Canonical: https://knowlpedia.com/functional-analysis/schatten-class-operator/

# Schatten-class operator

A compact Hilbert-space operator whose singular-value sequence is summable to a specified power.

Let \(H\) and \(K\) be [Hilbert spaces](https://knowlpedia.com/linear-algebra/hilbert-space/) and let
\(0<p<\infty\). A **Schatten-class operator of order \(p\)** is a
[compact operator](https://knowlpedia.com/linear-algebra/compact-operator/) \(T:H\to K\) whose
[singular values](https://knowlpedia.com/functional-analysis/singular-values/) satisfy
\[
\sum_{n\geq 1}s_n(T)^p<\infty.
\]
The space of all such operators is denoted \(\mathcal S^p(H,K)\), or
\(\mathcal L^p(H,K)\), and carries the norm or quasinorm
\[
\lVert T\rVert_p=\left(\sum_{n\geq1}s_n(T)^p\right)^{1/p}.
\]
It is a norm when \(p\geq1\) and a quasinorm when \(0<p<1\). When \(H=K\),
\(\mathcal S^p(H)\) is a [two-sided ideal](https://knowlpedia.com/algebra-rings/two-sided-ideal/) in the bounded operators on \(H\).
Thus Schatten membership is a quantitative compactness condition.

## Ideal and inclusion properties

If \(A\) and \(B\) are bounded operators of compatible sizes and
\(T\in\mathcal S^p\), then
\[
\lVert ATB\rVert_p\leq\lVert A\rVert\,\lVert T\rVert_p\,\lVert B\rVert.
\]
For \(0<p<q<\infty\), one has
\(\mathcal S^p\subseteq\mathcal S^q\subseteq\mathcal K\). Hölder's inequality
also has an operator-ideal form: if \(1/r=1/p+1/q\), then products of
\(\mathcal S^p\)- and \(\mathcal S^q\)-operators lie in
\(\mathcal S^r\), with the usual qualification \(r\geq1\) when a Banach norm
is desired.

## Distinguished cases

The class \(\mathcal S^1\) is the trace class, and \(\mathcal S^2\) is the
Hilbert--Schmidt class. For a diagonal compact operator on \(\ell^2\) with
diagonal \((\lambda_n)\), membership in \(\mathcal S^p\) is exactly the
condition \((\lambda_n)\in\ell^p\). [Finite-rank operators](https://knowlpedia.com/functional-analysis/finite-rank-operator/) belong to every
\(\mathcal S^p\) and are dense there in the \(p\)-norm.

## Conventions and scope

For \(0<p<1\), “Schatten class” is sometimes replaced by “quasi-Schatten
ideal” to emphasize the lack of a norm. Some authors extend the notation by setting
\(\mathcal S^\infty(H)=\mathcal K(H)\) with the [operator norm](https://knowlpedia.com/linear-algebra/operator-norm/). Others reserve
“Schatten class” for finite \(p\), as in the core definition. These
conventions should be stated when endpoints or subunit exponents matter.

## References

1. Barry Simon, *Trace Ideals and Their Applications*, 2nd ed., American Mathematical Society, 2005. [DOI record](https://doi.org/10.1090/surv/120). Relevant: Chapters 1--2 on singular values, Schatten ideals, and ideal inequalities.
