Supergravity backgrounds for deformations of AdS_n x S^n supercoset string models

We consider type IIB supergravity backgrounds corresponding to the deformed AdS_n x S^n supercoset string models of the type constructed in arXiv:1309.5850 which depend on one deformation parameter \k. In AdS_2 x S^2 case we find that the deformed metric can be extended to a full supergravity solution with non-trivial dilaton, RR scalar and RR 5-form strength. The solution depends on a free parameter (called a) that should be chosen as a particular function of \k to correspond to the deformed supercoset model. In AdS_3 x S^3 case the full solution supported by the dilaton, RR scalar and RR 3-form strength exists only in the two special cases of a=0 and a=1. We conjecture that there may be a more general one-parameter solution supported by several RR fields that for particular a=a(\k) corresponds to the supercoset model. In the most complicated deformed AdS_5 x S^5 case we were able to find only the expressions for the dilaton and the RR scalar. The full solution is likely to be supported by a combination of the 5-form and 3-form RR field strengths. We comment on the singularity structure of the resulting metric and exact dilaton field.


Introduction
Integrability of string sigma model is a key feature that allows to determine the string spectrum in non-trivial curved backgrounds. The study of integrable deformations of the most-symmetric AdS 5 ×S 5 model underlying AdS/CFT correspondence [1] is thus an important avenue of research that may also shed light on hidden symmetries of dual gauge theories. Recently, a novel one-parameter integrable deformation of the AdS 5 ×S 5 supercoset model was constructed in [2] (see also [3,4,5,6,7,8]). This model generalizes some previously known low-dimensional bosonic integrable models [9,10,11,12,13].
The corresponding target space type IIB supergravity background has no space-time supersymmetry and the bosonic isometry is reduced from SO(2, 4)×SO (6) to its Cartan subgroup [SO (2)] 3 ×[SO(2)] 3 , i.e. most of the symmetry of the original AdS 5 ×S 5 space becomes hidden (or "q-deformed"). Starting with a specific parametrization of the bosonic part of the deformed supercoset model [2] the corresponding 10d metric and B-field were found explicitly in [3]. However, extracting the associated RR field strengths that should promote the deformed metric to an exact supergravity solution from the fermionic part of the supercoset action turns out to be challenging even in the simpler low-dimensional AdS 2 ×S 2 and AdS 3 ×S 3 models [4].
Our aim here will be to find the deformed AdS n ×S n type IIB backgrounds by (i) starting with the deformed metric as given by the bosonic part of the supercoset model and (ii) solving the supergravity equations directly to find the expressions of all other fields required to support this metric as an exact solution. Finding "matter" fields supporting a given metric via Einstein equations is not a standard GR problem; the solution may not exist or, if it exists, it may not be unique. The present case is complicated also by the absence of supersymmetry and non-abelian isometries. We shall see that the solutions will have a rather unusual feature: while the string-frame metric is a direct sum of the deformed AdS n and S n parts, this will no longer be so for the dilaton and the RR fields -they will not factorize and thus "tie" the AdS n and S n parts together (as what fermion part of supercoset model does).
Having found a supergravity solution with the required deformed AdS n ×S n metric, one is still to decide if it is the one that actually corresponds to the integrable deformed supercoset model of [2]. As we shall see below, the solution for the dilaton and RR fluxes supporting a given deformed metric is not unique: in AdS 2 ×S 2 case there is a one-parameter a-family of solutions, and the same is expected to be the case also in the AdS 3 ×S 3 and AdS 5 ×S 5 cases. One is then to choose a as a function of the the deformation parameter κ 1 in order to match the supercoset model. This choice may be aided by consideration of the two special limits discussed in [4]: (i) κ = ∞ or "maximal deformation limit": in this case the deformed AdS n ×S n supercoset model becomes T-dual to "double Wick rotation" of the undeformed AdS n ×S n model, i.e. it has dS n ×H n target space supported by an imaginary n-form RR flux; (ii) κ = i (combined with a rescaling of coordinates and string tension) or "pp-wave limit": in this case the target-space metric becomes of pp-wave type and the problem of finding the supporting dilaton and fluxes simplifies.
We shall start in section 2 by finding a one-parameter type IIB solution with the metric being that of the κ-deformation of the AdS 2 ×S 2 ×T 6 one [4]. It corresponds to a solution of 4d supergravity obtained by compactification on 6-torus with only the dilaton, RR scalar and the RR 2-form being non-trivial. Guided by the two special limits mentioned above we shall argue that for a special value of the free parameter a = a(κ) = κ −1 η = κ −2 ( √ κ 2 + 1 − 1) the resulting background should corresponds to the κ-deformation of the AdS 2 ×S 2 supercoset model.
In section 3 we shall consider the κ-deformation of the AdS 3 ×S 3 ×T 4 space supported by the RR 3-form flux. Compactifying on 4-torus we shall use the truncated 6d action containing the metric, dilaton, RR scalar and RR 3-form field. Starting with the κ-deformed AdS 3 ×S 3 metric [4] we will find again a one-parameter family of solutions of the three scalar equations. However, only two special members of this family (with a = 0 and a = 1) will have extensions to solutions of the full set of 6d supergravity equations if only one RR 3-form field is assumed to be non-zero. The existence of the complete solution with an arbitrary parameter a (that may be chosen again as a(κ) to match the deformed supercoset model) appears to require more RR field strengths to be non-zero, a possibility which remains to be studied. We shall also present the analogs of the a = 0 and a = 1 solutions in the case of 2-parameter (κ + , κ − ) deformation of the AdS 3 ×S 3 supercoset [14] with the metric corresponding to the 2-parameter Fateev model [10] for deformations of AdS 3 and S 3 .
Guided by the low-dimensional examples, in section 5 we shall address the problem of promoting the κ-deformed AdS 5 ×S 5 metric and the B-field found in [3] to the full type IIB supergravity solution. An additional complication is that the 10d metric (and thus also other background fields) contains a non-trivial dependence on two extra angular coordinates. We will present two special solutions to the equations for the dilaton and the RR scalar which are the counterparts of the a = 0 and a = 1 solutions in the AdS 3 ×S 3 case. Here we will not able to find the corresponding 5-form flux and it appears likely that the full solution should exist only when also the RR 3-form flux is non-zero. Some comments on the singularity properties of the deformed AdS n ×S n backgrounds will be included in section 5. In appendix A we will give the form of the relevant supergravity equations in different dimensions and discuss truncations of the 10d supergravity action. In appendix B we will review the algebraic Rainich conditions on Maxwell stress tensor in 4 dimensions.
2 Deformation of AdS 2 ×S 2 In this section we shall extend the metric of the deformation of the AdS 2 ×S 2 ×T 6 space [4] to a 10d type IIB supergravity solution. Similarly to the undeformed background [15], this solution is a direct 10d lift of the corresponding solution of 4d supergravity obtained by compactification on 6-torus: the 5-form field strength F 5 is given by the product of the non-trivial 2-form field strength F 2 in 4 dimensions and the canonical hermitian 3-form of T 6 . 2 The background fields (metric, dilaton, RR scalar and 1-form potential) will depend on a free parameter a. We shall conjecture that for a special choice of a = a(κ) the resulting background should correspond to the superstring sigma model which is the κ-deformation of the AdS 2 ×S 2 supercoset model based on PSU(1, 1|2)/U(1)×U(1). As a check, we shall show that in the special limits of κ = ∞ (a = 0) and κ = i (a = 1) we indeed reproduce the expressions expected from the deformed supercoset construction.

One-parameter family of solutions
Let us recall that the AdS 2 ×S 2 ×T 6 solution can be obtained as a limit of 10d type IIB solution describing four intersecting stacks of D3-branes (see, e.g., [15] and refs. there). Upon reduction on T 6 to four dimensions it is supported by a two-form field strength F 2 . The compactification of type II supergravity to four dimensions on T 6 in general contains a large number of scalar and vector fields, some of which describe the deformations of the compact space. Since by construction the deformation acts only on the supercoset part of the geometry we may assume that the fields that should be nonvanishing are not related to T 6 . The minimal choice is the metric, dilaton, the RR scalar C and the vector A (with F 2 = dA as its field strength; the latter may represent several identified components of the 10d fields, cf. Appendix A). The Lagrangian for 4d supergravity restricted to these fields is 3 The simplest solution is the AdS 2 ×S 2 Bertotti-Robinson one with Φ and C being constant and Here c 1 , c 2 are reflecting the freedom of U (1) electromagnetic duality rotations. Our aim will be to find Φ, C and F 2 that promote the deformed AdS 2 ×S 2 metric [4] to an exact solution of the theory (2.1). Here L is the (inverse) curvature scale (that we shall often set to 1 in what follows) and κ is the parameter of deformation away from the symmetric AdS 2 ×S 2 point. The Ricci tensor and the curvature scalar of the metric g A ⊕ g S in (2.4) can be written as (2.6) The equations of motion following from (2.1) are given in the appendix A; we shall focus first on the trace of the Einstein equation, the equation for the RR scalar and the equation for the dilaton that can be organized as (cf. (A.7)) A way to solve this system is to consider first a particular limit: a small κ expansion combined with a particular rescaling of the coordinates κ → 0 , with fixed κρ , κ −1 t , (2.8) in which the S 2 part of the metric becomes undeformed while the deformation of the AdS 2 part remains non-trivial, i.e. (2.9) The perturbative expansion in κ respecting the symmetry κ → λκ, ρ → ρ λ , t → λt of the metric (2.9) should then be an expansion in powers of κρ: (κρ) n f n (r), C = (κρ) n g n (r). (2.10) Substituting this into (2.7) and summing up the perturbative series, we find the most general solution corresponding to the metric (2.9). The solution depends on one free parameter a: Note that here the combination C 2 + 4e −2Φ that should be a harmonic function according to (2.7) is simply a constant (2.12) Going back to the general case of the metric (2.4) and requiring that the solution of (2.7) should have the same property (2.12) leads to a similar one-parameter solution for the scalar fields. It is then easy to find also the solution for the vector potential A 4 14) This solution depends on the parameter a and also on a trivial parameter c 1 reflecting again the freedom of U (1) electromagnetic duality which is the symmetry of the equations following from (2.1) (we may always assume that c 1 = c 2 = 1 √ 2 without loss of generality). Let us note that the solution for the scalar C is of course defined up to a constant. Using this the special solution corresponding to a = 0 can be written as 5 Another special case corresponds to a = 1:

Symmetries and limits of the solution
The free parameter a should be fixed in order to establish a relation to the supercoset model which depends just on κ. To understand possible dependence of a on κ let us now discuss some properties and limits of the solution in (2.13)-(2.16). It turns out that it is invariant under certain sequences of dualities and analytic continuations: Perform T-dualities along t and ϕ directions. 4 One may check that the candidate F2 implied by the form of the Maxwell stress tensor appearing on the righthand side of the Einstein's equations for the given metric and the scalar fields obeys the algebraic Rainich condition [16,17,18,19] (see appendix B), i.e. there should indeed exist a vector field sourcing this geometry. 5 Note that the infinite shift of the RR scalar effectively makes C 2 + 4e −2Φ a function of the coordinates. Even for a = 0 one can perform a constant shift to have C = 0 for κ = 0, making C 2 + 4e −2Φ somewhat complicated. For a = 0 the shift is required.
3. Replace the 2-form potential, appearing after the T-dualities, by an axion via tensor-scalar duality in 4d. 4. Rescale the dilaton, the axion, and the Maxwell field to make e −2Φ = 1 when r = ρ = 0.
The transformation B has an implication for the large κ limit of (2.13)-(2.16): if we want to send κ to infinity while keeping the metric finite, then L/κ must remain fixed. In the transformation B this corresponds to sending ℓ = κ −1 to zero while keepingL fixed; eq. (2.19) then implies that such limit leads to imaginary fluxes unless a = 0. To see this more explicitly, let us consider the large κ limit of (2.13)-(2.16) for a = 0: Here we kept all coordinates fixed and rescaled the exponent of the dilaton and the RR fluxes by a free parameter γ. It is clear that no real value of this parameter makes e Φ positive while keeping C real. This argument breaks down only for a = 0, when the expression for C is to be modified by an infinite constant shift. For a = 0 we get Setting L = iκ, γ = 1/L, we find AdS 2 ×S 2 in the inverted coordinates. This suggests that the parameter a should vanish in this large κ limit.
A different way of taking the large κ limit of (2.13)-(2.16) is found by rescaling the coordinates and L as follows (the variables with tildes are to be kept fixed) Taking the limit κ, L → ∞ in (2.13) we then get (omitting tildes) [6] The T-dualities in t and ϕ applied to this metric give dS 2 ×H 2 space which is naturally a solution with a constant dilaton. Then the simplest choice for the dilaton that represents a solution together with the metric (2.23) should be [6] On the other hand, in the limit (2.22) the dilaton in (2.14) becomes To match (2.24) we should thus set a = 0. This suggests that if a is a function of κ then one should have a(κ → ∞) → 0. Another useful limit corresponds to setting κ = i in (2.13)-(2.16) (see [4]). Then metric (2.4),(2.13) becomes flat, and the dilaton (2.14) takes the form This expression is real only if a = 1 suggesting that one should have a(κ → i) → 1. In this case the dilaton becomes constant as appropriate for a "minimal" choice of the dilaton solution in the case a flat metric. There is also a special way of taking this κ = i limit by combining it with a rescaling of the coordinates This leads to a pp-wave 4d metric [4]. In this limit the F 2 flux in (2.16) diverges unless again a = 1.

Choice of a(κ)
In the previous subsection we discussed the natural values of a for the two special values of κ: (i) κ = ∞: in this limit the metric is related (T-dual) to an analytic continuation of AdS 2 ×S 2 and the simplest choice is to set a(∞) = 0.
(ii) κ = i: the assumption that Φ and C should remain real within the family of solutions parametrized by a implies that a(κ) should satisfy a(i) = 1.
Let us now propose a particular function a(κ) which has the required limits and is also consistent with the structure of the supercoset action.The deformed supercoset action of [2] depends naturally on combination of the projectors and P k are projectors on the supergroup elements with i k charge under Z 4 transformations. The string sigma model action and thus the background fields should then contain the two parameters κ and η entering simply as a ratio. We conjecture that the solution (2.13)-(2.16) with a(κ) given by should correspond to the AdS 2 ×S 2 supercoset model. Then in agreement with the above discussion of the two special limits.
3 Deformation of AdS 3 ×S 3 In the previous section we constructed a supergravity solution that should represent the background underlying the κ-deformed AdS 2 ×S 2 supercoset model. The important ingredient was the existence of a one-parameter family of solutions with the free parameter a which was then fixed to be a specific function of κ to match the corresponding limits of the supercoset construction.
In this section we shall attempt to follow the same strategy for the κ-deformation of AdS 3 ×S 3 ×T 4 space supported by RR 3-form flux. Compactifying on 4-torus we shall use the effective 6d action containing the dilaton Φ, RR scalar C and RR 3-form field strenght F 3 . Starting with the κ-deformed AdS 3 ×S 3 metric [4] we will find again a one-parameter family of solutions of the three scalar equations of the 6d theory. It will turn out, however, that only two members of this family -the analogs of the a = 0 and a = 1 solutions in (2.17) and (2.18) -can be extended to solutions of the full set of 6d equations if one assumes that in addition to Φ and C only one RR 3-form is non-zero.
It is likely that there should exist a more general solution (with an additional F 5 field in 10d or an extra F 3 field in 6d) parametrized by an arbitrary a that should match the supercoset model for a special choice of a = a(κ).

One-parameter family of solutions of the scalar equations
We shall start with the following "minimal" 6d Lagrangian representing a reduction and truncation of type IIB 10d supergravity on 4-torus. As discussed in appendix A, consistent truncation leads to the Lagragian supplemented by an additional constraint (A.13) (see Appendix A): The equations of motion for (3.1) are also given in appendix A. The simplest solution is AdS 3 ×S 3 supported by the self-dual F mnk (with Φ and C being trivial). We will be interested in finding a solution for which the metric is given by the κ-deformed AdS 3 ×S 3 metric implied by the supercoset construction [3,4] As in the previous section we shall first focus on the three scalar equations: the trace of the Einstein's equation, the RR scalar one and the dilaton one that can be organized as (cf. (A.7) and (2.7)) We begin by first solving them perturbatively in the small κ limit with κρ, κ −1 t being fixed as in (2.8),(2.9) when the metric becomes (3.5) Expanding in powers of κρ as in (2.10) we find a unique solution (regular at κρ = 0) which depends on one parameter a: This small κ solution can be generalized to the arbitrary κ solution of the three scalar equations (3.4): The same choice (2.29) for a(κ) as in the AdS 2 ×S 2 case then gives us a solution which is consistent with both κ = ∞ and κ = i limits. It is interesting to note a relation between the quadratic polynomial P in the deformed AdS 2 ×S 2 solution (2.13)-(2.16) and the quartic polynomial P 2 in (3.9). If we define the analog of P ≡ P − in (2.14) with 6 b → −b as P + then we observe that P 2 can be written as a product of P + and P − , i.e.
Attempting to extend this one-parameter solution of the scalar equations to a solution of the full set of 6d equations following from (3.1)-(3.2) using an ansatz-based approach suggests that this is possible only for the special values 0 and 1 of the parameter a. We shall also see another indication of this obstruction from the algebraic constraints on the 3-form stress tensor discussed in the next subsection.

Existence of a field strength for a given stress tensor: Rainich conditions
The question we are facing is how to find a 3-form flux supporting (together with the scalar fields) a given metric through the Einstein equation, i.e. how to find a solution for the flux given a specific form of its stress tensor. In general, the question is when some field configuration (i.e. some metric as well as other fields) can be sourced by an n-form field in d = 2n dimensions.
In the four dimensional Einstein-Maxwell theory this question was addressed long ago [16,17,18,19]: in order for some stress tensor T mn implied by the Einstein's equations to be generated by a Maxwell field strength T mn should be traceless and also its third power should be traceless as well (a brief derivation of this fact is given in appendix B).
Here we find the analogous conditions in six dimensions (the generalization to higher dimensions is also straightforward). Let us consider the stress tensor of a 3-form field strength Direct calculation shows that it satisfies Thus given a six-dimensional background (metric, dilaton, etc.) and computing the effective stress tensor T mn in the right-hand side of the Einstein equation that should be representing the contribution of the 3-form field, this T mn should satisfy eq. (3.12) in order for F mnk to exist. This is a necessary condition, which in general may not be a sufficient one. Some additional constraints may appear for special choices of the field strength. For example, for an (imaginary)-self-dual field strength we find that 4d : T m n = 0 , 6d : A similiar analysis implies that the necessary conditions that some 10d symmetric 2nd rank tensor may be the stress tensor of a five-form field strength are tr T = 0, tr T 3 = 0, tr T 5 = 0, tr T 7 = 0, tr T 9 = 0 . (3.14) These relations hold, in particular, for AdS 5 × M 5 solutions, where M 5 is an Einstein space.

Complete solutions
Starting with the metric (3.5) and the dilaton (3.6) and RR scalar (3.7) one can find explicitly the expected stress tensor for the 3-form RR field Direct calculation shows that it satisfies the non-trivial (last two) relations in (3.12) only for the special values a = 0, 1 of the parameter in (3.7). Thus (3.7) can be supported by the 3-form flux only in these two special cases. It may be possible to go around this problem by allowing for two non-vanishing independent 3-form fields in the reduced 6d Lagrangian (3.1). We will not attempt to study this option here. The corresponding small κ limit solutions (supplementing the metric (3.5)) are:

17)
These small-κ limit solutions can be extended to solutions with general κ by making ansätze that dress the solutions (3.16) and (3.17) with numerator and denominator functions of κ, r and ρ which become unity in the small κ limit. The resulting exact solutions are found to be (cf. (2.17),(2.18)) a = 0 : There are also solutions with flipped signs of t, χ, ϕ, ψ.
Let us note that, as in the 4d case (2.3), the undeformed AdS 3 ×S 3 metric (i.e. (3.3) with κ = 0) can be supported by a one-parameter family of 2-form potentials However, this freedom does not extend to the case of κ = 0 with nontrivial Φ and C. This is related to a different structure of the "electro-magnetic" duality group that acts on the 3-form field strength: in 4d this is SO(2) that rotates (c 1 , c 2 ) and in 6d this is Z 2 .

Symmetries and limits of the solution
Let us now discuss some properties of the solutions (3.8),(3.9) and (3.18), (3.19) corresponding to the metric (3.3).
A. Swap of the coordinates on the sphere The metric (3.3) is invariant under swapping of the angles on S 3 together with a redefinition of r: One can check that the scalar fields in (3.8),(3.9) remain invariant provided one also transforms a as a → 1 − a 1 + aκ 2 .

B. T-dualities
As in the AdS 2 ×S 2 case, we can perform a sequence of transformations: 1. T-dualize along t and ϕ directions.
This sequence maps the a = 0 solution (3.18) back to itself (after an appropriate rescaling of coordinates). The a = 1 background (3.19) is mapped into a solution with imaginary fluxes, which cannot be made real by further analytic continuations.

C. Inversion of coordinates
As in the AdS 2 ×S 2 case, the limit κ = ∞ simplifies after a sequence of duality transformations and analytic continuations: 1. Rewrite (2.13)-(2.16) in terms of x ≡ 1/ρ and y ≡ 1/r.
One can show that then the RR fields become complex unless a = 0.
Thus as in the AdS 2 ×S 2 case the large-κ limit appears to prefer the a = 0 solution. At the same time, the κ = i or pp-wave limit [4] appears to prefer the a = 1 solution. Namely, if we consider again the limit (2.27) then C 2 in (3.18) diverges, while C 2 in (3.19) remains finite and real, i.e. a = 1 : Comparing this with eqs. (3.28) and (3.29) in [4] giving the two-form potential in this limit and accounting for the coordinate change we find a perfect match. The value a = 1 for κ = i is also singled out by comparing the corresponding limits of the dilatons in (3.18) and (3.19) with the expression for the natural value of the dilaton found directly in this limit in [4] (see eqs. We conclude that, as in the AdS 2 ×S 2 case, the limits κ = ∞ and κ = i appear to select two different values of a, suggesting that there should exist an interpolating solution with a = a(κ). While the six-dimensional Rainich conditions discussed in sec. 3.2 rule out such a solution supported by a single 3-form flux, a preliminary investigation suggests that there may exist a 6d supergravity solution with two different 3-form fields being non-zero.

A generalization: 2-parameter deformation
It was shown in [4] that the κ-deformation of the AdS 3 ×S 3 metric corresponds to a special case of the general 2-parameter Fateev model [10] which is also the same as the 2-parameter family of classically integrable bi-Yang-Baxter sigma models constructed in [11,13]. The corresponding deformed AdS 3 ×S 3 metric can be written as For κ − = 0, κ + = κ we get back to the metric (3.3). There is no B-field. The supercoset model with this bosonic part was constructed in [14]. Similarly to the case of the κ-deformed AdS 3 ×S 3 metric, it should thus be possible to extend the metric (3.25) to a full supergravity solution.
Indeed, we found the following generalizations of the a = 0 (3.18) and a = 1 (3.19) solutions with both κ + and κ − : As in the AdS 2 ×S 2 and AdS 3 ×S 3 cases discussed above, it is natural to expect that there should exist a one-parameter family of solutions including (3.26) and (3.27) as special cases. Solutions (3.26) and (3.27) are interchanged by the transformation The invariance of the metric (3.25) under the map (3.28) was noted in [14].

Deformation of AdS 5 ×S 5
The extension of the κ-deformed AdS 5 ×S 5 metric and B-field [3] to a full supergravity solution turns out to be more challenging than in the above lower-dimensional cases. This is due, in particular, to the lack of isometries, i.e. a non-trivial dependence on the two extra angular coordinates. While we will not find a complete solution, in this section we shall discuss some of its features and draw analogies with the AdS 2 ×S 2 and AdS 3 ×S 3 cases. Assuming a particular structure of the RR fluxes we shall find two different solutions to the scalar equations which are the counterparts of the a = 0 (3.18) and a = 1 (3.19) solutions in the AdS 3 ×S 3 case (we shall thus refer to them as the "a = 0" and "a = 1" solutions). To construct them it will be useful to switch to a T-dual frame where there is no B-field. We shall find that in this frame both solutions have vanishing RR scalar, C = 0. However, in contrast to the AdS 3 ×S 3 case we have been unable to find a one-parameter family connecting these two special solutions. Moreover, the 10d algebraic Rainich conditions discussed in sec. 3.2 imply that these solutions cannot be supported solely by a 5-form flux, i.e. one should excite other fluxes as well. We leave the study of this possibility for the future.
Our starting point will be the deformed AdS 5 ×S 5 metric and B-field corresponding [3] to the κ-deformed supercoset model of [2] 7 where we used the shorthand notation s x = sin x , c x = cos x.
To put this background on the equal footing with the above low-dimensional solutions, it is convenient to remove the NS-NS B-field by performing the T-duality along the two compact coordinates φ 1 and ψ 1 which gives a non-diagonal metric In the absence of the B-field the dilaton equation is We shall assume that in addition to the metric and the dilaton only the RR scalar C and the 5-form field F 5 = C 4 are excited. Then for a given metric the scalars C and Φ must satisfy an over-constrained system of the three equations -(4.4) as well as the RR scalar equation and the trace of the Einstein equation: In the general case we may then parametrize the dilaton as (cf. (2.14),(3.6)) e −2Φ = (1 − κ 2 ρ 2 )(1 + κ 2 r 2 )(ρc ζ rc θ ) 2 P 4 (ρ, r, ζ, θ) , (4.8) where P 4 is expected to have a polynomial dependence on ρ and r as well as a polynomial dependence on the trigonometric functions of ζ and θ.
Remarkably, as in the AdS 3 ×S 3 case (cf. (3.16), (3.17)), here we find two special solutions with C = 0: Undoing the T-duality, in the original frame (4.1),(4.2) the expressions for the dilaton become a = 0 : Let us now consider the κ → ∞ and the κ → i limits [4] of the above expressions for the dilaton: κ = ∞: As discussed in [4], in the κ → ∞ limit the natural solution for the dilaton is expected to be a product of factors depending separately on the AdS 5 and S 5 coordinates. Taking κ → ∞ in (4.11) we indeed find a factorization 8 At the same time, the limit of the a = 1 expression (4.12) does not factorize for κ → ∞. κ = i: In the κ → i limit the dilaton may also be expected to factorize [4]. 9 However, this does not happen for the a = 0 expression (4.11). At the same time, the κ → i limit of the a = 1 dilaton (4.12) does factorize a = 1 : (4.14) Thus, as in the lower-dimensional cases, it seems natural to expect the existence of a one-parameter family of solutions with a = a(κ) chosen so that a(i) = 1 and a(∞) = 0. At the same time, it is possible to check that the algebraic Rainich conditions (3.14) for existence of the F 5 flux are not satisfied by the stress tensor containing the contribution of the T-dual frame dilatons eqs. (4.9) and (4.10) only. This indicates that one should look for more general solutions with several RR fields excited. This is analogous to our earlier observation that the a-family of scalar field solutions in the deformed AdS 3 ×S 3 case (3.8),(3.9) cannot be supported by just one 3-form RR field strength.

Some properties of the deformed backgrounds
While we did not find the full solution in the deformed AdS 5 ×S 5 case some of its properties are already evident from the form of the metric and the dilaton and are shared with the corresponding AdS 2 ×S 2 and AdS 3 ×S 3 solutions. All these deformed backgrounds represent a novel class of non-supersymmetric type IIB supergravity solutions which have factorized string-frame metric but non-factorized dilaton and RR fields.
For all the three deformed string-frame metrics (2.4),(3.3),(4.1) in dimensions 4, 6 and 10 there is a (naked) curvature singularity at ρ = κ −1 (cf. (2.6)). The integrability of the underlying sigma models [9,10,12,2] implies, in particular, that it should be possible to find the explicit form of the corresponding geodesics and study their approach to the singularity. At the same time, concentrating on the point-like limit may be misleading: one may need to investigate if the string probes "see" the singularity. For example, attempting to probe it with a long spinning folded string shows that the fold-points of the string remain at some finite distance from the singularity [20,21]. 10 Since the string-frame dilaton equation is independent of the RR fluxes, the singularities of the dilaton are determined by the singularities of the metric and the NSNS B-field. From the dilaton equation (or the exact solutions (2.14),(3.8), (3.18), (3.19),(4.11),(4.12)) one concludes that near this point e Φ → (1 − κ 2 ρ 2 ) −1/2 → ∞. This means that the effective string coupling blows up, suggesting that one cannot study the near-singularity region using string perturbation theory. 11 This conclusion may, however, be premature: due to lack of supersymmetry the leading-order supergravity solution may receive non-trivial α ′ corrections that may smear the singularity out in both the metric and the dilaton. Clarifying this issue requires a better understanding of the underlying deformed supercoset model at the quantum level.
It is interesting to note that while both the deformed metric g mn and the dilaton Φ are singular, in all AdS n ×S n cases the "T-duality invariant" volume density e −2Φ √ −g is regular at ρ = κ −1 . For example, if one performs a formal T-duality along the time t direction, in, e.g., (2.4) one gets a regular metric with a horizon at ρ = κ −1 and with the T-dual string coupling e Φ vanishing at that point. While this time-like T-duality is a formal transformation (the resulting type IIA background will have complex fluxes) this may be suggesting a hidden regularity of the original type IIB background. 12

A Equations of motion and embedding into 10d supergravity
In sections 2 and 3 we discussed d = 4 and d = 6 supergravities truncated to two scalar fields (the dilaton and the RR scalar) and one d/2-form field. The corresponding actions may be written as where the coefficients (e 2 , e 3 ) are d = 4 : e 2 = 1, e 3 = 0; d = 6 : e 2 = 0, e 3 = 1 .
The equations of motion coming from this action are It is convenient to separate the trace of the Einstein equation and combine it with the other scalar equations. Using the fact that the trace of the stress tensor of the d/2-form in d dimensions vanishes, we have from (A.3) Then (A.5) with (A.6) give which may be used in place of any of the three scalar equations in (A.5) and (A.6).
In the 10d case with non-vanishing B-field we get the following forms of the scalar equations Let us now review the embedding of the four-and six-dimensional systems (A.1), (A.2) in 10D supergravity. The undeformed AdS 3 ×S 3 solution can be embedded in type IIB supegravity by identifying F 3 in (A.1) with RR 3-form field strength in ten dimensions. To embed the deformed 6d solution, we also identify C with the RR scalar in ten dimensions, i.e. the starting point is the following truncated 10d action (A.10) To perform the reduction to 6d, we write the ten-dimensional metric as 13 ds 2 10 = g mn dx m dx n + e A dy i dy i , (A.11) where y i are flat coordinates on T 4 . The standard dimensional reduction on the 4-torus then gives (see, e.g., [25]) This reduces to (A.1) for A = 0, but equation of motion for A leads to an additional constraint: This relation is satisfied by (3.18), (3.19), (3.26), (3.27).
The undeformed AdS 2 ×S 2 solution can be embedded in type II 10d supergravity in two different ways [15], which are related by T-dualities. In the absence of the Kalb-Ramond field, the action for type IIA supergravity is Choosing the ansatz (z i are 3 complex coordinates of 6-torus) and reducing on the 6-torus we find To have a solution with A = 0, we must set To write the action we relax the self-duality conditions by replacing F 5 with The dimensional reduction of the type IIB action then gives This coincides with (A.1) for configurations with A = 0, and the equation of motion for A does not introduce additional constraints.
To summarize, we have demonstrated that the reduction of type II 10d supergravity reproduces the 4d action (A.1), but its 6d counterpart must be supplemented by the constraint (A.13).

B Rainich conditions in four dimensions
As discussed in section 3, to test whether a given stress-energy tensor can be sourced by a particular type of flux, we need a generalization of the Rainich condition to higher dimensions. To review the original condition in 4d, let us start with Maxwell stress tensor The first condition is obvious, while to prove the second one, we can go to the orthonormal frame and perform a (coordinate-dependent) rotation to put F mn into a blog-diagonal form where we used that tr T = 0. Then using (B.2) we conclude that tr(T 3 ) T = 0, implying (B.4). 14 Here T 3 stands for T k n T m k T r m , etc.