Lecture 48: Laplacian equation

IIT Kanpur July 201
20 followers
Sep 26, 2024
Welcome back. In the last lecture, we have seen that if we are considering F is in source
the heat equation on r cross 0 infinity and u of x 0 is equal to F of x on r, then u of
x t which is equal to F convolution of the heat kernel h t of x where h t of x is equal
to 1 by square root of 4 pi t e to the power minus x square by 4 t, this is going to be
a solution. This is what we had observed and also we have observed that u x t converges
to F uniformly as t goes to 0 as well as we have seen that not only this in the square
mean also u x t converges to F of x, okay. So, now what we are asking about that whether
this is going to be unique or not. So, as you can see that if F is in source space and
h t also we have taken it to be in the source space. So, for each fixed t, u of x t is in
source space. For each t greater than 0, u of x t this belongs to the source space
of r in the space variable. So, what we will observe that this is actually in certain sense
this is uniform over t. So, what do I mean by that? So, let us write it as a proposition
and for each t bigger than 0, supremum over x belongs to R and 0 is less than t less than
mod of x to the power k and mod of del L by del x L u of x t. This is finite. That means
this in some sense this is uniform over t for all k L belongs to N, okay. So, first we will show
for one u and for the derivatives the proof is exactly the same. So, now look at mod of u of
x t. This is less or equal to integral minus infinity to infinity mod of F of x minus of y
and then this is h t is a positive non positive function. So, this is dy. So, now I break this
y integral into two parts. Mod of y is less or equal to mod of x by 2 and integral of mod of y
is greater than mod x by 2. So, now first let us look at this is I1 plus I2. Now for I1 we know
that F is in Schwarz space. Thus mod of F of x minus of y is less or equal to some constant which
depends on N and 1 plus mod of x minus y to the power N. Now where is our y? y is less or equal
to x. So, look at that x minus of y is greater or equal to mod x minus mod y and mod y is less
or equal to mod x by 2. Therefore, this is greater or equal to mod x by 2. So, here we can get some
other constant dN 1 plus mod of x to the power N. This is for every N because of the rapid decay
of F what we are going to get this. So, therefore if I hit with a polynomial on x, so this is under
control. Now for the second part, so for 2 we have mod of y is greater than mod x by 2. Now
you look at H t of y. This is equal to e to the power minus y square by 4 t. Now this is I have
missed out a factor of 4 pi t 1 half. This is less or equal to 1 by 4 pi t square root and e to the
power minus mod x square by 16 t e to the power as you can see that as t goes to 0 this function
goes to 0 much more rapidly than square root of t. Therefore, this is a bounded function.
So, hence what we have got, hence u of x t is less or equal to some for any N some cN 1 plus
mod x to the power N and plus some constant c by square root of t e to the power minus some
constant x square by t. Now this is of course if I am hitting with a polynomial with x and
then this will be bounded over the uniform over t. Hence, similar calculation for derivatives
of u on the space variable will give us the result. That is a small observation what we have seen.
So, now let us get back to the uniqueness. So, now for the uniqueness, so what we are
dealing with? We are dealing with a equation which is this is our equation. Suppose there
exist two solutions for star that is u1 and u2. So, take v is equal to u1 minus of u2. So,
vt is equal to u1 t minus u2 of t which is equal to u1 of xx minus u2 of xx which is equal to v
of xx and v of x0 which is equal to u1 of x0 minus u2 of x0 which is equal to fx minus fx.
This is equal to 0. So, essentially if u1 and u2 are the solution of the two equation and then v
is a solution of this equation. So, this essentially says that if we can prove that vt equal to vxx,
v of x0 is equal to 0 on the boundary on r cross 0 infinity on r, then v identically equal to 0
is the only solution. If we can prove this, then we are able to show the uniqueness of the heat
equation. So, that is what we are now going to prove. This is theorem. Suppose and throughout
we are assuming that f is in the source space. Suppose u satisfies the following one u continuous
on r2 plus ut is equal to uxx t greater than 0 and u of x, 0 is equal to 0. Thirdly, u dot t
belongs to source space uniform on t because that is what we have seen that f convolution of h t
that is uniform on t, then u is identically equal to 0. That means that is convolving with the heat
kernel with this condition when f is in the source space, then what we are going to get
that it is in fact unique. And the proof is very cute and this technique is very powerful used in
PDE. So, how to prove this? We want to show that u of xt is equal to 0. And so, if we show that
e of t which I will define integral minus infinity to infinity mod of u of xt square d t,
then if et is equal to 0, then for every x. So, then what you are going to get that u of
xt is equal to 0. This is et is dx. If et is equal to 0 for every t, then what we are going
to get u xt is equal to 0. If et is equal to 0 for every t, then u of xt this is equal to 0 for
all x belongs to R and for all t belongs to 0 infinity. So, that is what we want.
Now, how to show that et is equal to 0? So, what we first observe is that actually e of 0 is nothing
but minus infinity to infinity mod of u of x 0 square dx. This u of x 0 is equal to 0 is given
to us. Therefore, this is equal to 0. Now, of course, e is a map from 0 infinity to R positive.
This is let us say 0 infinity. Now, u of 0 is 0. Now, if you can show that e is a decreasing
function on t, therefore e of t is going to be 0. So, claim e is a decreasing function.
So, how does one show that a function on t is a decreasing function? That means I need
to calculate e prime of t. If we can show that e prime of t is less or equal to 0, then
e is going to be a decreasing function. So, let us try to calculate what is e prime of
d by dt e of t. So, this is equal to integral minus infinity to infinity, we can put the
integral inside mod of u of xt square, then this is dx. Now, this is minus infinity to
infinity del by del t of mod of u of xt into u of x comma t bar dx. So, this is going to
rather del by del t of u of xt into u bar of xt plus u of xt into del by del t of u
bar of xt, this is dx. Now, what we know? We know our equation as u t is equal to u
x x. So, we will replace this minus infinity to infinity, this is u x x, this part and
then this is u bar plus u. Now, obviously if u t is equal to u x x, then u bar of t
is equal to u bar u x x bar. Therefore, this is u into u x x bar, this of dx.
Now, in the first integral what we do is that this is minus infinity to infinity u x x u bar
plus minus infinity to infinity u of u x x bar. So, by applying integration by parts, apply
integration by parts, so what we get? This is u x u bar minus infinity to infinity minus integral
minus infinity to infinity u x and then the derivative is u x bar plus in the second term
also, this is u into u x bar, this is minus infinity to infinity then minus of minus integral
u x and then u x bar, this is dx. Now, remember that u is a Schwarz class function.
Therefore, u x is also in the Schwarz class function. So, both u and u x, they are going
to vanish at plus infinity and minus infinity. So, here what we have got, this is minus infinity
to infinity mod of u x square dx. Now, this is a non-negative quantity and heated with
negative this. So, this is 0. Hence, E is decreasing. Therefore, E t which is equal
to less or equal to E of 0 which is equal to 0 and we know that E t is always greater
or equal to 0 because we have defined it to be the integral minus infinity to infinity
mod of u x t whole square dx. So, this will imply that E t is equal to 0, that is and
u is a continuous function. So, this will imply that u of x t is equal to 0, that is
uniformly over t. So, now this is u is identically equal to 0, that is what we wanted. So, what
we are saying is that if f is the initial value problem, if f is the boundary function
and u satisfies the heat equation and the condition we are imposing f is in the Schwarz
we are getting is that this is the solution is unique, okay.
So, now we will apply the similar kind of technique now for some other PDE. So, let
us take our attention to the PDE of the form. This is again we are looking on R2 plus which
is equal to here I will make a little change of the notation x is in R and y is greater
than 0. This is the upper half plane, okay. Consider delta of u which is defined to be
u xx plus u yy. So, now if we are looking at this is equal to 0 on and u of x, 0 this
is equal to f of x. This is the equation we are considering. So, there u xx was equal
to u y that is the heat equation and this is the Laplacian equation what we are considering
in the upper half plane, okay.
So, again we play the game of Fourier transform by taking the Fourier transform on the x variable
on both the side what we are going to get is that by taking this is u xx hat at xi,
y plus this is going to again change this is what u yy hat at xi, y this is equal to
0. This implies that this is minus 2 pi i xi square and u hat of xi, y plus here as
we have seen the time variable differential operator commute with the Fourier transform
on the space side. So, here we will get del square by del y square of u hat at xi, y.
Again by defining our phi of y to be u hat xi, y for each fixed xi, this will translate
it to minus 4 pi square xi square phi of y plus phi double prime of y which is equal
to this is equal to 0. Now, this is a second order ODE and with the condition given that
phi of 0 this is nothing but f hat of 0. Now, this second order ODE is going to have
a solution for each fixed xi, this is a xi what we know that e to the power minus 2 pi
mod xi, y plus some b constant depends on xi because for we are playing this game for
each fixed xi 2 pi mod xi, y. Remember that y is positive and the kind of solution we
are looking for is a bounded solution and this is no more a bounded not bounded.
So, therefore we can ignore this. So, this is we can take b xi to be 0, this is phi of
y a xi e to the power minus 2 pi mod xi, y. We know exactly that phi of 0 is equal to
f hat of sorry phi of 0 is f hat of xi. Therefore, this a xi this is going to be equal to by
this condition what we get is that this is equal to f hat of xi e to the power minus
2 pi mod xi, y. And now this if p of y if I am defining to be a function such that for
each fixed y is minus 2 pi mod xi, y then this is f convolution of p y hat at xi. Again
what we get is that the solution it is going to be then u of x t this is equal to f convolution
of p y x y is equal to f convolution of p y of x.
Now what we need to find what this p y is it looks quite familiar to us when we have
dealt with the Poisson kernel as a good kernel in the series case. And in the next lecture
we are going to see what is the form of this Poisson of this p y such that p y hat of xi
is e to the power minus 2 pi mod xi by y. Thank you.
