Professional Documents
Culture Documents
and Solutions
Igor Yanovsky
1
Partial Dierential Equations Igor Yanovsky, 2005 2
Contents
1 Trigonometric Identities 6
3 Separation of Variables:
Quick Guide 9
5 First-Order Equations 10
5.1 Quasilinear Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.2 Weak Solutions for Quasilinear Equations . . . . . . . . . . . . . . . . . 12
5.2.1 Conservation Laws and Jump Conditions . . . . . . . . . . . . . 12
5.2.2 Fans and Rarefaction Waves . . . . . . . . . . . . . . . . . . . . . 12
5.3 General Nonlinear Equations . . . . . . . . . . . . . . . . . . . . . . . . 13
5.3.1 Two Spatial Dimensions . . . . . . . . . . . . . . . . . . . . . . . 13
5.3.2 Three Spatial Dimensions . . . . . . . . . . . . . . . . . . . . . . 13
6 Second-Order Equations 14
6.1 Classication by Characteristics . . . . . . . . . . . . . . . . . . . . . . . 14
6.2 Canonical Forms and General Solutions . . . . . . . . . . . . . . . . . . 14
6.3 Well-Posedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
7 Wave Equation 23
7.1 The Initial Value Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 23
7.2 Weak Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
7.3 Initial/Boundary Value Problem . . . . . . . . . . . . . . . . . . . . . . 24
7.4 Duhamels Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
7.5 The Nonhomogeneous Equation . . . . . . . . . . . . . . . . . . . . . . . 24
7.6 Higher Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
7.6.1 Spherical Means . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
7.6.2 Application to the Cauchy Problem . . . . . . . . . . . . . . . . 26
7.6.3 Three-Dimensional Wave Equation . . . . . . . . . . . . . . . . . 27
7.6.4 Two-Dimensional Wave Equation . . . . . . . . . . . . . . . . . . 28
7.6.5 Huygens Principle . . . . . . . . . . . . . . . . . . . . . . . . . . 28
7.7 Energy Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7.8 Contraction Mapping Principle . . . . . . . . . . . . . . . . . . . . . . . 30
8 Laplace Equation 31
8.1 Greens Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
8.2 Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
8.3 Polar Laplacian in R2 for Radial Functions . . . . . . . . . . . . . . . . 32
8.4 Spherical Laplacian in R3 and Rn for Radial Functions . . . . . . . . . . 32
8.5 Cylindrical Laplacian in R3 for Radial Functions . . . . . . . . . . . . . 33
8.6 Mean Value Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
8.7 Maximum Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
8.8 The Fundamental Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 34
8.9 Representation Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
8.10 Greens Function and the Poisson Kernel . . . . . . . . . . . . . . . . . . 42
Partial Dierential Equations Igor Yanovsky, 2005 4
9 Heat Equation 45
9.1 The Pure Initial Value Problem . . . . . . . . . . . . . . . . . . . . . . . 45
9.1.1 Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . 45
9.1.2 Multi-Index Notation . . . . . . . . . . . . . . . . . . . . . . . . 45
9.1.3 Solution of the Pure Initial Value Problem . . . . . . . . . . . . . 49
9.1.4 Nonhomogeneous Equation . . . . . . . . . . . . . . . . . . . . . 50
9.1.5 Nonhomogeneous Equation with Nonhomogeneous Initial Condi-
tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
9.1.6 The Fundamental Solution . . . . . . . . . . . . . . . . . . . . . 50
10 Schr
odinger Equation 52
12 Problems: Shocks 75
1 Trigonometric Identities
L
nx mx 0 n = m
cos cos dx =
cos(a + b) = cos a cos b sin a sin b L L L L n=m
L
cos(a b) = cos a cos b + sin a sin b nx mx 0 n = m
sin(a + b) = sin a cos b + cos a sin b sin sin dx =
L L L L n=m
sin(a b) = sin a cos b cos a sin b L
nx mx
sin cos dx = 0
L L L
cos(a + b) + cos(a b)
cos a cos b =
2
sin(a + b) + sin(a b)
sin a cos b =
L
nx mx 0 n = m
2 cos cos dx = L
0 L L 2 n=m
cos(a b) cos(a + b)
sin a sin b = L
0 n = m
2 nx mx
sin sin dx = L
0 L L 2 n=m
cos 2t = cos2 t sin2 t
sin 2t = 2 sin t cos t
L
0 n = m
1 1 + cos t einx eimx dx =
cos2 t = 0 L n=m
2 2
1 1 cos t
sin2 t =
L
0 n = 0
2 2 einx dx =
0 L n=0
1 + tan2 t = sec2 t
x sin x cos x
2
cot t + 1 = csc t 2 sin2 x dx =
2 2
x sin x cos x
eix + eix cos2 x dx = +
cos x = 2 2
2
eix eix tan2 x dx = tan x x
sin x =
2i
cos2 x
sin x cos x dx =
2
ex + ex
cosh x =
2
ex ex ln(xy) = ln(x) + ln(y)
sinh x = x
2 ln = ln(x) ln(y)
y
d ln xr = r lnx
cosh x = sinh(x)
dx
d
sinh x = cosh(x)
dx ln x dx = x ln x x
cosh2 x sinh2 x = 1 x2 x2
x ln x dx = ln x
2 4
du 1 u
= tan1 + C
a2
+ u2 a a 2
ez dz =
du u
= sin1 + C R
a2 u2 a z2
e 2 dz = 2
R
Partial Dierential Equations Igor Yanovsky, 2005 7
a b 1 1 d b
A= , A =
c d det(A) c a
Partial Dierential Equations Igor Yanovsky, 2005 8
X X = 0
Boundary conditions Eigenvalues n Eigenfunctions Xn
n 4
X(0) = X(L) = 0, X (0) = X (L) = 0 sin n
L x n = 1, 2, . . .
n
L
4
X (0) = X (L) = 0, X (0) = X (L) = 0 L cos n
L x n = 0, 1, 2, . . .
Partial Dierential Equations Igor Yanovsky, 2005 9
X (t) Y ()
= = .
X(t) Y ()
Y () + Y () = 0.
uxx + uyy + k2 u = 0.
Wave Equation: utt uxx = 0. X Y
= + k 2 = c2 .
X (x) T (t) X Y
= = .
X(x) T (t) X + c2 X = 0,
X + X = 0. Y + (k2 c2 )Y = 0.
utt + 3ut + u = uxx .
T T X
+3 +1 = = .
T T X uxx + uyy + k2 u = 0.
X + X = 0.
Y X
utt uxx + u = 0. = + k 2 = c2 .
Y X
T X Y + c2 Y = 0,
+1 = = .
T X X + (k2 c2 )X = 0.
X + X = 0.
utt + ut = c2 uxx + uxxt, ( > 0)
X
= ,
X
1 T T
T X
+ = 1 + .
c2 T c2 T c2 T X
4th Order: utt = k uxxxx.
X 1 T
= = .
X k T
X X = 0.
5 First-Order Equations
5.1 Quasilinear Equations
Consider the Cauchy problem for the quasilinear equation in two variables
f (s) b(f (s), g(s), h(s)) g (s) a(f (s), g(s), h(s)) = 0.
x = h(s)t + s, y = t, z = h(s).
x = h(s)t + s
along which u has the constant value u = h(s). Two characteristics x = h(s1 )t + s1
and x = h(s2 )t + s2 intersect at a point (x, t) with
s2 s1
t= .
h(s2 ) h(s1 )
1
y and t are interchanged here
Partial Dierential Equations Igor Yanovsky, 2005 11
h (s)
ux = h (s)(1 ux t) ux =
1 + h (s)t
The smallest t for which this happens corresponds to the value s = s0 at which h (s)
has a minimum (i.e.h (s) has a maximum). At time T = 1/h (s0 ) the solution u
experiences a gradient catastrophe.
Partial Dierential Equations Igor Yanovsky, 2005 12
ut + f (u)x = 0, (5.3)
This is an example of a conservation law. Notice that (5.4) implies (5.3) if u is C 1 , but
(5.4) makes sense for more general u.
Consider a solution of (5.4) that, for xed t, has a jump discontinuity at x = (t).
We assume that u, ux , and ut are continuous up to . Also, we assume that (t) is C 1
in t.
Taking a < (t) < b in (5.4), we obtain
b
d
u dx + u dx + f (u(b, t)) f (u(a, t))
dt a
b
= (t)ul ((t), t) (t)ur ((t), t) + ut (x, t) dx + ut(x, t) dx
a
+ f (u(b, t)) f (u(a, t)) = 0,
where ul and ur denote the limiting values of u from the left and right sides of the shock.
Letting a (t) and b (t), we get the Rankine-Hugoniot jump condition:
f (ur ) f (ul )
(t) = .
ur ul
We need to complete to a strip. Find (s) and (s), the initial conditions for p(s, t)
and q(s, t), respectively:
F (f (s), g(s), h(s), (s), (s)) = 0
h (s) = (s)f (s) + (s)g (s)
The characteristic equations are
dx dy
= Fp = Fq
dt dt
dz
= pFp + qFq
dt
dp dq
= Fx Fz p = Fy Fz q
dt dt
We need to have the Jacobian condition. That is, in order to solve the Cauchy problem
in a neighborhood of , the following condition must be satised:
f (s) Fq [f, g, h, , ](s) g (s) Fp [f, g, h, , ](s) = 0.
We need to complete to a strip. Find 1 (s1 , s2 ), 2 (s1 , s2 ), and 3 (s1 , s2 ), the initial
conditions for p1 (s1 , s2 , t), p2 (s1 , s2 , t), and p3 (s1 , s2 , t), respectively:
F f1 (s1 , s2 ), f2 (s1 , s2 ), f3 (s1 , s2 ), h(s1 , s2 ), 1 , 2 , 3 = 0
h f1 f2 f3
= 1 + 2 + 3
s1 s1 s1 s1
h f1 f2 f3
= 1 + 2 + 3
s2 s2 s2 s2
The characteristic equations are
dx1 dx2 dx3
= Fp1 = Fp2 = Fp3
dt dt dt
dz
= p1 Fp1 + p2 Fp2 + p3 Fp3
dt
dp1 dp2 dp3
= Fx1 p1 Fz = Fx2 p2 Fz = Fx3 p3 Fz
dt dt dt
Partial Dierential Equations Igor Yanovsky, 2005 14
6 Second-Order Equations
6.1 Classification by Characteristics
Consider the second-order equation in which the derivatives of second-order all occur
linearly, with coecients only depending on the independent variables:
to nd y = 3x + c1 and y = 2x + c2 .
Partial Dierential Equations Igor Yanovsky, 2005 15
Problem (F03, #4). Find the characteristics of the partial dierential equation
and then show that it can be transformed into the canonical form
( 2 + 4)u + u = 0
whence and are suitably chosen canonical coordinates. Use this to obtain the general
solution in the form
g( ) d
u(, ) = f () + 1
( 2 + 4 ) 2
where f and g are arbitrary functions of and .
6.3 Well-Posedness
Problem (S99, #2). In R2 consider the unit square dened by 0 x, y 1.
Consider
a) ux + uyy = 0;
b) uxx + uyy = 0;
c) uxx uyy = 0.
Prescribe data for each problem separately on the boundary of so that each of these
problems is well-posed. Justify your answers.
Proof. The initial / boundary value problem for the HEAT EQUATION is well-
posed:
ut = u x , t > 0,
u(x, 0) = g(x) x ,
u(x, t) = 0 x , t > 0.
Existence - by eigenfunction expansion.
Uniqueness and continuous dependence on the data -
by maximum principle.
The method of eigenfunction expansion and maximum
principle give well-posedness for more general problems:
ut = u + f (x, t)
x , t > 0,
u(x, 0) = g(x) x ,
u(x, t) = h(x, t) x , t > 0.
uy + auxx + buyy = 0
Proof.
ab < 0 two sets of characteristics hyperbolic.
Relabeling the variables (y t), we have
a 1
utt + uxx = ut .
b b
The solution of the
a equation is ofthe form
u(x, t) = F (x + b t) + G(x ab t).
Existence of the solution to the initial/boundary
value problem is given by the method of separation
of variables (expansion in eigenfunctions)
and by the parallelogram rule.
Uniqueness is given by the energy method.
Need initial conditions u(x, 0), ut (x, 0).
Prescribe u or ux for each of the two boundaries.
7 Wave Equation
The one-dimensional wave equation is
u = 0. (7.2)
The general solution of (7.2) is u(, ) = F ()+G(), where F and G are C 1 functions.
Returning to the variables x, t we nd that
Using these expressions in (7.3), we obtain dAlemberts Formula for the solution
of the initial value problem (7.4):
x+ct
1 1
u(x, t) = (g(x + ct) + g(x ct)) + h() d.
2 2c xct
u(A) + u(C) =
= F (k1 ) + G(k3) + F (k2 ) + G(k4 )
= u(B) + u(D),
0
ut (x, 0) = 0. Ut (x, 0, s) = f (x, s)
an + n an = fn (t) an + n an = 0 t
an (0) = 0 an (0, s) = 0 an (t) = an (t s, s) ds.
0
an (0) = 0 an (0, s) = fn (s)
by adding together dAlemberts formula and Duhamels principle gives the solution:
x+ct t x+c(ts)
1 1 1
u(x, t) = (g(x + ct) + g(x ct)) + h() d + f (, s) d ds.
2 2c xct 2c 0 xc(ts)
Partial Dierential Equations Igor Yanovsky, 2005 26
Note that for a radial function u = u(r), we have Mu = u, so the equation provides the
Laplacian of u in spherical coordinates.
Suppose that u(x, t) solves (7.9). We can view t as a parameter and take the spherical
mean to obtain Mu (x, r, t), which satises
2 1 1
Mu (x, r, t) = utt (x + r, t)dS = c2 u(x + r, t)dS = c2 Mu (x, r, t).
t2 n ||=1 n ||=1
Invoking the Darboux equation, we obtain the Euler-Poisson-Darboux equation:
2 2 n1
Mu (x, r, t) = c2 + Mu (x, r, t).
t2 r 2 r r
2
2
2
rMu (x, r, t) = c rMu (x, r, t) .
t2 r 2
For each xed x, consider V x (r, t) = rMu (x, r, t) as a solution of the one-dimensional
wave equation in r, t > 0:
2 x 2
2
V (r, t) = c V x (r, t),
t2 r 2
V x (r, 0) = rMg (x, r) Gx(r), (IC)
Vtx (r, 0) = rMh (x, r) H (r),x
(IC)
V (0, t) = lim rMu (x, r, t) = 0 u(x, t) = 0.
x
(BC)
r0
Gx (0) = H (0) = 0.
x
We may extend Gx and H x as odd functions of r and use dAlemberts formula for
V x (r, t):
1 x 1 r+ct x
V x (r, t) = G (r + ct) + Gx (r ct) + H () d.
2 2c rct
Since Gx and H x are odd functions, we have for r < ct:
r+ct ct+r
G (r ct) = G (ct r)
x x
and x
H () d = H x () d.
rct ctr
After some more manipulations, we nd that the solution of (7.9) is given by the
Kirchhos formula:
1 t
u(x, t) = t g(x + ct)dS + h(x + ct)dS .
4 t ||=1 4 ||=1
3
For x Rn :
1
f (x + t)dS = n1 f (x + y)dy
t ||=1 t |y|t
f (x + y)dy = tn1 f (x + t)dS
t |y|t ||=1
Partial Dierential Equations Igor Yanovsky, 2005 29
d(T x, T y) d(x, y) x, y X.
8 Laplace Equation
Consider the Laplace equation
u = 0 in Rn (8.1)
u = f in Rn . (8.2)
ux vx dxdy = vux n1 ds vuxx dxdy n = (n1 , n2 ) R2
uxk v dx = uvnk ds uvxk dx n = (n1 , . . ., nn ) Rn .
v u
u 2 v dx = u ds v ds + uv dx.
n n
2 2
v u v u
u v v u dx = u v ds + u v ds.
n n n n
Partial Dierential Equations Igor Yanovsky, 2005 32
1 2 1
u = rur r = + u.
r r 2 r r
2 n1
u = + u.
r 2 r r
In R3 : 4
1 1 2 2
u = 2 r 2 ur r = ru rr = + u.
r r r 2 r r
4
These formulas are taken from S. Farlow, p. 411.
Partial Dierential Equations Igor Yanovsky, 2005 33
1 2 1
u = rur r = + u.
r r 2 r r
for all .
We can derive, (8.6) for any given n. For intance, when n = 3, we have:
2
K + K = 0.
r
Let
1
K = w(r),
r
1 1
K = w 2 w,
r r
1 2 2
K = w 2 w + 3 w.
r r r
Plugging these into , we obtain:
1
w = 0, or
r
w = 0.
Thus,
w = c1 r + c2 ,
1 c2
K = w(r) = c1 + .
r r
See the similar problem, F99, #2, where the fundamental solution for ( I) is
found in the process.
Partial Dierential Equations Igor Yanovsky, 2005 35
Suppose v(x) 0 for |x| R and let = BR (0); for small > 0 let
= B (0).
K(|x|) is harmonic (K(|x|) = 0) in . Consider Greens identity ( =
B (0)):
v K(|x|)
v K(|x|)
K(|x|)v dx = K(|x|) v dS + K(|x|) v dS.
n n B(0) n n
=0, since v0 f or xR
c2
lim K(|x|)v dx = K(|x|)v dx. Since K(r) = c1 + is integrable at x = 0.
0 r
On B (0), K(|x|) = K(). Thus, 5
v
v
c2
K(|x|) dS = K() dS c 1 + 42 max v 0, as 0.
B(0) n B(0) n
K(|x|) c2
v(x) dS = 2
v(x) dS
B(0) n B(0)
c2 c2
= 2
v(0) dS + 2
[v(x) v(0)] dS
B(0) B(0)
c2
= 2 v(0) 42 + 4c2 max v(x) v(0)
xB(0)
0, (v is continuous)
= 4c2 v(0) v(0).
1
Thus, taking 4c2 = 1, i.e. c2 = 4 , we obtain
K(|x|)v dx = lim K(|x|)v dx = v(0),
0
1
that is K(r) = 4r is the fundamental solution of .
5
In R3 , for |x| = ,
c2
K(|x|) = K() = c1 + .
K(|x|) K() c2
= = 2, (since n points inwards.)
n r
n points toward 0 on the sphere |x| = (i.e., n = x/|x|).
Partial Dierential Equations Igor Yanovsky, 2005 36
Show that the Fundamental Solution of the Laplace Operator is given by.
1
2 log r if n = 2
K(x) = 1 2n
(8.7)
(2n)n r if n 3.
lim K(|x|)v dx = K(|x|)v dx. Since K(r) is integrable at x = 0.
0
6
Note that for |x| = ,
1
2
log if n = 2
K(|x|) = K() = 1 2n
(2n)n if n 3.
1
K(|x|) K() 2
if n = 2 1
= = 1 = , (since n points inwards.)
n r n n1
if n 3, n n1
7
The last integral tends to 0 as 0:
1 u(y)
1 1 u(y) 1
[u(y) u(x)] dS u(y) u(x) + dS
2 2
B(x) n B(x) B (x) n
4 max u(y) u(x) + 4 max u(y) .
yB(x) y
0, (u continuous in ) 0, (|u| is f inite)
7
Note that for points y on B (x),
1 1 1 1
= and = 2.
|x y| n |x y|
Partial Dierential Equations Igor Yanovsky, 2005 39
Representation Theorem, n = 2.
Let be bounded domain in R2 and let n be the unit exterior normal to . Let
u C 2 (). Then the value of u at any point x is given by the formula
1 1 u(y)
u(x) = u(y) log |x y| dy + u(y) log |x y| log |x y| dS.
(8.11)
2 2 n n
Proof. Consider the Greens identity:
w u
(uw wu) dy = u w dS,
n n
where w is the harmonic function
w(y) = log |x y|,
which is singular at x . In order to be able to apply Greens identity, we consider
a new domain :
= B (x).
Since u, w C2 ( ), Greens identity can be applied. Since w is harmonic (w = 0)
in and since = B (x), we have
u(y) log |x y| dy (8.12)
u(y)
= u(y) log |x y| log |x y| dS
n n
u(y)
+ u(y) log |x y| log |x y| dS.
B (x) n n
We will show that formula (8.11) is obtained by letting 0.
lim u(y) log |x y| dy = u(y) log |x y| dy. since log |x y| is integrable at x = y.
0
The rst integral on the right of (8.12) does not depend on . Hence, the limit as 0
of the second integral on the right of (8.12) exists, and in order to obtain (8.11), need
u(y)
lim u(y) log |x y| log |x y| dS = 2u(x).
0 B(x) n n
u(y) 1 u(y)
u(y) log |x y| log |x y| dS = u(y) log dS
B (x) n n B(x) n
1 1 u(y)
= u(x) dS + [u(y) u(x)] log dS
B(x) B(x) n
1 u(y)
= 2u(x) + [u(y) u(x)] log dS.
B(x) n
Partial Dierential Equations Igor Yanovsky, 2005 40
8
The last integral tends to 0 as 0:
u(y)
1 u(y) 1
[u(y) u(x)] log dS u(y) u(x) + log dS
B(x) n
B(x) B(x) n
2 max u(y) u(x) + 2 log max u(y) .
yB (x) y
0, (u continuous in ) 0, (|u| is f inite)
8
Note that for points y on B (x),
1
log |x y| = log and log |x y| = .
n
Partial Dierential Equations Igor Yanovsky, 2005 41
Representation Theorems, n > 3 can be obtained in the same way. We use the
Greens identity with
1
w(y) = ,
|x y|n2
which is a harmonic function in Rn with a singularity at x.
The rst integral on the right of (8.15) does not depend on . Hence, the limit as 0
of the second integral on the right of (8.15) exists, and in order to obtain (8.13), need
K(x y) u(y)
lim u(y) K(x y) dS = u(x).
0 B(x) n n
Partial Dierential Equations Igor Yanovsky, 2005 42
K(x y) u(y) K() u(y)
u(y) K(x y) dS = u(y) K() dS
B (x) n n B(x) n n
K() K() u(y)
= u(x) dS + [u(y) u(x)] K() dS
B(x) n B (x) n n
1 1 u(y)
= n1
u(x) dS n1
[u(y) u(x)] dS K() dS
n B(x) n B (x) B(x) n
1 1 u(y)
= n1
u(x)n n1
n1
[u(y) u(x)] dS K() dS.
n B (x) B(x) n
n
u(x)
9 The last two integrals tend to 0 as 0:
1 u(y)
[u(y) u(x)] dS K() dS
n1 n
n B(x) B(x)
1
max u(y) u(x)n n1 + K() max u(y)n n1 .
n n1 yB (x) y
0, (u continuous in ) 0, (|u| is f inite)
G(x, ) = 0 x .
9
Note that for points y on B (x),
1
2
log if n = 2
K(x y) = K() = 1
(2n)n
2n if n 3.
1
K(x y) K() 2
if n = 2 1
= = 1
= , (since n points inwards.)
n r n n1
if n 3, n n1
Partial Dierential Equations Igor Yanovsky, 2005 43
G is called the Greens function and is useful in satisfying Dirichlet boundary conditions.
The Greens function is dicult to construct for a general domain since it requires
solving the Dirichlet problem = 0 in , (x) = K(x ) for x , for each
.
From (8.18) we nd 10
G(x, )
u() = u(x) dS.
n
Thus if we know that the Dirichlet problem has a solution u C 2 (), then we can
calculate u from the Poisson integral formula (provided of course that we can compute
G(x, )).
10
If we did not assume u = 0 in our derivation, we would have (8.13) instead of (8.16), and an
extra term in (8.17), which would give us a more general expression:
G(x, )
u() = G(x, ) u dx + u(x) dS.
n
Partial Dierential Equations Igor Yanovsky, 2005 44
where is a bounded domain and is a (complex) number. The values of for which
(8.19) admits a nontrivial solution u are called the eigenvalues of in , and the
solution u is an eigenfunction associated to the eigenvalue . (The convention
u + u = 0 is chosen so that all eigenvalues will be positive.)
9 Heat Equation
The heat equation is
ut = ku for x , t > 0, (9.1)
with initial and boundary conditions.
= (i) u
().
(i) = (i1 )1 (in )n .
Partial Dierential Equations Igor Yanovsky, 2005 46
2
|u(x)| dx = u()|2 d.
|
Also,
u(x) v(x) dx = ()
u v() d.
The properties (9.2) and (9.3) make it very natural to consider the fourier transform
on a subspace of L1 (Rn ) called the Schwartz class of functions, S, which consists of the
smooth functions whose derivatives of all orders decay faster than any polynomial, i.e.
u) = u; that is,
Fourier Inversion Theorem (McOwen). If u S, then (
1 1
u(x) = n
ix
e () d =
u u) (x).
ei(xy) u(y) dy d = (
(2) 2 Rn (2)n R2n
u(x a)() = eia u
().
Delta function:
1 1
(x)() = eix (x) dx = , since u(x) = (x y) u(y) dy .
2
R 2 R
1
(x a)() = eia () = eia . (using result from Shift)
2
Convolution:
(f g)(x) = f (x y)g(y) dy,
Rn
1 ix 1
(f g)() = n e f (x y) g(y) dy dx = n eix f (x y) g(y) dy dx
(2) 2 Rn Rn (2) 2 Rn Rn
1 i(xy) iy
= n e f (x y) dx e g(y) dy
(2) 2 Rn Rn
1 n
= n eiz
f (z) dz eiy g(y) dy = (2) 2 f()g().
(2) 2 Rn Rn
(f
n
g)() = (2) 2 f()
g().
x2 2
e 2 () = e 2 .
Multiplication by x:
1 d
ixu() = eix ixu(x) dx = ().
u
2 R d
d
xu(x)() = i u ().
d
Partial Dierential Equations Igor Yanovsky, 2005 48
1 ix
1 ix 1
xux (x)() = e xux (x) dx = e xu (i)eix x + eix u dx
2 R 2 2 R
=0
1 1
= i eix x u dx eix u dx
2 R 2 R
d d
= i xu(x)() u () = i i u () u () = u () u
().
d d
d
xux (x)() = () u
u ().
d
ax2 1 2
e 2 () = e 2a , (Gaussian)
a
1
b
eibx f (ax)() = f ,
a
a
1, |x| L 1 2 sin(L)
f (x) = f
(x)() = ,
0, |x| > L, 2
1 2a
e
a|x| () = , (a > 0)
2 a + 2
2
1 2 a||
() = e , (a > 0)
a + x2
2 2a
2 1
H(a |x|)() = sin a,
1
1
H(x)() = () + ,
2 i
2 1
H(x)
H(x) () = , (sign)
i
1() = 2().
11
Results with marked with were taken from W. Strauss, where the denition of Fourier Transform
is dierent. An extra multiple of 12 was added to each of these results.
Partial Dierential Equations Igor Yanovsky, 2005 49
Uniqueness of solutions for the pure initial value problem fails: there are nontrivial
solutions of (9.4) with g = 0. 13 Thus, the pure initial value problem for the heat
equation is not well-posed, as it was for the wave equation. However, the nontrivial
solutions are unbounded as functions of x when t > 0 is xed; uniqueness can be
regained by adding a boundedness condition on the solution.
12
Identity (Evans, p. 187.) :
|x|2
2
n
2
eix|| t d = e 4t .
Rn t
13
The following function u satises ut = uxx for t > 0 with u(x, 0) = 0:
1 dk 2
u(x, t) = x2k k e1/t .
(2k)! dt
k=0
Partial Dierential Equations Igor Yanovsky, 2005 50
is given by
t
u(x, t) = y, t) g(y) dy +
K(x y, t s) f (y, s) dy ds.
K(x
Rn 0 Rn
We call K a fundamental solution for the initial value problem. The solution of
(9.7) is then given by convolution in the space variables:
u(x, t) = K(x y, t) g(y) dy.
Rn
Partial Dierential Equations Igor Yanovsky, 2005 51
For operators of the form t L, the fundamental solution of the initial value problem,
K(x, t) as dened in (9.8), coincides with the free space fundamental solution, which
satises
t L K(x, t) = (x, t),
provided we extend K(x, t) by zero to t < 0. For the heat equation, consider
|x|2
1 4t
(4t) n/2 e t>0
K(x, t) = (9.9)
0 t 0.
14
Note, for the operator L = /t, the adjoint operator is L = /t.
Partial Dierential Equations Igor Yanovsky, 2005 52
10 Schr
odinger Equation
Problem (F96, #5). The Gauss kernel
1 (xy)2
G(t, x, y) = 1 e
4t
(4t) 2
is the fundamental solution of the heat equation, solving
Gt = Gxx, G(0, x, y) = (x y).
By analogy with the heat equation, nd the fundamental solution H(t, x, y) of the
Schr
odinger equation
Ht = iHxx, H(0, x, y) = (x y).
Show that your expression H(x) is indeed the fundamental solution for the
Schr
odinger equation. You may use the following special integral
ix2
e 4 dx = i4.
Proof. Remark: Consider the initial value problem for the Schr
odinger equation
ut = iu x Rn , t > 0,
u(x, 0) = g(x) x Rn .
If we formally replace t by it in the heat kernel, we obtain the Fundamental
Solution of the Schr odinger Equation: 15
1 |x|2
4it
H(x, t) = n e (x Rn , t = 0)
(4it) 2
1 |xy|2
u(x, t) = n e 4it g(y) dy.
(4it) 2 Rn
In particular, the Schr
odinger equation is reversible in time, whereas the heat equation
is not.
Solution: We have already found the fundamental solution for the heat equation
using the Fourier transform. For the Schr odinger equation is one dimension, we have
(, t) = i 2 u
u (, t),
t
2
(, t) = Cei t .
which is an ordinary dierential equation in t, with the solution u
(, 0) =
The initial condition u g() gives
2
(, t) = g() ei t ,
u
2
1 i 2 t
u(x, t) = g() ei t
= g e
2
1
1 2
= g ei t eix d
2 2 R
1 2
= g eixi t d = (need some work) =
2 R
|x|2
1 1 |xy|2
= g e 4it = e 4it g(y) dy.
4it 4it R
15
Evans, p. 188, Example 3.
Partial Dierential Equations Igor Yanovsky, 2005 53
16
Note, for the operator L = /t, the adjoint operator is L = /t.
Partial Dierential Equations Igor Yanovsky, 2005 54
x2 ux + xyuy = u2
which passes through the curve u = 1, x = y 2 . Determine where this solution becomes
singular.
fx + x2 yfy + f = 0
f (x = 0, y) = y 2
and solve it by the method of characteristics. Discuss the properties of the solution; in
particular investigate the behavior of |ux(, t)| for t .
ut + u2 ux = 0, t > 0,
u(0, x) = 2 + x.
Proof. Solved
Partial Dierential Equations Igor Yanovsky, 2005 56
ut + uux = x, t0
u(x, 0) = f (x), < x < .
Thus,
x = 16 f 2 (s)e6y + 16 f 2 (s) + s, 1 z 2 6y 1 z 2 z2 z2
x= e + + s = + s,
f (s) = ez3y 6 e6y 6 e6y 6e6y 6
z2 z2
s = x 6y + .
6e 6
z 2
z 2 3y
z = f x 6y + e .
6e 6
u2 u2 3y
u(x, y) = f x 6y + e .
6e 6
Partial Dierential Equations Igor Yanovsky, 2005 58
where a > 0 is constant. Solve until the rst appearance of discontinuous derivative
and determine that critical time.
b) Consider the equation
u3
ut + = cu. (11.2)
3 x
How large does the constant c > 0 has to be, so that a smooth solution (with no discon-
tinuities) exists for all t > 0? Explain.
Proof. a) Characteristic form: ut + u2 ux = 0. : (s, 0, h(s)).
dx dy dz
= z2, = 1, = 0.
dt dt dt
x = h(s)2 t + s, y = t, z = h(s).
u(x, y) = h(x u2 y) (11.3)
The characteristic projection in the xt-plane17 passing through the point (s, 0) is the
line
x = h(s)2 t + s
2a2 ex (1 2ex ), x<0 x = ln( 12 ) = ln(2), x < 0
f (x) = 2 x x
=0
2a e (1 2e ), x > 0 x = ln(2), x>0
1 1
a2
f (ln( 12 )) = 2a2 eln( 2 ) (1 eln( 2 ) ) = 2a2 ( 12 )( 12 ) = 2 , x<0
2
f (ln(2)) = 2a2 ( 12 )(1 12 ) = a2 , x>0
1 2
t=
= 2
min{2h(s)h (s)} a
u3x
ut + =0 (11.4)
3
for t > 0, < x < with initial data
a(1 ex ), x<0
u(x, 0) = h(x) = x
a(1 e ), x > 0
dx 2 a2 e2s a2 e2s t + c4 (s) a2 e2s t + s
= Fp = p = x(s, t) = x=
dt a2 e2s a2 e2s t + c5 (s) a2 e2s t + s
dy
= Fq = 1 y(s, t) = t + c1 (s) y = t
dt 3 3s
dz 3 a3 e3s a 3e = 23 a3 e3s , x<0
= pFp + qFq = p + q = 3 3s a3 e3s 2 3 3s
dt a e + 3 = 3a e , x>0
2 3 3s 2 3 3s
3 a e t a(1 e ),
s
z(s, t) = 3 a e t + c6 (s), x<0
z =
x<0
23 a3 e3s t + c7 (s), x>0 23 a3 e3s t a(1 es ), x>0
dp aes , x<0
= Fx Fz p = 0 p(s, t) = c2 (s) p=
dt aes , x>0
Partial Dierential Equations Igor Yanovsky, 2005 61
3 e3s
dq a 3 , x<0
= Fy Fz q = 0 q(s, t) = c3 (s) q= a3 e3s
dt 3 , x>0
Thus,
2 3 3s
3 a e y a(1 e ),
s x<0
u(x, y) =
23 a3 e3s y a(1 es ), x>0
where s is dened as
a2 e2s y + s, x<0
x=
a2 e2s y + s, x > 0.
u3x
ut + = cu. (11.5)
3
Proof. Rewrite the equation as
u3x
F (x, y, u, ux, uy ) = + uy + cu = 0,
3
p3
F (x, y, z, p, q) = + q + cz = 0.
3
is parameterized by : (s, 0, h(s), (s), (s)).
We need to complete to a strip. Find (s) and (s), the initial conditions for p(s, t)
and q(s, t), respectively:
F (f (s), g(s), h(s), (s), (s)) = 0,
(s)3
+ (s) + ch(s) = 0,
3
(s)3
(s)3 3 + ca(1 ex ), x < 0
(s) = ch(s) = 3
(s) x
3 + ca(1 e ), x > 0
3
dx
= Fp = p2
dt
dy
= Fq = 1
dt
dz
= pFp + qFq = p3 + q
dt
dp
= Fx Fz p = cp
dt
dq
= Fy Fz q = cq
dt
We can proceed solving the characteristic equations with initial conditions above.
Partial Dierential Equations Igor Yanovsky, 2005 63
ut + u3 ux = 0,
a(1 ex ), for x < 0
u(x, 0) =
a(1 ex ), for x > 0
where a > 0 is a constant. Determine the rst time when a shock forms.
dx dy dz
= z3, = 1, = 0.
dt dt dt
x = h(s)3 t + s, y = t, z = h(s).
u(x, y) = h(x u3 y) (11.6)
The characteristic projection in the xt-plane18
passing through the point (s, 0) is the line
x = h(s)3 t + s
along which u has a constant value u = h(s).
Characteristics x = h(s1 )3 t + s1 and x = h(s2 )3 t + s2 intersect at a point (x, t) with
s2 s1
t= .
h(s2 )3 h(s1 )3
From (11.6), we have
h (s)
ux = h (s)(1 3u2 ux t) ux =
1 + 3h(s)2 h (s)t
Hence for 3h(s)2 h (s) < 0, ux becomes innite at the positive time
1
t= .
3h(s)2 h (s)
The smallest t for which this happens corresponds to the value s = s0 at which
h(s)2 h (s) has a minimum (i.e.h(s)2 h (s) has a maximum). At time T = 1/(3h(s0 )2 h (s0 ))
the solution u experiences a gradient catastrophe.
Therefore, need to nd a minimum of
3a2 (1 ex )2 aex = 3a3 ex (1 ex )2 , x<0
f (x) = 3h(x)2 h (x) = 2 x 2 x 3 x x 2
3a (1 e ) ae = 3a e (1 e ) , x > 0
3a3 ex (1 ex )2 ex 2(1 ex )ex = 3a3 ex (1 ex )(1 3ex ), x<0
f (x) = 3
x x 2 x x x
3 x x x
=0
3a e (1 e ) + e 2(1 e )e = 3a e (1 e )(1 + 3e ), x > 0
x = 0, x = ln 3, x < 0,
The zeros of f (x) are We check which ones give the minimum of f (x) :
x = 0, x = ln 3, x > 0.
3
f (0) = 3a3 , f ( ln 3) = 3a3 13 (1 13 )2 = 4a9 , x<0
3
f (0) = 3a3 , f (ln 3) = 3a3 13 (1 13 )2 = 4a9 , x>0
18
y and t are interchanged here
Partial Dierential Equations Igor Yanovsky, 2005 64
1 1 1
t= 2
= = 3.
min{3h(s) h (s)} min f (s) 3a
Partial Dierential Equations Igor Yanovsky, 2005 65
b) Now consider
ut + u3 ux + cu = 0
with the same initial data and a positive constant c. How large does c need to be in
order to prevent shock formation?
uy + a(x)ux = 0,
u(x, 0) = h(x).
Give an example of an (unbounded) smooth a(x) for which the solution of the Cauchy
problem is not unique.
b) Find a class of initial data such that this problem has a global solution for all t.
Compute the critical time for the existence of a smooth solution for initial data, f ,
which is not in the above class.
Problem (F96, #6). Find an implicit formula for the solution u of the initial-value
problem
x
u(x, y) = ; solution is smooth for all positive time y.
y+1
b) is parameterized by : (s, 0, s).
dx x x
= z = s x = st + s s = = .
dt 1t 1y
dy
= 1 y = t,
dt
dz
= 0 z = s.
dt
x
u(x, y) = ; solution blows up at time y = 1.
y1
Partial Dierential Equations Igor Yanovsky, 2005 71
dx s+1
= (x + 1)2 x= 1,
dt (s + 1)t 1
dy
= 1 y = t,
dt
dz s+1
= x= 1,
dt (s + 1)t 1
z = ln|(s + 1)t 1| t + c1 (s),
z = ln|(s + 1)t 1| t + f (s).
In region I, characteristics are of the form
s+1
x= 1.
(s + 1)y 1
Thus, region I is bounded above by the line
1 x
x= 1, or y= .
y 1 x+1
xxyy
Since t = y, s = xy+y+1 ,
we have
x xy y
x xy y
u(x, y) = ln + 1 y 1 y + f ,
xy + y + 1 xy + y + 1
1
x xy y
u(x, y) = ln y+f .
xy + y + 1 xy + y + 1
For region II, is parameterized by : (0, s, g(s)).
dx 1
= (x + 1)2 x= 1,
dt t1
dy
= 1 y = t + s,
dt
dz 1
= x= 1,
dt t1
z = ln|t 1| t + c2 (s),
z = ln|t 1| t + g(s).
19
Variable t as a third coordinate of u and variable t used to parametrize characteristic equations
are two dierent entities.
Partial Dierential Equations Igor Yanovsky, 2005 72
Since t = x
s = y x+1
x+1 ,
x
, we have
x
x
x
u(x, y) = ln 1 +g y .
x+1 x+1 x+1
x
Note that on y = x+1 , both solutions are equal if f (0) = g(0).
Partial Dierential Equations Igor Yanovsky, 2005 73
20
Variable t as a third coordinate of u and variable t used to parametrize characteristic equations
are two dierent entities.
Partial Dierential Equations Igor Yanovsky, 2005 74
xux + (x + y)uy = 1
12 Problems: Shocks
Example 1. Determine the exact solution to Burgers equation
1
ut + u2 = 0, t>0
2 x
x = h(s)t + s.
x = h(s)t + s.
For Burgers equation, for a rarefaction fan emanating from (s, 0) on xt-plane, we have:
ul , t ul ,
xs
u(x, t) = xs
t , ul t ur ,
xs
t ur .
xs
ur ,
1, x < t 1,
x+1
t 1 < x < 1, i.e. 1 < x+1
t , t <0
u(x, t) = 0, 1 < x < 1,
x1 x1
t , 1 < x < t + 1, i.e. 0< t <1
1, x > t + 1.
Partial Dierential Equations Igor Yanovsky, 2005 78
Partial Dierential Equations Igor Yanovsky, 2005 79
x = h(s)t + s.
x = h(s)t + s.
x+1
u(x, t) = .
t+1
Another way to nd a solution on the left half-plane is to solve for s to nd
xt xt x+1
s= . Thus, u(x, t) = h(s) = 1 + s = 1 + = .
1+t 1+t t+1
F (ur ) F (ul ) 1 2
2 ur 12 u2l 0 12 ( x+1
t+1 )
2
1x+1
(t) = = = = ,
ur ul ur ul 0 t+1
x+1 2 t+1
dxs 1x+1
= ,
dt 2 t+1
x = c t + 1 1,
where u0 > 0.
x = h(s)t + s.
u0 t+1
1 u0 if x < 2 ,
For t > , u(x, t) = u0 t+1
u0 0 if x > 2 .
Proof. If u = f (x/t),
x x
x 1
ut = f 2 and ux = f .
t t t t
Hence, ut + a(u)ux = 0 implies that
x x
x
x 1
f 2 +a f f =0
t t t t t
or, assuming f is not identically 0 to rule out the constant solution, that
x x
a f = .
t t
This shows the functions a and f to be inverses of each other.
Partial Dierential Equations Igor Yanovsky, 2005 86
where h(x) is smooth function which vanishes for |x| large enough.
p2 x2
F (x, y, z, p, q) = +q+ = 0.
2 2
is parameterized by : (s, 0, h(s), (s), (s)).
We need to complete to a strip. Find (s) and (s), the initial conditions for p(s, t)
and q(s, t), respectively:
F (f (s), g(s), h(s), (s), (s)) = 0,
F (s, 0, h(s), (s), (s)) = 0,
(s)2 s2
+ (s) + = 0,
2 2
(s)2 + s2
(s) = .
2
ux + ux uy = 1,
u(x, 0) = f (x).
Proof. Solved.
In order to solve the Cauchy problem in a neighborhood of , need:
f (s) Fq [f, g, h, , ](s) g (s) Fp [f, g, h, , ](s) = 0,
1 h (s)
1 h (s) 0 1 + = 0,
h (s)
h (s) = 0.
Thus, h (s) = 0 ensures that the problem is noncharacteristic.
To show that one can solve y = (f (s))2 (x s) for (x, y) in a suciently small
neighborhood of (x0 , 0) with s(x0 , 0) = x0 , let
G(x, y, s) = (f (s))2 (x s) y = 0,
G(x0 , 0, x0) = 0,
Gr (x0 , 0, x0) = (f (s))2 .
Hence, if f (s) = 0, s, then Gs (x0 , 0, x0) = 0 and we can use the implicit function
theorem in a neighborhood of (x0 , 0, x0) to get
G(x, y, h(x, y)) = 0
and solve the equation in terms of x and y.
Partial Dierential Equations Igor Yanovsky, 2005 92
F (x, y, z, p, q) = p2 + q 2 1 = 0.
2
is parameterized by : (s, s2 , 0, (s), (s)).
We need to complete to a strip. Find (s) and (s), the initial conditions for p(s, t)
and q(s, t), respectively:
F (f (s), g(s), h(s), (s), (s)) = 0,
s2
F s, , 0, (s), (s) = 0,
2
(s) + (s)2 = 1.
2
1
Thus, s2 (s)2 + (s)2 = 1 (s)2 = .
s2 +1
ux + u2y + u2z = 1,
u(0, y, z) = y z.
We need to complete to a strip. Find 1 (s1 , s2 ), 2 (s1 , s2 ), and 3 (s1 , s2 ), the initial
conditions for p1 (s1 , s2 , t), p2 (s1 , s2 , t), and p3 (s1 , s2 , t), respectively:
F f1 (s1 , s2 ), f2 (s1 , s2 ), f3 (s1 , s2 ), h(s1 , s2 ), 1 , 2 , 3 = 0,
F 0, s1 , s2 , s1 s2 , 1 , 2 , 3 = 1 + 22 + 23 1 = 0,
1 + 22 + 23 = 1.
h f1 f2 f3
= 1 + 2 + 3 ,
s1 s1 s1 s1
s2 = 2 .
h f1 f2 f3
= 1 + 2 + 3 ,
s2 s2 s2 s2
s1 = 3 .
Thus, we have: 2 = s2 , 3 = s1 , 1 = s21 s22 + 1.
: 0
, s1 , s2 , s1 s2 , s21 s22 + 1, s2 , s1
x1 (s1 ,s2 ,0) x2(s1 ,s2 ,0) x3(s1 ,s2 ,0) z(s1 ,s2 ,0) p1 (s1 ,s2 ,0) p2(s1 ,s2 ,0) p3(s1 ,s2 ,0)
21
This problem is very similar to an already hand-written solved problem F95 #2.
Partial Dierential Equations Igor Yanovsky, 2005 94
x 1 = t
t = x1
t = x1
x = 2s t + s s = x 2s t s1 = x2 2x12x3
2 2 1 1 2 2 14x
1
2x1 x2
x3 = 2s1 t + s2
s2 = x3 2s1 t
s2 = x314x
2
z = (s2 + s2 + 1)t + s s z = (s2 + s2 + 1)t + s s 1
ux + uy + u3z = x + y + z,
u(x, y, 0) = xy.
Proof. Solved
Partial Dierential Equations Igor Yanovsky, 2005 95
Problem (S94, #1). Solve the following PDE for f (x, y, t):
ft + xfx + 3t2 fy = 0
f (x, y, 0) = x2 + y 2 .
Proof. Rewrite the equation as (x x1 , y x2 , t x3 , f u):
x1 ux1 + 3x23 ux2 + ux3 = 0,
u(x1 , x2, 0) = x21 + x22 .
We need to complete to a strip. Find 1 (s1 , s2 ), 2 (s1 , s2 ), and 3 (s1 , s2 ), the initial
conditions for p1 (s1 , s2 , t), p2 (s1 , s2 , t), and p3 (s1 , s2 , t), respectively:
F f1 (s1 , s2 ), f2 (s1 , s2 ), f3 (s1 , s2 ), h(s1 , s2 ), 1 , 2 , 3 = 0,
F s1 , s2 , 0, s21 + s22 , 1 , 2 , 3 = s1 1 + 3 = 0,
3 = s1 1 .
h f1 f2 f3
= 1 + 2 + 3 ,
s1 s1 s1 s1
2s1 = 1 .
h f1 f2 f3
= 1 + 2 + 3 ,
s2 s2 s2 s2
2s2 = 2 .
Thus, we have: 1 = 2s1 , 2 = 2s2 , 3 = 2s21 .
: s1 , s2 , 0 , s21 + s22 , 2s1 , 2s2 , 2s21
x1 (s1 ,s2 ,0) x2(s1 ,s2 ,0) x3(s1 ,s2 ,0) z(s1 ,s2 ,0) p1(s1 ,s2 ,0) p2 (s1 ,s2 ,0) p3 (s1 ,s2 ,0)
We need to complete to a strip. Find 1 (s1 , s2 ), 2 (s1 , s2 ), and 3 (s1 , s2 ), the initial
conditions for p1 (s1 , s2 , t), p2 (s1 , s2 , t), and p3 (s1 , s2 , t), respectively:
F f1 (s1 , s2 ), f2 (s1 , s2 ), f3 (s1 , s2 ), h(s1 , s2 ), 1 , 2 , 3 = 0,
F s1 , s2 , 0, s1s2 , 1 , 2 , 3 = s1 1 + s1 2 + 3 = 0,
3 = s1 (1 + 2 ).
h f1 f2 f3
= 1 + 2 + 3 ,
s1 s1 s1 s1
s2 = 1 .
h f1 f2 f3
= 1 + 2 + 3 ,
s2 s2 s2 s2
s1 = 2 .
Thus, we have: 1 = s2 , 2 = s1 , 3 = s21 s1 s2 .
: s1 , s2 , 0 , s1 s2 , s2 , s1 , s21 s1 s2
x1 (s1 ,s2 ,0) x2(s1 ,s2 ,0) x3(s1 ,s2 ,0) z(s1 ,s2 ,0) p1(s1 ,s2 ,0) p2 (s1 ,s2 ,0) p3 (s1 ,s2 ,0)
Partial Dierential Equations Igor Yanovsky, 2005 97
22
Variable t in the derivatives of characteristics equations and t in the solution f (x, y, t) are dierent
entities.
Partial Dierential Equations Igor Yanovsky, 2005 98
ut + ux + uy + u = 0 for t > 0
u(x, y, 0) = (x, y),
Since t = x3 , s1 = x1 x3 , s2 = x2 x3 , we have
u(x1 , x2 , x3 ) = (x1 x3 , x2 x3 )ex3 , or
u(x, y, t) = (x t, y t)et .
The solution satises the PDE and initial condition.24
23
Variable t as a third coordinate of u and variable t used to parametrize characteristic equations
are two dierent entities.
24
Chain Rule: u(x1 , x2 , x3 ) = (f (x1, x2 , x3 ), g(x1 , x2 , x3 )), then ux1 = f
f x1
+ g
g x1
.
Partial Dierential Equations Igor Yanovsky, 2005 99
Problem (F89, #4). Consider the rst order partial dierential equation
b) We need a compatibility condition between the initial and boundary values to hold
on y-axis (x = 0, t = 0):
u(x = 0, y, 0) = u(0, y, t = 0),
0 = 0.
Partial Dierential Equations Igor Yanovsky, 2005 101
Partial Dierential Equations Igor Yanovsky, 2005 102
For region I, (14.2) and (14.3) give two initial value problems (since any point in
region I can be traced back along both characteristics to initial conditions):
(1) (1) (2) (2)
vt 2vx = 0, vt + vx = 0,
v (1)(x, 0) = sin x; v (2)(x, 0) = 0.
which we solve by characteristics to get traveling wave solutions:
v (1)(x, t) = sin(x + 2t), v (2)(x, t) = 0.
Thus, for region I,
1 1 sin(x + 2t) sin(x + 2t)
U = V = = .
0 1 0 0
For region II, solutions of the form F (x + 2t) can be traced back to initial conditions.
Thus, v (1) is the same as in region I. Solutions of the form G(x t) can be traced back
to the boundary. Since from (14.4), u(2) = v (2), we use boundary conditions to get
u(2)(0, t) = t = G(t).
Hence, G(x t) = (x t).
Thus, for region II,
1 1 sin(x + 2t) sin(x + 2t) (x t)
U = V = = .
0 1 (x t) (x t)
Find an explicit solution for the following mixed problem for the system (14.5):
u(x, 0) f (x)
= for x > 0,
v(x, 0) 0
u(0, t) = 0 for t > 0.
You may assume that the function f is smooth and vanishes on a neighborhood of x = 0.
For region I, (14.6) and (14.7) give two initial value problems (since value at any
point in region I can be traced back along both characteristics to initial conditions):
(1) (1) (2) (2)
vt 3vx = 0, vt + 2vx = 0,
v (1)(x, 0) = 15 f (x); v (2)(x, 0) = 25 f (x).
which we solve by characteristics to get traveling wave solutions:
1 2
v (1)(x, t) = f (x + 3t), v (2)(x, t) = f (x 2t).
5 5
1 1
1 2 5 f (x + 3t) 5 f (x + 3t) + 45 f (x 2t)
Thus, for region I, U = V = = .
2 1 25 f (x 2t) 2 2
5 f (x + 3t) 5 f (x 2t)
For region II, solutions of the form F (x + 3t) can be traced back to initial conditions.
Thus, v (1) is the same as in region I. Solutions of the form G(x 2t) can be traced back
to the boundary. Since from (14.9),
u(1) = v (1) 2v (2), we have
1
u(1)(x, t) = F (x + 3t) 2G(x 2t) = f (x + 3t) 2G(x 2t).
5
The boundary condition gives
1
u(1)(0, t) = 0 = f (3t) 2G(2t),
5
1
2G(2t) = f (3t),
5
1
3
G(t) = f t ,
10 2
1
3
G(x 2t) = f (x 2t) .
10 2
1
1 1 3
1 2 5 f (x + 3t) 5 f (x + 3t) 5 f ( 2 (x 2t))
Thus, for region II, U = V = 1 3 = 2 1 3 .
2 1 10 f ( 2 (x 2t)) 5 f (x + 3t) + 10 f ( 2 (x 2t))
Problem (F94, #1; S97, #7). Solve the initial-boundary value problem
ut + 3vx = 0,
vt + ux + 2vx = 0
25
In S97, #7, the zero initial conditions are considered.
Partial Dierential Equations Igor Yanovsky, 2005 109
For region I, (14.11) and (14.12) give two initial value problems (since value at
any point in region I can be traced back along characteristics to initial conditions):
(1) (1)
(2) (2)
vt vx = 0, vt + 3vx = 0,
v (1)(x, 0) = 14 1 (x) + 14 2 (x); v (2)(x, 0) = 14 1 (x) + 34 2 (x),
which we solve by characteristics to get traveling wave solutions:
1 1 1 3
v (1)(x, t) = 1 (x + t) + 2 (x + t), v (2)(x, t) = 1 (x 3t) + 2 (x 3t).
4 4 4 4
Thus, for region I,
1
3 1 4 1 (x + t) + 14 2 (x + t)
U = V = 1 3
1 1 4 1 (x 3t) + 4 2 (x 3t)
1 31 (x + t) 32 (x + t) + 1 (x 3t) + 32 (x 3t)
= .
4 1 (x + t) + 2 (x + t) + 1 (x 3t) + 32 (x 3t)
For region II, solutions of the form F (x + t) can be traced back to initial conditions.
Thus, v (1) is the same as in region I. Solutions of the form G(x 3t) can be traced back
to the boundary. Since from (14.14),
u(1) = 3v (1) + v (2), we have
3 3
u(1)(x, t) = 1 (x + t) 2 (x + t) + G(x 3t).
4 4
The boundary condition gives
3 3
u(1)(0, t) = (t) = 1 (t) 2 (t) + G(3t),
4 4
3 3
G(3t) = (t) 1 (t) + 2 (t),
4 4
t 3
t 3
t
G(t) = 1 + 2 ,
3 4 3 4 3
x 3t 3
x 3t 3
x 3t
G(x 3t) = 1 + 2 .
3 4 3 4 3
Thus, for region II,
3 1 14 1 (x + t) + 14 2 (x + t)
U = V = 3 3
( x3t3 ) 4 1 ( 3 ) + 4 2 ( 3 )
x3t x3t
1 1
3 3 3 3
4 1 (x + t) 4 2 (x + t) + ( 3 ) 4 1 ( 3 ) + 4 2 ( 3 )
x3t x3t x3t
= .
14 1 (x + t) + 14 2 (x + t) + ( x3t 3 3
3 ) 4 1 ( 3 ) + 4 2 ( 3 )
x3t x3t
Partial Dierential Equations Igor Yanovsky, 2005 110
Problem (F91, #1). Solve explicitly the following initial-boundary value problem for
linear 22 hyperbolic system
ut = ux + vx
vt = 3ux vx ,
where b = 13 is a constant.
What happens when b = 13 ?
Proof. Let us change the notation (u u(1), v u(2)). Rewrite the equation as
1 1
Ut + Ux = 0, (14.15)
3 1
(1) !
(1)
u (x, 0) u0 (x)
U (x, 0) = = .
u(2)(x, 0) (2)
u0 (x)
The eigenvalues ofthe matrix A are 1 = 2, 2 = 2 and the corresponding eigen-
1 1
vectors are e1 = , e2 = . Thus,
1 3
2 0 1 1 1 3 1
= , = , 1 = .
0 2 1 3 4 1 1
Let U = V . Then,
Ut + AUx = 0,
Vt + AVx = 0,
Vt + 1 AVx = 0,
Vt + Vx = 0.
Thus, the transformed problem is
2 0
Vt + Vx = 0, (14.16)
0 2
!
(1) (2)
1 1 3 1 u(1)(x, 0) 1 3u0 (x) + u0 (x)
V (x, 0) = U (x, 0) = = .
4 1 1 u(2)(x, 0) 4 (1) (2)
u0 (x) u0 (x)
(14.17)
For region I, (14.16) and (14.17) give two initial value problems (since value at any
point in region I can be traced back along characteristics to initial conditions):
(1) (1)
(2) (2)
vt 2vx = 0, vt + 2vx = 0,
(1) (2) (1) (2)
v (1)(x, 0) = 34 u0 (x) + 14 u0 (x); v (2)(x, 0) = 14 u0 (x) 14 u0 (x),
which we solve by characteristics to get traveling wave solutions:
3 (1) 1 (2) 1 (1) 1 (2)
v (1)(x, t) = u0 (x + 2t) + u0 (x + 2t); v (2)(x, t) = u0 (x 2t) u0 (x 2t).
4 4 4 4
Thus, for region I,
!
3 (1) 1 (2)
1 1 u
4 0 (x + 2t) + u
4 0 (x + 2t)
U = V = 1 (1) 1 (2)
1 3
u
4 0 (x 2t) 4 u0 (x 2t)
!
3 (1) 1 (2) 1 (1) 1 (2)
u
4 0 (x + 2t) + u
4 0 (x + 2t) + u
4 0 (x 2t) u
4 0 (x 2t)
= 3 (1) 1 (2) 3 (1) 3 (2)
.
u
4 0 (x + 2t) + u
4 0 (x + 2t) u
4 0 (x 2t) + 4 u0 (x 2t)
For region II, solutions of the form F (x + 2t) can be traced back to initial conditions.
Thus, v (1) is the same as in region I. Solutions of the form G(x 2t) can be traced back
to the boundary. The boundary condition gives
u(1)(0, t) + bu(2)(0, t) = (t).
Using (14.19),
v (1)(0, t) + G(2t) + bv (1)(0, t) 3bG(2t) = (t),
(1 + b)v (1)(0, t) + (1 3b)G(2t) = (t),
3 1 (2)
(1)
(1 + b) u0 (2t) + u0 (2t) + (1 3b)G(2t) = (t),
4 4
(1) (2)
(t) (1 + b) 34 u0 (2t) + 14 u0 (2t)
G(2t) = ,
1 3b
(1) (2)
( 2t ) (1 + b) 34 u0 (t) + 14 u0 (t)
G(t) = ,
1 3b
3 (1) 1 (2)
( x2t
2 ) (1 + b) u
4 0 ((x 2t)) + u
4 0 ((x 2t))
G(x 2t) = .
1 3b
Thus, for region II,
3 (1) 1 (2)
1 1 u
4 0 (x + 2t)
(1) + u
4 0 (x + 2t)
U = V = (2)
1 3 ( x2t
2
)(1+b) 34 u0 ((x2t))+ 41 u0 ((x2t))
13b
(1) (2)
x2t
3 (1) 1 (2) ( 2 )(1+b) 34 u0 ((x2t))+ 14 u0 ((x2t))
u (x + 2t) + 4 u0 (x + 2t) +
= 4 0 (1)13b (2)
.
3 (1) 1 (2) 3( x2t )3(1+b) 34 u0 ((x2t))+ 14 u0 ((x2t))
4 u0 (x + 2t) + 4 u0 (x + 2t)
2
13b
The following were performed, but are arithmetically complicated:
Solutions for regions I and II satisfy (14.15).
Partial Dierential Equations Igor Yanovsky, 2005 113
ut = 3ux + 2vx
vt = vx v
in the region x 0, t 0. Which of the following sets of initial and boundary data
make this a well-posed problem?
a) u(x, 0) = 0, x 0
v(x, 0) = x2 , x 0
v(0, t) = t2 , t 0.
b) u(x, 0) = 0, x 0
v(x, 0) = x2 , x 0
u(0, t) = t, t 0.
c) u(x, 0) = 0, x 0
v(x, 0) = x2 , x 0
u(0, t) = t, t0
2
v(0, t) = t , t 0.
Proof. Rewrite the equation as Ut + AUx = BU . Initial conditions are same for
(a),(b),(c):
3 2 0 0
Ut + Ux = U,
0 1 0 1
(1)
u (x, 0) 0
U (x, 0) = = .
u(2)(x, 0) x2
The eigenvalues ofthe matrix A are 1 = 3, 2 = 1, and the corresponding eigen-
1 1
vectors are e1 = , e2 = . Thus,
0 2
3 0 1 1 1 1 2 1
= , = , = .
0 1 0 2 2 0 1
Let U = V . Then,
Ut + AUx = BU,
Vt + AVx = BV,
Vt + 1 AVx = 1 BV,
Vt + Vx = 1 BV.
Thus, the transformed problem is
3 0 0 1
Vt + Vx = V, (14.20)
0 1 0 1
1 1 2 1 0 x2 1
V (x, 0) = U (x, 0) = = . (14.21)
2 0 1 x2 2 1
Equation (14.20) gives traveling wave solutions of the form
For region I, (14.20) and (14.21) give two initial value problems (since a value at any
point in region I can be traced back along both characteristics to initial conditions):
(1) (1) (2) (2)
vt 3vx = v (2), vt + vx = v (2),
2 2
v (1)(x, 0) = x2 ; v (2)(x, 0) = x2 ,
which we do not solve here. Thus, initial conditions for v (1) and v (2) have to be dened.
Since (14.23) denes u(1) and u(2) in terms of v (1) and v (2), we need to dene two initial
conditions for U .
For region II, solutions of the form F (x + 3t) can be traced back to initial conditions.
Thus, v (1) is the same as in region I. Solutions of the form G(x t) are traced back to
the boundary at x = 0. Since from (14.23), u(2)(x, t) = 2v (2)(x, t) = 2G(x t), i.e.
u(2) is written in term of v (2) only, u(2) requires a boundary condition to be dened on
x = 0.
Thus,
a) u(2)(0, t) = t2 , t 0. Well-posed.
b) u(1)(0, t) = t, t 0. Not well-posed.
c) u(1)(0, t) = t, u(2) (0, t) = t2 , t 0. Not well-posed.
Partial Dierential Equations Igor Yanovsky, 2005 116
1 1 1 + 2 1 f (x) 1 (1 + 2)f (x) + g(x)
V (x, 0) = U (x, 0) = = .
2 2 1 + 2 1 g(x) 2 2 (1 + 2)f (x) g(x)
(14.25)
Equation (14.24) gives traveling wave solutions of the form:
v (1)(x, t) = F (x 2t), v (2)(x, t) = G(x + 2t). (14.26)
However, we can continue and obtain the solutions. We have two initial value problems
(1) (1) (2) (2)
vt + 2vx = 0, vt 2vx = 0,
(1+ 2) 1 (1+ 2) 1
v (1)(x, 0) = f (x) +
2 2 2 2
g(x); v (2)(x, 0) =
2 2
f (x)
2 2
g(x),
which we solve by characteristics to get traveling wave solutions:
(1) (1 + 2) 1
v (x, t) = f (x 2t) + g(x 2t),
2 2 2 2
(1 + 2) 1
v (2)(x, t) = f (x + 2t) g(x + 2t).
2 2 2 2
Partial Dierential Equations Igor Yanovsky, 2005 117
Thus,
a) No boundary conditions. Not well-posed.
b) u(1)(0, t) = h(x), u(2)(0, t) = k(x). Not well-posed.
c) u(1)(0, t) = h(x), u(2)(1, t) = k(x). Well-posed.
Partial Dierential Equations Igor Yanovsky, 2005 118
ft + gx = 0
gt + fx = 0
ht + 2hx = 0
a) f , g, h prescribed on t = 0, x 0; f, h prescribed on x = 0, t 0.
b) f , g, h prescribed on t = 0, x 0; f g, h prescribed on x = 0, t 0.
c) f + g, h prescribed on t = 0, x 0; f , g, h prescribed on x = 0, t 0.
For each of these 3 sets of data, determine whether or not the system is well-posed.
Justify your conclusions.
Proof. The third equation is decoupled from the rst two and can be considered sepa-
rately. Its solution can be written in the form
h(x, t) = H(x 2t),
and therefore, h must be prescribed on t = 0 and on x = 0, since the characteristics
propagate from both the x and t axis.
We rewrite the rst two equations as (f u1 , g u2 ):
0 1
Ut + Ux = 0,
1 0
(1)
u (x, 0)
U (x, 0) = .
u(2)(x, 0)
The eigenvalues ofthe matrix
A are 1 = 1, 2 = 1 and the corresponding eigen-
1 1
vectors are e1 = , e2 = . Thus,
1 1
1 0 1 1 1 1 1 1
= , = , = .
0 1 1 1 2 1 1
Let U = V . Then,
Ut + AUx = 0,
Vt + AVx = 0,
Vt + 1 AVx = 0,
Vt + Vx = 0.
Thus, the transformed problem is
1 0
Vt + Vx = 0, (14.28)
0 1
1 1 1 1 u(1) (x, 0)
V (x, 0) = U (x, 0) = . (14.29)
2 1 1 u(1) (x, 0)
Equation (14.28) gives traveling wave solutions of the form:
Proof. a) Let us change the notation (u u(1), v u(2)). Rewrite the equation as
1 a
Ut + Ux = 0, (14.32)
b 1
!
(1)
u(1)(x, 0) u0 (x)
U (x, 0) = = ,
u(2)(x, 0) (2)
u0 (x)
u(1)(0, t)
U (0, t) = = 0.
u(2)(0, t)
The eigenvalues of the matrix A are 1 = 1 ab, 2 = 1 + ab.
1 ab 0
= .
0 1 + ab
Let U = V , where is a matrix of eigenvectors. Then,
Ut + AUx = 0,
Vt + AVx = 0,
Vt + 1 AVx = 0,
Vt + Vx = 0.
Thus, the transformed problem is
1 ab 0
Vt + Vx = 0, (14.33)
0 1 + ab
ut + ux = 0
vt (1 cx2 )vx + ux = 0
We could also do similar analysis we have done in other problems on rst order sys-
tems involving nding eigenvalues/eigenvectors of the system and using the fact that
u(1)(x, t) is known at both boundaries (i.e. values of u(1)(1, t) can be traced back either
to initial conditions or to boundary conditions on x = 1).
Partial Dierential Equations Igor Yanovsky, 2005 122
ut + aux + bvx = 0
vt + cux + dvx = 0
u(x, 0) = u0 (x)
v(x, 0) = v0 (x).
Proof. a) Let us change the notation (u u(1), v u(2)). Rewrite the equation as
a b
Ut + Ux = 0, (14.35)
c d
(1) !
(1)
u (x, 0) u0 (x)
U (x, 0) = = .
u(2)(x, 0) (2)
u0 (x)
The system is hyperbolic if for each value of u(1) and u(2) the eigenvalues are real
and the matrix is diagonalizable, i.e. there is a complete set of linearly independent
eigenvectors. The eigenvalues of the matrix A are
a + d (a + d)2 4(ad bc) a + d (a d)2 + 4bc
1,2 = = .
2 2
We need (a d)2 + 4bc > 0. This also makes the problem to be diagonalizable.
Let U = V , where is a matrix of eigenvectors. Then,
Ut + AUx = 0,
Vt + AVx = 0,
Vt + 1 AVx = 0,
Vt + Vx = 0.
Thus, the transformed problem is
1 0
Vt + Vx = 0, (14.36)
0 2
u u u 9vx uxx
b) We want to solve +L = 0, L = . We have
v v v ux vxx
t
u u u u
= L = , i.e. = et . We can write the solution as
v t v v v
&
un (t)einx
U (x, t) = & = an e1 t v1 einx + bn e2 t v2 einx
vn (t)einx
n=
(n2 +3n)t 3i inx (n2 3n)t 3i
= an e e + bn e einx .
1 1
n=
ix
3i inx 3i inx e
U (x, 0) = an e + bn e = ,
1 1 0
n=
an = bn = 0, n = 1;
1 1
a1 + b1 = and a1 = b1 a1 = b1 = .
3i 6i
1 4t 3i 1 3i
U (x, t) = e eix + e2t eix
6i 1 6i 1
1 4t
2 (e + e2t)
= 1 4t 2t ) eix .
6i (e e
26 27
26
ChiuYens and Sung-Has solutions give similar answers.
27
Questions about this problem:
1. Needed to nd eigenfunctions, not eigenvectors.
2. The notation of L was changed. The problem statement incorporates the derivatives wrt. t into L.
3. Why can we write the solution in this form above?
Partial Dierential Equations Igor Yanovsky, 2005 126
ut ux = vt + vx = 0
in the diamond shaped region 1 < x + t < 1, 1 < x t < 1. For each of
the following boundary value problems state whether this problem is well-posed. If it is
well-posed, nd the solution.
a) u(x + t) = u0 (x + t) on x t = 1, v(x t) = v0 (x t) on x + t = 1.
b) v(x + t) = v0 (x + t) on x t = 1, u(x t) = u0 (x t) on x + t = 1.
Proof. We have
ut ux = 0,
vt + vx = 0.
u is constant along the characteristics: x + t = c1 (s).
Thus, its solution is u(x, t) = u0 (x + t).
It the initial condition is prescribed at x t = 1,
the solution can be determined in the entire region
by tracing back through the characteristics.
v is constant along the characteristics: x t = c2 (s).
Thus, its solution is v(x, t) = v0 (x t).
It the initial condition is prescribed at x + t = 1,
the solution can be determined in the entire region
by tracing forward through the characteristics.
Partial Dierential Equations Igor Yanovsky, 2005 127
ut + uux + (F ())x = 0,
t + (u)x = 0.
a) Assume that
where U0 , R0 > 0 are constants and f1 (x), g1 (x) are small perturbations. Lin-
earize the equations and given conditions for F such that the linearized problem is
well-posed.
b) Assume that U0 > 0 and consider the above linearized equations for 0 x 1,
t 0. Construct boundary conditions such that the initial-boundary value problem is
well-posed.
ut + uux + x = uxx ,
t + (u)x = 0
2 2
||u(x, t)||22 = 2
u (x, t) dx = inx
an (t)e an (t)e imx
dx
0 0 n= m=
2
4 +2n2 )t
= a2n (t) einx einx dx = 2 a2n (t) = 2 a2n (0)e2(n
n= 0 n= n=
2(n4 +2n2 )t 2 2t
2 2
an (0) e 2 2
e2(n 1) t
= 2 an (0) e
n= n= n= n=
||||2 = C1 , (convergent)
= C2 e2t||||2.
METHOD III: Can use Fourier transform. See ChiuYens solutions, that have both
Method II and III.
Partial Dierential Equations Igor Yanovsky, 2005 132
the L2 -norm of f .
a) Express ||dpf /dxp||2 in terms of the Fourier coecients of f .
b) Let q > p > 0 be integers. Prove that > 0, K = N (, p, q), constant, such that
dpf 2 dq f 2
p q + K||f ||2.
dx dx
c) Discuss how K depends on .
31
Proof. a) Let
f (x) = fn einx ,
p
d f
= fn (in)peinx ,
dxp
dpf 2 2 2
p inx 2
2
p = fn (in) e dx = |i2 |p fn np einx dx
dx 0 0
2
2
= fn np einx dx = 2 fn2 n2p .
0 n=0
b) We have
dpf 2 dq f 2
p q + K||f ||2,
dx dx
2 fn2 n2p 2 fn2 n2q + K 2 fn2 ,
n=0 n=0 n=0
2p 2q
n n K,
2p q
n (1 n ) K, some q > 0.
< 0, f or n large
Thus, the above inequality is true for n large enough. The statement follows.
31
Note:
L
inx imx 0 n = m
e e dx =
0 L n=m
Partial Dierential Equations Igor Yanovsky, 2005 133
32
Problem (S90, #5). Consider the ame front equation
ut + uux + uxx + uxxxx = 0
with 2-periodic initial data
u(x, 0) = f (x), f (x) = f (x + 2) C .
a) Determine the solution, if f (x) f0 = const.
b) Assume that
f (x) = 1 + g(x), 0 < 1, |g| = 1, g(x) = g(x + 2).
Linearize the equation. Is the Cauchy problem well-posed for the linearized equation,
i.e., do its solutions v satisfy an estimate
||v(, t)|| Ke(tt0 ) ||v(, t0)||?
c) Determine the best possible constants K, .
+ + u1 u1xx + u1 u1xxxx = 0,
dt 2 2 x
2 2
1 d 2 u21 2 2
u1 dx + + u1 u1xx dx + u1 u1xxxx dx = 0,
2 dt 0 20 0 0
=0
2 2 2 2 2
1 d 2 2
||u1 ||2 + u1 u1x u1x dx + u1 u1xxx u1x u1xx + u21xx dx = 0,
2 dt 0 0 0
0 0
=0 =0 =0
2 2
1 d
||u1 ||22 = u21x dx u21xx dx.
2 dt 0 0
32
S90 #5, #6, #7 all have similar formulations.
Partial Dierential Equations Igor Yanovsky, 2005 134
Thus,
2 2
1 d
||u1 ||22 = u21x dx u21xx dx
2 dt 0 0
2
2
22
= dx n2 an (t)einx dx
nan (t)e inx
0 0
2 2
= 2 n an (t) 2 n an (t)2 = 2
4
an (t)2 (n2 + n4 ) 0.
ut = ux + u4 for t > 0
u = u0 for t = 0
for 0 < x < 2. Dene the set A = {u = u(x) : u (k) = 0 if k < 0}, in which
{
u(k, t)} is the Fourier series of u in x on [0, 2].
a) If u0 A, show that u(t) A.
b) Find dierential equations for u
(0, t), u
(1, t), and u
(2, t).
Proof. a) Solving
ut = ux + u4
u(x, 0) = u0 (x)
by the method of characteristics, we get
u0 (x + t)
u(x, t) = 1 .
(1 3t(u0 (x + t))3 ) 3
Since u0 A, u
0k = 0 if k < 0. Thus,
kx
u0 (x) = 0k ei
u 2 .
k=0
Since
2
1 kx
k =
u u(x, t) ei 2 dx,
2 0
Partial Dierential Equations Igor Yanovsky, 2005 135
we have
kx
u(x, t) = k ei
u 2 ,
k=0
ut + U ux + uxx + uxxxx = 0,
u(x, 0) = f (x), f (x) = f (x + 2) C .
This can also been done using Fourier Transform. See ChiuYens solutions where the
above method and the Fourier Transform methods are used.
33
S90 #5, #6, #7 all have similar formulations.
Partial Dierential Equations Igor Yanovsky, 2005 137
34
Problem (S90, #7). Consider the nonlinear equation
Let w = ux , then
2 2 2
1 d
||w||2 = w 3 dx wwxx dx wwxxxx dx
2 dt 0 0 0
2 2 2 2
3 2 2
= w dx + wx dx wxx dx w 3 dx,
0 0 0 0
2
d
||ux||2 = u3x dx.
dt 0
Partial Dierential Equations Igor Yanovsky, 2005 139
1
u(x, t) = uA (x, t) + uD (x, t) = sin x cos ct + xt + t + (cos x cos x cos ct).
c2
35
Note the relationship: x , t s.
Partial Dierential Equations Igor Yanovsky, 2005 140
We can check that the solution satises equation (16.1). Can also check that uA , uD
satisfy
2 A 2 D
tt c uxx = 0,
uA tt c uxx = cos x,
uD
uA (x, 0) = sin x, uA
t (x, 0) = 1 + x; uD (x, 0) = 0, uDt (x, 0) = 0.
Partial Dierential Equations Igor Yanovsky, 2005 141
n 2
nx 2 nx
bn (t) sin +c bn (t) sin = 0, or
n=1
L n=1
L L
nc 2
bn (t) + bn (t) = 0,
L
whose general solution is
nct nct
bn (t) = cn sin + dn cos . (16.4)
L L
Also, bn (t) = cn ( nc
L ) cos L dn ( L ) sin L .
nct nc nct
4 cos(2n + 1)t sin(2n + 1)x
u(x, t) = .
n=0 (2n + 1)
Partial Dierential Equations Igor Yanovsky, 2005 144
2 (cos n 1) cos nt cos nx
u(x, t) = + .
2 n=1 n2
Partial Dierential Equations Igor Yanovsky, 2005 148
Example (McOwen 3.1 #4). Consider the initial boundary value problem
2
utt c uxx = 0 for x > 0, t > 0
u(x, 0) = g(x), ut (x, 0) = h(x) for x > 0 (16.22)
u(0, t) = 0 for t 0,
where g(0) = 0 = h(0). If we extend g and h as odd functions on < x < , show
that dAlemberts formula gives the solution.
To show that dAlemberts formula gives the solution to (16.23), we need to show that
the solution given by dAlemberts formula satises the boundary condition u(0, t) = 0.
1 1 x+ct
u
(x, t) = g(x + ct) + g(x ct)) +
( h() d,
2 2c xct
1 1 ct
u
(0, t) = (
g(ct) + g(ct)) + h() d
2 2c ct
1 1
= g(ct) g(ct)) + (H(ct) H(ct))
(
2 2c
1
= 0 + (H(ct) H(ct)) = 0,
2c
x
where we used H(x) =
h() d; and since h is odd, then H is even.
0
Example (McOwen 3.1 #5). Find in closed form (similar to dAlembets formula)
the solution u(x, t) of
2
utt c uxx = 0 for x, t > 0
u(x, 0) = g(x), ut (x, 0) = h(x) for x > 0 (16.24)
u(0, t) = (t) for t 0,
where g, h, C 2 satisfy (0) = g(0), (0) = h(0), and (0) = c2 g (0). Verify that
u C 2 , even on the characteristic x = ct.
Proof. As in (McOwen 3.1 #4), we can extend g and h to be odd functions. We want
to transform the problem to have zero boundary conditions.
Consider the function:
To show that dAlemberts formula gives the solution to (16.27), we need to show that
the solution given by dAlemberts formula satises the boundary condition ux (0, t) = 0.
1 1 x+ct
u
(x, t) = g(x + ct) + g(x ct)) +
( h() d.
2 2c xct
1 1
x (x, t) =
u g (x + ct) + g (x ct)) + [h(x
( ct)],
+ ct) h(x
2 2c
1 1
ux (0, t) = g (ct) + g (ct)) + [h(ct) h(ct)]
( = 0.
2 2c
Since g is even, then g is odd.
Proof. DAlemberts formula is derived by plugging in the following into the above
equation and initial conditions:
u(x, t) = F (x + ct) + G(x ct).
As in (Zachmanoglou 7.2), we can extend g and h to be even functions.
36
Similar to McOwen 3.1 #5. The notation in this problem is changed to be consistent with McOwen.
Partial Dierential Equations Igor Yanovsky, 2005 152
by rst nding a particular solution by separation of variables and then adding the
appropriate solution of the homogeneous PDE.
b) Since the particular solution is not unique, it will not be obvious that the solution
to the initial value problem that you have found in part (a) is unique. Prove that it is
unique.
It can be veried that the solution satises the above homogeneous PDE with the
boundary conditions. Thus, the complete solution is:
37
Alternatively, we could use Duhamels principle to nd the solution:
t x+(ts)
1
u(x, t) = f () cos s d ds.
2 0 x(ts)
b) The particular solution is not unique, since any constants a, b give the solution.
However, we show that the solution to the initial value problem is unique.
Suppose u1 and u2 are two solutions. Then w = u1 u2 satises:
wtt wxx = 0,
w(x, 0) = 0, wt (x, 0) = 0.
DAlemberts formula gives
x+t
1 1
w(x, t) = (g(x + t) + g(x t)) + h() d = 0.
2 2 xt
Thus, the solution to the initial value problem is unique.
37
Note the relationship: x , t s.
Partial Dierential Equations Igor Yanovsky, 2005 155
1 1
= , = .
2 2
1 1
w yw = w + 2ww ,
2 2
1 1
w + 2ww + yw + w = 0.
2 2
We can write the ODE as
1
w + 2ww + (yw) = 0.
2
Integrating it with respect to y, we obtain the rst order ODE:
1
w + w 2 + yw = c.
2
Partial Dierential Equations Igor Yanovsky, 2005 156
cf + 6f f + f = 0. (16.34)
cf + 3f 2 + f = a, (16.35)
cf f + 3f 2 f + f f = af .
c (f )2
f2 + f3 + = 0, or f = f c 2f .
2 2
The solution of this ODE is
c 2 c
f (x) = sech [ (x x0 )],
2 2
where x0 is the constant of integration. A solution of this form is called a soliton.
38
Evans, p. 174; Strauss, p. 367.
Partial Dierential Equations Igor Yanovsky, 2005 157
Thus, either |f | 0 f = 0, or
f n + c = 0. Since f 0 as x , we have c = 0 f = 0.
Thus, solitary waves do not exist if n is even.
Partial Dierential Equations Igor Yanovsky, 2005 159
When n = 1, we have
(f )2 = f 3 + cf 2 . (16.39)
We show that all conditions of the problem are satised provided c > 0, including
1
u = c sech2 c(x ct) , or
2
c 2
2 c c
f = c sech = = c cosh .
2 cosh2 [ 2 c ] 2
We have
c 3 c c c 3 c
f = 2c cosh sinh = c c cosh sinh ,
2 2 2 2 2
3 2
c sinh 2 c
2 ,
(f ) =
cosh6 2 c
c3
f3 = ,
cosh6 c
2
c3
cf 2 = .
cosh4 c
2
Also, f, f , f 0, as || , since
2
2 c c 2
f () = c sech = = c
c
c
0, as || .
2 cosh2 [ 2 c ] e[ 2 ] + e[ 2 ]
Similarly, f , f 0, as || .
39
cosh2 x sinh2 x = 1.
ex + ex ex ex
cosh x = , sinh x =
2 2
Partial Dierential Equations Igor Yanovsky, 2005 160
Problem (S00, #5). Look for a traveling wave solution of the PDE
of the form u(x, t) = v(x ct). In particular, you should nd an ODE for v. Under
the assumption that v goes to a constant as |x| , describe the form of the solution.
Proof. Since (u2 )x = 2uux, and (u2 )xx = 2u2x + 2uuxx, we have
utt + 2u2x + 2uuxx = uxxxx .
We look for a traveling wave solution
v + c2 v + v 2 = a(x ct) + b.
u(x, t) = (x st)
lim (y) = 0
y
+ = lim (y).
y
s 1
+ + 2 = 0.
2
1 = = 2 ,
s 1 s 1
2 = = + 2 = 2 1 + 21 .
2 2
1 = 2 = 0, 1 = 2 = 0, 1 = 2 = 0,
1 2
1 2
2 = 2 s 1 + 2 1 = 0; 2 = s 1 + 2 1 = 0; 2 = 1 1 (s 12 1 ) = 0.
1 = 2 = f (1 , 2 ),
s 1
2 = 2 1 + 21 = g(1 , 2 ).
2
Partial Dierential Equations Igor Yanovsky, 2005 162
For (1 , 2 ) = (0, 0) :
1
det(J|(0,0) I) = s = 2 + +
s
= 0.
'
2
= 2
4 2 .
s
2
If 4 >s R, < 0.
(0,0) is Stable Improper Node.
2
If 4 <s C, Re( ) < 0.
(0,0) is Stable Spiral Point.
For (1 , 2 ) = (2s, 0) :
1
det(J|(2s,0) I) = s = 2 + s
= 0.
'
2
= 2
4 s
2 + .
+ > 0, < 0.
(2s,0) is Untable Saddle Point.
b) Since
we may have
That is, a particle may start o at an unstable node (2s, 0) and as t increases, approach
the stable node (0, 0).
A phase diagram with (0, 0) being a stable spiral point, is shown below.
Partial Dierential Equations Igor Yanovsky, 2005 163
Partial Dierential Equations Igor Yanovsky, 2005 164
ut + f (u)x = uxx
where f is smooth and > 0. We seek traveling wave solutions to this equation,
i.e., solutions of the form u = (x st), under the boundary conditions
u uL and ux 0 as x ,
u uR and ux 0 as x +.
Find a necessary and sucient condition on f , uL , uR and s for such traveling waves
to exist; in case this condition holds, write an equation which denes implicitly.
40
For the solution for the second part of the problem, refer to Chiu-Yens solutions.
Partial Dierential Equations Igor Yanovsky, 2005 165
One class of solutions is the set of wavefront solutions. These have the form u(x, t) =
(x + ct), c 0.
Determine the ordinary dierential equation and boundary conditions which must
satisfy (to be of physical interest). Carry out a phase plane analysis of this equation,
and show that physically interesting wavefront solutions are possible if c 2, but not if
0 c < 2.
c = (1 ) + ,
c + 2 = 0,
(s) 0, as s ,
(s) 1, as s +,
0 1.
In order to nd and analyze the stationary points of an ODE above, we write it as a
rst-order system.
y1 = ,
y2 = .
y1 = = y2 ,
y2 = = c + 2 = cy2 y1 + y12 .
y1 = y2 = 0, y2 = 0,
2
y2 = cy2 y1 + y1 = 0; y1 (y1 1) = 0.
y1 = y2 = f (y1 , y2 ),
y2 = cy2 y1 + y12 = g(y1 , y2 ).
In order to classify a stationary point, need to nd eigenvalues of a linearized system
at that point.
( )
f f
y1 y2 0 1
J(f (y1 , y2 ), g(y1, y2 )) = = .
g
y
g
y
2y1 1 c
1 2
Partial Dierential Equations Igor Yanovsky, 2005 166
By looking at the phase plot, a particle may start o at an unstable node (0, 0) and as
t increases, approach the unstable node (1, 0).
Partial Dierential Equations Igor Yanovsky, 2005 167
Partial Dierential Equations Igor Yanovsky, 2005 168
t + x u = 0
t u + x (u) = x2 u
look for traveling wave solutions of the form (x, t) = (y = x st), u(x, t) = u(y =
x st). In particular
a) Find a rst order ODE for u.
b) Show that this equation has solutions of the form
u(y) = u0 + u1 tanh(y + y0 ),
b) Note that the ODE above may be written in the following form:
u + Au2 + Bu = C,
which is a nonlinear rst order equation.
Partial Dierential Equations Igor Yanovsky, 2005 169
ft + fx = g 2 f 2
gt gx = f 2 g
a) Find a system of ODEs that describes traveling wave solutions of the PDE
system; i.e. for solutions of the form f (x, t) = f (x st) and g(x, t) = g(x st).
b) Analyze the stationary points and draw the phase plane for this ODE system in the
standing wave case s = 0.
16.5 Dispersion
Problem (S97, #8). Consider the following equation
ut = (f (ux ))x uxxxx , f (v) = v 2 v, (16.40)
with constant .
a) Linearize this equation around u = 0 and nd the principal mode solution of the
form et+ikx . For which values of are there unstable modes, i.e., modes with = 0
for real k? For these values, nd the maximally unstable mode, i.e., the value of k with
the largest positive value of .
b) Consider the steady solution of the (fully nonlinear) problem. Show that the resulting
equation can be written as a second order autonomous ODE for v = ux and draw the
corresponding phase plane.
Proof. a) We have
ut = (f (ux))x uxxxx ,
ut = (u2x ux )x uxxxx ,
ut = 2ux uxx uxx uxxxx .
However, we need to linearize (16.40) around u = 0. To do this, we need to linearize f .
u2
f (u) = f (0) + uf (0) + f (0) + = 0 + u(0 1) + = u + .
2
Thus, we have
ut = uxx uxxxx .
Consider u(x, t) = et+ikx .
et+ikx = (k2 k4 )et+ikx ,
= k2 k4 .
To nd unstable nodes, we set = 0, to get
1
= .
k2
To nd the maximally unstable mode, i.e., the value of k with the largest positive
value of , consider
(k) = k2 k4 ,
(k) = 2k 4k3 .
To nd the extremas of , we set = 0. Thus,the extremas are at
1
k1 = 0, k2,3 = .
2
To nd if the extremas are maximums or minimums, we set = 0:
(k) = 2 12k2 = 0,
(0) = 2 > 0 k = 0 is the minimum.
1 1
= 4 < 0 k = is the maximum unstable mode.
2 2
1 1
= is the largest positive value of .
2 4
Partial Dierential Equations Igor Yanovsky, 2005 172
b) Integrating , we get
u2x ux uxxx = 0.
Let v = ux . Then,
v 2 v vxx = 0, or
v2 v
v = .
In order to nd and analyze the stationary points of an ODE above, we write it as a
rst-order system.
y1 = v,
y2 = v .
y1 = v = y2 ,
v2 v y 2 y1
y2 = v = = 1 .
y1 = y2 = 0, y2 = 0,
y12 y1
y2 = = 0; y1 (y1 1) = 0.
y1 = y2 = f (y1 , y2 ),
2
y y1
y2 = 1 = g(y1 , y2 ).
In order to classify a stationary point, need to nd eigenvalues of a linearized system
at that point.
( )
f f
y1 y2 0 1
J(f (y1 , y2 ), g(y1, y2 )) = g g = 2y1 1 .
y1 y2 0
'
For (y1 , y2 ) = (0, 0), = 1 .
If < 0, R, ' + > 0, < 0. (0,0) is Unstable Saddle Point.
1
If > 0, = i C, Re( ) = 0. (0,0) is Spiral Point.
'
For (y1 , y2 ) = (1, 0), = 1 .
'
If < 0, = i 1 C, Re( ) = 0. (1,0) is Spiral Point.
If > 0, R, + > 0, < 0. (1,0) is Unstable Saddle Point.
Partial Dierential Equations Igor Yanovsky, 2005 173
Partial Dierential Equations Igor Yanovsky, 2005 174
Proof.
dE u
=0 = (utt u)ut dx = utt ut dx ut ds + u ut dx
dt n
1 2 1 2
= (ut ) dx + |u| dx + a(x)u2t ds.
2 t 2 t
Thus,
1
a(x)u2t dx = u2t + |u|2 dx.
2 t
0
In order to prove that the given E(t) 0 from scratch, take its derivative with respect
to t:
dE
(t) = ut utt + u ut dx
dt
u
= ut utt dx + ut ds ut u dx
n
Proof. We have
dE 1
(t) = 2
u u
t tt + u u t dx
dt c (x)
1 u
= ut utt dx + ut ds ut u dx
c2 (x) n
1 1 2
= ut 2 utt u dx + ut dx
c (x) (x)
=0
1 2 > 0, if (x) > 0, x ,
= ut dx =
(x) < 0, if (x) < 0, x .
Partial Dierential Equations Igor Yanovsky, 2005 176
Problem (F92, #2). Let Rn . Let u(x, t) be a smooth solution of the following
initial boundary value problem:
Proof. Take the derivative of E(t) with respect to t. Note that the boundary integral
is 0 by Huygens principle.
dE
(t) = u ut + ututt dx
dt 3
R
u
= ut ds ut u dx + ututt dx
R3 n R3 R3
=0
= ut (u + utt) dx = ut (a(x)ut) dx = a(x)u2t dx 0.
R3 R3 R3
Thus, dE
dt 0 E(t) E(0), i.e. E(t) is a decreasing function of t.
Partial Dierential Equations Igor Yanovsky, 2005 178
Problem (W03, #8). a) Consider the damped wave equation for high-speed waves
(0 < << 1) in a bounded region D
2 utt + ut = u
with the boundary condition u(x, t) = 0 on D. Show that the energy functional
E(t) = 2 u2t + |u|2 dx
D
b) Consider the solution to the boundary value problem in part (a) with initial data
u (x, 0) = 0, ut (x, 0) = f (x), where f does not depend on and < 1. Use part
(a) to show that
|u(x, t)|2 dx 0
D
c) Show that the result in part (b) does not hold for = 1. To do this consider
the case where f is an eigenfunction of the Laplacian, i.e. f + f = 0 in D and
f = 0 on D, and solve for u explicitly.
Proof. a)
dE 2
= 2 ututt dx + 2u ut dx
dt
D D
2 u
= 2 ututt dx + 2 ut ds 2uut dx
D n
D
D
=0, (u=0 on D)
2
= 2 ( utt u)ut dx = = 2 |ut |2 dx 0.
D D
Thus, E(t) E(0), i.e. E(t) is nonincreasing.
c) If = 1,
E (0) = 2(1) f (x)2 dx = f (x)2 dx.
D D
Sincef is independent of , E(0) does not approach 0 as 0. We can not conclude
that D |u(x, t)|2 dx 0.
Partial Dierential Equations Igor Yanovsky, 2005 179
ftt fxx = f (1 + f 2 )1
for x [0, 1], with f (x = 0, t) = f (x = 1, t) = 0 and with smooth initial data f (x, t) =
f0 (x).
a) Find an energy integral E(t) which is constant in time.
b) Show that |f (x, t)| < c for all x and t, in which c is a constant.
Hint: Note that
f 1 d
2
= log(1 + f 2 ).
1+f 2 df
Proof. a) Since f (0, t) = f (1, t) = 0, t, we have ft (0, t) = ft (1, t) = 0. Let
1
dE
= 0 = ftt fxx + f (1 + f 2 )1 ft dx
dt 0
1 1 1
f ft
= fttft dx fxx ft dx + 2
dx
0 0 0 1+f
1 1 1
1 f ft
= fttft dx [fx ft ]0 + fx ftx dx + dx
0 0 0 1+f
2
=0
1 1 1
1 2 1 2 1
= (ft ) dx + (fx ) dx + (ln(1 + f 2 )) dx
0 2 t 0 2 t 0 2 t
1 d 1 2
= ft + fx2 + ln(1 + f 2 ) dx.
2 dt 0
Thus,
1
1
E(t) = ft2 + fx2 + ln(1 + f 2 ) dx.
2 0
b) We want to show that f is bounded. For smooth f (x, 0) = f0 (x), we have
1 1
E(0) = ft (x, 0)2 + fx (x, 0)2 + ln(1 + f (x, 0)2 ) dx < .
2 0
Since E(t) is constant in time, E(t) = E(0) < . Thus,
1 1 2 1 1 2
ln(1 + f ) dx ft + fx2 + ln(1 + f 2 ) dx = E(t) < .
2 0 2 0
Hence, f is bounded.
Partial Dierential Equations Igor Yanovsky, 2005 180
Problem (S98, #4). a) Let u(x, y, z, t), < x, y, z < be a solution of the
equation
utt + ut = uxx + uyy + uzz
u(x, y, z, 0) = f (x, y, z), (16.41)
ut (x, y, z, 0) = g(x, y, z).
Here f , g are smooth functions which vanish if x2 + y 2 + z 2 is large enough. Prove
that it is the unique solution for t 0.
b) Suppose we want to solve the same equation (16.41) in the region z 0, <
x, y < , with the additional conditions
u(x, y, 0, t) = f (x, y, t)
uz (x, y, 0, t) = g(x, y, t)
b)
Partial Dierential Equations Igor Yanovsky, 2005 182
Problem (F94, #8). The one-dimensional, isothermal uid equations with viscosity
and capillarity in Lagrangian variables are
vt ux = 0
ut + p(v)x = uxx vxxx
in which v(= 1/) is specic volume, u is velocity, and p(v) is pressure. The coecients
and are non-negative.
Find an energy integral which is non-increasing (as t increases) if > 0 and con-
served if = 0.
Hint: if = 0, E = u2 /2 P (v) dx where P (v) = p(v).
b) We have
u2t
+ Q(ux ).
F (ut , ux) =
2
41 For F to be convex, the Hessian matrix of partial derivatives must be positive denite.
41
A function f is convex on a convex set S if it satises
f (x + (1 )y) f (x) + (1 )f (y)
for all 0 1 and for all x, y S.
If a one-dimensional function f has two continuous derivatives, then f is convex if and only if
f (x) 0.
In the multi-dimensional case the Hessian matrix of second derivatives must be positive semi-denite,
that is, at every point x S
yT 2 f (x) y 0, for all y.
The Hessian matrix is the matrix with entries
2 f (x)
[2 f (x)]ij .
xi xj
For functions with continuous second derivatives, it will always be symmetric matrix: fxi xj = fxj xi .
Partial Dierential Equations Igor Yanovsky, 2005 184
c) We have
d 1
F (ut , ux ) dx = 0,
dt 0
1
F (ut , ux ) dx = const,
0
1 1
F (ut , ux ) dx = F (ut (x, 0), ux(x, 0)) dx,
0 0
1
2 1
2
ut v0
+ Q(ux) dx = + Q(u0x ) dx.
0 2 0 2
42
Problem (S96, #8). Let u(x, t) be the solution of the Korteweg-de Vries equation
ut + uux = uxxx, 0 x 2,
with 2-periodic boundary conditions and prescribed initial data
u(x, t = 0) = f (x).
a) Prove that the energy integral
2
I1 (u) = u2 (x, t) dx
0
is independent of the time t.
b) Prove that the second energy integral,
2
1 2 1
I2 (u) = ux (x, t) + u3 (x, t) dx
0 2 6
is also independent of the time t.
c) Assume the initial data are such that I1 (f ) + I2 (f ) < . Use (a) + (b) to prove
that the maximum norm of the solution, |u| = supx |u(x, t)|, is bounded in time.
Hint: Use the following inequalities (here, |u|p is the Lp -norm of u(x, t) at xed time
t):
|u|2 (|u|22 + |ux |22 ) (one of Sobolevs inequalities);
6
|u|33 |u|22 |u| (straightforward).
Proof. a) Multiply the equation by u and integrate. Note that all boundary terms are
0 due to 2-periodicity.
uut + u2 ux = uuxxx ,
2 2 2
uut dx + u2 ux dx = uuxxx dx,
0 0 0
2
1 d 2 2 1 2 3 2
u dx +
(u )x dx = uuxx 0 ux uxx dx,
2 dt 0 3 0 0
1 d 2 2 1 3 2 1 2 2
u dx + u 0 = (ux )x dx,
2 dt 0 3 2 0
1 d 2 2 1 2
u dx = u2x 0 = 0.
2 dt 0 2
2
I1 (u) = u2 dx = C.
0
2 2
Thus, I1 (u) = 0 u (x, t) dx is independent of the time t.
Alternatively, we may dierentiate I1 (u):
2 2
dI1 d 2 2
(u) = u dx = 2uut dx = 2u(uux + uxxx) dx
dt dt 0 0 0
2 2 2 2
2 3 2
= 2
2u ux dx + 2uuxxx dx =
(u )x dx + 2uuxx 0 2ux uxx dx
0 0 0 3 0
2
2 2 2
= u3 0 (u2x )x dx = u2x 0 = 0.
3 0
42
Also, see S92, #7.
Partial Dierential Equations Igor Yanovsky, 2005 186
dI2 d 2
1 2 1 3 1
(u) = ux + u dx = ux uxt + u2 ut dx =
dt dt 0 2 6 0 2
We dierentiate the original equation with respect to x:
ut = uux + uxxx
utx = (uux )x + uxxxx .
2
1 2 2
= ux ((uux )x + uxxxx ) dx + u (uux + uxxx ) dx
0 2 0
2 2
1 2 3 1 2 2
= ux (uux)x dx + ux uxxxx dx u ux dx + u uxxx dx
0 0 2 0 2 0
2 2
2 2
=
ux uux 0 +
uxxuux dx + ux uxxx 0 uxx uxxx dx
0 0
1 2
u4 1 2 1 2
dx + u2 uxx0 2uux uxx dx
2 0 4 x 2 2 0
2 2 2
1 u4 2
= uxx uux dx uxx uxxx dx uux uxx dx
0 0 2 4 0 0
2 2 2
2 2
= uxx uxxx dx = uxx 0 + uxxx uxx dx = uxxx uxx dx = 0,
0 0 0
2 2
since 0 uxxuxxx dx = + 0 uxx uxxx dx. Thus,
2
1 2 1 3
I2 (u) = ux (x, t) + u (x, t) dx = C,
0 2 6
and I2 (u) is independent of the time t.
1 2 1 3 1 1
I2 (u) = ux + u dx = ||ux||22 + ||u||33.
0 2 6 2 6
Using given inequalities, we have
1
||u||2 (||u||22 + ||ux||22 ) I1 (u) + 2I2 (u) ||u||33
6 6 3
2
I1 (u) + I2 (u) + ||u||2 ||u|| I1 (u) + I2 (u) + I1 (u)||u||
6 3 18 6 3 18
= C + C1 ||u||.
||u||2 C + C1 ||u||,
||u|| C2 .
Thus, ||u|| is bounded in time.
Also see Energy Methods problems for higher order equations (3rd and
4th) in the section on Gas Dynamics.
Partial Dierential Equations Igor Yanovsky, 2005 187
Proof.
We may use the Kirchhos formula:
1 t
u(x, t) = t g(x + ct) dS + h(x + ct) dS
4 t ||=1 4 ||=1
1 2 2
= t (x1 + ct1 ) + (x2 + ct2 ) dS + 0 =
4 t ||=1
We may solve the problem by Hadamards method of descent, since initial con-
ditions are independent of x3 . We need to convert surface integrals in R3 to domain
integrals in R2 . Specically, we need to express the surface measure on the upper half
2
of the unit sphere S+ in terms of the two variables 1 and 2 . To do this, consider
'
f (1 , 2 ) = 1 12 22 over the unit disk 12 + 22 < 1.
'
d1 d2
dS = 1 + (f1 )2 + (f2 )2 d1 d2 = .
1 12 22
Partial Dierential Equations Igor Yanovsky, 2005 188
1 g(x1 + ct1 , x2 + ct2 ) d1 d2
u(x1 , x2 , t) = 2t
4 t 12 +22 <1 1 12 22
t h(x1 + ct1 , x2 + ct2 ) d1d2
+ 2
4 2 + 2 <1 1 12 22
1 2
1 (x1 + t1 )2 + (x2 + t2 )2
= 2t d1 d2 + 0,
4 t 12 +22 <1 1 12 22
1 x21 + 2x1 t1 + t2 12 + x22 + 2x2 t2 + t2 22
= t d1 d2
2 t 12 +22 <1 1 12 22
1 tx21 + 2x1 t2 1 + t3 12 + tx22 + 2x2 t2 2 + t3 22
= d1 d2
2 t 12 +22 <1 1 12 22
1 x21 + 4x1 t1 + 3t2 12 + x22 + 4x2 t2 + 3t2 22
= d1 d2
2 12 +22 <1 1 12 22
1 (x21 + x22 ) + 4t(x1 1 + x2 2 ) + 3t2 (12 + 22 )
= d1 d2
2 12 +22 <1 1 12 22
1 2 d d 4t x + x2 2
= (x1 + x22 ) 1 2 + 1 1 d1 d2
2 2 2 2
1 1 2 2 2 2 2 1 12 22
1 +2 <1 1 +2 <1
3t2 2 + 22
+ 1 d1 d2 =
2 12 +22 <1 1 12 22
2 1
1 2 2 d1 d2 1 2 2 r dr d
= (x1 + x2 ) = (x1 + x2 )
2 12 +22 <1 1 12 22 2 0 0 1 r2
2 1 1
1 2 2 du d
= (x1 + x22 ) 2 1 u = 1 r 2 , du = 2r dr
2 0 0 u2
2
1 2
= (x + x22 ) 1 d = x21 + x22 .
2 1 0
1 2
4t x1 1 + x2 2 4t 1 2 x1 1 + x2 2
= d1 d2 = d1 d2
2 12 +22 <1 1 1 2 2 2 2 1 122 1 12 22
= 0.
3t2 12 + 22 3t2 2 1 (r cos )2 + (r sin )2
= d1 d2 = r drd
2 12 +22 <1 1 12 22 2 0 0 1 r2
3t2 2 1 r3
= drd u = 1 r 2 , du = 2r dr
2 0 0 1 r2
2
3t2 2 t2 2
= d = d = 2t2 .
2 0 3 0
Check:
u(x, y, z, 0) = x2 + y 2 .
ut (x, y, z, t) = (x + t) + (y + t) (x t) (y t),
ut (x, y, z, 0) = 0.
utt (x, y, z, t) = 4,
ux (x, y, z, t) = (x + t) + (x t),
uxx (x, y, z, t) = 2,
uy (x, y, z, t) = (y + t) + (y t),
uyy (x, y, z, t) = 2,
uzz (x, y, z, t) = 0,
utt = uxx + uyy + uzz .
Partial Dierential Equations Igor Yanovsky, 2005 190
2
c T M 1
M 1 2M c
= + + .
2 3/4 c2 t 3/4 c2 T t c2 t
u(x, t) C1 /t for t > 2T .
For t 2T :
1 2T M + M + 4M cT d1 d2
|u(x, t)| = '
2 ||2<c2 t2 2 c2 t2
1 c||
2 t2
ct
1 r dr/c2t2
= (2T M + M + 4M ct)2 '
2 0 2
1 cr2 t2
M (2T + 1 + 4cT ) 1 du M (2T + 1 + 4cT ) M (2T + 1 + 4cT )2T
= 1/2
= 2 .
2 0 u 2 t
Letting C = max(C1 , M (2T + 1 + 4cT )2T ), we have |u(x, t)| C(x)/t.
area of intersection of the sphere of radius ct with the support of the functions g and
h.
Partial Dierential Equations Igor Yanovsky, 2005 192
Problem (S95, #6). Spherical waves in 3-d are waves symmetric about the origin;
i.e. u = u(r, t) where r is the distance from the origin. The wave equation
utt = c2 u
then reduces to
1 2
2
utt = urr + ur . (16.43)
c r
a) Find the general solutions u(r, t) by solving (16.43). Include both the incoming waves
and outgoing waves in your solutions.
b) Consider only the outgoing waves and assume the nite out-ux condition
for all t. The wavefront is dened as r = ct. How is the amplitude of the wavefront
decaying in time?
Proof. a) We want to reduce (16.43) to the 1D wave equation. Let v = ru. Then
vtt = rutt,
vr = rur + u,
vrr = rurr + 2ur .
Thus, (16.43) becomes
1 1 1
2
vtt = vrr ,
c r r
1
vtt = vrr ,
c2
vtt = c2 vrr ,
which has the solution
v(r, t) = f (r + ct) + g(r ct).
Thus,
1 1 1
u(r, t) = v(r, t) = f (r + ct) + g(r ct) .
r r
r
incoming, (c>0) outgoing, (c>0)
Problem (S00, #8). a) Show that for a smooth function F on the line, while
u(x, t) = F (ct + |x|)/|x| may look like a solution of the wave equation utt = c2 u
in R3 , it actually is not. Do this by showing that for any smooth function (x, t) with
compact support
u(x, t)(tt ) dxdt = 4 (0, t)F (ct) dt.
R3 R R
Note that, setting r = |x|, for any function w which only depends on r one has
w = r 2 (r 2 wr )r = wrr + 2r wr .
b) If F (0) = F (0) = 0, what is the true solution to utt = u with initial conditions
u(x, 0) = F (|x|)/|x| and ut(x, 0) = F (|x|)/|x|?
c) (Ralston Hw) Suppose u(x, t) is a solution to the wave equation utt = c2 u in
R3 R with u(x, t) = w(|x|, t) and u(x, 0) = 0. Show that
F (|x| + ct) F (|x| ct)
u(x, t) =
|x|
for a function F of one variable.
Proof. a) We have
u (tt ) dxdt = lim dt u (tt ) dx
R3 R 0 R |x|>
u
= lim dt (utt u) dx + u dS .
0 R |x|> |x|= n n
The nal equality is derived by integrating by parts twice in t, and using Greens
theorem:
u v
(vu uv) dx = v u ds.
n n
Since dS = 2 sin d d and n
= r
, substituting u(x, t) = F (|x| + ct)/|x|
gives:
u (tt ) dxdt = 4F (ct) dt.
R3 R R
Thus, u is not a weak solution to the wave equation.
b)
c) We want to show that v(|x|, t) = |x|w(|x|, t) is a solution to the wave equation in
one space dimension and hence must have the from v = F (|x| + ct) + G(|x| ct). Then
we can argue that w will be undened at x = 0 for some t unless F (ct) + G(ct) = 0
for all t.
We work in spherical coordinates. Note that w and v are independent of and . We
have:
1 1
vtt(r, t) = c2 w = c2 2 (r 2 wr )r = c2 2 (2rwr + r 2 wrr ),
r r
rwtt = c2 rwrr + 2wr .
Thus we see that vtt = c2 vrr , and we can conclude that
v(r, t) = F (r + ct) + G(r ct) and
F (r + ct) + G(r ct)
w(r, t) = .
r
Partial Dierential Equations Igor Yanovsky, 2005 195
limr0 w(r, t) does not exist unless F (ct) + G(ct) = 0 for all t. Hence
F (ct + r) + G(ct r)
w(r, t) = , and
r
F (ct + |x|) + G(ct |x|)
u(x, t) = .
|x|
Partial Dierential Equations Igor Yanovsky, 2005 196
K(x) = (x)
The fundamental solution for the Laplace operator is
1
2 log |x| if n = 2
K(x) = 1 2n
(2n)n
|x| if n 3.
45
Greens function is useful in satisfying Dirichlet boundary conditions.
Partial Dierential Equations Igor Yanovsky, 2005 197
G(x,)
where H(x, ) = nx is the Poisson kernel.
Thus if we know that the Dirichlet problem has a solution u C 2 (), then we can
calculate u from the Poisson integral formula (provided of course that we can compute
G(x, )).
Partial Dierential Equations Igor Yanovsky, 2005 198
If g(x ) is bounded and continuous for x Rn1 , then u() is C and harmonic in Rn+
and extends continuously to Rn+ such that u( ) = g( ).
Partial Dierential Equations Igor Yanovsky, 2005 199
u = ux1 x1 + ux2 x2 = 0 x ,
ux2 (x1 , 0) = f (x1 ) < x1 < .
1
Proof. a) Notation: x = (x, y), = (x0 , y0 ). Since K(x ) = 2 log |x |, n = 2.
First, we nd the Greens function. We have
1
K(x ) = log (x x0 )2 + (y y0 )2 .
2
Let G(x, ) = K(x ) + (x).
Since the problem is Neumann, we need:
G(x, ) = (x ),
G
y ((x, 0), ) = 0.
1
G((x, y), ) = log (x x0 )2 + (y y0 )2 + ((x, y), ),
2
G 1 y y0
((x, y), ) = + y ((x, y), ),
y 2 (x x0 )2 + (y y0 )2
G 1 y0
((x, 0), ) = + y ((x, 0), ) = 0.
y 2 (x x0 )2 + y02
Let
a
((x, y), ) = log (x x0 )2 + (y + y0 )2 . Then,
2
G 1 y0 a y0
((x, 0), ) = 2 + = 0.
y 2
2 (x x0 ) + y0 2 (x x0 )2 + y02
Thus, a = 1.
1 1
G((x, y), ) = log (x x0 )2 + (y y0 )2 + log (x x0 )2 + (y + y0 )2 .
2 2
46
G u
(uG G u ) dx = u G dS
n n
=0
=0
46
Note that for the Dirichlet problem, we would have gotten the - sign instead of + in front of
.
Partial Dierential Equations Igor Yanovsky, 2005 200
Since u
n = u
= f (x), we have
(y)
u (x ) dx = G((x, y), ) f (x) dx,
u() = G((x, y), ) f (x) dx.
For y = 0, we have
' '
1 2 1
G((x, y), ) = 2
log (x x0 ) + y0 + log (x x0 )2 + y02
2 ' 2
1
= 2 log (x x0 )2 + y02
2
1
= log (x x0 )2 + y02 .
2
Thus,
1
u() = log (x x0 )2 + y02 f (x) dx.
2
b) Show that this solution is bounded in if and only if f (x1 ) dx1 = 0.
Proof. For x ,
|x (0)| |x (2)| = |x (1)| |x (3)|,
|x (1) | |x (3)|
|x (0)| = .
|x (2)|
But (0) = , so for n = 2,
1 1 |x (1)| |x (3)|
G(x, ) = log |x | log .
2 2 |x (2)|
G(x, ) = 0, x .
G = (x )
Problem (S96, #3). Construct a Greens function for the following mixed Dirichlet-
Neumann problem in = {x = (x1 , x2 ) R2 : x1 > 0, x2 > 0}:
2u 2u
u = + = f, x ,
x21 x22
ux2 (x1 , 0) = 0, x1 > 0,
u(0, x2) = 0, x2 > 0.
1
Proof. Notation: x = (x, y), = (x0 , y0 ). Since K(x ) = 2 log |x |, n = 2.
1
K(x ) = log (x x0 )2 + (y y0 )2 .
2
Let G(x, ) = K(x ) + (x).
At (0, y), y > 0,
'
1
G (0, y), = log x20 + (y y0 )2 + (0, y) = 0.
2
Also,
1
1 2 2(y y0 )
Gy (x, y), = + wy (x, y)
2 (x x0 )2 + (y y0 )2
1 y y0
= + wy (x, y).
2 (x x0 )2 + (y y0 )2
At (x, 0), x > 0,
1 y0
Gy (x, 0), = + wy (x, 0) = 0.
2 (x x0 )2 + y02
We have
a
((x, y), ) = log (x + x0 )2 + (y y0 )2
2
b
+ log (x x0 )2 + (y + y0 )2
2
c
+ log (x + x0 )2 + (y + y0 )2 .
2
Using boundary conditions, we have
'
1
0 = G((0, y), ) = log x20 + (y y0 )2 + (0, y)
2
' ' ' '
1 a b c
= log x20 + (y y0 )2 + log x20 + (y y0 )2 + log x20 + (y + y0 )2 + log x20 + (y + y0 )2 .
2 2 2 2
Thus, a = 1, c = b. Also,
1 y0
0 = Gy ((x, 0), ) = + wy (x, 0)
2 (x x0 )2 + y02
1 y0 (1) y0 b y0 (b) y0
= 2
2
+ 2
+ .
2
2 (x x0 ) + y0 2
2 (x + x0 ) + y0 2
2 (x x0 ) + y0 2 (x + x0 )2 + y02
Thus, b = 1, and
1 1
G((x, y), ) = 2 2
log (x x0 ) + (y y0 ) + (x) = log (x x0 )2 + (y y0 )2
2 2
Partial Dierential Equations Igor Yanovsky, 2005 203
log (x + x0 )2 + (y y0 )2 2 2 2 2
+ log (x x0 ) + (y + y0 ) log (x + x0 ) + (y + y0 ) .
/ .
|x | a ||
= for |x| = a. |x | = |x |. (17.2)
|x | || a
From (17.2) we conclude that for x (i.e. |x| = a),
2
1 ||
log a |x | if n = 2
K(x ) =
n2 (17.3)
a K(x ) if n 3.
||
Dene for x, :
K(x ) 1 log || |x | if n = 2
2 a
G(x, ) =
K(x ) a n2 K(x ) if n 3.
||
Proof. METHOD :
( aI)u = f ( bI)u = f
u = Ka f u = Kb f
fundamental solution kernel
*af
=K
u =K
u *b f if u L2 ,
( u = f
aI)u = (||2 a) ( u = f
bI)u = (||2 b)
1 1
u = 2 f() =
u f()
( + a) ( 2 + b)
K *a = 1 *b = 1
K
2 + a 2 + b
( aI)( bI)u = f,
2
(a + b) + abI u = f,
1
= 2
u f() new f(),
= K
( + a)( 2 + b)
new = 1 1
1 1 1 b K
a),
K = 2 + 2 = (K
( + a)( 2 + b)
2 ba +b +a ba
1
Knew = (Kb Ka),
ba
1 1
c1 = , c2 = .
ba ba
n = 3 is not relevant (may be used to assume Ka, Kb L2 ).
For 2 , a = 0, b = 1 above, or more explicitly
(2 )u = f,
u = f,
( 4 + 2 )
1
=
u f,
( 4+ 2)
= 1 1 1 1 1 K
0.
K 4 2
= 2 2 = 2 + 2 = K
( + ) ( + 1) +1
Partial Dierential Equations Igor Yanovsky, 2005 206
METHOD :
For u C0 (Rn ) we have:
u(x) = Ka(x y) ( aI) u(y) dy,
Rn
u(x) = Kb(x y) ( bI) u(y) dy.
Rn
Let
u(x) = c1 ( bI) (x), for
u(x) = c2 ( aI) (x), for
Thus,
w = c1 er + c2 er ,
1 er er
K = w(r) = c1 + c2 .
r r r
Suppose v(x) 0 for |x| R and let = BR (0); for small > 0 let
= B (0).
Note: ( I)K(|x|) = 0 in . Consider Greens identity ( = B (0)):
v K(|x|)
v K(|x|)
K(|x|)v vK(|x|) dx = K(|x|) v dS + K(|x|) v dS
n n B(0) n n
=0, since v0 f or xR
We add K(|x|) v dx + v K(|x|) dx to LHS to get:
v K(|x|)
K(|x|)( I)v v ( I)K(|x|) dx = K(|x|) v dS.
B(0) n n
= 0, in
er er
lim K(|x|)( I)v dx = K(|x|)( I)v dx. Since K(r) = c1 + c2 is integrable at x = 0.
0 r r
On B (0), K(|x|) = K(). Thus, 47
v e
v
e 2
v 0, as 0.
K(|x|) dS = K() dS c 1 + c 2 4 max
B(0) n B(0) n
1
K(|x|) 1
v(x) dS = c1 e + c2 e + 2 c1 e + c2 e v(x) dS
B (0) n B (0)
1
1
= c1 e + c2 e
+ 2 c1 e + c2 e v(x) dS
B(0)
1
1
= c1 e + c2 e + 2 c1 e + c2 e v(0) dS
B(0)
1
1
+ c1 e + c2 e
+ 2 c1 e + c2 e [v(x) v(0)] dS
B (0)
1
2 c1 e + c2 e v(0) 42
4(c1 + c2 )v(0) = v(0).
1
Thus, taking c1 = c2 , we have c1 = c2 = 8 , which gives
K(|x|)( I)v dx = lim K(|x|)( I)v dx = v(0),
0
47
In R3 , for |x| = ,
e e
K(|x|) = K() = c1 + c2 .
K(|x|) K()
e e
e e 1 1
= = c1 2 c2 2 = c1 e + c2 e + 2 c1 e + c2 e ,
n r
since n points inwards. n points toward 0 on the sphere |x| = (i.e., n = x/|x|).
Partial Dierential Equations Igor Yanovsky, 2005 208
1 er er 1
that is K(r) = 8 r + r = 4r cosh(r) is the fundamental solution of
( I).
1
By part (a), ab (Ka Kb ) is a fundamental solution of ( aI)( bI).
1 1
Here, the fundamental solution of ( 0I)( 1I) is 1 (K0 K) = 4r +
1 1
4r cosh(r) = 4r 1 cosh(r) .
Partial Dierential Equations Igor Yanovsky, 2005 209
is a solution to ( + k2 )u = f .
1 cos k|x|
Proof. For v C0 (Rn ), we want to show that for K(|x|) = 4 |x| ,
we have ( + k2 )K = , i.e.
K(|x|) ( + k2 )v(x) dx = v(0).
Rn
Suppose v(x) 0 for |x| R and let = BR (0); for small > 0 let
= B (0).
( + k2 )K(|x|) = 0 in . Consider Greens identity ( = B (0)):
v K(|x|)
v K(|x|)
K(|x|)v vK(|x|) dx = K(|x|) v dS + K(|x|) v dS
n n B(0) n n
=0, since v0 f or xR
We add k2 K(|x|) v dx v k2 K(|x|) dx to LHS to get:
2 2 v K(|x|)
K(|x|)( + k )v v ( + k )K(|x|) dx = K(|x|) v dS.
B(0) n n
= 0, in
2 cos kr
lim K(|x|)( + k )v dx = K(|x|)( + k2 )v dx. Since K(r) = is integrable at x = 0.
0 4r
On B (0), K(|x|) = K(). Thus, 48
v cos k
v
K(|x|) dS = K() dS 42
max v 0, as 0.
n
B(0) B(0) n 4
48
In R3 , for |x| = ,
cos k
K(|x|) = K() = .
4
K(|x|) K() 1
k sin k cos k 1
cos k
= = 2
= k sin k + ,
n r 4 4
since n points inwards. n points toward 0 on the sphere |x| = (i.e., n = x/|x|).
Partial Dierential Equations Igor Yanovsky, 2005 210
K(|x|) 1
cos k
v(x) dS = k sin k + v(x) dS
B (0) n B(0) 4
1
cos k
= k sin k + v(x) dS
4 B (0)
1
cos k 1
cos k
= k sin k + v(0) dS k sin k + [v(x) v(0)] dS
4 B (0) 4 B(0)
1
cos k 1
cos k
= k sin k + v(0) 42 k sin k + [v(x) v(0)] 42
4 4
0, (v is continuous)
cos k v(0) v(0).
Thus,
2
K(|x|)( + k )v dx = lim K(|x|)( + k2 )v dx = v(0),
0
1 cos kr
that is, K(r) = 4 r is the fundamental solution of + k2 .
1 ikr
Proof. Denote K(|x|) = 4r e , a fundamental solution. Thus, ( + k2 )K = .
Let x0 be any point and = BR (x0 ); for small > 0 let
= B (x0 ).
( + k2 )K(|x|) = 0 in . Consider Greens identity ( = B (x0 )):
2 2 K u
K u
u ( + k )K K( + k )u dx = u K dS + u K dS .
n n n n
B (x0 )
=0 u(x0 ), as 0
(It can be shown by the method previously used that the integral over B (x0 ) ap-
proaches u(x0 ) as 0.) Taking the limit when 0, we obtain
K u eik|xx0 | eik|xx0 | u
u(x0 ) = u K dS = u dS
n n r 4|x x0 | 4|x x0 | r
eik|xx0 | eik|xx0 | eik|xx0 | u
= u ik iku dS
r 4|x x0 | 4|x x0 | 4|x x0 | r
1
= O( ); (can be shown)
|x|2
1
1
1
1
2
= O O 4R O O 4R2 = 0.
R R2 R R2
Taking the limit when R , we get u(x0 ) = 0.
Partial Dierential Equations Igor Yanovsky, 2005 211
b) Let
1
K(x ) = |x |2 log |x | 1 .
8
We use the method of images to construct the Greens function.
Let G(x, ) = K(x ) + (x). We need G(x, ) = 0 and G n (x, ) = 0 for x .
2 w
Consider w (x) with w (x) = 0 in , w (x) = K(x ) and n (x) = Kn (x )
on . Note, we can nd the Greens function for the upper-half plane, and then
make a conformal map onto the domain.
Partial Dierential Equations Igor Yanovsky, 2005 213
2 w = F (x1 , x2 ).
1 1 2 1
Hint: In polar coordinates, =
r r (r r ) + r 2 2 ; for example, V = 2 (1 + ln(r)).
The fundamental solution V (x) for 2 is the distribution satisfying: 2 V (r) = (r).
1 1 1 1
2 V = (V ) = (1 + log r) = (1 + log r) = r(1 + log r)r r
2 2 2 r
1 1
1 1 1
= r = (1)r = 0 for r = 0.
2 r r r 2 r
Thus, 2 V (r) = (r) V is the fundamental solution.
The approach above is not rigorous. See the next page that shows that
V dened above is the Fundamental Solution of 2 .
The solution of
2 = F (x),
if given by
2 1
(x) = V (x y) (y) dy = |x y|2 log |x y| F (y) dy.
R2 8 R2
Partial Dierential Equations Igor Yanovsky, 2005 214
v
v 1
K(|x|)
dS = K() dS (1 + log )2 max |v| 0, as 0.
n
B(0) B (0) n 2 x
K(|x|) 1 1
v dS = log v(x) dS
B(0) n B(0) 4 8
1
log + 2 max |v| 0, as 0.
4 2 xB(0)
49
Note that for |x| = ,
1 2 1
K(|x|) = K() = log , K = (1 + log ),
8 2
K(|x|) K() 1 1 K K 1
= = log , = = .
n r 4 8 n r 2
Partial Dierential Equations Igor Yanovsky, 2005 215
2
K(|x|) v dx = lim K(|x|)2 v dx = v(0).
0
Partial Dierential Equations Igor Yanovsky, 2005 216
t
R(t) = R0 e 2 .
Partial Dierential Equations Igor Yanovsky, 2005 218
Problem (F01, #3). Let u = u(x, t) solve the following PDE in three spatial di-
mensions
u = 0
for R1 < r < R(t), in which r = |x| is the radial variable, with boundary conditions
u(r = R(t), t) = 0, and u(r = R1 , t) = 1.
In addition assume that R(t) satises
dR u
= (r = R)
dt r
with initial condition R(0) = R0 in which R0 > R1 .
a) Find the solution u(x, t).
b) Find an ODE for the outer radius R(t).
Proof. a) Rewrite the equation in spherical coordinates (n = 3, radial functions):
2 2 1
u = 2
+ u = 2 (r 2 ur )r = 0.
r r r r
(r 2 ur )r = 0,
r 2 ur = c1 ,
c1
ur = ,
r2
c1
u(r, t) = + c2 .
r
Using boundary conditions, we have
c1 c1
u(R(t), t) = + c2 = 0 c2 = ,
R(t) R(t)
c1
u(R1 , t) = + c2 = 1.
R1
This gives
R1 R(t) R1
c1 = , c2 = .
R1 R(t) R1 R(t)
R1 R(t) 1 R1
u(r, t) = + .
R1 R(t) r R1 R(t)
b) We have
R1 R(t) 1 R1
u(r, t) = + ,
R1 R(t) r R1 R(t)
u R1 R(t) 1
= 2,
r R1 R(t) r
dR u R1 R(t) 1 R1
= (r = R) = 2
= (from )
dt r R1 R(t) R(t) (R1 R(t)) R(t)
Thus, an ODE for the outer radius R(t) is
dR R1
dt = (R(t)R1) R(t) ,
R(0) = R0 , R0 > R1 .
Partial Dierential Equations Igor Yanovsky, 2005 219
provided
uv dx = f v dx
for all test functions v H02 (). Prove that for each f L2 (), there exists a unique
weak solution for this problem. Here, H02 () is the closure of all smooth
functions in
which vanish on the boundary and with nite H 2 norm: ||u||22 = (u2xx + u2xy +
u2yy ) dxdy < .
Hint: use Lax-Milgram lemma.
50
We have
2
|a(u, v)|2 = uv dx (u)2 dx (v)2 dx ||u||2H 2() ||v||2H 2() .
0 0
a(v, v) = (v)2 dx ||v||H02() .
1
1
2 2 2 2
|L(v)| = f v dx |f | |v| dx f dx v dx
= ||f ||L2() ||v||L2() ||f ||L2() ||v||H 2() .
0
50
Cauchy-Schwarz Inequality:
1 1
|(u, v)| ||u||||v|| in any norm, for example |uv|dx ( u2 dx) 2 ( v 2 dx) 2 ;
1 1
|a(u, v)| a(u, u) 2 a(v, v) 2 ;
1 1
|v|dx = |v| 1 dx = ( |v|2 dx) 2 ( 12 dx) 2 .
Poincare Inequality:
||v||H 2() C (v)2 dx.
Greens formula:
(u)2 dx = (u2xx + u2yy + 2uxx uyy ) dxdy = (u2xx + u2yy 2uxxy uy ) dxdy = (u2xx + u2yy + 2|uxy |2 ) dxdy ||u||2H 2().
0
Partial Dierential Equations Igor Yanovsky, 2005 223
17.5 Uniqueness
Problem. The solution of the Robin problem
u
+ u = , x
n
for the Laplace equation is unique when > 0 is a constant.
Proof. Let u1 and u2 be two solutions of the Robin problem. Let w = u1 u2 . Then
w = 0 in ,
w
+ w = 0 on .
n
Consider Greens formula:
u
u v dx = v ds v u dx.
n
Setting w = u = v gives
2 w
|w| dx = w ds w w dx .
n
=0
u q(x)u = 0 in .
Setting w = u = v gives
w
|w|2 dx = w ds w w dx.
n
=0, Dirichlet or Neumann
Partial Dierential Equations Igor Yanovsky, 2005 224
2
|w| dx = q(x)w 2 dx .
0 0
Thus, w 0, and u1 u2 . Hence, the solution to the Dirichlet and Neumann problems
are unique.
2 u = f in D,
u = u = 0 on D
is unique.
u = u2 and u| = 0.
Prove that u 0.
b) What can you say about u(x) when the condition u(x) 0 in is dropped?
Partial Dierential Equations Igor Yanovsky, 2005 225
b) If we dont know that u(x) 0, then u can not be nonnegative on the entire
domain . That is, u(x) < 0, on some (or all) parts of . If u is nonnegative on ,
then u 0.
Partial Dierential Equations Igor Yanovsky, 2005 226
Note that
n
u
k u dx = u u dx = nu2 ds u u dx,
k=1 xk
=0
and thus,
u u dx = 0.
Partial Dierential Equations Igor Yanovsky, 2005 227
u = f in D,
u
u(x1 , 0) = (0, x2) = 0
x1
can have only one bounded solution.
b) Find an explicit Greens function for this boundary value problem.
Setting w = u = v gives
2 w
|w| dx = w ds w w dx,
n
D
D
D
w w w
|w|2 dx = w ds + w ds + w ds w w dx
D Rx1 n Rx2 n |x|>R n D
w w w
= w(x1 , 0) ds + w(0, x2) ds + w
ds w w dx,
Rx1 x2 Rx2 x1 |x|>R n
=0 =0
D
=0
=0
|w|2 dx = 0 |w|2 = 0 w = const.
D
Since w(x1 , 0) = 0 w 0. Thus, u1 = u2 .
Problem (F98, #4). In two dimensions x = (x, y), dene the set a as
a = +
in which
+ = {|x x0 | a} {x 0}
= {|x + x0 | a} {x 0} = +
and x0 = (1, 0). Note that a consists of two components when 0 < a < 1
and a single component when a > 1. Consider the Neumann problem
2 u = f, x a
u/n = 0, x a
in which
f (x) dx = 1
+
f (x) dx = 1
a) Show that this problem has a solution for 1 < a, but not for 0 < a < 1.
(You do not need to construct the solution, only demonstrate solveability.)
= f (x) dx + f (x) dx = 1 1 = 0.
+
Thus, the problem has a solution for 1 < a.
Thus,
1
max |u| .
a |L|
As a 1 (L 0) maxa |u| .
Partial Dierential Equations Igor Yanovsky, 2005 230
u + a(x)u = f (x) in D,
u=0 on D.
a) Assuming that |a(x)| is small enough, prove the uniqueness of the classical solution.
b) Prove the existence of the solution in the Sobolev space H 1 (D) assuming that f
L2 (D).
Note: Use Poincare inequality.
Proof. a) By Poincare Inequality, for any u C01 (D), we have ||u||22 C||u||22.
Consider two solutions of the Dirichlet problem above. Let w = u1 u2 . Then, w
satises
w + a(x)w = 0 in D,
w=0 on D.
ww + a(x)w 2 = 0,
ww dx + a(x)w 2 dx = 0,
|w|2 dx + a(x)w 2 dx = 0,
1
a(x)w 2 dx = |w|2 dx w 2 dx, (by Poincare inequality)
C
1
a(x)w 2 dx w 2 dx 0,
C
1
|a(x)| w 2 dx w 2 dx 0,
C
1
|a(x)| w 2 dx 0.
C
If |a(x)| < C1 w 0.
b) Consider
F (v, u) = (vu + a(x)vu) dx = vf (x) dx = F (v).
in which r 2 = x2 + y 2 .
2
a) Show that = er satises L = 0.
b) Use this to show that the equation
Lu = f in
u
= on
n
has a solution only if
f dx = ds(x).
2 2 +y2 2 2
b) We have = er = ex = ex ey . From part (a),
L = 0,
2 2 2 2 2
= n = (x , y ) n = (2xex ey , 2yex ey ) n = 2er (x, y) (y, x) = 0.
n
51
Consider two equations:
Lu = u 4(r 2 + 1)u,
L = 4(r 2 + 1).
Multiply the rst equation by and the second by u and subtract the two equations:
Lu = u 4(r 2 + 1)u,
uL = u 4(r 2 + 1)u,
Lu uL = u u.
Then, we start from the LHS of the equality we need to prove and end up with RHS:
f dx = Lu dx = (Lu uL) dx = (u u) dx
u u
= ( u ) ds = ds = ds.
n n n
51
The only shortcoming in the above proof is that we assume
n = (y, x), without giving an expla-
nation why it is so.
Partial Dierential Equations Igor Yanovsky, 2005 232
We have
(Lu) v dx = a (x)D u v dx = a (x) v D u dx
||m ||m
||
= (1)
D (a(x) v) u dx = (1)||D (a(x) v) u dx
||m ||m
L (v)
= L (v) u dx,
The operator
L (v) = (1)||D (a(x) v)
||m
52
L = L (Lu|v) = (u|L v) = (u|Lv).
Partial Dierential Equations Igor Yanovsky, 2005 233
Proof. a) We have
Lu = u = 0
u
=0 on x = 0
x
u u
=0 on x = y.
x y
The operator L is self-adjoint if:
(u Lv v Lu) dx = 0.
v u
(u Lv v Lu) dx = (uv vu) dx = u v ds
n n
v u
v u
= u v ds + u v ds
x=0 n n x=y n n
= u (v n) v (u n) ds + u (v n) v (u n) ds
x=0 x=y
= u (vx , vy ) (1, 0) v (ux , uy ) (1, 0) ds
x=0
+ u (vx , vy ) (1/ 2, 1/ 2) v (ux, uy ) (1/ 2, 1/ 2) ds
x=y
= u (0, vy ) (1, 0) v (0, uy ) (1, 0) ds
x=0
= 0
+ u (vy , vy ) (1/ 2, 1/ 2) v (uy , uy ) (1/ 2, 1/ 2) ds
x=y
uv
vu
= y ( 1) y ( 1) ds = 0.
x=y 2 2
need
Partial Dierential Equations Igor Yanovsky, 2005 234
Problem (F99, #1). Suppose that u = 0 in the weak sense in Rn and that there
is a constant C such that
|u(y)| dy < C, x Rn .
{|xy|<1}
If |xy|<1 |u(y)| dy < C x Rn , we have |u(x)| < C in Rn .
Since u is harmonic and bounded in Rn , u is constant by Liouvilles theorem. 53
53
Liouvilles Theorem: A bounded harmonic function dened on Rn is a constant.
Partial Dierential Equations Igor Yanovsky, 2005 236
Problem (S01, #1). For bodies (bounded regions B in R3 ) which are not perfectly
conducting one considers the boundary value problem
u
3
0 = (x)u = (x)
xj xj
j=1
u = f on B.
The function (x) is the local conductivity of B and u is the voltage. We dene
operator (f ) mapping the boundary data f to the current density at the boundary by
u
(f ) = (x) ,
n
and /n is the inward normal derivative (this formula denes the current density).
a) Show that is a symmetric operator, i.e. prove
g(f ) dS = f (g) dS.
B B
b) Use the positivity of (x) > 0 to show that is negative as an operator, i.e., prove
f (f ) dS 0.
B
Proof. a) Let
(x)u = 0 on , (x)v = 0 on ,
u=f on . v=g on .
u v
(f ) = (x) , (g) = (x) .
n n
Since /n is inward normal derivative, Greens formula is:
u
v
(x) dS v (x)u dx = v (x)u dx.
n
=g
We have
u
g(f ) dS = g(x) dS = v (x)u dx v (x)u dx
n
=0
v
= u(x) dS + u (x)v dx
n
=0
v
= f (x) dS = f (g) dS.
n
b) We have (x) > 0.
u
f (f ) dS = u(x) dS = u (x)u dx (x)u u dx
n
=0
= (x)|u|2 0.
0
Partial Dierential Equations Igor Yanovsky, 2005 237
Problem (S01, #4). The Poincare Inequality states that for any bounded domain
in Rn there is a constant C such that
2
|u| dx C |u|2 dx
= {(x, y) : 0 x a, 0 y b}.
with C = 1/1 .
V = 0 in R3 /B
where B is the boundary of B and n is the outer unit normal to it (i.e. the normal
pointing toward innity).
c) Suppose that B B. Show that C(B ) C(B).
Setting W = u = v gives
2 W
|W | dx = W
ds W W dx = 0.
B B n B
=0 =0
b & c) For (b)&(c), see the solutions from Ralstons homework (a few pages
down).
Partial Dierential Equations Igor Yanovsky, 2005 240
Problem (W03, #2). Let L be the second order dierential operator L = a(x)
in which x = (x1 , x2 , x3 ) is in the three-dimensional cube C = {0 < xi < 1, i = 1, 2, 3}.
Suppose that a > 0 in C. Consider the eigenvalue problem
Lu = u for xC
u=0 for x C.
Proof. a) We have
u a(x)u = u.
Multiply the equation by u and integrate:
uu a(x)u2 = u2 ,
uu dx a(x)u dx = u2 dx,
2
u 2
u ds |u| dx a(x)u dx = u2 dx,
2
n
=0
(|u|2 + a(x)u2 ) dx
= < 0.
2
u dx
u a(x)u = u. (17.5)
v a(x)v = v. (17.6)
Multiply (17.5) by v and (17.6) by u and subtract equations from each other
vu a(x)uv = uv,
uv a(x)uv = uv.
vu uv = ( )uv.
Integrating over gives
vu uv dx = ( ) uv dx,
u v
v u dx = ( ) uv dx.
n n
=0
Since = , u and v are orthogonal on .
Partial Dierential Equations Igor Yanovsky, 2005 241
a) Show that 2 is self-adjoint on |x| < 1 with the following boundary conditions on
|x| = 1:
u = 0,
u = 0.
Proof. a) We have
Lu = 2 u = 0
u = 0 on |x| = 1
u = 0 on |x| = 1.
The operator L is self-adjoint if:
(u Lv v Lu) dx = 0.
(u Lv v Lu) dx = (u 2 v v 2 u) dx
v u
= u ds u (v) dx v ds + v (u) dx
n n
=0 =0
u v
= v ds + uv dx + u ds vu dx = 0.
n n
=0 =0
Partial Dierential Equations Igor Yanovsky, 2005 243
Show that
d r
S(r) = V (r).
dr 3
Hint: Rewrite S(r) as an integral over the unit sphere before dierentiating; i.e.,
1
S(r) = (4) u(rx ) dx.
|x |=1
1 1 2 1
S(r) = u(x) dSr = u(rx ) r dS1 = u(rx ) dS1 .
4r 2 |x|=r 4r 2 |x |=1 4 |x |=1
dS 1 u 1 u 1 u
= (rx ) dS1 = (rx ) dS1 = 2
(x) dSr
dr 4 |x |=1 r 4 |x|=1 n 4r |x|=r n
1
= u dx.
4r 2 |x|r
where we have used Greens identity in the last equality. Also
r 1
V (r) = u dx.
3 4r 2 |x|r
Volume integrals: = r in Rn :
h(x + ) d = h(x + r) rn d.
| |<r ||<1
Partial Dierential Equations Igor Yanovsky, 2005 244
Partial Dierential Equations Igor Yanovsky, 2005 245
55
Proof. a) We can express the integral in Spherical Coordinates:
R 2
f (x) dx = f (, , r) r 2 sin d d dr.
|x|R 0 0 0
R 2
d d
f (x) dx = f (, , r) r 2 sin d d dr = ???
dr |x|R dr 0 0 0
2
= f (, , r) R2 sin d d
0 0
= f (x) dS.
|x|=R
55
Dierential Volume in spherical coordinates:
d3 = 2 sin d d d.
dS = 2 sin d d.
Partial Dierential Equations Igor Yanovsky, 2005 246
b&c) We have
d 1 d 1 2 1 d
f (x) dS = f (rx ) r dS1 = f (rx ) dS1
dr 4r 2 |x|=r dr 4r 2 |x |=1 4 dr |x |=1
1 f 1 f
= (rx ) dS1 = (rx ) dS1
4 |x |=1 r 4 |x |=1 n
1 f 1
= (x) dS = f n dS
4r 2 |x|=r n 4r 2 |x|=r
1
= f dx.
4r 2 |x|r
Greens formula was used in the last equality.
Alternatively,
2
d 1 d 1 2
f (x) dS = f (, , r) r sin d d
dr 4r 2 |x|=r dr 4r 2 0 0
2
d 1
= f (, , r) sin d d
dr 4 0 0
2
1 f
= (, , r) sin d d
4 0 0 r
2
1
= f n sin d d
4 0 0
2
1
= f n r 2 sin d d
4r 2 0 0
1
= f n dS
4r 2 |x|=r
1
= f dx.
4r 2 |x|=r
d) Since f is biharmonic (i.e. f is harmonic), f has a mean value property. From
(c), we have 56
d 1 1 r 1
f (x) ds(x) = f (x) dx = f (x) dx
dr 4r 2 |x|=r 4r 2 |x|r 3 43 r 3 |x|r
r
= f (0).
3
1 r2
f (x) ds(x) = f (0) + f (0).
4r 2 |x|=r 6
56
Note that part (a) was not used. We use exactly the same derivation as we did in S95 #4.
Partial Dierential Equations Igor Yanovsky, 2005 247
57
57
Check with someone about the last result.
Partial Dierential Equations Igor Yanovsky, 2005 249
a) u > 0 in R3 /;
b) if (x) is a smooth function such that (x) = 1 for |x| > R and (x) = 0 near ,
then for |x| > R,
1 ((u))(y)
u(x) = dy.
4 R3 / |x y|
b) For n = 3,
1 1 1
K(|x y|) = |x y|2n = .
(2 n)n 4 |x y|
Since (x) = 1 for |x| > R, then for x / BR , we have (u) = u = 0. Thus,
1 ((u))(y)
dy
4 R3 / |x y|
1 ((u))(y)
= dy
4 BR / |x y|
1
1 1 1
= y y (u) dy u dSy
4 BR / |x y| 4 (BR/) n |x y|
1
1 1
1 1 1
= u dy + u dSy u dSy
4 BR / |x y| 4 (BR/) n |x y| 4 (BR/) n |x y|
1 1 u
= ??? = u(x) u dSy dSy
4R2 B 4R B n
0, as R 0, as R
= u(x).
Ralston Hw. a) Suppose that u is a smooth function on R3 and u = 0 for |x| > R.
1
If limx u(x) = 0, show that you can write u as a convolution of u with the 4|x|
and prove that limx |x|u(x) = 0 exists.
b) The conductor potential for R3 is the solution to the Dirichlet problem v =
0. The limit in part (a) is called the capacity of . Show that if 1 2 , then the
capacity of 2 is greater or equal the capacity of 1 .
Proof. a) If we dene
1 u(y)
v(x) = dy,
4 R3 |x y|
then (u v) = 0 in all R3 , and, since v(x) 0 as |x| , we have lim|x| (u(x)
v(x)) = 0. Thus, u v must be bounded, and Liouvilles theorem implies that it is
identically zero. Since we now have
1 |x| u(y)
|x|u(x) = dy,
4 R3 |x y|
and |x|/|x y| converges uniformly to 1 on {|y| R}, it follows that
1
lim |x|u(x) = u(y) dy.
|x| 4 R3
b) Note that part (a) implies that the limit lim|x| |x|v(x) exists, because we can
apply (a) to u(x) = (x)v(x), where is smooth and vanishes on , but (x) = 1 for
|x| > R.
Let v1 be the conductor potential for 1 and v2 for 2 . Since vi as |x| and
vi = 1 on i , the max principle says that 1 > vi (x) > 0 for x R3 i . Consider
v2 v1 . Since 1 2 , this is dened in R3 2 , positive on 2 , and has limit 0 as
|x| . Thus, it must be positive in R3 2 . Thus, lim|x| |x|(v2 v1 ) 0.
DR = {(x, y) : (x x0 )2 + (y y0 )2 R2 }
then
2
1
u(x0 , y0 ) u(x0 + R cos , y0 + R sin ) d.
2 0
Proof. Let
2
1
M (x0 , R) = u(x0 + R cos , y0 + R sin ) d,
2 0
w(r, ) = u(x0 + R cos , y0 + R sin ).
Dierentiate M (x0 , R) with respect to R:
2
d 1
M (x0 , R) = wr (R, )R d,
dr 2R 0
58
See McOwen, Sec.4.3, p.131, #1.
Partial Dierential Equations Igor Yanovsky, 2005 251
59
59
See ChiuYens solutions and Sung Has solutions (in two places). Nicks solutions, as started above,
have a very simplistic approach.
Partial Dierential Equations Igor Yanovsky, 2005 252
Proof. a) If u(x) is not less than sup u for all x , then the set
K = {x : u(x) = sup u}
is nonempty. This set is closed because u is continuous. We will show it is also open.
This implies that K = because is connected. Thus u is constant on .
Let x0 K. Since is open, > 0, s.t. B (x0 ) = {x Rn : |x x0 | } . Let
sup u = M . By the mean value property, for 0 r
1 1
M = u(x0 ) = u(x0 + r)dS , and 0 = (M u(x0 + r))dS .
A(S n1 ) ||=1 A(S n1 ) ||=1
Sinse M u(x0 + r) is a continuous nonnegative function on , this implies M u(x0 +
r) = 0 for all S n1 . Thus u = 0 on B (x0 ).
b) Since u v has the mean value property in the open interior of Br (), by part
a) it satises the maximum principle. Since it is continuous on Br (), its supremum
over the interior of Br () is its maximum on Br (), and this maximum is assumed at a
point x0 in Br (). If x0 in the interior of Br (), then u v is constant ant the constant
must be zero, since this is the value of u v on the boundary. If x0 is on the boundary,
then u v must be nonpositive in the interior of Br ().
Applying the same argument to v u, one nds that it is either identically zero or
nonpositive in the interior of Br (). Thus, u v 0 on Br ().
Yes, it does follow that u harmonic in . Take v in the preceding to be the harmonic
function in the interior of Br () which agrees with u on the boundary. Since u = v on
Br (), u is harmonic in the interior of Br (). Since is open we can do this for every
. Thus u is harmonic in .
Partial Dierential Equations Igor Yanovsky, 2005 253
Ralston Hw. Assume is a bounded open set in Rn and the Greens function, G(x, y),
for exists. Use the strong maximum principle, i.e. either u(x) < sup u for all x ,
or u is constant, to prove that G(x, y) < 0 for x, y , x = y.
Proof. G(x, y) = K(x, y) + (x, y). For each x , f (y) = (x, y) is continuous on ,
thus, bounded. So |(x, y)| Mx for all y . K(x y) as y x. Thus,
given Mx , there is > 0, such that K(x y) < Mx when |x y| = r and 0 < r .
So for 0 < r the Greens function with x xed satises, G(x, y) is harmonic on
Br (x), and G(x, y) 0 on the boundary of Br (x). Since we can choose r as
small as we wish, we get G(x, y) < 0 for y {x}.
Problem (W03, #6). Assume that u is a harmonic function in the half ball
D = {(x, y, z) : x2 + y 2 + z 2 < 1, z 0} which is continuously dierentiable, and satis-
es u(x, y, 0) = 0. Show that u can be extended to be a harmonic function in the whole
ball. If you propose and explicit extension for u, explain why the extension is harmonic.
Ralston Hw. Show that a bounded solution to the Dirichlet problem in a half
space is unique. (Note that one can show that a bounded solution exists for any
given bounded continuous Dirichlet data by using the Poisson kernel for the half space.)
Proof. We have to show that a function, u, which is harmonic in the half-space, con-
tinuous, equal to 0 when xn = 0, and bounded, must be identically 0. We can extend
u to all of n-space by dening
u(x , xn ) = u(x , xn )
for xn < 0. This extends u to a bounded harmonic function on all of n-space (by the
problem above). Liouvilles theorem says u must be constant, and since u(x , 0) = 0,
the constant is 0. So the original u must be identically 0.
Proof. The condition lim|x|0 |x|u(x) = 0 is necessary, because harmonic functions are
continuous.
To prove the converse, let v be the function which is continuous on {|x| a/2},
harmonic on {|x| < a/2}, and equals u on {|x| = a/2}. One can construct v using the
Poisson kernel. Since v is continuous, it is bounded, and we can assume that |v| M .
Since lim|x|0 |x|u(x) = 0, given > 0, we can choose , 0 < < a/2 such that
< |x|u(x) < when |x| < . Note that u, v 2/|x|, and v + 2/|x| are harmonic
Partial Dierential Equations Igor Yanovsky, 2005 254
on {0 < |x| < a/2}. Choose b, 0 < b < min(, a/2), so that /b > M . Then on both
{|x| = a/2} and {|x| = b} we have v 2/|x| < u(x) < v + 2/|x|. Thus, by
max principle these inequalities hold on {b |x| a/2}. Pick x with 0 < |x| a/2.
u(x) = v(x). v is the extension of u on {|x| < a/2}, and u is extended on {|x| < a}.
Partial Dierential Equations Igor Yanovsky, 2005 255
60
0
In calculations, we use:
ey dy = 0
ey dy, and g(y) = g(y).
Partial Dierential Equations Igor Yanovsky, 2005 256
61
Even extensions are always continuous. Not true for odd extensions. g odd is continuous if g(0) =
0.
62
0
In calculations, we use:
ey dy = 0
ey dy, and g(y) = g(y).
Partial Dierential Equations Igor Yanovsky, 2005 257
ft = fxx t0
f (t = 0, x) = f0 (x)
sup |f (x, t)| sup |f0 (x)| sup |fx (x, t)| sup |f0x (x)|.
x x x x
1 y2 y dy
sup |f0 (x)| e 4t dy z = , dz =
x 4t R 4t 4t
1 2
sup |f0 (x)| ez 4t dz
x 4t R
1 2
= sup |f0 (x)| ez dz = sup |f0 (x)|.
x R x
=
63
The Greens function for the heat equation on an innite domain; derived in R. Haberman using
the Fourier transform.
Partial Dierential Equations Igor Yanovsky, 2005 258
1 2(x y) (xy)2 1 d (xy)2
ux (x, t) = e 4t f0 (y) dy = e 4t f0 (y) dy
4t R 4t 4t R dy
1 (xy)2
4t
1 (xy)2
= e f0 (y) + e 4t f0y (y) dy,
4t 4t R
=0
2
2
1 (xy) 1
sup |u(x, t)| sup |f0x(x)| e 4t dy = sup |f0x(x)| ez 4t dz
x 4t x R 4t x R
= sup |f0x(x)|.
x
These inequalities are equalities when f0 (x) and f0x (x) are constants, respectively.
Partial Dierential Equations Igor Yanovsky, 2005 259
Problem (S01, #5). a) Show that the solution of the heat equation
ut = uxx , < x <
with square-integrable initial data u(x, 0) = f (x), decays in time, and there is a constant
independent of f and t such that for all t > 0
3
1
2
max |ux(x, t)| t 4 |f (x)|2 dx .
x x
b) Consider the solution of the transport equation t + ux = 0 with square-integrable
initial data (x, 0) = 0 (x) and the velocity u from part (a). Show that (x, t) remains
square-integrable for all nite time
1
|(x, t)|2 dx eCt 4 |0(x)|2 dx,
R R
where C does not depend on 0 .
Proof. a) The solution to the one-dimensional homogeneous heat equation is
1 (xy)2
u(x, t) = e 4t f (y) dy.
4t R
Take the derivative with respect to x, we get 64
1 2(x y) (xy)2 1 (xy)2
ux (x, t) = e 4t f (y) dy = 3 (x y)e 4t f (y) dy.
4t R 4t 4t 2 R
2
1 (xy)
|ux (x, t)| 3 (x y)e 4t f (y) dy (Cauchy-Schwarz)
4t 2 R
1
(xy)2
1
xy dy
(x y)2 e 2t dy ||f ||L2(R)
2
3 z = , dz =
4t 2 R 2t 2t
1
2 3
z2
12
= 3 z (2t) 2e dz ||f ||L2(R)
4t 2 R
(2t) 4
1
3
2 z2 2
= 3 z e dz ||f ||L2(R)
4t 2 R
M <
34 1 3
= Ct M ||f ||L2(R) = t 4 ||f ||L2(R) .
2
b) Note:
1
(xy)
2
1
(xy)2
1
e 2t dy ||f ||L2(R)
2
max |u| = max e 4t f (y) dy
x x 4t R 4t R
1 1
ez2 2t dz 2 ||f ||L2(R) xy dy
z = , dz =
4t R 2t 2t
1
1
(2t) 4 2 1
ez dz ||f ||L2(R) = Ct 4 ||f ||L2(R).
2
= 1 1
2 2 t 2 R
=
65
64
1 1
Cauchy-Schwarz: |(u, v)| ||u||||v|| in any norm, for example |uv|dx ( u2 dx) 2 ( v 2 dx) 2
65
See Yanas and Alans solutions.
Partial Dierential Equations Igor Yanovsky, 2005 260
M z2 z dz
|v(y, t)| e 4t dz s = , ds =
4t R 4t 4t
M 2 M 2
= es 4t ds = es ds = M.
4t R R
Integral in converges uniformly lim = lim. For = (a):
1 (yz)2 1 z2
4t
v(y, t) = e (z) dz = e 4t (y z) dz
4t 4t
1 2
= es (y s 4t) 4t ds
4t
1 2
= es (y s 4t) ds.
Partial Dierential Equations Igor Yanovsky, 2005 261
0
1 s2 1 2
lim v(y, t) = e lim (y s 4t) ds + es lim (y s 4t) ds
t+ 0 t+ t+
0
1 2 1 2 1 1
= es c ds + es b ds = c + b
0 2 2
c+b
= .
2
Partial Dierential Equations Igor Yanovsky, 2005 262
Problem. Consider
ut = kuxx + Q, 0<x<1
u(0, t) = 0,
u(1, t) = 1.
Show that such solutions, with initial data g L2 (Rn ), are unique, even when c is
negative.
Proof. McOwen. Consider v(x, t) = ect u(x, t). The transformed problem is
vt = v in Rn (0, )
(18.2)
v(x, 0) = g(x) on Rn .
Since bounded solutions of (18.3) are unique, and since w is a nontrivial solution, w is
unbounded. Thus, v is unbounded, and therefore, the bounded solution v is unique.
Partial Dierential Equations Igor Yanovsky, 2005 264
ut = u u3 in R2 , t 0
u(x, y, 0) = 0 in
u(x, y, t) = 0 in , t 0.
ut u = 0
a) For a xed domain , show M is a constant in time if the boundary conditions are
u/n = 0.
b) Suppose that = (t) is evolving in time, with a boundary that moves at velocity
v, which may vary along the boundary. Find a modied boundary condition (in terms
of local quantities only) for u, so that M is constant.
Hint: You may use the fact that
d
f (x, t) dx = ft (x, t) dx + n v f (x, t) dl,
dt (t) (t) (t)
Proof. a) We have
ut u = 0, on
u
n = 0, on .
d
We want to show that dt M (t) = 0. We have 66
d d u
M (t) = u(x, t) dx = ut dx = u dx = ds = 0.
dt dt n
d
b) We need dt M (t) = 0.
d d
0 = M (t) = u(x, t) dx = ut dx + n v u ds
dt dt
(t) (t) (t)
u
= u dx + n v u ds = ds + n v u ds
(t) (t) (t) n (t)
= u n ds + n v u ds = n (u + vu) ds.
(t) (t) (t)
Thus, we need:
n (u + vu) ds = 0, on .
66
The last equality below is obtained from the Greens formula:
u
u dx = ds.
n
Partial Dierential Equations Igor Yanovsky, 2005 266
Problem (S95, #3). Write down an explicit formula for a function u(x, t) solving
ut + b u + cu = u in Rn (0, )
(18.4)
u(x, 0) = f (x) on Rn .
Proof. Consider
u(x, t) = ex+tv(x, t).
Thus,
2 2
1 |xy|2
b
x( |b|4 +c)t b
x( |b|4 +c)t b
u(x, t) = e 2 v(x, t) = n e2 e 4t e 2 y f (y) dy.
(4t) 2 Rn
Partial Dierential Equations Igor Yanovsky, 2005 267
hT = hXX .
satises
ut + xu = uxx .
Plugging these into , most of the terms cancel out. The remaining two terms cancel
because KT = KXX .
Given that g(x) is bounded and continuous on x , we establish that 67
lim K(x y, t) g(y) dy = g(x).
t0
Furthermore, if |x x0 |
2 and |y x0 | , then
1
|y x0 | |y x| + |y x| + |y x0 |.
2 2
1
Thus, |y x| 2 |y x0 |. Consequently,
= + 2||g||L K(x y, t) dy
RB (x0 )
C |xy|2
+ e 4t dy
t RB (x0 )
C |yx0 |2
+ e 16t dy
t RB (x0 )
C r2
= + e 16t r dr + 0 as t 0+ .
t
Hence, if |x x0 | <
2 and t > 0 is small enough, |u(x, t) g(x0)| < 2.
67
Evans, p. 47, Theorem 1 (c).
Partial Dierential Equations Igor Yanovsky, 2005 270
T 2T T
4t = + 2u . (18.7)
t u2 u
Solutions of (18.7) that depend on u alone are called similarity solutions. 68
T T u T
= + ,
t u t
T T u
= ,
x u x
2T
T
T u
2 T u u T 2 u 2 T
u 2
= = = + = .
x2 x x x u x u2 x x x2
u u2 x
=0
68
This is only the part of the qual problem.
Partial Dierential Equations Igor Yanovsky, 2005 271
Problem (W02, #8). a) Find an explicit solution of the following Cauchy problem
2 2
u
t2
xu2 = f (t, x),
(19.2)
u(0, x) = 0, u x (0, x) = 0.
b) Use part (a) to prove the uniqueness of the solution of the Cauchy problem
2 2
u
t2
xu2 + q(t, x)u = 0,
(19.3)
u(0, x) = 0, u x (0, x) = 0.
Here f (t, x) and q(t, x) are continuous functions.
u|1 = , u|2 = ,
where 1 , 2 are two adjacent sides D. Here a(x, t), and are continuous functions.
Prove the uniqueness of the solution of this Goursat problem.
For small enough , the operator K is a contraction. Thus, there exists a unique xed
point of K, and f = Kf , where f is the unique solution.
Partial Dierential Equations Igor Yanovsky, 2005 273
is given by
t
u(x, t) = y, t) g(y) dy +
K(x y, t s) f (y, s) dy ds
K(x
Rn 0 Rn
where
|x|2
1
n e
4t for t > 0,
K(x, t) = (4t) 2
0 for t 0.
Problem (S97, #3). a) Let Q(x) 0 such that x= Q(x) dx = 1,
and dene Q = 1 Q( x ). Show that (here denotes convolution)
In particular, let Qt (x) denote the heat kernel (at time t), then
c) Fix t > 0. Let {un (x, t)}, n = 1, 2, . . . the xed point iterations for the solution of
(20.2)
t
un+1 (x, t) = Qt (x) f (x) + Qts (x) u2n (x, s) ds. (20.3)
0
Let Kn (t) = sup0mn ||um (x, t)||L . Using (a) and (b) show that
t
||un+1 (x, t) un (x, t)||L 2 sup Kn ( ) ||un (x, s) un1 (x, s)||L ds.
0 t 0
Conclude that the xed point iterations in (20.3) converge if t is suciently small.
Proof. a) We have
||Q(x) w(x)||L =
Q (x y)w(y) dy Q (x y)w(y) dy
1
x y
||w|| Q (x y) dy = ||w|| Q dy
1
y
y dy
= ||w|| Q dy z = , dz =
= ||w|| Q(z) dz = ||w(x)||.
Partial Dierential Equations Igor Yanovsky, 2005 275
x 2
Qt (x) = 1 e 4t
, the heat kernel. We have 69
4t
||Qt(x) w1 (x) Qt (x) w2 (x)||L = Qt (x y)w1 (y) dy Qt (x y)w2 (y) dy
1 (xy)2 (xy)2
= e 4t w1 (y) dy e 4t w2 (y) dy
4t
2
1 (xy)
4t
e w1 (y) w2 (y) dy
4t
1 (xy)2
w1 (y) w2 (y) e 4t dy
4t
xy dy 1 2
z = , dz = = w1 (y) w2 (y) ez 4t dz
4t 4t 4t
1
=
w1 (y) w2 (y) 2
ez dz
= w1 (y) w2 (y) .
69
Note:
1 (xy)2 1 2 1 2
Qt (x) dx = e 4t dy = ez 4t dz = ez dz = 1.
4t 4t
Partial Dierential Equations Igor Yanovsky, 2005 276
b) Consider
ut = uxx + u2 ,
u(x, 0) = f (x).
We will show that the solution of this equation satises
t
u(x, t) = Qt (x) f (x) + Qts (x) u2 (x, s) ds.
0
t t
Qts (x) u2 (x, s) ds = Qts (x y) u2 (y, s) dy ds
0 0 R
t
= Qts (x y) us (y, s) uyy (y, s) dy ds
0 R
t
d d
= Qts (x y)u(y, s) Qts (x y) u(y, s) Qts (x y)uyy (y, s) dy ds
ds ds
0 R
= Q0 (x y)u(y, t) dy Qt(x y)u(y, 0) dy
R R
t
d d2
Qts (x y) u(y, s) + 2 Qts (x y)u(y, s) dy ds
0 R ds dy
= 0, since Qt satisf ies heat equation
= u(x, t) Qt(x y)f (y) dy Note: lim Q(x, t) = 0 (x) = (x).
R t0+
= u(x, t) Qt(x) f (x). lim R Q(x y, t)v(y) dy = v(0).
t0+
c) Let t
un+1 (x, t) = Qt (x) f (x) + Qts (x) u2n (x, s) ds.
0
t
2
||un+1 (x, t) un (x, t)||L = Qts (x) un (x, s) un1 (x, s) ds
2
0
t
2
Qts (x) u (x, s) u2 (x, s) ds
n n1
0
t
u2 (x, s) u2
n n1 (x, s) ds
0
(a)
t
un (x, s) un1 (x, s) un (x, s) + un1 (x, s) ds
0
t
sup un (x, s) + un1 (x, s) un (x, s) un1 (x, s) ds
0 t 0
t
2 sup Kn ( ) ||un (x, s) un1 (x, s)||L ds.
0 t 0
Also, ||un+1 (x, t) un (x, t)||L 2t sup Kn ( ) ||un (x, s) un1 (x, s)||L .
0 t
Partial Dierential Equations Igor Yanovsky, 2005 277
Proof. The space of continuous bounded functions forms a complete metric space so
the contraction mapping
principle
applies.
First, let v(x, t) = w x2 , t , then
ut = uxx + f (u, w)
wt = wxx + g(u, w).
These initial value problems have the following solutions (K is the heat kernel):
t
u(x, t) =
K(x y, t) u0 (y) dy + y, t s) f (u, w) dy ds,
K(x
R n 0 R n
t
w(x, t) = y, t) w0(y) dy +
K(x y, t s) g(u, w) dy ds.
K(x
Rn 0 Rn
By the Lipshitz conditions,
|f (u, w)| M1 ||u||,
|g(u, w)| M2 ||w||.
Now we can show the mappings, as dened below, are contractions:
t
T1 u = K(x y, t) u0 (y) dy +
y, t s) f (u, w) dy ds,
K(x
R n 0 R n
t
T2 w = y, t) w0(y) dy +
K(x y, t s) g(u, w) dy ds.
K(x
Rn 0 Rn
t
|T1 (un ) T1 (un+1 )| y, t s) f (un , w) f (un+1 , w) dy ds
K(x
0 Rn
t
M1 K(x
y, t s) un un+1 dy ds
0 Rn
t
M1
sup un un+1 y, t s)dy ds
K(x
0 x Rn
t
M1 sup un un+1 ds M1 t sup un un+1
0 x x
< sup un un+1 for small t.
x
We used the Lipshitz condition and R K(x y, t s) dy = 1.
Thus, for small t, T1 is a contraction, and has a unique xed point. Thus, the solution
is dened as u = T1 u.
Similarly, T2 is a contraction and has a unique xed point. The solution is dened as
w = T2 w.
Partial Dierential Equations Igor Yanovsky, 2005 279
= {(x, t) U : x or t = 0}.
Problem (S98, #7). Prove that any smooth solution, u(x, y, t) in the unit box
= {(x, y) | 1 x, y 1}, of the following equation
Now
max u = max (v + t) max v + T max{ max v(1, 1, t), max f (x, y)} + T
[0,T ] [0,T ] [0,T ] 0tT (x,y)
max{ max u(1, 1, t), max f (x, y)} + T.
0tT (x,y)
with a > 0, b > 0, e > 0, for (x, y) , with a bounded open set in R2 .
a) Show that u cannot have a positive maximum or a negative minimum in the in-
terior of .
b) Use this to show that the only function u satisfying Lu = 0 in , u = 0 on
and u continuous on is u = 0.
Proof. a) For an interior (local) maximum or minimum at an interior point (x, y), we
have
ux = 0, uy = 0.
Suppose u has a positive maximum in the interior of . Then
u > 0, uxx 0, uyy 0.
With these values, we have
auxx + buyy + cux + duy eu = 0,
0 0 =0 =0 <0
b) Since u can not have positive maximum in the interior of , then max u = 0 on .
Since u can not have negative minimum in the interior of , then min u = 0 on .
Since u is continuous, u 0 on .
Partial Dierential Equations Igor Yanovsky, 2005 282
Proof. Assume u(x, y) = X(x)Y (y), then substitution in the PDE gives X Y + XY =
0.
X Y
= = .
X Y
From X + X = 0, we get Xn (x) = an cos nx + bn sin nx. Boundary conditions
give
u(0, y) = X(0)Y (y) = 0
X(0) = 0 = X().
u(, y) = X()Y (y) = 0
u(x, y) = a
n sin nx sinh ny,
n=1
which saties the equation and the three homogeneous boundary conditions. The
boundary condition at y = gives
u(x, ) = g(x) = an sin nx sinh n,
n=1
g(x) sin mx dx = an sinh n
sin nx sin mx dx = am sinh m.
0 0 2
n=1
Partial Dierential Equations Igor Yanovsky, 2005 283
2
a
n sinh n = g(x) sin nx dx.
0
Partial Dierential Equations Igor Yanovsky, 2005 284
Proof. Assume u(x, y) = X(x)Y (y), then substitution in the PDE gives X Y + XY =
0.
X Y
= = .
X Y
Consider X + X = 0.
If = 0, X0 (x) = a0 x + b0 .
If > 0, Xn (x) = an cos nx + bn sin nx.
Boundary conditions give
ux (0, y) = X (0)Y (y) = 0
X (0) = 0 = X ().
ux (, y) = X ()Y (y) = 0
u(x, y) = a
0 y + a
n cos nx sinh ny,
n=1
which saties the equation and the three homogeneous boundary conditions. The fourth
boundary condition gives
0 +
u(x, ) = g(x) = a an cos nx sinh n,
n=1
Partial Dierential Equations Igor Yanovsky, 2005 285
&
0 + 0 2 ,
g(x) dx = 0 a n=1 an cos nx sinh n dx = a
0 &
0 g(x) cos mx dx = n=1 an sinh n 0 cos nx cos mx dx = 2 am sinh m.
1
a
0 = 2 g(x) dx,
0
2
a
n sinh n = g(x) cos nx dx.
0
Partial Dierential Equations Igor Yanovsky, 2005 286
Find a bounded solution u(x, y) and show that u(x, y) 0 when |x| + y .
Proof. Assume u(x, y) = X(x)Y (y), then substitution in the PDE gives X Y + XY =
0.
X Y
= = .
X Y
Consider X + X = 0.
If = 0, X0 (x) = a0 x + b0 .
If > 0, Xn (x) = an cos n x + bn sin n x.
Since we look for bounded solutions as |x| , we have a0 = 0.
Consider Y n Y = 0.
If n = 0, Y0 (y) = c0 y +d0 .
If n > 0, Yn (y) = cn e n y + dn e n y .
Since we look for bounded solutions as y , we have c0 = 0, dn = 0. Thus,
0 +
u(x, y) = a e n y
an cos
n x + bn sin n x .
n=1
Initial condition gives:
f (x) = uy (x, 0) u(x, 0) =
a0 ( n + 1) an cos n x + bn sin n x .
n=1
f (x) C0 (R1 ), i.e. has compact support [L, L], for some L > 0. Thus the coecients
n , bn are given by
a
L
f (x) cos n x dx = ( n + 1) an L.
L
L
f (x) sin n x dx = ( n + 1)bnL.
L
70
Thus, u(x, y) 0 when |x| + y .
70
Note that if we change the roles of X and Y in , the solution we get will be unbounded.
Partial Dierential Equations Igor Yanovsky, 2005 287
r 2 urr + rur + u = 0.
Let r = et , u(r(t), ).
ut = ur rt = et ur ,
utt = (et ur )t = et ur + e2t urr = rur + r 2 urr .
Thus, we have
utt + u = 0.
Let u(t, ) = X(t)Y (), which gives X (t)Y () + X(t)Y () = 0.
X (t) Y ()
= = .
X(t) Y ()
The coecients an , bn for n 1 are determined from the Fourier series for h().
a0 is not determined by h() and therefore may take an arbitrary value. Moreover,
Partial Dierential Equations Igor Yanovsky, 2005 288
2
the constant term in the Fourier series for h() must be zero [i.e., 0 h()d = 0].
Therefore, the problem is not solvable for an arbitrary function h(), and when it is
solvable, the solution is not unique.
Partial Dierential Equations Igor Yanovsky, 2005 289
r 2 urr + rur + u = 0.
Let r = et , u(r(t), ).
ut = ur rt = et ur ,
utt = (et ur )t = et ur + e2t urr = rur + r 2 urr .
Thus, we have
utt + u = 0.
Let u(t, ) = X(t)Y (), which gives X (t)Y () + X(t)Y () = 0.
X (t) Y ()
= = .
X(t) Y ()
u(r, ) = a
0 + an cos n + bn sin n).
r n (
n=1
a1 cos + b1 sin ).
u(r, ) = r(
Partial Dierential Equations Igor Yanovsky, 2005 292
u = 0 in r < 1
u/r = f () on r = 1,
beginning with the Fourier series for f (you may assume that f is continuously dif-
ferentiable). Give your answer as a power series in x1 + ix2 plus a power series in
x1 ix2 . There is a necessary condition on f for this boundary value problem to be
solvable that you will nd in the course of doing this.
u(r, ) = a
0 + an cos n + bn sin n).
r n (
n=1
Since
ur (r, ) = an cos n + bn sin n),
nr n1 (
n=1
the boundary condition gives
ur (1, ) = an cos n + bn sin n) = f ().
n (
n=1
2
1
an = f () cos n d,
n 0
2
bn = 1
f () sin n d.
n 0
2
0 is not determined by f () (since 0 f () d = 0). Therefore, it may take an
a
arbitrary
2 value. Moreover, the constant term in the Fourier series for f () must be zero
[i.e., 0 f ()d = 0]. Therefore, the problem is not solvable for an arbitrary function
f (), and when it is solvable, the solution is not unique.
uxx + uyy = 0
u=f for x2 + y 2 = 1.
u(r, ) = a
0 + r n (
an cos n + bn sin n).
n=1
we need
2
2
lim r f0 + r 2n (fn cos n + fn sin n)
= 0.
r
n=1 need
Thus, the conditions are
fn , fn = 0, n = 0, 1, 2.
Partial Dierential Equations Igor Yanovsky, 2005 296
1 1 2
Hint: Use the formula =
r r (r r ) + r 2 2 for the Laplacian in polar coordinates.
X (t) Y ()
= = .
X(t) Y ()
From Y () + Y () = 0, we get Yn () = an cos + bn sin .
Boundary conditions give
un (r, 0) = 0 = Xn (r)Yn (0) = 0, Yn (0) = 0,
un (r, ) = 0 = Xn (r)Yn () = 0, Yn () = 0.
Thus, 0 = Yn (0) = an , and Yn () = bn sin = 0 = n n = n2 .
Thus, Yn () = bn sin n, n = 1, 2, . . ..
With these values of n we solve X (t) n2 X(t) = 0.
If n = 0, X0 (t) = c0 t + d0 . X0 (r) = c0 log r + d0 .
If n > 0, Xn (t) = cn ent + dn ent Xn (r) = cn r n + dn r n .
We have,
u(r, ) = Xn (r)Yn () = cn r n + dn r n ) sin n.
(
n=1 n=1
4 r
u(r, ) = sin .
3r 3
Partial Dierential Equations Igor Yanovsky, 2005 298
X (t) Y ()
= = .
X(t) Y ()
u(r, ) = Xn (r)Yn () = cn r n + dn r n ) sin n.
(
n=1 n=1
X (t) Y ()
= = .
X(t) Y ()
From Y () + Y () = 0, we get Yn () = an cos + bn sin .
Boundary conditions give
un (r, 0) = Xn (r)Yn (0) = 0 Yn (0) = 0,
un (r, ) = Xn (r)Yn ( ) = 0 Yn ( ) = 0.
2 2 2
Yn () = an n sin n + bn n cos n . Thus, Yn (0) = bn n = 0
bn = 0.
Yn ( 2 ) = an n sin n 2 = 0 n 2 = n n = (2n)2 .
Thus, Yn () = an cos(2n), n = 0, 1, 2, . . ..
In particular, Y0 () = a0 t + b0 . Boundary conditions give Y0 () = b0 .
With these values of n we solve X (t) (2n)2 X(t) = 0.
If n = 0, X0 (t) = c0 t + d0 . X0 (r) = c0 log r + d0 .
2nt
If n > 0, Xn (t) = cn e + dn e 2nt Xn (r) = cn r 2n + dn r 2n .
u(r, ) = c0 log r + d0 + cn r 2n + dn r 2n ) cos(2n).
(
n=1
c0
f1 () = ur (a, ) = +2 cn a2n1 + dn a2n1 ) cos(2n),
n(
a
n=1
c0
f2 () = ur (b, ) = +2 cn b2n1 + dn b2n1 ) cos(2n).
n(
b
n=1
Proof. No, since the boundary conditions are Neumann. The solution is unique only
up to a constant.
Partial Dierential Equations Igor Yanovsky, 2005 301
Problem (S99, #4). Let u(x, y) be harmonic inside the unit disc,
with boundary values along the unit circle
1, y > 0
u(x, y) =
0, y 0.
r 2 urr + rur + u = 0.
With r = et , we have
utt + u = 0.
Let u(t, ) = X(t)Y (), which gives X (t)Y () + X(t)Y () = 0.
X (t) Y ()
= = .
X(t) Y ()
u(r, ) = a
0 + an cos n + bn sin n).
r n (
n=1
and the coecients an and bn are determined from the above equation.
71
71
See Yanas solutions, where Greens function on a unit disk is constructed.
Partial Dierential Equations Igor Yanovsky, 2005 302
We now plug this equation into with inhomogeneous term and obtain
Xn(t) sin[(2n 1)] (2n 1)2 Xn (t) sin[(2n 1)] = f (t, ),
n=1
Xn (t) (2n 1)2 Xn (t) sin[(2n 1)] = f (t, ),
n=1
2
Xn (t) (2n 1)2 Xn (t) = f (t, ) sin[(2n 1)] d,
4 0
4 2
Xn(t) 2
(2n 1) Xn (t) = f (t, ) sin[(2n 1)] d.
0
The solution to this equation is
Xn (t) = cn e(2n1)t + dn e(2n1)t + Unp (t), or
(2n1) (2n1)
Xn (r) = cn r + dn r + unp(r),
where unp is the particular solution of inhomogeneous equation.
u must be nite at r = 0 cn = 0, n = 1, 2, . . .. Thus,
u(r, ) = dn r (2n1) + unp(r) sin[(2n 1)].
n=1
Note the similar problem on 2D Poisson equation on a square domain. The prob-
lem is used by rst nding the eigenvalues and eigenfunctions of the Laplacian, and
then expanding f (x, y) in eigenfunctions, and comparing coecients of f with the gen-
eral solution u(x, y).
Here, however, this could not be done because of the circular geometry of the domain.
In particular, the boundary conditions do not give enough information to nd explicit
representations for m and n . Also, the condition u = 0 for x > 0, y > 0, x2 +y 2 = 1
Partial Dierential Equations Igor Yanovsky, 2005 304
72
ChiuYens solutions have attempts to solve this problem using Greens function.
Partial Dierential Equations Igor Yanovsky, 2005 305
Proof. Assume u(x, t) = X(x)T (t), then substitution in the PDE gives XT X T = 0.
X T
= = .
X T
From X + X = 0, we get Xn (x) = an cos nx + bn sin nx. Boundary conditions
give
u(0, t) = X(0)T (t) = 0
X(0) = X() = 0.
u(, t) = X()T (t) = 0
Thus, Xn (0) = an = 0, and Xn (x) = bn sin nx, n = n2 , n = 1, 2, . . . .
With these values of n , we solve T +n2 T = 0 to nd Tn (t) = cn sin nt+dn cos nt.
Thus,
u(x, t) = cn sin nt + dn cos nt sin nx,
n=1
ut (x, t) = cn cos nt ndn sin nt sin nx.
n
n=1
Initial conditions give
1 = u(x, 0) = dn sin nx,
n=1
0 = ut (x, 0) = n
cn sin nx.
n=1
By orthogonality, we may multiply both equations by sin mx and integrate:
sin mx dx = dm ,
2
0
cn ,
0 dx = n
0 2
which gives the coecients
4
dn =
2
(1 cos n) = n , n odd,
and cn = 0.
n 0, n even,
Proof. Assume u(x, t) = X(x)T (t), then substitution in the PDE gives
XT + 3XT + XT = X T,
T T X
+3 +1 = = .
T T X
From X + X = 0, Xn (x) = an cos n x + bn sin n x. Boundary conditions
give
u(0, t) = X(0)T (t) = 0
X(0) = X(1) = 0.
u(1, t) = X(1)T (t) = 0
Thus, Xn (0) = an = 0, and Xn (x)
= bn sin n x.
Xn (1) = bn sin n = 0. Hence, n = n, or n = (n)2 , n = 1, 2, . . ..
32 t 5 5
Tn (t) = e cn cos (n)2 t + dn sin (n)2 t .
4 4
5
32 t 5
u(x, t) = X(x)T (t) = e cn cos (n)2 t + dn sin (n)2 t sin nx.
4 4
n=1
Initial conditions give
0 = u(x, 0) = cn sin nx.
n=1
By orthogonality, we may multiply this equations by sin mx and integrate:
1
1
0 dx = cm cm = 0.
0 2
Partial Dierential Equations Igor Yanovsky, 2005 307
Thus,
5
32 t
u(x, t) = dn e (n)2 t sin nx.
sin
4
n=1
5 5
5
3 32 t 2 32 t 2 2
ut (x, t) = dn e sin (n) t + dn e (n) cos (n) t sin nx,
2 4 4 4
n=1
5
x sin(2x) = ut (x, 0) = dn (n)2 sin nx.
4
n=1
By orthogonality, we may multiply this equations by sin mx and integrate:
1
1
5
x sin(2x) sin(mx) dx = dm (m)2 ,
0 2 4
1
2
dn = ' x sin(2x) sin(nx) dx.
2 5
(n) 4 0
5
32 t
u(x, t) = e dn sin (n)2 t sin nx.
4
n=1
Problem (F04, #1). Solve the following initial-boundary value problem for the wave
equation with a potential term,
utt uxx + u = 0 0 < x < , t < 0
u(0, t) = u(, t) = 0 t>0
u(x, 0) = f (x), ut (x, 0) = 0 0 < x < ,
where
x if x (0, /2),
f (x) =
x if x (/2, ).
The answer should be given in terms of an innite series of explicitly given functions.
Proof. Assume u(x, t) = X(x)T (t), then substitution in the PDE gives
XT X T + XT = 0,
T X
+1 = = .
T X
From X + X = 0, Xn (x) = an cos n x + bn sin n x. Boundary conditions
give
u(0, t) = X(0)T (t) = 0
X(0) = X() = 0.
u(, t) = X()T (t) = 0
Thus, Xn (0) = an = 0, and Xn (x)= bn sin n x.
Xn () = bn sin n = 0. Hence, n = n, or n = n2 , n = 1, 2, . . ..
u(x, t) = X(x)T (t) = cn cos 1 + n2 t + dn sin 1 + n2 t sin nx.
n=1
ut (x, t) = cn ( 1 + n2 ) sin 1 + n2 t + dn ( 1 + n2 ) cos 1 + n2 t sin nx.
n=1
Initial conditions give
f (x) = u(x, 0) = cn sin nx.
n=1
0 = ut(x, 0) = dn ( 1 + n2 ) sin nx.
n=1
By orthogonality, we may multiply both equations by sin mx and integrate:
f (x) sin mx dx = cm ,
2
0
0 dx = dm 1 + m2 ,
0 2
which gives the coecients
2 2 2 2
cn = f (x) sin nx dx = x sin nx dx + ( x) sin nx dx
0 0
2
2 1 2 1 2 2 1 1
= x cos nx + cos nx dx + cos nx + x cos nx cos nx dx
n 0 n 0 n 2
n 2
n
2
2 n 1 n 1
= cos + 2 sin 2 sin 0
2n 2 n 2 n
2 n n 1 1 n
+ cos n + cos + cos n cos 2 sin n + 2 sin
n n 2 n 2n 2 n n 2
2 1 n 2 1 n 4 n
= sin + sin = sin
n2 2 n2 2 n2 2
0, n = 2k
0, n = 2k
= 4 =
2, n = 4m + 1 n1
4
n
(1) 2 n2 , 4
n = 2k + 1.
n2 , n = 4m + 3
dn = 0.
u(x, t) = cn cos 1 + n2 t sin nx.
n=1
Partial Dierential Equations Igor Yanovsky, 2005 309
Proof. First, we need to obtain function v that satises vt = vxx and takes 0
boundary conditions. Let
nx
n 2
Xn (x) = bn sin , n = .
2 2
Partial Dierential Equations Igor Yanovsky, 2005 310
n 2
With these values of n , we solve T + 2 T = 0 to nd
n 2
Tn (t) = cn e( 2
) t
.
n 2 nx
v(x, t) = Xn (x)Tn (t) = cn e( 2
) t
sin .
2
n=1 n=1
2 0 n is even,
nx
cn = (x2 2x) sin dx = 32
0 2 (n)3 n is odd.
32 ( n )2 t nx
v(x, t) = 3
e 2 sin .
(n) 2
n=2k1
lim u(x, t) = x + 1.
t+
Partial Dierential Equations Igor Yanovsky, 2005 311
Find v(x) - the limit of u(x, t) when t . Show that v(x) is one of the ininitely
many solutions of the stationary problem
vxx = 0 0<x<L
vx (0) = vx (L) = A.
Proof. First, we need to obtain function v that satises vt = vxx and takes 0
boundary conditions. Let
Xn (x) = an sin x + bn cos x.
Hence, Xn (0)
= bn n= 0 bn = 0; and Xn (x) = an cos x.
Xn (L) = an sin L = 0 L = n n = ( n 2
L ) .
nx
n 2
Xn (x) = an cos , n = .
L L
n 2
With these values of n , we solve T + L T = 0 to nd
n 2
T0 (t) = c0 , Tn (t) = cn e( L
) t
, n = 1, 2, . . ..
n 2 nx
v(x, t) = Xn (x)Tn (t) = c0 + cn e( L
) t
cos .
L
n=1 n=1
L L
AL2 1 L AL
c0 =
L (f (x) Ax) dx = f (x) dx c0 = f (x) dx ,
0 0 2 L 0 2
L
L nx 1 L nx
cn = (f (x) Ax) cos dx cn = (f (x) Ax) cos dx.
2 0 L L 0 L
1 L
AL n 2 nx
v(x, t) = f (x) dx + cn e( L ) t cos .
L 0 2 n
L
We now use equation (25.4) to convert back to function u:
u(x, t) = v(x, t) + Ax.
1 L
AL n 2 nx
u(x, t) = f (x) dx + cn e( L ) t cos + Ax.
L 0 2 n
L
To show that v(x) is one of the ininitely many solutions of the stationary problem
vxx = 0 0<x<L
vx (0) = vx (L) = A,
we can solve the boundary value problem to obtain v(x, t) = Ax+b, where b is arbitrary.
Partial Dierential Equations Igor Yanovsky, 2005 313
Proof. We rst nd u1 , the solution to the homogeneous heat equation (no F (x, t)).
Let w(x) be the solution of the Dirichlet problem:
w = 0, x
w(x) = g(x), x
and let v(x, t) be the solution of the IBVP for the heat equation with homogeneous
BC:
vt = v,
x , t 0
v(x, 0) = f (x) w(x), x
v(x, t) = 0, x , t > 0.
Then u1 (x, t) satises
u1 (x, t) = v(x, t) + w(x).
Proof. First, we need to obtain function v that satises vt = vxx and takes 0
boundary conditions. Let
sin 1 1
v(x, t) = u(x, t) + (x 1) + (cos x + sin x)et . (25.8)
cos 1
Initial condition tranforms to:
sin 1 1 sin 1 1
v(x, 0) = u(x, 0) + (x 1) + (cos x + sin x) = (x 1) + (cos x + sin x).
cos 1 cos 1
The new problem is
vt = vxx ,
sin 11
v(x, 0) = (x 1) + (cos x + sin x),
cos 1
v(0, t) = 0, vx (1, t) = 0.
From X + X = 0, we get Xn (x) = an cos x + bn sin x.
Using the rst boundary condition, we have
X(0) = 0.
v(0, t) = X(0)T (t) = 0
Hence, Xn (0) = an = 0, and Xn (x) = bn sin x. We also have
vx (1, t) = X (1)T (t) = 0 X (1) = 0.
Xn (x) = bn cos x,
Xn (1) = bn cos = 0,
cos = 0,
= n + .
2
Thus,
2
Xn (x) = bn sin n + x, n = n + .
2 2
2
With these values of n , we solve T + n + 2 T = 0 to nd
2
Tn (t) = cn e(n+ 2 ) t .
v(x, t) = Xn (x)Tn (t) = bn sin n + x e(n+ 2 )2 t.
2
n=1 n=1
Problem (F02, #6). The temperature of a rod insulated at the ends with an ex-
ponentially decreasing heat source in it is a solution of the following boundary value
problem:
2t
ut = uxx + e g(x) for (x, t) [0, 1] R+
ux (0, t) = ux (1, t) = 0
u(x, 0) = f (x).
2 2 bn
u(x, t) = cn en t e2t
cos nx.
2 n2 2
n=0
bn
u(x, 0) = cn cos nx = f (x),
2 n2 2
n=0
u = f (x) x
u
(s) = g(s) s .
n
Show that your condition is necessary for a solution to exist.
ux (0) = 1 ux (2) = 0 ?
b) With
v(x, t) = u(x, t) cos x
the problem above transforms to
vt = vxx
vx (0, t) = vx(2, t) = 0
v(x, 0) = cos 2x.
We solve this problem for v using the separation of variables. Let v(x, t) = X(x)T (t),
which gives XT = X T .
X T
= = .
X T
From X +X = 0, we get X
n (x) = a
n cos x + b n sin x.
Xn (x) = n an sin x + n bn cos x.
Using boundary conditions, we have
vx (0, t) = X (0)T (t) = 0
X (0) = X (2) = 0.
vx (2, t) = X (2)T (t) = 0
Hence, Xn (0) = n bn= 0, and Xn (x) =an cos n x.
Xn (2) = n an sin n 2 = 0 n = n2 n = ( n2 )2 . Thus,
nx
n 2
Xn (x) = an cos , n =
2 2
Partial Dierential Equations Igor Yanovsky, 2005 319
n 2
With these values of n , we solve T + 2 T = 0 to nd
n 2
Tn (t) = cn e( 2 ) t .
n 2 nx
v(x, t) = Xn (x)Tn (t) = n e( 2 )
a t
cos .
2
n=0 n=0
c) Set ut = 0. We have
uxx + cos x = 0 x [0, 2]
ux (0) = 1, ux (2) = 0.
uxx = cos x,
ux = sin x + C,
u(x) = cos x + Cx + D.
Boundary conditions give:
1 = ux (0) = C,
0 = ux (2) = C contradiction
There exists no steady state solution.
We may use the result we obtained in part (a) with uxx = cos x = f (x). We
need
u
f (x) dx = ds,
n
2
cos x dx = ux (2) ux (0) =
1 .
0 given
=0
Partial Dierential Equations Igor Yanovsky, 2005 320
Proof. We can solve the second equation for v and then use the value of v to solve the
rst equation for u. 73
We have
vt = 2vxx, 0 x , t > 0
v(x, 0) = sin x,
v(0, t) = v(, t) = 0.
Assume v(x, t) = X(x)T (t), then substitution in the PDE gives XT = 2X T .
T X
=2 = .
T X
' '
From X
+
2X= 0, we get Xn (x) = an cos + bn sin
2x
2 x.
Boundary conditions give
v(0, t) = X(0)T (t) = 0
X(0) = X() = 0.
v(, t) = X()T (t) = 0
'
Thus, Xn (0) = an = 0, and Xn (x) = bn sin 2 x.
' '
Xn () = bn sin 2 = 0. Hence 2 = n, or = 2n2 .
73
Note that if the matrix was fully inseparable, we would have to nd eigenvalues and eigenvectors,
just as we did for the hyperbolic systems.
Partial Dierential Equations Igor Yanovsky, 2005 321
We have
1 2t
ut = uxx 2 e sin x, 0 x , t > 0
u(x, 0) = sin x,
u(0, t) = u(, t) = 0.
&
Let u(x, t) = n=1 un (t) sin nx. Plugging this into the equation, we get
1
un (t) sin nx + n2 un (t) sin nx = e2t sin x.
2
n=1 n=1
For n = 1:
1
u1 (t) + u1 (t) = e2t .
2
Combining homogeneous and particular solution of the above equation, we obtain:
1
u1 (t) = e2t + c1 et .
2
For n = 2, 3, . . .:
un (t) + n2 un (t) = 0,
2
un (t) = cn en t .
Thus,
1
2 1
2
u(x, t) = e2t + c1 et sin x + cn en t sin nx = e2t sin x + cn en t sin nx.
2 2
n=2 n=1
From initial condition, we get
1
u(x, 0) = sin x + cn sin nx = sin x.
2
n=1
Thus, c1 = 12 , cn = 0, n = 2, 3, . . ..
1
u(x, t) = sin x (e2t + et ).
2
X Y
+ + = 0.
X Y
Letting = 2 + 2 and using boundary conditions, we nd the equations for X and
Y:
X + 2 X = 0 Y + 2 Y = 0
X(0) = X(a) = 0 Y (0) = Y (b) = 0.
The solutions of these one-dimensional eigenvalue problems are
m n
m = n =
a b
mx ny
Xm (x) = sin Yn (y) = sin ,
a b
where m, n = 1, 2, . . . . Thus we obtain solutions of (26.1) of the form
m2 n2 mx ny
mn = 2 + umn (x, y) = sin sin ,
a2 b2 a b
where m, n = 1, 2, . . . .
Observe that the eigenvalues {mn } m,n=1 are positive. The smallest eigenvalue 11
has only one eigenfunction u11 (x, y) = sin(x/a) sin(y/b); notice that u11 is positive
in . Other eigenvalues may correspond to more than one choice of m and n; for
example, in the case a = b we have nm = nm . For this , there are two linearly
independent eigenfunctions. However, for a particular value of there are at most
nitely many linearly independent eigenfunctions. Moreover,
b a b a
mx ny m x n y
umn (x, y) umn (x, y) dx dy = sin sin sin sin dx dy
0 0 0 0 a b a b
n y
a b
2 0 sin ny
b sin b dy ab
4 if m = m and n = n
= =
0 0 if m = m or n = n .
In particular, the {umn } are pairwise
orthogonal. We could normalize each umn by a
scalar multiple (i.e. multiply by 4/ab) so that ab/4 above becomes 1.
1 if m = n
n (x)m(x) dx =
0 if m = n.
Partial Dierential Equations Igor Yanovsky, 2005 324
( + )u = 0 in D
u=0 on D.
2 u(x, y) 2 u(x, y)
+ + u(x, y) = 0 (26.2)
x2 y 2
in the strip {(x, y), 0 < y < , < x < +} with boundary conditions
Find all bounded solutions of the boundary value problem (26.4), (26.5) when
a) = 0, b) > 0, c) < 0.
Proof. a) = 0. We have
uxx + uyy = 0.
Assume u(x, y) = X(x)Y (y), then substitution in the PDE gives
X Y + XY = 0.
Boundary conditions give
u(x, 0) = X(x)Y (0) = 0
Y (0) = Y () = 0.
u(x, ) = X(x)Y () = 0
Method I: We have
X Y
= = c, c > 0.
X Y
From X + cX = 0, we have Xn (x) = an cos
cx +
bn sin cx.
From Y cY = 0, we have Yn (y) = cn e cy + dn e cy .
Y (0) = cn + dn = 0 cn = dn .
Partial Dierential Equations Igor Yanovsky, 2005 326
Y () = cn e c cn e c = 0 cn = 0 Yn (y) = 0.
u(x, y) = X(x)Y (y) = 0.
b) > 0. We have
X Y
+ + = 0.
X Y
Letting = 2 + 2 , and using boundary conditions for Y , we nd the equations:
X + 2 X = 0 Y + 2 Y = 0
Y (0) = Y () = 0.
The solutions of these one-dimensional eigenvalue problems are
Xm (x) = am cos m x + bm sin m x.
n = n, Yn (y) = dn sin ny, where m, n = 1, 2, . . ..
u(x, y) = umn (x, y) = (am cos m x + bm sin m x) sin ny.
m,n=1 m,n=1
c) < 0. We have
uxx + uyy + u = 0,
u(x, 0) = 0, u(x, ) = 0.
u 0 is the solution to this equation. We will show that this solution is unique.
Let u1 and u2 be two solutions, and consider w = u1 u2 . Then,
w + w = 0,
w(x, 0) = 0, w(x, ) = 0.
Multiply the equation by w and integrate:
ww + w 2 = 0,
ww dx + w 2 dx = 0,
w
w ds |w|2 dx + w 2 dx = 0,
n
=0
|w| dx = w 2 dx .
2
0 0
Partial Dierential Equations Igor Yanovsky, 2005 327
X Y
+ + k2 = 0.
X Y
Letting k2 = 2 + 2 and using boundary conditions, we nd:
X + 2 X = 0, Y + 2 Y = 0.
X(0) = X (a) = 0.
The solutions of these one-dimensional eigenvalue problems are
(m 12 )
m = ,
a
(m 12 )x
Xm (x) = sin Yn (y) = cn cos n y + dn sin n y,
a
where m, n = 1, 2, . . . . Thus we obtain solutions of the form
2
2 (m 12 ) (m 12 )x
kmn = +n2 , umn (x, y) = sin cn cos n y+dn sin n y ,
a a
where m, n = 1, 2, . . . .
(m 12 )x
u(x, y) = umn (x, y) = sin cn cos n y + dn sin n y .
a
m,n=1 m,n=1
We can take an alternate approach and prove the second part of the question. We
have
X Y + XY + k2 XY = 0,
Y X
= + k 2 = c2 .
Y X
We obtain Yn (y) = cn cos cy + dn sin cy. The second equation gives
X + k2 X = c2 X,
X + (k2 c2 )X = 0,
c2 k2 x c2 k2 x
Xm (x) = am e + bm e .
Thus, Xm (x) is bounded only if k2 c2 > 0, (if k2 c2 = 0, X = 0, and Xm (x) =
am x + bm , BCs give Xm (x) = x, unbounded), in which case
Xm (x) = am cos k2 c2 x + bm sin k2 c2 x.
Partial Dierential Equations Igor Yanovsky, 2005 329
2 u(x, y) 2 u(x, y)
+ + k2 u(x, y) = 0, (26.4)
x2 y 2
where < x < +, 0 < y < , k > 0 is a constant,
Proof. We have
uxx + uyy + k2 u = 0,
X Y + XY + k2 XY = 0,
X Y
= + k 2 = c2 .
X Y
We obtain Xm (x) = am cos cx + bm sin cx. The second equation gives
Y + k2 Y = c2 Y,
Y + (k2 c2 )Y = 0,
c2 k2 y c2 k2 y
Yn (y) = cn e + dn e .
Thus, Yn (y) is bounded only if k2 c2 > 0, (if k2 c2 = 0, Y = 0, and Yn (y) = cn y+dn ,
BCs give Y 0), in which case
Yn (y) = cn cos k2 c2 y + dn sin k2 c2 y.
Boundary conditions
give Yn (0) = cn = 0.
Yn () = dn sin k2 c2 = 0 k 2 c2 = n k2 c2 = n2
2 2 2
k = n + c , n = 1, 2, . . .. Hence, k > n, n = 1, 2, . . ..
Partial Dierential Equations Igor Yanovsky, 2005 330
u(x, y) = Xm (x)Yn (y) = sin ny Xm (x).
m,n=1 m,n=1
Partial Dierential Equations Igor Yanovsky, 2005 331
McOwen, 4.4 #7; 266B Ralston Hw. Show that the boundary value problem
a(x)u + b(x)u = u in
u=0 on
has only trivial solution with 0, when b(x) 0 and a(x) > 0 in .
which implies = 0. This gives the useful information that for the eigenvalue problem74
a(x)u + b(x)u = u
u/n = 0,
= 0 is an eigenvalue, its eigenspace is the set of constants, and all other s are
positive.
74
In Ralstons Hw#7 solutions, there is no - sign in front of a(x)u below, which is probably a
typo.
Partial Dierential Equations Igor Yanovsky, 2005 332
Similar Problem II: If 0, we show that the only solution to the problem below
is the trivial solution.
u + u = 0 in
u=0 on
uu dx + u2 dx = 0,
u 2
u
ds |u| dx +
u2 dx = 0.
n
=0 0
u = f in
u=0 on ,
&
Proof. Writing u = cn n and inserting into u = f , we get
n cn n = an n (x).
n=1 n=1
Thus, cn = an /n , and
an n (x)
u(x) = .
n=1
n
u = f (x)
u(0) = 0, u(L) = 0,
=
(0) = 0, (L) = 0.
The eigenvalues are n = (n/L)2 , and the corresponding eigenfunctions are sin(nx/L),
n = 1, 2, . . .. & &
Writing u = cn n = cn sin(nx/L) and inserting into u = f , we get
n 2 nx
cn sin = f (x),
L L
n=1
L
n 2 L
nx mx mx
cn sin sin dx = f (x) sin dx,
0 L L L 0 L
n=1
n 2 L L
nx
cn = f (x) sin dx,
L 2 0 L
L
2 0 f (x) sin(nx/L) dx
cn = .
L (n/L)2
Partial Dierential Equations Igor Yanovsky, 2005 334
L
2 f () sin(nx/L) sin(n/L) d
0
u(x) = cn sin(nx/L) = ,
n=1
L (n/L)2
2 sin(nx/L) sin(n/L)
L
u= f () d.
0 L (n/L)2
n=1
= G(x,)
See similar, but more complicated, problem in Sturm-Liouville Problems (S92, #2(c)).
Partial Dierential Equations Igor Yanovsky, 2005 335
n (x)n (y)
G(x, y) .
n
n=1
75
See the section: ODE - Integral Equations.
Partial Dierential Equations Igor Yanovsky, 2005 336
f (x, y) C 2 , f (x, y) = 0 if x = 0, x = a, y = 0, y = b,
2 mx ny
f (x, y) = cmn sin sin .
ab m,n=1 a b
cmn
u(x, y) = mn (x, y),
n=1
mn
Partial Dierential Equations Igor Yanovsky, 2005 337
Assume the solution u(x, t) may be expanded in the eigenfunctions with coecients
&
depending on t: u(x, t) = n=1 un (t)n (x). This implies
un (t)n (x) = n un (t)n (x),
n=1 n=1
un (t) + n un (t) = 0 for each n.
Since n > 0, this ordinary dierential equation has general solution
un (t) = An cos n t + Bn sin n t. Thus,
u(x, t) = An cos n t + Bn sin n t n (x),
n=1
ut (x, t) = n An sin n t + n Bn cos n t n (x),
n=1
u(x, 0) = An n (x) = g(x),
n=1
ut (x, 0) = n Bn n (x) = h(x).
n=1
Comparing with , we obtain
bn
An = an , Bn = .
n
Thus, the solution is given by
bn
u(x, t) = an cos n t + sin n t n (x),
n=1
n
Partial Dierential Equations Igor Yanovsky, 2005 339
with
an = g(x)n(x) dx,
bn = h(x)n (x) dx.
Partial Dierential Equations Igor Yanovsky, 2005 340
X Y
+ + = 0.
X Y
Letting = 2 + 2 and using boundary conditions, we nd the equations for X and
Y:
X + 2 X = 0 Y + 2 Y = 0
X(0) = X(a) = 0 Y (0) = Y (b) = 0.
The solutions of these one-dimensional eigenvalue problems are
m n
m = n =
a b
mx ny
Xm (x) = sin Yn (y) = sin ,
a b
where m, n = 1, 2, . . .. Thus we obtain eigenvalues and normalized eigenfunctions of
the Laplacian:
m2 n2 2 mx ny
mn = 2 + mn (x, y) = sin sin ,
a2 b2 ab a b
where m, n = 1, 2, . . . .
Second, we solve the Wave Equation (28.1) using the space eigenfunctions.
For g, h C 2 () with g = h = 0 on , we have eigenfunction expansions 76
g(x) = an n (x) and h(x) = bn n (x).
n=1 n=1
&
Assume u(x, t) = n=1 un (t)n (x). This implies
un (t) + n un (t) = 0 for each n.
76
In 2D, n is really mn , and x is (x, y).
Partial Dierential Equations Igor Yanovsky, 2005 341
bmn
u(x, t) = amn cos mn t + sin mn t mn (x),
m,n=1
mn
McOwen, 4.4 #3; 266B Ralston Hw. Consider the initial-boundary value problem
utt = u + f (x, t)
for x , t > 0
u(x, t) = 0 for x , t > 0
u(x, 0) = 0, ut (x, 0) = 0 for x .
Use Duhamels principle and an expansion of f in eigenfunctions to obtain a (formal)
solution.
Proof. a) We expand u in terms of the Dirichlet eigenfunctions of Laplacian in
.
n + n n = 0 in , n = 0 on .
Assume
u(x, t) = an (t)n (x), an (t) = n (x)u(x, t) dx.
n=1
f (x, t) = fn (t)n (x), fn (t) = n (x)f (x, t) dx.
n=1
an (t) = n (x)utt dx = n (u + f ) dx = n u dx + n f dx
= n u dx + n f dx = n n u dx + n f dx = n an (t) + fn (t).
fn
an (0) = n (x)u(x, 0) dx = 0.
an (0) = n (x)ut(x, 0) dx = 0.
77 Thus, we have an ODE which is converted and solved by Duhamels principle:
an + n an = fn (t)
a n + n an = 0 t
an (0) = 0 an (0, s) = 0 a n (t) = an (t s, s) ds.
0
an (0) = 0 an (0, s) = fn (s)
With the anzats a n (t, s) = c1 cos n t + c2 sin n t, we get c1 = 0, c2 = fn (s)/ n ,
or
sin n t
an (t, s) = fn (s) .
n
Duhamels principle gives
t t
sin( n (t s))
an (t) = n (t s, s) ds =
a fn (s) ds.
0 0 n
n (x) t
u(x, t) = fn (s) sin( n (t s)) ds.
n=1
n 0
u
77
We used Greens formula: n n u
n ds = (n u n u) dx.
n
h(x) = hn n (x), hn = n (x)h(x) dx.
n=1
un (t) = n (x)utt dx = n (a(t)uxx + f ) dx = a(t) n uxx dx + n f dx
Note: The initial data is not 0; therefore, the Duhamels principle is not applicable.
Also, the ODE is not linear in t, and its solution is not obvious. Thus,
u(x, t) = un (t)n (x),
n=1
u1tt + u1xx = 0,
u2tt + u2xx = 0,
u1 (0, t) = u1 (, t) = 0, u2 (0, t) = u2 (, t) = 0,
u1 (x, 0) = g1 (x), u1t (x, 0) = h1 (x); u2 (x, 0) = g2 (x), u2t (x, 0) = h2 (x).
Note that the equation is elliptic, and therefore, the maximum principle holds.
In order to prove that the solution does not depend continuously on the initial data
g, h, we need to show that one of the following conditions holds:
max |u1 u2 | > max |g1 g2 |,
max |ut1 ut2 | > max |h1 h2 |.
That is, the dierence of the two solutions is not bounded by the dierence of initial
data.
By the method of separation of variables, we may obtain
u(x, t) = (an cos nt + bn sin nt) sin nx,
n=1
u(x, 0) = an sin nx = g(x),
n=1
ut (x, 0) = nbn sin nx = h(x).
n=1
Not complete.
We also know that for elliptic equations, and for Laplace equation in particular, the
value of the function u has to be prescribed on the entire boundary, i.e. u = g on
, which is not the case here, making the problem under-determined. Also, ut is
prescribed on one of the boundaries, making the problem overdetermined.
Partial Dierential Equations Igor Yanovsky, 2005 346
Assume the solution u(x, t)&may be expanded in the eigenfunctions with coecients
depending on t: u(x, t) = n=1 un (t)n (x). This implies
un (t)n (x) = n un (t)n (x),
n=1 n=1
un (t) + n un (t) = 0, which has the general solution
un (t) = An en t. Thus,
u(x, t) = An en t n (x),
n=1
u(x, 0) = An n (x) = g(x).
n=1
Comparing with , we obtain An = an . Thus, the solution is given by
u(x, t) = an en tn (x),
n=1
with an = g(x)n (x) dx.
Also
n t
u(x, t) = an e n (x) = g(y)n(y) dy en t n (x)
n=1 n=1
= en t n (x)n (y) g(y) dy
n=1
K(x,y,t), heat kernel
Partial Dierential Equations Igor Yanovsky, 2005 347
where m, n = 1, 2, . . . .
Second, we solve the Heat Equation (29.1) using the space eigenfunctions.
For g C 2 () with g = 0 on , we have eigenfunction expansion
g(x) = an n (x).
n=1
&
Assume u(x, t) = n=1 un (t)n (x). This implies
un (t) + n un (t) = 0, which has the general solution
un (t) = An en t. Thus,
u(x, t) = An en t n (x),
n=1
u(x, 0) = An n (x) = g(x).
n=1
Partial Dierential Equations Igor Yanovsky, 2005 348
u(x, t) = amn emn t mn (x),
m,n=1
with mn , mn given above and amn = g(x)mn (x) dx.
Partial Dierential Equations Igor Yanovsky, 2005 349
ut = uxx + uyy
X Y
+ + = 0.
X Y
Letting = 2 + 2 and using periodic BCs, we nd the equations for X and Y :
X + 2 X = 0 Y + 2 Y = 0
X(0) = X(2) Y (0) = Y (2).
The solutions of these one-dimensional eigenvalue problems are
m = m n = n
imx
Xm (x) = e Yn (y) = einy ,
where m, n = . . . , 2, 1, 0, 1, 2, . . .. Thus we obtain eigenvalues and normalized eigen-
functions of the Laplacian:
where m, n = . . . , 2, 1, 0, 1, 2, . . ..
Second, we solve the&Heat Equationimx using the space eigenfunctions.
Assume u(x, y, t) = m,n= umn (t)e einy . This implies
umn (t) + (m2 + n2 )umn (t) = 0, which has the general solution
2 +n2 )t
un (t) = cmn e(m . Thus,
2 +n2 )t
u(x, y, t) = cmn e(m eimx einy .
m,n=
Partial Dierential Equations Igor Yanovsky, 2005 350
u(x, y, 0) = cmn eimx einy = f (x, y),
m,n=
2 2 2 2
f (x, y)eimxeiny dxdy = cmn eimx einy eim x ein y dxdy
0 0 0 0
m,n=
2
= 2 cmn einy ein y dy = 4 2 cmn .
0 n=
2 2
1
cmn = f (x, y)eimx einy dxdy = fmn .
4 2 0 0
Partial Dierential Equations Igor Yanovsky, 2005 351
b) Show that the integral u2 (x, y, t) dxdy is decreasing in t, if f is not constant.
Proof. We have
ut = uxx + uyy
Multiply the equation by u and integrate:
uut = uu,
1 d 2
u = uu,
2 dt
1 d 2 u
u dxdy = uu dxdy = u ds |u|2 dxdy
2 dt n
=0, (periodic BC)
= |u|2 dxdy 0.
Equality is obtained only when u = 0 u = constant f = constant.
If f is not constant, u2 dxdy is decreasing in t.
Partial Dierential Equations Igor Yanovsky, 2005 352
You may call the relevant eigenfunctions n (x) and assume that they are known.
Hence, (x) = A sin(x ) + A cos(x ). The second boundary condition gives
(1) + (1) = 0 = A sin( ) + A cos( ) + A cos( ) A sin( )
= A (1 ) sin( ) + 2 cos( )
A = 0 (since A = 0 implies B = 0 and = 0, which is notan eigenfunction). Therefore,
(1 ) sin( ) = 2 cos( ), and thus tan( ) = 21 .
un (t) = An en t. Thus,
u(x, t) = An en t n (x),
n=1
u(x, 0) = An n (x) = f (x).
n=1
Comparing with , we have An = an . Thus, the solution is given by
u(x, t) = an en tn (x),
n=1
with
1
an = f (x)n (x) dx.
0
Partial Dierential Equations Igor Yanovsky, 2005 354
Problem (W03, #3); 266B Ralston Hw. Let be a smooth domain in three
dimensions and consider the initial-boundary value problem for the heat equation
ut = u + f (x)
for x , t > 0
u/n = 0 for x , t > 0
u(x, 0) = g(x) for x ,
g/n = 0 for x .
g(x) = gn n (x), gn = n (x)g(x) dx.
n=1
an (t) = n (x)ut dx = n (u + f ) dx = n u dx + n f dx
= n u dx + n f dx = n n u dx + n f dx = n an + fn .
fn
an (0) = n (x)u(x, 0) dx = n g dx = gn .
78
Thus, we solve the ODE:
an + n an = fn
an (0) = gn .
On , u
n
= 0; n
n
= 0 since eigenfunctions are Neumann.
Partial Dierential Equations Igor Yanovsky, 2005 355
fn n t fn
u(x, t) = (f1 t + g1 )1 (x) + gn e + n (x).
n=2
n n
fn n
If f1 = 0 f (x) dx = 0 , lim u(x, t) = g1 1 + .
t n
n=2
If f1 = 0 f (x) dx = 0 , lim u(x, t) f1 1 t.
t
Partial Dierential Equations Igor Yanovsky, 2005 356
Problem (S97, #2). a) Consider the eigenvalue problem for the Laplace operator
in R2 with zero Neumann boundary condition
uxx + uyy + u = 0 in
u
n = 0 on .
ut = u + f (x, y)
u
u(x, y, 0) = 0, u| = 0.
n
What is the behavior of u(x, y, t) as t ?
c) Consider the problem with Neumann boundary conditions
vxx + vyy + f (x, y) = 0 in
v
n v = 0 on .
When does a solution exist? Find this solution, and nd its relation with the behavior
of lim u(x, y, t) in (b) as t .
79
b) We expand u in terms of the Neumann eigenfunctions of Laplacian in .
n
n + n n = 0 in , = 0 on .
n
u(x, y, t) = an (t)n (x, y), an (t) = n (x, y)u(x, y, t) dx.
n=1
an (t) = n (x, y)ut dx = n (u + f ) dx = n u dx + n f dx
= n u dx + n f dx = n n u dx + n f dx = n an + fn .
fn
an (0) = n (x, y)u(x, y, 0) dx = 0.
80 Thus, we solve the ODE:
an + n an = fn
an (0) = 0.
fn n t fn
u(x, t) = f1 1 t + e + n (x).
n n
n=2
fn n
If f1 = 0 f (x) dx = 0 , lim u(x, t) = .
t n
n=2
If f1 = 0 f (x) dx = 0 , lim u(x, t) f1 1 t.
t
& &
Assume v(x, y) = n=0 an n (x, y). Since we have f (x, y) = n=0 fn n (x, y), we
obtain
n an n + fn n = 0,
n=0 n=0
n an n + fn n = 0,
fn
an = .
n
& fn
v(x, y) = n=0 ( n )n (x, y).
Partial Dierential Equations Igor Yanovsky, 2005 360
ut = u 2 u
ut = amn (t)ei(mx+ny) ,
m,n
u = uxx + uyy = (m2 + n2 ) amn (t)ei(mx+ny) ,
m,n
2 u = uxxxx + 2uxxyy + uyyyy = (m4 + 2m2 n2 + n4 ) amn (t)ei(mx+ny)
m,n
= (m2 + n2 )2 amn (t)ei(mx+ny) .
m,n
When = 2, we have
2 2 2 2
u(x, t) = mn e(m +n )[2(m +n )]t ei(mx+ny) .
m,n
2 2
||u(t = 0)|| = |u(x, y, t = 0)|2 dxdy 1/2 < .
0 0
Derive an explicit bound on ||u(t)|| and show that it stays nite for all t.
b) If a = 0, construct the normal modes for (29.4); i.e. nd all solutions of the form
u(x, y, t) = et+ikx+ily .
c) Use these normal modes to construct a solution of (29.4) with a = 0 for the initial
data
1 1
u(x, y, t = 0) = 1 ix + .
1 2e 1 12 eix
k
k
1 1
= eix + eix
2 2
k=0 k=0
1 ikx 1 ikx
= e + e ,
2k 2k
k=0 k=0
1 ikx 1 ikx
= 2+ k
e + k
e .
2 2
k=1 k=1
1 1
a0 = 2; ak = , k > 0; ak = , k<0
2k 2k
1
a0 = 2; 0.
ak = |k| , k =
2
+
1 (k2 +b)t+ikx
u(x, y, t) = 2ebt + |k|
e .
k=, k=0
2
81
81
Note a similar question formulation in F92 #3(b).
Partial Dierential Equations Igor Yanovsky, 2005 364
Problem (S00, #3). Consider the initial-boundary value problem for u = u(x, y, t)
ut = u u
for (x, y) [0, 2]2, with periodic boundary conditions and with
u(x, y, 0) = u0 (x, y)
in which u0 is periodic. Find an asymptotic expansion for u for t large with terms
tending to zero increasingly rapidly as t .
Since u0 is periodic,
2 2
1
u0 (x, y) = u0mn ei(mx+ny) , u0mn = u0 (x, y) ei(mx+ny) dxdy.
m,n
4 2 0 0
Problem (S02, #4). Use the Fourier transform on L2 (R) to show that
du
+ cu(x) + u(x 1) = f (30.1)
dx
has a unique solution u L2 (R) for each f L2 (R) when |c| > 1 - you may assume
that c is a real number.
Proof. u L2 (R). Dene its Fourier transform u by
1
() =
u eix u(x) dx for R.
2 R
*
du
() = i u().
dx
We can nd u(x 1)() in two ways.
Let u(x 1) = v(x), and determinte v():
y
1 1
u(x 1)() = v() = eix v(x) dx = ei(y+1) u(y) dy
2 R 2 R
1
= eiy ei u(y) dy = ei u
().
2 R
We can also write the denition for u () and substitute x 1 later in calculations:
1 iy 1
() =
u e u(y) dy = ei(x1) u(x 1) dx
2 R 2 R
1
= eix ei u(x 1) dx = ei u(x 1)(),
2 R
u(x 1)() = ei u ().
Substituting into (30.1), we obtain
i u() + ei u
u() + c () = f(),
f()
() =
u .
i + c + ei
f()
= 1 f B,
u(x) = i
= fB
i + c + e 2
= 1
where B ,
i + c + ei
1 1 eix
B= = d.
i + c + ei 2 R i + c + ei
For |c| > 1, u u()) and this is unique by the
() exists for all R, so that u(x) = (
Fourier Inversion Theorem.
Problem (F96, #3). Find the fundamental solution for the equation
Hint: The Fourier transform converts this problem into a PDE which can be solved
using the method of characteristics.
1 ix
1 ix 1
xux (x)() = e xux (x) dx = e xu (i)eix x + eix u dx
2 R 2 2 R
=0
1 1
= i eix x u dx eix u dx
2 R 2 R
d d
= i xu(x)() u () = i i u () u () = u () u
().
d d
d
xu x (x)() = () u
u ().
d
Plugging these into (30.2), we get:
d
(, t) = 2 u
u (, t) u (, t) u
(, t) ,
t d
2
t = u
u + ,
u + u
u u = ( 2 1)
t u.
dx
= x x = c1 et , (c1 = xet )
dt
dy
= 1 y = t + c2 ,
dt
dz dz
= (x2 1)z = (c21 e2t 1)z = (c21 e2t 1)dt
dt z
1 x2 x2 x2
log z = c21 e2t + t + c3 = + t + c3 = + y c2 + c3 z = ce 2 +y .
2 2 2
Partial Dierential Equations Igor Yanovsky, 2005 369
Thus, we have
2
+t
(, t) = ce 2
u .
We use Inverse Fourier Tranform to get u(x, t): 82
1 1 2
u(x, t) = ix
e u (, t) d = eix ce 2 +t d
2 R 2 R
c 2 c 2
= et eix e 2 d = et eix+ 2 d
2 2
R R
c 2ix+ 2
c t (+ix)2 x2
= e t
e 2 d = e e 2 d e 2
2 R 2 R
c x2 y2 c x2 x2
= et e 2 e 2 dy = et e 2 2 = c et e 2 .
2 R 2
x2
u(x, t) = c et e 2 .
Check:
x2
ut = c et e 2 ,
x2
ux = c et xe , 2
x2 x2
uxx = c et e 2 + x2 e 2 .
Thus,
ut = uxx xux ,
x2 x2 x2 x2
c et e 2 = c et e 2 + x2 e 2 x c et xe 2 .
82
We complete the square for powers of exponentials.
Partial Dierential Equations Igor Yanovsky, 2005 370
u
.
= ik u
xk
Thus, the equation becomes:
&
t + i nk=1 ak (t)k u
u + a0 (t)
u = 0,
(, 0) = f (),
u
or
t + i a(t) u
u + a0 (t)
u = 0,
u
t = i a(t) + a0 (t) u .
with solution:
This is an ODE in u
t
(, 0) = c = f().
(, t) = ce 0 (i a(s)+a0 (s)) ds ,
u u Thus,
t
(, t) = f() e 0 (i a(s)+a0 (s)) ds .
u
Use the Inverse Fourier transform to get u(x, t):
t (f g)(x)
(, t) = f() e 0 (i a(s)+a0 (s)) ds
u(x, t) = u = n ,
(2) 2
t
g() = e 0 (i a(s)+a0 (s)) ds .
where
1 1 t
eix e 0 (i a(s)+a0 (s)) ds d.
g(x) = n e
ix
g() d = n
(2) 2 Rn (2) 2 Rn
(f g)(x) 1 i(xy)
t (i a(s)+a
0 (s)) ds
u(x, t) = n = n
e e 0 d f (y) dy.
(2) 2 (2) Rn Rn
b) Use Duhamels Principle and the result from (a).
t
u(x, t) = U (x, t s, s) ds, where U (x, t, s) solves
0
U
n
U
+ ak (t) + a0 (t)U = 0,
t xk
k=1
U (x, 0, s) = f (x, s).
Partial Dierential Equations Igor Yanovsky, 2005 371
t
t
1 ts
u(x, t) = U (x, t s, s) ds = ei(xy) e 0 (i a(s)+a0 (s)) ds d f (y, s) dy ds.
0 (2)n 0 Rn Rn
Partial Dierential Equations Igor Yanovsky, 2005 372
83
Problem (S93, #2). a) Dene the Fourier transform
f () = eix f (x) dx.
where a is a real constant, what f (x) does the inversion theorem give?
b) Show that
f (x b) = eib f
(x),
where b is a real constant. Hence, using part (a) and Parsevals theorem, show that
1 sin a(x + z) sin a(x + ) sin a(z )
dx = ,
x+z x+ z
where z and are real constants.
= eiy ib
e f (y) dy = eib f().
R
sin ax
With f (x) = (from (a)), we have
x
1 sin a(x + z) sin a(x + s) 1
dx = f (x + z)f (x + s) dx (x = x + s, dx = dx)
x+z x+s
1
= f (x + z s)f (x ) dx (Parsevals)
1 1
= f (x+ z s)f (x ) d part (b)
2
1
= f() ei(zs) f() d
2 2
a
1 2
=
f () ei(zs) d
2 2 a
a
1
= 2 ei(zs) d
2 2 a
1 a i(zs)
= e d
2 a
1 i(zs) =a
= e
2i(z s) =a
We had
sin y( )
f() = .
2y( )
We make change of variables: = . Then, = + . We have
sin y( )
f() = f( + ) = , or
( )
sin y( )
f( + ) = .
( )
We will also use the following result.
Let f( + a) =
g(), and determinte g() :
1 1
f( + a) = g() = eix( a) f( ) d
e g() d =
ix
2 R 2 R
ixa
= e f (x).
Using these results, we have
1 sin( )y sin( )y 1 2
d = ( 2y) f () f( + ) d
( ) ( )
= 2y f (x) e()ixf (x) dx
= 2y f (x)2 e()ix dx
y 2ix
e
= 2y e()ix dx
y 4y
1 y ()ix
= e dx
2 y
1 ()ixx=y
= e
2i( ) x=y
1 ()iy
= e e()iy
2i( )
sin( )y
= .
Partial Dierential Equations Igor Yanovsky, 2005 376
gk =
N hk .
n > 0:
an (y) = cn eny ,
an (0) = cn =
gn , (Dirichlet)
an (0) = ncn =
hn . (Neumann)
n
gn =
hn .
n < 0:
an (y) = bn eny ,
an (0) = bn =
gn , (Dirichlet)
an (0) = nbn = hn . (Neumann)
gn =
n hn .
gn =
|n| hn , n = 0.
n = 0 : a0 (y) = c0 ,
a0 (0) = c0 =
g0 , (Dirichlet)
a0 (0) = 0 =
h0 . (Neumann)
Note that solution f (x, y) may be written as
1
inx inx
f (x, y) = an (y) e = a0 (y) + an (y) e + an (y) einx
n= n= n=1
1
= c0 + bn eny einx + cn eny einx
n= n=1
&1 &
g0 + gn e e +
n= n=1
ny inx
gn eny einx , (Dirichlet)
= &1 &
c0 + n= hnn eny einx + hn ny inx
n=1 n e e . (Neumann)
f(, y) = c1 ey + c2 ey .
For > 0, c1 = 0; for < 0, c2 = 0.
>0: f(, y) =
c2 ey , fy (, y) = c2 ey ,
1 1
c2 = f(, 0) = eix f (x, 0) dx = eix g(x) dx =
g(), (Dirichlet)
2 2
1 1
c2 = fy (, 0) = eix
fy (x, 0) dx = eix h(x) dx =
h(). (Neumann)
2 2
g() = h().
Partial Dierential Equations Igor Yanovsky, 2005 378
g() =
|| h().
Partial Dierential Equations Igor Yanovsky, 2005 379
Problem (F97, #3). Consider the Dirichlet problem in the half-space xn > 0,
n 2:
u
u + a + k2 u = 0, xn > 0
xn
u(x , 0) = f (x ), x = (x1 , . . ., xn1 ).
84
Proof. Denote = ( , n). Transform in the rst n 1 variables:
2u
u
| |2 u
( , xn) + 2
( , xn ) + a ( , xn) + k2 u
( , xn ) = 0.
xn xn
Thus, the ODE and initial conditions of the transformed problem become:
xn xn + a
u uxn + (k2 | |2 )
u = 0,
( , 0) = f( ).
u
With the anzats u = cesxn , we obtain s2 + as + (k2 | |2 ) = 0, and
a a2 4(k2 | |2 )
s1,2 = .
2
Choosing only the negative root, we obtain the solution: 85
a a2 4(k2 | |2 )
u
( , xn ) = c( ) e 2
xn
. ( , 0) = c = f( ).
u Thus,
a a2 4(k2 | |2 )
( , xn ) = f( ) e
u 2
xn
.
Parsevals theorem gives:
||u||2L2(Rn1 ) = u||2L2 (Rn1 )
|| = u( , xn)|2 d
|
Rn1
aa2 4(k2 | |2 ) x 2 2
= f( ) e 2 n
d f( ) d
Rn1 Rn1
= 2 2 n1
||f|| = ||f ||2L2(Rn1 ) C,
L (R )
84
Note that the last element of x = (x , xn ) = (x1 , . . . , xn1 , xn ), i.e. xn , plays a role of time t.
As such, the PDE may be written as
u + utt + aut + k2 u = 0.
85
Note that a > 0 should have been provided by the statement of the problem.
Partial Dierential Equations Igor Yanovsky, 2005 380
where a(t), b(t), c(t) are continuous functions on [0, +], a(t) > 0 for t > 0.
G
n
G
b) (x1 , . . . , xn , y1 , . . . , yn , t) = ak (t) for t > 0,
t xk
k=1
G(x1 , . . . , xn, y1 , . . . , yn , 0) = (x1 y1 )(x2 y2 ) . . . (xn yn ).
b) Denote x = (x1 , . . . , xn ),
y = (y1 , . . . , yn ). Transform in x:
, 1
G( y, t) = n eix G(x, y, t) dx.
(2) 2 Rn
The equation is transformed to an ODE, that can be solved:
n
t(,
G y, t) = ,
ak (t) ik G( y , t),
k=1
t &n
,
G( y, t) = c ei 0[ k=1 ak (s) k ] ds .
Partial Dierential Equations Igor Yanovsky, 2005 381
where f is an integrable function (i.e. f L1 (Rn )), satisfying f (x) = 0 for |x| R.
Solve (30.4) by Fourier transform, and prove the following results.
a) There 2 n
is a solution of (30.4) belonging to L (R ) if n > 4. 2 n
b) If Rn f (x) dx = 0, there is a solution of (30.4) belonging to L (R ) if n > 2.
Proof.
u = f,
() = f(),
|| u2
1
() = 2 f(),
u Rn ,
||
f()
u(x) = .
||2
a) Then
1 1
|f()|2 2 |f()|2 |f()|2 2
||
u||L2 (Rn) = d d + d .
Rn ||4 ||<1 || 4
||1 ||
4
A B
If n > 4,
A 2 dSn1 = ||f||2 < .
|f| 2
Sn1
1
||u||L2(Rn) = ||
u||L2 (Rn ) = (A + B) 2 < .
Partial Dierential Equations Igor Yanovsky, 2005 383
b) We have
f()
1 ix f ()
u(x, t) = = n e d
||2 (2) 2 ||2
ix
1 e 1 iy
= n n e f (y) dy d
(2) 2 ||2 (2) 2
ei(xy)
1
= f (y) d dy
(2)n ||2
1
1 ei(xy)r n1
= f (y) r dS n1 dr dy
(2)n 0 Sn1 r2
1
1
= f (y) e i(xy)r n3
r dS n1 dr dy.
(2)n 0 Sn1
M < , if n>2.
1
|u(x, t)| = M f (y) dy < .
(2)n
Partial Dierential Equations Igor Yanovsky, 2005 384
Problem (F02, #7). For the right choice of the constant c, the function
F (x, y) = c(x + iy)1 is a fundamental solution for the equation
u u
+i =f in R2 .
x y
Find the right choice of c, and use your answer to compute the Fourier transform
(in distribution sense) of (x + iy)1 .
Proof. 86
= +i i .
x y x y
1
F1 (x, y) = 2 log |z| is the fundamental solution of the Laplacian. z = x + iy.
F1 (x, y) = ,
+i i F (x, y) = .
x y x y
hx + ihy = ei(x1 +y2 ) .
Suppose h = h(x1 + y2 ) or h = cei(x1+y2 ) .
c i1 ei(x1 +y2 ) i2 2 ei(x1 +y2 ) = ic(1 i2 ) ei(x1 +y2 ) ei(x1 +y2 ) ,
ic(1 i2 ) = 1,
1
c = ,
i(1 i2 )
1
h(x, y) = ei(x1 +y2 ) .
i(1 i2 )
Integrate by parts:
1 1
1
() = ei(x1 +y2 ) +i dxdy
x + iy R2 i(1 i2 ) x y (x + iy) 0
1 1
= = .
i(1 i2 ) i(2 + i1 )
86
Alan solved in this problem in class.
Partial Dierential Equations Igor Yanovsky, 2005 385
31 Laplace Transform
If u L1 (R+ ), we dene its Laplace transform to be
L[u(t)] = u# (s) = est u(t) dt (s > 0).
0
In practice, for a PDE involving time, it may be useful to perform a Laplace transform
in t, holding the space variables x xed.
The inversion formula for the Laplace transform is:
c+i
1 # 1
u(t) = L [u (s)] = est u# (s) ds.
2i ci
Example: f (t) = 1.
1 t= 1
L[1] = est 1 dt = est = for s > 0.
0 s t=0 s
Example: f (t) = eat .
st at
1 (as)t t= 1
L[e ] =
at
e e dt = e(as)t dt = e = for s > a.
0 0 as t=0 sa
1 1
Convolution: We want to nd an inverse Laplace transform of s s2 +1
.
1 t
1 1
L 2 = f g = 1 sin t dt = 1 cos t.
s s + 1
0
L[f ] L[g]
Thus, the transformed problem is: sL[u] f (x) = L[u]. Writing v(x) = L[u], we
have
v + sv = f in U.
Thus, the solution of this equation with RHS f is the Laplace transform of the solution
of the heat equation with initial data f .
Partial Dierential Equations Igor Yanovsky, 2005 387
Example: f (t) = tn .
tn est n
st n n
L[t ] =
n
e t dt = + est tn1 dt = L[tn1 ]
0 s 0 s 0 s
n n1 n! n!
= L[tn2 ] = . . . = n L[1] = n+1 .
s s s s
Partial Dierential Equations Igor Yanovsky, 2005 389
To obtain u(x, t), we take the inverse Laplace transform of u# (x, s):
u(x, t) = L1 [u# (x, s)] = L1 g # (s) e
s+a x
= gf
L[g] L[f ]
1 c+i
1 s+a x st s+a x
= gL e = g e e ds ,
2i ci
t 1 c+i
u(x, t) = g(t t ) est e s+a x ds dt .
0 2i ci
Partial Dierential Equations Igor Yanovsky, 2005 390
Problem (F04, #8). The function y(x, t) satises the partial dierential equation
y 2y
x + + 2y = 0,
x xt
and the boundary conditions
y(x, 0) = 1, y(0, t) = eat ,
where a 0. Find the Laplace transform, y(x, s), of the solution, and hence derive
an expression for y(x, t) in the domain x 0, t 0.
Problem (F90, #1). Using the Laplace transform, or any other convenient method,
solve the Volterra integral equation
x
u(x) = sin x + sin(x y)u(y) dy.
0
Problem (F91, #5). In what follows, the Laplace transform of x(t) is denoted
either by x(s) or by Lx(t). Show that, for integral n 0,
n!
L(tn ) = .
sn+1
Hence show that
1
LJ0 (2 ut) = eu/s ,
s
where
(1)n ( 1 z)2n 2
J0 (z) =
n!n!
n=0
Proof.
tn est n
st n
L[tn ] = e tn
dt = + est tn1 dt = L[tn1 ]
s 0 s s
0
g f 0
=0
n
n 1 n2 n! n!
= L[t ] = . . . = n L[1] = n+1 .
s s s s
(1)n un tn
(1)nun
(1)n un
LJ0 (2 ut) = L = L[tn ] =
n!n! n!n! n!sn+1
n=0 n=0 n=0
n
u n
1 (1) 1 u
= = e s.
s n=0 n! s s
1 u
L J0 (2 ut) x(u) du = L[J0 (2 ut)] x(u) du = e s x(u) du
0 0 s 0
1 #
1
= x ,
s s
where
x# (s) = eus x(u) du.
0
Partial Dierential Equations Igor Yanovsky, 2005 393
32.4 H
older Inequality
|uv| dx ||u||p||v||q,
1 1
which holds for u Lp () and v Lq (), where p + q = 1. In particular, this shows
uv L1 ().
Partial Dierential Equations Igor Yanovsky, 2005 394
Note that C 1 () is not a Banach space since ||u||1, need not be nite for u C 1 ().
1
2 2
2
||u||1,2 = < u, u >1 = (|u| + |u| ) dx (32.2)
when these expressions are dened and nite. For example, (32.1) and (32.2) are dened
for functions in C01 (). However, C01 () is not complete under the norm (32.2), and so
does not form a Hilbert space.
Divergence Theorem
A n dS = dx
div A
Trace Theorem
Poincare Inequality
up Cup 1p
2
|u(x)| dx C |u(x)|2 dx u C01 (), H01,2 () i.e. p = 2
u u p up u H01,p()
Partial Dierential Equations Igor Yanovsky, 2005 395
1
u = u(x) dx (Average value of u over ), || is the volume of
||
Notes
u u u
= u n = n1 + n2 |u|2 = u2x1 + u2x2
n x1 x2
|u|
|u| dx = u dx
u
a+b a2 + b2 ||u||2 + ||u||2
ab ab ||u||||u||
2 2 2
u2
uu = ( )
2
2
(uxy ) dx = uxxuyy dx u H02 () square
Problem (F04, #6). Let q C01 (R3 ). Prove that the vector eld
1 q(y)(x y)
u(x) = dy
4 R3 |x y|3
enjoys the following properties: 87
a) u(x) is conservative;
b) div u(x) = q(x) for all x R3 ;
c) |u(x)| = O(|x|2 ) for large x.
Furthermore, prove that the proverties (1), (2), and (3) above determine the vector eld
u(x) uniquely.
Then,
1
u(x) = q(y) V (x y) dy,
4 R3
1
curl u(x) = q(y) curlx V (x y) dy.
4 R3
(x1 , x2 , x3 )
curl V (x) = curl 3
(x21 + x22 + x23 ) 2
32 2x2 x3 32 2x3 x2 32 2x3 x1 32 2x1 x3 32 2x1 x2
= 5 5 , 5 5 , 5
(x21 + x22 + x23 ) 2 (x21 + x22 + x23 ) 2 (x21 + x22 + x23 ) 2 (x21 + x22 + x23 ) 2 (x21 + x22 + x23 ) 2
= (0, 0, 0).
87
McOwen, p. 138-140.
Partial Dierential Equations Igor Yanovsky, 2005 396
1
Thus, curl u = 4 R3 q(y) 0 dy = 0, and
u(x) is conservative.
1
b) Note that the Laplace kernel in R3 is 4r .
1 q(y)(x y) 1 q(r) r q(r)
u(x) = 3
dy = 3
r dr = dr = q.
4 R3 |x y| 4 R3 r R3 4r
c) Consider
1 q(y)
F (x) = dy.
4 R3 |x y|
F (x) is O(|x|1 ) as |x| .
Note that u = F , which is clearly O(|x|2 ) as |x| .