Добавил:
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:

Schechter Minimax Systems and Critical Point Theory (Springer, 2009)

.pdf
Скачиваний:
40
Добавлен:
22.08.2013
Размер:
2.13 Mб
Скачать

10.4. Some applications

 

 

115

where

 

 

 

 

 

 

 

t

 

(10.42)

F(x, t) =

0

f (x, s)ds.

 

From (10.33) it is readily verified that G is continuously differentiable on D and

(10.43)

(G (u), v)/2 = a(u, v) ( f (u), v),

u, v D,

where we write

f (u) in place of f (x, u) (cf., e.g., [112]). We note that G (u) satisfies

(10.5) on E = D : for, if uk u weakly in D, then a(uk , v) a(u,2v) for each

v D, and there is a renamed subsequence such that V uk

V u in L ( ). Since

V (x) > 0 in , there is a renamed subsequence such that uk

u a.e. in . Thus,

(10.44)

f (x, uk )v f (x, u)v a.e.

 

Since

| f (x, uk )v| ≤ |V uk V v| + W |V v|

and the right-hand side converges to |V uV v|+ W |V v| in L1( ), we see that ( f (uk ), v)

( f (u), v). Thus, G (u) satisfies (10.5). I claim that

(10.45)

G(v) → −∞

as v D → ∞,

v N,

(10.46)

G(w) → ∞

as w D → ∞,

w M,

where u D

= | A|1/2u . To prove (10.45), let

{vk } be a sequence such that

ρk = vk D → ∞, and let v˜k = vk k . Then v˜k D = 1, and there is a renamed subsequence that converges weakly in D and a.e. in to a function v˜ D (again using the properties of V ). I claim that

(10.47)

 

 

 

2

F(x, vk )dxk2 α(v˜)

 

 

and

 

 

 

 

 

 

 

 

 

 

(10.48)

G(vk )/ρk2 → −1 α(v˜) = ( v˜ 2D 1) ( v˜ 2D + α(v˜)) 0.

To see this, note that

 

 

 

 

 

 

 

 

 

2F(x, vk )

=

2F(x, vk )

v2

α (v+)2

+

α (v)2

a.e.

 

ρk2

 

 

 

 

vk2

˜k −→

+ ˜

˜

 

Moreover, by (10.33),

|F(x, vk )|k2 C(|V v˜k |2 + W |V v˜k |k ) −→ C|V v˜|2

in L1( ). This implies (10.47) and (10.48). Now, the only way the right-hand side of (10.48) can vanish is if

(10.49)

v˜ D = 1

116

 

 

10. Weak Linking

and

 

 

 

 

 

(10.50)

 

 

( Av˜, v˜) = α(v˜).

 

 

Let

 

 

(v) = ( Av, v) α(v).

 

 

Then

 

 

 

 

 

 

(v) 0, v N.

 

 

˜

=

˜

=

 

˜

0, i.e.,

If (v)

 

0, then v

is a maximum point for on N. This means that (v )

 

that

( Av˜, v) α(v˜, v) = 0, v N.

Thus, v˜ is a solution of (10.38). However, this contradicts (10.49). Hence, the righthand side of (10.48) must be negative. This proves (10.45). The proof of (10.46) is similar. Note that (10.46) implies

b0 = inf G > −∞,

M

for if G(wk ) b0, {wk } M, then wk D C by (10.46). But then (10.33) implies that G is bounded on bounded sets in D. From (10.45) we see that there is an R such that (10.1) holds with A = N BR , B = M, Q = N BR , and F = P, the orthogonal projection of D onto N, p = 0. If S is a finite-dimensional subspace of D such that P S = {0}, let S0 = P S. Clearly, (10.6) is satisfied. We can now apply Theorem 10.2 to conclude that there is a sequence {uk } D such that

(10.51)

G(uk ) c, b0 c a1, G (uk ) 0.

Assume first that ρk

= uk D → ∞, and write u˜ k = uk k . Then u˜k D = 1.

Consequently, there is a renamed subsequence such that u˜k u˜ weakly in D and a.e. in (using the properties of V ). Since

| f (x, uk )v|k ≤ |V u˜k V v| + W |V v|k

and the right-hand side converges in L1( ), we see by (10.34) that

(10.52)

 

 

( f (uk ), v)/ρk α(u˜ , v).

 

 

Thus,

(u

 

 

 

 

 

 

 

(G

k

), v)/2ρ

a(u, v)

α(u, v), v

 

D.

 

 

k

˜

˜

 

Since this limit vanishes by (10.51), we see that u˜ is a solution of (10.38). Hence, u˜ ≡ 0 by hypothesis. Let u˜k = v˜k + w˜ k , where v˜k N, w˜ k M. Arguments similar to that given in the proof of (10.47) show that

( f (uk ), v˜k )/ρk α(u˜ , v˜)

and

( f (uk ), w˜ k )/ρk α(u˜, w),˜

10.4. Some applications

117

where u˜ = v˜ + w˜ . Since v˜k 2D + w˜ k 2D = 1, we have for a renamed subsequence

(G

(u

), v

)/2ρ

γ

1

α(u, v ),

 

k

˜k

k

 

˜ ˜

(G (uk ), wk )/2ρk

γ2

α(u, w),

 

 

˜

 

 

 

˜ ˜

where γ2 + γ1 = 1. By (10.51), both of these expressions must vanish. But this cannot happen if u˜ ≡ 0. Thus, the ρk must be bounded, and consequently there is a renamed subsequence of {uk } such that uk u weakly in D and a.e. in . As before, this implies ( f (uk ), v) ( f (u), v) for v D. Hence,

(G (uk ), v) (G (u), v), v D.

By (10.51), this limit must vanish for each v in D. We conclude that G (u) = 0 and that u is a solution of (10.39).

 

2

 

 

 

Rn and let A

λ0

>

 

 

2

As another application, let

 

 

 

0 be a self-adjoint operator

on L

 

( ) with compact resolvent. Let F(x, s, t) be a function on × R such that

(10.53)

f (x, s, t) = F/∂ t,

g(x, s, t) = F/∂ s

 

are Carath´eodory functions satisfying

 

 

 

 

 

 

(10.54)

| f (x, s, t)| + |g(x, s, t)| ≤ C(|s| + |t| + 1),

s, t R,

 

and

 

 

 

 

 

 

 

 

 

 

 

(10.55)

f (x, s, t)/ρ α(x)s˜ + β(x)t˜,

 

 

(10.56)

g(x, s, t)/ρ γ (x)s˜ + δ(x)t˜

 

 

 

as ρ2 = s2 + t2 → ∞, ss˜, tg˜. We wish to solve the system

(10.57)

Av = f (x, v, w),

Aw = g(x, v, w),

v, w D( A).

 

We assume that the only solution of

 

 

 

 

 

 

 

(10.58)

Av = αv + βw,

Aw = γ v + δw

 

 

is v

w 0 in . If λ0 is an eigenvalue of A, we assume that corresponding

eigenfunctions are = 0 a.e. on . Finally, we assume that

 

 

 

(10.59)

β + γ α δ ≤≡ 2λ0,

β + γ + α + δ ≤≡ 2λ0.

 

118 10. Weak Linking

We have

Theorem 10.6. Under the above assumptions, system (10.57) has a solution.

Proof. Let D = D( A1/2), E = D × D. Then D becomes a Hilbert space with norm given by

(10.60)

u 2E = ( Av, v) + ( Aw, w), u = (v, w) D.

We define

 

 

(10.61)

G(u) = ( Av, w)

F(x, v, w) dx, u D.

In view of (10.54), G C1(D, R) and

 

(10.62)

(G (u), h) = ( Av, h2) + ( Aw, h1) ( f (u), h2) (g(u), h1)

for u = (v, w) E, h = (h1, h2)

E, and we write f (u), g(u) in place of

f (x, v, w),

g(x, v, w), respectively. It follows from (10.62) that u E is a solu-

tion of (10.57) if u satisfies

 

(10.63)

G (u) = 0.

Let N denote the subspace of E consisting of those u = (v, w) for which w = −v and let M consist of those for which w = v. Then M = N . We note that

(10.64)

G(v, v) → −∞ as v D → ∞

and

 

(10.65)

G(w, w) → +∞ as w D → ∞.

To prove (10.64), let {vk } D be such that ρk = vk D → ∞, and take v˜k = vk k .

Since v˜k D 2= 1, there is a renamed subsequence such that v˜k v˜ weakly in N,

strongly in L ( ), and a.e. in such that

 

(10.66) 2G(vk , vk )/ρk2 = −2 v˜k 2D 2

F(x, vk , vk )dxk2

→ −2 [γ δ α + β]v˜2dx

= − [2λ0 + β + γ α δ]v˜2dx + 2λ0 v˜ 2 2.

In view of (10.59), this is negative unless λ0 v˜ 2 = 1. But this would mean that v˜ E0), the eigenspace of λ0. If λ0 is not an eigenvalue, then we conclude immediately that

(10.67)

lim sup G(v, v)/ v 2D < 0

 

v D →∞

10.4. Some applications

119

and (10.64) holds. If λ0 is an eigenvalue, then v˜ = 0 a.e. by hypothesis. But then the last integral in (10.66) must be positive by (10.59). Again, this implies (10.67) and (10.64). A similar argument implies (10.65). Now that we have (10.64) and (10.65), we know that (10.1) holds for A = N BR and B = M when R is sufficiently large. If we let F be the projection of E onto N, we can take Q = N BR and p = 0.

We can conclude from Theorem 10.2 that there is a sequence {uk }

E satisfying

(10.2), (10.3). I claim that

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(10.68)

 

 

 

ρk = uk E C.

 

 

 

 

 

 

 

 

 

To see this, assume that ρk → ∞, and let u˜k

 

= uk k .2

Then there is a renamed

subsequence such that u˜k u˜ weakly in E, strongly in L

( ), and a.e. in . If h =

(h1, h2) E, then, by (10.62),

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(10.69) (G (u

k

), h)/ρ

( Av

, h

2

)

+

( Aw

, h

1

)

( f (u

k

), h

)/ρ

k

(h(u

k

), h

)/ρ ,

 

k =

˜k

 

 

˜ k

 

 

 

 

 

2

 

 

1 k

where uk = (vk , wk ). Taking the limit and applying (10.54) – (10.56) we see that u˜ = (v˜, w)˜ is a solution of (10.58). By hypothesis, u˜ ≡ 0. On the other hand, there is a renamed subsequence such that v˜k D a, w˜ k D b with a2 + b2 = 1. Moreover,

(10.70)

(G

k

˜k

,

− ˜k

k

= −

˜k D +

k

˜k

k

(u

), (v

v

))/ρ

 

2 v 2

 

( f (u

), v

)/ρ

 

 

 

 

 

 

 

 

(g(uk ), v˜k )/ρk

 

 

 

 

 

 

 

→ − 2a2 + [α + β γ δ]v˜2dx

and

 

 

˜

˜

 

= ˜

D

 

 

˜

 

 

 

 

 

 

 

 

(10.71)

(G (uk ), (wk , wk ))/ρk

2

wk

2

( f (uk ), wk )/ρk

(g(uk ), wk )/ρk

2b2 [α + β + γ + δ]w˜ 2dx.

By (10.2), both these limits are 0. Since a and b cannot both vanish, the same is true of v˜ and w˜ . Hence, u˜ ≡ 0, and this contradiction shows that (10.68) does not hold. We can now use the usual procedures to show that {uk } has a renamed subsequence

converging weakly in E to a function u, strongly in L2( ), and a.e. in . Then this limit is a solution of (10.63) and consequently of (10.57).

For another application, let A, B be positive, self-adjoint operators on L2( ) with compact resolvents, where Rn. Let F(x, v, w) be a function on × R2 such that

(10.72)

f (x, v, w) = F/∂ v, g(x, v, w) = F/∂ w

are Carath´eodory functions satisfying

(10.73)

| f (x, v, w)| + |g(x, v, w)| ≤ C0(|v| + |w| + 1), v, w R,

120

10. Weak Linking

and

 

(10.74)

f (x, t y, tz)/t α+(x)v+ α(x)v+ β+(x)w+ β(x)w,

(10.75)

g(x, t y, tz)/t γ+(x)v+ γ(x)v+ δ+(x)w+ δ(x)w

as t → +∞, y v, z w, where a± = max(±a, 0). We wish to solve the system

(10.76)

Av = − f (x, v, w),

(10.77)

Bw = g(x, v, w).

Let λ00) be the lowest eigenvalue of A(B). We assume that the only solution of

(10.78)

Av = α+v+ αv+ β+w+ βw,

(10.79)

Bw = γ+v+ γv+ δ+w+ δw

is v = w = 0. We have

Theorem 10.7. Assume that eigenfunctions of λ0 are = 0 a.e. on ,

(10.80)

α±(x) ≥≡ −λ0,

x ,

and

 

 

(10.81)

2F(x, 0, t) μ0t2 + W (x),

x , t R,

where W (x) L1( ). Then the system (10.76), (10.77) has a solution.

Proof. Let D = D( A1/2) × D(B1/2). Then D becomes a Hilbert space with norm given by

(10.82)

u 2D = ( Av, v) + (Bw, w), u = (v, w) D.

We define

 

 

(10.83)

G(u) = b(w) a(v) 2

F(x, v, w) dx, u D,

where

 

 

(10.84)

a(v) = ( Av, v),

b(w) = (Bw, w).

Then G C1(D, R) and

 

(10.85)

(G (u), h)/2 = b(w, h2) a(v, h1) ( f (u), h1) (g(u), h2),

where we write f (u), g(u) in place of f (x, v, w), g(x, v, w), respectively. It is readily seen that the system (10.76), (10.77) is equivalent to

(10.86) G (u) = 0.

10.4. Some applications

121

We let N be the set of those (v, 0) D and M the set of those (0, w) D. Then M, N are orthogonal closed subspaces such that

(10.87)

D = M N.

If we define

 

(10.88)

Lu = 2(v, w), u = (v, w) D

then L is a self-adjoint, bounded operator on D. Also,

(10.89)

G (u) = Lu + c0(u),

where

 

(10.90)

c0(u) = −( A1 f (u), B1g(u))

is compact on D. This follows from (10.73) and the fact that A and B have compact resolvents. Now, by (10.81),

(10.91)

G(0, w) b(w) μ0 w 2

W (x) dx, (0, w) M.

Thus,

 

 

 

(10.92)

M G ≥ −

W (x)dx.

inf

 

 

On the other hand, (10.80) implies

 

 

(10.93)

sup G → −∞ as R → ∞.

 

NBR

 

 

To see this, let (vk , 0) be any sequence in N such that ρk2 = a(vk ) → ∞. Then

(10.94) G(vk , 0)/ρk2 = −a(v˜k ) 2 F(x, vk , 0)dxk2,

where v˜k = vk k . Note that a(v˜k ) = 1. Thus, there is a renamed subsequence v˜k v˜ weakly in N, strongly in L2( ), and a.e. in such that

G(vk , 0)/ρ2

→ −

1

 

α

 

(v

+ )2

+

α (v

)2

}

dx

 

 

 

k

 

{

+

˜

˜

 

 

 

 

 

 

= −

 

{

0

+

α

+

)(v+)2

+

0

+

α

)(v)2

}

dx

 

 

 

 

˜

 

 

 

 

˜

 

+

This is negative unless λ0 v˜ 2 the eigenspace of λ0. Thus, v˜ vanish by (10.80). Hence,

λ0 v˜ 2 1.

= 1. Since a(v˜) 1, this would mean that v˜ E0),= 0 a.e. by hypothesis. But then the integral cannot

(10.95)

lim sup G(0, v)/a(v) < 0,

a(v )→∞

122

10. Weak Linking

and (10.93) holds. Since N is an invariant subspace of L, we can apply Theorem 10.2 to conclude that there is a sequence {uk } D such that (10.2) and (10.3) hold. Let uk = (vk , wk ). I claim that

(10.96)

 

 

ρk2 = a(vk ) + b(wk ) C.

 

 

 

To see this, assume that ρk

→ ∞, and let u˜k

= uk k .2

Then there is a renamed

subsequence such that u˜ k u˜ weakly in D, strongly in L

( ), and a.e. in . If h =

(h1, h2) D, then

 

 

 

 

 

 

 

 

 

(10.97)

 

2b(wk, h2)

 

2a(vk , h1)

 

2( f (uk ), h1)/ρk

 

2(g(uk ), h2)/ρk .

(G (uk ), h)/ρk

=

 

˜

 

˜

 

 

 

Taking the limit and applying (10.73) – (10.75), we see that u˜ = (v˜, w)˜ is a solution of

(10.78)

(10.79).

Hence,

u˜

 

=

 

0

by

 

hypothesis.

 

 

 

 

On

the

other hand,

since

a

˜

 

+

 

 

˜

 

 

=

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

˜

 

→ ˜

˜

 

b

 

(vk )

 

b(wk ) 1, there is a renamed subsequence such that a(vk )

 

 

a, b(wk )

 

˜

 

 

˜ +

b

=

1. Thus, by (10.74), (10.75), and (10.85),

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

with a

 

˜

 

 

 

 

 

 

 

˜k

 

 

 

k = −

 

˜ k

 

 

 

 

 

 

 

 

 

 

 

 

 

˜k

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(10.98)

 

(G

(u

k

 

 

 

 

 

 

 

)

 

( f (u

k

 

 

 

 

 

 

k

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

), (v

, 0))/2ρ

 

 

 

 

a (v

 

 

 

 

 

), v

 

 

)/ρ

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a

 

 

 

 

(α v+

α v

+

β

w+

β

w)vdx

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

→ − ˜ −

 

 

 

 

 

+

˜

 

 

 

 

 

˜

 

 

 

+

˜

 

 

˜ ˜

 

and

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(10.99)

 

(G

(u

k

), (0, w

))/2ρ

 

 

b (w

 

)

(g(u

k

), w

 

)/ρ

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

˜ k

 

 

 

k =

 

 

 

 

 

 

 

 

 

 

 

 

 

˜ k

 

 

 

 

k

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

b

 

 

 

 

 

v+

 

γ

v

+

δ

w+

δ

w)wdx.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

˜

 

 

 

 

 

+ ˜

 

 

 

 

 

 

˜

 

 

+

˜

 

˜ ˜

 

 

Thus, by (10.42),

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(10.100)

 

 

 

 

 

 

 

 

 

a

 

 

 

v+

 

 

v

 

+

β

+

w+

β

w)v dx

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

˜ = −

 

 

 

+ ˜

 

 

˜

 

 

 

˜

 

 

 

 

˜ ˜

 

 

 

 

 

 

 

 

 

and

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(10.101)

 

 

 

 

 

 

 

 

 

b

 

+

v+

γ

 

 

 

+

δ

w+

 

δ

w)wdx.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

˜ =

 

 

˜

 

 

− ˜

 

 

 

 

+

˜

 

 

 

 

 

 

˜ ˜

 

 

 

 

 

 

˜ ≡

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

˜

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0. This

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

b is not zero, we see that we cannot have u

Since one of the two numbers a,

˜

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

contradiction proves (10.96). Once this is known, we can use the usual procedures to show that there is a renamed subsequence such that uk u in D, and u satisfies (10.86).

Corollary 10.8. In Theorem 10.7 we can replace (10.80), (10.81) with

(10.102)

2F(x, v, 0) ≥ −λ0v2 W (x), x , v R,

provided eigenfunctions of μ0 are = 0 a.e. in and

(10.103)

δ±(x) ≤≡ μ0.

10.4. Some applications

123

Proof. We interchange the roles of M and N in the proof of Theorem 10.7.

Theorem 10.9. If λ0 is simple and the eigenfunctions of λ0 and μ0 are bounded and= 0 a.e. in , we can replace (10.81) in Theorem 10.7 with

(10.104)

2F(x, t, 0) ≥ −λ1t2, t R,

(10.105)

2F(x, s, t) μ0t2 λ0s2, |t| + |s| ≤ δ,

where λ1 is the next eigenvalue of A and δ > 0. Moreover, system (10.76), (10.77) has a nontrivial solution.

Proof. Let N be the orthogonal complement of N0 = {ϕ0} in N. Then N = N N0 and

(10.106)

 

G(v , 0) = −a(v ) 2

F(x, v , 0) dx ≤ −a(v ) + λ1 v 2 0, v N ,

and

 

(10.107)

G(v, 0) → −∞ as a(v) → ∞

by (10.80) [cf. (10.93)]. Let M0 be the subspace of M spanned by the eigenfunctions of B corresponding to μ0, and let M be its orthogonal complement in M. Since N0 and M0 are contained in L( ), there is a positive constant ρ such that

(10.108)

a(y) ρ2

y δ/4,

y N0,

(10.109)

b(h) ρ2

h δ/4,

h M0,

where δ is the number given in (10.105). If

 

(10.110)

a(y) ρ2, b(w) ρ2, |y(x)| + |w(x)| ≥ δ,

we write w = h + w , h M0, w M , and

 

(10.111) δ ≤ |y(x)| + |w(x)| ≤ |y(x)| + |h(x)| + |w (x)| ≤ (δ/2) + |w (x)|.

Thus,

 

(10.112)

|y(x)| + |h(x)| ≤ δ/2 ≤ |w (x)|

and

 

(10.113)

|y(x)| + |w(x)| ≤ 2|w (x)|.

124

 

 

 

 

10. Weak Linking

Now, by (10.105) and (10.113),

 

 

 

G(y, w) =b(w) a(y) 2

F(x, y, w)dx

 

b(w) a(y)

 

{μ0w2 λ0 y2}dx

 

 

 

|y|+|w|

 

c0

 

(|y| + |w| + 1)dx

 

|y|+|w|

 

 

 

b(w) a(y) μ0 w 2 + λ0 y 2 c1

|w |4dx

 

 

 

 

2|w |

 

b(w ) μ0 w 2 c2b(w )2

 

 

μ0

 

 

 

1

 

c2b(w ) b(w )

 

μ1

 

when

a(y) ρ2,

b(w) ρ2,

 

 

 

where μ1 is the next eigenvalue of B after μ0. If we reduce ρ accordingly, we can find a positive constant ν such that

(10.114)

G(y, w) νb(w ), a(y) ρ2, b(w) ρ2.

I claim that either

(a) (15.48), (15.49) has a nontrivial solution or (b) there is an

> 0 such that

 

(10.115)

G(y, w) , a(y) + b(w) = ρ2.

To see this, suppose (10.115) did not hold. Then there would be a sequence {yk , wk } such that a(yk ) + b(wk ) = ρ2 and G(yk , wk ) 0. If we write wk = wk + hk , wk

M , hk M0, then (10.114) tells us that b(wk ) 0. Thus, a(yk ) + b(hk ) ρ2.

Since N0, M0 are finite-dimensional, there is a renamed subsequence such that yk y

in N0 and hk h in M0. By (10.108) and (10.109), y δ/4 and h δ/4. Consequently, (10.105) implies

(10.116)

2F(x, y, h) μ0h2 λ0 y2.

Since

 

 

(10.117)

G(y, h) = b(h) a(y) 2

F(x, y, h) dx = 0,

we have

 

 

(10.118)

{2F(x, y, h) + λ0 y2 μ0h2}dx = 0.

Соседние файлы в предмете Экономика