noc21 - ma33 - lec07

IIT Kanpur July 201
20 followers
1 Video View·Dec 3, 2023
In this lecture, we are going to talk about fundamental solutions, right. Now, before
we move on, we are going to talk about Laplace equation in this video, Laplace equation.
So, this is our continuation video on Laplace equation. And essentially, let us just clear
it up what we exactly want to achieve here. So, essentially, we are going to talk about
this is what we need to understand, the well posedness, well posedness of the Dirichlet
problem, Dirichlet problem given by, I will explain why I call it a Dirichlet problem
given by minus Laplacian of u equals to f in omega, as usual omega is an open subset
of R n and u restricted to the boundary of omega is g, ok. This is the problem. So, as
you can understand, this is a Poisson problem, right. It is a Poisson equation, because Laplacian
of u equals to 0 is called the Poisson equation, if you remember, ok. This is after a mathematician
by the name of Poisson and a French mathematician. And if f equals to 0, that is a Laplace equation.
If you have an arbitrary f, that is a Poisson equation. And u, when you restrict u to the
boundary of omega, that is going to be some function g. Now, here, here f and g are given
smooth functions, ok, given smooth functions. I mean, here I am not exactly specifying what
sort of functions are there, but let us just assume there are smooth functions, ok.
So, why it is called Dirichlet? Let us understand this thing. See, let us say you are given
a Poisson equation, right, minus Laplacian of u equals to f, yeah. With this, you can
specify different type of boundary conditions, yeah. So, first of all, let us look at the
equation on the boundary conditions. So, different boundary conditions, different boundary conditions.
Let us assume that you are given a Poisson equation, which looks like this, Laplacian
of u equals to f, ok. Let us call that 1. So, 1 plus this boundary condition that u
restrict to the boundary is g, g is some function. So, essentially you are given the information
on u in the domain. So, Laplacian of u, you know that it is equals to f in omega and the
boundary, on the boundary, the unknown function u will look like g, yeah.
If this sort of boundary condition is given, we call it a Dirichlet boundary condition,
Dirichlet boundary, yeah, Dirichlet boundary. Now, let us say that 1 is given plus this
sort of boundary condition is given, del u del gamma at the boundary, ok, that is equals
to g, let us say, ok, where gamma is the unit outward normal, outward, ok, to del omega,
ok. If that is the case, then this sort of boundary condition is called the Neumann condition,
ok. It is called a, how do I put it? So, that, this is called a Neumann condition, Neumann,
ok. There are other boundary conditions also.
Let me give you another example. So, let us say that is 0, 1, that is 0, 2 and 3. There
are other boundary conditions, you can just specify, but for now, we will just give one
other condition. So, 1 plus, ok, so obviously this equation is there, equation plus a boundary
condition, ok. So, boundary conditions look like this, something like, let us say, u plus
del u del gamma, del gamma, ok. So, this whole thing restricted to del omega, let us
say that is equals to g, ok. So, you have this equation and you have this boundary condition
u plus del u del gamma restricted to the boundary is g, yeah. This sort of boundary condition
is called a mixed boundary condition, mixed boundary condition, ok.
Now, there are other conditions also. So, let us say in some part of the boundary, let
us say, so let us say this is your omega, ok, that is your omega, let us say this is
your omega and that is your boundary, right. Now, it may happen that in this part of the
boundary, let us put it in red color, yeah, it may happen that you are given a boundary
condition like in this part of the boundary, it is g and in this part of the boundary,
so u is g here on the boundary, ok, and del u del eta, let us say that is some h, some
function h on this part of the boundary, ok, on this part, ok. It may happen that you
are given this equation and this sort of boundary condition, yeah, that may also happen.
So, there are different options, but for now we will just restrict ourselves to the Dirichlet
boundary condition, yeah. So, let us see what happens. So, Dirichlet, Dirichlet problem,
we say it is a Dirichlet problem, yeah, ok. How does the Dirichlet problem look like?
Let us say that Laplacian of u equals to f in omega, ok, and u restricted to the boundary
del omega, this is going to be g and we will assume, and we will assume f is in C of omega
and g is in C omega del omega, ok, the continuous on the boundary and f is continuous on omega.
So, this is the only condition which I am just taking.
Now, the thing is this, what do we need to do here? If you remember properly, when working
with this sort of problem, we said that our main motivation to study this sort of problem
is the following. We need to, we need to understand that, of course, ok, here, whenever
we are doing omega is assumed to be a open connected set in R n, clear. So, that is always
assumed. Now, f is continuous and g is continuous on the boundary. We want to talk about the
well-posedness of the problem, well-posedness of, let us say that is your 1 of 1, ok, well-posedness
of 1, ok. Now, first of all, let us start with this
thing. See, we will not directly attack this problem, ok, we will take it part by part.
So, first of all, what we are going to do is we are going to assume, yeah, we are going
to assume that, so, well-posedness, what does well-posedness, if you remember, first thing
is existence, right, existence. So, first of all, you have to find that if there is
a solution or not. Number 2, uniqueness, that is, that if there is a solution, it is unique.
And number 3 is the stability. Stability is, I mean, how does the solution behave? So,
the solution must be small when the initial data is small. So, basically, let, let me
write it as stability of initial condition, stability of initial condition, let us put
it this way, ok, initial data, ok. Now, let us look at what are the problems
associated to this, ok. So, let us, see, most of you guys, I am quite sure, you know, let
us say if I am asking you to solve this problem, Laplacian of u equals to 0 in omega, clear,
and omega is given to be something like this, let us say in two dimension, yeah, this is just a
motivation, motivation. In two dimension, let us say it will look like this, right, a, b cross c,
d, yeah. So, in this thing, I want to solve this equation, yeah. So, in this rectangle, let us say,
I do not know, maybe something like this. So, this is a, b cross c, d, this rectangle,
let us say, this is omega, yeah, this is omega. Now, if I am asking you to solve this problem,
what do you do? You can use a separation of variable, right, and what is it? Just roughly
speaking, it will look like this. So, you are basically looking, trying to find a, you know,
solution u of x, y, which will look like some function of x times some function of y, ok.
Now, if, of course, you are thinking of this as a C 2 function. Now, since this is a solution,
you are thinking of this as a solution of this. So, you can calculate u x x, u y y,
and after that put it in that equation. Given the boundary condition, you can
use Fourier series to find out what the solution is, yeah, you can do that, ok.
Now, let us say, from here, if I am asking you to, so, this can be done, right. So,
this is separation of variable, separation of variable. So, as you know in the introductory
video itself, I have told that we are not going to study about separation of variable in this
course, right, because, I mean, that is not, to be very frank, it is not very interesting thing
to do. I mean, it is just a method to solve it, very significant method, but I mean,
nonetheless, just one method to solve this. Now, the thing is, let us say you are, you want,
I am asking you to solve this problem, and omega is in A B cross C, yeah, how do you solve this
thing? So, if you remember your separation of variable methods, what you do is, first of all,
you try to find, try to find the eigenvalues, try to solve the eigenvalue problem, eigenvalue
problem, problem, Laplacian of u equals to lambda u, ok, minus Laplacian of u in omega, that is the
omega A B cross C D, and u restricted to the boundary detail omega is g, ok. You just solve,
you just try to solve this eigenvalue problem, and then, I mean, you can use the data from here
to actually, I mean, you can use the, you know, function from here, the solutions, and you can
construct a solution for f, you can construct solution for 1. So, using, using the linearly
independent eigenfunctions, so once you solve the, this equation, let us say this is your 2,
once you solve 2, you get some eigenfunctions, right. So, when you use this eigenfunctions,
I am getting the circular, ok, so using the linearly independent eigenfunctions, ok,
um, one can construct the solution of 1, right, you, you remember, yeah, so that is the separation
of variable. Now, what is the restriction with this method? See, if this method works in any
given domain, then we do not have anything to do, right, I mean, there is nothing to do here,
I, you can just use separation of variable to solve this problem, and you can look at the
explicit solution, you use the explicit solution to, I mean, that is the point, right, once you
get an explicit solution, you do not have to do anything, existence, uniqueness, stability,
all of this can be easily checked. But the problem is this, separation of variable,
please remember this thing, yeah, why we were doing this thing. The main restriction,
restriction of separation of variable, this method, method is that it works
on certain type of domains, type of domains. So, most, if you remember, just work out one
problem using separation of variable, you will understand. See, if it is not in this form,
A B cross C D, yeah, not in a rectangular form, then this separation of variable would not work.
You can think of, I mean, of course, in the, you know, other coordinate system, you can take,
think of a circle also on a disc and all, you can actually solve it, but you are essentially,
you know, whenever you are doing this change of variable to radial coordinate or spherical
cylindrical coordinates, you are essentially doing something like this. So, essentially,
for a domain which looks like a rectangular kind of domain or like a spherical domain,
that sort of domain, separation of variable works, ok. But most of the problems which we
encounter, that is not a, I mean, not a given domain like this, right. I mean, it can be any
arbitrary domain. What do you do there? That is the question, ok. So, let us look at what do we
do there. See, what we do is this. Question is, well posedness for an arbitrary domain.
I hope I have made it clear what, why we need to study this. See, again, let me recall,
the point is this. Using separation of variable, you can, of course, get an explicit solution,
right, yeah. But the problem is the domain has to be very specific and to be very precise,
the domain has to be a rectangular domain, yeah, or a circular domain, that sort of thing,
a cylindrical domain. But most of the times in applications, when you talk about domains,
it may not look like that, right. Now, question is, how do you talk about well
posedness for this? So, you have to find an explicit solution. Can you always find an
explicit solution? The question is, can one always derive an explicit solution on an
arbitrary domain, ok. So, this is the rephrased version of this, yeah, well posedness,
the rephrased version. But can you find an explicit solution for an arbitrary domain,
ok. So, the answer is this. Theoretically, it is possible, but practically you cannot find.
So, practically, practically not possible, not possible, ok. Theoretically, it is possible,
we will talk about that, we will use Green's function to do that, ok.
So, let us just explore more. So, let us just say that, so there are three components, right,
of well posedness, existence, uniqueness and the stability with respect to the initial data,
ok. Let us start about, talk about uniqueness first, yeah, let us just get that out of the way.
So, consider one, consider this problem. Let us start with Laplacian u equals to f in omega,
u equals to g on the boundary, ok. We are interested, we are interested,
interested in the question of uniqueness first, question of uniqueness. Now, please understand
that we are not, we have not talked about existence because that is a difficult part.
First of all, let us say that we will assume that there is a solution. We want to see that
if there is one solution, yeah, can we find other solutions from there for this problem,
ok, that is the question. See, if we can find, if we can solve this problem,
if you take f equals to 0, then the Laplace equation is also done, right,
ok. So, this is a much general problem. So, the question is, can, so assume,
therefore, let us just put it this way, assume one admits, admits a solution u in C 2 of omega
bar, clear. Let us say that one admits a solution of u in C 2 of omega, ok, then that solution,
that solution is unique, ok. So, this is a theorem. I do not know how to, let us say theorem 1,
ok. So, what it is essentially saying is this. See, first of all, we need to study well-posedness,
right. We do not know whether that problem has a solution or not, but we are talking about
uniqueness first, yeah. In the well-posedness part, three part are there. We just want to
talk about uniqueness first. We are saying that if there is one solution, yeah, at least one
solution, if you can just show that there is at least one solution, you can show that that
solution is unique. So, we are assuming that there is at least one solution will show that
you cannot find any other solution. That is only one solution, ok. There is only one solution to
work. So, how do you find it? Let us just take, look at the proof of this thing, ok.
Now, I hope you understand what I meant by one solution. So, there is a,
um, ok, maybe before I go on doing this thing, maybe I can talk about what do I mean by solution
of this thing, yeah. It will be better that way. So, just a solution of one, sorry, I,
I just messed up the order a little bit. Solution of one, what do I mean by a solution of one,
then, then we can talk about this. So, solution of one, as you can understand
that u which is in C 2 omega bar, ok, is called, called a solution of one if
Laplacian of u at the point x is f of x, ok, for all x in omega and u restricted to boundary. So,
u of x equals to g of x for x on the boundary of omega, ok. Just a simple thing. It is only
that u has to be in C 2 of omega bar for now, ok. This is what we mean by a solution.
So, we are saying that if you can find a solution for this problem, then the solution is,
how do I prove this thing, yeah. So, let, let u tilde be any other solution,
any other solution of one, clear. So, define, define w to be u minus u tilde,
ok. See, initially we have one solution u, I am also assuming that u tilde is another solution
of one, ok and you define w equals to u minus u tilde, yeah. So, clearly, clearly w is in
C 2 of omega bar, right. See, u is a solution and definitely it is in C 2 of omega bar, right.
u tilde is also a solution and what do I mean by a solution? A solution to this one is a C 2
of omega bar function which satisfies this. So, if u tilde is another solution,
u tilde has to be in C 2 omega bar, yeah. So, u and u tilde are both in C 2 omega bar,
you can see that the u minus u tilde, the difference of those two which is defined by w,
ok, which is defined by w, that is also in C 2 of omega bar and clearly w is in C 2 omega and
satisfies, satisfies Laplacian of w equals to 0 in omega and w equals to 0 on the boundary.
Do you agree with me? See, if Laplacian of w, Laplacian is a linear operator, right. So,
Laplacian of w is Laplacian of u minus Laplacian of u tilde. Laplacian of u is also f, Laplacian
of u tilde is also f. So, Laplacian of w is going to be 0 and moreover, u is g on del omega and u
tilde is also g on del omega. So, w is, so w is 0 on the, 0 on the boundary, right, ok.
So, see, essentially if we define it like this, w is going to solve the Dirichlet problem,
w equals to 0, but not the Poisson problem. So, it is essentially a Laplace equation,
here w solves the Laplace equation, ok. Now, what happens? When it does that,
then you take a integration by parts, ok, again integration by parts, use integration by parts,
what does it do? See, if you multiply this with the w, yeah, multiply both sides with w, ok.
So, let us say w times Laplacian of w integral over omega, that is going to be 0, yes,
because Laplacian of, I mean, let us assume that w is not 0, yeah. See, if w is, so let us assume
that this is the problem given to us, yeah, if we can show that w equals to 0 is the only possible
solution of this, then we are done, right. Now, because if w is 0, u is equals to u delta and
you are done, right. So, there is only one solution. Now, let us assume that w is not 0.
So, we can just multiply it by w. So, w Laplacian w, that is also 0,
ok. Even if it is 0, it does not matter, but anyways. So, integral omega w Laplacian w
equals to 0. Now, if you do an integration by parts here, what will it look like? Omega gradient
of w square, ok, dx, this is with respect to dx, I forgot to write dx, dx. This is equals to 0,
yeah, because there is a boundary term, but in the boundary it is 0, w 0. So,
the boundary term is not there. So, this is equals to 0, if you remember, yeah.
So, this is integration by parts. If you do not know this, please look at the first lecture,
ok, there this integration by parts is done. So, this will give you, see, gradient w square,
this is a positive function, right. You are integrating a positive function, yeah, minus,
you can just throw it outside, it is not a problem. So, you are integrating a positive function over
omega and you are saying that is equals to 0. So, the positive function gradient of w has to be 0,
right. So, what does that mean? It means the gradient of w is 0 in omega, right.
Thus, the integral of a positive function is 0. So, the function has to be 0, yeah,
the function has to be 0, ok. Now, so if the function, sorry, yeah, so the function has to
be 0. So, here the function is gradient of w. So, gradient of w is 0, yeah, and it is given that w
is equals to 0 on the boundary, ok. What is w? w is a C 2 omega bar function. So,
it is continuous till the boundary, till the boundary of omega, right. On the boundary,
it is 0 and the derivative of w is 0 on the interior omega. What does that say? It says
that w has to be 0 in omega, right. So, that will imply, that implies w is 0 in omega, yeah.
See, if w is non-zero somewhere in omega, yeah, it is a constant, that constant has to be there
up till the boundary because, you know, gradient of w is 0 in omega, yeah. If the gradient is 0,
then the function has to be constant. If it is a constant in omega and the constant is not 0,
so on the boundary, it cannot take 0, yeah. See, if w is a constant which is non-zero
in the interior of the domain, that cannot be 0 on the boundary because w is continuous function,
right. So, w has to be 0 on omega, in omega and that will imply u is equivalent to u tilde in
omega, yeah. And hence, there is a unique solution. So, you do not have to worry about this.
See, we also proved that the Laplace equation has a unique solution, right. So, essentially what it
is saying is this. If you have a Poiseuille equation which looks like this, Laplace and u
equals to f in omega and u equals to g on the boundary, obvious, of course, f and g are two
nice functions. If you are given something like this, then these functions admit a, I mean,
you see, I do not know if there is a solution or not. If you just give me one solution,
then I can guarantee you that that solution is a unique solution, yeah.
So, uniqueness is out of the way. I mean, we do not care about uniqueness anymore.
Now, the question is, if we can find such a solution or not, ok, and this is the place where
it gets a little complicated, ok. Let me give you a characterization of solutions and then from there,
we will do a, let us, construction to find solutions, ok. So, this is a characterization
of harmonic functions, characterization, not harmonic functions, I mean, you know,
I mean, solutions of the Poiseuille equation. So, characterization of solutions to
the Poiseuille equation, ok. We want a characterization of this thing.
So, this is actually, this characterization, this is not very, very important for this course,
but I just want you to know how this works. But this is the basis. So, essentially,
let me give you a small note. This is the basis of modern PDE, ok. If you want to do modern PDE,
this is what you need to do. I mean, this is all we do actually, ok. This is the basis of modern
PDE and is a branch and is a topic, I think you can say, is a topic called variational principle,
variational principle in calculus of variation, calculus of variation, ok.
What I meant by this is the below thing, whatever I am doing right now, the characterization of
solutions, this, for this course, it may not be very, very important, but I mean, of course,
we use this thing, but the thing is not very important, but this is most important if you
are looking for a modern approach to PDE, ok. So, whatever we are doing is like 100 to 200 years old,
ok, in this course, yeah, but if you want to do the modern part, this is the modern part.
So, just to give you the taste of that modern part, I am just doing this part, ok. So,
essentially, you are given this Poisson equation. So, let me write it down. Let us say,
Laplacian of u equals to 0 equals to f in omega, ok, u equals to g on the boundary, yeah,
our usual problem, we are calling it 1. So, this is your 1.
Now, I want to talk about the solution of this problem, ok, and while talking about the solution
of this problem, I will define, define the energy functional, ok. So, what I meant by this is,
I will define a new object. For now, you just think of this as a new object called I of w,
ok, and that will be given by integral over omega half gradient w square minus w f, ok, dx.
Maybe I can put it in a small dx, ok. So, that is I of w, yeah, and it is not like,
I mean, it is coming from heavens, it is not like that. There is a proper scientific way of
how this is coming, and this is, that is what we are going to, I mean, explain now.
Essentially, let me just put down where w is, where w belongs to this set A,
ok, and how do we define this set A? It is defined as all the functions in C 2 omega bar,
ok, such that, so you define A like this, yeah, you define a set A and you take w like this,
such that w equals to g on the boundary, ok. So, you are looking at a C 2 omega bar function
such that w equals to g on the boundary, and you look at, I mean, you define a I,
which is from this set to R, ok. So, you define I. Hence, I is defined from this set to R,
ok, is a linear functional. Functional is just a function, we call it a functional,
because it is a function from a function space to R. It is taking a function and putting it into,
giving you back a real number, that is why it is called a functional, ok.
Sorry, it is not a linear functional, it is a function, ok, we call it a energy functional.
So, essentially this actually, so, I mean, encapsulates the total energy of the system,
ok, but let us not go there, let us just talk about what is so special about this and what
is the relation between one and this energy functional, yeah. So, for that, we are going
to talk about a theorem, we are going to prove a theorem. So, it is called a Dirichlet theorem,
Dirichlet principle, principle, ok. What it says, Dirichlet principle, it says that assume,
let u is in C 2 omega bar, ok, and it solves, solves 1 here, then I of u, which is the minimum,
ok, w is in A, I of w, ok. So, what it is saying is, if you have a solution,
it is saying that this C, let us say, I am looking for a solution of this problem, right.
Let us assume that you guys already know what the solution is. If there is a solution of this thing,
that solution actually minimizes this functional, is it clear? This functional is some object,
right. If there is a solution, that solution will be in this A, of course, it is, yeah.
And you see, what it is saying is, if you, if you put that u in I, so I, if you put
u in that I, in that function, then that will give you I of u, yeah, and what is I of u,
that will be the minimum of all these I's, ok. So, it minimizes the, any solution of 1,
yeah, we have seen that any solution of 1, there is only one solution to this problem, right.
So, what it is saying is, that solution actually minimizes the energy functional,
this minimizes this particular stuff I w, ok. So, you can take any w from this set A,
does not matter, that w, the value of w is always greater than equal the value of I at the point u,
where u is the solution of this problem, right. There is a converse part also. Conversely,
this is very important, conversely, if, if u is in A, ok, and satisfies, satisfies,
let us call this thing as 2, ok, this 2, the minimum of this is I of u, satisfies 2, ok,
then u solves 1. So, very important, yeah, and very, very exciting thing.
What it is saying is this, let us say that you have a function, if you have somehow a solution,
so first of all, it is saying that you give me a solution, ok. Once you give me a solution to 1,
once you give me a solution to this problem 1, you guys already know that if there is a solution,
it has to be a unique solution, right. So, once you give that unique solution to me,
I can say that that solution minimizes this functional I w, which is defined by this,
that solution minimizes. Conversely, if you can find a minimizer to this solution,
so basically it is saying that if you can find a u in A, in this set, which satisfies this,
which satisfies this, it means that, see, it means that there is a function u, where the
minimum is attained, where I is attaining its minimum, that, that is u, u, ok.
So, basically it is saying that if there is a function u, where I is attained,
sorry, the minimum is attained, then you can say that that function solves 1, ok. So,
both sides are true, this is called a Dirichlet principle. So, essentially, you see, if you want
to solve the problem, this problem, the Poisson equation in an arbitrary domain, what you can do
is, you just look at the energy functional and try to minimize the energy functional, ok.
If you can, let us say, somehow manage to find a minimizer to this energy functional,
that, then that minimizer will turn out to be a solution to this problem, ok. So, let us
prove this theorem. So, choose, I will prove this part first. So, we choose w in A, w in A,
then from 1, we have, we have c, if I take everything together, Laplacian of u minus f
ok, and u minus w, w I am taking from A and u solves this 1, right. So, I am taking u minus w.
So, this is on omega, ok. So, you see, Laplacian of u equals to f in omega. So,
this particular thing is 0, irrespective of what, where w is, u and w are in C 2 of omega,
in A, u and w is in A, right, ok. So, u minus w is also in A, right. So, that is, sorry,
u minus w is not in A, u minus w. So, u minus w is not in A, ok. Please remember this thing.
So, basically, but it is in C 2 of omega bar, but u minus w is in C 2 of omega bar,
ok. Why it is not in A? Because on the boundary, this is going to be 0, right, but on A,
the boundary u minus w has to be g, if it is, has to be in A, ok. So, it is not in A,
but it is in C 2. So, basically, the integral has to be 0, why? Because this is 0,
Laplacian of u minus Laplacian of u is equal to f in omega, right, ok, f in, you can see.
So, let us write it like minus Laplacian. This is just a convention, nothing else, yeah,
minus Laplacian u equals to f, u equals to g. So, see what I meant by this is. So,
it means minus Laplacian of u minus f is 0 in omega, yeah, so that if you multiply it by u
minus w and integrate it over and over, that is going to be 0, ok. Now, we use integration
by parts, integration by parts. If we use integration by parts, what happens? It is
minus Laplacian of u times u minus w minus integral over omega f u minus w equals 0,
ok. So, what does that, I mean, there is a dx, in all of this there is a dx, I am just
forgetting to write it down, there is a d of x, right. Now, if you use a integration by parts
here, so this is not a integration by part, it is just whatever I write, here I am using
integration by parts, ok. So, that will give you gradient of u dot gradient of u minus w dx
minus equals to integral over omega f u minus integral over omega f w,
again I forgot to write dx, dx, ok, right. So, I have this. Now, let us just look at,
ok. Now, what happens, see if you are doing integration by parts, yeah, let us just write
it like this, this is minus, this is equals to 0. So, if you are doing integration by parts,
there is a boundary term, right, but u minus w is 0 on the boundary, that is why we just
multiplied by u minus w, ok, u minus w that is 0 on the boundary, right, ok. So, there is a
boundary term if you remember in integration by parts, that boundary term is not there, ok.
So, now what we do is, what it implies is this, see this implies that gradient of u square,
ok, gradient of u dot gradient of u, gradient of u square minus integral omega u f dx, dx,
ok, this is equals to, I will take that part here. So, integral gradient of u dot gradient of w,
I hope it is fine, minus integral omega u w dx, right, again I am always forgetting dx, dx is
there, ok. Now, if this is there, see what does that say, this says that, let us do this thing a
little bit, you see, this is, here I am using this thing, you guys already know a b is less than a
square by 2 plus b square by 2, right, that is just a easy algebraic inequality. So, from here
what do we get, we get integral omega gradient u square dx minus integral omega u f dx, ok,
this is less than equals to half integral omega gradient u square, ok, plus half omega gradient
w square minus integral omega u w dx, yeah. How is it coming? See, integral over omega
gradient u dot gradient, so integral over omega gradient u dot gradient w, ok, this is dominated
by integral over omega gradient, mod gradient u dot mod gradient w, yeah, and then you use this
inequality to get this, right, ok. So, once you get this, if you take this part here,
it will give you half integral omega gradient u square minus integral omega u f dx, I mean,
please forget me, I am always forgetting dx. So, this is less than equals to, I mean,
if you want you do not write dx, that is not a problem, I am always forgetting dx.
So, this is half integral omega gradient of w square dx minus integral omega u w dx,
ok. So, if you remember what is it, this is I of u, this is I of u and that is less than equal
I of, this is w, right, ok. And since, since u is in A, yeah, u is in A, right, so therefore,
minimum I of u, see w is any arbitrary, if you remember, see w is an arbitrary point in A,
right, and u is in C 2 omega such that u on the boundary is g. So, u is definitely in A,
right. So, you see u is in A, that implies I of u is equals to the minimum of I w,
w is in A, yeah. See, it is saying that this is the infimum, but u, since u is in A,
that infimum is achieved and that is why this is the minimum here, ok.
So, you have seen that if there is a solution, that solution is the minimum of this energy
function. Let us look at the converse part. Conversely, ok, let us say I of u is less than
equal I of w, ok, and this w is in A, yeah, it is given, yeah, it is given to you. So,
sorry, I should write it as a minimum of I w, I mean, I want to write this thing. Conversely,
I am assuming this, I will show that you, so maybe I can write it as. Conversely, let us say
this is given to you, minimum of w in A I of w. Now, I want to show that you solve the equation.
How do I solve it? So, you fix, fix a v in C C infinity omega, ok. So, basically C C infinity
omega, I am not quite sure if I defined it, it is the set of all, all compactly supported,
compactly supported smooth functions, ok, smooth functions in omega. Compactly supported means,
what I meant by this is, so basically support of a function is a set of all those x such that,
ok, f of x is not 0, ok, the closure of this thing. If you take the closure of this thing,
that set, this is called the support of, this is called the support of f, yeah, this is the set,
support of f is the set where f of x is non-zero, the closure of that set, ok.
So, you are basically, and if that, this kind of thing is contained in omega, yeah, if you are
looking at a function such that this happens, then we say that the function has a compact
support, yeah, in omega. So, essentially you can think of this, I mean, let us say a function which
is 0 on the boundary and some non-zero part in the interior, that kind of function has a compact
support, yeah. For example, let us say that is your boundary, let us say this is your omega,
ok, one dimension, let us just look at one dimension and this sort of function are you
talking about, ok. So, this has a compact support, yeah, because it is non-zero in a compact set,
in this set, if you take the closure, that is a compact set and it is 0 everywhere.
So, basically the compact set is contained in omega and this is a, I mean, example of a function
which has a compact support. So, you start with a v which has a compact support in omega and it
is smooth, C infinity, ok, so infinitely revaluable and define, define I of tau to be
capital I of u plus tau v, ok. u is this minimizer, it is given to us, v is a C C infinity function,
I am just taking the convex combination of those two, u plus tau v, ok, and where is tau?
Tau is in R, clear. Now, you see, I want to look at what I does, see, u plus tau v, do you think
it is in A? Of course, it is in A, right, for every tau in R, of course, it is. Why? Because,
you see, this is the point, why we are, why we do want a compact support function, because
v is 0 on the boundary for a compact support function, right. So, on the boundary, u is
basically g, ok, and since v is C C infinity, it is also C 2, yeah, so essentially this whole thing
is in A, u plus tau v. I, this function, therefore, I has a minimum, minimum at tau equals to 0,
right, because I of u, at the point u, I attains a minimum. So, the small i, this will attain a
minimum at tau equals to 0, because at tau equals to 0, capital I is u, right. So, capital I of u,
that is where the minimum is attained, so that is why small i attains its minimum at 0.
So, if it is, attains its minimum at 0, I can talk about I prime of 0, that is equals to 0,
right, prime is with respect to tau. So, I prime is d d tau, right, ok. So, of course,
the derivative exists, right, the derivative exists and it is, I mean, you can, of course,
show it is differentiable, it is not a problem, ok. If you are not convinced,
please check that the derivative exists here, ok. So, I prime 0 is 0. Now, if I prime 0 is 0,
see what, what is happening, if that is 0. I of tau, ok, this is equals to half integral
gradient of u plus tau v square minus integral u plus tau v times f dx, right, this is what happens
and that is equals to half integral gradient u square plus tau gradient u dot gradient v,
again I am forgetting dx, all that, ok, so let me do it like this, omega plus tau square by 2
gradient v square minus u plus tau v f dx, ok, so this is the case.
So, therefore, 0 equals to I prime of tau, I prime of 0 essentially,
see I attains minimum at 0, so I prime of 0 is 0, so that will be giving, you just take the
derivative with respect to tau, if you take that, this is going to be 0, this is, this is remaining
and this is remaining, right. So, it will look like integral over omega gradient u dot gradient
v, ok, dx minus integral over omega v f dx, ok. This is equals to integral over omega minus
Laplacian of u minus f dx, right. Please understand, I of tau will look like this,
right. So, if you take the derivative, that derivative with respect to tau, that will go
inside the integral because this integral is with respect to x, yeah, it has nothing to do with tau.
So, I can take the derivative inside, if I am taking the derivative, this term is not there,
this is, this term is only, I mean the derivative of this term with respect to tau is only gradient
u dot gradient v, which is this and here it is tau, tau times this one and tau at the point 0,
this term would not be there, ok. So, the only term which will be there is minus v f, which is
the derivative of this term is v f, so minus v f, right. So, it is this.
Now, you see, this holds, this identity, ok. So, let us say that is 3, 3 holds for all v
in C C infinity omega, right. If this holds for all v in C C infinity omega,
that will imply Laplacian of u equals to f in omega, ok, because C, if this holds for C,
so this is a small exercise if you want, exercise. So, let us say u is in C 2 of omega bar be such
that u times phi, this is dx is 0 over omega for all phi in C C infinity omega, that will imply
u has to be 0 in omega, ok, that is a small example. So, it is basically saying that if you
have a C 2 function such that integral over omega u phi dx is 0 for all C C infinity functions,
then it means that u is 0 in omega. So, please do this thing, this I will put it in an assignment,
ok. So, if this is true, then that will imply that Laplacian u equals to f and hence you see
u, u is in alpha, so u is g on the boundary, it is always given and we have also showed that
Laplacian u equals to f in the omega and hence u solves the, therefore, u solves 1, ok.
So, what we proved is if you want to solve 1, this equation, if you want to solve the 1,
then you have to find a minimizer of this functional and vice versa,
ok. With this, we are going to end this lecture.
