THE GOLD STANDARD · TOPOLOGY
Eliminating μ₃ changes μ₄—and changes what counts as invariant
A careful guide to higher-operation gauge changes, secondary cocycles, residual indeterminacy and the obstruction quotients that keep the answer independent of choices.
Choose a reading route
Chapter contents
- 1. Higher operations depend on choices, but obstructions should not
- 2. The minimal-model tower starts with μ₂
- 3. The first obstruction is the class of μ₃
- 4. Gauge elimination changes the next operation
- 5. Why the secondary class may live in a quotient
- 6. Worked bookkeeping example in a toy complex
- 7. Changed-context transfer: what if [μ₃] is non-zero?
- 8. A∞ versus C∞ and Harrison compatibility
- 9. From cochains to topological invariants
- 10. Limits, misconceptions and frequently asked questions
- 11. Evidence checks for an obstruction calculation
- 12. Sources and connected reading
1. Higher operations depend on choices, but obstructions should not
A differential graded algebra can transfer to a minimal A∞ or C∞ structure on its cohomology. The binary operation μ₂ is the familiar cohomology product. Higher operations μ₃, μ₄ and beyond record homotopy information that the product alone forgets.
The individual formulas depend on contraction data and on the chosen minimal-model coordinates. A change of coordinates—often called a gauge transformation—can alter a higher operation without changing the represented homotopy type. The real question is therefore not whether one displayed μ₃ or μ₄ vanishes, but whether its class can be removed by an allowed gauge.
- Fix the algebraic category: associative A∞ or commutative C∞.
- Fix grading, suspension and sign conventions.
- Identify the deformation complex: Hochschild for A∞, Harrison for C∞ in characteristic zero.
- Track both coboundaries and residual gauge actions before naming an obstruction.
2. The minimal-model tower starts with μ₂
In a minimal model μ₁ = 0, so cohomology carries no remaining differential. The Stasheff identities become a sequence of equations organised by arity. At low arity they say that μ₂ is associative and that μ₃ behaves like a cocycle in the cochain complex determined by μ₂.
For a commutative-up-to-homotopy C∞ structure, the higher cochains vanish on shuffle combinations, placing them in the Harrison subcomplex. This restriction matters: a Hochschild primitive may exist even when no compatible Harrison primitive exists.
Kadeishvili packages the higher operations as a twisting cochain satisfying a Maurer–Cartan-type equation. That viewpoint makes both obstruction classes and gauge changes visible in one calculation.
3. The first obstruction is the class of μ₃
Let δ denote the Hochschild or Harrison differential induced by μ₂. The arity-three identity gives δμ₃ = 0 in the usual minimal setting, so μ₃ determines a cohomology class [μ₃].
A gauge transformation whose first non-trivial component is a binary cochain h changes μ₃ by a coboundary, schematically μ₃ ↦ μ₃ − δh. Signs and normalisations vary with convention, but the invariant statement does not: μ₃ can be eliminated exactly when its class vanishes in the relevant deformation cohomology.
Fiorenza, Kawai, Lê and Schwachhöfer show that for their low-dimensional Poincaré DGCA setting a distinguished Harrison class [μ₃] is the only formality obstruction. That is a theorem with connectivity and dimension hypotheses, not a universal claim about every space.
4. Gauge elimination changes the next operation
Suppose [μ₃] = 0 and choose h with δh = μ₃. Applying the corresponding gauge makes μ′₃ = 0. It also changes μ₄. The corrected operation μ′₄ contains the original μ₄ plus terms built from h, μ₂ and μ₃. The exact signs depend on the bar or suspension convention.
This correction is essential. Simply setting a displayed μ₃ formula to zero while leaving μ₄ untouched mixes two coordinate systems. The resulting μ₄ need not satisfy the arity-four identity of the new gauge.
After μ′₃ = 0, the next Stasheff equation usually reduces to δμ′₄ = 0. The corrected μ′₄ is then a cocycle and becomes a candidate secondary obstruction.
5. Why the secondary class may live in a quotient
The primitive h is rarely unique. If h solves δh = μ₃, then h + z is another solution whenever δz = 0. A further higher gauge component can also alter μ′₄ by a coboundary. Consequently, changing the elimination data can move the μ₄ cocycle inside a family.
A well-defined secondary invariant must forget both ordinary boundaries and the shifts produced by residual gauges that preserve μ′₃ = 0. Schematic notation is [μ′₄] in Z⁴/(B⁴ + R⁴), where Z⁴ is the cocycle space, B⁴ the coboundaries and R⁴ the residual-gauge image. The actual degrees and subspaces depend on the algebra and convention.
| Object | Role | What must be quotiented |
|---|---|---|
| μ₃ | First higher cocycle | δ-exact changes |
| h | Choice eliminating μ₃ | Closed alternatives h + z |
| μ′₄ | Corrected fourth operation | Coboundaries and residual shifts |
| Secondary class | Gauge-stable information | Class in the appropriate subquotient |
6. Worked bookkeeping example in a toy complex
Consider a simplified arity-four cocycle space Z⁴ with basis e₁, e₂, e₃. Suppose ordinary boundaries span e₁ and residual gauges that keep μ₃ eliminated span e₂. After choosing one primitive h, the corrected operation is μ′₄ = 3e₁ + 2e₂ + e₃.
In the quotient Z⁴/(span(e₁) + span(e₂)), only the e₃ coordinate survives. The secondary obstruction is therefore the non-zero class represented by e₃.
Choose a different primitive and higher gauge. The representative might become 8e₁ − 2e₂ + e₃. Its raw coefficients have changed, but its quotient class is still e₃. This is the point of the obstruction quotient: it removes coordinate motion while preserving the part that no allowed change can erase.
| Representative | Boundary part | Residual-gauge part | Surviving class |
|---|---|---|---|
| 3e₁ + 2e₂ + e₃ | 3e₁ | 2e₂ | e₃ |
| 8e₁ − 2e₂ + e₃ | 8e₁ | −2e₂ | e₃ |
| e₁ + 4e₂ | e₁ | 4e₂ | 0 |
7. Changed-context transfer: what if [μ₃] is non-zero?
Now change the first assumption. If [μ₃] ≠ 0, no compatible h satisfies δh = μ₃, so the gauge slice μ′₃ = 0 does not exist. The secondary μ₄ construction above is not available in that form.
One may still study the full Maurer–Cartan element, filtered deformation problems or coupled equations involving μ₃ and μ₄. But it is misleading to quote a standalone μ₄ class as though the first obstruction had already vanished.
Transfer check: before computing any secondary invariant, document the vanishing of the primary class, the primitive used, the residual gauge group and the proof that changing those choices changes only the quotient representative.
8. A∞ versus C∞ and Harrison compatibility
An A∞ structure uses the Hochschild deformation complex. A C∞ structure encodes homotopy commutativity, so its higher operations lie in the Harrison part under suitable characteristic-zero hypotheses. The obstruction group can therefore shrink when commutative compatibility is imposed.
This distinction affects elimination. A cochain h that kills μ₃ in the full Hochschild complex may fail the shuffle conditions required for a C∞ gauge. Formality in the associative category and formality in the commutative category are related but should not be conflated.
Always state the coefficient field. Harrison decompositions and averaging arguments can behave differently in positive characteristic.
9. From cochains to topological invariants
Higher operations connect to Massey products, but the relationship is not one-to-one. A local Massey product depends on a defining system and can be undefined or indeterminate. A transferred μ₃ is a global operation in a chosen minimal model, while its cohomology class captures gauge-stable information.
The Bianchi–Massey tensor illustrates why subquotients matter. Crowley and Nordström define it on a subquotient of a fourth tensor power and show, within a dimension and connectivity range, that it detects rational homotopy information invisible to ordinary Massey products.
A good computation keeps three layers separate: a cochain formula, its class after coboundaries, and the further quotient needed to remove defining-system or gauge indeterminacy.
10. Limits, misconceptions and frequently asked questions
If one transferred model has μ₃ = 0, is the space formal?
Not in general. Higher operations may remain. Formality follows only when all relevant higher data can be removed, or when a theorem shows that no later obstruction can occur under stated hypotheses.
Is μ₄ automatically a cocycle?
Not before checking the Stasheff identity in the chosen gauge. After μ₃ has been compatibly eliminated, the corrected μ₄ often becomes a cocycle.
Why not report the raw coefficients of μ₄?
They depend on basis, transfer and gauge choices. An invariant statement identifies the cohomology or quotient class that survives allowed changes.
Does vanishing of all ordinary Massey products imply formality?
No in general. Bianchi–Massey-type invariants provide examples of information that can remain when ordinary Massey products vanish.
A misconception worth removing
Gauge elimination is not deletion by notation. It is a coordinated change of higher operations and morphism components that must continue to satisfy the full identities.
11. Evidence checks for an obstruction calculation
- Verify arity, internal degree and signs in one declared convention.
- Check δμ₃ = 0 directly.
- Exhibit h and verify δh = μ₃ before claiming elimination.
- Recompute the transformed μ′₄, including quadratic correction terms.
- Check δμ′₄ = 0 in the new gauge.
- Describe all changes from alternative primitives and higher gauge components.
- Form the correct quotient and test representative independence.
- Use computer algebra only after encoding graded signs and shuffle relations explicitly.
The calculation is complete only when a second reader can change the primitive and recover the same quotient class.
12. Sources and connected reading
Primary and specialist sources: Kadeishvili, Structure of A∞-algebra and Hochschild and Harrison cohomology; Fiorenza, Kawai, Lê and Schwachhöfer, Almost formality of manifolds of low dimension; Roitzheim and Whitehouse on uniqueness of A∞ structures and Hochschild cohomology; and Crowley and Nordström on the Bianchi–Massey tensor and rational homotopy type.
Continue through The Gold Standard of Topology, the Gold Standard of Education and the How X Works hub.
Specialist route: Massey products and defining-system indeterminacy, the Bianchi–Massey tensor, higher formality obstructions, and transferred μ₃ and μ₄ operations.