logo

Compact Integral Operators 📂Banach Space

Compact Integral Operators

Theorem1

For a given compact interval J=[a,b]J = [a, b], let KK be a function continuous on J×JJ \times J. Let X=C[a,b]X = C[a, b] be the space of continuous functions. Then, the integral operator T:XXT : X \to X with kernel KK is a compact linear operator.

(Tx)(s)=abK(s,t)x(t)dt,xX (Tx)(s) = \int\limits_{a}^{b} K(s, t) x(t) dt,\qquad \forall x \in X

Proof

Since the integral operator is linear and bounded, we only need to show it is compact.

Compact Operator

A linear operator T:XYT : X \to Y is called a compact operator if for every bounded subset MXM \subset X, T(M)\overline{T(M)} is compact implies TT.

Norm of Continuous Function Space

The norm of the continuous function space C[a,b]C[a, b] is defined as follows:

x:=maxt[a,b]x(t),xC[a,b] \left\| x \right\| := \max\limits_{t \in [a, b]} \left| x(t) \right|,\qquad x \in C[a, b]

Assume that the sequence XX in {xn}\left\{ x_{n} \right\} is bounded. In other words, there exists cc such that for all nNn \in \mathbb{N}, xnc\left\| x_{n} \right\| \le c holds. Let yn=Txny_{n} = Tx_{n}. Since TT is bounded, then

yn=TxnTxn \left\| y_{n} \right\| = \left\| Tx_{n} \right\| \le \left\| T \right\| \left\| x_{n} \right\|

Therefore, {yn}\left\{ y_{n} \right\} is also bounded. Now, we will show that {yn}\left\{ y_{n} \right\} is equicontinuous. Since the kernel KK is continuous and J×JJ \times J is compact, KK is uniformly continuous. Hence, for any given ε>0\varepsilon \gt 0, there exists δ>0\delta \gt 0 that satisfies the following:

s1,s2J and tJs1s2<δ    K(s1,t)K(s2,t)<ε(ba)c \forall s_{1}, s_{2} \in J \text{ and } t \in J \qquad \left| s_{1} - s_{2} \right| \lt \delta \implies \left| K(s_{1}, t) - K(s_{2}, t) \right| \lt \frac{\varepsilon}{(b-a)c}

Therefore, for every s1,s2s_{1}, s_{2} and all nNn \in \mathbb{N}, the following holds:

yn(s1)yn(s2)=abK(s1,t)xn(t)dtabK(s2,t)xn(t)dt=ab[K(s1,t)K(s2,t)]xn(t)dtabK(s1,t)K(s2,t)xn(t)dtabε(ba)ccdt=ε \begin{align*} \left| y_{n}(s_{1}) - y_{n}(s_{2}) \right| &= \left| \int\limits_{a}^{b} K(s_{1}, t)x_{n}(t) dt - \int\limits_{a}^{b} K(s_{2}, t)x_{n}(t) dt \right| \\ &= \left| \int\limits_{a}^{b} \left[ K(s_{1}, t) - K(s_{2}, t) \right] x_{n}(t) dt \right| \\ &\le \int\limits_{a}^{b} \left| K(s_{1}, t) - K(s_{2}, t) \right| \left| x_{n}(t) \right| dt \\ &\le \int\limits_{a}^{b} \frac{\varepsilon}{(b-a)c} \cdot c dt \\ &= \varepsilon \end{align*}

As a result, {yn}\left\{ y_{n} \right\} is equicontinuous. Since a bounded equicontinuous sequence of functions has a convergent subsequence, {yn}\left\{ y_{n} \right\} has a convergent subsequence.

Auxiliary Theorem

Let XX and YY be normed spaces. Let T:XYT : X \to Y be a linear operator. Then, the following two propositions are equivalent:

  1. TT is a compact operator.
  2. TT maps “every bounded sequence in XX” to “a sequence in YY that has a convergent subsequence”.

Therefore, since an arbitrary bounded sequence {xn}\left\{ x_{n} \right\} is mapped by TT to a sequence {yn}\left\{ y_{n} \right\} that has a convergent subsequence, TT is a compact operator.


  1. Erwin Kreyszig, Introductory Functional Analysis with Applications (1978), p454-455 ↩︎