Equivariant Postnikov Towers and the Slice Filtration

#1. Non-Equivariant Postnikov Towers|1. Non-Equivariant Postnikov Towers
#2. Equivariant Postnikov Towers: Unstable|2. Equivariant Postnikov Towers: Unstable
#3. Why the Naive Stable Analogue Fails|3. Why the Naive Stable Analogue Fails
#4. The Slice Filtration|4. The Slice Filtration
#5. The Slice Spectral Sequence|5. The Slice Spectral Sequence
#6. Comparison and the HHR Application|6. Comparison and the HHR Application
- #6.1 The Three-Way Comparison Table|6.1 The Three-Way Comparison Table
- #6.2 The Slices of the Real Cobordism Spectrum|6.2 The Slices of the Real Cobordism Spectrum
#7. CW-Postnikov Duality and Eckmann-Hilton|7. CW-Postnikov Duality and Eckmann-Hilton
#References|References

1. Non-Equivariant Postnikov Towers 📐

We begin with a careful review of the classical theory. This section establishes the precise statements that will be upgraded in the equivariant and stable settings.

1.1 Postnikov Sections: Definition and Characterization

Throughout this section, $X$ is a based, connected topological space (or spectrum, when noted).

Definition (Postnikov Section). A Postnikov section of $X$ at level $n$ is a space $P_n X$ equipped with a map $p_n: X \to P_n X$ such that:

$(p_n)_*: \pi_k(X) \xrightarrow{\sim} \pi_k(P_n X)$ is an isomorphism for all $k \leq n$,
$\pi_k(P_n X) = 0$ for all $k > n$,
$X \simeq \operatorname{holim}_n P_n X$ (convergence of the tower).

The natural transformations $P_n X \to P_{n-1} X$ are induced by the universal property: since $\pi_k(P_{n-1} X) = 0$ for $k \geq n$ and $\pi_k(P_n X) = \pi_k(X)$ for $k \leq n-1$, there is an essentially unique compatible map between sections.

The collection $\{\cdots \to P_2 X \to P_1 X \to P_0 X\}$ is the Postnikov tower of $X$.

Duality with CW complexes The Postnikov tower is, in a precise sense, dual to the cellular (CW) filtration. The CW filtration is a colimit of skeleta $X^0 \hookrightarrow X^1 \hookrightarrow X^2 \hookrightarrow \cdots$ where each attachment kills only finitely many cells; the Postnikov tower is a limit of stages $\cdots \to P_2 X \to P_1 X \to P_0 X$ where each stage kills infinitely many cells. The axiomatization of Postnikov towers in a triangulated category is called a t-structure; the dual axiomatization of CW filtrations is a weight structure.

1.2 Construction via the Small Object Argument

There is a canonical point-set construction. The naive attempt — given $\alpha \in \pi_{n+1}(X)$, form a pushout \[\begin{tikzcd} S^{n+1} \arrow[r, "\alpha"] \arrow[d, hook] & X \arrow[d] \\ D^{n+2} \arrow[r] & X \cup_\alpha D^{n+2} \end{tikzcd}\] to kill $\alpha$ — is not functorial: it depends on the choice of representative $\alpha$.

The fix is the small object argument. Fix $n \geq 0$. Define:

\[X^{(n+1)} := X \cup \left( \coprod_{\alpha: S^{n+1} \to X} D^{n+2} \right),\]

where we attach a disk along every map $S^{n+1} \to X$ simultaneously. Iterate: form $X^{(N)}$ by attaching disks along all maps $S^N \to X^{(N-1)}$ for each $N > n+1$. Set

\[P_n X := \operatorname{colim}_{N \geq n+1} X^{(N)}.\]

This is functorial in $X$: a map $f: X \to Y$ induces maps between the pushout diagrams at each stage by post-composition.

Why the iteration is necessary Attaching a disk to kill $\alpha \in \pi_{n+1}(X)$ may create new elements in $\pi_{n+2}$ (via the attaching map $S^{n+1} \to X$, viewed as a class in the relative homotopy). One must iterate to kill all these new contributions. The colimit converges because at each stage we only affect homotopy groups in degrees $> n$.

Proposition. The map $X \to P_n X$ constructed above satisfies the Postnikov section axioms. The proof follows by induction on the relative homotopy long exact sequence and the observation that the colimit is filtered, so the homotopy groups of $P_n X$ are computed as the colimit of the homotopy groups.

1.3 The Fiber Sequences and k-Invariants

Proposition (Fiber Sequence). For each $n \geq 1$, there is a homotopy fiber sequence

\[K(\pi_n X, n) \longrightarrow P_n X \longrightarrow P_{n-1} X.\]

Proof sketch. Let $F_n$ be the homotopy fiber of $P_n X \to P_{n-1} X$. From the long exact sequence on homotopy groups:

\[\cdots \to \pi_k(F_n) \to \pi_k(P_n X) \to \pi_k(P_{n-1} X) \to \pi_{k-1}(F_n) \to \cdots\]

By the Postnikov section axioms: - For $k < n$: $\pi_k(P_n X) \cong \pi_k(X) \cong \pi_k(P_{n-1} X)$, so $\pi_k(F_n) = 0$. - For $k = n$: $\pi_n(P_n X) \cong \pi_n(X)$ and $\pi_n(P_{n-1} X) = 0$, so $\pi_n(F_n) \cong \pi_n(X)$. - For $k > n$: $\pi_k(P_n X) = 0$ and $\pi_k(P_{n-1} X) = 0$, so $\pi_k(F_n) = 0$.

Thus $F_n$ has exactly one nonzero homotopy group, $\pi_n(F_n) \cong \pi_n(X)$, in degree $n$, so $F_n \simeq K(\pi_n X, n)$. $\square$

The map $P_n X \to P_{n-1} X$ is classified by an element

\[k^{n+2}(X) \in H^{n+2}(P_{n-1} X;\, \pi_n X),\]

called the (n+2)-th) k-invariant of $X$. Here we use the bijection $[P_{n-1} X, K(\pi_n X, n+1)] \cong H^{n+2}(P_{n-1} X; \pi_n X)$ (noting that $K(\pi_n X, n+1)$ is the classifying space for $H^{n+1}(-; \pi_n X)$, so a map $P_{n-1} X \to K(\pi_n X, n+1)$ — whose homotopy fiber is $K(\pi_n X, n)$ — is classified by a class in $H^{n+2}$). Iterating the k-invariants completely determines the Postnikov tower, and hence the homotopy type of $X$, up to extensions.

[!EXAMPLE]- k-invariant for the 2-sphere For $X = S^2$, the first nontrivial k-invariant lives in $H^4(K(\mathbb{Z},2); \mathbb{Z}) \cong \mathbb{Z}$, generated by the fundamental class. This k-invariant records the Hopf fibration $S^3 \to S^2$; its nontriviality reflects $\pi_3(S^2) \cong \mathbb{Z}$.

Exercise 1 (Computational) This exercise computes the first two nontrivial Postnikov sections of $S^2$ explicitly, grounding the abstract fiber sequence story in a familiar example.

Show $P_1(S^2) \simeq *$ and $P_2(S^2) \simeq K(\mathbb{Z}, 2) = \mathbb{CP}^\infty$. What is the map $S^2 \to \mathbb{CP}^\infty$?
The k-invariant $k^4 \in H^4(\mathbb{CP}^\infty; \mathbb{Z})$ for $P_3(S^2) \to P_2(S^2) = \mathbb{CP}^\infty$ classifies the extension. Compute $H^4(\mathbb{CP}^\infty; \mathbb{Z})$ and identify the k-invariant. What does its nontriviality say about $\pi_3(S^2)$?

[!TIP]- Solution to Exercise 1 (a) $\pi_1(S^2) = 0$, so $P_1(S^2)$ is contractible. $P_2(S^2)$ has $\pi_2 = \mathbb{Z}$ and all other homotopy groups zero, so $P_2(S^2) \simeq K(\mathbb{Z}, 2) = \mathbb{CP}^\infty$. The map $S^2 \to \mathbb{CP}^\infty$ is the inclusion of the bottom cell (the standard embedding of $S^2 \cong \mathbb{CP}^1 \hookrightarrow \mathbb{CP}^\infty$).

$H^*(\mathbb{CP}^\infty; \mathbb{Z}) = \mathbb{Z}[c]$ with $|c| = 2$, so $H^4(\mathbb{CP}^\infty; \mathbb{Z}) \cong \mathbb{Z}$, generated by $c^2$. The k-invariant $k^4 \in \mathbb{Z}$ is the generator $\pm 1$ (it is nontrivial, since $\pi_3(S^2) \cong \mathbb{Z} \neq 0$). A trivial k-invariant would give $P_3(S^2) \simeq \mathbb{CP}^\infty \times K(\mathbb{Z}, 3)$, which is wrong. The Hopf fibration $\eta: S^3 \to S^2$ represents the generator of $\pi_3(S^2)$ and the nontriviality of $k^4$ is exactly the statement that $\eta$ does not extend to $\mathbb{CP}^\infty$.

1.4 Postnikov Towers as Localization

The Postnikov section $P_n$ has a clean categorical description: it is $n$-truncation, or localization with respect to the class of maps that are $\pi_k$-isomorphisms for $k \leq n$ and $\pi_k$-surjections for $k = n+1$.

More precisely, $P_n$ is the left adjoint (in the homotopy category) to the inclusion of $n$-truncated spaces (spaces with $\pi_k = 0$ for $k > n$) into all spaces. The unit of this adjunction is the map $X \to P_n X$.

This localization perspective is crucial: it will generalize directly to the equivariant stable setting, where the slice sections $P^n$ are defined as localizations with respect to an appropriate class of spectra.

1.5 The Atiyah-Hirzebruch Spectral Sequence

For a generalized (co)homology theory $E_*$, the Postnikov tower of a space $X$ gives rise to the Atiyah-Hirzebruch spectral sequence (AHSS). The key observation is that the fiber sequences

\[\Sigma^n H\pi_n(X) \to \Sigma^\infty P_n X \to \Sigma^\infty P_{n-1} X\]

(after suspending to spectra) form an exact couple whose associated spectral sequence has:

\[E_2^{p,q} = H^p(X;\, E_q(*)) \Implies E^{p+q}(X).\]

The $E_2$-page is the Bredon cohomology of $X$ with coefficients in the homotopy groups of $E$; the spectral sequence converges to the $E$-cohomology of $X$. The analogous construction in the equivariant stable setting — replacing the Postnikov tower with the slice tower — produces the slice spectral sequence.

t-structures The abstract framework for Postnikov towers in a triangulated (or stable $\infty$-) category is a t-structure $(\mathcal{C}_{\geq 0}, \mathcal{C}_{\leq 0})$: a pair of full subcategories closed under appropriate suspensions with the property that every object $X$ fits into a unique fiber sequence $X_{\geq n} \to X \to X_{\leq n-1}$. The slice filtration will be a t-structure on $\mathrm{Sp}^G$ (though with integer indexing replaced by a more refined dimension function).

2. Equivariant Postnikov Towers: Unstable 💡

We now pass to the equivariant setting, working in $G\mathbf{Top}$ (see concepts/equivariant-stable-homotopy/g-spaces-and-equivariant-maps|G-Spaces and Equivariant Maps for foundations). The key insight — due to Elmendorf — is that the equivariant Postnikov tower is completely determined by the ordinary Postnikov towers of all the fixed-point spaces simultaneously.

2.1 Equivariant Homotopy Groups as Coefficient Systems

Recall from concepts/equivariant-stable-homotopy/g-spaces-and-equivariant-maps|G-Spaces and Equivariant Maps that the equivariant homotopy groups of a based $G$-space $X$ are defined, for each closed subgroup $H \leq G$, as

\[\pi_n^H(X) := \pi_n(X^H),\]

the $n$-th homotopy group of the $H$-fixed-point space.

Definition (Coefficient System). A coefficient system for $G$ is a functor $\underline{M}: \mathcal{O}_G^{op} \to \mathbf{Ab}$, where $\mathcal{O}_G$ is the orbit category of $G$ (objects: orbits $G/H$ for closed $H \leq G$; morphisms: $G$-equivariant maps $G/H \to G/K$, corresponding to elements $g \in G$ with $gHg^{-1} \leq K$).

The equivariant homotopy groups assemble into a coefficient system: define

\[\underline{\pi}_n X: \mathcal{O}_G^{op} \to \mathbf{Ab}, \qquad G/H \mapsto \pi_n(X^H).\]

The functoriality is as follows: a $G$-map $G/H \to G/K$ (corresponding to conjugation by some $g$) induces a map $X^K \to X^H$ on fixed-point spaces (by precomposition), and hence a map $\pi_n(X^K) \to \pi_n(X^H)$ — these are the restriction maps of the coefficient system.

Coefficient systems vs. Mackey functors A coefficient system has only restriction maps (contravariant functoriality). A Mackey functor additionally has transfer maps (covariant functoriality) satisfying a double-coset formula. Stably, homotopy groups always form Mackey functors; unstably, they only form coefficient systems. This distinction is invisible non-equivariantly (where there are no proper subgroups to transfer along) but is fundamental in the equivariant stable setting. See Section 3.3 for more.

Exercise C (Computational) This exercise computes the coefficient system of equivariant homotopy groups for the orbits of $G$, identifying a basic class of $G$-spaces that are already their own Postnikov sections.

Let $G = C_2$. Recall the orbit category $\mathcal{O}_{C_2}$ has two objects: $C_2/e$ and $C_2/C_2$.

Regard each orbit $G/H$ as a $C_2$-space. Compute $\underline{\pi}_n(C_2/e)$ and $\underline{\pi}_n(C_2/C_2)$ for $n \geq 0$. That is, compute the abelian group at each object of $\mathcal{O}_{C_2}^{op}$ and the restriction map.
Show that both orbits $C_2/e$ and $C_2/C_2$ are already their own 0-th Postnikov sections: $P_0^{C_2}(G/H) \simeq G/H$ for $H \in \{e, C_2\}$.
Conclude that any discrete $C_2$-set $S$ (with $C_2$-action) satisfies $P_0^{C_2}(S) \simeq S$. Why does this follow from Elmendorf’s theorem without further computation?

[!TIP]- Solution to Exercise C (a) For $G/e = C_2$ (the free orbit, underlying set $\{e, \tau\}$ with $\tau$ acting by the nontrivial element): - $(C_2/e)^e = C_2$ (two points), $(C_2/e)^{C_2} = \emptyset$ (no fixed points — $\tau$ acts freely). - $\underline{\pi}_0(C_2/e)(C_2/e) = \pi_0(C_2) = \mathbb{Z}/2$ (two path components). - $\underline{\pi}_0(C_2/e)(C_2/C_2) = \pi_0(\emptyset)$: undefined (basepoint issue; as an unpointed set, $\emptyset$). - $\underline{\pi}_n(C_2/e) = 0$ for $n \geq 1$ (both $(C_2/e)^e = C_2$ and $(C_2/e)^{C_2} = \emptyset$ are discrete, so all higher homotopy groups vanish).

For $G/C_2 = *$ (the trivial orbit, one point with trivial action): - $(*) ^e = *$, $(*) ^{C_2} = *$. - $\underline{\pi}_0(*)(C_2/e) = \pi_0(*) = 0$, $\underline{\pi}_0(*)(C_2/C_2) = \pi_0(*) = 0$. - $\underline{\pi}_n(*) = 0$ for all $n$ (a point has trivial homotopy).

$P_0^{C_2}(X)$ is characterized by $\pi_k^H(P_0^{C_2}(X)) = 0$ for $k \geq 1$ and all $H$. Since both orbits have $\pi_k^H = 0$ for $k \geq 1$ (they are discrete), they are already 0-truncated. So $P_0^{C_2}(G/H) \simeq G/H$.

Any discrete $C_2$-set $S$ is a disjoint union of orbits $C_2/e$ and $C_2/C_2$, which are both already 0-truncated. Elmendorf’s theorem says $P_0^{C_2}(S)$ is computed objectwise on fixed-point sets: $(P_0^{C_2}(S))^H \simeq P_0(S^H)$. Since $S^H$ is discrete for each $H$, $P_0(S^H) \simeq S^H$. Hence $P_0^{C_2}(S) \simeq S$.

2.2 Construction via Elmendorf’s Theorem

Elmendorf’s Theorem (recalled from concepts/equivariant-stable-homotopy/g-spaces-and-equivariant-maps|G-Spaces and Equivariant Maps): There is a Quillen equivalence

\[G\mathbf{Top} \simeq \mathrm{Fun}(\mathcal{O}_G^{op}, \mathbf{Spaces}),\]

under which a $G$-space $X$ corresponds to the presheaf $G/H \mapsto X^H$.

This equivalence immediately gives us equivariant Postnikov sections for free:

Definition (Equivariant Postnikov Section). The equivariant Postnikov section $P_n^G X$ is the $G$-space corresponding, under Elmendorf’s theorem, to the presheaf

\[G/H \mapsto P_n(X^H).\]

Equivalently:

Proposition. $P_n^G X$ is uniquely characterized (up to weak $G$-equivalence) by: 1. $\pi_k^H(P_n^G X) \cong \pi_k^H(X)$ for all closed $H \leq G$ and $k \leq n$, 2. $\pi_k^H(P_n^G X) = 0$ for all closed $H \leq G$ and $k > n$, 3. $X \simeq \operatorname{holim}_n P_n^G X$.

The fixed-point formula $(P_n^G X)^H \simeq P_n(X^H)$ is the key identity: the equivariant Postnikov section is computed objectwise on fixed-point spaces.

As an application of Elmendorf Blumberg’s lecture notes explicitly identify the equivariant Postnikov tower as “another application of Elmendorf’s theorem.” The equivalence $G\mathbf{Top} \simeq \mathrm{Fun}(\mathcal{O}_G^{op}, \mathbf{Spaces})$ reduces any homotopy-theoretic construction on $G$-spaces — such as Postnikov sections — to the corresponding objectwise construction on presheaves of spaces. This is the power of the Elmendorf perspective.

The construction of Section 1.2 applies objectwise: form $(P_n^G X)^H$ by the small object argument applied to $X^H$, using all maps $S^k \to X^H$ for $k > n$. The equivariance is automatic because the construction is functorial in $X^H$, and the fixed-point functors $X \mapsto X^H$ are functorial in $G$-equivariant maps.

Exercise D (Proof) This exercise proves the equivariant Whitehead theorem, identifying the precise role of Elmendorf’s theorem in the argument.

Let $f: X \to Y$ be a $G$-equivariant map between $G$-CW complexes. Suppose $f$ induces isomorphisms $\pi_n^H(X) \xrightarrow{\sim} \pi_n^H(Y)$ for all closed subgroups $H \leq G$ and all $n \geq 0$.

Using Elmendorf’s theorem, translate the hypothesis into a statement about the induced map of presheaves $\mathcal{O}_G^{op} \to \mathbf{Spaces}$.
Apply the classical (non-equivariant) Whitehead theorem objectwise to conclude that $f^H: X^H \to Y^H$ is a homotopy equivalence for each $H$.
Conclude that $f$ is a $G$-homotopy equivalence. Where is the $G$-CW assumption used?

[!TIP]- Solution to Exercise D (a) Under Elmendorf’s equivalence, $f: X \to Y$ corresponds to a natural transformation $\hat{f}: \hat{X} \to \hat{Y}$ between presheaves, where $\hat{X}(G/H) = X^H$. The hypothesis $\pi_n^H(f): \pi_n(X^H) \xrightarrow{\sim} \pi_n(Y^H)$ for all $H$ and $n$ says precisely that $\hat{f}(G/H): X^H \to Y^H$ is a weak homotopy equivalence for each $H$ — i.e., $\hat{f}$ is an objectwise weak equivalence.

The classical Whitehead theorem states: a weak homotopy equivalence between CW complexes is a homotopy equivalence. Since $X$ is a $G$-CW complex, each fixed-point space $X^H$ is a CW complex (the $H$-fixed points of a $G$-CW complex form a sub-CW complex). Similarly $Y^H$. So the classical Whitehead theorem applies: $f^H: X^H \xrightarrow{\simeq} Y^H$ is a homotopy equivalence for each $H$.

An objectwise homotopy equivalence of presheaves is an equivalence in the functor category $\mathrm{Fun}(\mathcal{O}_G^{op}, \mathbf{Spaces})$. Under Elmendorf’s Quillen equivalence, this corresponds to a $G$-homotopy equivalence $X \xrightarrow{\simeq} Y$. The $G$-CW assumption is used in step (b): without it, $X^H$ and $Y^H$ need not be CW complexes, so the classical Whitehead theorem does not directly apply. (One can still conclude that $f$ is a weak $G$-equivalence using only the isomorphism on equivariant homotopy groups, but the stronger conclusion of a $G$-homotopy equivalence requires the CW structure.)

2.3 Equivariant Eilenberg-Mac Lane Spaces

Definition (Equivariant Eilenberg-Mac Lane Space). For a coefficient system $\underline{M}: \mathcal{O}_G^{op} \to \mathbf{Ab}$ and an integer $n \geq 1$, the equivariant Eilenberg-Mac Lane space $K(\underline{M}, n)$ is the $G$-space corresponding under Elmendorf’s theorem to the presheaf

\[G/H \mapsto K(\underline{M}(G/H),\, n).\]

That is, $(K(\underline{M},n))^H = K(\underline{M}(G/H), n)$.

Theorem (Representability). For a $G$-CW complex $X$ and a coefficient system $\underline{M}$:

\[[X,\, K(\underline{M}, n)]_G \cong H_G^n(X;\, \underline{M}),\]

where $H_G^*(X; \underline{M})$ denotes Bredon cohomology with coefficients in $\underline{M}$.

Proof sketch. Under Elmendorf’s equivalence, $G$-equivariant maps $X \to K(\underline{M},n)$ correspond to natural transformations between the corresponding presheaves. Since $K(\underline{M}(G/H),n)$ represents $H^n(-; \underline{M}(G/H))$ objectwise, and Bredon cohomology is defined as the cohomology of the chain complex $\mathrm{Hom}_{\mathcal{O}_G}(C_*^{cell}(X), \underline{M})$ (where $C_*^{cell}(X)(G/H) = H_*(X^H, (X^H)^{n-1})$ are the cellular chains), the result follows from the objectwise representability and the Elmendorf equivalence. $\square$

Bredon cohomology Bredon cohomology, introduced by Bredon in 1967, is the “correct” cohomology theory for $G$-spaces. Its coefficient systems take values in all of $\mathcal{O}_G$, not just individual subgroups. The equivariant Eilenberg-Mac Lane spaces show that Bredon cohomology classes have geometric representatives, exactly as in the non-equivariant case.

2.4 Fiber Sequences and Bredon k-Invariants

Proposition (Equivariant Fiber Sequence). For each $n \geq 1$, there is a homotopy fiber sequence of $G$-spaces:

\[K(\underline{\pi}_n X, n) \longrightarrow P_n^G X \longrightarrow P_{n-1}^G X.\]

Proof. Taking $H$-fixed points, we get the ordinary fiber sequence $K(\pi_n(X^H), n) \to P_n(X^H) \to P_{n-1}(X^H)$ for each $H$. Since Elmendorf’s theorem is a Quillen equivalence (in particular, it preserves fiber sequences), the objectwise fiber sequences assemble into a $G$-equivariant fiber sequence. $\square$

The fiber sequence is classified by a Bredon k-invariant

\[k^{n+2}(X) \in H_G^{n+2}(P_{n-1}^G X;\, \underline{\pi}_n X),\]

which lives in Bredon cohomology (not ordinary cohomology). This follows from the representability theorem: the map $P_{n-1}^G X \to K(\underline{\pi}_n X, n+1)$ classifying the fiber sequence corresponds to a class in $[P_{n-1}^G X, K(\underline{\pi}_n X, n+1)]_G \cong H_G^{n+2}(P_{n-1}^G X; \underline{\pi}_n X)$.

The role of coefficient systems vs. Mackey functors here In the unstable equivariant setting, the k-invariants live in Bredon cohomology with coefficient system coefficients — not Mackey functor coefficients. This is because the coefficient system $\underline{\pi}_n X$ only has restriction maps (it sees $X^K \to X^H$ for inclusions $H \hookrightarrow K$), not transfer maps. Stably, this upgrades: the $n$-slice involves Mackey functors with genuine transfer maps. See Section 3.3.

2.5 Non-Equivariant vs. Equivariant Comparison

The following table summarizes the parallel structure between the non-equivariant and equivariant unstable Postnikov towers.

Feature	Non-equivariant $\mathbf{Top}$	Equivariant $G\mathbf{Top}$
Homotopy groups	$\pi_n(X) \in \mathbf{Ab}$	$\underline{\pi}_n X: \mathcal{O}_G^{op} \to \mathbf{Ab}$ (coefficient system)
Postnikov section	$P_n X$, $\pi_k(P_n X) = \begin{cases}\pi_k X & k\leq n \\ 0 & k > n\end{cases}$	$P_n^G X$, same for each $\pi_k^H$
Construction	Small object argument on $X$	Small object argument objectwise on each $X^H$
Eilenberg-Mac Lane	$K(A, n)$, $A \in \mathbf{Ab}$	$K(\underline{M}, n)$, $\underline{M}: \mathcal{O}_G^{op} \to \mathbf{Ab}$
Fiber sequence	$K(\pi_n X, n) \to P_n X \to P_{n-1} X$	$K(\underline{\pi}_n X, n) \to P_n^G X \to P_{n-1}^G X$
k-invariant	$k^{n+2} \in H^{n+2}(P_{n-1} X; \pi_n X)$	$k^{n+2} \in H_G^{n+2}(P_{n-1}^G X; \underline{\pi}_n X)$ (Bredon)
Key tool	Whitehead theorem	Elmendorf’s theorem + objectwise Whitehead

The formal structure is identical; the coefficients upgrade from abelian groups to coefficient systems, and ordinary cohomology upgrades to Bredon cohomology.

Exercise 2 (Computational) This exercise computes the coefficient system of equivariant homotopy groups for a basic $C_2$-space, and identifies the Postnikov section.

Let $G = C_2$ and $X = S^1$ with the reflection action $\tau \cdot e^{i\theta} = e^{-i\theta}$.

Compute the coefficient system $\underline{\pi}_1(X): \mathcal{O}_{C_2}^{op} \to \mathbf{Ab}$. What abelian groups and restriction map does it consist of?
Show that $P_1^{C_2}(S^1_{\mathrm{refl}}) \simeq S^1_{\mathrm{refl}}$ (i.e., the space is already its own first Postnikov section).
What is the equivariant Eilenberg-Mac Lane space $K(\underline{\pi}_1(X), 1)$?

[!TIP]- Solution to Exercise 2 (a) Recall $(S^1)^{C_2} = \{1, -1\} \cong S^0$ (the two real points). So: - $\underline{\pi}_1(X)(C_2/e) = \pi_1(S^1) = \mathbb{Z}$, with $C_2$ acting by $-1$ (reflection reverses orientation, hence acts by inversion on $\pi_1$). - $\underline{\pi}_1(X)(C_2/C_2) = \pi_1((S^1)^{C_2}) = \pi_1(S^0) = 0$. - Restriction map: $0 \to \mathbb{Z}$. The coefficient system is $\underline{\pi}_1(S^1_{\mathrm{refl}}) = (0 \leftarrow \mathbb{Z}^-)$, where $\mathbb{Z}^-$ denotes $\mathbb{Z}$ with $C_2$ acting by $-1$.

For $k > 1$: $\pi_k^e(S^1) = \pi_k(S^1) = 0$ and $\pi_k^{C_2}(S^1) = \pi_k(S^0) = 0$ (for $k \geq 2$). So $\pi_k^H(S^1_{\mathrm{refl}}) = 0$ for all $k > 1$ and all $H$, meaning $S^1_{\mathrm{refl}}$ is already 1-truncated. Thus $P_1^{C_2}(S^1_{\mathrm{refl}}) \simeq S^1_{\mathrm{refl}}$.

$K(\underline{\pi}_1(X), 1) = K(\underline{M}, 1)$ where $\underline{M}(C_2/e) = \mathbb{Z}^-$ and $\underline{M}(C_2/C_2) = 0$. Under Elmendorf: $(K(\underline{M},1))^e = K(\mathbb{Z}, 1) = S^1$ and $(K(\underline{M},1))^{C_2} = K(0,1) = *$. So $K(\underline{M}, 1) \simeq S^1_{\mathrm{antipodal}}$ — the circle with $C_2$ acting antipodally (no fixed points, and the underlying $K(\mathbb{Z},1)$-structure is recovered).

3. Why the Naive Stable Analogue Fails ⚠️

Passing to genuine $G$-spectra indexed by a complete universe $\mathcal{U}$, the situation becomes subtler. The integer-indexed Postnikov filtration fails to capture the correct equivariant structure.

3.1 The Integer-Indexed Stable Postnikov Tower

In the non-equivariant stable category $\mathrm{Sp}$, the Postnikov tower works as follows. Given a spectrum $E$, define $P^n E$ by attaching $\Sigma^{n+1}$-cells (cones on Moore spaces $M(\pi_{n+1}E, n+1)$) to kill $\pi_{n+1}$, then iterating. The construction is indexed by the integers: attaching $\Sigma^k$-cells for $k > n$.

Naively, one might attempt the same in $\mathrm{Sp}^G$ (genuine $G$-spectra): for each subgroup $H \leq G$, attach cells of the form $\Sigma^\infty_G G/H_+ \wedge \Sigma^{n+1} M$ to kill $\pi_{n+1}^H$. This defines an integer-indexed filtration

\[\cdots \to \tau_{\leq n} E \to \tau_{\leq n-1} E \to \cdots\]

where $\pi_k^H(\tau_{\leq n} E) = \begin{cases} \pi_k^H(E) & k \leq n \\ 0 & k > n \end{cases}$ for all $H$.

3.2 Non-Nullhomotopic Maps Between Representation Spheres

The integer-indexed Postnikov filtration has a fundamental deficiency: it is not compatible with the richer sphere spectrum structure of $\mathrm{Sp}^G$.

The Key Issue. In $\mathrm{Sp}^G$, the sphere spectrum is $S = \{S^V\}_{V \in \mathcal{U}}$ (the sphere $G$-spectrum indexed over the complete universe $\mathcal{U}$ of representations). For representations $V \subset W$, there is a canonical map $S^V \to S^W$ induced by the inclusion $V \hookrightarrow W$. This map need not be null-homotopic in the $G$-equivariant stable category, even when $\dim W > \dim V$.

Example. Let $G = C_2$ and let $\sigma$ denote the sign representation of $C_2$ (the 1-dimensional real representation where the generator acts by $-1$). The map

\[S^\sigma \longrightarrow S^{2\sigma}\]

induced by the inclusion $\sigma \hookrightarrow 2\sigma = \sigma \oplus \sigma$ is not null-homotopic in $\mathrm{Sp}^{C_2}$.

Why? Nonequivariantly, $S^\sigma \simeq S^1$ and $S^{2\sigma} \simeq S^2$, so the map is a degree-1 map $S^1 \to S^2$, which is null-homotopic. But equivariantly, one must check the $C_2$-fixed-point level: $(S^\sigma)^{C_2} = S^0$ and $(S^{2\sigma})^{C_2} = S^0$, and the induced map $(S^\sigma)^{C_2} \to (S^{2\sigma})^{C_2}$ is a degree-1 map $S^0 \to S^0$, which is not null-homotopic. Since a $G$-equivariant map is null-homotopic iff its fixed-point maps are all null-homotopic, this map is non-null.

More generally: any map $S^V \to S^W$ with $V \subseteq W$ is non-null-homotopic at the $H$-fixed-point level precisely when $W^H \neq V^H$, i.e., when $W - V$ has $H$-fixed summands. If $V$ contains a trivial summand, all such maps are null-homotopic (since we can use the trivial direction to contract); if $V$ has no trivial summands, non-null maps abound.

Why this breaks the integer Postnikov tower Suppose we build an integer-indexed Postnikov section $\tau_{\leq n} E$ by attaching integer-suspension cells. A slice cell $G_+ \wedge_H S^{m\rho_H}$ of dimension $m|H|$ might not be “seen” by any integer-suspension cell of the right dimension, because the representation sphere $S^{m\rho_H}$ has non-trivial equivariant structure that $\Sigma^{m|H|}$ (with trivial $G$-action) does not capture. The cells of the integer Postnikov tower are “too coarse” — they see the underlying non-equivariant homotopy but miss the representation-theoretic data.

Exercise 3 (Computational) This exercise verifies the non-nullhomotopy of $S^\sigma \to S^{2\sigma}$ directly from fixed-point data, making the failure of the integer Postnikov tower concrete.

Let $G = C_2$ and $\sigma$ the sign representation.

Show that $S^\sigma \simeq S^1$ non-equivariantly and $S^{2\sigma} \simeq S^2$ non-equivariantly. Conclude that the map $S^\sigma \to S^{2\sigma}$ is null non-equivariantly.
Compute $(S^\sigma)^{C_2}$ and $(S^{2\sigma})^{C_2}$. Show the induced map $(S^\sigma)^{C_2} \to (S^{2\sigma})^{C_2}$ is the identity $S^0 \to S^0$ (degree 1), hence not null.
Explain why parts (a) and (b) together show that the map $S^\sigma \to S^{2\sigma}$ is non-null in $\mathrm{Sp}^{C_2}$ but null in $\mathrm{Sp}$ (with trivial $C_2$-action).

[!TIP]- Solution to Exercise 3 (a) Non-equivariantly, $S^\sigma = (\mathbb{R}_\sigma)^+ \simeq S^1$ (1-point compactification of $\mathbb{R}$). The map $\mathbb{R}_\sigma \hookrightarrow \mathbb{R}_\sigma \oplus \mathbb{R}_\sigma$ induces $S^1 \to S^2$, which is null (as $[S^1, S^2] = \pi_1(S^2) = 0$).

$(S^\sigma)^{C_2}$: $C_2$ fixes $0 \in \mathbb{R}$ and $\infty$, giving $(S^\sigma)^{C_2} = \{0, \infty\} \cong S^0$. Similarly $(S^{2\sigma})^{C_2} = \{0, \infty\} \cong S^0$. The map $\sigma \hookrightarrow 2\sigma$ sends $v \mapsto (v, 0)$; on the 1-point compactifications, $0 \mapsto 0$ and $\infty \mapsto \infty$, which is the identity map $S^0 \to S^0$, of degree 1, hence not null.

A map of $G$-spectra is null iff ALL of its fixed-point maps are null (for all $H \leq G$). The $e$-level gives the non-equivariant map (null by (a)); but the $C_2$-level gives degree 1 on $S^0$ (non-null by (b)). Since at least one fixed-point level is non-null, the $C_2$-equivariant map is non-null.

3.3 Mackey Functors vs. Coefficient Systems

Stably, equivariant homotopy groups always form Mackey functors, not merely coefficient systems.

Definition (Mackey Functor). A Mackey functor $\underline{M}$ for $G$ is a pair of functors

\[\underline{M}^*: \mathcal{O}_G^{op} \to \mathbf{Ab}, \qquad \underline{M}_*: \mathcal{O}_G \to \mathbf{Ab}\]

(i.e., a contravariant part encoding restriction maps $\mathrm{res}_H^K: \underline{M}(G/K) \to \underline{M}(G/H)$ and a covariant part encoding transfer maps $\mathrm{tr}_H^K: \underline{M}(G/H) \to \underline{M}(G/K)$) satisfying: 1. $\underline{M}^*(G/H) = \underline{M}_*(G/H)$ for all $H$ (same underlying abelian groups), 2. The Mackey double-coset formula: for $H, K \leq G$, \[\mathrm{res}_H^G \circ \mathrm{tr}_K^G = \sum_{[g] \in H \backslash G / K} \mathrm{tr}_{H \cap gKg^{-1}}^H \circ c_g \circ \mathrm{res}_{g^{-1}Hg \cap K}^K,\] where $c_g$ is conjugation by $g$.

Theorem. For a genuine $G$-spectrum $E \in \mathrm{Sp}^G$, the homotopy groups $\underline{\pi}_n^{Mky}(E)(G/H) := \pi_n^H(E) = [G/H_+ \wedge S^n, E]_G$ assemble into a Mackey functor, with: - Restriction: induced by the $G$-map $G/H_+ \to G/K_+$ (for $H \leq K$), - Transfer: induced by the norm map / umkehr map in the stable category.

The transfer maps arise from the covariant functoriality of the smash product in $\mathrm{Sp}^G$ — they have no counterpart in the unstable setting because the suspension spectrum functor $\Sigma^\infty_G$ does not preserve covariant functoriality of orbit sets.

The integer Postnikov filtration only captures the coefficient system structure (restriction maps); it misses the transfer maps. The slice filtration, built from representation spheres, is sensitive to the full Mackey functor structure.

Exercise F (Computational) This exercise makes the Burnside Mackey functor explicit for $G = C_2$, concretely illustrating the distinction between coefficient systems and Mackey functors.

The Burnside Mackey functor $\underline{A}$ assigns to each $G$-orbit the Burnside ring: $\underline{A}(G/H) = A(H)$, the Grothendieck group of finite $H$-sets.

Compute $\underline{A}(C_2/e) = A(e)$ and $\underline{A}(C_2/C_2) = A(C_2)$ explicitly as abelian groups, listing generators.
Write down the restriction map $\mathrm{res}^{C_2}_e: A(C_2) \to A(e)$ and the transfer map $\mathrm{tr}_e^{C_2}: A(e) \to A(C_2)$ on generators.
Verify the double-coset formula: $\mathrm{res}^{C_2}_e \circ \mathrm{tr}_e^{C_2} = \sum_{[g] \in e \backslash C_2 / e} c_g = \mathrm{id} + \tau_*$, where $\tau$ is the nontrivial element. Conclude $\mathrm{res} \circ \mathrm{tr}([1]) = 2 \in A(e) = \mathbb{Z}$.
Explain why the constant coefficient system $\underline{\mathbb{Z}}$ (with restriction $\mathrm{id}: \mathbb{Z} \to \mathbb{Z}$ and no transfer) is not a Mackey functor: what transfer would be forced by the double-coset formula?

[!TIP]- Solution to Exercise F (a) $A(e) = $ Grothendieck group of finite sets $= \mathbb{Z}$, generated by $[*]$ (one point). $A(C_2) = $ Grothendieck group of finite $C_2$-sets. The indecomposable $C_2$-sets are $[C_2/C_2] = [*]$ (trivial) and $[C_2/e] = [C_2]$ (free orbit). So $A(C_2) \cong \mathbb{Z}^2$, with generators $[C_2/C_2]$ and $[C_2/e]$.

Restriction $\mathrm{res}: A(C_2) \to A(e)$: forget the $C_2$-action and take the underlying set. $[C_2/C_2] = [*] \mapsto [*] = 1$ and $[C_2/e] = [C_2] \mapsto [C_2] = 2[*] = 2$.

Transfer $\mathrm{tr}: A(e) \to A(C_2)$: “induce” the $e$-set to a $C_2$-set. $\mathrm{tr}([*]) = [C_2 \times_e *] = [C_2/e]$ (the free $C_2$-orbit on one point). So $\mathrm{tr}(n) = n \cdot [C_2/e]$.

$\mathrm{res}(\mathrm{tr}([*])) = \mathrm{res}([C_2/e]) = 2$. The double-coset formula: $e \backslash C_2 / e = \{[e], [\tau]\}$ (two cosets). Each contributes $c_g \circ \mathrm{res}^e_{e \cap e} = \mathrm{id}$, so $\mathrm{res} \circ \mathrm{tr} = \mathrm{id} + \mathrm{id} = 2\cdot\mathrm{id}$ on $A(e) = \mathbb{Z}$. So $\mathrm{res}(\mathrm{tr}(1)) = 2$. ✓

For $\underline{\mathbb{Z}}$ to be a Mackey functor with $\underline{\mathbb{Z}}(C_2/e) = \underline{\mathbb{Z}}(C_2/C_2) = \mathbb{Z}$ and restriction $= \mathrm{id}$, the double-coset formula forces $\mathrm{res} \circ \mathrm{tr} = 2\cdot\mathrm{id}$ (as above), so the transfer $\mathrm{tr}: \mathbb{Z} \to \mathbb{Z}$ would need to satisfy $\mathrm{id} \circ \mathrm{tr} = 2\cdot\mathrm{id}$, i.e., $\mathrm{tr}(n) = 2n$. This is the multiplication-by-2 transfer. The “naively constant” Mackey functor has $\mathrm{tr} = 2$; the constant coefficient system ignores this map entirely. They agree as coefficient systems but differ as Mackey functors.

Exercise G (Proof) This exercise identifies the transfer map in stable homotopy concretely, showing why stable homotopy groups carry Mackey functor structure while unstable ones do not.

Let $G = C_2$, and consider the stable transfer map $\mathrm{tr}_e^{C_2}: \pi_n^e(X) \to \pi_n^{C_2}(X)$ for $X \in \mathrm{Sp}^{C_2}$.

The stable transfer for the covering $p: C_2/e \to C_2/C_2$ is a map of $G$-spectra $\Sigma^\infty_G C_{2+} \to \Sigma^\infty_G S^0$ (the “norm map” / “umkehr map”). Identify the counit of the adjunction $\Sigma^\infty_G(C_{2+} \wedge -) \dashv \mathrm{Res}^{C_2}_e$ and explain how it gives a map $\Sigma^\infty_G C_{2+} \to S^0$ in $\mathrm{Sp}^{C_2}$.
Explain why the corresponding unstable transfer does not exist: why is there no natural map $\Sigma(C_2/e)_+ \to S^0$ in $G\mathbf{Top}$ that deserves to be called a transfer? (Hint: consider what happens to fixed-point sets.)
Conclude: the suspension spectrum functor $\Sigma^\infty_G: G\mathbf{Top} \to \mathrm{Sp}^G$ does not preserve the absence of transfers — passing to spectra creates transfer maps that did not exist unstably.

[!TIP]- Solution to Exercise G (a) The adjunction $G_+ \wedge_H (-): \mathrm{Sp}^H \rightleftharpoons \mathrm{Sp}^G: i_H^*$ has counit $\epsilon: G_+ \wedge_H i_H^* X \to X$. For $H = e$ and $X = S^0$: $\epsilon: G_+ \wedge_e S^0 = \Sigma^\infty_G C_{2+} \to S^0$. This is the stable fold map (fold $C_2$ copies of $S^0$ together), and it induces the transfer $\mathrm{tr}_e^{C_2}: \pi_n(X^e) \to \pi_n^{C_2}(X)$ on homotopy groups by precomposition.

The unstable analogue would be a map $C_{2+} = C_2 \sqcup \{+\} \to S^0$ in based $C_2$-spaces. The only based maps $C_{2+} \to S^0$ send $C_2$ to the basepoint, i.e., collapse everything — there is no “fold” in the unstable category that respects the $C_2$-action nontrivially. Checking fixed points: $(C_{2+})^{C_2} = \{+\}$ (the basepoint, since $C_2$ acts freely on $C_2$) and $(S^0)^{C_2} = S^0$. A $C_2$-equivariant map $\{+\} \to S^0$ must land at the basepoint, giving no information. The unstable transfer would need to “pick up” elements of $\pi_n(X^e)$ and push them to $\pi_n^{C_2}(X)$ via a norm construction, but the pointset-level norm is not available without the stable structure.

The suspension spectrum functor is symmetric monoidal (and in particular smash-product-preserving), and the smash product in $\mathrm{Sp}^G$ supports the norm/transfer via the counit of the induction–restriction adjunction. The unstable category $G\mathbf{Top}$ has induction $G_+ \wedge_H -$, but its counit is only a unstable map of spaces, and the fold map $C_{2+} \to S^0$ is only well-defined in spectra (where $S^0$ is the unit for $\wedge$, not just a space). Thus $\Sigma^\infty_G$ creates transfers that do not exist in $G\mathbf{Top}$.

3.4 The Remedy: Representation-Indexed Filtration

The resolution, due to Hill-Hopkins-Ravenel, is to replace the integer indexing with a representation-dimension indexing. Instead of attaching cells $G/H_+ \wedge \Sigma^n D^1$ to kill integer-degree homotopy, one uses slice cells — cells built from representation spheres $S^{m\rho_H}$ — to filter by representation dimension.

The philosophy: the “correct” notion of dimension for the slice filtration is not the integer dimension of a cell, but the representation-theoretic dimension, which counts each irreducible representation with multiplicity according to its role in the regular representation.

4. The Slice Filtration 🔑

We now develop the slice filtration of Hill-Hopkins-Ravenel. Throughout, $G$ is a finite group and $\mathrm{Sp}^G$ denotes the category of genuine $G$-spectra.

4.1 Slice Cells

Definition (Regular Representation). For a finite group $H$, the regular representation $\rho_H$ is the real representation $\mathbb{R}[H]$ with $H$ acting by left multiplication. Its dimension is $|\rho_H| = |H|$.

Definition (Slice Cells, HHR §4). For a subgroup $H \leq G$ and integer $m \geq 0$, the slice cells are the $G$-spectra:

Even slice cell of dimension $m|H|$: \[\widetilde{S}^{m|H|} := G_+ \wedge_H S^{m\rho_H},\] where $S^{m\rho_H}$ is the one-point compactification of $m\rho_H = \rho_H^{\oplus m}$, and $G_+ \wedge_H -$ is the induction from $H$ to $G$.
Odd slice cell of dimension $m|H| - 1$: \[\widetilde{S}^{m|H|-1} := \Sigma^{-1}(G_+ \wedge_H S^{m\rho_H}).\]

The dimension of $G_+ \wedge_H S^{m\rho_H}$ is $m|H|$ because $\dim_\mathbb{R}(m\rho_H) = m \cdot |H|$, and suspension by a representation of dimension $d$ raises underlying (non-equivariant) dimension by $d$.

[!EXAMPLE]- Slice cells for G = C2 Let $G = C_2$, $H = C_2$. The regular representation is $\rho_{C_2} = \mathbb{R} \oplus \sigma$ where $\sigma$ is the sign representation ($\dim \rho_{C_2} = 2$). The even slice cells are: - $m=0$: $G_+ \wedge_{C_2} S^0 = S^0$ (dimension 0) - $m=1$: $G_+ \wedge_{C_2} S^{\rho_{C_2}} = G_+ \wedge_{C_2} S^{1+\sigma}$ (dimension 2) - $m=2$: $G_+ \wedge_{C_2} S^{2\rho_{C_2}}$ (dimension 4)

For $H = e$ (trivial subgroup), $\rho_e = \mathbb{R}$ (trivial 1-dimensional representation), $|H| = 1$. The slice cells are $G_+ \wedge_e S^m = \Sigma^\infty_G G_+ \wedge S^m$ — just the free $G$-spectrum on $S^m$. These are the “ordinary” suspension cells.

Surprisingly, the slice cells for $H = e$ are exactly the cells that appear in the ordinary (integer-indexed) Postnikov tower. The extra structure comes from the nontrivial subgroup cells.

Why the regular representation? The choice of $\rho_H$ (rather than some other representation) is not arbitrary. The regular representation has the property that $\rho_H^{C_2} = 0$ for all proper subgroups $H' < H$ of $G$ that don’t fix $H$ — this ensures that the slice cells behave well under restriction and induction. More precisely, the regular representation is the unique (up to multiples) representation whose character is the delta function at the identity of $H$, which makes the slice cells “concentrated” at the right level of the subgroup lattice.

Exercise 4 (Computational) This exercise enumerates all $C_2$-slice cells of low dimension, making the slice filtration concrete for the simplest non-trivial group.

For $G = C_2$, list all slice cells $G_+ \wedge_H S^{m\rho_H}$ and their odd counterparts $\Sigma^{-1} G_+ \wedge_H S^{m\rho_H}$ of dimension $\leq 4$. For each, identify: - The subgroup $H$ and integer $m$ - The underlying non-equivariant space - The equivariant structure ($C_2$-action)

[!TIP]- Solution to Exercise 4 For $G = C_2$, the subgroups are $H = e$ and $H = C_2$.

$H = e$ (free cells, $\rho_e = \mathbb{R}^1$ trivial, $|\mathrm{e}| = 1$, dim $= m$):

$m$ Cell $C_{2+} \wedge S^m$ Dim $C_2$-action

0 $C_{2+} \wedge S^0 = \Sigma^\infty_+ C_2$ 0 Free (permutes two copies)

1 $C_{2+} \wedge S^1$ 1 Free

2 $C_{2+} \wedge S^2$ 2 Free

3 $C_{2+} \wedge S^3$ 3 Free

4 $C_{2+} \wedge S^4$ 4 Free

$H = C_2$ (regular rep cells, $\rho_{C_2} = 1 + \sigma$, $|C_2| = 2$, even dim $= 2m$):

$m$ Even cell $S^{m\rho_{C_2}}$ Dim Odd cell $\Sigma^{-1}S^{m\rho_{C_2}}$ Odd dim

1 $S^{1+\sigma} = S^1 \wedge S^\sigma$ 2 $S^\sigma$ 1

2 $S^{2+2\sigma}$ 4 $S^{1+2\sigma}$ 3

So at dimension 1: two slice cells — the free $C_{2+} \wedge S^1$ and the representation-sphere $S^\sigma$. At dimension 2: two cells — $C_{2+} \wedge S^2$ and $S^{1+\sigma}$. This multiplicity at each dimension (one free, one from the regular representation) is a key feature of $C_2$-equivariant stable homotopy theory.

\(m\)	Cell \(C_{2+} \wedge S^m\)	Dim	\(C_2\)-action
0	\(C_{2+} \wedge S^0 = \Sigma^\infty_+ C_2\)	0	Free (permutes two copies)
1	\(C_{2+} \wedge S^1\)	1	Free
2	\(C_{2+} \wedge S^2\)	2	Free
3	\(C_{2+} \wedge S^3\)	3	Free
4	\(C_{2+} \wedge S^4\)	4	Free

\(m\)	Even cell \(S^{m\rho_{C_2}}\)	Dim	Odd cell \(\Sigma^{-1}S^{m\rho_{C_2}}\)	Odd dim
1	\(S^{1+\sigma} = S^1 \wedge S^\sigma\)	2	\(S^\sigma\)	1
2	\(S^{2+2\sigma}\)	4	\(S^{1+2\sigma}\)	3

4.2 Slice Null and Slice Positive Spectra

Definition (Blumberg/HHR, Definition 5.2.2). Let $X \in \mathrm{Sp}^G$.

$X$ is slice $n$-null (also written: slice $\leq n-1$, or slice $< n$) if $\mathrm{Map}(\widetilde{S}, X)$ is contractible as a $G$-space for every slice cell $\widetilde{S}$ of dimension greater than $n$.
$X$ is slice $n$-positive (also written: slice $> n$, or slice $\geq n+1$) if $\mathrm{Map}(\widetilde{S}, X)$ is contractible as a $G$-space for every slice cell $\widetilde{S}$ of dimension at most $n$.

Here $\mathrm{Map}(\widetilde{S}, X)$ is the $G$-equivariant mapping space (the internal hom in $\mathrm{Sp}^G$, evaluated as a $G$-space).

Contractibility is as G-spaces The condition is that $\mathrm{Map}(\widetilde{S}, X)$ is contractible as a $G$-space, meaning $\mathrm{Map}(\widetilde{S}, X)^H \simeq *$ for all closed $H \leq G$. This is stronger than just asking the underlying (non-equivariant) mapping space to be contractible. In particular, it encodes the mapping spaces out of all representations simultaneously.

Note the parallel with the non-equivariant case: a spectrum $E$ is $n$-truncated (Postnikov $\leq n$) iff $\mathrm{Map}(S^k, E) \simeq *$ for all $k > n$, and $n$-connective iff $\mathrm{Map}(S^k, E) \simeq *$ for all $k \leq n$. The slice conditions replace $\{S^k\}$ with $\{\widetilde{S}^k\}$ (slice cells).

4.3 Relationship to Ordinary Connectivity

Proposition (HHR, Proposition 4.11; Blumberg Exercise 5.2.3). Let $X$ be a $G$-spectrum. Then:

$X$ is slice $0$-positive if and only if $X$ is $(-1)$-connected (i.e., $\pi_k^H(X) = 0$ for $k \leq -1$ and all $H \leq G$).
$X$ is slice $(-1)$-positive if and only if $X$ is $(-2)$-connected.
$X$ is slice $0$-null if and only if $X$ is $0$-coconnected (i.e., $\pi_k^H(X) = 0$ for $k \geq 1$ and all $H \leq G$).
$X$ is slice $(-1)$-null if and only if $X$ is $(-1)$-coconnected.

Proof sketch for (1). $X$ is slice $0$-positive iff $\mathrm{Map}(\widetilde{S}, X) \simeq *$ for all slice cells of dimension $\leq 0$. The slice cells of dimension $\leq 0$ are: $\widetilde{S}^0 = G_+ \wedge_H S^0$ (the $0$-dimensional even cells, $m=0$) and $\widetilde{S}^{-1} = \Sigma^{-1}(G_+ \wedge_H S^0)$ (for $m=0$ and various $H$). Mapping out of $G_+ \wedge_H S^0$ gives $\mathrm{Map}_G(G_+ \wedge_H S^0, X) \simeq \mathrm{Map}_H(S^0, i_H^* X) \simeq X^H$, and mapping out of $\Sigma^{-1}(G_+ \wedge_H S^0)$ gives $\Omega X^H$. These are contractible iff $X^H \simeq *$ and $\pi_0(X^H) = 0$, i.e., iff $\pi_k^H(X) = 0$ for $k \leq -1$. $\square$

This proposition shows that, for small dimensions, the slice conditions precisely recover ordinary connectivity conditions. The slice filtration is a genuine generalization of the connectivity filtration.

The general pattern The connectivity statements hold in the range where slice cells are “the same” as ordinary suspension cells — i.e., for $H = e$, where $G_+ \wedge_e S^m = \Sigma^\infty_G G_+ \wedge S^m$ are just free $G$-spectra. At higher dimensions, the slice cells for nontrivial $H$ introduce additional conditions beyond ordinary connectivity.

Exercise 5 This exercise verifies the connectivity–slice dimension correspondence for Eilenberg-Mac Lane spectra, connecting the abstract slice conditions to familiar objects.

Let $G = C_2$ and $H\underline{\mathbb{Z}}$ the Eilenberg-Mac Lane $C_2$-spectrum for the constant Mackey functor $\underline{\mathbb{Z}}$ (sending both $C_2/e$ and $C_2/C_2$ to $\mathbb{Z}$).

Show $H\underline{\mathbb{Z}}$ is slice 0-null: verify that $\mathrm{Map}_{C_2}(\tilde{S}, H\underline{\mathbb{Z}})$ is contractible for every slice cell $\tilde{S}$ of dimension $> 0$.
Is $H\underline{\mathbb{Z}}$ slice 0-positive? Explain.

[!TIP]- Solution to Exercise 5 (a) Slice cells of dimension $> 0$ include $C_{2+} \wedge S^n$ ($n > 0$) and $S^{m\rho_{C_2}}$ ($m \geq 1$, dim $= 2m \geq 2$). For the free cells: $\mathrm{Map}_{C_2}(C_{2+} \wedge S^n, H\underline{\mathbb{Z}}) \simeq \mathrm{Map}(S^n, (H\underline{\mathbb{Z}})^e) \simeq \mathrm{Map}(S^n, H\mathbb{Z})$, which is contractible for $n > 0$ since $\pi_n(H\mathbb{Z}) = 0$. For representation cells: $\mathrm{Map}_{C_2}(S^{m\rho}, H\underline{\mathbb{Z}}) \simeq \Omega^{m\rho} H\underline{\mathbb{Z}}$, which is contractible since $H\underline{\mathbb{Z}}$ is 0-truncated (all homotopy groups vanish in positive degrees). So $H\underline{\mathbb{Z}}$ is slice 0-null. ✓

No. $H\underline{\mathbb{Z}}$ is NOT slice 0-positive. Slice 0-positive requires $\mathrm{Map}_{C_2}(\tilde{S}, H\underline{\mathbb{Z}}) \simeq *$ for all slice cells of dimension $\leq 0$. But $\mathrm{Map}_{C_2}(S^0, H\underline{\mathbb{Z}}) \simeq H\underline{\mathbb{Z}}(S^0)$, which is not contractible (it has $\pi_0 = \mathbb{Z}$). So $H\underline{\mathbb{Z}}$ is slice 0-null but NOT slice 0-positive — it sits in the “heart” at slice level 0, which is consistent with it being the 0-slice of the sphere spectrum $P_0^0(S^0) \simeq H\underline{\mathbb{Z}}$.

4.4 Characterization via Filtrations

Proposition (HHR, Proposition 4.15; Blumberg Proposition 5.2.4). A $G$-spectrum $X$ is slice $n$-positive if and only if, up to weak $G$-equivalence, $X$ admits a filtration

\[X^0 \subseteq X^1 \subseteq X^2 \subseteq \cdots \subseteq X\]

such that: - $X \simeq \operatorname{hocolim}_i X^i$, - each cofiber $X^i / X^{i-1}$ is a wedge of slice cells of dimension $> n$.

This is the analogue of the CW approximation theorem: slice $n$-positive spectra are exactly those built from slice cells in dimensions $> n$.

Parallels with Postnikov and CW Compare: in $\mathrm{Sp}$, an $n$-connective spectrum is one built from cells $\Sigma^k$ for $k > n$. The slice $n$-positivity replaces $\{\Sigma^k : k > n\}$ with $\{\widetilde{S}^k : k > n\}$. Since slice cells include the ordinary suspension cells (via $H = e$), slice $n$-positivity is in general a stronger condition than $n$-connectivity.

4.5 The Slice Sections and the Slice Tower

Definition (Slice Section $P^n X$). For $X \in \mathrm{Sp}^G$, the $n$-th slice section $P^n X$ is the localization of $X$ with respect to the class of slice $n$-null spectra. Concretely:

There is a natural map $X \to P^n X$,
$P^n X$ is slice $n$-null,
The homotopy fiber $P_n^{n+1} X := \mathrm{fib}(X \to P^n X)$ is slice $n$-positive.

The existence of $P^n X$ follows from Bousfield localization theory applied in $\mathrm{Sp}^G$; the class of slice $n$-null spectra is a localizing subcategory (closed under arbitrary homotopy colimits and retracts).

Subtlety in the construction The naive attempt — kill all mapping spaces from slice cells of dimension $> n$ — runs into the same issue as Section 1.2: one must iterate. An additional subtlety in the equivariant case is the existence of nontrivial maps $S^V \to S^W$ for $V \subset W$ representations. If $V$ contains a trivial summand, all such maps are null-homotopic (one can deform via the trivial direction), but if $V$ has no trivial summands, these maps are genuinely nontrivial and must be handled separately in the small object argument. This is the precise obstruction identified in Section 3.2, and it forces the use of representation spheres rather than integer spheres in the construction.

Definition (Slice Tower). The slice tower of $X$ is the sequence of maps

\[\cdots \longrightarrow P^{n+1} X \longrightarrow P^n X \longrightarrow P^{n-1} X \longrightarrow \cdots\]

obtained from the natural maps $X \to P^n X$ and the functoriality of localization. The maps $P^{n+1} X \to P^n X$ arise because any slice $n$-null spectrum is also slice $(n+1)$-null (trivially: if all mapping spaces from cells of dimension $> n$ are contractible, then a fortiori all mapping spaces from cells of dimension $> n+1$ are contractible).

Theorem. $X \simeq \operatorname{holim}_n P^n X$.

This is the convergence theorem for the slice tower, analogous to $X \simeq \operatorname{holim}_n P_n X$ for the Postnikov tower. The proof uses the fact that $X$ is built from slice cells (by Proposition 4.4) and that the localization maps are compatible with the filtration.

4.6 The n-Slice and an Explicit Example

Definition (n-Slice). The $n$-slice of $X$ is

\[P_n^n X := \mathrm{fib}(P^n X \to P^{n-1} X).\]

This is the “associated graded piece” of the slice tower at level $n$: it captures the “purely dimension-$n$” part of $X$.

Proposition (Blumberg Exercise 5.2.5; HHR Proposition 4.20). The $(-1)$-slice of $X$ is

\[P_{-1}^{-1} X \simeq \Sigma^{-1} H\underline{\pi}_{-1}(X),\]

where $\underline{\pi}_{-1}(X)$ is the $(-1)$-st homotopy group of $X$ viewed as a Mackey functor (with $\underline{\pi}_{-1}(X)(G/H) = \pi_{-1}^H(X)$, with restriction and transfer maps), and $H\underline{M}$ is the Eilenberg-Mac Lane $G$-spectrum for a Mackey functor $\underline{M}$.

Proof sketch. From the long exact sequence on homotopy Mackey functors associated to the fibration $P_{-1}^{-1} X \to P^{-1} X \to P^{-2} X$: - $P^{-1} X$ is slice $(-1)$-null, so $\pi_k^H(P^{-1} X) = 0$ for $k \geq 0$ (by Proposition 4.3, part 3). - $P^{-2} X$ is slice $(-2)$-null, so $\pi_k^H(P^{-2} X) = 0$ for $k \geq -1$. - From the long exact sequence, $\pi_k^H(P_{-1}^{-1} X) \cong \begin{cases} \pi_{-1}^H(X) & k = -1 \\ 0 & k \neq -1 \end{cases}$.

A $G$-spectrum with exactly one nonzero homotopy Mackey functor (in degree $-1$) is, by the equivariant Whitehead theorem, equivalent to $\Sigma^{-1} H\underline{\pi}_{-1}(X)$, where $H\underline{M}$ denotes the Eilenberg-Mac Lane spectrum for the Mackey functor $\underline{M}$. $\square$

Eilenberg-Mac Lane G-spectra For a Mackey functor $\underline{M}$, the Eilenberg-Mac Lane $G$-spectrum $H\underline{M}$ is the unique (up to weak equivalence) $G$-spectrum with $\underline{\pi}_0(H\underline{M}) \cong \underline{M}$ and $\underline{\pi}_k(H\underline{M}) = 0$ for $k \neq 0$. It represents Bredon cohomology with Mackey functor coefficients: $[X, H\underline{M}]_G \cong H_G^0(X; \underline{M})$. The stable Eilenberg-Mac Lane spectra extend the unstable $K(\underline{M}, n)$ to the full stable category and encode both restriction and transfer maps.

The 0-slice For a connective $G$-spectrum $X$, the $0$-slice is $P_0^0 X \simeq H\underline{\pi}_0(X)$ — the Eilenberg-Mac Lane spectrum on the $0$-th homotopy Mackey functor. This is the exact stable analogue of the fact that the $0$-th Postnikov section $P_0 X$ is a $K(\pi_0 X, 0)$.

Exercise I (Proof) This exercise proves that the $(-1)$-slice is a desuspended Eilenberg-Mac Lane spectrum, anchoring the slice tower from below and making the analogy with Postnikov towers precise.

Let $X \in \mathrm{Sp}^G$ be any $G$-spectrum with $\underline{\pi}_{-1}(X) \neq 0$.

Using the long exact sequence in homotopy Mackey functors for the fiber sequence $P_{-1}^{-1} X \to P^{-1} X \to P^{-2} X$, show: \[\underline{\pi}_k(P_{-1}^{-1} X) \cong \begin{cases} \underline{\pi}_{-1}(X) & k = -1 \\ 0 & k \neq -1. \end{cases}\]
Show that any $G$-spectrum with exactly one nonzero homotopy Mackey functor $\underline{M}$ concentrated in degree $n$ is equivalent to $\Sigma^n H\underline{M}$. (Use the equivariant Whitehead theorem and the fact that $H\underline{M}$ represents Bredon cohomology.)
Combine (a) and (b) to conclude $P_{-1}^{-1} X \simeq \Sigma^{-1} H\underline{\pi}_{-1}(X)$.

[!TIP]- Solution to Exercise I (a) Recall from §4.3: $P^{-1} X$ is slice $(-1)$-null, so $\underline{\pi}_k(P^{-1} X) = 0$ for $k \geq -1$ (by Proposition 4.3 part 4 — slice $(-1)$-null iff $(-1)$-coconnected, i.e., $\pi_k = 0$ for $k \geq -1$). And $P^{-2} X$ is slice $(-2)$-null, so $\underline{\pi}_k(P^{-2} X) = 0$ for $k \geq -2$. From the long exact sequence of the fibration $P_{-1}^{-1} X \to P^{-1} X \to P^{-2} X$: \[\cdots \to \underline{\pi}_k(P_{-1}^{-1} X) \to \underline{\pi}_k(P^{-1} X) \to \underline{\pi}_k(P^{-2} X) \to \underline{\pi}_{k-1}(P_{-1}^{-1} X) \to \cdots\] For $k \geq 0$: both $\underline{\pi}_k(P^{-1} X) = 0$ and $\underline{\pi}_k(P^{-2} X) = 0$, so $\underline{\pi}_k(P_{-1}^{-1} X) = 0$. For $k = -1$: $\underline{\pi}_{-1}(P^{-1} X) = 0$ and $\underline{\pi}_{-1}(P^{-2} X) = \underline{\pi}_{-1}(X)$ (the map $X \to P^{-2} X$ is an isomorphism on $\pi_{-1}$ since $P^{-2}$ kills nothing below $-2$). The exact sequence gives $\underline{\pi}_{-1}(P_{-1}^{-1} X) \cong \underline{\pi}_{-1}(X)$. For $k \leq -2$: $\underline{\pi}_k(P^{-2} X) = 0$ forces $\underline{\pi}_k(P_{-1}^{-1} X) = 0$.

Let $Y$ have $\underline{\pi}_n(Y) = \underline{M}$ and $\underline{\pi}_k(Y) = 0$ for $k \neq n$. The map $Y \to \Sigma^n H\underline{M}$ classifying the fundamental class (the identity on $\underline{\pi}_n$) is a map of $G$-spectra inducing isomorphisms on all $\underline{\pi}_k^H$ for all $H$ and all $k$. By the equivariant Whitehead theorem for spectra, this is a weak equivalence. (The spectrum-level Whitehead theorem holds for $G$-spectra: a map in $\mathrm{Sp}^G$ is a weak equivalence iff it induces isomorphisms on $\pi_k^H$ for all $k \in \mathbb{Z}$ and all $H \leq G$.)

Part (a) gives $P_{-1}^{-1} X$ has a single nonzero homotopy Mackey functor $\underline{\pi}_{-1}(X)$ in degree $-1$. By (b), $P_{-1}^{-1} X \simeq \Sigma^{-1} H\underline{\pi}_{-1}(X)$. $\square$

Exercise 6 (Computational) This exercise computes the slices of the sphere $G$-spectrum in the lowest degrees, making the slice tower concrete.

Let $G = C_2$ and $S^0$ the sphere $C_2$-spectrum (with trivial $C_2$-action).

Compute $P^0 S^0$ (the slice 0-section). What does the slice tower $S^0 \to P^0 S^0$ look like?
Identify the 0-slice $P_0^0 S^0$. By the result of §4.6 adapted to degree 0, this should be an Eilenberg-Mac Lane spectrum. Which one?
What is the $(-1)$-slice $P_{-1}^{-1} S^0$?

[!TIP]- Solution to Exercise 6 (a) $P^0 S^0$ is the 0-slice section of $S^0$: the Bousfield localization of $S^0$ with respect to slice 0-null spectra. Since $S^0$ is $(-1)$-connected (it has $\pi_n = 0$ for $n < 0$) and the slice 0-null condition corresponds to 0-coconnectedness, the map $S^0 \to P^0 S^0$ is the unit map $S^0 \to \tau_{\leq 0} S^0$ (Postnikov truncation to degree 0). Since $\pi_0^{C_2}(S^0) = \underline{\mathbb{Z}}$ (the constant Mackey functor valued in $\mathbb{Z}$, since the sphere is the unit), $P^0 S^0 \simeq H\underline{\mathbb{Z}}$.

The 0-slice $P_0^0 S^0$ = homotopy fiber of $P^0 S^0 \to P^{-1} S^0$. Since $S^0$ is $(-1)$-connected, $P^{-1} S^0 \simeq *$ (there’s nothing in degree $< 0$). So $P_0^0 S^0 \simeq P^0 S^0 \simeq H\underline{\mathbb{Z}}$.

By §4.6, $P_{-1}^{-1} S^0 = \Sigma^{-1} H\underline{\pi}_{-1}(S^0)$. Since $\pi_{-1}(S^0) = 0$ as a Mackey functor (the sphere spectrum is $(-1)$-connected), $P_{-1}^{-1} S^0 \simeq *$.

4.7 Behavior Under Change Functors

Proposition (HHR). The slice cells, and hence the slice filtration, are preserved by the following change-of-group functors: - Restriction: $i_K^*: \mathrm{Sp}^G \to \mathrm{Sp}^K$ (restrict to a subgroup $K \leq G$), - Induction: $G_+ \wedge_K -: \mathrm{Sp}^K \to \mathrm{Sp}^G$ (induce from $K$ to $G$), - Norm: $N_K^G: \mathrm{Sp}^K \to \mathrm{Sp}^G$ (the Hill-Hopkins-Ravenel norm).

Proof sketch for restriction. A slice cell $\widetilde{S} = G_+ \wedge_H S^{m\rho_H}$ restricts under $i_K^*$ to $i_K^*(G_+ \wedge_H S^{m\rho_H}) \cong (G/H)_+ \wedge_\emptyset S^{m\rho_H|_K}$. By Mackey’s double coset formula, $(i_K^* G_+ \wedge_H -) \cong \bigvee_{[g] \in K\backslash G/H} K_+ \wedge_{K \cap gHg^{-1}} (-)^{c_g}$. Each term $K_+ \wedge_{K \cap gHg^{-1}} S^{m \cdot g \cdot \rho_H|_{K \cap gHg^{-1}}}$ is a $K$-slice cell (since $g \cdot \rho_H|_{K \cap gHg^{-1}}$ contains the regular representation of $K \cap gHg^{-1}$ as a summand). $\square$

The compatibility of the slice filtration with all three change functors is crucial for HHR’s proof: it allows slice computations to be moved between groups and norms to be taken without losing the slice structure.

Exercise H (Computational) This exercise verifies that restriction commutes with the slice tower and uses this to compute an explicit example, making the change-of-group compatibility concrete.

Let $G = C_2$ and $H = e$ (trivial subgroup), and let $i_e^*: \mathrm{Sp}^{C_2} \to \mathrm{Sp}$ denote restriction to the trivial subgroup (forgetting equivariance).

Show directly from the definitions that $i_e^* P^n X \simeq P^n (i_e^* X)$ for any $G$-spectrum $X$. That is, restriction to the trivial subgroup commutes with the slice section. (Use that $i_e^*$ of a slice $C_2$-cell of dimension $m$ is a wedge of non-equivariant cells of total dimension $m$.)
Compute $i_e^*(H\underline{\mathbb{Z}})$ — the underlying non-equivariant spectrum of the Eilenberg-Mac Lane $C_2$-spectrum for the constant Mackey functor $\underline{\mathbb{Z}}$. Use the fact that $(H\underline{\mathbb{Z}})^e = H\mathbb{Z}$ to identify $i_e^*(H\underline{\mathbb{Z}}) \simeq H\mathbb{Z}$.
Use (a) and (b) to compute $i_e^*(P^n H\underline{\mathbb{Z}})$ for all $n$. For which $n$ is $P^n H\underline{\mathbb{Z}}$ non-trivially equivariant?

[!TIP]- Solution to Exercise H (a) The slice section $P^n X$ is characterized by: (i) $P^n X$ is slice $n$-null, and (ii) the fiber of $X \to P^n X$ is slice $n$-positive. Applying $i_e^*$: since $i_e^*$ of a slice cell $G_+ \wedge_H S^{m\rho_H}$ of dimension $m|H|$ is $(G/H)_+ \wedge S^{m\rho_H|_e}$, which is a wedge of copies of $S^{m|H|}$ (all equivariant structure forgotten). A spectrum is slice $n$-null (resp. $n$-positive) iff mapping spaces from all slice cells of dim $> n$ (resp. $\leq n$) are contractible; after applying $i_e^*$, this reduces to the ordinary connectivity conditions for $i_e^* X$ in $\mathrm{Sp}$. One checks the universal property: $i_e^* P^n X$ satisfies the characterization of $P^n(i_e^* X)$, hence $i_e^* P^n X \simeq P^n(i_e^* X)$.

The underlying spectrum of $H\underline{\mathbb{Z}}$ is $(H\underline{\mathbb{Z}})^e = H\mathbb{Z}$ (the ordinary Eilenberg-Mac Lane spectrum). This is because the fixed-point functor $(-)^e = i_e^*$ on $\mathrm{Sp}^{C_2}$ forgets the $C_2$-action, and the fixed-point space of an EM spectrum for the constant Mackey functor is the ordinary EM spectrum. So $i_e^*(H\underline{\mathbb{Z}}) \simeq H\mathbb{Z}$.

By (a): $i_e^*(P^n H\underline{\mathbb{Z}}) \simeq P^n(H\mathbb{Z})$.

For $n \geq 0$: $H\mathbb{Z}$ is 0-truncated, so $P^0(H\mathbb{Z}) \simeq H\mathbb{Z}$ and $P^n(H\mathbb{Z}) \simeq H\mathbb{Z}$ for all $n \geq 0$.

For $n < 0$: $P^n(H\mathbb{Z}) \simeq *$ (since $H\mathbb{Z}$ is 0-connected, it is already in the positive part; the $n$-section for $n < 0$ kills it entirely).

So the underlying spectrum $i_e^* P^n H\underline{\mathbb{Z}}$ is $H\mathbb{Z}$ for $n \geq 0$ and $*$ for $n < 0$. The equivariant content of $P^n H\underline{\mathbb{Z}}$ is non-trivial for all $n \geq 0$ (it retains the $C_2$-action on $H\underline{\mathbb{Z}}$ via the transfer maps in $\underline{\mathbb{Z}}$), but the underlying non-equivariant spectrum is always just $H\mathbb{Z}$.

[!TIP]- Intuition for norm compatibility The norm $N_K^G$ is the equivariant analogue of the tensor product (not the direct sum). For the slice filtration to be compatible with norms, one needs the dimensions to multiply correctly: a $K$-spectrum of slice dimension $d$ has norm of slice dimension $[G:K] \cdot d$ (since the norm raises the “representation dimension” by the index). This multiplicativity is exactly what the regular representation provides: $\rho_{G|_K}^{\otimes [G:K]}$ contributes to $\rho_G$, matching dimensions.

5. The Slice Spectral Sequence 📐

The slice tower, like any filtered object in a stable $\infty$-category, gives rise to a spectral sequence. This is the slice spectral sequence, the equivariant analogue of the Atiyah-Hirzebruch spectral sequence.

5.1 Construction from the Slice Tower

Let $X \in \mathrm{Sp}^G$ and fix a subgroup $H \leq G$. The slice tower

\[\cdots \to P^{n+1} X \to P^n X \to P^{n-1} X \to \cdots\]

is a filtration of $X$ in $\mathrm{Sp}^G$. Applying $\pi_*^H(-)$ and using the long exact sequences associated to the fiber sequences

\[P_n^n X \longrightarrow P^n X \longrightarrow P^{n-1} X,\]

we obtain an exact couple, hence a spectral sequence.

Definition (Slice Spectral Sequence). The slice spectral sequence for $X$ is the spectral sequence with:

\[E_1^{n,*} = \pi_*^G(P_n^n X) \Implies \pi_*^G(X),\]

where $P_n^n X$ is the $n$-slice of $X$. The differentials $d_r: E_r^{n,*} \to E_r^{n-r, *+r-1}$ are induced by the attaching maps in the slice tower.

More precisely, since the slices $P_n^n X$ are Eilenberg-Mac Lane-type spectra (built from $H\underline{M}$’s), the $E_2$-page can often be computed in terms of Bredon cohomology of the slices. One writes

\[E_2^{p,q} = \pi_p^G(P_q^q X) \Implies \pi_{p+q}^G(X).\]

RO(G)-grading For genuine $G$-spectra, the homotopy groups are $RO(G)$-graded: $\pi_\alpha^H(X) = [S^\alpha, X]_G$ for $\alpha \in RO(G)$ (the real representation ring of $G$). The slice spectral sequence is naturally $RO(G)$-graded, and its full strength requires tracking contributions from all representation spheres. The integer-graded version (collapsing $RO(G)$ to $\mathbb{Z}$) is a shadow of the full spectral sequence.

5.2 Dugger’s C2 Construction and KR-Theory

The first instance of the slice spectral sequence was constructed by Dugger for $G = C_2$ and Atiyah’s KR-theory.

KR-theory is the $C_2$-equivariant cohomology theory $KR^{*,*}(X)$ defined by Atiyah, with $C_2$ acting by complex conjugation. Non-equivariantly, $KR^*(X) \cong K^*(X)$; at the $C_2$-fixed level, $KR^*(X^{C_2}) \cong KO^*(X^{C_2})$ (real K-theory).

Dugger’s motivation was the deep analogy between $C_2$-equivariant homotopy theory and motivic homotopy theory over $\mathbb{R}$: both have a “weight” grading (corresponding to the sign representation $\sigma$ for $C_2$, or the Tate twist for motivic) in addition to the topological grading.

Theorem (Dugger). There is a conditionally convergent spectral sequence

\[E_2^{p,q} = H^{p,\, r-q/2}(X;\, \underline{\mathbb{Z}}) \Implies KR^{p+q,\, r}(X),\]

where $X$ is a $C_2$-space, $\underline{\mathbb{Z}}$ is the constant Mackey functor valued in $\mathbb{Z}$, and $KR^{p+q, r}$ is the $(p+q)$-th $KR$-group with motivic weight $r$. The left-hand side uses bigraded cohomology reflecting the $C_2$-equivariant structure.

The motivic analogy Dugger’s construction was explicitly motivated by the analogous spectral sequence in motivic homotopy theory: \[E_2^{p,q} = H^{p,q}(X; \mathbb{Z}) \Implies K^{p+q}(X)_\mathbb{Z}^{\wedge}\] (the motivic AHSS for algebraic K-theory). The $C_2$-equivariant slice spectral sequence is the “topological realization” of the motivic one, via the Betti realization functor sending motivic spectra over $\mathbb{R}$ to $C_2$-equivariant spectra.

Exercise 7 This exercise checks that the Dugger slice spectral sequence reduces to the classical AHSS in the non-equivariant limit.

In Dugger’s slice spectral sequence $E_2^{p,q} = H^{p, r-q/2}(X; \underline{\mathbb{Z}}) \Rightarrow KR^{p+q,r}(X)$, set $G = C_2$ acting trivially on a space $X$ (so $X$ is a space with trivial $C_2$-action).

What does $KR^{*,*}(X)$ reduce to when $C_2$ acts trivially on $X$?
What does the bigraded cohomology $H^{p, r-q/2}(X; \underline{\mathbb{Z}})$ reduce to in this case?
Verify the spectral sequence reduces to the classical AHSS $E_2^{p,q} = H^p(X; KU^q(*)) \Rightarrow KU^{p+q}(X)$.

[!TIP]- Hint for Exercise 7 When $C_2$ acts trivially, $KR^{*,*}(X) \cong KO^*(X) \oplus KU^*(X)$ splits (roughly), with the bigraded structure collapsing. The bigraded cohomology $H^{p,q}(X; \underline{\mathbb{Z}})$ with trivial $C_2$-action reduces to ordinary cohomology $H^p(X; \mathbb{Z})$ (the weight grading collapses). Comparing dimensions gives the AHSS.

5.3 The General HHR Slice Spectral Sequence

The full generality of the slice spectral sequence, for $G = C_{2^n}$ and arbitrary $G$-spectra, was developed by Hill-Hopkins-Ravenel in their proof of the Kervaire invariant one theorem.

Theorem (HHR). For $G = C_{2^n}$ and $X \in \mathrm{Sp}^G$, the slice tower gives a conditionally convergent spectral sequence

\[E_2^{p,q} = \pi_p^G(P_q^q X) \Implies \pi_{p+q}^G(X),\]

which is the slice spectral sequence for $X$.

The key properties established by HHR: 1. Convergence: the tower converges conditionally (and strongly under finiteness hypotheses). 2. $E_2$-page identification: the slices $P_q^q X$ are Eilenberg-Mac Lane spectra (wedges of $H\underline{M}$’s) in favorable cases, making the $E_2$-page computable in terms of Mackey functor cohomology. 3. Multiplicative structure: when $X$ is a commutative $G$-ring spectrum, the slice spectral sequence is multiplicative (a spectral sequence of rings).

The key feature distinguishing the slice spectral sequence from the ordinary AHSS is its sensitivity to the $RO(G)$-graded homotopy groups. Non-equivariantly (with trivial $G$-action), the slice tower and integer Postnikov tower coincide, and the spectral sequence reduces to the ordinary AHSS:

\[E_2^{p,q} = H^p(X; \pi_q E) \Implies E^{p+q}(X).\]

6. Comparison and the HHR Application 🔑

6.1 The Three-Way Comparison Table

The following table compares the three tiers of Postnikov/slice theory.

Feature	Non-equivariant $\mathrm{Sp}$	Equivariant unstable $G\mathbf{Top}$	Equivariant stable $\mathrm{Sp}^G$
Cells	$S^n$ (integer spheres)	$G/H_+ \times S^n$ (orbit cells)	$G_+ \wedge_H S^{m\rho_H}$ (slice cells)
Dimension indexing	Integer $n$	Integer $n$	Representation dim $m\|H\|$
Filtration sections	$P_n X$ (Postnikov)	$P_n^G X$ (equivariant Postnikov)	$P^n X$ (slice section)
Fiber / slice	$K(\pi_n X, n)$	$K(\underline{\pi}_n X, n)$	$P_n^n X \simeq \Sigma^n H\underline{M}$
Coefficients	$\pi_n(X) \in \mathbf{Ab}$	$\underline{\pi}_n X$ (coeff. system)	$\underline{\pi}_n^{Mky}(X)$ (Mackey functor)
k-invariants	$H^{n+2}(-; \pi_n X)$	$H_G^{n+2}(-; \underline{\pi}_n X)$ (Bredon)	Mackey functor cohomology
Spectral sequence	—	Bredon AHSS	Slice SS
$E_2$-page	$H^p(X; \pi_q E)$	$H_G^p(X; \underline{\pi}_q E)$	$\pi_p^G(P_q^q X)$
Construction tool	Small object argument	Elmendorf + objectwise	Bousfield localization
Key theorem	Whitehead	Elmendorf’s theorem	HHR (Prop. 4.11, 4.15, 4.20)

Summary of the upgrade chain As one passes from non-equivariant to equivariant unstable to equivariant stable, each layer adds structure: - Non-equivariant → equivariant unstable: coefficients upgrade from abelian groups to coefficient systems (add restriction maps); k-invariants upgrade from ordinary to Bredon cohomology. - Equivariant unstable → equivariant stable: coefficients upgrade from coefficient systems to Mackey functors (add transfer maps); cells upgrade from orbit cells to slice cells (add representation spheres); the indexing upgrades from integers to representation dimensions.

6.2 The Slices of the Real Cobordism Spectrum

The central computation in HHR’s proof of the Kervaire invariant one theorem is the identification of the slices of the Real cobordism spectrum $MU_\mathbb{R}$.

Construction. The Real cobordism spectrum $MU_\mathbb{R}$ is a $C_2$-equivariant spectrum; it is the Thom spectrum of the $C_2$-equivariant complex cobordism, with $C_2$ acting by complex conjugation. The full $C_{2^n}$-equivariant spectrum used by HHR is the norm:

\[MU^{((C_{2^n}))} := N_{C_2}^{C_{2^n}} MU_\mathbb{R},\]

the Hill-Hopkins-Ravenel norm of $MU_\mathbb{R}$ from $C_2$ to $C_{2^n}$.

Theorem (HHR, Theorem 6.1). The slices of $MU^{((C_{2^n}))}$ are:

\[P_{2k}^{2k} MU^{((C_{2^n}))} \simeq \bigvee H\underline{\mathbb{Z}} \wedge S^{k\rho_G},\]

where the wedge is over an explicit finite set of representation spheres, and $H\underline{\mathbb{Z}}$ is the Eilenberg-Mac Lane $G$-spectrum for the constant Mackey functor $\underline{\mathbb{Z}}$. Odd slices vanish.

This explicit identification is the “hard part” of the HHR proof. Once the slices are known, the slice spectral sequence converges to $\pi_*^G(MU^{((C_{2^n}))})$, and the differentials can be analyzed to prove the vanishing of Kervaire invariant one elements $\theta_j$ for $j \geq 7$.

Open: the general slice computation The full slice structure of $MU^{((G))}$ for groups $G$ beyond $C_{2^n}$ is not known in general. The slice machinery extends to all finite groups $G$ (and more generally to compact Lie groups), but the explicit slice computations depend on detailed knowledge of the representation theory of $G$. This remains an active area of research.

Historical note: Dugger’s priority The first piece of the slice spectral sequence was worked out by Dugger for $C_2$. His motivation was purely the $C_2$-motivic analogy: he sought a spectral sequence converging to $KR$-theory whose $E_2$-page was motivic cohomology. HHR recognized that this construction generalized to a filtration of all $G$-spectra indexed by representation dimension, and that the resulting spectral sequence for $MU^{((C_{2^n}))}$ would be powerful enough to settle the Kervaire invariant problem.

7. CW-Postnikov Duality and Eckmann-Hilton 💡

The Postnikov tower and the CW filtration are not merely analogous — they are dual in a precise sense that goes by the name of Eckmann-Hilton duality. Understanding this duality clarifies both the structure of the Postnikov tower and why the equivariant stable analogue (the slice filtration) breaks the classical picture.

7.1 The Classical Duality

Every CW complex $X$ has two canonical filtrations:

CW filtration (bottom-up): $X^{(0)} \hookrightarrow X^{(1)} \hookrightarrow X^{(2)} \hookrightarrow \cdots \xrightarrow{\sim} X$, a homotopy colimit along cofibrations, with cofibers $X^{(n)}/X^{(n-1)} \simeq \bigvee_\alpha S^n$.
Postnikov tower (top-down): $X \to \cdots \to P_2 X \to P_1 X \to P_0 X$, a homotopy limit along fibrations, with fibers $K(\pi_n X, n) \to P_n X \to P_{n-1} X$.

The duality is the observation that one is obtained from the other by reversing all arrows:

CW filtration	Postnikov tower
Built by pushouts (cofibrations)	Built by pullbacks (fibrations)
$X \simeq \hocolim_n X^{(n)}$	$X \simeq \holim_n P_n X$
Cells: $S^n \simeq D^n/S^{n-1}$ (cofibers)	Fibers: $K(\pi_n X, n)$
Attaching maps: $\phi_\alpha: S^{n-1} \to X^{(n-1)}$	k-invariants: $k^{n+2}: P_{n-1} X \to K(\pi_n X, n+1)$
Classification: $[\phi_\alpha] \in \pi_{n-1}(X^{(n-1)})$	Classification: $[k^{n+2}] \in H^{n+2}(P_{n-1} X;\, \pi_n X)$
Cells probe: $[S^n, X] = \pi_n(X)$	Fibers probe: $[X, K(G,n)] = H^n(X; G)$
“Dual cells”: Moore spaces $M(G, n)$ with $H_n = G$	“Dual cells”: Eilenberg-Mac Lane spaces $K(G, n)$ with $\pi_n = G$
Abstract: weight structure (Bondarko)	Abstract: t-structure (Beilinson-Bernstein-Deligne)

The deep reason for the duality is that $S^n$ and $K(\mathbb{Z}, n)$ are Eckmann-Hilton duals: - $S^n = \Sigma^n S^0$ is the cogroup object in $\mathrm{Ho}(\mathbf{Top}_*)$: it carries a comultiplication $S^n \to S^n \vee S^n$ (pinch map), making $[S^n, X] = \pi_n(X)$ into a group. - $K(\mathbb{Z}, n) = \Omega^n K(\mathbb{Z}, 0)$ is the group object: it carries a multiplication $K(\mathbb{Z},n) \times K(\mathbb{Z},n) \to K(\mathbb{Z},n)$, making $[X, K(\mathbb{Z}, n)] = H^n(X; \mathbb{Z})$ into a group.

Eckmann-Hilton Duality Eckmann-Hilton duality is the meta-principle that any homotopy-theoretic construction has a dual obtained by reversing all arrows. Concretely:

Primal	Dual
Cofibration $A \hookrightarrow X$	Fibration $E \twoheadrightarrow B$
Cofiber $X/A$ (pushout)	Fiber $F$ (pullback)
Suspension $\Sigma X$	Loop space $\Omega X$
Sphere $S^n = \Sigma^n S^0$	Eilenberg-Mac Lane $K(G, n) = \Omega^{-n} HG$
Co-H-space (comultiplication)	H-space (multiplication)
Homotopy groups $\pi_n(X) = [S^n, X]$	Cohomology $H^n(X;G) = [X, K(G,n)]$
Moore space $M(G,n)$: $\tilde{H}_n = G$	Eilenberg-Mac Lane $K(G,n)$: $\pi_n = G$
CW filtration / weight structure	Postnikov tower / t-structure

The duality is not an equivalence: the primal and dual constructions live in different parts of the homotopy category. But every theorem about one has a dual theorem about the other.

t-Structures and Weight Structures In a triangulated category $\mathcal{C}$: - A t-structure $(\mathcal{C}^{\leq 0}, \mathcal{C}^{\geq 0})$ axiomatizes the Postnikov filtration: every object $X$ fits into a unique triangle $\tau_{\geq 1} X \to X \to \tau_{\leq 0} X$. The heart $\mathcal{C}^{\leq 0} \cap \mathcal{C}^{\geq 0}$ is an abelian category (e.g., abelian groups in $\mathrm{Sp}$). - A weight structure $(\mathcal{C}^{w \leq 0}, \mathcal{C}^{w \geq 0})$ (Bondarko, 2010) axiomatizes the CW filtration: every object fits into a (non-unique) triangle $X' \to X \to X''$ with $X' \in \mathcal{C}^{w \leq 0}$ and $X'' \in \mathcal{C}^{w \geq 0}$. The heart $\mathcal{C}^{w=0} = \mathcal{C}^{w \leq 0} \cap \mathcal{C}^{w \geq 0}$ is an additive (but generally not abelian) category.

The axioms of a weight structure are obtained from those of a t-structure by reversing the inclusion directions and removing exactness requirements. In $\mathrm{Sp}$, the standard t-structure has heart $\mathbf{Ab}$ (= $H\mathbb{Z}$-modules), and the standard weight structure has heart generated by the sphere spectrum $S^0$.

In spectra, the duality is especially clean. For $E \in \mathrm{Sp}$ and the standard t-structure:

\[\tau_{\geq n+1} E \longrightarrow E \longrightarrow \tau_{\leq n} E \longrightarrow \Sigma \tau_{\geq n+1} E\]

is a cofiber sequence. The cofiber $\tau_{\leq n} E / \tau_{\leq n-1} E \simeq H\pi_n(E)[n]$ is an Eilenberg-Mac Lane spectrum. Dually, the weight filtration has associated graded $\bigoplus_n S^n \otimes \pi_n^{cell}(E)$ built from spheres.

Exercise L (Proof) This exercise makes the Eckmann-Hilton duality between spheres and Eilenberg-Mac Lane spectra precise at the level of t-structures and weight structures, identifying the generating objects of each heart.

Work in $\mathrm{Sp}$ (non-equivariant spectra) with the standard t-structure $(\tau_{\geq 0}, \tau_{\leq 0})$.

Show that the heart $\mathcal{C}^{\heartsuit} = \tau_{\geq 0} \mathrm{Sp} \cap \tau_{\leq 0} \mathrm{Sp}$ is equivalent to $\mathbf{Ab}$ (abelian groups). Identify the functor realizing this equivalence explicitly. (Hint: $\pi_0: \mathcal{C}^{\heartsuit} \to \mathbf{Ab}$.)
Show that $H\mathbb{Z} \in \mathrm{Sp}$ is the unique (up to equivalence) spectrum in the heart with $\pi_0(H\mathbb{Z}) = \mathbb{Z}$, and that every object in the heart is a product of copies of $H\mathbb{Z}[n]$ for varying $n$ (i.e., the heart is generated by $H\mathbb{Z}$).
For the standard weight structure on $\mathrm{Sp}$ (whose heart $\mathcal{C}^{w=0}$ consists of retracts of wedges of $S^0$), show that the sphere spectrum $S^0$ generates the heart, and that its Eckmann-Hilton dual is $H\mathbb{Z}$ in the sense that $[S^n, H\mathbb{Z}] = \pi_n(H\mathbb{Z}) = \delta_{n,0} \mathbb{Z}$ and $[H\mathbb{Z}, S^n] = H^n(H\mathbb{Z}; \mathbb{Z})$.
Why is the duality between t-structure heart and weight structure heart exact in $\mathrm{Sp}$ but not (with integer cells and slices) in $\mathrm{Sp}^G$?

[!TIP]- Solution to Exercise L (a) The functor $\pi_0: \tau_{\geq 0}\mathrm{Sp} \cap \tau_{\leq 0}\mathrm{Sp} \to \mathbf{Ab}$ is an equivalence. Every object $E$ in the heart has $\pi_k(E) = 0$ for $k \neq 0$ and $\pi_0(E) \in \mathbf{Ab}$. The inverse is $A \mapsto HA$ (Eilenberg-Mac Lane spectrum). This is an equivalence of categories: $\pi_0(HA) = A$ and a map $E \to E'$ in the heart is determined by $\pi_0(E) \to \pi_0(E')$ (since both spectra are connective and coconnective, the Postnikov section map $E \to H\pi_0(E)$ is an equivalence).

$H\mathbb{Z}$ is the spectrum with $\pi_k(H\mathbb{Z}) = \delta_{k,0}\mathbb{Z}$; it is the unit of the equivalence in (a) for $A = \mathbb{Z}$. Every abelian group $A$ is a quotient (or extension) of free groups, hence $HA$ is built from copies of $H\mathbb{Z}$ in the derived sense. More precisely, the heart $\mathbf{Ab}$ is generated by $\mathbb{Z}$ (every abelian group is a quotient of a free one), so the heart of $\mathrm{Sp}$ is generated by $H\mathbb{Z}$.

The weight structure heart $\mathcal{C}^{w=0}$ consists of retracts of $\bigvee_\alpha S^0$; its generating object is $S^0$. The duality: $[S^n, H\mathbb{Z}] = \pi_n(H\mathbb{Z}) = \delta_{n,0}\mathbb{Z}$ (maps from the CW-generator to the t-structure generator concentrate in degree 0). Dually, $[H\mathbb{Z}, S^n]$ computes the cohomology of $H\mathbb{Z}$ with $\mathbb{Z}$ coefficients via $H^n(H\mathbb{Z}; \mathbb{Z}) = [H\mathbb{Z}, H\mathbb{Z}[n]]$, which is $\mathbb{Z}$ for $n=0$ and more complex for $n > 0$ (the cohomology of $K(\mathbb{Z},0) = \mathbb{Z}$ in spaces is trivial, but the stable cohomology is $\mathbb{Z}$ concentrated in degree 0). The key point: $S^0$ and $H\mathbb{Z}$ are “dual” in the sense that $[S^0, H\mathbb{Z}] = \mathbb{Z}$ and $H\mathbb{Z}$ is built from $S^0$ via the unit map $S^0 \to H\mathbb{Z}$.

In $\mathrm{Sp}$, the sphere spectrum is the only “basic” cell — $S^n = \Sigma^n S^0$ for all $n$. So the weight structure heart (generated by $S^0$) and the t-structure heart (generated by $H\mathbb{Z}$) are genuinely dual: one generates by suspensions of a single object, the other by desuspensions of the Eilenberg-Mac Lane spectrum for $\pi_0(S^0) = \mathbb{Z}$.

In $\mathrm{Sp}^G$, there are two families of basic cells: integer-graded cells $G/H_+ \wedge S^n$ and slice cells $G_+ \wedge_H S^{m\rho_H}$. The integer CW cells generate the weight structure of the naive (integer) Postnikov tower; the slice cells generate the weight structure of the slice filtration. These are genuinely different categories of cells (they use different representations), so there is no single “equivariant sphere spectrum” that simultaneously generates both. The duality splits into two separate dualities that are not related by a single Eckmann-Hilton symmetry.

[!EXAMPLE]- Duality for $E = H\mathbb{Z}$ The Eilenberg-Mac Lane spectrum $H\mathbb{Z}$ is both slice 0-null AND in the heart of the weight structure: it has $\pi_n(H\mathbb{Z}) = 0$ for $n \neq 0$ (so the Postnikov tower collapses) and is built from a single cell $S^0$ (so the CW filtration is trivial). This extreme simplicity is why $H\mathbb{Z}$ appears as both the fibers of Postnikov towers and the cells of Moore complexes.

7.2 The Equivariant Unstable Setting: Duality Holds Objectwise

In $G\mathbf{Top}$, the CW-Postnikov duality holds perfectly, mediated by Elmendorf’s theorem.

Equivariant CW cells and their duals. The generating equivariant CW cells are $G/H_+ \wedge S^n$ (i.e., $G/H \times D^n$ with boundary $G/H \times S^{n-1}$). Their Eckmann-Hilton duals are the equivariant Eilenberg-Mac Lane spaces $K(\underline{M}, n)$ for coefficient systems $\underline{M}$.

The duality is precise: the equivariant CW cells detect the same homotopy-theoretic data as the Postnikov fibers:

\[[G/H_+ \wedge S^n,\, X]_G = [S^n, X^H] = \pi_n(X^H) = \pi_n^H(X)\]

and

\[[X,\, K(\underline{M}, n)]_G = H_G^n(X;\, \underline{M}) \quad \text{(Bredon cohomology)}.\]

The equivariant CW cells probe the fixed-point homotopy groups $\pi_n^H(X)$; the equivariant Eilenberg-Mac Lane spaces represent Bredon cohomology with coefficient systems valued in $\pi_n^H(X)$.

Via Elmendorf, both reduce to objectwise classical duality: for each $H$, the $H$-fixed-point level of the G-CW filtration gives the CW filtration of $X^H$, and the $H$-fixed-point level of the equivariant Postnikov tower gives $P_n(X^H)$. The CW-Postnikov duality for $X^H$ holds by the classical result, and these piece together equivariantly.

Why Elmendorf Makes Duality Easy The Elmendorf equivalence $G\mathbf{Top} \simeq \mathrm{Fun}(\mathcal{O}_G^{op}, \mathbf{Spaces})$ reduces all homotopy-theoretic questions about $G$-spaces to objectwise questions about spaces. Since CW-Postnikov duality holds in $\mathbf{Spaces}$, it holds in $\mathrm{Fun}(\mathcal{O}_G^{op}, \mathbf{Spaces})$ (objectwise), and hence in $G\mathbf{Top}$. This is the “cost” of the Elmendorf perspective: it makes equivariant homotopy theory look like ordinary homotopy theory with extra coefficients, at the expense of not immediately seeing the representation-theoretic structure.

7.3 The Equivariant Stable Setting: The Duality Breaks

In the equivariant stable category $\mathrm{Sp}^G$, the clean CW-Postnikov duality of the unstable setting breaks down. This is precisely why the slice filtration is necessary and non-trivial.

The naive integer-graded duality. There is still an integer-graded t-structure on $\mathrm{Sp}^G$ (the naive Postnikov filtration of §3.1), and its dual weight structure uses integer-indexed CW cells $G/H_+ \wedge S^n$. This duality holds, but it is the wrong duality for genuine equivariant stable homotopy theory: it misses all the representation-theoretic structure.

The slice t-structure and its dual. The slice filtration defines a t-structure on $\mathrm{Sp}^G$ (or more precisely, a semi-t-structure with non-standard truncation properties) whose “truncation objects” are the slice sections $P^n X$. The “heart” at each level consists of the slices $P_n^n X$ — Eilenberg-Mac Lane-type spectra built from $H\underline{M} \wedge S^V$ for Mackey functors $\underline{M}$ and representation spheres $S^V$.

The dual of the slice t-structure would be a weight structure whose generating objects are the slice cells $G_+ \wedge_H S^{m\rho_H}$. This exists formally (by general machinery), but the “slice CW complexes” built from representation-sphere cells are much harder to work with than either integer-graded CW complexes or ordinary equivariant CW complexes.

The key asymmetry. In $\mathrm{Sp}$, the non-equivariant Postnikov t-structure and CW weight structure are symmetric because $S^n = \Sigma^n S^0$ — all cells are desuspensions of the same sphere spectrum. In $\mathrm{Sp}^G$:

Integer-graded CW cells: $G/H_+ \wedge S^n$ use spheres with trivial $G$-representation.
Slice cells: $G_+ \wedge_H S^{m\rho_H}$ use the regular representation $\rho_H$.

These are genuinely different cells, and there is no single “equivariant sphere spectrum” that generates both. The slice filtration is sensitive to the regular representation in a way that integer-graded CW complexes are not.

Exercise M (Computational) This exercise exhibits an explicit $C_2$-spectrum that lies in the integer-graded Postnikov filtration but not in the correct slice filtration, making the divergence of the two theories concrete.

Let $G = C_2$ and consider $X = \Sigma^1 H\underline{\mathbb{Z}}$ (the suspension of the Eilenberg-Mac Lane $C_2$-spectrum for the constant Mackey functor).

Show that $X = \Sigma H\underline{\mathbb{Z}}$ is in the integer-graded Postnikov filtration at level 1: $\pi_k^H(X) = 0$ for $k > 1$ and all $H \leq C_2$. (This means $X \simeq \tau_{\leq 1} X$ in the naive integer Postnikov tower.)
Show that $X$ is not slice 1-null. Specifically, find a slice cell $\widetilde{S}$ of dimension $> 1$ such that $\mathrm{Map}_{C_2}(\widetilde{S}, X)$ is not contractible.
Conclude that the integer-graded Postnikov filtration and the slice filtration give different answers for $X$: integer-Postnikov declares $X$ to be in level $\leq 1$, but the slice filtration does not. Explain in one sentence what extra structure the slice filtration detects.

[!TIP]- Solution to Exercise M (a) $\pi_k^H(\Sigma H\underline{\mathbb{Z}}) = \pi_{k-1}^H(H\underline{\mathbb{Z}})$. Since $H\underline{\mathbb{Z}}$ has $\pi_j^H = \delta_{j,0}\mathbb{Z}$ for all $H \leq C_2$ (it is 0-truncated and 0-connected), we get $\pi_k^H(\Sigma H\underline{\mathbb{Z}}) = \delta_{k,1}\mathbb{Z}$. In particular, $\pi_k^H(X) = 0$ for $k > 1$, so $X$ is in the integer Postnikov level 1: $\tau_{\leq 1}^{\mathrm{int}} X \simeq X$. ✓

Consider the even slice cell $\widetilde{S} = S^{1+\sigma} = G_+ \wedge_{C_2} S^{\rho_{C_2}}$ of dimension 2 (see Exercise 4). Compute: \[\mathrm{Map}_{C_2}(S^{1+\sigma}, \Sigma H\underline{\mathbb{Z}}) \simeq \Omega^{1+\sigma}(\Sigma H\underline{\mathbb{Z}}) \simeq \Omega^\sigma H\underline{\mathbb{Z}}.\] Now $\Omega^\sigma H\underline{\mathbb{Z}} \simeq \mathrm{Map}_{C_2}(S^\sigma, H\underline{\mathbb{Z}})$. Since $S^\sigma$ has $(S^\sigma)^{C_2} = S^0$ and $(S^\sigma)^e = S^1$, a $C_2$-map $S^\sigma \to H\underline{\mathbb{Z}}$ involves a map $S^0 \to (H\underline{\mathbb{Z}})^{C_2} = H\mathbb{Z}$ and a compatible map $S^1 \to (H\underline{\mathbb{Z}})^e = H\mathbb{Z}$. The map $S^0 \to H\mathbb{Z}$ picks a class in $\pi_0(H\mathbb{Z}) = \mathbb{Z}$ (non-contractible), so $\mathrm{Map}_{C_2}(S^\sigma, H\underline{\mathbb{Z}})$ has $\pi_0 \cong \mathbb{Z} \neq 0$. Hence $\mathrm{Map}_{C_2}(\widetilde{S}, X)$ is not contractible, and $X$ is not slice 1-null (the slice cell of dimension 2 “sees” it).

The integer-graded Postnikov filtration only tracks $\pi_k^H(X)$ for integer $k$, seeing that $X$ has homotopy concentrated in degree 1. The slice filtration additionally detects representation-theoretic structure: specifically, the $RO(C_2)$-graded homotopy groups $\pi_\alpha^H(X)$ for $\alpha \in RO(C_2)$. The class detected by $S^\sigma \to H\underline{\mathbb{Z}}$ lives in $\pi_{-\sigma}(H\underline{\mathbb{Z}}) = \pi_0(H\underline{\mathbb{Z}}) = \mathbb{Z}$ (a representation-graded homotopy group), which the integer filtration is blind to.

The Failure of Naive Equivariant Eckmann-Hilton Duality In the equivariant stable category, the Eckmann-Hilton duality between CW filtrations and Postnikov towers fails to extend cleanly from the integer-graded setting to the representation-graded setting. Specifically:

The dual of the integer-graded Postnikov tower is the integer-graded CW filtration: this duality works, but it is the wrong theory (it misses representation-theoretic data, as explained in §3).
The slice t-structure IS the “right” Postnikov analogue, but its dual (the slice weight structure built from $G_+ \wedge_H S^{m\rho_H}$) is not the same as the integer-graded CW weight structure.

The lesson: the CW-Postnikov duality is not merely a formal consequence of t/weight structure duality in $\mathrm{Sp}^G$. It depends on the identification of cells and fibers, which in the equivariant stable setting requires a choice: integer or representation-graded? The slice filtration makes the representation-graded choice, and its CW dual is less studied.

Open: Slice Weight Structure The weight structure dual to the slice t-structure — whose heart consists of direct sums of slice cells $G_+ \wedge_H S^{m\rho_H}$ — has not been fully developed in the literature. One expects a “slice cellular approximation theorem” saying every $G$-spectrum is weakly equivalent to one built from slice cells. The precise formulation and its consequences (e.g., for duality in $\mathrm{Sp}^G$) remain open problems.

References

Reference Name	Brief Summary	Link to Reference
M392C Lecture Notes (Blumberg, notes by Debray)	Primary source; §1.3 for equivariant Postnikov towers, §5.2 for the full slice filtration and slice spectral sequence	https://adebray.github.io/lecture_notes/m392c_EHT_notes.pdf
On the Non-Existence of Elements of Kervaire Invariant One (Hill, Hopkins, Ravenel)	Original slice filtration paper; Propositions 4.11, 4.15, 4.20 are the foundational results on slice null/positive and the slice tower	https://arxiv.org/abs/0908.3724
The Equivariant Slice Filtration: A Primer (Hill)	Best expository introduction to slice cells and the slice tower; highly recommended as a companion to HHR	https://arxiv.org/abs/1107.3582
Equivariant Homotopy and Cohomology Theory (May et al.)	Classical treatment of equivariant Postnikov towers, coefficient systems, Bredon cohomology, and equivariant Eilenberg-Mac Lane spaces	https://www.math.uchicago.edu/~may/BOOKS/alaska.pdf
Equivariant Stable Homotopy Theory — Handbook (Greenlees, May)	Survey covering Mackey functors, RO(G)-grading, the stable equivariant category, and change-of-group functors	https://www.math.uchicago.edu/~may/PAPERS/Newthird.pdf
Equivariant Cohomology Theories (Bredon)	Original definition of Bredon cohomology and coefficient systems; foundational for the equivariant Postnikov theory	https://link.springer.com/book/10.1007/BFb0082690
Weight Structures and Motives (Bondarko)	Defines weight structures as the categorical dual of t-structures; gives the abstract framework for CW-Postnikov duality in triangulated categories	arXiv:0704.4003

Feature	Non-equivariant \(\mathbf{Top}\)	Equivariant \(G\mathbf{Top}\)
Homotopy groups	\(\pi_n(X) \in \mathbf{Ab}\)	\(\underline{\pi}_n X: \mathcal{O}_G^{op} \to \mathbf{Ab}\) (coefficient system)
Postnikov section	\(P_n X\), \(\pi_k(P_n X) = \begin{cases}\pi_k X & k\leq n \\ 0 & k > n\end{cases}\)	\(P_n^G X\), same for each \(\pi_k^H\)
Construction	Small object argument on \(X\)	Small object argument objectwise on each \(X^H\)
Eilenberg-Mac Lane	\(K(A, n)\), \(A \in \mathbf{Ab}\)	\(K(\underline{M}, n)\), \(\underline{M}: \mathcal{O}_G^{op} \to \mathbf{Ab}\)
Fiber sequence	\(K(\pi_n X, n) \to P_n X \to P_{n-1} X\)	\(K(\underline{\pi}_n X, n) \to P_n^G X \to P_{n-1}^G X\)
k-invariant	\(k^{n+2} \in H^{n+2}(P_{n-1} X; \pi_n X)\)	\(k^{n+2} \in H_G^{n+2}(P_{n-1}^G X; \underline{\pi}_n X)\) (Bredon)
Key tool	Whitehead theorem	Elmendorf’s theorem + objectwise Whitehead

Feature	Non-equivariant \(\mathrm{Sp}\)	Equivariant unstable \(G\mathbf{Top}\)	Equivariant stable \(\mathrm{Sp}^G\)
Cells	\(S^n\) (integer spheres)	\(G/H_+ \times S^n\) (orbit cells)	\(G_+ \wedge_H S^{m\rho_H}\) (slice cells)
Dimension indexing	Integer \(n\)	Integer \(n\)	Representation dim \(m\|H\|\)
Filtration sections	\(P_n X\) (Postnikov)	\(P_n^G X\) (equivariant Postnikov)	\(P^n X\) (slice section)
Fiber / slice	\(K(\pi_n X, n)\)	\(K(\underline{\pi}_n X, n)\)	\(P_n^n X \simeq \Sigma^n H\underline{M}\)
Coefficients	\(\pi_n(X) \in \mathbf{Ab}\)	\(\underline{\pi}_n X\) (coeff. system)	\(\underline{\pi}_n^{Mky}(X)\) (Mackey functor)
k-invariants	\(H^{n+2}(-; \pi_n X)\)	\(H_G^{n+2}(-; \underline{\pi}_n X)\) (Bredon)	Mackey functor cohomology
Spectral sequence	—	Bredon AHSS	Slice SS
\(E_2\)-page	\(H^p(X; \pi_q E)\)	\(H_G^p(X; \underline{\pi}_q E)\)	\(\pi_p^G(P_q^q X)\)
Construction tool	Small object argument	Elmendorf + objectwise	Bousfield localization
Key theorem	Whitehead	Elmendorf’s theorem	HHR (Prop. 4.11, 4.15, 4.20)

CW filtration	Postnikov tower
Built by pushouts (cofibrations)	Built by pullbacks (fibrations)
\(X \simeq \hocolim_n X^{(n)}\)	\(X \simeq \holim_n P_n X\)
Cells: \(S^n \simeq D^n/S^{n-1}\) (cofibers)	Fibers: \(K(\pi_n X, n)\)
Attaching maps: \(\phi_\alpha: S^{n-1} \to X^{(n-1)}\)	k-invariants: \(k^{n+2}: P_{n-1} X \to K(\pi_n X, n+1)\)
Classification: \([\phi_\alpha] \in \pi_{n-1}(X^{(n-1)})\)	Classification: \([k^{n+2}] \in H^{n+2}(P_{n-1} X;\, \pi_n X)\)
Cells probe: \([S^n, X] = \pi_n(X)\)	Fibers probe: \([X, K(G,n)] = H^n(X; G)\)
“Dual cells”: Moore spaces \(M(G, n)\) with \(H_n = G\)	“Dual cells”: Eilenberg-Mac Lane spaces \(K(G, n)\) with \(\pi_n = G\)
Abstract: weight structure (Bondarko)	Abstract: t-structure (Beilinson-Bernstein-Deligne)

Primal	Dual
Cofibration \(A \hookrightarrow X\)	Fibration \(E \twoheadrightarrow B\)
Cofiber \(X/A\) (pushout)	Fiber \(F\) (pullback)
Suspension \(\Sigma X\)	Loop space \(\Omega X\)
Sphere \(S^n = \Sigma^n S^0\)	Eilenberg-Mac Lane \(K(G, n) = \Omega^{-n} HG\)
Co-H-space (comultiplication)	H-space (multiplication)
Homotopy groups \(\pi_n(X) = [S^n, X]\)	Cohomology \(H^n(X;G) = [X, K(G,n)]\)
Moore space \(M(G,n)\): \(\tilde{H}_n = G\)	Eilenberg-Mac Lane \(K(G,n)\): \(\pi_n = G\)
CW filtration / weight structure	Postnikov tower / t-structure