{"id":"4277e169-6413-4f59-831c-704ad83a73a0","arxiv_id":"hep-th/0406102","paper_version":3,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"T-fold string backgrounds, glued by T-duality transition functions, can be reformulated as conventional geometry on a doubled T^{2n} fibration in which physical spacetime is a locally chosen null T^n slice.","lead":"String theory may be consistent on 'T-fold' backgrounds whose patches are glued by T-duality rather than by ordinary coordinate changes, so no single spacetime manifold exists globally. Hull shows such backgrounds can be rewritten as ordinary geometry on a doubled torus, with our spacetime appearing as a slice of it.","discovery_kind":"unclear","skeptic_critique":{"model":"selfhosted/deepseek-v4.1-flash","headline":"The load-bearing gap is quantum, not classical: the claim that physics is independent of polarization (§4) requires an O(n,n;Z)-invariant chiral-boson measure, which §8 concedes is unresolved. The doubled geometry is sound; the 'string formulation' is not yet demonstrated.","rationale":"The reader's weakest_assumption identifies the right load-bearing concern: the construction is a classical sigma-model statement, and its claim to describe strings rests on the self-duality constraint being imposable consistently in the quantum theory. The paper acknowledges this in §8, so it is not a new finding; what remains unresolved is the precise quantum mechanism, and the abstract's 'formulated' is stronger than §8 supports. I would add one precision: the claim at the end of §4 that physics is independent of polarization is exactly the statement that the chiral-boson measure is O(n,n;Z)-invariant, which is the sharp form of the quantization problem and is not demonstrated. On the classical side, the paper holds up well. The O(n,n;Z) covariance of (2.5) is straightforward to check; the product-structure formalism of §5 is a correct characterization of maximal isotropic splittings; and the §6 argument that local T^n slices fail to glue is a standard structure-group non-reducibility statement, more rigorous than the reader credits. The n=1 Buscher check in §3.1 is a real external benchmark. The reader's second criticism — that §6 computes no obstruction and gives no worked example — is partly answered by §3.2, which is a worked n=1 example of the failure to glue, and by the fact that the obstruction is the non-reducibility of O(n,n;Z) to GL(n;Z), which is a standard topological statement. The general-n Buscher rules in §2.5 are asserted rather than derived, but they are cited from [15-17], so per the review rules that is not a load-bearing concern. The paper's honesty in §8 and footnote 1 is a point in its favor. My read does not change the verdict: ACCEPT with MODERATE confidence is appropriate, with the caveat that the abstract should be read as a classical formulation claim rather than a demonstrated quantum one. The concrete test above would settle whether the polarization-independence claim survives at one loop.","tokens_in":25306,"tokens_out":10435,"duration_ms":206805,"concrete_test":"For the n=1 T-fold of §3.2 (circle of radius R with monodromy R -> 1/R, equivalently the asymmetric orbifold of [12,13]), compute the one-loop torus partition function in the doubled formalism with the constraint (2.5) imposed via one of the cited schemes (e.g. Floreanini-Jackiw as in Tseytlin [31,32]), including the measure-induced dilaton shift. Compare it to the known partition function of the corresponding T-fold CFT. Agreement, including the dilaton shift and modular invariance, would establish polarization-independence at one loop; a mismatch in the phase or in the dilaton shift would show the concern lands.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's classical geometry is internally coherent: the self-duality constraint (2.5) is O(n,n)-covariant, the product-structure characterization of polarizations in §5 is a clean bijection, and the §6 topological argument (the structure group O(n,n;Z) does not reduce to GL(n;Z), so no global null T^n sub-bundle exists) is essentially the standard non-reducibility statement, not merely qualitative. The n=1 Buscher check in §3.1 is a genuine external benchmark. The load-bearing concern is the step from this classical rewriting to the claim that 'strings in T-fold backgrounds are formulated' in the enlarged space (abstract:s2). The doubled lagrangian L_d is classically O(n,n)-invariant, and §2.4 shows classical equivalence to the standard sigma model by solving the constraint to eliminate ~X. But the interpretive claim at the end of §4 — that the physics is independent of the choice of polarization — is a quantum statement. It requires that the path-integral measure for the self-dual chiral bosons be invariant under O(n,n;Z) changes of polarization, including the measure-induced dilaton shift (1.5). The paper asserts this and cites Tseytlin [31,32] for the measure transformation, but does not demonstrate it in the T-fold setting where no global polarization exists. §8 acknowledges the quantization is open and points to competing chiral-boson schemes [32-40], noting that conformal and Lorentz anomalies must cancel and modular invariance is an additional constraint. So the concern is not that the classical construction is wrong; it is that the central physical claim is a quantum claim the paper does not establish. The n=1 check verifies the classical Buscher rules, not the quantum measure.","agreement_with_reader":"agree"},"referee_report":{"model":"selfhosted/deepseek-v4.1-flash","summary":"The manuscript proposes a formulation of string theory on 'T-fold' backgrounds, i.e. backgrounds whose local n-torus fibrations are glued by transition functions in O(n,n;Z) rather than by diffeomorphisms of T^n. The construction replaces the T^n bundle over the base N by a T^{2n} bundle ~M with structure group O(n,n;Z), carrying the O(n,n)-invariant metric 2 dX^i dX~_i, and imposes a covariant self-duality constraint (2.5) on the world-sheet one-forms P^I that halves the doubled degrees of freedom. Section 2 shows that on choosing a polarization the doubled lagrangian reduces to a conventional sigma model with metric G and B-field (2.45), (2.49)-(2.50); section 3 verifies this explicitly for n=1 and recovers the Buscher rules (3.10)-(3.11); section 4 shows that T-duality can be viewed either as an active transformation of (H,J) at fixed polarization or as a passive change of polarization at fixed (H,J); section 5 characterizes polarizations as integral product structures R with R^2=1, pseudo-hermitian with respect to L; section 6 gives the topological criterion distinguishing geometric from non-geometric backgrounds (existence of a global null T^n sub-bundle of ~M); and section 7 extends the discussion to open strings and D-branes, where a Dp-brane in one patch can appear as a Dp'-brane in another. Section 8 discusses quantization, U-duality generalisations, and relations to F-theory.","tokens_in":25680,"tokens_out":5640,"duration_ms":116524,"significance":"If the construction holds, it converts a qualitative statement ('this background has no global geometry') into a precise one ('the local null T^n subspaces of the T^{2n} bundle do not glue into a global sub-bundle'), and it makes T-duality a statement about the choice of physical subspace rather than about a transformation of the background. That is a genuine conceptual advance and it organizes a large body of earlier examples (duality-twisted reductions, T-duals of flux compactifications, asymmetric orbifolds, mirrors of T^3-fibrations) under one geometric framework. The paper also earns credit for explicit internal checks rather than assertion: the n=1 case in section 3.1 is worked through and reproduces the standard Buscher rules (3.10)-(3.11), the reduction of the doubled lagrangian to the conventional sigma model is given in equations (2.45)-(2.50), and the product-structure characterization of polarizations in section 5 is a clean bijection rather than a heuristic. The construction contains no adjustable parameters and no fitted quantities; the only inputs are the O(n,n;Z) transition data and the coset form of H. The D-brane discussion in section 7 yields a concrete, check-1-","major_comments":[{"comment":"The list of objects that are declared tensors on the doubled bundle ~M in (6.3)-(6.4) covers P, L, H and J, but not the dilaton. This matters because Eq. (1.5) shows that under an O(n,n;Z) transition function e^Φ picks up a factor (det G'/det G)^{1/4}, and det G is polarization-dependent, so Φ as defined is not a function on ~M. Since the abstract's central claim is that 'the global structure involves the doubled geometry', the paper should state explicitly what the dilaton is on ~M: for instance a section of a line bundle whose transition function is that determinant factor, or a shifted dilaton absorbing a determinant of the fibre metric, or a function on N only with the determinant factor attributed to the measure. As written, the reader cannot tell whether the doubled space carries the full background data or only the (H,J) part of it, and this is load-bearing for the global claim.","section":"§6, Eqs. (6.3)-(6.4); §8","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The equivalence between the constrained doubled model and the standard sigma model is the bridge on which everything else rests, and it is asserted 'by an argument similar to that given in [27]' rather than shown. In particular, the claim that the Y-equations of the doubled lagrangian (2.3) with the constraint (2.5) reproduce the Y-equations of (2.48) with the identifications (2.49)-(2.50) is not verified in the text. A short explicit check, or at least a statement of which field equations were compared and how the k, v terms of (2.46) enter, would make the equivalence verifiable by the reader. The explicit n=1 treatment in section 3.1 partly compensates, but the general-n step is the one that carries the paper's claim.","section":"§2.4, Eqs. (2.45)-(2.50)","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The abstract states that strings in T-fold backgrounds 'are formulated' in the enlarged space, whereas section 8 is explicit that the quantization of the constrained doubled model is open and that competing chiral-boson schemes [32-40] must still be shown to give a consistent, modular-invariant theory. The authors acknowledge this limitation, so the request is only that the abstract signal the classical/local character of the construction, for instance by saying that the backgrounds are 'described' or 'given a geometric formulation' locally.","section":"Abstract","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The final sentence reads 'the physical D-brane is the part of the part of the physical space lying in the generalised D-brane subspace'; the duplicated 'part of the' should be removed.","section":"Abstract","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The symbol Φ is used both for the coset 'vielbein' converting O(n,n) indices to GL(n) indices in (4.4) and for the dilaton throughout. The text flags the clash parenthetically, but a distinct symbol for the coset representative (for example V or P, with the dilaton reserved for Φ) would remove an ambiguity that recurs in sections 4, 5 and 8.","section":"§4, Eq. (4.4) and §8","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The active/passive discussion describes the n=1 T-duality as 'rotating the vector through π/2' or 'rotating the axes through -π/2'. For n=1 the relevant group O(1,1;Z) is discrete, so the rotation language is only heuristic; saying that the discrete T-duality exchanges the two cycles of T^2, and that the two viewpoints differ in whether the exchange is applied to the geometry or to the polarization, would be more accurate without losing the intuition.","section":"§3.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The boundary conditions (7.1)-(7.2), the vanishing of the boundary term via (7.3), and the compatibility condition (7.4) with the self-duality constraint are stated without derivation, and the step from (7.4) to the normal form (7.5) is asserted. A sentence or two explaining why (7.4) is the correct compatibility condition, and why it forces the block form (7.5), would make section 7 self-contained.","section":"§7, Eqs. (7.1)-(7.5)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The phrase 'B is a gerb-connection' should read 'gerbe connection'; the same paragraph also uses 'gerbe connection' correctly, so the two spellings should be harmonized.","section":"§1, p. 3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"There are typographical errors that should be corrected: 'funcitons' in the sentence following (6.6), 'bane' in the sentence introducing the Dp_Y-brane in section 7, and 'somewehere' in the discussion of U-folds in section 8.","section":"§6 and §8","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Reference [30] (Floreanini and Jackiw, Phys. Rev. Lett. 59 (1987) 1873) duplicates reference [36] exactly; the two should be consolidated. References [23], [24] and [39] do not appear to be cited anywhere in the text and should either be cited or removed.","section":"References","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The world-sheet is taken flat at (2.3) and the Fradkin-Tseytlin term is deferred to section 8. A forward reference at (2.3) would spare the reader the impression that the coupling to world-sheet curvature has been overlooked, particularly since the dilaton transformation (1.5) is a measure effect.","section":"§2.1, Eq. (2.3)","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"minor_revision","confidential_remarks":"This is a foundational contribution that will be cited widely, and the classical construction is carefully cross-checked against the Buscher rules and against the standard sigma model. The revision requested is limited: the dilaton's status on the doubled space should be stated, and the general-n equivalence argument in section 2.4 should be made explicit. Neither point threatens the central claim, and I would not send the paper back for further refereeing if the authors address them in a response and a short added paragraph."},"author_rebuttal":null,"desk_editor":{"model":"selfhosted/deepseek-v4.1-flash","letter":"The thing to take away: Hull recasts a T-fold as a geometric T^{2n} bundle over the same base, with physical spacetime a locally chosen null T^n slice of each fibre. T-duality then becomes a change of which slice is physical, and 'non-geometric' becomes the precise statement that the local slices do not glue into a global T^n sub-bundle. That reframing is the contribution, and it is the reason people still cite this paper. It is not a repackaging of his earlier T-fold work [1,10,11]; the polarization machinery in §5 is new and it is what makes the rest work. The product-structure characterization (integral R with R^2=1, pseudo-hermitian w.r.t. L) is a clean bijection, and the D-brane discussion in §7 falls out of it without strain, including the p -> p' change across a transition function, which is the sort of thing that usually has to be bolted on by hand. The n=1 case in §3.1 is a real external check: the Buscher rules (3.10),(3.11) come back out of the doubled formalism, and §2.4 does the elimination of ~X explicitly rather than gesturing at it. I would not call §6 qualitative, by the way — the argument that O(n,n;Z) does not reduce to GL(n;Z) is the standard non-reducibility statement and it is enough for the claim being made. A worked example with an explicit obstruction would be nice, not necessary. The soft spot is the one the paper itself flags. 'Physics is independent of polarization' is a quantum statement, and it needs an O(n,n;Z)-invariant chiral-boson measure including the dilaton shift (1.5). §8 concedes the quantization of the self-duality constraint is open and lists the competing schemes [32-40]. So the abstract's 'are formulated' is a classical formulation; the n=1 check verifies the classical Buscher rules, not the measure. That is a genuine gap, but it is stated plainly and it is the right gap for a framework paper to leave. Footnote 1 is similarly honest about the T^3 H-flux example not being a CFT as it stands. Who should read it: anyone working on doubled geometry, generalized geometry, or non-geometric flux backgrounds — this is the conceptual ancestor of a lot of that. Worth a serious referee, and worth citing.","headline":"The geometric core is right and has aged well; the 'string formulation' in the title is a classical statement, and Hull says so in §8.","tokens_in":26436,"tokens_out":1071,"would_cite":true,"duration_ms":22939,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":["11.25.-w","11.25.Mj","11.25.Uv"],"model":"selfhosted/deepseek-v4.1-flash","headline":"Hull proposes that string backgrounds glued together by T-duality, which have no global spacetime, are secretly ordinary manifolds once each torus fibre is doubled to carry both coordinates and their duals, with physical spacetime emerging,","keywords":["T-folds","non-geometric string backgrounds","doubled geometry","T-duality","O(n,n;Z) transition functions","flux compactifications","D-branes","self-duality constraint"],"falsifier":"Compute the partition function of the constrained doubled model for a simple T-fold, for instance the $n=1$ case with a T-duality monodromy or the $T^3$ with $H$-flux taken through two T-dualities, and compare it with the known modular-invariant partition function of the corresponding asymmetric orbifold; a mismatch in the spectrum, or a failure of the conformal and Lorentz anomalies to cancel, would show that the doubled geometry is a classical bookkeeping device rather than a formulation of string theory on T-folds.","tokens_in":26559,"feed_emoji":"🌀","tokens_out":3899,"duration_ms":75607,"temperature":0.7,"pith_summary":"The paper asks what a string background means when its coordinate patches are glued not by diffeomorphisms but by T-duality, the symmetry that trades a circle of radius $R$ for one of radius $1/R$. Hull's answer is to double the fibre: replace the local $n$-torus $T^n$ by a $T^{2n}$ carrying both the coordinates and their duals, on which the duality transition functions act as ordinary diffeomorphisms, and then impose a self-duality constraint that halves the doubled coordinates back to the right count of degrees of freedom. In the enlarged space the background is an honest manifold, and ordinary spacetime reappears only as a locally chosen null $T^n$ slice of each $T^{2n}$ fibre. The payoff is a sharp criterion: a background is geometric exactly when those local slices glue into a global $T^n$ sub-bundle, and non-geometric when they do not, so that 'non-geometric' stops being a slogan and becomes a statement about the topology of the doubled bundle.","feed_headline":"Double the torus, and non-geometry becomes geometry","feed_subtitle":"Backgrounds glued by T-duality have no global spacetime, but they do have a global doubled torus.","key_machinery":"The load-bearing device is the doubled $\\sigma$ model: a world-sheet Lagrangian $L_d = -1/2 H_{IJ} P^I /wedge *P^J + P^I /wedge *J_I + L(Y)$ in which the $2n$ one-form momenta $P^I$ on $T^{2n}$ obey $dP^I = 0$ and are subject to the self-duality constraint above, with $H$ a positive-definite metric on $T^{2n}$ built from a vielbein of $O(n,n)/(O(n) times O(n))$ so that it is parameterised by the usual $n$-torus metric and $B$-field, $E = G + B$. Solving the constraint for $n$ of the coordinates reproduces the standard Buscher $\\sigma$-model, which is how the formalism is checked against known T-duality rules. The second device is the polarization, encoded as an integral product structure $R in GL(2n;Z)$ with $R^2 = 1$, whose $+1$ and $-1$ eigenspaces are the physical $T^n$ and the dual $tilde{T}^n$; the three structures $R$, $S = LH$ and $I = SR$ satisfy an $SL(2,R)$ algebra, and the subgroup of $O(n,n)$ preserving all three is $O(n)$. This is what makes 'choosing a spacetime' a precise, local, and generally non-integrable operation.","core_discovery":"The central claim is that a T-fold, a background whose fibres are locally $n$-tori and whose transition functions lie in $O(n,n;Z)$ rather than in the geometric subgroup $GL(n;Z)$, can be reformulated as a string propagating on a genuine $T^{2n}$ bundle over the same base, subject to a self-duality constraint on the doubled momenta. The constraint, $P^I = L^{IJ} * (H_{JK}P^K + J_J)$ with $S^2 = 1$ for $S = LH$, is the covariant generalisation of the statement that half the doubled coordinates are left-movers and half are right-movers, with the split itself depending on position in the base. Choosing a physical spacetime then amounts to choosing a polarization: a maximal null $T^n$ subspace of each $T^{2n}$ fibre, equivalently an integral product structure $R$ with $R^2 = 1$ that is pseudo-hermitian with respect to the $O(n,n)$ metric. T-duality acts by changing that choice rather than by changing the background, and the physics is asserted to be independent of it. For a geometric background the locally chosen $T^n$ slices fit together into a $T^n$ bundle over the base; for a T-fold they do not, and there is a doubled manifold but no spacetime manifold.","pith_inferences":["If 'non-geometric' means 'the local polarizations fail to glue', then the failure should be measurable by a cohomology class on the base, and different non-geometric flux backgrounds ought to be distinguishable by that class rather than only by their local field content.","Because the doubled total space is a genuine manifold with a genuine $O(n,n;Z)$ bundle structure, standard topological tools, characteristic classes and index theorems, become available for non-geometric backgrounds that had none before; that seems the most direct route to invariants of T-folds.","The identification of a polarization with a pure spinor of $Spin(n,n)$ suggests that generalized complex and generalized Kähler geometry, developed for the geometric case, may classify which polarizations are admissible and which obstructions are possible.","A natural check outside the paper's scope is whether the doubled description survives when the fibre is not a torus: if the mechanism is really about null subspaces and self-duality, some analogue should exist for Calabi-Yau or K3 fibrations, and if none does, the construction is more tied to tori than the discussion implies."],"forward_implications":["T-duality becomes a passive operation: instead of transforming the background fields by the Buscher rules, one keeps the doubled geometry fixed and rotates the polarization, and the statement that the physics is T-duality invariant becomes the statement that no polarization is preferred.","For open strings, a generalized D-brane wraps a null $T^n$ of each $T^{2n}$ fibre, and the physical D-brane is the intersection of that Dirichlet subspace with the physical slice, so a $Dp$-brane in one patch can be glued to a $Dp'$-brane with $p' \\neq p$ in an overlapping patch.","Compactification on a T-fold still yields an ordinary low-energy field theory on the base, even in cases where no field-theory description exists on the total space.","The construction extends formally to transition functions in U-duality, mirror symmetry, and to M-theory, where the doubled torus is enlarged to include coordinates conjugate to membrane, fivebrane, and Kaluza-Klein-monopole wrapping charges.","The $n=1$ case shows the mechanism concretely: gluing a circle of radius $R$ to one of radius $1/R$ produces a smooth doubled $T^2$ bundle whose physical $S^1$ sub-bundle does not exist globally."],"supporting_citations":[{"why":"Supplies the T-fold compactification with duality twists, the class of backgrounds the paper sets out to reformulate.","marker":"[1]"},{"why":"Gives the T-dual of a flux compactification whose transition functions leave the geometric subgroup, the motivating example.","marker":"[3]"},{"why":"Provides the explicit $O(2,2;Z)$ monodromy matrix used as the worked non-geometric example.","marker":"[5]"},{"why":"The Buscher T-duality rules for $G$ and $B$ that the doubled formalism must reproduce when the constraint is solved.","marker":"[15]"},{"why":"The nonlinear $O(n,n)$ transformation of $E = G + B$ that defines how the doubled metric transforms between patches.","marker":"[16]"},{"why":"The doubled-torus formalism of Cremmer, Julia, Lu and Pope, on which the doubled action and constraint are built.","marker":"[27]"},{"why":"The chiral-boson quantization scheme named in the discussion as one route to imposing the self-duality constraint quantum mechanically.","marker":"[36]"},{"why":"Pure spinors and generalized complex structures, the framework invoked to describe a polarization as a choice of pure spinor of $Spin(n,n)$.","marker":"[42]"}],"fun_headline_variants":["T-fold strings live on a doubled torus, not a spacetime","Non-geometric backgrounds get a global doubled geometry","T-duality changes the torus slice, not the background","Doubled torus bundles make T-fold transitions geometric","T-folds have no global spacetime but a doubled torus"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The whole construction is a classical $\\sigma$-model rewriting, and it describes strings only if the self-duality constraint $P^I = L^{IJ} * (H_{JK}P^K + J_J)$ can be imposed consistently in the quantum theory; the paper concedes in its discussion that this is unresolved, pointing to competing chiral-boson quantization schemes and to the cancellation of conformal and Lorentz anomalies and to modular invariance as outstanding requirements.","fun_headline_variants_meta":{"raw":{"variants":["T-fold strings live on a doubled torus, not a spacetime","Non-geometric backgrounds get a global doubled geometry","T-duality changes the torus slice, not the background","Doubled torus bundles make T-fold transitions geometric","T-folds have no global spacetime but a doubled torus"]},"model":"selfhosted/deepseek-v4.1-flash","effort":"low","attempts":1,"cost_usd":0.0,"raw_usage":{"total_tokens":1819,"prompt_tokens":1087,"completion_tokens":732,"prompt_tokens_details":null,"completion_tokens_details":{"reasoning_tokens":663}},"tokens_in":1087,"tokens_out":732,"duration_ms":13133,"image_pages":[],"temperature":1.0,"reasoning_budget":50,"reasoning_tokens":663,"transport_provider":"selfhosted","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-10-29T16:43:54.131185+00:00","model_set":{"reader":"selfhosted/deepseek-v4.1-flash"},"falsifier":"Compute the partition function of the constrained doubled model for a simple T-fold, for instance the $n=1$ case with a T-duality monodromy or the $T^3$ with $H$-flux taken through two T-dualities, and compare it with the known modular-invariant partition function of the corresponding asymmetric orbifold; a mismatch in the spectrum, or a failure of the conformal and Lorentz anomalies to cancel, would show that the doubled geometry is a classical bookkeeping device rather than a formulation of string theory on T-folds.","supporting_citations":[{"cited_title":"Duality Twists, Orbifolds, and Fluxes","cited_arxiv_id":"hep-th/0210209","evidence_quote":"Supplies the T-fold compactification with duality twists, the class of backgrounds the paper sets out to reformulate."},{"cited_title":"New Supersymmetric String Compactifications","cited_arxiv_id":"hep-th/0211182","evidence_quote":"Gives the T-dual of a flux compactification whose transition functions leave the geometric subgroup, the motivating example."},{"cited_title":"Massive IIA String Theory and Matrix Theory Compactification","cited_arxiv_id":"hep-th/0303173","evidence_quote":"Provides the explicit $O(2,2;Z)$ monodromy matrix used as the worked non-geometric example."},{"cited_title":"Phys. Lett. B 194, 59","cited_arxiv_id":null,"evidence_quote":"The Buscher T-duality rules for $G$ and $B$ that the doubled formalism must reproduce when the constraint is solved."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The nonlinear $O(n,n)$ transformation of $E = G + B$ that defines how the doubled metric transforms between patches."},{"cited_title":"Dualisation of Dualities, I","cited_arxiv_id":"hep-th/9710119","evidence_quote":"The doubled-torus formalism of Cremmer, Julia, Lu and Pope, on which the doubled action and constraint are built."},{"cited_title":"Phys. Rev. Lett. 59, 1873","cited_arxiv_id":null,"evidence_quote":"The chiral-boson quantization scheme named in the discussion as one route to imposing the self-duality constraint quantum mechanically."},{"cited_title":"Hitchin, Generalized Calabi-Yau manifolds , Quart","cited_arxiv_id":null,"evidence_quote":"Pure spinors and generalized complex structures, the framework invoked to describe a polarization as a choice of pure spinor of $Spin(n,n)$."}],"review_version":1}