Canonical: https://knowlpedia.com/operator-algebras/spectral-permanence/

# Spectral permanence

A C-star-subalgebra computes the same spectrum as its ambient C-star-algebra when units are treated consistently.

Let \(B\subseteq A\) be a
[\(C^*\)-subalgebra](https://knowlpedia.com/operator-algebras/cstar-subalgebra/). If \(A\) and \(B\)
are unital and share the same identity, then every \(b\in B\) has
\[
\sigma_B(b)=\sigma_A(b),
\]
where the spectra are those of
[Banach-algebra elements](https://knowlpedia.com/functional-analysis/banach-algebra-spectrum/).
Equivalently, whenever \(b\) is invertible in \(A\), its inverse already lies
in \(B\). For nonunital inclusions, the same assertion holds after passing to
the compatible unitizations. This theorem is **spectral permanence**. Its
unit convention is essential: viewing a nonunital subalgebra as though it
shared the ambient unit can change the status of zero in the spectrum.

## Proof idea

For \(b\in B\), invertibility of \(b\) in \(A\) implies that
\(b^*b\) is [bounded below](https://knowlpedia.com/real-analysis/bounded-below/) by a positive scalar. Continuous functional
calculus inside the \(C^*\)-algebra generated by \(b^*b\) then produces
\((b^*b)^{-1}\) in \(B\), and
\[
b^{-1}=(b^*b)^{-1}b^*.
\]
Applying the same reasoning to shifts \(b-\lambda1\) gives equality of
spectra.

## Consequences

Positivity, spectral radius, and
[continuous functional
calculus](https://knowlpedia.com/operator-algebras/continuous-functional-calculus/) for an element of \(B\) may be computed intrinsically in \(B\) or
in the ambient algebra \(A\). In particular, if \(b\in B\) is normal and
\(f\) is continuous on its spectrum, then \(f(b)\in B\). This is one reason norm-closed
\(*\)-subalgebras are the natural subobjects in \(C^*\)-algebra theory.

## Conventions and scope

**Warning.** Spectral permanence is not true for arbitrary dense
\(*\)-subalgebras or for arbitrary Banach subalgebras. Such a subalgebra is
called spectrally invariant or inverse-closed only when this conclusion is
separately verified. The theorem also does not say that every inverse in the
ambient algebra of an element outside \(B\) must belong to \(B\).

## References

1. Gerard J. Murphy, *C*-Algebras and Operator Theory*, Academic Press, 1990. [DOI record](https://doi.org/10.1016/C2009-0-22289-6). Relevant: §2.1 on \(C^*\)-subalgebras and spectral permanence.
2. Gert K. Pedersen, *C*-Algebras and Their Automorphism Groups*, 2nd ed., Academic Press, 2018. [DOI record](https://doi.org/10.1016/C2016-0-03431-9). Relevant: §1.2 on spectra and functional calculus.
