---
title: Homogeneous Functors in Calculus
url: https://www.emergentmind.com/topics/homogeneous-functors
type: topic
---

# Homogeneous Functors in Calculus

Homogeneous functors are the “pure-degree” pieces that occur when a functor is approximated by a tower of polynomial or excisive approximations. Across Goodwillie calculus, Weiss calculus, manifold calculus, and algebraic theories of polynomial functors, the common pattern is that one first constructs successive polynomial stages and then isolates the \(d\)-th or \(n\)-th layer as the fibre between adjacent stages. In that sense, a homogeneous functor is the analogue of a single monomial term in ordinary calculus, but realized as a functor together with its coherence data. The modern literature gives several classification theorems for such layers, ranging from equivariant spectra in Weiss calculus to configuration-space representations in manifold calculus and module categories over Schur superalgebras in the theory of strict polynomial functors [2508.03808] [1703.09632] [1708.02642] [1807.06120] [1302.0042].

## 1. Core definitions and recurring pattern

In Goodwillie calculus, an endofunctor \(F\colon \mathrm{Sp}\to \mathrm{Sp}\) is \(n\)-excisive if it carries strongly cocartesian \((n+1)\)-cubes in spectra to homotopy cartesian cubes. Goodwillie’s tower
\[
\cdots \to P_nF \to P_{n-1}F \to \cdots \to P_0F
\]
provides universal \(n\)-excisive approximations, and the \(n\)-th homogeneous layer is
\[
D_nF := \mathrm{fib}(P_nF \to P_{n-1}F).
\]
A functor is \(n\)-homogeneous when it is \(n\)-excisive and \(P_{n-1}F\simeq *\); equivalently \(F\simeq D_nF\) [2412.12281] [1703.09632].

Weiss calculus uses the same formal pattern for functors
\[
F\colon \mathrm{Vect}_k \to \mathrm{Sp},
\qquad k\in\{\mathbb R,\mathbb C,\mathbb H\}.
\]
A functor is \(d\)-polynomial if for every \(V\in \mathrm{Vect}_k\) the canonical map
\[
F(V)\to \lim_{0\neq U\subset k^{d+1}} F(V\oplus U)
\]
is an equivalence. Equivalently, \(F\) is \(d\)-polynomial exactly when its \((d+1)\)-st cross-effect vanishes. The Weiss–Taylor tower
\[
\cdots \to P_dF \to P_{d-1}F \to \cdots \to P_0F
\]
then defines the \(d\)-th homogeneous layer by
\[
D_dF=\mathrm{fib}(P_dF\to P_{d-1}F),
\]
and a \(d\)-homogeneous functor is characterized by being \(d\)-polynomial with \(P_{d-1}(D_dF)\simeq 0\) [2508.03808].

In manifold calculus, for a contravariant functor \(F\colon \mathcal O(M)^{op}\to \mathcal C\), polynomiality of degree \(\le k\) is expressed by homotopy-cartesianness of the puncturing cube associated to \(k+1\) pairwise disjoint closed subsets of an open set \(U\subset M\). The Taylor approximation is
\[
T_kF(U)=\underset{V\in \mathcal O_k(U)}{\mathrm{holim}}\,F(V),
\]
and \(F\) is homogeneous of degree \(k\) if it is polynomial of degree \(\le k\) and \(T_{k-1}F\) is trivial [1708.02642] [1807.06120].

A useful point of comparison is that these definitions all isolate the top nonvanishing stage of a polynomial approximation, but the ambient categories and geometric meaning differ substantially. This suggests that “homogeneous functor” is best understood as a family of parallel notions rather than a single context-free definition.

## 2. Goodwillie and Weiss homogeneous layers

Goodwillie’s tower may be described as a sequence of left exact localizations, and the homogeneous layer \(D_nF\) is not merely \(n\)-excisive but also \((n-1)\)-reduced. In the higher-topos-theoretic treatment of Anel, Biedermann, Finster, and Joyal, fibrewise orthogonality controls the \(n\)-excisive modality and leads to structural results such as the pushout–product theorem for \(P_m\)- and \(P_n\)-equivalences, a Blakers–Massey theorem for the Goodwillie tower, and a delooping theorem for homogeneous layers. In particular, every \(n\)-homogeneous functor \(D_nF\) admits a canonical \(n\)-fold delooping in the slice \(\infty\)-topos over \(P_0F\) [1703.09632].

For reduced \(d\)-excisive endofunctors of spectra landing in \( \mathbb Q \)-local spectra, rationalization causes the tower to split:
\[
F(X)\simeq_{\mathbb Q}\prod_{n=0}^d L_nF(X),
\]
equivalently
\[
\mathrm{Exc}_d(\mathrm{Sp})_{\mathbb Q}\simeq \prod_{n=1}^d \mathrm{Homog}_n(\mathrm{Sp})_{\mathbb Q}.
\]
The \(n\)-th homogeneous layer is described by the derivative spectrum with symmetric-group action,
\[
D_nF(X)\simeq (\partial_nF\wedge X^{\wedge n})_{h\Sigma_n},
\]
and rationally this yields an algebraic model
\[
\mathrm{Exc}_d(\mathrm{Sp})_{\mathbb Q}\simeq \prod_{n=1}^d \mathrm{Ch}(\mathbb Q[\Sigma_n]).
\]
The homogeneous layers therefore become the basic algebraic constituents of rational excisive functors [2412.12281].

In Weiss calculus, the homogeneous layer has a more geometric input: finite-dimensional inner-product spaces and orthogonal splittings. The details emphasize that the tower is built from orthogonal embeddings, while the classifying side is governed by orthogonal epimorphisms. The \(d\)-th layer is again the fibre of \(P_dF\to P_{d-1}F\), but its classification takes a particularly rigid form in terms of equivariant spectra, as discussed below [2508.03808].

## 3. Classification in Weiss calculus and comparison with Goodwillie theory

A central result of recent work in Weiss calculus is a Dwyer–Rezk-type classification: the \(\infty\)-category of reduced, filtered-colimit-preserving \(d\)-polynomial functors
\[
F\colon \mathrm{Vect}_k\to \mathrm{Sp}
\]
is equivalent to the \(\infty\)-category
\[
\mathrm{Fun}(\mathrm{OEpi}_{\le d},\mathrm{Sp}),
\]
where \(\mathrm{OEpi}_{\le d}\) has objects \(k^0,k^1,\dots,k^d\) and morphisms given by surjective linear isometries, i.e. orthogonal epimorphisms. Under this equivalence, a functor is determined by its values on \(k^n\) for \(n\le d\) together with the coherence encoded by orthogonal epimorphisms [2508.03808].

The proof follows the “Schwede–Shipley plus Yoneda” pattern. One first shows that \(\mathrm{Poly}^d(\mathrm{Vect}_k,\mathrm{Sp})\) is compactly generated by the objects
\[
S^{n(-)}\colon V\mapsto \Sigma^\infty S^{k^n\otimes V},
\qquad n=0,1,\dots,d.
\]
An \(\infty\)-categorical Schwede–Shipley theorem then identifies the polynomial category with spectral presheaves on the full spectral subcategory generated by these objects, and a cross-effect calculation shows that this small spectral category is the free spectral enrichment of \(\mathrm{OEpi}_{\le d}\). The resulting classification also identifies mapping spectra in the polynomial category with mapping spectra of the corresponding diagrams [2508.03808].

The homogeneous case is extracted by forcing the lower stages to vanish. A diagram in \(\mathrm{Fun}(\mathrm{OEpi}_{\le d},\mathrm{Sp})\) that is trivial on \(k^0,\dots,k^{d-1}\) is supported only at \(k^d\), and since \(\mathrm{OEpi}_{\le d}\) has no nontrivial morphisms out of \(k^d\) except automorphisms, one obtains
\[
\mathrm{Homog}^d(\mathrm{Vect}_k,\mathrm{Sp}) \simeq \mathrm{Fun}(B\mathrm{Aut}(d),\mathrm{Sp}).
\]
Concretely, a Borel–\(\mathrm{Aut}(d)\)-spectrum \(\Theta\) gives the homogeneous functor
\[
V\mapsto (S^{k^d\otimes V}\wedge \Theta)_{h\mathrm{Aut}(d)}.
\]
This recovers the classical identification of homogeneous Weiss functors with equivariant spectra [2508.03808].

The comparison with Goodwillie theory is formal but not identical. In Goodwillie calculus, finite sets of size \(\le d\) and surjections are replaced in Weiss calculus by Euclidean spaces \(k^n\) and orthogonal epimorphisms. Both theories use compact generators and enriched Yoneda methods, but the Weiss case exhibits less transfer structure: polynomial Weiss functors are not genuine Mackey functors on \(\mathrm{OEpi}_{\le d}\) but only presheaves, whereas in spectra-to-spectra settings one encounters full Mackey or biset phenomena. A common misconception is therefore that the Weiss classification is a literal orthogonal analogue of all Goodwillie features; the formal resemblance is exact at the level of the classification statement, but the Mackey-theoretic content differs [2508.03808].

## 4. Manifold calculus, configuration spaces, and classifying spaces

In manifold calculus, homogeneous functors arise from contravariant functors on the poset \(\mathcal O(M)\) of open subsets of a smooth manifold \(M\). The polynomial degree-\(\le k\) condition is defined using puncturing by \(k+1\) pairwise disjoint closed subsets, and the \(k\)-th Taylor approximation is obtained by homotopy right Kan extension from the full subposet \(\mathcal O_k(M)\) of opens diffeomorphic to at most \(k\) disjoint balls. A homogeneous functor of degree \(k\) is polynomial of degree \(\le k\) with trivial \((k-1)\)-st stage [1708.02642] [1807.06120].

A key geometric fact is that the subposet \(\mathcal O^{=k}(M)\) of opens diffeomorphic to exactly \(k\) disjoint balls is modeled on the unordered configuration space
\[
\mathrm{Conf}(k,M),
\]
and the nerve \(N(\mathcal O^{=k}(M))\) is weakly equivalent to \(\mathrm{Conf}(k,M)\). This leads to a weak equivalence between the category of homogeneous cofunctors of degree \(k\) and the category of linear cofunctors on \(\mathcal O(\mathrm{Conf}(k,M))\), provided the target simplicial model category has a zero object [1708.02642].

For very good homogeneous functors—those that send isotopy equivalences to isomorphisms—Songhafouo Tsopmè and Stanley prove a sharper statement:
\[
\mathrm{VGHF}_k(M;C)\simeq \mathrm{Fun}^{ctr}(\Pi(F_k(M)),C),
\]
where \(F_k(M)\) is the unordered configuration space of \(k\) points in \(M\), \(\Pi(F_k(M))\) its fundamental groupoid, and \(C\) a category with zero object and all small limits. If \(F_k(M)\) is connected, this becomes an equivalence with representations of \(\pi_1(F_k(M))\) in \(C\). In this formulation, homogeneous functors behave like configuration-space local systems [1705.01202].

A second line of classification fixes a fibrant–cofibrant object \(A\) in a simplicial model category \(\mathcal M\) and considers homogeneous functors of degree \(k\) that take a disjoint union of \(k\) balls to an object weakly equivalent to \(A\). Tsopméné and Stanley construct a space \(\hat A\) and show that the set of weak-equivalence classes of such functors is in bijection with
\[
[F_k(M),\hat A].
\]
Subsequent work proves that \(\hat A\) is weakly equivalent to \(B\mathrm{haut}(A)\), the classifying space of the simplicial monoid of self weak equivalences of \(A\). This identifies the classification space intrinsically and extends Weiss’s topological classification to arbitrary simplicial model categories [1807.06120] [2606.25142].

These results make precise the intuition that manifold-calculus homogeneous functors are higher-order local systems parametrized by configuration spaces rather than merely by the manifold itself.

## 5. Algebraic and categorical versions outside homotopy calculus

The term “homogeneous” also appears in purely categorical theories of polynomial functors. In a locally cartesian closed category \(\mathcal E\), a polynomial from \(I\) to \(J\) is a diagram
\[
I \xleftarrow{s} E \xrightarrow{p} B \xrightarrow{t} J
\]
inducing the functor
\[
P=\Sigma_t\Pi_p\Delta_s\colon \mathcal E/I\to \mathcal E/J.
\]
Walker defines homogeneous and monomial terms of such a polynomial using a negation or image-complement operator in the slices \(\mathcal E/B\). Under the hypotheses that \(\mathcal E\) is locally cartesian closed, has disjoint finite coproducts, and finite coproducts are extensive, the \(n\)-th homogeneous term exists for every finite \(n\), and the inclusion of order-\(n\) homogeneous polynomials into all polynomials admits a right adjoint
\[
\mathrm{inc}_n\dashv \mathrm{homterm}_n.
\]
Thus the order-\(n\) homogeneous polynomials form a coreflective sub-bicategory [2208.13088].

This setting is conceptually parallel to Goodwillie- and Weiss-style towers, but the technical mechanism is different. Instead of cross-effects and excision, the construction uses distributivity pullbacks, negation, and the dense–closed orthogonal factorization system induced by a strict initial object. The same framework also defines derivatives of polynomial functors as right adjoints to multiplication by the linear polynomial, and after localizing at dense monomorphisms every polynomial admits such a derivative [2208.13088].

A further algebraic instance appears in the theory of homogeneous strict polynomial functors on vector superspaces. Axtell introduces the categories
\[
\mathrm{Pol}^I_{d,k}= \mathrm{Fct}_k(\mathbb I_d,\mathrm{svec}_k),\qquad
\mathrm{Pol}^{II}_{d,k}= \mathrm{Fct}_k(\mathbb{II}_d,\mathrm{svec}_k),
\]
whose objects are homogeneous strict polynomial functors of degree \(d\). For \(m,n\ge d\) in type I, and \(n\ge d\) in type II, evaluation induces equivalences with supermodule categories over the Schur superalgebras \(S(m|n,d)\) and \(Q(n,d)\), respectively. In type II, the representable functor \(\Gamma^d\) is projective, its endomorphism superalgebra is the Sergeev superalgebra \(W(d)=C(1)\wr \Sigma_d\), and this yields an exact Sergeev duality functor
\[
\mathrm{Hom}_{\mathrm{Pol}^{II}_d}(\Gamma^d,-)\colon \mathrm{Pol}^{II}_{d,k}\to W(d)\text{-smod}.
\]
Here “homogeneous” does not refer to a fibre in a Taylor tower, but to degree-\(d\) strict polynomiality [1302.0042].

## 6. Structural themes, examples, and interpretive cautions

Several structural themes recur across these theories. First, homogeneous functors are usually controlled by a smaller classifying object than arbitrary polynomial functors: \(B\mathrm{Aut}(d)\) in Weiss calculus, \(\Pi(F_k(M))\) or \(B\mathrm{haut}(A)\) in manifold calculus, and \(\Sigma_n\)-action data in Goodwillie’s stable setting [2508.03808] [1705.01202] [2606.25142] [2412.12281].

Second, homogeneous layers often furnish the basic decomposition of more complicated functors. In Goodwillie theory this is explicit in the fibre sequence
\[
D_nF\to P_nF\to P_{n-1}F,
\]
and in the rational spectral case every reduced \(d\)-excisive functor splits as a product of its homogeneous layers. In Weiss calculus, the vanishing of \(P_{d-1}F\) removes all data below dimension \(d\), leaving only the top stratum \(k^d\) and its automorphism action. In manifold calculus, vanishing on fewer than \(k\) balls forces the essential information onto exactly \(k\)-component opens, hence onto configuration spaces [2412.12281] [2508.03808] [1708.02642].

Third, the literature makes clear that homogeneous functors are not synonymous with linear functors, except in degree \(1\). Degree \(1\) recovers locally constant or linear behavior in several settings—for example, \(F_{1,A}(O;\mathcal M)/\!\sim \cong [M,\hat A]\) in manifold calculus, and \(\mathrm{Exc}_1(\mathrm{Sp})_{\mathbb Q}\simeq \mathrm{Sp}_{\mathbb Q}\simeq \mathrm{Ch}(\mathbb Q)\) in the rational spectral setting—but higher degrees encode genuinely multi-variable or multi-point interactions [1807.06120] [2412.12281].

Finally, a plausible implication of the comparison results is that homogeneous functors form the most robust point of contact among distinct calculi of functors. The definitions of polynomiality vary—from cocartesian cubes, to orthogonal splittings, to punctured open sets, to distributivity pullbacks—but the homogeneous stage repeatedly emerges as the object that is most classifiable, most representation-theoretic, and most amenable to explicit models.

Source: https://www.emergentmind.com/topics/homogeneous-functors