Principal of least degeneracy
Higher order matching and nonlinear equations
Lesson 16
A boundary layer of thickness O ( √ ε)
Let us consider the example εddx2f2 + x2dxdf − f = 0 with
x ∈ (0, 1), f(0) = 1, f(1) = 2. We will look for the one
term expansion. Since x2 > 1 on (0, 1) we assume that the boundary layer is at x = 0. Is something strange up to now? Thus
x2df0out
dx − f0out = 0, f0out(1) = 2.
Separating variables leads to
Z df0out f0out =
Z dx
x2 ⇒ f0out = Ce−1/x = 2e1−1/x.
Thickness O( √ ε)
We suppose that s = x/ε, then the new equation is
d2f
ds2 + ε2s2 df
ds − εf = 0, f(0) = 1
then
d2f0in
ds2 = 0 ⇒ f0in = As + 1.
Applying matching condition, we get
xlim→0 2e1−1/x = lim
s→∞
As + 1.
What is wrong?
Thickness O( √ ε)
Let us try the stretched variable s = x/εp, then
df
dx = 1 εp
df
ds, d2f
dx2 = 1 ε2p
d2f ds2 . ε1−2p d2f
ds2 + εps2 df
ds − f = 0.
Let us try to keep the maximum number of terms in the equation. Taking p = 1/2 we keep the second derivative and f. Then
d2f0in
ds2 − f0in = 0, f0in(0) = 1, ⇒ f0in = Aes + (1 − A)e−s.
Thickness O( √ ε)
Applying Prandtl’s matching condition, we get
xlim→0 2e1−1/x = lim
s→∞
Aes + (1 − A)e−s
that implies A = 2. Thus
f0comp = 2e1−1/x + e−x/√ε.
If p < 1/2 then f0in = 0 which does not satisfy the boundary condition at s = 0.
An interior boundary layer
εd2f
dx2 +xdf
dx+xf = 0 with x ∈ (−1, 1), f(−1) = e, f(1) = 2e−1. We expect the boundary layer at x = 0, two outer
solution f0+out = f0+ and f0−out = f0−. We have
df0+
dx + f0+ = 0, f0+(1) = 2e−1 ⇒ f0+ = 2e−x, df0−
dx + f0− = 0, f0−(−1) = e ⇒ f0− = e−x.
An interior boundary layer
The stretched variable s = x/εp leads to
ε1−2p d2f
ds2 + sdf
ds + εpsf = 0.
Taking p = 1/2, we get
d2f0in
ds2 + sdf0in
ds = 0.
We make a substitution w = dfds0in, then
dw + sw = 0 ⇒ w = Ae−s2/2 ⇒
An interior boundary layer
f0in = A
Z s
0
e−t2/2 dt + f0in(0).
We write R0s e−t2/2 dt = √12 R0s/√2 e−u2 du = B erf(s/√2). Then
f0in = B erf(s/√
2) + f0in(0).
The Prandtl’s condition gives
xlim→0+ 2e−x = lim
s→+∞
B erf(s/√
2) +f0in(0) ⇒ B+f0in(0) = 2,
xlim→0− e−x = lim
s→+∞
B erf(s/√
2)+f0in(0) ⇒ −B+f0in(0) = 1.
An interior boundary layer
B = 1/2, f0in(0) = 3/2.
Summarizing
f0+ = 2e−x, x > O(√ ε) f0in = 1/2 erf(x/√
2ε) + 3/2, x = O(√ ε) f0− = e−x, x < −O(√
ε)
The composition solution can not be produced by the standard way. We observe
An interior boundary layer
f0in
x > O(√
ε)
= 1/2 + 3/2 + sm. t., since erf ∼ 1,
f0in
x < −O(√
ε)
= −1/2 + 3/2 +sm. t., since erf ∼ −1, We conclude that
f0comp =
1/2 erf(x/√
2ε) + 3/2
e−x.
An interior boundary layer
Higher order matching
We shall work with the old example
εd2f
dx2 + df
dx = 2x + 1, x ∈ (0, 1), f(0) = 1, f(1) = 4, fex = x2 + x + 2 − e−x/ε + ε
2(1 − x) − 2e−x/ε
Using the stretched variable s = x/ε we get
Higher order matching
d2f
ds2 + df
ds = 2ε2s + ε, f(s, ε) ∼ f0(s) + εf1(s) + . . . , ε0 d2f0in
ds2 + df0in
ds = 0, f0in(0) = 1 ⇒ f0in = A + (1 − A)e−s ε1 d2f1in
ds2 + df1in
ds = 1, f1in(0) = 0 ⇒ f1in = B − Be−s + s ε2 d2f2in
ds2 + df2in
ds = 2s, f2in(0) = 0 ⇒ f2in = C − Ce−s + s2 − 2s εn d2fnin
ds2 + dfnin
ds = 0, fnin(0) = 0 ⇒ fnin = Dn − Dne−s,
Higher order matching
fin = A + (1 − A)e−s + ε(B − Be−s + s) +ε2(C − Ce−s + s2 − 2s) +
X∞
n=3
εnDn(1 − e−s).
The Prandtl’s condition can not be applied, because of existence of unbounded terms εs and ε2(s2 − 2s). They represent x and x2 − 2εx. Let us write the exact
solution in terms of s:
Higher order matching
fex = x2 + x + 2 − e−x/ε + ε
2(1 − x) − 2e−x/ε
= ε2s2 + εs + 2 − e−s + ε 2(1 − εs) − 2e−s
= 2 − e−s + ε(2 − 2e−s + s) + ε2(s2 − 2s).
Comparing with fin we get A = 2, B = 2, C = 0, Dn = 0,
n = 3, 4, . . .. How we can define A, B, C, Dn without of exact solution?
Higher order matching
Let us look on the Prandtl’s condition from the other point of view. Let x = O(ε1/2), then
fin = A+(1−A)e−s = A+o(1), fout = x2+x+2−e−x/ε = 2+o(1).
We recuperated A = 2. Let us do the same, but in slightly general form. We represent s = xε in two
variables x = tεα then s = ε1t−α . Let us represent two term expansion in terms of new variable t = x/εα,
α ∈ (0, 1) for t fixed. We get
fout = x2 + x + 2 + ε2(1 − x) = 2 + εαt + ε2αt2 + 2ε − 2ε1+αt
Higher order matching
Now we use s = ε1t−α and fined two term expansion of
fin in terms of ε1t−α .
fin = A + ε(B + s) = A + εαt + εB.
Matching with
fout = 2 + εαt + ε2αt2 + 2ε − 2ε1+αt
we have to ask ε2α = o(ε), or α > 1/2. We get A = 2 and
B = 2.
Higher order matching
The composition solutions is
f2compterm = f2outterm + f2interm − f2matchterm
where f2compterm is given by terms of order up to O(ε). In our case it is 2 + x + 2ε. We have
fcomp = x2 + x + 2 + 2ε(1 − x) +2 − e−x/ε + ε(2 − 2e−x/ε + x/ε)
−2 − x − 2ε = x2 + x + 2 − e−x/ε + ε(2 − 2x − 2e−x/ε).
Nonlinear examples
εd2f
dx2 + df
dx + f2 = 0, x ∈ (0, 1), f(0) = 2, f(1) = 1/2.
df0out
dx + (f0out)2 = 0, f0out(1) = 1/2 ⇒ f0out = 1 x + 1. s = x
ε ⇒ d2fin
ds2 + dfin
ds + ε(fin)2 = 0 d2f0in
ds2 + df0in
ds = 0 ⇒ f0in = A + (2 − A)e−s
Nonlinear examples
The Prandtl’s condition implies
xlim→0
1
x + 1 = lim
s→∞
A + (2 − A)e−s ⇒ A = 1.
fcomp = 1
x + 1 + e−x/ε.
Nonlinear examples
εd2f
dx2 + 2f df
dx − 4f = 0, x ∈ (0, 1), f(0) = 0, f(1) = 4.
2f0outdf0out
dx − 4f0out = 0, f(1) = 4 ⇒ f0out = 2x + 2 s = x
ε ⇒ d2f0in
ds2 +2f0indf0in
ds = 0 ⇒ d ds
df0in
ds +(f0in)2
= 0 df0in
ds + (f0in)2 = c = a2
since f0in(0) = 0 and lim
x 0 f0out = 2 and f0in increase
Nonlinear examples
• R df0in
a2−(f0in)2 = R
ds
• −2a1 R df0in
f0in−a + 2a1 R df0in
f0in+a = s + b
• − ln |f0in − a| + ln |f0in + a| = 2as + 2ab
• f0in−a
f0in+a = Ae2as, A = ±e2ab
• A = 1, f0in = aee22asas−+11 = a tanh as
• lim
x→0 f0out = lim
s→∞
f0in ⇒ a = 2
Nonlinear examples
• R df0in
a2−(f0in)2 = R
ds
• −2a1 R df0in
f0in−a + 2a1 R df0in
f0in+a = s + b
• − ln |f0in − a| + ln |f0in + a| = 2as + 2ab
• f0in−a
f0in+a = Ae2as, A = ±e2ab
• A = 1, f0in = aee22asas−+11 = a tanh as
• lim
x→0 f0out = lim
s→∞
f0in ⇒ a = 2
Nonlinear examples
• R df0in
a2−(f0in)2 = R
ds
• −2a1 R df0in
f0in−a + 2a1 R df0in
f0in+a = s + b
• − ln |f0in − a| + ln |f0in + a| = 2as + 2ab
• f0in−a
f0in+a = Ae2as, A = ±e2ab
• A = 1, f0in = aee22asas−+11 = a tanh as
• lim
x→0 f0out = lim
s→∞
f0in ⇒ a = 2
Nonlinear examples
• R df0in
a2−(f0in)2 = R
ds
• −2a1 R df0in
f0in−a + 2a1 R df0in
f0in+a = s + b
• − ln |f0in − a| + ln |f0in + a| = 2as + 2ab
• f0in−a
f0in+a = Ae2as, A = ±e2ab
• A = 1, f0in = aee22asas−+11 = a tanh as
• lim
x→0 f0out = lim
s→∞
f0in ⇒ a = 2
Nonlinear examples
• R df0in
a2−(f0in)2 = R
ds
• −2a1 R df0in
f0in−a + 2a1 R df0in
f0in+a = s + b
• − ln |f0in − a| + ln |f0in + a| = 2as + 2ab
• f0in−a
f0in+a = Ae2as, A = ±e2ab
• A = 1, f0in = aee22asas−+11 = a tanh as
• lim
x→0 f0out = lim
s→∞
f0in ⇒ a = 2
Nonlinear examples
• R df0in
a2−(f0in)2 = R
ds
• −2a1 R df0in
f0in−a + 2a1 R df0in
f0in+a = s + b
• − ln |f0in − a| + ln |f0in + a| = 2as + 2ab
• f0in−a
f0in+a = Ae2as, A = ±e2ab
• A = 1, f0in = aee22asas−+11 = a tanh as
• lim
x→0 f0out = lim
s→∞
f0in ⇒ a = 2
Nonlinear examples
f0in = a tanh 2s
fcomp = 2x + 2 tanh(2x/ε).