03.07.16 · modern-geometry / gauge-theory

$B$ -field as a gerbe connection

shipped3 tiersLean: none

Anchor (Master): Brylinski *Loop Spaces, Characteristic Classes and Geometric Quantization* (Birkhäuser 1993) §5 (gerbe connective and curving structures); Murray 1996 *J. London Math. Soc.* 54, 403-416 (bundle gerbe formulation); Freed-Hopkins-Teleman 2003/2007/2011 (gerbes and loop groups); Bouwknegt-Mathai 2000 *J. High Energy Phys.* 03:007 (twisted K-theory of D-branes); Polchinski 1995 *Phys. Rev. Lett.* 75, 4724-4727 (D-brane charge)

Intuition [Beginner]

A magnetic field in three dimensions pushes on electric charges, and the electric charge couples to the magnetic vector potential along the worldline it traces out in spacetime. A fundamental string sweeps out a two-dimensional surface in spacetime — its worldsheet — and the natural object that couples to a surface is not a one-form but a two-form. The string-theory $B$ -field is exactly this object: a two-form on spacetime whose integral over the worldsheet contributes a phase to the string's action.

The catch is that the $B$ -field is not a single global two-form. On each patch of spacetime you can write down a local two-form, but on overlaps the local pieces do not match up by a function the way magnetic vector potentials do. They match up by a line bundle's transition data. The right object that organises this gluing — one rung above a line bundle in the ladder of geometric structures — is called a $U (1)$ -gerbe.

Why does the gluing matter? Because the field strength $H = d B$ is a closed three-form whose integral over a closed three-cycle is quantised in integer units, and the integer reflects the gerbe class, not the local two-form. This is the analogue of magnetic flux quantisation, one dimension up.

Visual [Beginner]

A schematic showing a string worldsheet $Σ$ embedded in a spacetime $M$ with three coordinate patches indicated. On each patch a local two-form is written; on each pair-wise overlap a line bundle's transition data sits where a function would for ordinary gauge theory; on each triple overlap a $U (1)$ -valued function ensures the data is consistent. An arrow points from the gerbe-data picture to a three-form $H$ on $M$ labelled "curvature", with a small inset showing the integer flux quantisation of $H$ over a closed three-cycle.

The picture conveys why a single global two-form cannot tell the whole story: the global content is in the way local two-forms patch, and that patching is the gerbe.

Worked example [Beginner]

Take the spacetime $M = S^{3}$ and ask: what gerbes does it carry? Compute the integer flux of the curvature for the level-one case.

Step 1. Set up the topology. $S^{3}$ is the three-sphere. Its third cohomology group with integer coefficients is $H^{3} (S^{3}, Z) = Z$ . A choice of integer is called the level $k$ .

Step 2. Pick the level $k = 1$ . The gerbe of level one carries a curvature three-form $H$ on $S^{3}$ whose total flux over $S^{3}$ equals $2 π$ . Concretely, normalise the round metric so that the volume of $S^{3}$ is $2 π^{2}$ . Then $H$ can be chosen as a uniform multiple of the volume form: $H = (1/ π) vol_{S^{3}}$ , and a check shows the total of $H$ across $S^{3}$ equals $(1/ π) (2 π^{2}) = 2 π$ , matching the flux quantisation condition with integer one.

Step 3. Cover $S^{3}$ by two patches: the north-pole neighbourhood $U_{N}$ and the south-pole neighbourhood $U_{S}$ , both topologically discs. On each disc the closed three-form $H$ admits a local two-form primitive $B_{N}$ on $U_{N}$ and $B_{S}$ on $U_{S}$ . On the equatorial overlap $U_{N} \cap U_{S}$ — which is topologically a cylinder $S^{2} \times R$ — the difference $B_{N} - B_{S}$ is a closed two-form whose period over the equatorial $S^{2}$ is $2 π$ .

Step 4. Read the moral. The difference $B_{N} - B_{S}$ being closed but having non-zero period means it is not a one-form's exterior derivative globally: a line bundle with first Chern class equal to one sits on the equatorial $S^{2}$ and provides the patching data. The gerbe structure is the recipe that records this patching.

What this tells us: the level-one $B$ -field on $S^{3}$ is the simplest non-identity $U (1)$ -gerbe with connection. Higher levels $k$ multiply the curvature by $k$ , and the equatorial line bundle's first Chern class becomes $k$ .

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $M$ be a smooth manifold and let $U = {U_{α}}$ be a good open cover. A $U (1)$ -gerbe with connective structure and curving on $M$ is the following data, modulo refinement of cover and gauge equivalence.

On each patch $U_{α}$ , a 2-form $B_{α} \in Ω^{2} (U_{α})$ .

On each double overlap $U_{α β} = U_{α} \cap U_{β}$ , a 1-form $A_{α β} \in Ω^{1} (U_{α β})$ , with $A_{α β} + A_{β α} = 0$ .

On each triple overlap $U_{α β γ}$ , a smooth function $g_{α β γ} : U_{α β γ} \to U (1)$ , totally antisymmetric in its three indices.

These data satisfy three descent equations:

(i) $g_{β γ δ} g_{α γ δ}^{- 1} g_{α β δ} g_{α β γ}^{- 1} = 1$ on each quadruple overlap.

(ii) $A_{α β} + A_{β γ} + A_{γ α} = - i g_{α β γ}^{- 1} d g_{α β γ}$ on each triple overlap.

(iii) $B_{β} - B_{α} = d A_{α β}$ on each double overlap.

The triple $(B_{α}, A_{α β}, g_{α β γ})$ is the Čech-Deligne cocycle of the gerbe in the smooth Deligne complex $Z (3)_{D}^{\infty} = (\underline{Z} \to Ω^{0} \to Ω^{1} \to Ω^{2})$ . Equivalence is by adding the Čech-Deligne coboundary of a degree-2 cochain.

The field strength is the globally defined 3-form $H \in Ω^{3} (M)$ obtained from $H ∣_{U_{α}} = d B_{α}$ . Equation (iii) shows that $d B_{β} = d B_{α}$ on overlaps, so $H$ glues to a closed global 3-form. Its de Rham class equals $2 π$ times the image of the Dixmier-Douady class $[g_{α β γ}] \in H^{3} (M, \underline{Z}) = H^{3} (M, Z)$ under the comparison $H^{3} (M, Z) \otimes R \to H^{3} (M, R)$ realised by de Rham cohomology. See 03.06.09 for the classifying statement.

The gauge transformations of the $B$ -field are described by a 1-form $Λ_{α}$ on each $U_{α}$ and a $U (1)$ -valued function $h_{α β}$ on each overlap satisfying $h_{β γ} h_{α γ}^{- 1} h_{α β} = 1$ on triple overlaps (so $(h_{α β})$ defines a line bundle $L$ on $M$ with first Chern class in $H^{2} (M, Z)$ ). The action is $$ B_\alpha \to B_\alpha + d\Lambda_\alpha, \qquad A_{\alpha\beta} \to A_{\alpha\beta} + \Lambda_\beta - \Lambda_\alpha - i,h_{\alpha\beta}^{-1} d h_{\alpha\beta}, \qquad g_{\alpha\beta\gamma} \to g_{\alpha\beta\gamma} \cdot h_{\beta\gamma} h_{\alpha\gamma}^{-1} h_{\alpha\beta}. $$ The 1-form $Λ_{α}$ is the 0-form-shift gauge symmetry, recovering the naïve $B \to B + d Λ$ when $L$ is the identity line bundle. The line bundle $L$ encodes the genuinely 2-form gauge symmetry: $B$ -field gauge transformations are themselves classified up to homotopy by line bundles.

Counterexamples to common slips

The local 2-forms $B_{α}$ are not the components of a global 2-form: equation (iii) shows they differ on overlaps by $d A_{α β}$ , and $A_{α β}$ is not pure gauge when $g_{α β γ}$ is non-vanishing on triple overlaps.
A gauge transformation $B \to B + d Λ$ with a globally defined $Λ \in Ω^{1} (M)$ is the simple case. The full gauge group is the line-bundle-shifted version, and the line bundle is itself a 2-form gauge parameter modulo line-bundle isomorphism.
The Dixmier-Douady class $[H /2 π] \in H^{3} (M, Z)$ is integral, not just real: the integrality is the descent condition (i) on the triple-overlap cocycle, exactly as the first Chern class of a line bundle is the integrality coming from the double-overlap cocycle.
The $B$ -field is not classified by $H^{3} (M, R)$ : two $B$ -fields whose 3-curvatures are de Rham cohomologous can sit on inequivalent gerbes. The torsion subgroup of $H^{3} (M, Z)$ records gerbes invisible to the curvature.

Key theorem with proof [Intermediate+]

Theorem (gerbe interpretation of the $B$ -field; Brylinski 1993 §5). Let $M$ be a smooth manifold. The set of equivalence classes of $U (1)$ -gerbes with connective structure and curving on $M$ is in natural bijection with the smooth Deligne hypercohomology group $H^{3} (M, Z (3)_{D}^{\infty})$ . The forgetful map to $H^{3} (M, Z)$ assigning a gerbe its Dixmier-Douady class fits into a short exact sequence $$ 0 \to H^2(M, \mathbb{R}/\mathbb{Z}) \to H^3(M, \mathbb{Z}(3)D^\infty) \to \Omega^3\mathbb{Z}(M) \times_{H^3(M, \mathbb{R})} H^3(M, \mathbb{Z}) \to 0, $$ where $Ω_{Z}^{3} (M)$ is the group of closed 3-forms with integral periods and the fibre product is along the de Rham comparison.

Proof sketch. The smooth Deligne complex $Z (3)_{D}^{\infty}$ on $M$ is the complex of sheaves $$ \underline{\mathbb{Z}} \hookrightarrow \Omega^0 \xrightarrow{d} \Omega^1 \xrightarrow{d} \Omega^2, $$ placed in degrees $0, 1, 2, 3$ . Its hypercohomology in degree 3 is computed by the Čech-Deligne double complex on a good cover. A degree-3 cocycle has components $(B_{α}, A_{α β}, c_{α β γ}, n_{α β γ δ})$ where $c_{α β γ} : U_{α β γ} \to R /2 π Z$ and $n_{α β γ δ} : U_{α β γ δ} \to Z$ are subject to the descent equations (i)-(iii) of the formal definition (with $g_{α β γ} = exp (i c_{α β γ})$ ) together with the additional constraint that $c_{β γ δ} - c_{α γ δ} + c_{α β δ} - c_{α β γ} = 2 π n_{α β γ δ}$ . This last equation is the integrality of the Dixmier-Douady class: it states that the Čech cocycle $g_{α β γ}$ , viewed as a class in $H^{2} (M, \underline{U (1)})$ , lifts to a class in $H^{3} (M, \underline{Z})$ via the exponential short exact sequence $0 \to \underline{Z} \to \underline{R} \to \underline{U (1)} \to 0$ . The bijection with $U (1)$ -gerbes is by reading off the cocycle data directly.

The short exact sequence is the standard differential-cohomology exact sequence for the Deligne complex of weight 3, proved by writing the smooth Deligne complex as a homotopy pullback $Z (3)_{D}^{\infty} ≃ \underline{Z} [3] \times_{Ω_{R}^{\geq 3}} Ω_{R}^{\leq 2}$ in the derived category and computing hypercohomology. The kernel $H^{2} (M, R / Z)$ records the flat gerbes — those whose 3-curvature vanishes but whose Dixmier-Douady class can carry torsion. $□$

Bridge. The theorem builds toward the entire architecture of higher gauge theory and identifies the $B$ -field with the connective + curving data of a $U (1)$ -gerbe. The foundational reason it holds is exactly that the smooth Deligne complex on $M$ classifies connections on higher principal bundles, with the level $p$ controlling the rank of the gauge field: $p = 2$ recovers ordinary $U (1)$ -gauge theory (the Kostant-Weil isomorphism 03.06.08), $p = 3$ recovers gerbes, and the general $p$ -form gauge theory of $(p - 1)$ -branes follows the same pattern. The central insight is that the Kalb-Ramond $B$ -field is exactly the $p = 3$ analogue of the electromagnetic vector potential — putting these together with the descent equations (i)-(iii), one identifies the local 2-form $B_{α}$ with the higher-rank potential and the curvature $H = d B$ with the higher-rank field strength. The bridge is the recognition that 2-form gauge transformations are themselves gauge fields one rung lower, so the gauge group of $B$ -field transformations is itself a stack and the gerbe is its classifying object. This same pattern appears again in 03.10.03 pending (the WZW model), where the Wess-Zumino term picks up its globally well-defined incarnation only through the gerbe interpretation, and the integer level $k$ of the WZW model is exactly the Dixmier-Douady class of the level- $k$ gerbe on the target group. Connections to the sister gauge theory 03.07.05 (Yang-Mills): both are connections, but on different categorical levels — Yang-Mills lives on a principal bundle (a level-1 object), the $B$ -field lives on a gerbe (a level-2 object).

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

Verify equation (ii) of the descent system as a consistency condition on triple overlaps: deriving $d (A_{α β} + A_{β γ} + A_{γ α})$ and using (iii) twice, show that the right-hand side $- i g^{- 1} d g$ is the unique closed 1-form representative.

Hint

Compute $d (A_{α β} + A_{β γ} + A_{γ α})$ using $d A_{α β} = B_{β} - B_{α}$ on each overlap and observe the telescoping sum.

Answer

Using (iii) on each pairwise overlap: $d A_{α β} = B_{β} - B_{α}$ , $d A_{β γ} = B_{γ} - B_{β}$ , $d A_{γ α} = B_{α} - B_{γ}$ . Summing, $$ d(A_{\alpha\beta} + A_{\beta\gamma} + A_{\gamma\alpha}) = (B_\beta - B_\alpha) + (B_\gamma - B_\beta) + (B_\alpha - B_\gamma) = 0. $$ So $A_{α β} + A_{β γ} + A_{γ α}$ is a closed 1-form on the triple overlap $U_{α β γ}$ , and the descent equation (ii) identifies it with the $U (1)$ -Maurer-Cartan form $- i g^{- 1} d g$ of $g = g_{α β γ}$ . Both sides are closed; the descent condition fixes the gauge representative. Rubric: full credit for the telescoping computation and the Maurer-Cartan identification.

Exercise 4 (medium, short-answer).

Compute the gerbe holonomy $exp (2 π i \int_{Σ} B)$ for a closed worldsheet $Σ ≅ S^{2}$ mapped into $M = S^{3}$ by the inclusion of an equatorial 2-sphere, with $S^{3}$ carrying the level- $k$ gerbe. Express the answer in terms of $k$ .

Hint

The 2-sphere $Σ$ bounds a hemisphere $D \subset S^{3}$ . By Stokes (carefully applied through the patches), the worldsheet holonomy can be reduced to $exp (i \int_{D} H)$ . The integral is half the total flux.

Answer

$exp (i \cdot π k) = (- 1)^{k}$ . The equatorial $S^{2} \subset S^{3}$ bounds a hemisphere $D$ . The gerbe holonomy on a closed surface $Σ = \partial D$ equals $exp (i \int_{D} H)$ , and $\int_{D} H = (1/2) \int_{S^{3}} H = π k$ since the hemisphere is half of $S^{3}$ . So the holonomy is $exp (iπ k) = (- 1)^{k}$ . For $k = 1$ the equatorial worldsheet picks up a minus sign; this is the trace of the non-triviality of the gerbe. Rubric: full credit for Stokes' theorem applied to $H = d B$ across patches and the half-flux computation.

Exercise 6 (medium, short-answer).

State the descent equations (i)-(iii) of a $U (1)$ -gerbe with curving in the language of differential characters (Cheeger-Simons): a degree-3 differential character on $M$ is a homomorphism from the group of smooth 3-cycles to $R / Z$ whose value on boundaries integrates a global 3-form.

Hint

A differential character of degree $p$ is the data of a smooth function $χ : Z_{p} (M) \to R / Z$ and a global form $ω \in Ω^{p + 1} (M)$ with $χ (\partial c) = \int_{c} ω mod Z$ .

Answer

A degree-3 differential character is a pair $(χ, H)$ with $χ : Z_{3} (M) \to R / Z$ a homomorphism and $H \in Ω_{\geq int}^{4} (M)$ wait — the convention is $H$ in degree $p + 1 = 4$ for character degree $p = 3$ ; correcting: degree-2 character on $Z_{2} (M) \to R / Z$ for ordinary line bundles with curvature $F \in Ω^{2}$ , degree-3 character on $Z_{3} (M) \to R / Z$ — actually the standard convention pairs a degree- $p$ character with $Z_{p - 1}$ — different sources use different indexing. The gerbe-with-curving data $(B_{α}, A_{α β}, g_{α β γ})$ assemble to a degree-3 Cheeger-Simons character $χ : Z_{2} (M) \to R / Z$ with $χ (\partial c) = \int_{c} H mod Z$ , where $H$ is the curvature 3-form. The Dixmier-Douady class is the integral lift of $χ$ via the universal coefficient sequence; descent (i)-(iii) is the cocycle realisation of $χ$ . Rubric: full credit for the character-curvature pairing and identifying $H$ as the curvature in the Cheeger-Simons sense.

Exercise 7 (hard, short-answer).

For $M = SU (2) ≅ S^{3}$ with the standard Cartan-Killing metric normalised so that $vol (SU (2)) = 2 π^{2}$ , write down the level- $k$ Wess-Zumino-Witten action for a smooth map $ϕ : Σ \to SU (2)$ in the form $S_{W Z W} [ϕ] = \int_{Σ} ϕ^{*} B + (k /12 π) \int_{B} ϕ^{*} H$ , identify the integer level $k$ with the Dixmier-Douady class of the level- $k$ gerbe on $SU (2)$ , and explain why the action is well-defined modulo $2 π Z$ regardless of which 3-manifold $B$ with $\partial B = Σ$ is chosen.

Hint

Two choices $B$ and $B^{'}$ of bounding 3-manifolds differ by a closed 3-manifold $B - B^{'}$ . The integral $\int_{B - B^{'}} ϕ^{*} H$ is a flux of $H$ over a closed 3-cycle, hence an integer multiple of $2 π$ once divided by $2 π k$ and times $k$ .

Answer

The level- $k$ WZW action is $$ S_{WZW}[\phi] = \frac{k}{4\pi}\int_\Sigma \mathrm{tr}(\phi^{-1} d\phi \wedge \star \phi^{-1} d\phi) + \frac{k}{24\pi^2}\int_B \mathrm{tr}((\phi^{-1} d\phi)^3), $$ where $B$ is any 3-manifold with $\partial B = Σ$ and $ϕ$ extended over $B$ . The 3-form $ω = (1/24 π^{2}) tr ((g^{- 1} d g)^{3})$ is the bi-invariant generator of $H^{3} (SU (2), Z) = Z$ — it integrates to $1$ over the fundamental class. So $ϕ^{*} H = k ϕ^{*} ω$ defines the level- $k$ gerbe's curvature. Two choices $B, B^{'}$ of bounding 3-manifolds satisfy $\partial (B - B^{'}) = 0$ , making $B - B^{'}$ a closed 3-cycle; the integral $\int_{B - B^{'}} k ϕ^{*} ω$ lies in $2 π k Z$ by the integer-flux quantisation. Dividing by the action's normalisation $k / (24 π^{2})$ gives a contribution of $2 π \cdot k \cdot (integer)$ , hence the exponentiated action $exp (i S_{W Z W})$ is well-defined for $k \in Z$ . The level $k$ is the Dixmier-Douady class of the level- $k$ gerbe on $SU (2)$ — see 03.06.09. Rubric: full credit for the WZ-term construction, the bounding-3-manifold independence argument, and the $k \in Z$ quantisation conclusion.

Exercise 8 (hard, short-answer).

Sketch the argument that D-brane charge in a spacetime $M$ with $B$ -field of Dixmier-Douady class $[H] \in H^{3} (M, Z)$ takes values in twisted K-theory $K^{*} (M, [H])$ rather than ordinary K-theory $K^{*} (M)$ , following Bouwknegt-Mathai 2000 and the worldsheet anomaly argument of Freed-Witten 1999.

Hint

The Freed-Witten anomaly is the worldsheet anomaly that arises when a D-brane wraps a submanifold $W \subset M$ on which the restricted $B$ -field has non-vanishing Dixmier-Douady class. Cancellation requires a twisted bundle on $W$ .

Answer

A D-brane wraps a submanifold $W \subset M$ carrying a vector bundle $E \to W$ that classifies the Chan-Paton degrees of freedom. The open-string worldsheet anomaly cancellation (Freed-Witten 1999) requires $W_{3} (W) + [H ∣_{W}] = 0$ in $H^{3} (W, Z)$ , where $W_{3}$ is the third integral Stiefel-Whitney class. When $[H ∣_{W}] \neq = 0$ , the bundle $E$ cannot be an ordinary $U (N)$ -bundle on $W$ — it must be a twisted vector bundle, i.e. a bundle for the gerbe of class $[H ∣_{W}]$ rather than the identity gerbe. The Grothendieck group of such twisted bundles is the twisted K-theory $K^{0} (W, [H ∣_{W}])$ , defined by Bouwknegt-Mathai as the K-theory of the twisted PU(H)-bundle (or equivalently the K-theory of sections of a continuous-trace $C^{*}$ -algebra bundle with Dixmier-Douady class $[H ∣_{W}]$ ). D-brane charge sits in $K^{*} (M, [H])$ via the pushforward $ι_{!} : K^{0} (W, [H ∣_{W}]) \to K^{*} (M, [H])$ along the embedding $ι : W ↪ M$ . The level-1 abelian case recovers the integer charge of a D-brane in $K^{0} (M) = K^{0} (M, 0)$ ; non-vanishing $[H]$ deforms this to the twisted K-theory of the spacetime. See 03.06.09 for the Dixmier-Douady class and 03.08.01 for K-theory. Rubric: full credit for the Freed-Witten anomaly statement and the twisted-bundle identification.

Lean formalization [Intermediate+]

Mathlib has no formalisation of $U (1)$ -gerbes, no smooth Deligne complex, and no Dixmier-Douady class as the obstruction to a gerbe being the identity. The intended formalisation reads schematically:

import Mathlib.Geometry.Manifold.SmoothManifoldWithCorners
import Mathlib.Geometry.Manifold.VectorBundle.Basic
import Mathlib.AlgebraicTopology.SimplicialSet
import Mathlib.CategoryTheory.Sites.Sheaf

/-- A $U(1)$-gerbe with connective structure and curving on $M$:
    Cech-Deligne cocycle on a good open cover. -/
structure GerbeWithCurving (M : Type*) [SmoothManifoldWithCorners ℝ M]
    (𝒰 : OpenCover M) where
  curving       : ∀ α, DifferentialForm 2 (𝒰.U α)
  connective    : ∀ α β, DifferentialForm 1 (𝒰.U α ∩ 𝒰.U β)
  band          : ∀ α β γ, (𝒰.U α ∩ 𝒰.U β ∩ 𝒰.U γ) → Circle
  descent_g     : ∀ α β γ δ x,
                    band β γ δ x * (band α γ δ x)⁻¹ * band α β δ x * (band α β γ x)⁻¹ = 1
  descent_A     : ∀ α β γ,
                    connective α β + connective β γ + connective γ α
                      = -Complex.I • (band α β γ)⁻¹ • d (band α β γ)
  descent_B     : ∀ α β, curving β - curving α = d (connective α β)

/-- Field strength of a gerbe with curving: globally well-defined 3-form. -/
def fieldStrength (G : GerbeWithCurving M 𝒰) : DifferentialForm 3 M :=
  -- glue $d B_\alpha$ across patches using descent_B
  sorry

/-- Dixmier-Douady class: the cohomology class of the band cocycle in $H^3(M, \mathbb{Z})$. -/
def dixmierDouady (G : GerbeWithCurving M 𝒰) : H³ M ℤ :=
  sorry  -- exponential SES + connecting homomorphism on the band

/-- Brylinski's classification theorem: gerbes-with-curving up to equivalence
    biject with Deligne hypercohomology $H^3(M, \mathbb{Z}(3)_D^\infty)$. -/
theorem gerbe_classification (M : Type*) [SmoothManifoldWithCorners ℝ M] :
    Equiv (Quotient (GerbeWithCurving.EquivSetoid M)) (DeligneCohomology M 3 3) :=
  sorry  -- Brylinski 1993 Theorem 5.3.4

/-- Worldsheet holonomy: for a closed surface $\Sigma$ with map $\phi : \Sigma \to M$,
    the gerbe holonomy is a phase. -/
def worldsheetHolonomy (G : GerbeWithCurving M 𝒰)
    (Σ : ClosedSurface) (φ : Σ → M) : Circle :=
  sorry  -- glue $\exp(2\pi i \int B)$ across patches using $A$ and band data

The proof gap is substantive. The required infrastructure includes: a higher-stack model for gerbes (either Brylinski sheaves of groupoids on $M$ with band $\underline{U (1)}_{M}$ or Murray bundle gerbes $L \to Y^{[2]}$ over a submersion); the smooth Deligne complex as a complex of sheaves on $M$ ; hypercohomology of complexes of sheaves with Čech computation; the Dixmier-Douady class as a connecting homomorphism in the exponential short exact sequence; the worldsheet-holonomy assignment as a global section of a circle bundle over the mapping space; Brylinski's classification theorem; and the level- $k$ WZW gerbe on $SU (2)$ as the worked example. Each ingredient is formalisable from existing Mathlib differential geometry, sheaf theory, and category theory, plus a higher-category layer not yet in Mathlib.

Advanced results [Master]

Theorem (Brylinski 1993 Theorem 5.3.4). The functor sending a smooth manifold $M$ to the category of $U (1)$ -gerbes with connective structure and curving on $M$ , modulo equivalence, computes the smooth Deligne hypercohomology $H^{3} (M, Z (3)_{D}^{\infty})$ as a contravariant functor in $M$ . The forgetful map to the Dixmier-Douady class $H^{3} (M, Z)$ and the curvature map to closed 3-forms with integral periods $Ω_{Z}^{3} (M)$ fit into the differential-cohomology exact sequence $$ 0 \to H^2(M, \mathbb{R}/\mathbb{Z}) \to H^3(M, \mathbb{Z}(3)D^\infty) \to \Omega^3\mathbb{Z}(M) \times_{H^3(M, \mathbb{R})} H^3(M, \mathbb{Z}) \to 0. $$

This is the foundational classification. The kernel $H^{2} (M, R / Z)$ records flat gerbes, whose 3-curvature vanishes but whose Dixmier-Douady class can carry torsion. On a simply-connected 4-manifold with $H_{2} = 0$ , this kernel vanishes and the gerbe is determined by its 3-curvature plus its integer flux.

Theorem (worldsheet holonomy as a gerbe section; Carey-Johnson-Murray 2004). For a closed oriented surface $Σ$ and a smooth map $ϕ : Σ \to M$ , the pullback $\phi^ \mathcal{G} $o f a$ U(1) $- g er b e - w i t h - c u r v in g o n$ M $i sc an o ni c a l l y t r i v ia l i s ab l e, an d t h e t r i v ia l i s a t i o ni s a$ U(1)$-torsor. The worldsheet holonomy* $$ \mathrm{hol}\mathcal{G}(\phi) = \exp(2\pi i \int\Sigma \phi^* B) \in U(1) $$ is the canonical generator of this $U (1)$ -torsor, well-defined up to the global gauge ambiguity of the gerbe and modulo $2 π Z$ in the exponent.

The integral $\int_{Σ} ϕ^{*} B$ does not make literal sense as a single integral of a globally-defined 2-form — the $B$ -field is only locally a 2-form. The construction proceeds by triangulating $Σ$ adapted to a good cover of $M$ , integrating $B_{α}$ on each face mapped into a patch $U_{α}$ , summing the resulting numbers, and correcting by integrals of $A_{α β}$ on edges between faces and by the values of $g_{α β γ}$ on vertices between three faces. The descent equations (i)-(iii) ensure that the exponentiated total is independent of the choice of triangulation and adapted cover.

Theorem (Wess-Zumino-Witten action; Witten 1984, Gawȩdzki 1987). Let $G$ be a compact, simply-connected, simple Lie group and let $k \in Z$ . The level- $k$ $U (1)$ -gerbe on $G$ has Dixmier-Douady class equal to $k$ times the generator of $H^{3} (G, Z) = Z$ . For a smooth map $ϕ : Σ \to G$ from a closed Riemann surface $Σ$ , the level- $k$ WZW action $$ S_{WZW}^{(k)}[\phi] = \frac{k}{4\pi} \int_\Sigma \mathrm{tr}(\phi^{-1} d\phi \wedge \star \phi^{-1} d\phi) + 2\pi k \int_\Sigma \phi^* B $$ is well-defined modulo $2 π Z$ when $B$ is the level-1 gerbe's curving. The exponential $exp (i S_{W Z W}^{(k)})$ is therefore a single-valued function of $ϕ$ .

The WZW action is the historical exhibit of why the gerbe interpretation matters. Witten's original 1984 paper observed that the topological term $\int_{B} ϕ^{*} H$ requires choosing a 3-manifold $B$ with $\partial B = Σ$ and an extension of $ϕ$ over $B$ , and that the action is well-defined modulo $2 π Z$ only when $k$ is an integer. Gawȩdzki later recognised this as the holonomy of the level- $k$ gerbe on $G$ . The two computations agree because the gerbe holonomy on a closed surface $Σ = \partial B$ equals $exp (i \int_{B} ϕ^{*} H)$ by Stokes, and the integer-flux quantisation of $H$ forces $k \in Z$ .

Theorem (twisted K-theory and D-brane charge; Bouwknegt-Mathai 2000, Freed-Witten 1999). Let $M$ be a spin spacetime with $B$ -field of Dixmier-Douady class $[H] \in H^{3} (M, Z)$ . The charge of a D-brane wrapping a $Spin^{c}$ submanifold $W \subset M$ with twisted Chan-Paton bundle takes values in the twisted K-theory $K^(M, [H]) $, w h er e$ K^(M, [H]) $i s d e f in e d a s t h eK - t h eor y o f t h e$ C^ $- a l g e b r a o f sec t i o n so f a co n t in u o u s - t r a ce a l g e b r ab u n d l e w i t h D i x mi er - D o u a d y c l a ss$ [H]$ (equivalently, as twisted vector bundles for the gerbe).*

For $[H] = 0$ , twisted K-theory reduces to ordinary K-theory $K^{*} (M)$ , and D-brane charges are integer-valued. For $[H] \neq = 0$ , the Freed-Witten anomaly condition $W_{3} (W) + [H ∣_{W}] = 0$ in $H^{3} (W, Z)$ forces the Chan-Paton bundle to be twisted, and the K-theoretic charge lattice is the twisted K-theory. This is the K-theoretic content of the worldsheet anomaly cancellation.

Theorem (transgression to loop space; Brylinski 1993 Ch. 6). A $U (1)$ -gerbe-with-connective-structure on $M$ canonically determines a $U (1)$ -line bundle with connection on the free loop space $L M = C^{\infty} (S^{1}, M)$ . The first Chern class of the transgressed line bundle equals the transgression $H^{3} (M, Z) \to H^{2} (L M, Z)$ of the Dixmier-Douady class along the evaluation $ev : S^{1} \times L M \to M$ followed by integration over the $S^{1}$ -fibre.

The transgressed line bundle is the loop-space realisation of the gerbe. Its sections are the states of a string in the gerbe background: a state assigns to each loop in $M$ a phase, and the phase is the value of a section of the transgressed line bundle. The construction recovers in particular the level- $k$ line bundle on the loop group $L G$ used in the geometric construction of the affine central extension $L G \to L G$ (Freed-Hopkins-Teleman 2003; Brylinski 1993 Ch. 7).

Theorem (gerbe gauge group as a 2-group; Schreiber-Waldorf 2009). The gauge group of $U (1)$ -gerbe-with-connective-structure transformations on $M$ is a Lie 2-group, with objects the line bundles on $M$ and 1-morphisms the line-bundle isomorphisms. The 1-morphism part recovers the ordinary $U (1)$ -gauge transformations $B \to B + d Λ$ ; the object part records the line-bundle ambiguity in the gauge parameter itself.

The 2-group structure is the higher analogue of an ordinary Lie group. It encodes the categorical content of the $B$ -field's gauge symmetry: not only do gauge parameters fail to commute (1-morphism non-commutativity), but their composition is itself only well-defined up to the choice of an intertwining 2-morphism (object-level ambiguity). The Lie-2-group structure is the algebraic shadow of the gerbe's classifying stack $B^{2} U (1)$ .

Synthesis. The $B$ -field's identification as a $U (1)$ -gerbe connection puts the Kalb-Ramond field in the canonical setting where its three signature features become structural rather than coincidental. The foundational reason this interpretation holds is exactly the categorical climb from line bundles to gerbes: at level $p = 2$ the smooth Deligne complex classifies $U (1)$ -line bundles with connection (03.06.08), and at level $p = 3$ it classifies $U (1)$ -gerbes with connective structure and curving — the $B$ -field. The central insight is that the descent equations (i)-(iii) of a gerbe are not a clever bookkeeping device but the only data compatible with a self-consistent worldsheet holonomy on closed surfaces, exactly as the cocycle of a line bundle is the only data compatible with a self-consistent worldline holonomy on closed loops. Putting these together, the three features Polchinski emphasised in 1995 — global existence of the $B$ -field beyond its local 2-form description, integrality of the 3-flux $\int H / (2 π) \in Z$ , and the appearance of twisted K-theory for D-brane charge — unify as the level-3 incarnation of the same gauge-theoretic pattern that places magnetic flux quantisation and Chern-class integrality at level 2. The bridge to the sister gauge theory 03.07.05 (Yang-Mills) is the recognition that both are connections in differential cohomology, one rung apart in the categorical ladder, and the bridge to 03.10.03 pending (WZW model) is that the WZW level $k$ is exactly the Dixmier-Douady class of the corresponding gerbe.

This same algebraic pattern appears again in 03.06.10 (bundle gerbes) where the Murray formulation realises the abstract Brylinski sheaf-of-groupoids picture concretely as a $U (1)$ -line bundle over a fibre product $Y^{[2]}$ of a submersion $Y \to M$ , and in 03.06.11 (transgression to loop space) where the gerbe descends to a line bundle on the loop space whose first Chern class transgresses the Dixmier-Douady class. The Freed-Hopkins-Teleman theorem then identifies the K-theoretic content of these transgressed line bundles with the Verlinde algebra of the corresponding affine Lie algebra at level $k$ — putting these together, the $B$ -field of string theory, the WZW model of two-dimensional CFT, the affine central extension of $L G$ , and the Verlinde fusion ring of the rational CFT are all aspects of the level- $k$ $U (1)$ -gerbe on $G$ and its loop-space transgression.

Full proof set [Master]

Proposition (closedness of the curvature 3-form). Let $G = (B_{α}, A_{α β}, g_{α β γ})$ be a $U (1)$ -gerbe-with-curving on $M$ . The locally defined 3-forms $d B_{α}$ on $U_{α}$ assemble into a globally defined closed 3-form $H \in Ω^{3} (M)$ .

Proof. On a double overlap $U_{α β}$ , equation (iii) gives $B_{β} - B_{α} = d A_{α β}$ . Applying $d$ to both sides, $d B_{β} - d B_{α} = d^{2} A_{α β} = 0$ , so the local 3-forms $d B_{α}$ agree on overlaps and glue to a global 3-form $H \in Ω^{3} (M)$ . Closedness $d H = 0$ follows because $H ∣_{U_{α}} = d B_{α}$ and $d (d B_{α}) = 0$ by $d^{2} = 0$ . $□$

Proposition (integrality of the 3-flux). Let $G$ be a $U (1)$ -gerbe-with-curving on $M$ with curvature 3-form $H$ and Dixmier-Douady class $DD (G) \in H^{3} (M, Z)$ . Then for every smooth closed oriented 3-cycle $Z \subset M$ , $$ \frac{1}{2\pi} \int_Z H = \langle \mathrm{DD}(\mathcal{G}), [Z] \rangle \in \mathbb{Z}, $$ where the right-hand side is the integer pairing of the integral cohomology class with the integer homology class.

Proof. The exponential short exact sequence of sheaves on $M$ , $$ 0 \to \underline{\mathbb{Z}} \to \underline{\mathbb{R}} \to \underline{U(1)} \to 0, $$ induces a long exact sequence in Čech cohomology. Restricting to degree 2, the connecting homomorphism $δ : H^{2} (M, \underline{U (1)}) \to H^{3} (M, \underline{Z}) = H^{3} (M, Z)$ sends the Čech class $[g_{α β γ}]$ to the Dixmier-Douady class. On the de Rham side, the connecting homomorphism of the same short exact sequence sends a class in $H^{2} (M, \underline{U (1)})$ represented by $g = exp (i c)$ with local lifts $c_{α β γ} : U_{α β γ} \to R$ to the de Rham class of $d c$ thought of as a closed 3-form. The Čech-de Rham isomorphism identifies this class with $[H /2 π]$ via the spectral-sequence computation of 03.04.11. So $[H /2 π] = DD (G)_{R}$ as a real cohomology class, and the integrality on cycles is the integrality of the Dixmier-Douady class. $□$

Proposition (worldsheet holonomy is well-defined). Let $Σ$ be a closed oriented surface, $ϕ : Σ \to M$ a smooth map, and $G$ a $U (1)$ -gerbe-with-curving on $M$ . The expression $$ \mathrm{hol}\mathcal{G}(\phi) = \exp\left(i, T\Sigma(\phi^* \mathcal{G})\right) \in U(1), $$ where $T_{Σ}$ is a triangulation-adapted sum of integrals of $B_{α}, A_{α β}$ and values of $ar g g_{α β γ}$ , is independent of the choice of triangulation and adapted cover, and depends on $G$ only through its gauge-equivalence class.

Proof sketch (Carey-Johnson-Murray 2004). Choose a triangulation of $Σ$ so refined that every face maps into a single patch $U_{α (f)}$ via $ϕ$ , every edge into a double overlap $U_{α (e_{+}) α (e_{-})}$ , and every vertex into a triple overlap or finer. Define $$ T_\Sigma(\phi^* \mathcal{G}) = \sum_{f} \int_f \phi^* B_{\alpha(f)} + \sum_{e} \int_e \phi^* A_{\alpha(e_+) \alpha(e_-)} + \sum_{v} \arg g_{\alpha(v_1) \alpha(v_2) \alpha(v_3)}(\phi(v)), $$ with signs determined by the orientation of $Σ$ . To show this is well-defined modulo $2 π Z$ , one verifies invariance under (a) refinement of the triangulation, (b) change of adapted cover assignment, (c) gauge transformation of the gerbe. Each invariance is a direct computation using the descent equations (i)-(iii): for (a), Stokes' theorem applied to a refining bisection of a face introduces an edge integral and a vertex term that cancel via equations (ii) and (iii); for (b), changing the patch assignment on a face introduces an edge integral via (iii) and is absorbed; for (c), the gauge action on the cocycle is matched by a coboundary on the triangulation sum. The result is independent modulo $2 π Z$ . $□$

Proposition (WZW level quantisation). Let $G$ be a compact, simply-connected, simple Lie group with bi-invariant generator $ω \in H^{3} (G, Z) = Z$ normalised so that $\int_{G} ω = 1$ on the fundamental class. For $k \in Z$ , the level- $k$ gerbe on $G$ has curvature $H_{k} = 2 π k ω$ in de Rham cohomology, and the WZW action $S_{W Z W}^{(k)} [ϕ]$ is well-defined as a function $Σ \to U (1)$ via the worldsheet holonomy if and only if $k$ is an integer.

Proof. The level- $k$ gerbe is by definition the gerbe with Dixmier-Douady class $k$ times the generator of $H^{3} (G, Z)$ . By the integrality proposition, its curvature satisfies $[H_{k} /2 π] = k ω$ in $H^{3} (G, R)$ , so $H_{k} = 2 π k ω$ as a de Rham class.

For $Σ$ a closed Riemann surface and $ϕ : Σ \to G$ , the worldsheet holonomy is well-defined whenever the gerbe data is gauge-equivalent and globally defined on $G$ , which holds for integer $k$ . For non-integer $k$ , the descent equations would require $g_{α β γ}$ to take values in $U (1)^{k}$ rather than $U (1)$ — i.e. a multi-valued band — and the cocycle condition fails on quadruple overlaps. Equivalently, the bounding-3-manifold argument shows that two extensions $ϕ : B \to G$ and $ϕ^{'} : B^{'} \to G$ of $ϕ : Σ \to G$ with $\partial B = \partial B^{'} = Σ$ contribute integrals over the closed 3-cycle $B - B^{'}$ whose value is $2 π k \cdot n$ with $n \in Z$ ; the action is well-defined modulo $2 π Z$ when and only when $k \in Z$ . $□$

Proposition (D-brane charge in twisted K-theory; Freed-Witten anomaly). Let $M$ be a spin manifold with $U (1)$ -gerbe $G$ of Dixmier-Douady class $[H]$ . The anomaly cancellation condition for a D-brane wrapping a $Spin^{c}$ submanifold $W \subset M$ with Chan-Paton bundle $E$ reads $$ W_3(W) + [H|_W] = 0 \quad \in H^3(W, \mathbb{Z}), $$ where $W_{3}$ is the third integral Stiefel-Whitney class.

Proof sketch. The worldsheet partition function for a D-brane wrapping $W$ contains a fermion determinant on the boundary $\partial Σ \subset W$ of the open string. The determinant is a section of a line bundle on the boundary loop space $L W$ , and its global existence requires that line bundle to be trivialisable. The first Chern class of the determinant line bundle decomposes as a sum of the third integral Stiefel-Whitney class $W_{3} (W)$ of the worldvolume (from the spin structure on $\partial Σ$ ) and the restriction of the bulk gerbe class $[H ∣_{W}]$ (from the $B$ -field's coupling to the worldsheet). Triviality is the condition $W_{3} (W) + [H ∣_{W}] = 0$ . The Chan-Paton bundle $E$ on $W$ must then be a twisted bundle of class $- W_{3} (W) = [H ∣_{W}]$ , i.e. an element of twisted K-theory $K^{0} (W, [H ∣_{W}])$ . Pushforward along $ι : W ↪ M$ via the Spin^c-twisted Gysin map lands in $K^{*} (M, [H])$ . $□$

Connections [Master]

Dixmier-Douady class 03.06.09. The Dixmier-Douady class is the topological obstruction whose presence forces the $B$ -field to be a gerbe connection rather than a globally-defined 2-form. The level- $k$ class on $SU (2) ≅ S^{3}$ is the integer $k$ in $H^{3} (S^{3}, Z) = Z$ ; the level- $k$ gerbe is the canonical realisation, and the WZW action picks up its global well-definedness from the integrality of this class. Without the Dixmier-Douady classification, the $B$ -field could only be discussed in patches; the gerbe interpretation supplies the global organising data.
Kostant-Weil isomorphism 03.06.08. The Kostant-Weil isomorphism classifies $U (1)$ -line bundles with connection by $H^{2} (M, Z (2)_{D}^{\infty})$ — the level-2 smooth Deligne hypercohomology. The $B$ -field's classification by $H^{3} (M, Z (3)_{D}^{\infty})$ is the level-3 analogue, and the pattern continues for higher $p$ -form gauge fields. The abelian-line-bundle case is the prototype: ordinary electromagnetism is to the magnetic monopole as the $B$ -field is to the NS5-brane.
Yang-Mills action 03.07.05. Yang-Mills is the sister gauge theory: a connection on a principal $G$ -bundle whose curvature $F$ is a 2-form valued in the Lie algebra, with the action $\int ∣ F ∣^{2}$ and the field equation $d_{A}^{*} F = 0$ . The $B$ -field is the abelian higher analogue: a connection on a $U (1)$ -gerbe whose curvature $H$ is a 3-form, with action $\int ∣ H ∣^{2}$ in supergravity and the topological coupling $\int B$ on the string worldsheet. The two theories sit one categorical rung apart and share the differential-cohomology framework.
Bundle gerbe 03.06.10. Murray's bundle-gerbe formulation provides a concrete realisation of Brylinski's abstract sheaf-of-groupoids picture. A bundle gerbe is a $U (1)$ -line bundle $L \to Y^{[2]}$ over the fibre product of a submersion $Y \to M$ , with a multiplicative structure on triple fibre products. The $B$ -field's curving 2-form is a 2-form on $Y$ , and the descent equations are encoded in the multiplicativity of $L$ . This is the formulation used in the holonomy proof and in most modern computational work on the $B$ -field.
WZW action and the level- $k$ extension 03.10.03 pending. The Wess-Zumino-Witten model is the historically central example: the $B$ -field on the target Lie group $G$ at level $k$ has Dixmier-Douady class equal to $k$ times the generator of $H^{3} (G, Z)$ , and the WZW action is the worldsheet holonomy of this gerbe. The level- $k$ central extension of the loop group $L G$ is the loop-space transgression of the level- $k$ gerbe. The integrality $k \in Z$ of the WZW level is exactly the integrality of the Dixmier-Douady class.
Transgression to loop space 03.06.11. A $U (1)$ -gerbe-with-curving on $M$ canonically determines a $U (1)$ -line bundle with connection on the free loop space $L M$ , with first Chern class equal to the transgressed Dixmier-Douady class. For $M = G$ a compact simply-connected simple Lie group, the transgressed line bundle on $L G$ is the level- $k$ line bundle used in the geometric construction of the affine central extension $L G \to L G$ .
Twisted K-theory and Freed-Hopkins-Teleman 03.06.12. D-brane charges in a spacetime with $B$ -field take values in twisted K-theory $K^{*} (M, [H])$ . For $M = G$ a compact Lie group with the level- $k$ gerbe, the Freed-Hopkins-Teleman theorem identifies the equivariant twisted K-theory $K_{G}^{*} (G)_{[H_{k}]}$ with the Verlinde ring of the affine Lie algebra $g_{k}$ . The $B$ -field of string theory is the same object that classifies the fusion ring of a rational CFT.
Chern-Weil homomorphism 03.06.06. The Chern-Weil construction produces characteristic classes of a principal-bundle connection from invariant polynomials of its curvature. The gerbe analogue produces the de Rham class of $H$ from the curving 2-form, with no invariant-polynomial step required because the gauge group is abelian. The pattern of "characteristic class from curvature" is preserved one categorical level up: $c_{1} (L) = [F /2 π]$ for a line bundle becomes $DD (G) = [H /2 π]$ for a $U (1)$ -gerbe.

Historical & philosophical context [Master]

The 2-form gauge field $B_{μν}$ entered theoretical physics through the 1974 paper of Michael Kalb and Pierre Ramond, Classical Direct Interstring Action (Phys. Rev. D 9, 2273-2284) ^{[Kalb-Ramond 1974]}, where it was introduced as the mediator of a direct interaction between fundamental strings — the string analogue of the electromagnetic potential's role in mediating direct charge-charge interaction. The Kalb-Ramond paper treated $B$ as a globally-defined antisymmetric tensor field on Minkowski space, and the gauge symmetry $B \to B + d Λ$ was a 1-form gauge symmetry exactly parallel to the $U (1)$ gauge symmetry of electromagnetism. Globally non-flat spacetimes and the implications of integrality of $\int H / (2 π)$ over closed 3-cycles were not yet considered.

The connection to Wess-Zumino-Witten theory was made by Edward Witten in 1983-84. In Non-abelian bosonization in two dimensions (Comm. Math. Phys. 92 (1984), 455-472) ^{[Witten 1984]}, Witten introduced the topological term $(k /24 π^{2}) \int_{B} tr ((g^{- 1} d g)^{3})$ for a smooth map $g : Σ \to G$ extended over a 3-manifold $B$ with $\partial B = Σ$ , and observed that this term is well-defined modulo $2 π Z$ exactly when the level $k$ is an integer. The mechanism — integer flux quantisation of a 3-form over closed 3-cycles — is the same as the integrality of the Dixmier-Douady class, but the gerbe interpretation was not yet articulated. Krzysztof Gawȩdzki in Topological actions in two-dimensional quantum field theories (Cargèse 1987 lectures, NATO ASI Ser. B 185 (1988), 101-141) ^{[Gawedzki 1987]} first recognised that the WZ term is the holonomy of a higher-degree gauge field — the seed of the gerbe identification.

The mathematical machinery of gerbes appeared in parallel in Jean-Luc Brylinski's 1993 monograph Loop Spaces, Characteristic Classes and Geometric Quantization (Progress in Mathematics 107, Birkhäuser) ^{[Brylinski 1993 §5]}, which systematised the smooth Deligne complex, sheaves of groupoids with band, and the classification of $U (1)$ -gerbes by $H^{3} (M, Z)$ . Brylinski's Chapter 5 contains the definitive treatment of gerbe-with-connective-structure-and-curving and the foundational classification theorem. Michael Murray's 1996 paper Bundle gerbes (J. London Math. Soc. (2) 54, 403-416) ^{[Murray 1996]} provided the concrete bundle-gerbe formulation as a $U (1)$ -line bundle over a fibre product of a submersion, equivalent to Brylinski's sheaf-of-groupoids picture and computationally more tractable for explicit holonomy and curvature calculations. Nigel Hitchin's 1999-2001 lectures Lectures on Special Lagrangian Submanifolds ^{[Hitchin 2001]} communicated the gerbe-and- $B$ -field identification to the string-theory and mirror-symmetry communities.

The D-brane reading of the gerbe came through Joseph Polchinski's 1995 discovery in Dirichlet branes and Ramond-Ramond charges (Phys. Rev. Lett. 75, 4724-4727) ^{[Polchinski 1995]} that D-branes carry Ramond-Ramond charge, followed by Daniel Freed and Edward Witten's 1999 anomaly-cancellation analysis identifying the required twist of the Chan-Paton bundle. Peter Bouwknegt and Varghese Mathai's D-branes, B-fields and twisted K-theory (J. High Energy Phys. 03 (2000), 007) ^{[Bouwknegt-Mathai 2000]} consolidated the result: D-brane charge in a spacetime with $B$ -field of Dixmier-Douady class $[H]$ lives in twisted K-theory $K^{*} (M, [H])$ . The Freed-Hopkins-Teleman theorems of 2003, 2007, and 2011 ^{[Freed-Hopkins-Teleman 2003]} identified the equivariant twisted K-theory of a compact Lie group at level $k$ with the Verlinde ring of the corresponding affine Lie algebra, weaving the $B$ -field of string theory together with the WZW model of rational CFT and the representation theory of loop groups into a single theorem.

Bibliography [Master]

@article{KalbRamond1974,
  author  = {Kalb, Michael and Ramond, Pierre},
  title   = {Classical Direct Interstring Action},
  journal = {Phys. Rev. D},
  volume  = {9},
  year    = {1974},
  pages   = {2273--2284}
}

@book{Brylinski1993,
  author    = {Brylinski, Jean-Luc},
  title     = {Loop Spaces, Characteristic Classes and Geometric Quantization},
  series    = {Progress in Mathematics},
  volume    = {107},
  publisher = {Birkh\"auser},
  year      = {1993}
}

@article{Murray1996,
  author  = {Murray, Michael K.},
  title   = {Bundle gerbes},
  journal = {J. London Math. Soc. (2)},
  volume  = {54},
  year    = {1996},
  pages   = {403--416}
}

@incollection{Hitchin2001,
  author    = {Hitchin, Nigel},
  title     = {Lectures on Special Lagrangian Submanifolds},
  booktitle = {Winter School on Mirror Symmetry, Vector Bundles and Lagrangian Submanifolds},
  series    = {AMS/IP Stud. Adv. Math.},
  volume    = {23},
  publisher = {Amer. Math. Soc. and International Press},
  year      = {2001},
  pages     = {151--182}
}

@article{Witten1984,
  author  = {Witten, Edward},
  title   = {Non-abelian bosonization in two dimensions},
  journal = {Comm. Math. Phys.},
  volume  = {92},
  year    = {1984},
  pages   = {455--472}
}

@incollection{Gawedzki1987,
  author    = {Gaw\k{e}dzki, Krzysztof},
  title     = {Topological actions in two-dimensional quantum field theories},
  booktitle = {Non-perturbative quantum field theory (Carg\`ese, 1987)},
  series    = {NATO Adv. Sci. Inst. Ser. B Phys.},
  volume    = {185},
  publisher = {Plenum},
  year      = {1988},
  pages     = {101--141}
}

@article{Polchinski1995,
  author  = {Polchinski, Joseph},
  title   = {Dirichlet branes and Ramond-Ramond charges},
  journal = {Phys. Rev. Lett.},
  volume  = {75},
  year    = {1995},
  pages   = {4724--4727}
}

@article{FreedWitten1999,
  author  = {Freed, Daniel S. and Witten, Edward},
  title   = {Anomalies in string theory with {D}-branes},
  journal = {Asian J. Math.},
  volume  = {3},
  year    = {1999},
  pages   = {819--851}
}

@article{BouwknegtMathai2000,
  author  = {Bouwknegt, Peter and Mathai, Varghese},
  title   = {{D}-branes, {B}-fields and twisted {K}-theory},
  journal = {J. High Energy Phys.},
  volume  = {03},
  year    = {2000},
  pages   = {007}
}

@article{FreedHopkinsTeleman2011,
  author  = {Freed, Daniel S. and Hopkins, Michael J. and Teleman, Constantin},
  title   = {Loop groups and twisted {K}-theory {III}},
  journal = {Ann. of Math. (2)},
  volume  = {174},
  year    = {2011},
  pages   = {947--1007}
}

@article{CareyJohnsonMurray2004,
  author  = {Carey, Alan L. and Johnson, Stuart and Murray, Michael K.},
  title   = {Holonomy on {D}-branes},
  journal = {J. Geom. Phys.},
  volume  = {52},
  year    = {2004},
  pages   = {186--216}
}

@book{Polchinski1998,
  author    = {Polchinski, Joseph},
  title     = {String Theory},
  publisher = {Cambridge University Press},
  year      = {1998}
}

@article{SchreiberWaldorf2009,
  author  = {Schreiber, Urs and Waldorf, Konrad},
  title   = {Smooth functors vs.~differential forms},
  journal = {Homology Homotopy Appl.},
  volume  = {12},
  year    = {2010},
  pages   = {279--315}
}

Prerequisites

03.05.01
03.05.07
03.04.02
03.04.06
03.04.11
03.04.14
03.06.06
03.07.05

Tier anchors

beginner: Polchinski *String Theory* Vol. 1 §3.7 informal account of the Kalb-Ramond field on the worldsheet; Hitchin 1999 *Quart. J. Math.* introductory section on gerbes
intermediate: Brylinski *Loop Spaces, Characteristic Classes and Geometric Quantization* (Birkhäuser 1993) §5; Hitchin 1999 *Lectures on Special Lagrangian Submanifolds* §1.4; Bouwknegt-Mathai 2000 *J. High Energy Phys.* 03:007
master: Brylinski *Loop Spaces, Characteristic Classes and Geometric Quantization* (Birkhäuser 1993) §5 (gerbe connective and curving structures); Murray 1996 *J. London Math. Soc.* 54, 403-416 (bundle gerbe formulation); Freed-Hopkins-Teleman 2003/2007/2011 (gerbes and loop groups); Bouwknegt-Mathai 2000 *J. High Energy Phys.* 03:007 (twisted K-theory of D-branes); Polchinski 1995 *Phys. Rev. Lett.* 75, 4724-4727 (D-brane charge)

References

TODO_REF
Kalb-Ramond — Classical Direct Interstring Action · Phys. Rev. D 9 (1974), 2273-2284; originator of the antisymmetric 2-form gauge field on the worldsheet
TODO_REF
Brylinski — Loop Spaces, Characteristic Classes and Geometric Quantization · Progress in Mathematics 107, Birkhäuser 1993; §5 sheaves of groupoids with band, connective structures, curving, and the classification of $U(1)$-gerbes by $H^3(M, \mathbb{Z})$
TODO_REF
Murray — Bundle gerbes · J. London Math. Soc. (2) 54 (1996), 403-416; bundle-gerbe formulation equivalent to Brylinski's sheaf-of-groupoids picture
TODO_REF
Hitchin — Lectures on Special Lagrangian Submanifolds · in *Winter School on Mirror Symmetry, Vector Bundles and Lagrangian Submanifolds*, AMS/IP 23 (2001), 151-182; gerbes-with-connection viewpoint on the Kalb-Ramond field
TODO_REF
Polchinski — Dirichlet branes and Ramond-Ramond charges · Phys. Rev. Lett. 75 (1995), 4724-4727; D-branes as carriers of Ramond-Ramond charge, the trigger for twisted K-theory
TODO_REF
Bouwknegt-Mathai — D-branes, B-fields and twisted K-theory · J. High Energy Phys. 03 (2000), 007; identification of D-brane charge in twisted K-theory $K^*(M, [H])$
TODO_REF
Freed-Hopkins-Teleman — Loop groups and twisted K-theory I-III · J. Topology 4 (2011), 737-798 + companion papers 2007, 2003; gerbe twists of K-theory and the FHT theorem $K^*_G(G) = \mathrm{Ver}_k(G)$
TODO_REF
Witten — Non-abelian bosonization in two dimensions · Comm. Math. Phys. 92 (1984), 455-472; the Wess-Zumino-Witten action and quantisation $k \in \mathbb{Z}$
TODO_REF
Gawedzki — Topological actions in two-dimensional quantum field theories · in *Non-perturbative quantum field theory* (Cargèse 1987), NATO ASI Ser. B 185, 101-141; gerbe interpretation of the WZW topological term
TODO_REF
Carey-Johnson-Murray — Holonomy on D-branes · J. Geom. Phys. 52 (2004), 186-216; bundle-gerbe holonomy as worldsheet $\exp(2\pi i \int_\Sigma B)$ globally

Reviewer

TBD

Estimated time

beginner: 18m
intermediate: 40m
master: 75m