Now, since v is also solution of the second equation in (2.16), we have
where \( p(v) = k(v) a( \Phi (v)) \Phi ''(v) + k(v) b( \Phi (v)) ( \Phi '(v))^2 - l(v) a( \Phi (v) ) \Phi '(v) \).
By hypothesis \( p \ne 0 \), thus
where
Also note that \( F_i = \dfrac{u_{x_i}}{u_{x_n}} = \dfrac{v_{x_i}}{v_{x_n}}\).
So we apply Proposition 2.1 and we obtain that
Now, for the divergence of F:
Thus, from (2.19) and (2.20) the equation (2.23) becomes:
Therefore
\(\square \)
Notes: (1) It also holds that solutions of the Allen Cahn equations that satisfy the equipartition also satisfy \( div(\dfrac{\nabla u}{| \nabla u|}) =0 \) has been proved for more general type of equations (see Proposition 4.11 in [10]).
(2) We could see the fact that \( div_y F=0 \), can alternatively be obtained with calculations utilizing the stress-energy tensor (see [2], p.88), applied in the scalar case.
(3) If \( u: \Omega \subset {\mathbb {R}}^2 \rightarrow {\mathbb {R}} \), then Theorem 2.3 implies that \( F = \frac{u_x}{u_y} \) is solution of \( F_y + F F_x =0 \) and in addition \( F_x=0 \). This gives that the level sets of u are hyperplanes in any open and connected domain in \( {\mathbb {R}}^2 \). This property of solutions was known to hold for entire solutions in the case where \( u=v,\, a(u)=1=l(u), \, b(u)=0=k(u) \) and \( g'(u)=2f(u) \) to our setting (see [8, 20]).
3 Applications
As we will see now, in dimension 3 we can transform the Allen Cahn equation for the class of solutions that satisfy the equipartition of the energy, into the minimal surface equation of dimension 2 and then apply Bernstein’s result to conclude that the level sets of the solution are hyperplanes. The one dimensionality of entire solutions that satisfy the equipartition is a special case of Theorem 5.1 in [8] and it was also known by Modica and Mortola (see the final remark in [20]). However, the result in Corollary 3.1 below holds for any open subset of \( {\mathbb {R}}^n \), so by imposing the appropriate boundary conditions, utilizing the result in [13], we can obtain the result for any convex domain.
Corollary 3.1
Let \( u \in C^2 (\Omega ; {\mathbb {R}}) \) be a solution of \( \Delta u = W'(u) \) such that \( u_z >0 \), where \( \Omega \subset {\mathbb {R}}^3 \) is an open, convex set. If u satisfies
then there exists a function \( \psi \) such that \( \psi _y =- \dfrac{u_x}{u_z},\; \psi _x = \dfrac{u_y}{u_z} \) that satisfies the minimal surface equation
In particular, if \( \Omega = {{\mathbb {R}}}^3 \) or if \( \Omega \subset {\mathbb {R}}^3 \) and \( u_x=au_z,\; u_y =bu_z \) in \( {\mathbb {R}}^3 {\setminus } \Omega \), then the level sets of u are hyperplanes.
Proof
From Theorem 2.3 we have that \( div_{(x,y)} F =0 \), thus there exists some \( \psi = \psi (x,y,z) \,: F_1 = -\psi _y \) and \( F_2 = \psi _x \,.\)
As we noted in Remark 2.1, \( u(\Omega ) \cap \lbrace W = 0 \rbrace = \emptyset \, \) (by (3.1) and since \( u_z >0 \)).
So we set \( v = G(u) \), with \( G'(u) = \dfrac{1}{\sqrt{2W(u)}} \), thus
and \( v_x = F_1 v_z = \dfrac{F_1}{\sqrt{F_1^2 + F_2^2 +1}} \)
Also, by Proposition 2.1, F satisfy
and therefore, from the fact that \( v_{zx}=v_{xz} \) since \( v \in C^2( \Omega ) \), we obtain
Finally, if \( \Omega = {\mathbb {R}}^n \), by Berstein’s theorem (see Theorem 1.21 [9]) \( \psi \) must be a plane (in respect to the variables (x, y) , since \( \psi _{xx}= \psi _{xy} = \psi _{yy}=0 \)): \( {\psi }_x = b(z) \) and \( {\psi }_y =-a(z) \) (for some functions \( a,b: {\mathbb {R}} \rightarrow {\mathbb {R}} \)) \( \Rightarrow \psi (x,y,z) = b(z)x -a(z)y + c(z) \). This gives: \( F_1 = -{\psi }_y = a(z), \; F_2 = {\psi }_x =b(z) \)
Now we have
Differentiating the last equation with respect to y, z respectively (and utilizing \( H_x = \frac{a}{b} H_t\)), we obtain
thus, \( b=b_0 =constant \). Arguing similarly for \( G= G(s,y) \) we obtain \( a=a_0 = constant. \) Therefore,
where h is a solution of the ODE
In the case where \( \Omega \subset {\mathbb {R}}^3 \), we utilize Theorem 1.1 in [13] to obtain that \( \psi \) is linear in \( \Omega \) and similarly we conclude. \(\square \)
Now we will prove an analogue of Theorem 5.1 in [8] for subsolutions of the Allen Cahn equation and also, without excluding apriori some potential singularities of the solutions. The observation in Proposition 3.2 below, is to utilize the main result from [6].
Proposition 3.2
Let \( u: {\mathbb {R}}^n \rightarrow {\mathbb {R}} \) be a non constant, smooth subsolution of \( \Delta u = W'(u), \, W: {\mathbb {R}} \rightarrow [0, + \infty ) \), except perhaps on a closed set S of potential singularities with \( {\mathcal {H}}^1(S) =0 \) and \( {\mathbb {R}}^n \setminus S \) is connected, such that
where \( {\mathcal {H}}^1 \) is the Hausdorff 1-measure in \( {\mathbb {R}}^n. \)
Then
and g is such that \( g'' = W'(g). \)
Proof
First we see that W is strictly positive in \( u( {\mathbb {R}}^n {\setminus } S) \). Indeed, if there exists \( x_0 \in {\mathbb {R}}^n {\setminus } S \) such that \( W(u(x_0)) = 0 \), then u is a constant by Corollary 3.1 in [2] and since \( {\mathbb {R}}^n {\setminus } S \) is connected.
So let \( v = G(u) \), where \( G'(u) = \dfrac{1}{\sqrt{2W(u)}} \), then
so v is a smooth solution of the Eikonal equation except perhaps of a closed set S of potential singularities with \( {\mathcal {H}}^1 (S) =0 \). Thus from the result of [6], we have that \( v= a \cdot x +b,\; a \in {\mathbb {R}}^n,\, |a|=1, \; b \in {\mathbb {R}} \) or \( v = | x-x_0 | +c \) for some \( x_0 \in {\mathbb {R}}^n,\; c \in {\mathbb {R}}\).
Therefore,
where \( G: {\mathbb {R}} \rightarrow {\mathbb {R}} \), such that \( G' = \dfrac{1}{\sqrt{2W}} \).
If \( u = G^{-1}(d +c) \) where \( d(x)= |x-x_0|\), then
and also,
and thus, \( (G^{-1})'' = W'(G^{-1}) \), so we obtain
which contradicts the fact that W is strictly positive in \( u({\mathbb {R}}^n {\setminus } S) \).
Therefore \( u(x) = g (a \cdot x +b) \) where \( g = G^{-1} \). \(\square \)
Remark 3.3
- (1) :
-
In Proposition 3.2 above, radially symmetric solutions are excluded as we see in the proof, but as it is well known (see [18]) if f is smooth and \( u \in C^2 ( {\overline{\Omega }}) \) is a positive solution of \( - \Delta u = f(u) \) for \( x \in B_1 \subset {\mathbb {R}}^n \) that vanishes on \( \partial B_1 \), it holds that then u is radially symmetric. So radially symmetric solutions of the Allen–Cahn equations are incompatible with the equipartition even if we do not exclude apriori singularities.
- (2) :
-
Note that, in Theorem 2.3, if u, v are smooth entire solutions, by (2.20) in the proof and the monotonicity \( u_{x_n}>0 \), arguing as in the proof of Proposition 3.2 above we can conclude that \( u, \, v \) are one dimensional and the radially symmetric solutions are also excluded in this case.
4 The Allen Cahn system
4.1 Applications of the Equipartition
We begin by proposing a De Giorgi like result for the Allen Cahn systems for solutions that satisfy the equipartition of the energy or as an analogy of [8] in the vector case. First, the property that the level sets of a solution are hyperplanes can be expressed equivalently as \( \dfrac{u_{x_i}}{u_{x_n}} = c_i,\;\, i =1,\ldots ,n-1 \;\, (u: {\mathbb {R}}^n \rightarrow {\mathbb {R}}, \; u_{x_n}>0 \)), that is, if we consider \( v_i = \dfrac{u_{x_i}}{u_{x_n}},\; i=1,\ldots ,n,\;\, v_i: {\mathbb {R}}^n \rightarrow {\mathbb {R}} \), then
We can see the above statement as follows,
If \( v_i = c_i,\; i=1,\ldots ,n-1 \) then \( \nabla v_i =0 \Rightarrow rank( \nabla v_i) =0 <1\) Conversely, if \( rank( \nabla v_i) <1 \), that is \( rank( \nabla v_i) =0 \) since \( v_i: {\mathbb {R}}^n \rightarrow {\mathbb {R}} \), we have by Sard’s theorem that \( {\mathcal {L}}^1(v_i({\mathbb {R}}^n)) = 0,\; i=1,\ldots ,n-1 \) (see for example [24]) where \( {\mathcal {L}}^1 \) is the lebesgue measure on \( {\mathbb {R}}\). Thus, \({\mathcal {L}}^1(v_i({\mathbb {R}}^n)) =0 \Rightarrow v_i =c_i \) (constant) \(i=1,\ldots ,n-1\).
Now, we can generalize the above to the vector case as follows:
Let \( u: {\mathbb {R}}^n \rightarrow {\mathbb {R}}^m, \; u = (u_1,\ldots ,u_m),\; u_i = u_i (x_1,\ldots ,x_n) \), we consider the functions
and \( {\tilde{v}}^k = (v_{1k},\ldots ,v_{mk}),\; {\tilde{v}}^k: {\mathbb {R}}^n \rightarrow {\mathbb {R}}^m,\; k =1,\ldots ,n-1 \) and \( \nabla {\tilde{v}}^k: {\mathbb {R}}^n \rightarrow {\mathbb {R}}^{m \times n} \). Thus, if u is a solution of the Allen Cahn system, we could ask (under appropriate assumptions) whether \( rank ( \nabla {\tilde{v}}^k) < \min \lbrace n,m \rbrace = \mu \) (and by Sard’s Theorem we would have that \( {\mathcal {L}}^{\mu } ( {\tilde{v}}^k ({\mathbb {R}}^n)) =0 \), where \( {\mathcal {L}}^{\mu } \) is the Lebesgue measure in \( {\mathbb {R}}^{\mu })\).
Apart from u being a solution of the Allen Cahn system (and \( u_{i \, x_n}>0) \)) we should need further assumptions, as in the scalar case. The geometric analog in the vector case is far more complicated than in the scalar case. In particular, there is a relationship with minimizing partitions. However, one possible assumption would be that u also satisfies the equipartition, i.e. \( \dfrac{1}{2} | \nabla u |^2 = W(u) \). We will now prove that the above is true, at least for \( n=m=2 \), that is, if \( {\tilde{v}} = (v_1,v_2),\; v_i = \dfrac{u_{i \, x}}{u_{i \,y}} \) and \( u=(u_1,u_2) \) is a solution of the Allen–Cahn system that satisfy the equipartition, then \( rank(\nabla v) <2 \). In fact, we can obtain a quite stronger result about the structure of solutions in two dimensions, as stated in Theorem 4.1 that follows.
Theorem 4.1
Let \( u: {\mathbb {R}}^2 \rightarrow {\mathbb {R}}^2 \) be a smooth solution of
with \( u_{iy}>0,\; i =1,2 \) and \( W: {\mathbb {R}}^2 \rightarrow [0, +\infty ) \) smooth.
If u satisfies
Then
for some \( \;\, h: {\mathbb {R}}^2 \rightarrow {\mathbb {R}}. \)
In particular, \( {\mathcal {L}}^2(v({\mathbb {R}}^2 ))=0 \), where \( v= \left( \dfrac{u_{1x}}{u_{1y}}, \dfrac{u_{2x}}{u_{2y}}\right) . \)
Proof
We differentiate (4.2) with respect to \( x,\, y \)
and utilizing (4.1) we get
Now we define \( v_i:= \dfrac{u_{ix}}{u_{iy}} ,\; i=1,2 \) and by the second equation in (4.7) we have
similarly by the first equation in (4.7) we have
From (4.8), (4.9) and the assumption \( u_{iy}>0 ,\; i=1,2 \) we obtain that
Since \( det( \nabla v) = 0 \), we have that \( rank ( \nabla v)<2 \) and by Sard’s Theorem (see for example [24], p. 20) we have that \( {\mathcal {L}}^2(v({\mathbb {R}}^2 ))=0 \). By Theorem 1.4.14 in [24], since \( rank ( \nabla v)<2 \), we have that \( v_1,v_2 \) are functionally dependent, that is, there exists a smooth function \( h: {\mathbb {R}}^2 \rightarrow {\mathbb {R}} \) such that
Thus we have
so, together with (4.8), (4.9) we get
which gives
in the first case we also have
and therefore
where
In the second case we see that both equations of (4.4) are satisfied. \(\square \)
Note: If \( W(u_1,u_2) = W_1(u_1) +W_2(u_2) \), then (4.1) becomes
so, by analogy with the scalar case we should suppose \( u_{iy} >0 \) as we see in Theorem 4.1 above.
4.2 The Leray projection on the Allen–Cahn system
We begin with a calculation with which we will obtain an equation independent of the potential W.
Let \( u: {\mathbb {R}}^2 \rightarrow {\mathbb {R}}^2 \) be a smooth solution of the system
where \( W: {\mathbb {R}}^2 \rightarrow {\mathbb {R}} \), a \( C^2 \) potential.
From (4.17), differentiating over \( x ,\,y \) we obtain
and therefore
thus we have
Now we will apply the Helmholtz-Leray decomposition, that resolves a vector field u in \( {\mathbb {R}}^n \;\, (n = 2, 3) \) into the sum of a gradient and a curl vector. Regardless of any boundary conditions, for a given vector field u can be decomposed in the form
where \( div \, {\tilde{\sigma }} =0 \Leftrightarrow {\tilde{\sigma }}_{1 x} + {\tilde{\sigma }}_{2 y} =0 \) since we are in two dimensions, and thus \( {\tilde{\sigma }}_1 = - \sigma _y,\; {\tilde{\sigma }}_2 = \sigma _x \). So, we have that
for some \( \phi , \sigma : {\mathbb {R}}^2 \rightarrow {\mathbb {R}} \).
Utilizing now this decomposition of u, we obtain
Thus, if in particular we apply the Leray projection, \( v = {\mathbb {P}}(u) \), we have that \( v = {\tilde{\sigma }} \), that is, \( v = ( - \sigma _y, \sigma _x) \). So, from (4.21) we have
Note that a class of solutions to (4.22) is \( \sigma \) that satisfy
and we can solve explicitly in \( {\mathbb {R}}^2 \),
for arbitrary functions \( A,B,F,G: {\mathbb {R}} \rightarrow {\mathbb {R}} \).
In the first case, the Leray projection of the solution is of the form
and in the second case
Similarly, if we take the projection to the space of gradients, we have \( {\tilde{v}} = ( \phi _x, \phi _y) \) that will also satisfy
so again, the projection to the space of gradients of the solution will be of the form
Therefore, if we determine a class of potentials W, such that the solutions (or some solutions) are invariant under the Leray projection (or the projection to the space of gradients), we can obtain explicit solutions of the form (4.24) or (4.28). In the Appendix we give such examples.
Availability of data and materials
This declaration is not applicable.
References
Alama, S., Bronsard, L., Gui, C.: Stationary layered solutions in \( {\mathbb{R} }^2 \) for an Allen-Cahn system with multiple well potential. Calc. Var. 5, 359–390 (1997). (Springer-Verlag 1997)
Alikakos, N.D., Fusco, G., Smyrnelis, P.: Elliptic Systems of Phase Transition Type, Progress in Nonlinear Differential Equations and Their Applications, Birkhäuser (2018)
Ambrosio, L., Cabre, X.: Entire solutions of semilinear elliptic equations in \({{\mathbb{R} }}^3\) and a conjecture of De Giorgi. J. Am. Math. Soc. 13, 725 (2000)
Bombieri, E., De Giorgi, E., Giusti, E.: Minimal cones and the Bernstein problem. Invent. Math. 7, 243–268 (1969)
Cabre, X.: Uniqueness and stability of saddle-shaped solutions to the Allen-Cahn equation. Journal de Mathématiques Pures et Appliquées 98(3), 239–256 (2012)
Caffarelli, L., Crandall, M.: Distance functions and almost global solutions of Eikonal equations. Com. Part. Diff. Equ. 35(3), 391–414 (2010)
Caffarelli, L., Lin, F.: Singularly Perturbated elliptic systems and multi-valued harmonic functions with free boundaries. J. Am. Math. Soc. 21, 3 (2008)
Caffarelli, L., Garofalo, N., Segala, F.: A gradient bound for entire solutions of Quasi-Linear Equations and its consequences. Commun. Pure Appl. Math. XLVII, 1457–1473 (1994)
Colding, T.H., Minicozzi, W.P.: A couse in minimal surfaces. Grad. Sutd. Math. 121 (2011)
Danielli, D., Garofalo, N.: Properties of entire solutions of non-uniformly elliptic equations arising in geometry and in phase transitions. Calc. Var. (2002)
De Giorgi, E.: Convergence problems for functionals and operators. In: Proccedings of the International Meeting on Recent Methods in Non linear. Analysis, pp. 131–188 (1978)
Del Pino, M., Kowalczyk, M., Wei, J.: On the De Giorgi conjecture in dimension \( N \ge 9 \). Ann. Math. 174(3), 1485–1569 (2011)
Edelen, N., Wang, Z.: A Bernstein-type theorem for minimal graphs over convex domains, Ann. Inst. Poincare (2022)
Farina, A.: Finite-Energy solutions, Quantization Effects and Liouville-type results for a variant of the Ginzburg-Landau Systems in \( {\mathbb{R} }^N\). Diff. Integ. Equ. 11(6), 875–893 (1998)
Fusco, G.: Minimizing under relaxed symmetry constraints, Triple and N-junctions, Annali Di Scienze, Scuola Normale Superiore (2022)
Ghoussoub, N., Gui, C.: On a conjecture of De Giorgi and some related problems. Math. Ann. 311, 481–491 (1998)
Ghoussoub, N., Gui, C.: On the De Giorgi’s conjecture in dimensions 4 and 5. Ann. Math. (2) 157(1), 313–334 (2003)
Gidas, B., Ni, W.-M., Nirengerg, L.: Symmetry and related properties via the maximum principle. Commun. Math. Phys. 68, 209–243 (1979)
Majda, Andrew J., Bertozzi, Andrea L.: Vorticity and incompressible flow, Cambridge texts in Applied Mathematics (2002)
Modica, L., Mortola, S.: Some entire solutions in the plane of Nonlinear Poisson Equations, Instituto di Mathematica Leonida Tonelli, Via Derna I, 56100 Pisa (1980)
Modica, L.: \( \Gamma \)-convergence to minimal surfaces problem and global solutions of \( \Delta u = 2(u^3 - u) \), Proceedings of the International Meeting on Recent Methods in Nonlinear Analysis (Rome, 1978), pp. 223–244, Pitagora, Bologna (1979)
Modica, L.: A gradient bound and a liouville theorem for nonlinear Poisson equations. Commun. Pure Appl. Math. 38, 679–684 (1985)
Modica, L., Mortola, S.: Un esempio di \( \Gamma \)-convergenza (Italian). Boll. Un. Mat. Ital. B(5) 14(1), 285–299 (1977)
Narasimhan, R.: Analysis on Real and Complex Manifolds. North Holland (1985)
Nicholas, D.: Alikakos, Dimitrios Gazoulis, Almost entire solutions of the burgers equation. Electr. J. Diff. Equa. 2018(53), 1–6 (2018)
Savin, O.: Phase transitions, minimal surfaces and a conjecture of De Giorgi. Curr. Devel. Math. 2010, 59–113 (2009)
Savin, O.: Regularity of flat level sets in phase transitions. Ann. Math. 169, 41–78 (2009)
Smyrnelis, P.: Gradient estimates for semilinear elliptic systems and other related results. Proc. Royal Soc. Edinb. Sect. A 145(6), 1313–1330 (2015)
Acknowledgements
I would like to thank my advisor Professor Nicholas Alikakos for his guidance and inspiration, and also for motivating the study of implications of the equipartition in the Allen–Cahn system. Also, I would like to thank S. Papathanasiou and Professors A. Farina and P. Smyrnelis for their valuable comments on a previous version of this paper, which led to various improvements.
Funding
Open access funding provided by HEAL-Link Greece. No funding was received for conducting this study.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author have no conflict of interest to declare that are relevant to the content of this article.
Ethical approval
This declaration is not applicable.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A: Some examples of entire solutions of the Allen–Cahn system
We note that solutions of the form (4.24) and (4.28) are equivalent in the special case that (4.23) is satisfied. So in the class of solutions of (4.23) the Leray projection is, in some sense equivalent with the projection to the space of gradients. Suppose now that \( u = \nabla \phi \) for some \( \phi : {\mathbb {R}}^2 \rightarrow {\mathbb {R}} \), that is, a solution of the Allen–Cahn system remains invariant under the projection to the space of gradients. Then, as (4.27) we have
So a simple solution to (A.1) is
and \( u(x,y) = ( \phi _x, \phi _y) \), so in this case u has the form
for some \( f,g: {\mathbb {R}} \rightarrow {\mathbb {R}}\).
If u has the form (A.3), we can see that it also satisfies the equipartition. Indeed, (4.17) becomes
and the equipartition can be written as
(the system (A.1) remains equivalent if we add a constant to the potential)
First we note that solutions of the form (A.3) satisfy (4.4) in Theorem 4.1. Indeed, if u is of the form (A.3),
so the function \( h:{\mathbb {R}}^2 \rightarrow {\mathbb {R}} \) in (4.4) is \( h(s,t) = st -1 \). Also,
Now we will see some examples of solutions to the Allen Cahn system that are not in the form (4.3) in Theorem 4.1 (which are more similar to the ones in the scalar case). Some of the examples of such solutions are in the form (A.3) and for all solutions in this form the function h in (4.4) is, as mentioned above, \( h(s,t) = st -1\).
Example
[1] If \( W(u_1,u_2) = u_1 u_2 \), then
where \( \text {cosh} (t) = \dfrac{e^{t} + e^{-t}}{2} \), is a solution of \( \Delta u = W_u(u) \) that satisfies the equipartition and is of the form (4.4). A more general solution is
However, not all solutions in this form satisfy the equipartition. In this example the zero set of the potential is \( \lbrace W= 0 \rbrace = \lbrace u_1=0 \rbrace \cup \lbrace u_2=0 \rbrace \). Such potentials W belong in a class of potentials that have been thoroughly studied in [7].
Example
[2] If \( W(u_1,u_2)= \dfrac{[(u_1+u_2)^2-4]^2+[(u_1-u_2)^2-4]^2}{16}\), then
is a solution of \( \Delta u = W_u(u) \) that satisfies the equipartition (and is of the form (4.4) and \( h(s,t) = st-1 \)). In addition, u above connects all four phases of the potential W at infinity, that is
\(\lbrace W=0 \rbrace = \lbrace (2,0), (-2,0), (0, 2 ), (0, -2) \rbrace \).
This solution is a saddle solution (see [15]) and is invariant under rotations of \( \frac{\pi }{2} \) angle (i.e. \( u( \omega (x,y)) = \omega u(x,y) \), where \( \omega \) is the \( \frac{\pi }{2} \)-rotation matrix.
Also, another solution of \( \Delta u = W_u(u) \) for such potential is
for this solution the function h in (4.4) is \( h(s,t)=s+t -2 \) but u in (A.10) does not satisfy the equipartition. Thus, the class of solutions of the Allen–Cahn system that are of the form (4.4) in Theorem 4.1, is more general than that of solutions to the Allen–Cahn system that satisfy the equipartition. Note that u in (A.10) has the property that
and \( W(-u_1,u_2) = W(u_1,u_2) \). The general existence of solutions with property similar to (A.11) for potentials with such symmetry hypothesis can be found in [1].
More generally, if \( a^2 +b^2 =1= c^2 +d^2 \), then
solves (4.17) and we obtain infinitely many solutions which connect the four minima of W in sectors of variable angle.
Example
[3] If \( W(u_1,u_2) = u_1^2 + u_2^2 -1 \), then
is a solution of \( \Delta u = W_u(u) \), where \( a_i^2+b_i^2 =2,\; c_i \in {\mathbb {R}} \).
In this case, \( \lbrace W = 0 \rbrace = \lbrace u_1^2+u_2^2=1 \rbrace \).
Also, if \( W(u_1,u_2) = W(u_1^2+u_2^2) \) and \( W' <0 \), we have that
with \( a^2+b^2= -2W'(1) \), is a solution to \( \Delta u =W_u(u) \).
Appendix B: Entire solutions of the Euler equations
In this Appendix we will determine some smooth entire solutions of the 2D and 3D Euler equations and the pressure being a linear function with respect to the space variables.
We begin by illustrating an analogy for steady solutions of the incompressible Euler equations in two space dimensions and the De Giorgi conjecture.
Let \( u=(u_1,u_2): {\mathbb {R}}^2 \times (0, + \infty ) \rightarrow {\mathbb {R}}^2,\; u_i =u_i(x,t),\; x=(x_1,x_2) \) be a smooth solution of the Euler equations. The incompressibility condition \( div \, u =0 \) gives that there exists a (unique up to an additive constant) stream function \( \psi (x,t) \) such that
In addition, by Proposition 2.2 in [19], a stream function \( \psi \) on a domain \( \Omega \subset {\mathbb {R}}^2 \) defines a steady solution (i.e. time independent) of the 2D Euler equation on \( \Omega \) if and only if
So, if \( \psi \) is a bounded, entire solution such that \( \psi _{x_2} \ge 0 \), then by De Giorgi’s conjecture (see Theorem 1.1 in [16]) it holds that
Therefore we raise the following question.
Question: Let \( u: {\mathbb {R}}^2 \times (0, + \infty ) \rightarrow {\mathbb {R}}^2,\; (u = u(x,y,t)=(u_1,u_2) )\) be a smooth, bounded entire solution of the Isobaric 2D Euler equations
Is it true that then
where \( c_1 \beta + c_2 \gamma =0,\; c_1,c_2,{\tilde{c}}_1,{\tilde{c}}_2, \beta , \gamma \in {\mathbb {R}}\).
From the form of solution (B.2) we can obtain a solution of the 2D Euler equation with pressure being a linear function in respect to the space variables.
Let \( u: {\mathbb {R}}^3 \rightarrow {\mathbb {R}}^2,\; (u = u(x,y,t)=(u_1,u_2) )\) is such that
where \( A'(t)=a(t),\; a,b: {\mathbb {R}} \rightarrow {\mathbb {R}} \) and \( c_1, {\tilde{c}}_1, c_2, {\tilde{c}}_2,\beta , \gamma , \lambda , \xi \in {\mathbb {R}} \) are such that \( c_1 \beta + c_2 \gamma =0 \) and \( \lambda \beta + \xi \gamma =0\).
Then \( u = (u_1,u_2) \) satisfies
Now we give some examples of smooth entire solutions for the three dimensional Euler equations. If \( u = (u_1,u_2,u_3): {\mathbb {R}}^4 \rightarrow {\mathbb {R}}^3 \) where \( u_i = u_i(x,y,z,t) \) is such that
then \( u = (u_1,u_2,u_3) \) is an entire solution of the Euler equations, that is u satisfies
Note that from symmetry properties of the Euler equations and from (B.5) we can also have the following solution of (B.6):
and also,
Finally, another example of smooth entire solution of (B.6) is the following
(we can choose A such that \( A(0)=0 )\)
Therefore we conclude to the following result
Theorem B.1
Let \( u=(u_1,u_2,u_3),\; u_i, \quad p: {\mathbb {R}}^3 \times (0,+ \infty ) \rightarrow {\mathbb {R}} \) and consider the initial value problem
where \( g = (g_1,g_2,g_3) \) is either of the form
or
Then there exists a smooth, globally defined in \( t>0 \), solution of (B.10).
In particular, either u and p are given by (B.9) if the initial value g is of the form (B.11) or u and p are given by (B.5) if g is of the form (B.12).
The condition (B.12) could be easily modified in order to obtain the solutions given by (B.7) and (B.8).
Remark B.2
Such solutions can be extended to general dimensions, i.e. solutions of (2.8) and \( n \ge 4 \), together with the divergence free condition and a pressure being a linear function with respect to space variables.
Appendix C: Some examples of entire solutions of the Navier–Stokes equations
First we note that some solutions of the 3D Euler equations in Appendix B have the form \( u=(u_1, c_1u_1 + {\tilde{c}}_1,c_2u_1 + {\tilde{c}}_2) \), that is, we have linear dependence of the components of the solution. So, now we will determine some specific examples of solutions of the Navier–Stokes equations with linear dependent components.
Let \( u=(u_1,u_2),\; u_i=u_i(x,y,t): {\mathbb {R}}^2 \times (0,+ \infty ) \rightarrow {\mathbb {R}} \) defined as
then u is a solution of
Similarly in the three dimensional case, we give some examples of solutions of
Let \( g=g(s, \eta ,t),\; g: {\mathbb {R}}^2 \times (0, + \infty ) \rightarrow {\mathbb {R}} \) be a solution of
where \( \mu >0,\; c_1,c_2, {\tilde{c}}_1, {\tilde{c}}_2 \in {\mathbb {R}} \) and \( t>0 \).
Then \( u=(u_1,u_2,u_3),\; u_i: {\mathbb {R}}^3 \times (0,+ \infty ) \rightarrow {\mathbb {R}},\; i =1,2,3 \) defined as
is a solution of (C.3).
Therefore we conclude to the following
Proposition C.1
Let \( u=(u_1,u_2,u_3),\; u_i,\, p: {\mathbb {R}}^3 \times (0, + \infty ) \rightarrow {\mathbb {R}}^3 \) and consider the initial value problem
where \( h=(h_1,c_1h_1 + {\tilde{c}}_1, c_2h_1 +{\tilde{c}}_2) \) and \( h_1(x,y,z)= H(2c_1c_2 x - c_2y - c_1z),\; c_1,c_2, {\tilde{c}}_1, {\tilde{c}}_2 \in {\mathbb {R}} \) such that \( {\tilde{c}}_1 c_2 + c_1 {\tilde{c}}_2 =0 \) and H smooth.
Then there exists a smooth, globally defined in \( t>0 \), solution to (C.6).
In particular,
Remark C.2
We can also have the same result for a bit more general initial values h in Proposition C.1, as we can see from (C.4), (C.5). It suffices to have linear dependency of the components of h and \( h_1 \) above can also be for example of the form \( h_1(x,y,z) = H(2c_1c_2x-c_2y-c_1z,c_2y-c_1z) \).
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Gazoulis, D. A Relation of the Allen–Cahn equations and the Euler equations and applications of the equipartition. Nonlinear Differ. Equ. Appl. 30, 81 (2023). https://doi.org/10.1007/s00030-023-00888-2
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00030-023-00888-2