Weiss Calculus: Manifold & Orthogonal Calculi

• Weiss Calculus is a framework that approximates unstable geometric data via Taylor towers and polynomial functor techniques.
• It encompasses manifold calculus for contravariant functors on open subsets and orthogonal calculus for functors on finite-dimensional inner product spaces.
• Key applications include stable splittings and derivative computations in topology, enabling precise classification using configuration spaces and spectra.

Weiss calculus denotes a family of functor calculi associated with Michael Weiss and, in the manifold case, with Goodwillie and Weiss. In one standard usage it is the manifold calculus of contravariant functors on the poset of open subsets of a manifold; in another it is orthogonal calculus for functors out of finite-dimensional inner product spaces. In both usages the central constructions are polynomial approximations, Taylor towers, homogeneous layers, and derivatives, with classification theorems that replace an unstable functor by configuration-space data or spectra equipped with group actions (Munson, 2010, Tsopmene et al., 2017, Barnes et al., 5 Aug 2025).

1. Terminology and scope

The expression “Weiss calculus” is not completely uniform across the literature. “Manifold calculus,” “embedding calculus,” and “Weiss calculus” are used for the theory of contravariant functors

$F:\mathcal O(M)^{op}\to \mathsf{Top}$

on open subsets of a smooth manifold $M$ (Pryor, 2013, Munson, 2010). “Orthogonal calculus” is also described as Weiss calculus, especially for functors on finite-dimensional inner product spaces, and later work treats unitary calculus, calculus with Reality, and equivariant Weiss calculus as extensions of that framework (Barnes et al., 5 Aug 2025, Taggart, 2020, Bhattacharya et al., 2024).

Variant	Source category	Characteristic layer/classification
Manifold calculus	$\mathcal O(M)$, open subsets of a smooth manifold	homogeneous degree $k$ functors via configuration spaces $F_k(M)$ or $\binom{U}{k}$ (Tsopmene et al., 2017, Munson, 2010)
Orthogonal calculus	$\mathcal J^O$ or $\mathrm{Vect}_\Bbbk$	$n$-homogeneous functors via spectra with $O(n)$- or $M$0-action (Krannich et al., 2021, Barnes et al., 5 Aug 2025)
Unitary and Reality variants	$M$1, $M$2	homogeneous functors via $M$3- and $M$4-spectra (Taggart, 2020)
Equivariant Weiss calculus	representations over the orbit category of a finite group $M$5	homogeneous layers via orthogonal $M$6-spectra (Bhattacharya et al., 2024)

A distinct operator-theoretic usage also appears in the literature on generalized Weiss operators, where

$M$7

is studied as a factorized differential operator on $M$8 (Borisenok et al., 2012). This suggests a separate usage of the name from the functor-calculus tradition.

2. Manifold calculus of functors

In manifold calculus one fixes a smooth manifold $M$9 and considers the poset-category $\mathcal O(M)$0 of open subsets with morphisms given by inclusions. A cofunctor is a contravariant functor

$\mathcal O(M)$1

with $\mathcal O(M)$2 typically a simplicial model category (Tsopmene et al., 2017). The basic regularity hypothesis is that $\mathcal O(M)$3 be good: it is isotopy invariant, and for any increasing sequence $\mathcal O(M)$4, the canonical map

$\mathcal O(M)$5

is a weak equivalence (Tsopmene et al., 2017, Munson, 2010).

Polynomiality is a higher excision condition. A cofunctor $\mathcal O(M)$6 is polynomial of degree $\mathcal O(M)$7 if for every $\mathcal O(M)$8 and every collection of pairwise disjoint closed subsets $\mathcal O(M)$9, the canonical map

$k$0

is a weak equivalence (Tsopmene et al., 2017). The $k$1-th Taylor approximation is defined by restriction to small open sets and homotopy right Kan extension: $k$2 where $k$3 is the full subposet of open subsets diffeomorphic to a disjoint union of at most $k$4 balls (Tsopmene et al., 2017, Munson, 2010). The resulting tower

$k$5

is the manifold-calculus Taylor tower.

A central structural theorem is that polynomial functors are determined by their values on small opens. Weiss proved this for $k$6, and Pryor showed that one can replace $k$7 by more general subposets $k$8 built from any good basis $k$9 of ball-like opens without changing the notion of polynomial cofunctor (Pryor, 2013). Songhafouo Tsopméné and Stanley generalized this from spaces to arbitrary simplicial model categories, proving that a cofunctor is good and polynomial of degree $F_k(M)$0 exactly when it is objectwise weakly equivalent to the homotopy right Kan extension of an isotopy cofunctor on $F_k(M)$1 (Tsopmene et al., 2017).

Homogeneous functors isolate a single layer of the tower. If $F_k(M)$2 has a zero object, a cofunctor is homogeneous of degree $F_k(M)$3 if it is good, polynomial of degree $F_k(M)$4, and satisfies

$F_k(M)$5

The classification theorem identifies homogeneous degree $F_k(M)$6 cofunctors on $F_k(M)$7 with linear cofunctors on the unordered configuration space $F_k(M)$8: $F_k(M)$9 (Tsopmene et al., 2017). In the exposition centered on compactly supported sections, homogeneous degree $\binom{U}{k}$0 functors are described as

$\binom{U}{k}$1

where $\binom{U}{k}$2 is a fibration and the fiber over a configuration is the $\binom{U}{k}$3-th derivative of the functor (Munson, 2010).

3. Orthogonal calculus

In orthogonal calculus the source is geometric but linear rather than manifold-theoretic. One common formulation takes continuous functors

$\binom{U}{k}$4

where $\binom{U}{k}$5 is the category of finite-dimensional real inner product spaces and linear isometries (Krannich et al., 2021). Another formulation studies

$\binom{U}{k}$6

where $\binom{U}{k}$7 is $\binom{U}{k}$8, or $\binom{U}{k}$9, and $\mathcal J^O$0 is the category of finite-dimensional inner product spaces over $\mathcal J^O$1 and linear isometric embeddings (Barnes et al., 5 Aug 2025).

The orthogonal Taylor tower is built from the endofunctors

$\mathcal J^O$2

in the real formulation, or by the equivalent limit condition

$\mathcal J^O$3

in the spectrum-valued formulation (Taggart, 2020, Barnes et al., 5 Aug 2025). A functor is $\mathcal J^O$4-polynomial if this map is an equivalence for every $\mathcal J^O$5, and the $\mathcal J^O$6-th approximation $\mathcal J^O$7 is the universal $\mathcal J^O$8-polynomial approximation. The corresponding homogeneous layer is

$\mathcal J^O$9

The classification of homogeneous layers is one of the defining results of the subject. In the real case, if $\mathrm{Vect}_\Bbbk$0 is $\mathrm{Vect}_\Bbbk$1-homogeneous, then there is a spectrum $\mathrm{Vect}_\Bbbk$2 with $\mathrm{Vect}_\Bbbk$3-action such that

$\mathrm{Vect}_\Bbbk$4

(Krannich et al., 2021). In the $\mathrm{Vect}_\Bbbk$5-linear formulation, $\mathrm{Vect}_\Bbbk$6-homogeneous functors are classified by spectra with

$\mathrm{Vect}_\Bbbk$7

action, depending on the field (Barnes et al., 5 Aug 2025). A basic recognition theorem is that $\mathrm{Vect}_\Bbbk$8 is $\mathrm{Vect}_\Bbbk$9-polynomial if and only if its $n$0-st Weiss cross-effect vanishes (Barnes et al., 5 Aug 2025).

This classification is concrete enough to support explicit derivative calculations. For the functor $n$1, the first derivative is rationally equivalent to $n$2, and the second derivative is rationally equivalent to

$n$3

as an $n$4-spectrum (Krannich et al., 2021).

4. Algebraic classifications and stable Weiss towers

Recent work has made the algebraic content of orthogonal calculus substantially more explicit. A Dwyer–Rezk-style classification identifies the $n$5-category of $n$6-polynomial functors with a small functor category: $n$7 where $n$8 is the category of finite-dimensional inner product spaces of dimension at most $n$9 and orthogonal epimorphisms (Barnes et al., 5 Aug 2025). The same work proves the Morita-type description

$O(n)$0

showing that every $O(n)$1-polynomial functor is recovered by left Kan extension from its values on $O(n)$2 (Barnes et al., 5 Aug 2025).

Homogeneous functors admit an equally compact description. The category of $O(n)$3-homogeneous functors is equivalent to Borel $O(n)$4-spectra: $O(n)$5 Equivalently, if $O(n)$6 is a Borel $O(n)$7-spectrum, the associated homogeneous functor is

$O(n)$8

(Barnes et al., 5 Aug 2025).

A further structural advance comes from Koszul duality and categorical Fourier transforms in stable Weiss calculus. The derivatives of a functor $O(n)$9 are organized not merely as an orthogonal sequence $M$00, but as a right module over the Koszul dual category $M$01 (Malin et al., 2024). In that framework the $M$02-category of $M$03-polynomial functors is equivalent to the category of $M$04-truncated coalgebras over a comonad on derivative modules: $M$05 (Malin et al., 2024). The $M$06-th polynomial approximation is described by a homotopy pullback square involving the linear fat diagonal $M$07: $M$08 The paper interprets the comparison with a “fake” tower in terms of generalized norm maps and $M$09-Tate theory (Malin et al., 2024).

5. Unitary, Reality, equivariant, local, and monoidal variants

Unitary calculus replaces real inner product spaces by complex ones. Its input category is $M$10, the category of finite-dimensional complex inner product spaces and unitary embeddings, and an $M$11-polynomial functor is defined by

$M$12

(Taggart, 2020). Taggart’s model-categorical treatment constructs the $M$13-polynomial model structure, the $M$14-homogeneous model structure, the intermediate category $M$15, and Quillen equivalences

$M$16

(Taggart, 2019). Homogeneous unitary functors are classified by spectra with $M$17-action (Taggart, 2020).

Calculus with Reality is a $M$18-equivariant refinement built from the category $M$19 of complexifications $M$20, with complex conjugation giving the $M$21-action (Taggart, 2020). Its homogeneous functors are classified by spectra with $M$22-action, and the fundamental recovery theorem states that for every functor with reality $M$23,

$M$24

(Taggart, 2020). A later comparison shows that orthogonal calculus is recovered from calculus with Reality “up to a shift” by a suitable $M$25-fixed points functor, with

$M$26

(Taggart, 2022).

Equivariant Weiss calculus for a finite group $M$27 replaces dimensions by finite-dimensional $M$28-representations. In this theory Taylor approximations and derivatives are indexed by representations, homogeneous layers are classified by orthogonal $M$29-spectra, and the framework includes both restriction and fixed-point functors (Bhattacharya et al., 2024). For a representation $M$30, the $M$31-th derivative at infinity is an $M$32-system of orthogonal $M$33-spectra with naive $M$34-action, and $M$35-homogeneous functors are classified by a formula of the form

$M$36

(Bhattacharya et al., 2024).

Two further refinements alter the target homotopy theory rather than the source. First, local orthogonal calculus constructs $M$37-local polynomial and homogeneous model structures for a set $M$38 of maps, and proves that $M$39-local $M$40-homogeneous functors are equivalent to $M$41-local spectra with $M$42-action (Taggart, 2021). Second, monoidal orthogonal calculus studies lax symmetric monoidal functors $M$43 under Day convolution and proves that the Taylor approximations $M$44 are again lax symmetric monoidal (Hendrian, 2023). In that setting the derivative spectra carry $M$45-equivariant maps

$M$46

where

$M$47

is the Klein–Spivak dualising spectrum of $M$48 (Hendrian, 2023).

6. Applications, convergence, and significance

The paradigmatic manifold-calculus application is the embedding functor

$M$49

Its linear approximation satisfies

$M$50

and the Taylor tower converges to $M$51 in codimension $M$52, with

$M$53

$M$54-connected (Munson, 2010). The basis-independence theorem for special open sets means that these approximations can be computed using balls, cubes, simplices, or other convenient basis elements without changing their homotopy type (Pryor, 2013).

In orthogonal calculus, derivative computations feed directly into geometric topology. The rational first two derivatives of $M$55 determine the rational homotopy type of $M$56 in a range and the rational homotopy groups of $M$57 up to approximately $M$58 (Krannich et al., 2021). The same paper identifies the rational concordance stable range of the disc by

$M$59

(Krannich et al., 2021).

Equivariant Weiss calculus has also been used to produce stable splittings. An equivariant version of Weiss calculus yields a $M$60-equivariant stable splitting of $M$61 for $M$62, and by taking geometric fixed points one obtains a stable splitting of

$M$63

and in particular of $M$64 (Tynan, 2017).

Taken together, these developments show that Weiss calculus is not a single theorem but a family of Taylor theories for functors in geometry and homotopy theory. In manifold calculus the basic local model is a union of balls inside a manifold; in orthogonal and unitary calculus it is stabilization by direct sum with low-dimensional vector spaces; in the equivariant theory it is stabilization by representations. This suggests a common structural role: Weiss calculus organizes how unstable geometric input is approximated by polynomial stages whose layers admit stable, equivariant, and often explicitly classifiable models (Tsopmene et al., 2017, Barnes et al., 5 Aug 2025, Bhattacharya et al., 2024).