Marginal Stability, Chambers, and Wall Crossing

A wall of marginal stability is a real codimension-one locus where central charges of possible decay products align. On the wall, a BPS bound state can separate at no energy cost; across it, a protected index may jump. Wall crossing constrains that jump through charge pairings and phase ordering. It does not determine the entire BPS spectrum from algebra alone, and it says still less about unprotected excited states.

Required background. BPS particles and central charges fixes the charge and index conventions. Helpful background. Framed BPS states gives the halo interpretation, and Stokes phenomena provides a useful structural analogy for ordered discontinuity factors.

Locating a marginal-stability wall

Consider a possible decay

$$\gamma=\gamma_1+\gamma_2, \qquad Z_\gamma=Z_1+Z_2.$$

The BPS masses obey the triangle inequality

$$|Z_\gamma|\le |Z_1|+|Z_2|.$$

Marginality requires aligned, not anti-aligned, central charges:

$$\operatorname{Im}(Z_1\overline{Z_2})=0, \qquad \operatorname{Re}(Z_1\overline{Z_2})>0.$$

The first equation alone also includes anti-alignment, where $|Z_1+Z_2|=\bigl||Z_1|-|Z_2|\bigr|$ and the proposed two-body decay is not marginal in the same way.

Let

$$\kappa=\langle\gamma_1,\gamma_2\rangle\in\mathbb Z.$$

In the two-center approximation, a conventional orientation gives the separation

$$r_{12} =\frac{\kappa\,|Z_1+Z_2|} {2\operatorname{Im}(Z_1\overline{Z_2})}.$$

Thus the bound state exists only on the side where the right-hand side is positive, and its radius diverges as the wall is approached. Reversing the definition of the symplectic pairing or interchanging $1$ and $2$ reverses both signs without changing the physical chamber. The two-center separation and its stability interpretation follow from Denef 2000, arXiv:hep-th/0005049.

The primitive jump

Assume:

• $\gamma_1$ and $\gamma_2$ are primitive charges;
• only the two-center bound state becomes marginal on the chosen wall segment;
• their protected one-particle indices $\Omega(\gamma_i)$ are constant there;
• no additional aligned charge ray or scaling solution participates.

Define

$$\Delta_{\mathrm{st}}\Omega(\gamma_1+\gamma_2) =\Omega_{\mathrm{stable}}-\Omega_{\mathrm{empty}},$$

where “empty” names the adjacent chamber without this two-center bound state. In the second-helicity-supertrace convention used on the particle page,

$$\Delta_{\mathrm{st}}\Omega(\gamma_1+\gamma_2) =(-1)^{\kappa-1}|\kappa| \,\Omega(\gamma_1)\Omega(\gamma_2).$$

The factor $|\kappa|$ is the dimension of the relative angular-momentum multiplet, whose spin is $(|\kappa|-1)/2$. The sign is the helicity-supertrace sign. Authors who reverse the definition of the index or define $\Delta$ as final minus initial will display a different sign; the stable-minus-empty definition above removes that ambiguity here.

For an exact finite fixture, take $\kappa=2$ and two hypermultiplet constituents with $\Omega(\gamma_1)=\Omega(\gamma_2)=1$. Then

$$\Delta_{\mathrm{st}}\Omega(\gamma_1+\gamma_2)=-2.$$

The two orbital states form spin $1/2$ before the universal BPS multiplet is folded into the index. If a proposed calculation gives $+2$, it may be using the opposite index convention; if it gives magnitude other than two, it has lost the charge pairing or included extra states.

Chamber data are part of the answer

A reproducible wall-crossing statement records:

Input	What must be specified	Failure if omitted
Charge lattice	Basis, global form, and pairing convention	The integer $\kappa$ and allowed decays are ambiguous.
Central charges	Functions $Z_\gamma(u)$ and branch choices	The wall and phase order cannot be located.
Path	Initial chamber, oriented crossing, and endpoint	The sign and even the set of crossed walls are ambiguous.
Protected quantity	Unrefined or refined index and universal factors	Multiplicities from different conventions cannot be compared.
Spectrum near the wall	Every ray that becomes aligned	A primitive formula may be applied outside its hypotheses.
Completeness claim	Bounded charge sector or full spectrum	A local consistency check can be mistaken for a classification.

This information is especially important when several walls meet. A small deformation of the path can change their order; consistency requires the ordered product of jumps to be path-independent when no singularity is enclosed.

Ordered symplectomorphisms

Introduce formal Darboux variables $X_\gamma$ with multiplication determined by the charge lattice. In a classical sign convention, a BPS ray of charge $\gamma$ acts by

$$\mathcal K_\gamma: X_\beta\longmapsto X_\beta(1-X_\gamma)^{\langle\gamma,\beta\rangle\Omega(\gamma)}.$$

The Kontsevich–Soibelman statement is that the phase-ordered product of these transformations is locally constant as the vacuum varies, provided the angular sector boundaries do not cross BPS rays. When rays exchange order, the spectrum between them must change so that the total transformation remains the same. The general ordered-product identity is formulated in Kontsevich and Soibelman 2008, arXiv:0811.2435.

For $\langle\gamma_1,\gamma_2\rangle=1$, the elementary pentagon identity takes the schematic ordered form

$$\mathcal K_{\gamma_1}\mathcal K_{\gamma_2} =\mathcal K_{\gamma_2} \mathcal K_{\gamma_1+\gamma_2} \mathcal K_{\gamma_1},$$

with the displayed order tied to a chosen clockwise phase convention. The new middle factor is the bound state required on the other side. For larger pairing or nonprimitive charges, towers and rational invariants enter; the primitive formula alone is insufficient.

Framed derivation through halos

Place a line defect with core charge $\gamma_c$ and let bulk particles of charge $\gamma_h$ form halos. When the phase of $Z_{\gamma_h}$ crosses the phase selected by the defect, the halo radius passes through infinity. The framed generating function changes by a factor built from $\Omega(\gamma_h)$ and $\langle\gamma_c,\gamma_h\rangle$.

Now demand that the framed answer after transporting the defect between two vacua be independent of how the path is slightly deformed. The equality of the resulting ordered halo factors is the bulk wall-crossing identity. This argument clarifies both the strength and the limit of the result: it constrains protected indices because long multiplets cancel, and it assumes that every relevant BPS ray in the angular sector has been included. For the four-dimensional field-theory derivation, see Gaiotto, Moore, and Neitzke 2010, pp. 163–224, arXiv:0807.4723.

What wall crossing cannot establish

Wall-crossing identities are consistency equations. Given enough seed data, they can propagate an index to other chambers. They cannot, without those seeds and a completeness argument,

• prove that the proposed initial spectrum is complete;
• reconstruct long multiplets or unprotected degeneracies;
• determine a BPS wavefunction or its size far from the wall;
• justify a primitive formula when multiple or nonprimitive constituents align;
• cross a singular locus where the charge lattice or effective description itself changes without additional input.

An exact computation in a bounded charge cone is valuable evidence, but its scope should remain bounded to that cone.

Common pitfalls

Using phase equality without the positivity condition. Alignment and anti-alignment both set the imaginary part to zero. The real part distinguishes a marginal sum from a marginal difference.

Hiding the orientation of the jump. A bare $\Delta\Omega$ has an arbitrary sign. Define the chamber order or use stable-minus-empty as above.

Applying the primitive formula to a crowded ray. If $n\gamma_i$, other decompositions, or scaling solutions contribute, use the full ordered-product formula with the appropriate rational or refined invariants.

Exercises

Let $Z_1=1+i\epsilon$, $Z_2=2-i\epsilon$, and $\kappa=3$, with small real $\epsilon$.

1. Locate the wall.
2. Using the separation convention above, determine which sign of $\epsilon$ supports the two-center state to leading order.
3. If both constituents have index one, compute the stable-minus-empty primitive jump.

Solution

The wall is at $\epsilon=0$, where both central charges are positive real. To first order,

$$\operatorname{Im}(Z_1\overline{Z_2}) =\operatorname{Im}\bigl((1+i\epsilon)(2+i\epsilon)\bigr) =3\epsilon+O(\epsilon^2).$$

Since $\kappa=3>0$, the displayed $r_{12}$ is positive for $\epsilon>0$. The jump is

$$(-1)^{3-1}|3|=+3.$$

This determines a protected contribution in the two-center approximation, not a complete spectrum.

References

• Denef, Frederik. “Supergravity Flows and D-Brane Stability.” Journal of High Energy Physics 08 (2000): 050. arXiv:hep-th/0005049.
• Gaiotto, Davide, Gregory W. Moore, and Andrew Neitzke. “Four-Dimensional Wall-Crossing via Three-Dimensional Field Theory.” Communications in Mathematical Physics 299 (2010): 163–224. arXiv:0807.4723.
• Kontsevich, Maxim, and Yan Soibelman. “Stability Structures, Motivic Donaldson–Thomas Invariants and Cluster Transformations.” 2008. arXiv:0811.2435.