Which Pure Spinor Gauge Choice Is Real? Canonical Non-Invariance of Kinematic Numerators

Two computations of the same scattering amplitude can share the final answer and still disagree about every intermediate object. In early-universe cosmology this has a sharp form. Quantum initial conditions for inflation and canonical invariance showed that popular rules for choosing the quantum vacuum, such as minimizing the Hamiltonian at an initial time, change their answer when one merely rescales the field or adds a total derivative to the action. The paper calls this failure canonical non-invariance. The one rule that survives these changes is built directly from the covariant stress-energy tensor. We propose to apply the same test to a modern tool for computing amplitudes: the pure spinor formalism for ten-dimensional super-Yang-Mills (SYM) theory. This formalism adds constrained "ghost" spinor variables so that supersymmetry and Lorentz symmetry stay manifest. A composite operator, the b-ghost, then defines the gauge choice and the propagator.

The claim is as follows. The Siegel gauge used by 10D super-Yang-Mills scattering amplitudes from its pure spinor action is a family of gauges, not a single one. The b-ghost is fixed only by the condition that the BRST charge $Q$ (the operator whose cohomology picks out physical states) maps it to the Hamiltonian. Adding any $Q$-exact piece to $b$ gives another valid choice. Each representative yields different kinematic numerators: the momentum-and-polarization factors attached to each cubic diagram, whose Jacobi identities realize color-kinematics duality. We will map this family and determine which deformations keep the Jacobi identities and which destroy them. We will also test the cosmology lesson directly: does a prescription tied only to the Hamiltonian, rather than to a chosen $b$, remove the ambiguity? The proposal treats the cosmology direction as the method and pure spinor amplitude computation as the place where it is applied. The practical stakes are the size of the numerators and the short-distance divergences that currently afflict them.

Core Claim

Let $b_0$ be a b-ghost satisfying $\{Q,b_0\}=T$, where $T$ is the worldline Hamiltonian $P^2/2$. Any operator $\chi$ of the right ghost number generates a one-parameter family

$$ b_s = b_0 + s\,[Q,\chi], \qquad \{Q,b_s\}=T \ \text{for all } s, $$

and each $b_s$ defines its own Siegel gauge $b_s V=0$ for the master-action superfield $V$. Here $s$ is a real deformation parameter.

Our hypothesis has three parts, which we test separately:

  1. Amplitudes are invariant. Full amplitudes do not depend on $s$, as BRST invariance requires.
  2. Jacobi identities survive only for special $\chi$. The numerators $n_i(s)$ shift by generalized gauge transformations. The kinematic Jacobi identities hold only for those $\chi$ for which the bracket built from $b_s$ stays nilpotent up to terms that vanish on the physical states.
  3. The divergences depend on $s$. The short-distance divergences of nested b-ghosts noted by Ben-Shahar and Guillen also vary with $s$. If so, they come from the choice of representative and are not intrinsic to the formalism.

A second deformation class has the same structure. It consists of superfield redefinitions $V\to V+\alpha\,F(V)$ and the addition of $Q$-exact total terms to the master action. These are the direct analogues of the field rescalings and surface terms in the cosmology paper.

Two outcomes are possible, and both are informative. If the numerators are canonically invariant within some natural subclass, that subclass is the "stress-tensor-like" prescription, and it defines numerators that are meaningful without a choice of representative. If they are not invariant, the ambiguity becomes a resource. One can search the family for the representative that gives the smallest numerators, or finite isolated numerators, while the invariant amplitude certifies correctness.

Relation to Prior Work

The formalism we deform. 10D super-Yang-Mills scattering amplitudes from its pure spinor action supplies the method. It computes tree amplitudes from the pure spinor master action using perturbiners, which are generating-function solutions of the classical field equations. It shows that the Siegel gauge produces color-kinematics duality, with numerators built from nested b-ghosts. It fixes one b-ghost and does not ask how the numerators move when $b$ changes; that question is the one we pose.

The deformation directions, and the obstruction we inherit. Two 11D papers give competing b-ghosts:

These two papers give the first concrete pair of representatives whose numerators can be compared. The fact that $\{b,b\}$ vanishes only up to $Q$-exact terms is exactly what we expect to break Jacobi identities off the physical states.

The extension target. Tree-level 11D supergravity amplitudes from the pure spinor worldline and The pure spinor superparticle and 10D super-Yang-Mills amplitudes supply worldline versions of the same computation. The first reports unregularized poles at higher multiplicity. That defect is the second place where representative dependence may be hiding.

The analogy and its limit. The cosmology paper motivates the whole design. Its key finding is that minimizing an operator expectation value is representation-dependent, while a covariant stress-tensor criterion is not. That finding suggests the analogous prescription here: one that refers only to $T$.

We are honest about where the analogy may fail. The cosmology paper studies free fields and vacuum choice, not gauge fixing. The analogy is a hypothesis to test, not a theorem to borrow.

The axiomatic literature as a constraint. The indefinite-metric work in Scattering Theory for Quantum Fields with Indefinite Metric and Quantum Fields Obtained from Convoluted Generalized White Noise Never Have Positive Metric shows that interacting relativistic models in these settings live in spaces with negative-norm states. It reminds us that ghost sectors are not optional overhead; they are where interactions and covariance coexist. We do not attempt the Krein-space or LSZ analysis of the pure spinor superparticle in this proposal. That is a separate and much harder project.

Research Program

Step 1: Classify admissible deformations. For the 10D non-minimal formalism, we enumerate the low-order candidates for $\chi$ built from the non-minimal variables $\bar\lambda$ and $r$. We do the same for field redefinitions of $V$ that preserve ghost number and Lorentz covariance, and for $Q$-exact additions to the action. For each, we compute $\{b_s,b_s\}$ symbolically. This determines whether the kinematic bracket remains nilpotent exactly, only up to $Q$-exact terms, or not at all.

Step 2: Compute numerators along each family. We run the perturbiner recursion to obtain the 3-, 4- and 5-point tree numerators as explicit functions of $s$ and $\alpha$. On each family we record four quantities:

  • the full amplitude;
  • the Jacobi residual on each triple of diagrams;
  • the coefficient of the short-distance divergence in each isolated numerator;
  • the number of independent superspace terms in the numerators. This count is our concrete measure of computational complexity.

Step 3: Test the stress-tensor-anchored prescription. This is the analogue of the invariant cosmological rule. We define the propagator through quantities fixed by $T$ alone. An example is averaging over a $Q$-exact orbit with a covariant weight, or imposing a gauge condition stated through $\{Q,\cdot\}=T$ without choosing $b$. We then check whether numerators built this way are independent of the representative.

Step 4: Apply the same tests in 11D. The two published b-ghosts give a two-point family there. We compare the 4-point supergravity correlator and the pole structure at higher points between them.

Step 5: Validate the pipeline on a known case. We reproduce the known non-invariance of Hamiltonian diagonalization under field rescaling for the Mukhanov-Sasaki equation, and the invariance of the stress-tensor rule. This checks that our test for invariance detects the effect where it is known to exist.

Baselines

The do-nothing reference. The naive reference point is the published single-gauge computation of Ben-Shahar and Guillen: one b-ghost, Siegel gauge, perturbiner numerators. We will regenerate it with the same symbolic pipeline, the same superspace conventions and the same GAMMA-style Clifford algebra routines used in Simplified D = 11 pure spinor b ghost, so that term counts are comparable. No reported result was found on Lacuna for numerator term counts or divergence coefficients in that computation. Our first deliverable is therefore to establish those numbers. A deformed representative "beats" this baseline if it meets two conditions at the same multiplicity:

  • it keeps exact Jacobi identities;
  • it either reduces the independent term count or makes isolated numerators finite.

The strongest external competitor. This is the pure spinor worldsheet (CFT) computation, which uses operator products of vertex operators on the string worldsheet. Ben-Shahar and Guillen report that it agrees with their perturbiner amplitudes. We use it in two ways. It provides the correctness check for every member of our families, and its numerators provide an independent point in "numerator space" against which our family can be compared. Because both start from the same superspace building blocks, the comparison is like for like. No reported result was found on Lacuna for its numerator sizes either.

The 11D baseline. In 11D the baseline is the agreement between perturbiner and worldline 4-point amplitudes in Tree-level 11D supergravity amplitudes from the pure spinor worldline.

Ablations. These isolate the key ingredient by deforming one thing at a time:

  • only the b-ghost representative;
  • only superfield redefinitions;
  • only $Q$-exact action terms.

The ablations tell us which class carries the non-invariance. A further ablation removes the $T$-anchoring from Step 3 to check that this prescription, not merely averaging, is responsible for any invariance we find.

Why not just…?

Why not just invoke BRST invariance and stop, since amplitudes cannot depend on the gauge? That argument settles invariance of the full amplitude, and we expect to confirm it. It says nothing about the numerators, the kinematic algebra or the nested-b divergences. These are exactly the objects that determine computational cost and that Ben-Shahar and Guillen flag as unresolved. The real threat is the following outcome: every admissible deformation turns out to be an ordinary generalized gauge transformation that leaves both Jacobi identities and divergences unchanged. Step 2 detects this directly. If it happens, we report the negative result that pure spinor numerators are canonically rigid. That is useful in its own right, and the program then narrows to the 11D two-representative comparison.

Why not just fix light-cone gauge and avoid ghosts altogether? Equivalence of the 11D pure spinor and Brink-Schwarz-like superparticle cohomologies explains why the Brink-Schwarz-like superparticle cannot be quantized covariantly: its fermionic constraints mix first and second class. Light-cone gauge also hides the action-level origin of color-kinematics duality that Ben-Shahar and Guillen expose. Light-cone numerators would answer a different question. They cannot tell us whether the covariant prescription itself is ambiguous.

Why not just solve for color-kinematics numerators with a generic ansatz, independent of any action? That would give some valid numerators. It would not tell us which ones the pure spinor action produces, or why the divergences appear. The value of the action-based route is that each numerator traces to a specific b-ghost, so the ambiguity can be located and controlled. We will still compare our minimal term counts against an ansatz solution at 4 and 5 points as a sanity check on the complexity claim.

Why not just use the worldsheet formalism, which is already known to agree? In 10D it does agree and serves as our check. In 11D, however, there is no fundamental superstring to supply a worldsheet. That is why the worldline formalism and the non-minimal b-ghosts exist at all, as The 11D pure spinor ghost number zero vertex operator makes clear. The ambiguity we study matters most precisely there.

Decisive Checks

The proposal stands or falls on the 4- and 5-point 10D computation in Step 2. We look for three signatures in turn:

  1. Is there an ambiguity at all? If the numerators vary with $s$ at fixed amplitude, the ambiguity is real. If they do not, the canonical-invariance analogy is false for this formalism, and we say so.
  2. Is it an obstruction or a choice? If the Jacobi residual is nonzero exactly when $\{b_s,b_s\}$ fails to vanish off-shell, nilpotency of $b$ is identified as the obstruction to color-kinematics duality. This directly links the 11D result $\{b,b\}=Q\Omega$ to whether 11D numerators can be organized the same way. If instead the divergence coefficient depends on $s$ and vanishes for some representative, the regularization problem left open by Ben-Shahar and Guillen is solved by choosing the representative rather than by an external regulator.
  3. Does the cosmology lesson transfer? The final check is whether the $T$-anchored prescription gives numerators that are independent of the representative. If it does, the lesson that prescriptions built from the covariant Hamiltonian or stress tensor are the invariant ones carries over from inflationary vacua to amplitude computation.