Qual Problems, Fall 2006

Fall 2006 Qual, Part I: 1, 2, 3, 4; Part II: 1

I.1 Let G be an unbounded open set in (0, ∞). Define

D = { x ∈ (0, ∞) : nx ∈ G for infinitely many n}.

Prove that D is dense in (0, ∞).

Consider the sets
Dk = { x ∈ (0, ∞) : nx ∈ G for some n > k }.
We claim that Dk is open and dense for each k.
That Dk is open follows from the openness of G: If x ∈ Dk , then there is some n > k
such that nx ∈ G. G is open, so there is some e > 0 such that (nx − e, nx + e) ⊂ G. Thus
if |y − x | < e/n, we must have y ∈ Dk as well, and so Dk is open.
Moreover, that Dk is dense follows from the unboundedness of G: For x ∈ (0, ∞) and
e > 0, we claim that the set
∪n>k (nx − ne, nx + ne)
contains a half-line. This follows from the claim that if N is large enough, then

(n( x − e), n( x + e)) ∩ ((n + 1)( x − e), (n + 1)( x + e)) 6= ∅

for all n > N, because then we would have ( N ( x − e), ∞) is contained in that union. To
−e
see the claim, just observe that if n > x2e , then

n( x + e) − (n + 1)( x − e) = 2en − ( x − e) > ( x − e) − ( x − e) > 0,

so that the two intervals must overlap (since clearly (n + 1)( x + e) > n( x − e)). Thus the
set
∪n>k (nx − ne, nx + ne)
contains a half-line, and so must have non-trivial intersection with G. In particular, for
each x ∈ (0, ∞) and e > 0, there is some y ∈ ( x − e, x + e) and n > k such that ny ∈ G,
and so Dk is dense.
Observe now that

D = { x : nx ∈ G for infinitely many n} = ∩k Dk .

We showed above that each Dk is open and dense, and then the Baire category theorem
tells us that D must be dense. 

1

1]) such that each φn ∈ C0 ([0. 1]) but f ∈ / L2 ([0. Suppose that A : H1 → H2 is a continuous injective linear map. and so would have to be given by an element of L2 by Riesz-Fischer. We may thus find gn ∈ C ([0. 1]). Tgn | Tgn | so hn → h in L2 . Av j(k) → w strongly and so weakly. and weak limits are unique.I. This is the main idea of this problem. 1])-norm. so we must have Av = w. 1]). Tgn Observe first that Thn = 0. 1]) and such that Z 1 f ( x )φn ( x )dx = 0 for all n. and so we know that the unit ball is weakly compact. 1]) : f φ = 0} is a dense subset of L2 ([0. Suppose that {v j } is a bounded sequence in H1 such that Av j converges strongly in H2 to some element w. 1]) such that k gn k L2 = 1 but Tgn → ∞. This means that there must be a weakly convergent subsequence of the v j . 1]). 1]). 1]) (or indeed. so it would have to be continuous on all of L2 . 0 We begin with the observation that if S is a dense subspace of L2 ([0. 2 . This reduces the problem to showing that Z S = {φ ∈ C ([0. so we know that the set S must be dense in L2 . 1]) so this makes sense. 1]). R so this cannot be continuous with respect to the L2 ([0. finisihing the proof. Now note that Th | Th| k h − h n k L2 = k gn k L2 = → 0. 1]) is dense in L2 . Hilbert spaces are reflexive. Prove that there exists an element v ∈ H1 such that v j converges weakly to v and Av = w. Now. so we know that Av j(k) → Av weakly. so hn ∈ ker T. We already know that bounded operators are also continuous in the weak topol- ogy. any Hilbert space) then we may extract a complete orthonormal basis φn of L2 such that φn ∈ S. consider Th hn = h − gn . The secondary idea of this problem is that S is the kernel of a linear functional that is not continuous and so it cannot be closed (so we have a chance).  I. This tells us that we may approximate any continuous function in L2 - norm by an element of ker T. but f ∈ / L2 ([0.3 Let H1 and H2 be two separable Hilbert spaces. For φ ∈ C ([0. We’ll show that it is actually dense. Find a complete orthonormal basis {φn } for L2 ([0. say v j(k) → v weakly. let Tφ = f φ. given h ∈ C ([0.2 Suppose that f ∈ L1 ([0. We may approximate any L2 function by a continuous one. f ∈ L1 ([0. This is because C ([0.

In other words. 1]). It only remains to prove that v j → v weakly. Does Tk extend as a continuous linear functional to all of C0 ([0. We know from the binomial theorem that for n ≥ k. so we must have that u = v. and let 1 1 φ j ( x ) = φ ( j ( x − ) + ). 2. every subsequence of v j has a weakly convergent sub-subsequence that converges to v. k which tends to infinity as n → ∞. but supp φj ⊂ [ 12 − 1j .  3 . 1]) be a compactly supported function such that φ(0) = φ(1) = 0 and φ(1/2) = 1. and so cannot extend to a functional on C0 ([0. Indeed. 2. and so we must have that v j → v weakly. A is injective. we may find a weakly convergent sub-subsequence v j(k(i)) such that Av j(k(i)) → w. n ≥ k. we consider f n ( x ) = (1 − x )n . 1] so as to consider P ⊂ C0 ([0. 1])? Hint: Consider f n ( x ) = (1 − x )n . 21 + 1j ]. Indeed. let φ ∈ C0 ([0. Let T : C0 ([0. we have that k)kC0 ([0. but 1 k φ j k L2 = p . Extend φ to be 0 outside of [0. j so T cannot be bounded with respect to the L2 norm. where u is such that Au = w. where ak is the coefficient of x k . This is where we use the injectivity of the map A. and so v j(k(i)) → u weakly.  I. Tk ( f ) = ak . 1.1 ] f n = 1 because |1 − x | ≤ 1 for x ∈ [0. On the other hand. Define the linear functional on P. We then have that T (φj ) = 1 for all j. 2 2 so that φj ( 21 ) = 1 for all j. Let P denote the set of polynomials of arbitrary degree. and consider their restric- tions to [0. Is T continuous with respect to the L2 norm? Explain why or why not. As suggested by the hint.   n Tk ( f n ) = . 1]) → C be defined by T ( f ) = f (1/2).4 Some problems about linear functionals: 1. T is not continuous with respect to the L2 norm. 1]. 1]). for any subsequence of v j . Thus Tk is not bounded with respect to the C0 norm on P. 1].

First suppose that c ∈ `2 . let v j ∈ Hj be such that kv j k = c j . 1. let Ik x = ∑ Π Hj x. Now.H is the norm limit of the finite rank operators Ik . there is another orthogonal 4 . . j≤k where Π Hj is the orthogonal projector onto Hj . we know it must be bounded. Indeed. ) where each c j > 0. We denote this decomposition H = { H j } j=1 . Note then that ( I − Ik ) x = ∑ Π Hj x. so c ∈ `2 . if Ac. .II. and so Ac. Prove that every compact subset K ⊂ H is contained in some Ac. .H = {v : kv j k ≤ c j } ⊂ H. Thus there is some constant C such that k x k2 ≤ C for all x ∈ Ac. Thus each v ∈ H L can be written uniquely as v = ∑∞ j=1 v j with v j ∈ H j . H = ∞ ∞ j=1 H j . j≤k We know that uk ∈ Ac. We claim that the identity operator on Ac. and define the subset Ac.H is compact in H. j>k Using that x ∈ Ac. Thus I | Ac.H is compact.H gives us that k( I − Ik ) x k2 = ∑ kΠ Hj xk2 ≤ ∑ c2j → 0 j>k j>k as k → ∞ because c ∈ `2 . 2. Consider the sequence uk = ∑ v j .H must be compact. We will prove the statement that given any orthogonal decomposition of H into fi- nite dimensional subspaces and any compact set K ⊂ H. and that C ≥ k u k k2 = ∑ kv j k2 = ∑ c2j . c2 . Prove that c ∈ `2 if and only if Ac.H for some H and c ∈ `2 . 2.H can be ap- proximated by finite rank operators Ik . 1. For each j.H . j≤k j≤k so we must have that ∑ c2j ≤ C.1 Let H be a Hilbert space with an orthogonal decomposition into finite dimensional subspaces.H . Let c = ( c1 .

.decomposition H of H into finite dimensional subspaces. we will even be able to assume that c is the sequence c j = 1j . .H . Then 1 k x j k = k Π j x − Π j −1 x k ≤ k x − Π j x k + k x − Π j −1 x k ≤ . In particular. then we write x = ∑ x j . j so that K ⊆ A{1/j}. then k( I − Π j ) x k < 1/2( j + 1) for all x ∈ K. . there is some i with 1 k x − yi k < . we know that ∞ M H= Hi . K is compact. where x j ∈ Ej . suppose that we can find Nj such that if we let N j M Ej = Hi . Note that we immediately have an orthogonal decomposition ∞ M H= Ej . and a sequence c ∈ `2 such that K ⊂ Ac. We must thus only show that we may find such a sequence of Nj . we have that N j M xi = Hk k =1 5 . with each subspace being a finite sum of the original summands. Fix a j. . 8( j + 1) Moreover.{Ej } . In fact. . . Indeed. i = Nj−1 +1 L Nj and we let Π j be the orthogonal projection onto i =i Hi . . i =1 so by taking Nj large. x m ∈ Hi i =1 1 such that k xi − yi k < 8( j +1) for all i ≤ m. ym ∈ K such that for all x ∈ K. j =1 If x ∈ K. we may find N j M x1 . . so it is totally bounded and we may find y1 .

for i = 1. . there is some i with 1 k xi − x k ≤ . . . m and for each x ∈ K. 2( j + 1)  6 . we need only observe that 1 k x − Π j x k ≤ k x − xi k + kΠ j ( x − xi )k ≤ . 4( j + 1) To finish the proof of the claim. .