1 Definitions and basic properties
1.1 Sectoriality in the complex plane
Let \(X\) be a Banach space and \(A\) a (typically unbounded) linear operator with domain \(D(A)\subset X\). The operator \(A\) is called sectorial if its spectrum lies in a closed sector and the resolvent behaves in a controlled manner away from that sector.
A sector in the complex plane is typically written as \[
| \Sigma_\theta := \{ \lambda\in\mathbb{C}\setminus\{0\} : | \arg \lambda | \le \theta \}, |
|---|
\] for an angle \(\theta\in(0,\pi)\), with a corresponding open sector \(\Sigma_\theta^\circ\) using strict inequality. For sectoriality, one also introduces a shift: there exists a real number \(\omega\) (or, more generally, an element that translates the spectrum) such that \(A-\omega I\) has sectorial spectral inclusion.
Informally, sectoriality means that \(A\) does not have “wild” spectral directions; instead, it behaves as though it were the generator of diffusion-like dynamics or a positive operator in a suitable functional sense.
1.2 Resolvent estimates and angle of sectoriality
The defining property of sectorial operators is expressed through resolvent bounds. One version states that \(A\) is sectorial of angle \(\theta\) if for some \(\omega\in\mathbb{R}\) and some constant \(M\ge 1\), \[
| \|( \lambda I - A)^{-1}\| \le \frac{M}{ | \lambda-\omega | },\qquad \lambda\notin \omega+\Sigma_\theta^\circ. |
|---|
\] Equivalently, using a shifted operator \(B:=A-\omega I\), \[
| \|(\lambda I - B)^{-1}\|\le \frac{M}{ | \lambda | }\quad \text{for } \lambda\notin \Sigma_\theta^\circ. |
|---|
\] The angle of sectoriality is the infimum over all such angles \(\theta\) for which the estimate holds. Smaller angles correspond to stronger control and, in applications, faster decay and better regularity.
1.3 Compatibility with the operator domain
Sectoriality is a statement about the operator as an abstract object, but it must interact consistently with its domain. Since \(A\) is usually unbounded, the domain \(D(A)\) is the natural space for applying \(A\) directly, while the resolvent condition supplies additional information about how \(A\) acts indirectly through \((\lambda I-A)^{-1}\).
A key consequence is that \(D(A)\) is dense in \(X\) in standard settings and that the resolvent maps the whole space into \(D(A)\): \[ (\lambda I-A)^{-1}X \subset D(A),\qquad \lambda\in\rho(A), \] where \(\rho(A)\) is the resolvent set. This compatibility underlies functional calculus constructions and the definition of fractional powers.
1.4 Equivalent characterizations
Sectoriality admits multiple equivalent formulations, often phrased in terms of:
| - the existence of bounds for \(\| \lambda(\lambda I-A)^{-1}\|\) outside a sector; |
|---|
- holomorphic extension properties of the resolvent;
- estimates for powers \(A^k\) relative to resolvent norms (under additional assumptions);
- the generation of analytic semigroups in the semigroup approach.
In Hilbert spaces, sectoriality for certain self-adjoint operators can be read directly from positivity or coercivity assumptions. In Banach spaces, equivalent criteria are frequently stated using resolvent growth, bounded \(H^\infty\)-calculus variants, or square-function estimates, with the exact equivalences depending on the underlying operator class and geometry of the space.
2 Examples and model operators
2.1 Positive self-adjoint operators
| A classical source of sectorial operators is the positive self-adjoint case on a Hilbert space. If \(A\) is self-adjoint and satisfies \( \langle Ax, x\rangle \ge c\|x\|^2\) for some \(c\in\mathbb{R}\), then the spectrum of \(A\) is contained in a half-line, and resolvent bounds follow from the spectral theorem. After shifting by \(c\) (or another appropriate value), the resolvent estimate shows that \(A\) is sectorial with angle \(0\) or \(\pi/2\) depending on the chosen sector convention. |
|---|
This example motivates the general definition: the analytic structure of \(t\mapsto e^{-tA}\) and the behavior of \((\lambda I-A)^{-1}\) are governed by the spectral location and the reality of the spectrum.
2.2 Generators of analytic semigroups
A major motivation for the concept is its connection to analytic semigroups. If an operator \(A\) generates a bounded analytic \(C_0\)-semigroup \((T(t))_{t\ge 0}\), then \(A\) is sectorial (with an angle determined by the semigroup analyticity angle). Conversely, under standard hypotheses, sectoriality implies the generation of an analytic semigroup.
This equivalence explains why sectorial operators are central in evolution equations: analytic semigroups provide a robust way to solve linear initial value problems while capturing smoothing effects.
2.3 Laplace-type operators on function spaces
Elliptic differential operators, such as Laplace-type operators, often give rise to sectorial operators when realized on suitable Banach or Hilbert spaces. For instance, the Laplacian with boundary conditions (Dirichlet, Neumann, or more general elliptic operators) typically yields a sectorial operator on \(L^p\) spaces. The resolvent estimate reflects elliptic regularity and the fact that the operator’s spectrum lies in a controlled region.
Although the exact setting depends on the boundary conditions and the choice of function space, the guiding theme is consistent: the differential operator behaves like a diffusion operator, so its resolvent decays at a rate compatible with sectorial bounds.
2.4 Bounded perturbations and stability of sectoriality
Sectoriality is stable under certain perturbations. A standard principle is that if \(A\) is sectorial and \(K\) is a bounded operator that is sufficiently small or relatively bounded in an appropriate sense, then \(A+K\) remains sectorial (possibly with a different angle or constants).
| This stability is essential in applications where operators arise as “principal part + lower-order terms.” Resolvent identities allow one to transfer sectorial estimates from \(A\) to \(A+K\) by controlling \((I - K(\lambda I-A)^{-1})^{-1}\) for large \( | \lambda | \) outside the sector. |
|---|
3 Functional calculus for sectorial operators
3.1 Holomorphic functional calculus
Sectoriality enables a holomorphic functional calculus: for suitable holomorphic functions \(f\) defined on a sector containing the spectrum, one can define an operator \(f(A)\) by contour integrals involving the resolvent: \[ f(A) := \frac{1}{2\pi i}\int_\Gamma f(\lambda)(\lambda I-A)^{-1}\,d\lambda. \] Here \(\Gamma\) is a contour that winds around the spectrum and stays inside the region of holomorphy. Sectorial resolvent bounds ensure the integral converges in operator norm (or in weaker senses depending on assumptions).
This framework generalizes the spectral theorem’s prescription of \(f\) on eigenvalues, but it works for non-normal operators as well, provided sectoriality supplies enough resolvent control.
3.2 Cauchy integral representations
The functional calculus is built from Cauchy-type representations. The contour can be chosen in several equivalent ways: paths consisting of two rays inside the complement of the sector and an arc at infinity, arranged so that the resolvent is defined and estimates yield convergence.
A practical point is that the calculus is compatible with algebraic operations: for functions where products make sense, one often has \((fg)(A)=f(A)g(A)\) under appropriate conditions, reflecting the integral representation and Cauchy’s theorem.
3.3 Spectral mapping for sectorial operators
A key property is a form of spectral mapping. Under the functional calculus, one expects \[ \sigma(f(A)) = \overline{f(\sigma(A))}, \] for functions \(f\) holomorphic on a neighborhood of \(\sigma(A)\) with additional requirements depending on growth and the type of calculus used. Spectral mapping provides a bridge between analytic properties of \(f\) and the spectral behavior of derived operators.
In semigroup applications, choosing \(f(z)=e^{-tz}\) yields \(T(t)=e^{-tA}\) and clarifies how spectrum affects long-time behavior.
3.4 Boundedness of calculus under growth conditions
Beyond defining \(f(A)\), one investigates when the map \(f\mapsto f(A)\) is bounded with respect to function norms. Growth conditions on \(f\) near \(0\) and at infinity typically control whether the contour integral yields bounded operators and whether estimates depend continuously on \(f\).
In more advanced frameworks, bounded \(H^\infty\)-calculus criteria describe when \[
| \|f(A)\|\le C\|f\|_\infty |
|---|
\] for bounded holomorphic \(f\) on sectors. Such results are closely tied to whether the operator admits a robust functional calculus suited for maximal regularity and stability of fractional domain estimates.
4 Fractional powers and interpolation
4.1 Defining fractional powers
Fractional powers \(A^\alpha\) for \(\alpha\in(0,1)\) are defined using the sectorial calculus. One constructs \(A^\alpha\) through holomorphic functions \(f(z)=z^\alpha\) (with a branch cut aligned with the sector) and sets \(A^\alpha := f(A)\) in the sense of the functional calculus.
This produces an operator whose domain \(D(A^\alpha)\) is typically smaller than \(D(A)\) for \(\alpha<1\) and larger for \(\alpha>1\). The resulting scale of spaces is fundamental in regularity theory, where solutions are measured by how many “powers” of the operator they possess.
4.2 Balakrishnan and related formulas
A common explicit representation is the Balakrishnan formula, which for appropriate sectorial operators expresses fractional powers via an integral over the resolvent: \[ A^\alpha x = \frac{\sin(\pi \alpha)}{\pi}\int_0^\infty \lambda^{\alpha-1}\, A(\lambda I + A)^{-1}x\, d\lambda, \] in settings where \(A\) is sectorial and the integral converges on a suitable domain. Related formulas express \(A^{-\alpha}\) similarly, often involving \((\lambda I+A)^{-1}\) without the factor of \(A\).
Such formulas are valuable both conceptually and computationally: they reveal how the resolvent behavior translates into fractional smoothing.
4.3 Fractional domains and graph norms
Once \(A^\alpha\) is defined, one typically equips fractional domains \(D(A^\alpha)\) with graph norms such as \[
| \|x\|_{D(A^\alpha)} \sim \|x\| + \|A^\alpha x\|. |
|---|
\] In many applications, these norms are equivalent to interpolation norms and can be identified with classical smoothness spaces. For differential operators, fractional domains often match Sobolev or Besov-type regularity scales.
4.4 Interpolation spaces generated by sectorial operators
Sectorial operators naturally generate interpolation spaces. Under suitable assumptions, the spaces \(D(A^\alpha)\) coincide (up to norm equivalence) with real or complex interpolation spaces between \(X\) and \(D(A)\). These identifications are central in treating nonlinear problems via linear estimates: interpolation allows one to estimate intermediate regularity without solving directly at every order.
5 Analytic semigroups and evolution equations
5.1 Semigroup representation via functional calculus
For sectorial operators, the analytic semigroup is represented through the functional calculus. Specifically, one defines \[ T(t) = e^{-tA} \] using the holomorphic functional calculus with \(f(z)=e^{-tz}\). The semigroup property \(T(t+s)=T(t)T(s)\) follows from the multiplicative properties of the calculus for exponentials and the resolvent’s analytic structure.
This representation turns spectral information and resolvent bounds into concrete time-dependent estimates.
5.2 Smoothing estimates and decay rates
Analytic semigroups exhibit smoothing: for \(t>0\), the operator \(T(t)\) maps \(X\) into \(D(A^k)\) for appropriate \(k\), even when the initial data is only in \(X\). The degree of smoothing is tied to sectorial resolvent bounds, leading to estimates of the form \[
| \|A^\alpha T(t)\|\lesssim t^{-\alpha}, |
|---|
\] for \(\alpha\ge 0\), under standard normalization and sectorial angle assumptions.
Similarly, decay rates follow from the spectral location and the semigroup’s boundedness properties. These estimates are a primary tool in proving regularity and convergence of numerical approximations.
5.3 Maximal regularity (overview level)
Maximal regularity is a refined property describing how solutions of the inhomogeneous evolution equation inherit regularity from the forcing term. In the sectorial setting, maximal regularity is typically connected to boundedness properties of the associated operator families (often involving resolvent estimates integrated over sectors).
At an overview level, the takeaway is: sectoriality supplies the analytic structure needed for maximal regularity results, which in turn yield strong control over time derivatives and spatial operators in parabolic-type problems.
5.4 Applications to parabolic-type problems
Consider linear parabolic equations of the schematic form \[ u'(t)+Au(t)=f(t),\qquad u(0)=u_0, \] where \(A\) is sectorial. The mild solution is expressed via \[ u(t)=T(t)u_0+\int_0^t T(t-s)f(s)\,ds. \] Sectoriality ensures that \(T(t)\) is analytic and provides a mechanism to derive estimates in time and space. For diffusion-reaction systems, one often combines sectoriality of the principal diffusion operator with perturbation arguments for lower-order terms.
6 Operator perturbation theory
6.1 Additive perturbations
Perturbation theory studies how the sectorial properties of \(A\) change under replacing \(A\) by \(A+K\). With \(K\) bounded, the resolvent identity \[ (\lambda I-(A+K))^{-1}=(\lambda I-A)^{-1}\bigl(I-K(\lambda I-A)^{-1}\bigr)^{-1} \]
| allows one to bound the new resolvent for \(\lambda\) outside a sufficiently large region. If \(\|K(\lambda I-A)^{-1}\|\) is small there, sectoriality persists, possibly with altered constants. |
|---|
6.2 Multiplicative or similarity transforms
More general perturbations consider similarity transforms or multiplicative factors, such as \(B=S^{-1}AS\) for invertible \(S\). In that case, sectoriality can be transferred, but the resolvent norms and angles may change due to non-isometric behavior of \(S\). Multiplicative perturbations may also be treated via operator-valued function methods, again relying on resolvent control.
These techniques are important when passing between equivalent formulations of a model problem, for example after rescaling variables or applying a change of unknowns.
6.3 Perturbation of resolvent bounds
Even when the operator angle or constants shift, one tracks the key quantitative ingredient: the resolvent estimate in the complement of a sector. Perturbation results often yield inequalities of the form \[
| \|(\lambda I-(A+K))^{-1}\|\le \frac{M'}{ | \lambda | }, |
|---|
\] for \(\lambda\) outside a possibly wider sector, with \(M'\) depending on bounds for \(K\). The stability of the resolvent growth rate is what makes fractional power estimates and semigroup smoothing estimates robust.
6.4 Consequences for fractional powers
Once sectoriality is stable, the associated fractional powers behave predictably. In many settings, if \(A\) and \(A+K\) are both sectorial with compatible functional calculi, then their fractional domains align in the sense of norm equivalence on suitable interpolation scales.
These consequences allow one to incorporate perturbations into regularity results: for instance, a lower-order term may not change the nature of the regularity scale, even though it affects constants in estimates.
7 Discretization and numerical perspectives
7.1 Rational and polynomial approximations
Numerical methods often approximate the semigroup or fractional powers using rational or polynomial functions of \(A\). Because sectorial operators admit good holomorphic calculus, one can design approximants \(r_n(A)\) to \(f(A)\) for chosen analytic \(f\). Rational approximations are particularly effective for functions with branch points or slow decay, such as those related to \(A^{-\alpha}\).
7.2 Approximation of semigroups
To approximate \(T(t)=e^{-tA}\), methods may employ time stepping schemes (like implicit or exponential integrators) that can be expressed as \(r(A)\) acting on \(u_0\) and history terms. Sectorial resolvent bounds translate into convergence rates through estimates on the approximation error functions \(e^{-z}-r_n(z)\) over relevant regions.
A typical outcome is that the quality of approximation depends on how accurately the rational function matches the exponential on sectors where the resolvent is controlled.
7.3 Error bounds driven by sectorial estimates
Error analysis frequently reduces to bounding operators like \[
| \| (f(A)-r_n(A))x\| |
|---|
\] by controlling the scalar approximation error on the complex plane and using the functional calculus. Sectoriality provides the needed growth estimates to ensure that these bounds hold uniformly over discretization parameters.
Furthermore, smoothing estimates often improve error rates for sufficiently regular data: the analytic semigroup regularizes the solution, making high-frequency components less influential after positive time.
7.4 Practical considerations for computing fractional powers
Computing \(A^\alpha\) or \(A^{-\alpha}\) numerically is more delicate than applying integer powers. Practical approaches include:
- contour-integral quadrature based on resolvent evaluations;
- rational approximations to \(z^\alpha\) or \(z^{-\alpha}\);
- integral representations using semigroup or resolvent forms.
Sectorial estimates are crucial because they bound resolvent norms along contours or in quadrature regimes, which in turn controls stability and error propagation.
8 Further topics and references
8.1 Links to m-accretive and dissipative operators
There are close relationships between sectorial operators and classes defined by accretivity or dissipativity. In many Banach-space frameworks, \(m\)-accretive operators generate contractive semigroups, while dissipative operators relate to stability and resolvent bounds in half-planes. Under appropriate hypotheses, these properties imply sectoriality (possibly after shifts), connecting abstract functional analysis to semigroup theory and evolution equations.
8.2 Connections to PDE theory
In partial differential equations, sectorial operators appear as analytic realizations of elliptic operators. Fractional powers correspond to fractional smoothness scales, and sectorial semigroups govern parabolic evolution. This connection is used to justify regularity results, understand boundary effects, and treat coupled systems by operator splitting and perturbation methods.
8.3 Common research directions and generalizations
Research directions include:
- refined criteria for bounded \(H^\infty\)-functional calculus;
- maximal regularity and its consequences for nonlinear problems;
- extension to non-autonomous evolution equations where sectoriality varies with time;
- sectorial operators on more general spaces (e.g., UMD Banach spaces) where harmonic-analytic tools apply;
- computational methods that efficiently evaluate resolvent-based representations of fractional powers.