Banach algebras and spectrum
Study invertibility in complex normed algebras, derive the Neumann resolvent expansion, and distinguish spectral radius from norm.
On this page
The question for today
What you’ll understand
- Specify the field, completeness, multiplication bound, and unit in a Banach algebra.
- Calculate a resolvent using a convergent Neumann series.
- Distinguish spectrum, spectral radius, and operator norm in a nonnormal example.
An algebra remembers how its elements multiply; a Banach space remembers how limits behave. A Banach algebra combines the two so that convergent series can solve algebraic problems. Our default in this lesson is a nonzero complex unital Banach algebra with unit norm one and a submultiplicative norm. Explicitly naming the complex field matters: the theorem that every element has a nonempty spectrum uses complex analysis and can fail if one searches only for real spectral values.
Multiplication and limits must cooperate
A complex Banach algebra A is a complete complex normed vector space with associative bilinear multiplication satisfying the product norm bound. Multiplication need not commute. If a unit is present, it acts as an identity on both sides. Completeness turns an absolutely norm-convergent formal series into an actual algebra element. Submultiplicativity then lets products pass through its partial-sum limit, providing the analytical step behind identities that initially look purely algebraic.
For an element x of norm below one, the geometric series of its nonnegative powers converges. Multiplying a finite partial sum by one minus x leaves the identity minus the next power, whose norm tends to zero. Thus the limit is a two-sided inverse. This argument works in a noncommutative algebra because all powers involved come from the same element. It also proves invertibility is stable under sufficiently small perturbations of an invertible element.
The spectrum records failures of invertibility
The spectrum of a consists of complex numbers lambda for which lambda times the identity minus a is not invertible inside the declared algebra. Its complement is the resolvent set. If the modulus of lambda exceeds the norm of a, factor out lambda and apply the geometric-series inverse. This places the spectrum in a closed bounded disk. A deeper complex-analytic argument shows that in a nonzero complex unital Banach algebra the spectrum is nonempty and compact; this theorem is stated rather than reproved here.
In finite-dimensional matrix algebras this recovers the eigenvalues, because noninvertibility is equivalent to zero determinant. In infinite dimensions there may be spectral values that are not eigenvalues, so the invertibility definition is more robust. The ambient algebra is also part of the definition: an inverse must belong to it. A nonunital algebra is handled by adjoining an identity and taking the spectrum in its unitization; silently using a nonexistent identity would make the formula meaningless.
Growth of powers can be smaller than a one-step norm
The spectral-radius formula identifies the largest spectral modulus with the limit of the nth root of the norm of the nth power. In particular, spectral radius is at most norm, but equality need not hold. A nonzero nilpotent matrix has only zero in its spectrum and eventually has zero powers, while its operator norm is positive. This is one reason transient amplification cannot be read from eigenvalues alone in a nonnormal system.
The complex-field hypothesis has a simple diagnostic example. A real ninety-degree rotation matrix has determinant lambda squared plus one after subtracting a real scalar, so it has no real spectral value when considered over the real field. After complexification, its spectrum consists of i and minus i. Later C*-algebra structure restores equality of norm and spectral radius for normal elements, but that conclusion uses the adjoint and is not a generic property of Banach algebras.
Work it through
A nilpotent resolvent without an infinite computation
In the complex matrix algebra M₂(C) with operator norm, let N have top-right entry 2 and all other entries zero. Find its spectrum, norm, and (3I−N)⁻¹.
N²=0, and its characteristic polynomial is λ², so its spectrum is {0}. Acting on the second standard basis vector gives length 2, while every output has length at most twice the input length; hence its norm is 2.
The resolvent series at λ=3 truncates after N because all higher powers vanish. Therefore (3I−N)⁻¹=(1/3)I+(1/9)N.
The inverse matrix has diagonal entries 1/3 and top-right entry 2/9. Multiplying by 3I−N yields I+(1/3)N−(1/3)N−N²/9=I in either order.
The spectral radius is zero although the norm is two. The inverse is [[1/3,2/9],[0,1/3]], with the series condition 3>2 satisfied.
The living notebook
Explore the model
Interactive lab · matrix
A shear separates spectral radius from norm
Pause, reset, then set diagonal scale a=1 and shear b=2. The selected matrix is I+N from the example. Predict determinant one, both eigenvalues one, and largest singular value 1+√2, approximately 2.414. Compare the determinant and singular-value readouts. The lab only handles its triangular two-by-two family; it does not compute a Banach-algebra spectrum or a resolvent.
- Area scale · det A
- 1.4
- Singular values · σmax, σmin
- 1.597, 0.876
- Condition number · κ₂(A)
- 1.823
The selected matrix is A = [[a,b],[0,1]]. Playback interpolates geometry between the identity and A; readouts always describe A. Its determinant is a and its singular values come from AᵀA. Positive a makes every supported matrix invertible. This two-dimensional family is not a general eigensolver or a learned network. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.
Look forFor a nonnormal matrix, the largest spectral modulus and the operator norm can differ.
Pause & reason
Check your understanding
Your turn
Try it yourself
Let P be a nonzero projection in a complex unital C*-algebra, viewed as a Banach algebra. Find its possible spectral values and compute (2I−P)⁻¹.
Take a moment to make a prediction or work it out on paper. Open a hint when you need one.
Give me a hint
Use P²=P to test a linear expression in I and P. A projection need not equal the identity.
Compare with the solution
The polynomial identity P(P−I)=0 implies its spectrum is contained in {0,1}; 1 occurs because P is nonzero. Zero also occurs unless P=I. Direct multiplication gives (2I−P)((1/2)I+(1/2)P)=I, using P²=P. The two expressions commute because both are polynomials in P, so this is a two-sided inverse. If P=I, the same formula reduces to I.
Keep the idea
Take it with you
- A Banach-algebra inverse can be constructed by norm-convergent power series.
- Spectrum is defined through invertibility in a specified complex unital algebra or unitization.
- Spectral radius measures asymptotic power growth and can be strictly smaller than norm.
Read further
Original sources and deeper explanations to follow at your own pace.
Optional progress, saved in this browser when available. Revisit any lesson at any time.
Lesson not marked complete.