Teschl Errata

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Errata

G. Teschl,

Ordinary Differential
Equations and
Dynamical Systems
Graduate Studies in Mathematics, Vol. 140
American Mathematical Society, Providence, Rhode Island, 2012
The official web page of the book:
http://www.mat.univie.ac.at/~gerald/ftp/book-ode/

Please send comments and corrections to


Gerald.Teschl@univie.ac.at

Updated as of October 14, 2022

Errata
Changes appear in yellow. Line k+ (resp., line k−) denotes the kth line from
the top (resp., the bottom) of a page.
My thanks go to the following individuals who have contributed to this
list: Theresa Dvorak, Gudrun Szewieczek, William Jagy, Jonathan Eckhardt,
Sebastian Woblistin, Yu Jiang, Annemarie Luger, Johannes Wächtler, Daniel
Scherl, Kristoffer Varholm, Constantino Santos, Minjae Park, Teck-Cheong Lim,
Peter Elbau, Walter Schachermayer, Sebas Pedersen, Raphael Stuhlmeier, Eric
Wahlén, Lukas Peham, Jakob Holböck, Guillaume Berger, Janusz Mierczynski,
Batuhan Bayır, Mark Homs Dones, Boyuan Shi, Johannes Rotheneder, Iryna
Karpenko, Long Xiaohuan, Malkiel Shoshan, Serge Richard, Jock Annelle, Pablo
Amster, Ricardo J. Alonso-Blanco.
Page 19. Problem 1.26, second item: y(x0 +x) = y(x0 − x)
Page 30.  
ψ̇(t, x) = 1 − 2φ(t, x) ψ(t, x)− 1 − sin(2πt) . (1.80)

Page 40.
t0 +T0
L1 (T0 )m L1 (T0 )
Z
sup |x(t) − K m (x0 )(t)| ≤ e |f (s, x0 )|ds. (2.28)
t0 ≤t≤t0 +T0 m! t0

Page 45. 10+: ∆(t) = |φ(t, t0 , y0 ) − φ(t, s0 , y0 )| and use . . .

1
Page 51. Theorem 2.13: Suppose the IVP (2.10) has a unique local solution
for every (t0 , x0 ) ∈ U
Page 57. 4+: Taking m → ∞ we finally obtain
Page 62.
k−1
X j t j
exp(tJ) = exp(αtI) exp(tN ) = eαt N . (3.18)
j=0
j!

Page 64. Proof of Lemma 3.3; We are looking for generalized eigenvectors uk
such that exp(N )uk = uk + uk−1 , that is,
n
X 1
uk,l = uk−1,j , 1 ≤ k ≤ n, 1 ≤ j ≤ n,
l=j+1
(l − j)!

with u0 = 0.
..
.
n
X
S(l, k)s(k, j) = δj,l , 1 ≤ j, l ≤ n.
k=j

Then, by construction, U −1 exp(N )U = I+N and the claim follows.


Page 73. Problem 3.13: It should read deg(q(t))≤ deg(p(t)) + s where s is the
size of the largest Jordan block corresponding to the eigenvalue β. Moreover,
here is an extended hint: Rt
(Hint: Investigate (3.48) using the following fact: (t − s)m p(s)eβs ds =
q(t)eβt , where q(t) is a polynomial of degree deg(q) = deg(p) + m if β 6= 0 and
R t k βs
deg(q) = deg(p) + m + 1 if β = 0. To see this differentiate s e ds with
respect to β.)
Page 84. Line before (3.98): Let X(t) be the identity matrix with the first
column replaced by φ1 (t),
Page 89. Example:
2 2 2 2 2 2
et c̈(t) + 4tet ċ(t) + (2 + 4t2 )et c(t) − 2t et ċ(t) + 2tet c(t) − 2et c(t)
 

Page 90. Problem 3.34: Consider the equation ẍ + q0 (t)x= 0.


Page 90. Problem 3.37:
1
Rt
y(t) = Q(t)−1 x(t), Q(t) = e− n qn−1 (s)ds
.
n−2 n  
(n) −1
XX j
y + Q(t) qj (t)Q(j−k) (t)y (k) = 0,
k
k=0 j=k

2
Page 90. Problem 3.38:

ẏ + e−Q(t) y 2 + eQ(t) q0 (t)= 0.

Page 100. Paragraph after the proof of Theorem 3.23: . . . is constant. To this
end recall that Corollary 3.5 tells us when the system corresponding to B(t) = 0
is stable . Moreover, . . .
Page 101. Theorem 3.26:
δ
|x(t)| ≤ Ce−(α−b0 C)t |x0 |, |x0 | < , t ≥ 0. (3.168)
C

Page 106. Take


U = ((A − α)n−1 u, . . . , (A − α)u, u), (3.191)

Page 108. Lemma 3.34: Moreover, if A is real and all Jordan blocks correspond-
ing to negative eigenvalues come in pairs, then there is a real logarithm.
Also adapt the first paragraph of the proof according to: Since the eigenval-
ues of A2 are the squares of the eigenvalues of A (show this), it remains to show
that B is real if all Jordan blocks corresponding to negative eigenvalues come in
pairs. In this respect note that a pair of Jordan blocks J, J corresponding to a
negative eigenvalue can be regarded as a pair J, J ∗ of complex conjugate blocks
and hence is covered by the analysis below (as the case ϕ = π).
Page 117. Lemma 4.4: In this case we have α = − limz→0 z p(z) and the radius of
convergence for the Laurent series of p(z) equals the radius of the largest ball
on which h(z) is analytic and nonzero.
Page 120. Proof of Theorem 4.5: An argument that h2 (z) has the same radius
of convergence is missing:
If there is a second solution of the form u2 (z) = z α2 h2 (z) the same argument
can be used for h2 (z). Otherwise one has to use (4.56) below in place of (4.39).
It is also possible to use u2 (z) = c(z)u1 (z) with
 Z z   Z z 
c0 (z) = z −2α1 −p0 h1 (z)−2 exp − p̃(z) = u1 (z)−2 exp − p(z) ,

where p̃(z) = p(z)− pz0 is the analytic part of p(z) near 0. However the primitive
c(z) might have logarithmic terms at a zeor z0 of u1 (z). A computation shows
Rz 
exp − 0 p(z) u001 (z0 ) + p(z0 )u01 (z0 )

0 1
c (z) = + + ···
u01 (z0 )2 (z − z0 )2 z − z0

and since u001 (z0 ) + p(z0 )u01 (z0 ) = 0 we conclude that c has a first order pole at
z0 which will be removed by u1 (z).
Alternatively, note that the present theorem will also follow as a special case
of Theorem 4.13.

3
Page 123.
(−1)j
h1,2j = , h1,2j+1 = 0, (4.64)
4j j!(ν + 1)j

Page 124.
1
h2,2j = , j < n. (4.69)
4j j!(n − j)j

Page 124.

2n (n − 1)! γ − log(2)
Yn (z) = − u2 (z) + n−1 u1 (z)
π 2 πn!
n−1 j
2 z 1 X (−1) (n − j − 1)!  z 2j−n
= (γ + log( ))Jn (z) −
π 2 π j=0 j! 2

1 X (−1)j (Hj+n + Hj )  z 2j+n
− (4.75)
π j=0 j!(j + n)! 2

(−1)n
Page 126. Problem 4.5: Γ(z) = n!(z+n)
+ O(1).

Page 126. Problem 4.8:


j−1
1X
hj = − pj−k hk ,
j
k=0

Page 127. Problem 4.10: For (i) you can use that Γ(z) has no zeros and hence
Γ(z)−1 is an entire function.
Page 127. Problem 4.11 (iii): Jν+1 (z) − Jν−1 (z) = −2Jν0 (z)
Page 137. Last paragraph: For all eigenvalues α of A0 for which α + j is not an
eigenvalue for all j ∈ N,
Page 145. Problem 5.3: We have x ∈ [0, 1] and there is only one boundary
condition u(t, 1) = 0.
Page 145. Problem 5.5:

d2
m u(t, n) = −k(u(t, n + 1) − u(t, n))−k(u(t, n − 1) − u(t, n)),
dt2

Page 154.
Z x
y(x) = y(x0 )c(z, x, x0 ) + p(x0 )y 0 (x0 )s(z, x, x0 )− s(z, x, t)g(t)r(t)dt. (5.50)
x0

−u(x)v(x0 )+u(x0 )v(x)


s(z, x, x0 ) = . (5.51)
W (u, v)

4
Page 155. Problem 5.13:

q(x(y)) (p(x(y))r(x(y)))1/4 0
Q(y) = − p(x(y))((p(x((y))r(x(y)))−1/4 )0 .
r(x(y)) r(x(y))

Page 162. Lemma 5.12:


q(x)
min ≤ E0 . (5.77)
x∈[a,b] r(x)
q(x)
And accordingly in the proof if Lemma 5.12: Then we have Q(f ) ≥ minx∈[a,b] kf k2
r(x)
q(x)
and hence (5.73) implies Q(uj ) = Ej ≥ minx∈[a,b] .
r(x)

q(x)
Page 162. Proof of Lemma 5.13: for q0 < minx∈[a,b] .
r(x)

Page 166. Problem 5.22: with f (2j) (0) = f (2j) (1) = 0 for 0 ≤ j ≤ k.
Page 168. In fact, θb (λ, x) as defined in (5.89) is the Prüfer angle for −ub (λ, x),
but this will be of no importance for our purpose.
Page 169.
   
θa (λ, b) − β α − θb (λ, a)
#(−∞,λ) (L) = = − 1, (5.91)
π π

Moreover, note that we also have:


   
θa (λ, b) − β α − θb (λ, a)
#(−∞,λ] (L) = +1= ,
π π

Page 172. Replace “As in the case of Theorem 5.18 one proves” by ”As an
immediate consequence of Lemma 5.16 we obtain”
Page 174. Problem 5.29: suppose 0 ≤ α2 < α1 < π and show
Page 175. Problem 5.30:
 
0 1 1
W (u0 , u1 ) = (q1 − λ1 r1 − q0 + λ0 r0 )u0 u1 + − p0 u00 p1 u01 . (5.111)
p0 p1

Page 175. Problem 5.31:

cos(x + b) (−3/4 − ν 2 ) sin(x + b)


 
0 −5/2
u (x) = a + + O(x ) .
x1/2 2x3/2

Page 177. Proof of Lemma 5.29: Then q − λr > 0 and any positive solution u
of (5.43) with z = λ satisfies (pu0 )0 = (q − λr)u > 0.

5
Page 183. Problem 5.33:
(
W (z)−1 u+ (z, x)u− (z, y)±ρ+ (z)u− (z, x)u+ (z, y), y < x,
G± (z, x, y) =
1 ∓ ρ+ (z) u− (z, y)u+ (z, x)±ρ+ (z)u+ (z, y)u− (z, x), y > x,

with W (z) = W (u+ (z), u− (z)).


Page 184. Problem 5.35: (iii) c(z, `) = s0 (z, `).
Page 190. 9+: Furthermore, K j+1 ∈ C k (Uj , M ) for any
Page 192. First paragraph of Section 6.3: Otherwise x is called regular and
Φ(., x) : Ix ,→ M is an immersion.
Page 195. Replace the 3rd line in the Example by:

Since |f (x)| ≤ 2(1 + |x|) the vector field is complete by Theorem 2.17. The
zeros are
Page 204.
x
x21
Z

sign(x1 ) p = t, E= + U (x0 ). (6.47)
x0 2(E − U (ξ)) 2

Page 207. Problem 6.24:

∂ ∂3 ∂
u(t, x) + u(t, x) + 6u(t, x) u(t, x)= 0
∂t ∂x3 ∂x

Page 208. Problem 6.25 should be changed according to:


Show that if {(ẋ, x) ∈ M |E(ẋ, x) = E0 } is compact and contains no fixed
point, then it corresponds to a periodic orbit. Conclude that all solutions are
periodic if lim|x|→∞ U (x) = +∞ and U has a unique minimum.
Page 211 and 212. Thy y axes in Figure 7.2 and Figure 7.3 are placed incorrectly.
They should be on the left, such that all orbits are in the first quadrant.
Page 213. Proof of Theorem 7.4: Exchange the definitions of Q3 and Q4 .
Page 214. Problem 7.4: The trajectory enters Q3 and satisfies x(t) < x0 in Q3
since . . . where y(t) decreases, implying x(t)>x1 = 1−yλ
1
when . . . . If y2 ≥ y0 ,
that is, if
1+µ
λµ2 ((µλ)2 − 1)(x0 − ) > 0, (1)
1 + µλ

Page 218. Paragraph after (7.28): Clearly, for y0 sufficiently small we have
∆(y0 ) > 0.
Page 222. The proof of Lemma 7.13 only shows that ωσ (x) contains a regular
periodic orbit. However, the claim follows from Lemma 7.14 if ωσ (x) is con-
nected. Moreover, connectedness follows from compactness as in the proof of

6
Lemma 6.6. In particular, the connectedness assumption in Theorem 7.16 is
superfluous.
Page 225. First equation in the example:
 
y
f (x, y) = ,
−ηE(x, y)2 y − U 0 (x)

Page 231. Sentence before (8.9): Moreover, for any positively invariant neigh-
borhood U ⊆ W + (Λ) of Λ we have
Page 233. Problem 8.2: Let VR = {x ∈ M |L(x)<R} be a relatively compact
set
Page 235. Replace the last sentence by: Using the Routh–Hurwitz criterion one
can show that the two new fixed points are asymptotically stable for 1 < r <
σ(3+b+σ)
σ−b−1 if 1 + b < σ and 1 < r if 1 + b ≥ σ. For the classical values σ = 10,
b = 8/3 this gives 1 < r < 470/19 = 24.74.
Page 239 after (8.34): where r(t0 ) = r(t1 ) = 0, we see that a necessary condition
for q to be extremal is that
Page 240.  
d p
= +grads H(p, q), (8.46)
dt q

Page 242. Problem 8.12 should be changed to:


The Lagrangian of a relativistic particle in an external force field is given by
r
2 v2
L(v, q) = −mc 1 − 2 − U (q),
c
where c is the speed of light, m the (rest) mass of the particle, and U the
potential of the force field. Derive the equation of motions from Hamilton’s
principle. Derive the corresponding Hamilton equations.
Page 244.
∂S1
(P, Q) = (− (q, Q(p, q)), Q(p, q)). (8.63)
∂Q
Page 246.
1
H(p, q) = (pM −1 p + qW q) (8.77)
2
Page 246. Line before equation (8.78). Then the symplectic transform (P, Q) =
(V T M −1/2 p, V T M +1/2 q) (Problem 8.15) gives the decoupled system
Page 246.
n 2 n
X pj X
H(p, q) = + U0 (qj+1 − qj ), q0 = qn+1 = 0, (8.81)
j=1
2m j=0

7
Page 246. In order to better explain what Problem 8.18 is about, the text
between ”If we assume that the particles . . . of the Jacobian matrix of the
potential.” should be replaced by:
If we assume that the particles are coupled by springs, the potential would
be U0 (x) = k2 x2 , where k > 0 is the so called spring constant, and we have a
harmonic oscillator with
2 −1
 

−1 2 . . .
 

 
M = mI, W = k
 .. .. .. .

 . . . 
 .. 
 . 2 −1
−1 2

The motion is decomposed into n modes corresponding to the eigenvectors of


W , which are given by
 
sin(ηj )
 sin(2ηj ) 
r
2   πj
vj = .. , ηj = .
n+1 n+1

. 
sin(nηj )
η
The corresponding eigenvalues are mωj2 , where ωj2 = 2k 4k j 2
m (1−cos(ηj )) = m sin( 2 ) .
Consequently the j’th mode corresponds to the initial condition (p(0), q(0)) =
(0, v j ) and is given by

q j (t) = cos(ωj t)v j , pj (t) = −mωj sin(ωj t)v j .

mωj2
The energy of the j’th mode is H(pj , q j ) = 2 .

Page 247. Problem 8.16: Ignore the hint.


Page 249.
Z r+  −1/2 r  3/2
T µ 1 1  1 1 µ p
= − − dr = π , (8.97)
2 l0 r− r r+ r− r γ 1 − ε2

Page 257. Bottom of the page: The requirement x ∈ U (x0 ) could be made part
of the definition of M ±,α . Then the intersection with U (x0 ) can be dropped at
various later points.
Page 259. Theorem 9.3: there are neighborhoods U (x0 ) of x0 and U of 0 and a
function h+,α ∈ C k (E +,α ∩ U, E −,−α ) such that
Page 261. Theorem 9.4: there are neighborhoods U (x0 ) of x0 and U of 0 and
functions h± ∈ C k (E ± ∩ U, E 0 ⊕E ∓ ) such that

8
Page 261. Theorem 9.5: Delete equation (9.22) from the statement and add the
following remark after the theorem:
Note that while we always have M ± (x0 ) ⊆ W ± (x0 ) ∩ U (x0 ), equality might
not hold even in the case of a hyperbolic fixed point. In fact, W ± (x0 ) might
also contain points from M ∓ (x0 ) as the example of a homoclinic orbit shows.
Page 262. Theorem 9.6: there are neighborhoods U (x0 ) of x0 and U of 0 and
functions h± ∈ C k (E ± ∩ U × Λ, E 0 ⊕E ∓ ) such that
Page 265. After (9.32): We will investigate this equation in the Banach space
of bounded continuous functions C(Rn , Rn ) with the sup norm.
(Remark: Also the notation Cb (Rn , Rn ) might be more appropriate here.)

Page 266. Proof of Lemma 9.7: The equation A−1 ◦ ϕ ◦ ϑ = ϕ ◦ ϑ ◦ A−1 should
read f ◦ ϕ ◦ ϑ = ϕ ◦ ϑ ◦ f . This last equation implies ϕ ◦ ϑ = I + l, where l is a
solution of Ll(x) = g(x) − g(x + l(x)). Using the estimates for the inverse of L
and for g one obtains l ≡ 0 and thus ϕ is a homeomorphism.
Page 268. The very last equation on the bottom of the page is only true if Φt
is linear. Set Φt = etA + Gt , where Gt is bounded, and replace this equation by

ht = Φt ◦ ϕ ◦ e−tA − I = etA ◦ h ◦ e−tA + Gt ◦ ϕ ◦ e−tA ,

where both terms are bounded.


Page 268.
ϕ(x) = (x1 +x22 , x2 ). (9.38)

Page 270, Problem 9.13: and that I−A is invertible if kAk < 1
Page 273.
Cδ = {Bδ (k(t, λ)) |(t, λ) ∈ [−T, T ] × Λ} ⊂ V (9.52)

Page 275. κ should be a matrix and |κ| should be read as the corresponding
matrix norm.
Page 286. Line before (10.19) which is zero at p, positive for x 6= p, whose level
sets Sδ = {x ∈ U (p)|L(x) ≤ δ} are connected for sufficiently small δ, and such
that x ∈ U (p) implies f n (x) ∈ U (p) and
Page 288.
m−1
X
x(m) = Π(m, m0 )x0 + Π(m, j+1)g(j), (10.28)
j=m0
m0
X
x(m) = Π(m, m0 )x0 − Π(m, j+1)g(j), m < m0 . (10.29)
j=m−1

Page 290. Theorem 10.5: there are neighborhoods U (p) of p and U of 0 and
functions h± ∈ C k (E ± ∩ U, E 0 ⊕E ∓ ) such that

9
Page 298. Second sentence after (11.10): Change W s to W + .
Page 298. Third paragraph: A map f as above is called topologically transi-
tive if for any given nonempty open sets U, V ⊆ M there is an n ∈ N such that
f n (U ) ∩ V 6= ∅.
Page 300. Sentence after (11.14): Moreover, each of the intervals In,j =
2j 2j+1
[ 2n , 2n ] is mapped . . .
Page 300. Problem 11.3: The assumption that the set has no isolated points
needs to be added.
Page 301. Paragraph after (11.19): Moreover, since the endpoints of the subin-
−n
tervals of Λn are just given by Tµ ({0, 1}), we see . . .

Page 306. Lemma 11.10: the number of periodic points of period at most l is
equal to tr(Al ).
Page 308. Problem 11.10: the number of periodic orbits of period at most n.
Page 311. Theorem 11.20: The Hausdorff dimension of the repeller Λ of the
tent map Tµ for µ > 2 is
Page 316. Problem 11.13: It seems too difficult to give a counterexample. The
problem should be ignored.
∂ΦT
Page 318. Line after (12.7): Mx0 (t0 ) = exp(T Qx0 (t0 )) = ∂x (Φ(t0 , x0 ))
Page 323. Theorem 12.9: Delete equation (12.25) from the statement.
Page 324. Problem 12.1:

ẋ = −y + µ + σ(x2 + y 2 ) x, ẏ = x + µ + σ(x2 + y 2 ) y
 

Page 324.
∆(q, ε) = Φ(τ (q, ε), (0, q), ε) − q. (12.31)

Page 335. Change W s to W + and W u to W − in the picture and the text before
the picture.
Page 339.
Z ∞
M (t) = p0 (s) (δp0 (s) + γ cos(ω(s − t))) ds
−∞
4δ √ πω
= − 2πγω sech( ) sin(ωt). (13.26)
3 2
Thus the perturbed Duffing equation is chaotic for ε sufficiently small provided
Page 339. Problem 13.3:

q̇ = p, ṗ = − sin(q) + ε sin(t).

10

You might also like