Elementary Properties of Cesàro Operator in L^2

Analysis Functional Analysis Lp-space Operator Theory Hilbert Space

An example in elementary calculus

Consider a sequence of real or complex numbers {sn}. If sns, then

πn=s1++snns.

Here, πn is called the Cesàro sum of {sn}. The proof is rather simple. Given ε>0, there exists some N>0 such that |sns|<ε for all n>N. Therefore we can write

|πns|=|s1+s2++sNn+sN+1++snns|=|(s1s)+(s2s)++(sNs)n+(sN+1s)++(sns)n||s1++sNNsn|+Nnε

For fixed N, we can pick n big enough such that N/n<1/2 (i.e. n>2N) and

|s1++sNNsn|<12ε.

Hence πn converges to s. But the converse is not true in general. For example, if we put sn=(1)n, then it diverges but πn0. If πn converges, we say {sn} is Cesàro summable.

If we treat πn as an integration with respect to the counting measure, things become interesting. Why don’t we investigate the operator defined to be

C(f)(x)=1x0xf(t)dt.

In this blog post we investigate this operator in Hilbert space L2(0,).

The Cesàro operator

Put L2=L2(0,) relative to Lebesgue measure, and the Cèsaro operator C is defined as follows:

(Cf)(s)=1s0sf(t)dt.

Compactness and boundedness of this operator

From the example above, we shouldn’t expect C to be too normal or well-behaved, as convergence is not goes as expected. But fortunately it is at the very least continuous: due to Hardy’s inequality, we have C=2. I have organised several proofs of this. But C is not compact.

Here is the proof. Consider a family of functions {φA}A>0 where

φA=Aχ(0,1/A].

(I owe Oliver Diaz for this family of functions.) It’s not hard to show that φA=1. If we apply C on it we see

(CφA)(x)=1x0xAχ(0,1/A]dx=A(χ(0,1/A](x)+1Axχ(1/A,+))

Hence CφA=1+A2A. Meanwhile for B>A, we have

C(φBφA)(x)=(BA)χ(0,1/B](x)+(1BxA)χ(1/B,1/A](x)+(1B1A)1xχ(1/A,+)(x)

It follows that

|C(φBφA)|(x)(1A1B)1xχ(1/A,)(x).

If we compute the norm on the right hand side we get

C(φBφA)|1AB|.

As a result, if we pick fn=φ2n, then for any m>n we get

C(fmfn)|12nm|112.

Therefore, we find a sequence (fn) on the unit ball such that (Cfn) has no convergent subsequence.

Also we can find its adjoint operator:

Cf,g=0(1s0sf(t)dt)g(s)ds=0(t1sf(t)g(s)ds)dt=0f(t)(t1sg(s)ds)dt.

Hence the adjoint is given by

(Cf)(t)=t1sg(s)ds.

C is not compact as well. Further, another application of Fubini’s theorem shows that

CC=C+C=CC(IC)(IC)=I=(IC)(IC)

Hence IC is an isometry, and C is normal.

Bilateral shift, spectrum

In this section we study the spectrum of C and C, which will be derived from properties of bilateral shift, which comes from 2 space. For convenience we write N=Z0. This section can also help you understand the connection between L2(0,1) and L2(0,).

An operator U on a Hilbert space H is called a simple unilateral shift if H has a orthonormal basis {en} such that U(en)=en+1 for all nN. This is nothing but right-shift operator in the sense of basis. Besides, we call U a unilateral shift of multiplicity m if U is a direct sum of m simple unilateral shifts (note: m can be any cardinal number, finite or infinite).

If we consider the difference between N and Z, we have the definition of bilateral shift. An operator W on K is called a simple bilateral shift if K has a orthonormal basis {en} such that Wen=en+1 for all nZ. Besides, if we consider the subspace H which is spanned by {en}, we see W|H is simply a unilateral shift. Before we begin, we investigate some elementary properties of uni/bilateral shifts.

(Proposition 1) A simple unilateral shift U is an isometry.

Proof. Note (Uem,Uen)=(em+1,en+1)=δm+1,n+1=δmn=(em,en).

(Proposition 2) A simple bilateral shift W is unitary, hence is also an isometry.

Proof. Note (Wem,en)=(em+1,en)=δm+1,n=δm,n1=(em,W1en), from which it follows that W=W1.

Now let the Hilbert space K and its subspace H (invariant under W) be given. Consider the ‘orthonormal’ operator given by Ren=e(n+1). It follows that R is a unitary involution and

Re0=W1e0RH=HRW=W1R.

With these tools, we are ready for the most important theorems.

W=IC is a simple bilateral shift on K=L2.

Step 1 - Obtaining missing subspace, operator and basis

Here we put H=L2(0,1), which can be canonically embedded into L2(0,) in the obvious way (consider all L2 functions vanish outside (0,1)). It is natural to put this, as there are many similarities between L2(0,1) and L2(0,).

Explicitly,

(Wf)(x)=f(x)x1tf(t)dt,xL2(0,).

Also we claim the basis to be generated by e0=χ(0,1). First of all we show that (Wen)n0 is orthonormal. Note as we have proved, WW=(IC)(IC)=I. Without loss of generality we assume that mn and therefore

(em,en)=(Wme0,Wne0)=((W)nWme0,e0)=((WW)nWmne0,e0)=(Wmne0,e0).

If m=n, then (em,en)=(e0,e0)=1. Hence it is reduced to prove that (Wke0,e0)=0 for all k>0. First of all we have

(We0,e0)=(e0,e0)(Ce0,e0)=1(Ce0,e0)

meanwhile

(Ce0,e0)=01(x11tdt)dx=01(lnx)dx=(xlnx+x)|01=1

Hence We0e0. Suppose now we have (Wk1e0,e0)=0, then

(Wke0,e0)=(WWk1e0,e0)=((IC)Wk1e0,e0)=(Wk1e0,e0)(CWk1e0,e0)=(Wk1e0,Ce0)=01Wk1e0(x)1x(0xdt)dx=01Wk1e0(x)1xxdx=(Wk1e0,e0)=0.

Note Wke0L2(0,) always vanishes when x1: when we are doing inner product, [1,) is automatically excluded. With these being said, (Wne0)n0 forms a orthonormal set. By The Hausdorff Maximality Theorem, it is contained in a maximal orthonormal set. But since H=L2(0,1) is separable (admitting a countable basis, proof), (Wke0)k forms a basis of H. From now on we write {en}.

To find the involution R, note first W=IC is already unitary (also, if it is not unitary, then it cannot be a bilateral shift, we have nothing to prove), whose inverse or adjoint is W=IC as we have proved earlier. Hence we have

Re0=e1=(IC)e0=χ(0,1)1x0xdt=1xχ[1,)

But we have no idea what R is exactly. We need to find it manually (or we have to guess). First of all it shall be guaranteed that RH=H. Since H contains all L2 functions vanish on [1,), functions in RH should vanish on (0,1). It is natural to put R(f)(x)=g(x)f(1x) for the time being. g should be determined by e1. Note e0(1x)=χ[1,) almost everywhere, we shall put g(x)=1x. It is then clear that Re0=W1e0 and RH=H. For the third condition, we need to show that

WRW=R.

Note

WRW(f)=WR(f(x)x1tf(t)dt)=W(1xf(1x)+1x1/xf(t)dt)=1xf(1x)+1x1/xf(t)dt+x1t2f(1t)dt+x1t21/tf(u)du=0 by Fubini's theorem, similar to proving CC=C+C.=R(f).

Step 2 - With these, W in step 1 has to be a simple bilateral shift

This is independent to the spaces chosen. To finish the proof, we need a lemma:

Suppose K is a Hilbert space, H is a subspace and e0H. W is a unitary operator such that Wne0H for all n0 and {en=Wne0}n0 forms a orthonormal basis of H. R is a unitary involution on K such that

Re0=W1e0RH=HRW=W1R

then W is a simple bilateral shift.

Indeed, objects mentioned in step 1 fit in this lemma. To begin with, we write en=Wne0 for all nZ. Then {en} is an orthonormal set because for arbitrary m,nZ, there is a jZ such that m+j,n+j0. Therefore

(em,en)=(Wjem,Wjen)=(Wm+je0,Wn+je0)=(em+j,en+j)=δm+j,n+j=δm,n.

Since (e0,e1,) spans H, RH=H, we see (Re0,Re1,) spans H. But

Ren=RWne0=WnRe0=Wn1e0=en1,

hence {e1,e2,} spans H. By definition of W, it is indeed a bilateral shift. And our proof is now complete.

References

  • Walter Rudin, Functional Analysis.
  • Arlen Brown, P. R. Halmos, A. L. Shields, Cesàro operators.

Comments