Nonconstant solutions of homogeneous Poisson equation on domain without boundaryQuestion on harmonic functionsProve solutions to the reduced Helmholtz equation are uniqueCharacterize solutions to Laplace's and Poisson's equation in the unit square with periodic boundary conditionsBiharmonic equation boundary conditionslimit of harmonic functions Gilbarg and Trudinger page 27How to numerically solve the Poisson equation given Neumann boundary conditions?Laplace Equation with Tangential Derivative Prescribed on the BoundaryExistence of solution of Poisson equation for a square with homogeneous Dirichlet boundary conditionDomain solutions on partial differential equationsHarmonic functions, equivalence of boundary conditions with phenomena outside domain.
How to explain that I do not want to visit a country due to personal safety concern?
Can anyone tell me why this program fails?
Why did it take so long to abandon sail after steamships were demonstrated?
Why would a flight no longer considered airworthy be redirected like this?
Using "wallow" verb with object
Theorems like the Lovász Local Lemma?
Is it possible that AIC = BIC?
How to make healing in an exploration game interesting
Do I need life insurance if I can cover my own funeral costs?
At what level can a dragon innately cast its spells?
Is having access to past exams cheating and, if yes, could it be proven just by a good grade?
Official degrees of earth’s rotation per day
PTIJ: Who should pay for Uber rides: the child or the parent?
When do we add an hyphen (-) to a complex adjective word?
Why doesn't the EU now just force the UK to choose between referendum and no-deal?
Informing my boss about remarks from a nasty colleague
Why do passenger jet manufacturers design their planes with stall prevention systems?
How could a female member of a species produce eggs unto death?
Russian cases: A few examples, I'm really confused
How do anti-virus programs start at Windows boot?
Why are the outputs of printf and std::cout different
Ban on all campaign finance?
Professor being mistaken for a grad student
Provisioning profile doesn't include the application-identifier and keychain-access-groups entitlements
Nonconstant solutions of homogeneous Poisson equation on domain without boundary
Question on harmonic functionsProve solutions to the reduced Helmholtz equation are uniqueCharacterize solutions to Laplace's and Poisson's equation in the unit square with periodic boundary conditionsBiharmonic equation boundary conditionslimit of harmonic functions Gilbarg and Trudinger page 27How to numerically solve the Poisson equation given Neumann boundary conditions?Laplace Equation with Tangential Derivative Prescribed on the BoundaryExistence of solution of Poisson equation for a square with homogeneous Dirichlet boundary conditionDomain solutions on partial differential equationsHarmonic functions, equivalence of boundary conditions with phenomena outside domain.
$begingroup$
This answer on physics stack exchange says
For a compact domain without boundary (such as the surface of a sphere), you don't need any boundary conditions: there are no non-constant harmonic functions on such domains. (Slick proof of this: you can prove that harmonic functions never have local maxima or minima, but a nonconstant function on such a domain must have them -- in particular, it must have a global maximum and a global minimum somewhere.)
A harmonic function obeys
$$nabla^2 f = fracpartial^2partial x^2f + fracpartial^2partial y^2f = 0$$
And indeed, if $f$ had a local extremum the sign of $fracpartial^2partial x^2f$ and of $fracpartial^2partial y^2f$ would have to be the same, so the sum cannot be $0$, so $f$ cannot have a local extremum, so on a compact domain without boundary it must be constant.
I have two questions about this line of reasoning:
What about local extrema of the form $x^4 y^4$? The second derivative in both directions could be $0$, so I cannot exclude the possibility that a harmonic function could have an extremum like this.
Suppose we are not considering harmonic functions, but rather solutions of
$$nabla cdot left[beginmatrix m_11 & m_12 \ m_21 & m_22endmatrixright]nabla f = 0 labelatag1$$
where the $m_ij$ can depend on $x,y$. Under what conditions is it still true that there are no nonconstant solutions of $(refa)$ on compact domains without boundary? (presumably the $m_ij$ matrix needs to be invertible. In the case of interest to me, it is Hermitian. I could try to change coordinates to a coordinate system whose axes are the eigenvectors of the the $m_ij$ matrix, but since those eigenvectors are not guaranteed to be real that would involve introducing complex coordinates...).
pde harmonic-functions
$endgroup$
add a comment |
$begingroup$
This answer on physics stack exchange says
For a compact domain without boundary (such as the surface of a sphere), you don't need any boundary conditions: there are no non-constant harmonic functions on such domains. (Slick proof of this: you can prove that harmonic functions never have local maxima or minima, but a nonconstant function on such a domain must have them -- in particular, it must have a global maximum and a global minimum somewhere.)
A harmonic function obeys
$$nabla^2 f = fracpartial^2partial x^2f + fracpartial^2partial y^2f = 0$$
And indeed, if $f$ had a local extremum the sign of $fracpartial^2partial x^2f$ and of $fracpartial^2partial y^2f$ would have to be the same, so the sum cannot be $0$, so $f$ cannot have a local extremum, so on a compact domain without boundary it must be constant.
I have two questions about this line of reasoning:
What about local extrema of the form $x^4 y^4$? The second derivative in both directions could be $0$, so I cannot exclude the possibility that a harmonic function could have an extremum like this.
Suppose we are not considering harmonic functions, but rather solutions of
$$nabla cdot left[beginmatrix m_11 & m_12 \ m_21 & m_22endmatrixright]nabla f = 0 labelatag1$$
where the $m_ij$ can depend on $x,y$. Under what conditions is it still true that there are no nonconstant solutions of $(refa)$ on compact domains without boundary? (presumably the $m_ij$ matrix needs to be invertible. In the case of interest to me, it is Hermitian. I could try to change coordinates to a coordinate system whose axes are the eigenvectors of the the $m_ij$ matrix, but since those eigenvectors are not guaranteed to be real that would involve introducing complex coordinates...).
pde harmonic-functions
$endgroup$
$begingroup$
$x^4y^4$ is not a harmonic function
$endgroup$
– Dylan
Mar 6 at 8:32
$begingroup$
I know. But it is an example of a function that has a local minimum yet $partial_x^2 f + partial_y^2 f=0$ at that local minimum. If the answer on physics stack exchange is correct, then harmonic functions cannot have this type of local minimum. Why not?
$endgroup$
– Wouter
Mar 6 at 8:53
add a comment |
$begingroup$
This answer on physics stack exchange says
For a compact domain without boundary (such as the surface of a sphere), you don't need any boundary conditions: there are no non-constant harmonic functions on such domains. (Slick proof of this: you can prove that harmonic functions never have local maxima or minima, but a nonconstant function on such a domain must have them -- in particular, it must have a global maximum and a global minimum somewhere.)
A harmonic function obeys
$$nabla^2 f = fracpartial^2partial x^2f + fracpartial^2partial y^2f = 0$$
And indeed, if $f$ had a local extremum the sign of $fracpartial^2partial x^2f$ and of $fracpartial^2partial y^2f$ would have to be the same, so the sum cannot be $0$, so $f$ cannot have a local extremum, so on a compact domain without boundary it must be constant.
I have two questions about this line of reasoning:
What about local extrema of the form $x^4 y^4$? The second derivative in both directions could be $0$, so I cannot exclude the possibility that a harmonic function could have an extremum like this.
Suppose we are not considering harmonic functions, but rather solutions of
$$nabla cdot left[beginmatrix m_11 & m_12 \ m_21 & m_22endmatrixright]nabla f = 0 labelatag1$$
where the $m_ij$ can depend on $x,y$. Under what conditions is it still true that there are no nonconstant solutions of $(refa)$ on compact domains without boundary? (presumably the $m_ij$ matrix needs to be invertible. In the case of interest to me, it is Hermitian. I could try to change coordinates to a coordinate system whose axes are the eigenvectors of the the $m_ij$ matrix, but since those eigenvectors are not guaranteed to be real that would involve introducing complex coordinates...).
pde harmonic-functions
$endgroup$
This answer on physics stack exchange says
For a compact domain without boundary (such as the surface of a sphere), you don't need any boundary conditions: there are no non-constant harmonic functions on such domains. (Slick proof of this: you can prove that harmonic functions never have local maxima or minima, but a nonconstant function on such a domain must have them -- in particular, it must have a global maximum and a global minimum somewhere.)
A harmonic function obeys
$$nabla^2 f = fracpartial^2partial x^2f + fracpartial^2partial y^2f = 0$$
And indeed, if $f$ had a local extremum the sign of $fracpartial^2partial x^2f$ and of $fracpartial^2partial y^2f$ would have to be the same, so the sum cannot be $0$, so $f$ cannot have a local extremum, so on a compact domain without boundary it must be constant.
I have two questions about this line of reasoning:
What about local extrema of the form $x^4 y^4$? The second derivative in both directions could be $0$, so I cannot exclude the possibility that a harmonic function could have an extremum like this.
Suppose we are not considering harmonic functions, but rather solutions of
$$nabla cdot left[beginmatrix m_11 & m_12 \ m_21 & m_22endmatrixright]nabla f = 0 labelatag1$$
where the $m_ij$ can depend on $x,y$. Under what conditions is it still true that there are no nonconstant solutions of $(refa)$ on compact domains without boundary? (presumably the $m_ij$ matrix needs to be invertible. In the case of interest to me, it is Hermitian. I could try to change coordinates to a coordinate system whose axes are the eigenvectors of the the $m_ij$ matrix, but since those eigenvectors are not guaranteed to be real that would involve introducing complex coordinates...).
pde harmonic-functions
pde harmonic-functions
edited Mar 11 at 10:26
Wouter
asked Mar 6 at 8:12
WouterWouter
5,92421436
5,92421436
$begingroup$
$x^4y^4$ is not a harmonic function
$endgroup$
– Dylan
Mar 6 at 8:32
$begingroup$
I know. But it is an example of a function that has a local minimum yet $partial_x^2 f + partial_y^2 f=0$ at that local minimum. If the answer on physics stack exchange is correct, then harmonic functions cannot have this type of local minimum. Why not?
$endgroup$
– Wouter
Mar 6 at 8:53
add a comment |
$begingroup$
$x^4y^4$ is not a harmonic function
$endgroup$
– Dylan
Mar 6 at 8:32
$begingroup$
I know. But it is an example of a function that has a local minimum yet $partial_x^2 f + partial_y^2 f=0$ at that local minimum. If the answer on physics stack exchange is correct, then harmonic functions cannot have this type of local minimum. Why not?
$endgroup$
– Wouter
Mar 6 at 8:53
$begingroup$
$x^4y^4$ is not a harmonic function
$endgroup$
– Dylan
Mar 6 at 8:32
$begingroup$
$x^4y^4$ is not a harmonic function
$endgroup$
– Dylan
Mar 6 at 8:32
$begingroup$
I know. But it is an example of a function that has a local minimum yet $partial_x^2 f + partial_y^2 f=0$ at that local minimum. If the answer on physics stack exchange is correct, then harmonic functions cannot have this type of local minimum. Why not?
$endgroup$
– Wouter
Mar 6 at 8:53
$begingroup$
I know. But it is an example of a function that has a local minimum yet $partial_x^2 f + partial_y^2 f=0$ at that local minimum. If the answer on physics stack exchange is correct, then harmonic functions cannot have this type of local minimum. Why not?
$endgroup$
– Wouter
Mar 6 at 8:53
add a comment |
1 Answer
1
active
oldest
votes
$begingroup$
Why can't a nonconstant harmonic function have a local minimum? The averaging property. The value of a plane harmonic function at a point $x$ is equal to its average value on any disk centered at $x$.
So, then, if $f(x)le f(y)$ for all $y$ with $|y-x|_2 le r$, then
$$0=f(x)-f(x)=frac1pi r^2iint__2le rf(y)-f(x),dy ge 0$$
with equality only if the (continuous) nonnegative function we're integrating is identically zero in the disk $|y-x|_2le r$.
Now, I should note that it's nontrivial to define the Laplacian operator $Delta = nabla^2$ on the sphere or something like it; the partial derivative operators we built it from are based on a specific set of coordinates, and there's no single set of coordinates that defines the sphere. We can't fall back on the (coordinate-independent) exterior derivative, either; while the gradient map from $0$-forms to $1$-forms and the divergence map from $(n-1)$-forms to $n$-forms are both cases of the exterior derivative, the map we need to bridge the gap from $1$-forms to $n$-forms is strongly coordinate-dependent. Even in the $n=2$ case.
How do we fix this and get something we can work with? To get an idea of what's going on, note that the Hessian matrix of second derivatives transforms by a congruence relation: if $g(x)=f(Ax)$, then $H_g=A^TH_fA^T$. The Laplacian is the trace of this matrix. The Laplacian is the trace of this matrix, so we would like to restrict ourselves to transformations that preserve the trace. What does that? Orthogonal matrices; if $A^T=A^-1$, then that's a similarity relation, and the trace is fixed.
Orthogonal matrices preserve distances and inner products. If we want this generalized Laplacian to work out, we need a notion of inner products on our manifold. That's a Riemannian metric - a system of inner products on the tangent spaces. So, then, to find the Laplacian, choose local coordinates $x_1,x_2,dots,x_n$ so that the inner product at the point we're interested in is the Euclidean inner product. The Laplacian at that point is $fracpartial^2partial x_1^2+fracpartial^2partial x_2^2+cdots+fracpartial^2partial x_n^2$, and this is invariant under any isometry.
Now, looking back at our argument - the averaging property depends on working in a flat space. If the metric varies point-to-point in those local coordinates we found, so will the formula for the Laplacian - and the averaging property is no longer something we can explicitly write down. What do we get instead? The Laplacian becomes
$$Delta f(x,y) = a(x,y)fracpartial^2 fpartial x^2+2b(x,y)fracpartial fpartial xpartial y+c(x,y)fracpartial^2 fpartial y^2 = texttr(MH)$$
in our local coordinates, where the matrix $M=beginbmatrixa&b\b&cendbmatrix$ is always positive definite and $H$ is the Hessian matrix of second derivatives.
And there's your second part. We need it in order to deal with harmonic functions on any manifold that isn't flat. The extra condition we need falls right out, too - that matrix has to be positive definite everywhere. If $M$ is indefinite, we can move to a basis of eigenvectors and then set $H$ to a positive diagonal matrix while still having that operator zero, giving us a local minimum.
As for the difference between real symmetric and Hermitian? We're adding the off-diagonal elements of $M$ with their transposes anyway, so the antisymmetric imaginary part zeros out. There's no reason to go to complex coordinates unless we started with them.
This more general result follows from the Hopf maximum principle. On a closed disk in our local coordinate patch, the maximum and minimum must occur on the boundary of the disk, so the center can't be a local extremum unless the function is constant. Every point on the manifold has coordinates of this form, so there aren't any local extrema unless the function is constant.
$endgroup$
add a comment |
Your Answer
StackExchange.ifUsing("editor", function ()
return StackExchange.using("mathjaxEditing", function ()
StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix)
StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["$", "$"], ["\\(","\\)"]]);
);
);
, "mathjax-editing");
StackExchange.ready(function()
var channelOptions =
tags: "".split(" "),
id: "69"
;
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function()
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled)
StackExchange.using("snippets", function()
createEditor();
);
else
createEditor();
);
function createEditor()
StackExchange.prepareEditor(
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: true,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: 10,
bindNavPrevention: true,
postfix: "",
imageUploader:
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
,
noCode: true, onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
);
);
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f3137252%2fnonconstant-solutions-of-homogeneous-poisson-equation-on-domain-without-boundary%23new-answer', 'question_page');
);
Post as a guest
Required, but never shown
1 Answer
1
active
oldest
votes
1 Answer
1
active
oldest
votes
active
oldest
votes
active
oldest
votes
$begingroup$
Why can't a nonconstant harmonic function have a local minimum? The averaging property. The value of a plane harmonic function at a point $x$ is equal to its average value on any disk centered at $x$.
So, then, if $f(x)le f(y)$ for all $y$ with $|y-x|_2 le r$, then
$$0=f(x)-f(x)=frac1pi r^2iint__2le rf(y)-f(x),dy ge 0$$
with equality only if the (continuous) nonnegative function we're integrating is identically zero in the disk $|y-x|_2le r$.
Now, I should note that it's nontrivial to define the Laplacian operator $Delta = nabla^2$ on the sphere or something like it; the partial derivative operators we built it from are based on a specific set of coordinates, and there's no single set of coordinates that defines the sphere. We can't fall back on the (coordinate-independent) exterior derivative, either; while the gradient map from $0$-forms to $1$-forms and the divergence map from $(n-1)$-forms to $n$-forms are both cases of the exterior derivative, the map we need to bridge the gap from $1$-forms to $n$-forms is strongly coordinate-dependent. Even in the $n=2$ case.
How do we fix this and get something we can work with? To get an idea of what's going on, note that the Hessian matrix of second derivatives transforms by a congruence relation: if $g(x)=f(Ax)$, then $H_g=A^TH_fA^T$. The Laplacian is the trace of this matrix. The Laplacian is the trace of this matrix, so we would like to restrict ourselves to transformations that preserve the trace. What does that? Orthogonal matrices; if $A^T=A^-1$, then that's a similarity relation, and the trace is fixed.
Orthogonal matrices preserve distances and inner products. If we want this generalized Laplacian to work out, we need a notion of inner products on our manifold. That's a Riemannian metric - a system of inner products on the tangent spaces. So, then, to find the Laplacian, choose local coordinates $x_1,x_2,dots,x_n$ so that the inner product at the point we're interested in is the Euclidean inner product. The Laplacian at that point is $fracpartial^2partial x_1^2+fracpartial^2partial x_2^2+cdots+fracpartial^2partial x_n^2$, and this is invariant under any isometry.
Now, looking back at our argument - the averaging property depends on working in a flat space. If the metric varies point-to-point in those local coordinates we found, so will the formula for the Laplacian - and the averaging property is no longer something we can explicitly write down. What do we get instead? The Laplacian becomes
$$Delta f(x,y) = a(x,y)fracpartial^2 fpartial x^2+2b(x,y)fracpartial fpartial xpartial y+c(x,y)fracpartial^2 fpartial y^2 = texttr(MH)$$
in our local coordinates, where the matrix $M=beginbmatrixa&b\b&cendbmatrix$ is always positive definite and $H$ is the Hessian matrix of second derivatives.
And there's your second part. We need it in order to deal with harmonic functions on any manifold that isn't flat. The extra condition we need falls right out, too - that matrix has to be positive definite everywhere. If $M$ is indefinite, we can move to a basis of eigenvectors and then set $H$ to a positive diagonal matrix while still having that operator zero, giving us a local minimum.
As for the difference between real symmetric and Hermitian? We're adding the off-diagonal elements of $M$ with their transposes anyway, so the antisymmetric imaginary part zeros out. There's no reason to go to complex coordinates unless we started with them.
This more general result follows from the Hopf maximum principle. On a closed disk in our local coordinate patch, the maximum and minimum must occur on the boundary of the disk, so the center can't be a local extremum unless the function is constant. Every point on the manifold has coordinates of this form, so there aren't any local extrema unless the function is constant.
$endgroup$
add a comment |
$begingroup$
Why can't a nonconstant harmonic function have a local minimum? The averaging property. The value of a plane harmonic function at a point $x$ is equal to its average value on any disk centered at $x$.
So, then, if $f(x)le f(y)$ for all $y$ with $|y-x|_2 le r$, then
$$0=f(x)-f(x)=frac1pi r^2iint__2le rf(y)-f(x),dy ge 0$$
with equality only if the (continuous) nonnegative function we're integrating is identically zero in the disk $|y-x|_2le r$.
Now, I should note that it's nontrivial to define the Laplacian operator $Delta = nabla^2$ on the sphere or something like it; the partial derivative operators we built it from are based on a specific set of coordinates, and there's no single set of coordinates that defines the sphere. We can't fall back on the (coordinate-independent) exterior derivative, either; while the gradient map from $0$-forms to $1$-forms and the divergence map from $(n-1)$-forms to $n$-forms are both cases of the exterior derivative, the map we need to bridge the gap from $1$-forms to $n$-forms is strongly coordinate-dependent. Even in the $n=2$ case.
How do we fix this and get something we can work with? To get an idea of what's going on, note that the Hessian matrix of second derivatives transforms by a congruence relation: if $g(x)=f(Ax)$, then $H_g=A^TH_fA^T$. The Laplacian is the trace of this matrix. The Laplacian is the trace of this matrix, so we would like to restrict ourselves to transformations that preserve the trace. What does that? Orthogonal matrices; if $A^T=A^-1$, then that's a similarity relation, and the trace is fixed.
Orthogonal matrices preserve distances and inner products. If we want this generalized Laplacian to work out, we need a notion of inner products on our manifold. That's a Riemannian metric - a system of inner products on the tangent spaces. So, then, to find the Laplacian, choose local coordinates $x_1,x_2,dots,x_n$ so that the inner product at the point we're interested in is the Euclidean inner product. The Laplacian at that point is $fracpartial^2partial x_1^2+fracpartial^2partial x_2^2+cdots+fracpartial^2partial x_n^2$, and this is invariant under any isometry.
Now, looking back at our argument - the averaging property depends on working in a flat space. If the metric varies point-to-point in those local coordinates we found, so will the formula for the Laplacian - and the averaging property is no longer something we can explicitly write down. What do we get instead? The Laplacian becomes
$$Delta f(x,y) = a(x,y)fracpartial^2 fpartial x^2+2b(x,y)fracpartial fpartial xpartial y+c(x,y)fracpartial^2 fpartial y^2 = texttr(MH)$$
in our local coordinates, where the matrix $M=beginbmatrixa&b\b&cendbmatrix$ is always positive definite and $H$ is the Hessian matrix of second derivatives.
And there's your second part. We need it in order to deal with harmonic functions on any manifold that isn't flat. The extra condition we need falls right out, too - that matrix has to be positive definite everywhere. If $M$ is indefinite, we can move to a basis of eigenvectors and then set $H$ to a positive diagonal matrix while still having that operator zero, giving us a local minimum.
As for the difference between real symmetric and Hermitian? We're adding the off-diagonal elements of $M$ with their transposes anyway, so the antisymmetric imaginary part zeros out. There's no reason to go to complex coordinates unless we started with them.
This more general result follows from the Hopf maximum principle. On a closed disk in our local coordinate patch, the maximum and minimum must occur on the boundary of the disk, so the center can't be a local extremum unless the function is constant. Every point on the manifold has coordinates of this form, so there aren't any local extrema unless the function is constant.
$endgroup$
add a comment |
$begingroup$
Why can't a nonconstant harmonic function have a local minimum? The averaging property. The value of a plane harmonic function at a point $x$ is equal to its average value on any disk centered at $x$.
So, then, if $f(x)le f(y)$ for all $y$ with $|y-x|_2 le r$, then
$$0=f(x)-f(x)=frac1pi r^2iint__2le rf(y)-f(x),dy ge 0$$
with equality only if the (continuous) nonnegative function we're integrating is identically zero in the disk $|y-x|_2le r$.
Now, I should note that it's nontrivial to define the Laplacian operator $Delta = nabla^2$ on the sphere or something like it; the partial derivative operators we built it from are based on a specific set of coordinates, and there's no single set of coordinates that defines the sphere. We can't fall back on the (coordinate-independent) exterior derivative, either; while the gradient map from $0$-forms to $1$-forms and the divergence map from $(n-1)$-forms to $n$-forms are both cases of the exterior derivative, the map we need to bridge the gap from $1$-forms to $n$-forms is strongly coordinate-dependent. Even in the $n=2$ case.
How do we fix this and get something we can work with? To get an idea of what's going on, note that the Hessian matrix of second derivatives transforms by a congruence relation: if $g(x)=f(Ax)$, then $H_g=A^TH_fA^T$. The Laplacian is the trace of this matrix. The Laplacian is the trace of this matrix, so we would like to restrict ourselves to transformations that preserve the trace. What does that? Orthogonal matrices; if $A^T=A^-1$, then that's a similarity relation, and the trace is fixed.
Orthogonal matrices preserve distances and inner products. If we want this generalized Laplacian to work out, we need a notion of inner products on our manifold. That's a Riemannian metric - a system of inner products on the tangent spaces. So, then, to find the Laplacian, choose local coordinates $x_1,x_2,dots,x_n$ so that the inner product at the point we're interested in is the Euclidean inner product. The Laplacian at that point is $fracpartial^2partial x_1^2+fracpartial^2partial x_2^2+cdots+fracpartial^2partial x_n^2$, and this is invariant under any isometry.
Now, looking back at our argument - the averaging property depends on working in a flat space. If the metric varies point-to-point in those local coordinates we found, so will the formula for the Laplacian - and the averaging property is no longer something we can explicitly write down. What do we get instead? The Laplacian becomes
$$Delta f(x,y) = a(x,y)fracpartial^2 fpartial x^2+2b(x,y)fracpartial fpartial xpartial y+c(x,y)fracpartial^2 fpartial y^2 = texttr(MH)$$
in our local coordinates, where the matrix $M=beginbmatrixa&b\b&cendbmatrix$ is always positive definite and $H$ is the Hessian matrix of second derivatives.
And there's your second part. We need it in order to deal with harmonic functions on any manifold that isn't flat. The extra condition we need falls right out, too - that matrix has to be positive definite everywhere. If $M$ is indefinite, we can move to a basis of eigenvectors and then set $H$ to a positive diagonal matrix while still having that operator zero, giving us a local minimum.
As for the difference between real symmetric and Hermitian? We're adding the off-diagonal elements of $M$ with their transposes anyway, so the antisymmetric imaginary part zeros out. There's no reason to go to complex coordinates unless we started with them.
This more general result follows from the Hopf maximum principle. On a closed disk in our local coordinate patch, the maximum and minimum must occur on the boundary of the disk, so the center can't be a local extremum unless the function is constant. Every point on the manifold has coordinates of this form, so there aren't any local extrema unless the function is constant.
$endgroup$
Why can't a nonconstant harmonic function have a local minimum? The averaging property. The value of a plane harmonic function at a point $x$ is equal to its average value on any disk centered at $x$.
So, then, if $f(x)le f(y)$ for all $y$ with $|y-x|_2 le r$, then
$$0=f(x)-f(x)=frac1pi r^2iint__2le rf(y)-f(x),dy ge 0$$
with equality only if the (continuous) nonnegative function we're integrating is identically zero in the disk $|y-x|_2le r$.
Now, I should note that it's nontrivial to define the Laplacian operator $Delta = nabla^2$ on the sphere or something like it; the partial derivative operators we built it from are based on a specific set of coordinates, and there's no single set of coordinates that defines the sphere. We can't fall back on the (coordinate-independent) exterior derivative, either; while the gradient map from $0$-forms to $1$-forms and the divergence map from $(n-1)$-forms to $n$-forms are both cases of the exterior derivative, the map we need to bridge the gap from $1$-forms to $n$-forms is strongly coordinate-dependent. Even in the $n=2$ case.
How do we fix this and get something we can work with? To get an idea of what's going on, note that the Hessian matrix of second derivatives transforms by a congruence relation: if $g(x)=f(Ax)$, then $H_g=A^TH_fA^T$. The Laplacian is the trace of this matrix. The Laplacian is the trace of this matrix, so we would like to restrict ourselves to transformations that preserve the trace. What does that? Orthogonal matrices; if $A^T=A^-1$, then that's a similarity relation, and the trace is fixed.
Orthogonal matrices preserve distances and inner products. If we want this generalized Laplacian to work out, we need a notion of inner products on our manifold. That's a Riemannian metric - a system of inner products on the tangent spaces. So, then, to find the Laplacian, choose local coordinates $x_1,x_2,dots,x_n$ so that the inner product at the point we're interested in is the Euclidean inner product. The Laplacian at that point is $fracpartial^2partial x_1^2+fracpartial^2partial x_2^2+cdots+fracpartial^2partial x_n^2$, and this is invariant under any isometry.
Now, looking back at our argument - the averaging property depends on working in a flat space. If the metric varies point-to-point in those local coordinates we found, so will the formula for the Laplacian - and the averaging property is no longer something we can explicitly write down. What do we get instead? The Laplacian becomes
$$Delta f(x,y) = a(x,y)fracpartial^2 fpartial x^2+2b(x,y)fracpartial fpartial xpartial y+c(x,y)fracpartial^2 fpartial y^2 = texttr(MH)$$
in our local coordinates, where the matrix $M=beginbmatrixa&b\b&cendbmatrix$ is always positive definite and $H$ is the Hessian matrix of second derivatives.
And there's your second part. We need it in order to deal with harmonic functions on any manifold that isn't flat. The extra condition we need falls right out, too - that matrix has to be positive definite everywhere. If $M$ is indefinite, we can move to a basis of eigenvectors and then set $H$ to a positive diagonal matrix while still having that operator zero, giving us a local minimum.
As for the difference between real symmetric and Hermitian? We're adding the off-diagonal elements of $M$ with their transposes anyway, so the antisymmetric imaginary part zeros out. There's no reason to go to complex coordinates unless we started with them.
This more general result follows from the Hopf maximum principle. On a closed disk in our local coordinate patch, the maximum and minimum must occur on the boundary of the disk, so the center can't be a local extremum unless the function is constant. Every point on the manifold has coordinates of this form, so there aren't any local extrema unless the function is constant.
answered yesterday
jmerryjmerry
14.1k1629
14.1k1629
add a comment |
add a comment |
Thanks for contributing an answer to Mathematics Stack Exchange!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
Use MathJax to format equations. MathJax reference.
To learn more, see our tips on writing great answers.
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f3137252%2fnonconstant-solutions-of-homogeneous-poisson-equation-on-domain-without-boundary%23new-answer', 'question_page');
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
$begingroup$
$x^4y^4$ is not a harmonic function
$endgroup$
– Dylan
Mar 6 at 8:32
$begingroup$
I know. But it is an example of a function that has a local minimum yet $partial_x^2 f + partial_y^2 f=0$ at that local minimum. If the answer on physics stack exchange is correct, then harmonic functions cannot have this type of local minimum. Why not?
$endgroup$
– Wouter
Mar 6 at 8:53