hep-th/0412193 NSF-KITP-04-134 PUPT-2146 MIT-CTP-3577 Cascading RG Flows from 5 New Sasaki-Einstein Manifolds 0 0 2 n a C. P. Herzog , Q. J. Ejaz , I. R. Klebanov J † ∗ ‡ 1 2 Kavli Institute for Theoretical Physics Center for Theoretical Physics † ∗ University of California Massachusetts Institute of Technology 4 v Santa Barbara, CA 93106, USA Cambridge, MA 02139, USA 3 [email protected] [email protected] 9 1 2 Joseph Henry Laboratories 1 ‡ 4 Princeton University, Princeton, NJ 08544, USA 0 / [email protected] h t - p e h Abstract : v i X In important recent developments, new Sasaki-Einstein spaces Yp,q and conformal gauge r theories dual to AdS Yp,q have been constructed. We consider a stack of N D3-branes a 5 × and M wrapped D5-branes at the apex of a cone over Yp,q. Replacing the D-branes by their fluxes, we construct asymptotic solutions for all p and q in the form of warped products of the cone and R3,1. We show that they describe cascading RG flows where N decreases logarithmically with the scale. The warp factor, which we determine explicitly, is a function of the radius of the cone and one of the coordinates on Yp,q. We describe the RG cascades in the dual quiver gauge theories, and find an exact agreement between the supergravity and the field theory β-functions. We also discuss certain dibaryon operators and their dual wrapped D3-branes in the conformal case M = 0. December 2004 1 Introduction An interesting generalization of the basic AdS/CFT correspondence [1, 2, 3] results from studying branes at conical singularities [4, 5, 6, 7, 8]. Consider a stack of N D3-branes placed at the apex of a Ricci-flat 6-d cone Y whose base is a 5-d Einstein manifold X . 6 5 Comparing the metric with the D-brane description leads one to conjecture that type IIB string theory on AdS X with N units of 5-form flux, is dual to the low-energy limit of 5 5 × the world volume theory on the D3-branes at the singularity. 5 Well-known examples of X are the orbifolds S /Γ where Γ is a discrete subgroup of 5 SO(6) [4]. In these cases X has the local geometry of a 5-sphere. Constructions of the 5 dual gauge theories for Einstein manifolds X which are not locally equivalent to S5 are also 5 possible. The simplest example is X = T1,1 = (SU(2) SU(2))/U(1) [6]. The dual gauge 5 × theory is the conformal limit of the world volume theory on a stack of N D3-branes placed at the apex of the conifold [6, 7], which is a cone over T1,1. This = 1 superconformal N gauge theory is SU(N) SU(N) with bifundamental fields A ,B , i,j = 1,2, and a quartic i j × superpotential. Recently, a new infinite class of Sasaki-Einstein manifolds Yp,q of topology S2 S3 was × discovered [9, 10]. Following progress in [11], the = 1 superconformal gauge theories N dual to AdS Yp,q were ingeniously constructed in [12]. These quiver theories have gauge 5 × groups SU(N)2p, bifundamental matter, and marginal superpotentials involving both cubic and quartic terms. These constructions generalize the SCFT on D3-branes placed at the apex of the complex cone over dP [13], corresponding to Y2,1 [11]. Impressive comparisons 1 of the conformal anomaly coefficients between the AdS and the CFT sides were carried out for dP in [14], and in full generality in [12]. 1 In this paper we address a number of further issues concerning the gauge/gravity duality involving the Yp,q spaces. We match the spectra of dibaryon operators in the gauge theory with that of wrapped D3-branes in the string theory. Next, we consider gauge theories that arise upon addition of M wrapped D5-branes at the apex of the cone. Our discussion generalizes that given in [15, 16] for the Y2,1 case. We show that these gauge theories can undergo duality cascades, and construct the dual warped supergravity solutions with (2,1) flux.1 As a preliminary, in the next two sections we review the gauge theory duals for and the geometry of these Yp,q spaces. 1The duality cascade was first developed for the conifold in [17, 18] and later generalized in [19, 15, 16]. 1 b) a) Y Y Y U U U U V Z c) Z V Y Y Y Y U U U U U Y Y V Y V Figure 1: Shown are a) the unit cell σ; b) the unit cell τ; and c) the quiver for Y4,3, στ˜σσ˜. 2 The Conformal Surface of Y p,q Gauge Theories In this section, we review the construction of the Yp,q gauge theories and argue that they flow to an IR conformal “fixed surface” of dimension two. That this surface has dimension two will be more or less clear from the gravity side where the two free complex parameters are C ie−φ and (C ie−φB ). − S2 2 − 2 As derived in [R12], the quivers for these Yp,q gauge theories can be constructed from two basic units, σ and τ. These units are shown in Figure 1. To construct a general quiver for Yp,q, we define some basic operations with σ and τ. First, there are the inverted unit cells, σ˜ and τ˜, which are mirror images of σ and τ through a horizontal plane. To glue the cells together, we identify the double arrows corresponding to the Uα fields on two unit cells. The arrows have to be pointing in the same direction for the identification to work. So for instance we may form the quiver στ˜ = τ˜σ, but στ is not allowed. In this notation, the first unit cell is to be glued not only to the cell on the right but also to the last cell in the chain. A general quiver might look like σσ˜στ˜τσ˜ . (1) In general, a Yp,q quiver consists of p unit cells of which q are of type σ. The Yp,p−1 gauge theories will have only one τ type unit cell, while the Yp,1 theories will have only one σ type 2 unit cell. Each node of the quiver corresponds to a gauge group while each arrow is a chiral field transforming in a bifundamental representation. For the Yp,q spaces, there are four types of bifundamentals labeled Uα, Vα, Y, and Z where α = 1 or 2. To get a conformal theory, we take all the gauge groups to be SU(N). Later in this paper, when we add D5-branes, we will change the ranks of some of the gauge groups and break the conformal symmetry. The superpotential for this quiver theory is constructed by summing over gauge invariant operators cubic and quartic in the fields Uα, Vα, Y, and Z. For each σ unit cell in the gauge theory, we add two cubic terms to the superpotential of the form ǫ UαVβY and ǫ UαVβY . (2) αβ L αβ R Here, the indices R and L specify which group of Uα enter in the superpotential, the Uα on the right side or the left side of σ. The trace over the color indices has been suppressed. For each τ unit cell, we add the quartic term ǫ ZUαYUβ . (3) αβ R L An analysis of the locus of conformal field theories begins with counting the fundamental degrees of freedom which are in this case the 2p gauge couplings and the p+q superpotential couplings (assuming an unbroken SU(2) symmetry for the Uα and Vα). We will assume all the gauge groups have equal ranks. There are in total 3p+q fields and thus 3p+q anomalous dimensions which we can tune to get a conformal theory. We think of the 3p+q β-functions as functions of the 3p+ q anomalous dimensions which are in turn functions of the 3p+ q coupling strengths, β (γ (g )). j i k Let us check that one set of solutions ofβ = 0 involves setting the anomalous dimensions j of all the Z fields equal, the anomalous dimensions of all the Y fields equal, and similarly for the Uα and Vα. Instead of working with the anomalous dimensions γ, of the fields, we find it convenient to work with the R charges, R , R , R , and R . (For superconformal Y Z U V gauge theories, recall that 2(1+γ) = 3R.) There are p + q β-functions for the superpotential couplings. p q of the β functions − vanish when R +R +2R = 2 and are associated with loops in the τ unit cells, while the Z Y U remaining 2q vanish when R +R +R = 2 and are associated to loops in the σ unit cells. U Y V There are 2p β-functions for the gauge couplings. 2q of these couplings are associated with the σ unit cells, and the beta functions for these couplings vanish when 2 = R + U R +R while the remaining 2p 2q belong to the τ unit cells and vanish when R +R + V Y Z Y − 2R = 2. Thus the gauge coupling β-functions contain exactly the same information as the U superpotential β-functions. It could be that there are more solutions to setting the β = 0 which involve more generic j values for the anomalous dimensions. However, such solutions would require even more 3 degeneracy among the 3p + q β-functions, which is unlikely. Assuming we have found the most general solution of β = 0 (which we have checked for dP but should be checked in j 1 general), we have found that only 3p+q 4+2 of the β = 0 are linearly independent. Thus j − there is seemingly a two dimensional plane in the space of allowed anomalous dimensions which produce conformal field theories. Of course we know that a-maximization [20] will pick out the right anomalous dimensions. However, there is a different way of looking at these 3p + q 2 linearly independent − β-functions. They place 3p +q 2 constraints on the 3p+ q couplings, leaving a space of − conformal theories with two complex dimensions. By construction, this space preserves the SU(2) U(1) U(1) global flavor symmetry of the Yp,q. If we allow a breaking of this × × symmetry, then there may exist additional exactly marginal superpotential deformations (see [21]). 3 Review of the Y p,q geometry The Yp,q spaces are topologically S2 S3, and the Sasaki-Einstein metric on them takes the × form [9, 10] 1 y 1 v(y) 2 2 2 2 2 2 dΩ = − (dθ +sin θdφ )+ dy + (dψ cosθdφ) Yp,q 6 w(y)v(y) 9 − 2 +w(y)[dα+f(y)(dψ cosθdφ)] (4) − where 2(b y2) w(y) = − , (5) 1 y − b 3y2+2y3 v(y) = − , (6) b y2 − b 2y +y2 f(y) = − . (7) 6(b y2) − For the metric to be complete, 1 p2 3q2 b = − 4p2 3q2 . (8) 2 − 4p3 − p The coordinate y is allowed to range between the two smaller roots of the cubic b 3y2+2y3: − 1 y = 2p 3q 4p2 3q2 , (9) 1 4p − − − (cid:16) p (cid:17) 1 y = 2p+3q 4p2 3q2 . (10) 2 4p − − (cid:16) p (cid:17) 4 The three roots of the cubic satisfy y + y + y = 3/2, so the biggest root, which we will 1 2 3 need later in the paper, is 1 y = 2p+2 4p2 3q2 . (11) 3 4p − (cid:16) p (cid:17) The period of α is 2πℓ where q ℓ = . (12) −4p2y y 1 2 The remaining coordinates are allowed the following ranges: 0 θ < π, 0 φ < 2π, and ≤ ≤ 0 ψ < 2π. ≤ The volume of Yp,q is given by q(2p+ 4p2 3q2)ℓπ3 Vol(Yp,q) = − . (13) p3p2 4 Dibaryons and New 3-Cycles We will identify some new supersymmetric 3-cycles in the Yp,q geometry, but first recall that Martelli and Sparks [11] identified two supersymmetric 3-cycles, denoted Σ and Σ in their 1 2 paper. These three cycles are obtained by setting y = y or y = y respectively. At these 1 2 values for y, the circle parametrized by ψ shrinks to zero size, and the three cycles can be thought of as a U(1) bundle parametrized by α over the round S2 parametrized by θ and φ. Martelli and Sparks [11] computed the R-charges of the dibaryons corresponding to D3- branes wrapped on Σ and Σ . In general, these R-charges are given by the formula [22] 1 2 πN Vol(Σ ) i R(Σ ) = . (14) i 3 Vol(Yp,q) From this general formula, it follows that N R(Σ ) = 4p2 +2pq +3q2 +(2p q) 4p2 3q2 , (15) 1 3q2 − − − (cid:16) p (cid:17) N R(Σ ) = 4p2 2pq+3q2 +(2p+q) 4p2 3q2 . (16) 2 3q2 − − − (cid:16) p (cid:17) These R-charges should correspond to operators det(Y) and det(Z) made out of the bifun- damental fields that are singlet under the global SU(2) symmetry. Dividing these dibaryon R-charges by N, we observe a perfect match with the R-charges of the Y and Z singlet fields determined from gauge theory by Benvenuti, Franco, Hanany, Martelli, and Sparks [12], R = R(Σ )/N and R = R(Σ )/N.2 Y 1 Z 2 2The gauge theory computation for Y2,1 was performed earlier by [14]. 5 Here we show which 3-cycles correspond to the dibaryons made out of the SU(2) doublet fields Uα and Vα.3 Such dibaryons carry spin N/2 under the global SU(2). On the string side, the wrapped D3-brane should therefore have an SU(2) collective coordinate (see [22] for an analogous discussion in the case of T1,1). The only possibility is that this SU(2) is precisely the SU(2) of the round S2 in the metric. Therefore, the 3-cycles corresponding to these dibaryons should be localized at a point on the S2. Now recall from the gauge theory analysis of [12] that R = (2p(2p 4p2 3q2))/3q2 , (17) U − − R = (3q 2p+p 4p2 3q2)/3q . (18) V − − p Before proceeding, note that R = R +R . So if we determine which cycle Σ corresponds V U Z 3 to Uα, we can deduce that Vα is just a sum of Σ and Σ . 3 2 As discussed above, the three cycle Σ should correspond to fixing a point on the S2 and 3 integrating over the fiber. Setting φ = θ = const, the induced metric on this three cycle becomes 1 v 2 2 2 2 ds = dy + dψ +w(dα+fdψ) . (19) wv 9 We can characterize this 3-cycle more precisely. The metric on Σ can be thought of as 3 a principal U(1) bundle over an S2 where the S2 is parametrized by y and ψ. A principal U(1) bundle over S2 is a Lens space S3/Z where k is given by the first Chern class c of k 1 the fibration. The A = fdψ is a connection one-form on the U(1) bundle. Because α ranges from 0 to 2πℓ, dA = 2πc /ℓ. Integrating c over the S2 yields 1 1 f(y ) f(y ) 2 1 c = − = p . (20) 1 ZS2 ℓ − In other words, our Σ is the Lens space S3/Z . In [11], Σ and Σ were identified as the 3 p 1 2 Lens spaces S3/Zp+q and S3/Zp−q respectively. We find 4π2ℓ Vol(Σ ) = √gdydαdψ = (y y ) , (21) 3 2 1 Z 3 − where we have used the fact from (19) that √g = 1/3. Plugging into the formula for the R-charge, indeed R(Σ ) = NR . 3 U We now imagine that Vα corresponds to adding the cycles Σ and Σ together. Indeed, 3 2 these two cycles intersect along a circle at y = y . 2 3We would like to thank S. Benvenuti and J. Sparks for discussions about these new 3-cycles. A similar analysis will likely appear in a revision of [12]. 6 We also check that Σ is a supersymmetric cycle, or in other words that the form 1J J, 3 2 ∧ where J is the Kaehler form on the cone over Yp,q, restricts to the induced volume form on the cone over Σ . More formally, we are checking that Σ is calibrated by 1J J. 3 3 2 ∧ From Martelli and Sparks (2.24) [11], we find that 1 y 2 J = r − sinθdθ dφ 6 ∧ 1 1 2 + rdr (dψ cosθdφ) d(yr ) dα+ (dψ cosθdφ) , 3 ∧ − − ∧(cid:18) 6 − (cid:19) and hence 1 1 2 J = rdr dψ +d(yr ) dα+ dψ . (22) |Σ3 3 ∧ ∧(cid:18) 6 (cid:19) Thus we find that 1 r3 J J = dr dψ dy dα (23) 2 ∧ (cid:12) 3 ∧ ∧ ∧ (cid:12)Σ3 (cid:12) as expected. (cid:12) 5 Warped Solutions with (2,1) Flux The first step in constructing supersymmetric warped solutions for these Yp,q spaces is con- structing a harmonic (2,1) form Ω . We begin by rewriting the metric so that locally we 2,1 have a U(1) fiber over a Kaehler-Einstein manifold. From (2.17) of [11], we have dΩ2 = (eθ)2 +(eφ)2 +(ey)2 +(eβ)2 +(eψ)2 (24) Yp,q where we have defined the one forms 1 y 1 y eθ = − dθ , eφ = − sinθdφ , (25) r 6 r 6 1 √wv ey = dy , eβ = (dβ +cosθdφ) , (26) √wv 6 1 eψ = (dψ cosθdφ+y(dβ +cosθdφ)) . (27) 3 − In terms of the original coordinates β = 6α ψ. Here, the ψ is a coordinate on the local − − U(1) fiber. There is then a local Kaehler form, denoted J by [11], on the Kaehler-Einstein base: 4 J = eθ eφ +ey eβ. (28) 4 ∧ ∧ 7 Based on [15], we expect to be able to construct Ω from a (1,1) form ω using this local 2,1 Kaehler-Einstein metric such that ω = ω, dω = 0, and ω J = 0. We guess that 4 4 ∗ − ∧ ω = F(y)(eθ eφ ey eβ) . (29) ∧ − ∧ The form ω is clearly anti-selfdual and orthogonal to J . Using a complex basis of one-forms 4 constructed in (2.27) of [11], it is not hard to check that ω is indeed a (1,1) form. The condition dω = 0 then implies that 1 F(y) = . (30) (1 y)2 − Further, we construct a (2,1) form from the wedge product of a (1,0) form and ω: dr Ω = K +ieψ ω . (31) 2,1 (cid:18) r (cid:19)∧ We have introduced a normalization constant K for later convenience. We have checked that dΩ = 0 and Ω = iΩ . 2,1 6 2,1 2,1 ∗ Next, we analyze Ω (32) 2,1 Z Σ i for the three three-cycles i = 1, 2, 3. We find that 8iπ2ℓ y 1 Ω = K , (33) 2,1 ZΣ1 − 3 1−y1 8iπ2ℓ y 2 Ω = K , (34) 2,1 ZΣ2 − 3 1−y2 4iπ2ℓ 1 1 Ω = K . (35) 2,1 ZΣ3 − 3 (cid:18)1−y2 − 1−y1(cid:19) The ratios between these integrals look superficially to be irrational. However, the ratios must be rational, and we find that if we set 9 2 2 K = (p q ) (36) 8π2 − then Ω = i( p+q) , (37) 2,1 Z − − Σ1 Ω = i(p+q) , (38) 2,1 Z − Σ2 Ω = ip . (39) 2,1 Z − Σ3 8 Now, to construct a supergravity solution, we take the real RR F and NSNS H forms 3 3 to be i ′ iK Ω = F + H , (40) 2,1 3 3 g s dr F = KK′eψ ω ; H = g KK′ ω , (41) 3 3 s − ∧ r ∧ ′ where we have introduced another normalization constant K . In particular, F should be 3 quantized such that 2 ′ F = 4π αM(p q) (42) 3 Z − Σ1 where M is the number of D5-branes. Thus we find that K′ = 4π2α′M. (See [23] for our normalization conventions.) 5.1 Derivation of Five-Form Flux For the metric and F we take the usual ansatz with the warp factor h, 5 ds2 = h−1/2dx2 +h1/2(dr2 +r2dΩ2 ) , (43) 4 Yp,q −1 4 −1 4 g F = d(h ) d x+ [d(h ) d x] . (44) s 5 ∧ ∗ ∧ Due to the appearance of the y-dependent factor F(y) in the (2,1) flux, it is inconsistent to assume thathis afunction ofr only. Instead, similar tothe gravity duals offractionalbranes on the Z orbifold [24], h is a function of two variables, r and y. For q p the y-dependence 2 ≪ can be ignored, and the warp factor approaches that found for the warped conifold in [17]. On the other hand, for p q p we find that h gets sharply peaked near y = 1, and the − ≪ solutions approach the gravity duals of fractional branes in orbifold theories [25, 26, 24]. Thus, the warped solutions we find with the Yp,q serve as interesting interpolations between the conifold and the orbifold cases. More explicitly, the first term in (44) is ∂h ∂h h−2 dr+√wv ey d4x . (45) − (cid:18)∂r ∂y (cid:19)∧ Working out its Hodge dual, and substituting into the equation dF = H F , (46) 5 3 3 ∧ we find the second order PDE ∂ ∂h ∂ ∂h C 5 3 (1 y) r r (1 y)wv = , (47) − − ∂r (cid:18) ∂r(cid:19)− ∂y (cid:18) − ∂y(cid:19) r(1 y)3 − 9