Finite temperature corrections to tachyon mass in intersecting D-branes

We continue with the analysis of finite temperature corrections to the Tachyon mass in intersecting branes which was initiated in [1]. In this paper we extend the computation to the case of intersecting D3 branes by considering a setup of two intersecting branes in flat-space background. A holographic model dual to BCS superconductor consisting of intersecting D8 branes in D4 brane background was proposed in [2]. The background considered here is a simplified configuration of this dual model. We compute the one-loop Tachyon amplitude in the Yang-Mills approximation and show that the result is finite. Analyzing the amplitudes further we numerically compute the transition temperature at which the Tachyon becomes massless. The analytic expressions for the one-loop amplitudes obtained here reduce to those for intersecting D1 branes obtained in [1] as well as those for intersecting D2 branes.


Introduction
Applications of AdS/CFT to various condensed matter systems and QCD at strong coupling have been an active field of research for the past few years. Specific areas where these holographic techniques have been applied with varied degree of success are in the study of quantum liquids at strong coupling [3][4][5][6][7][8][9][10], including studies of phase transitions such as in superconductors [11][12][13].
A holographic model for a BCS superconductor was proposed in [2]. The idea behind the proposal was to apply the holographic techniques for studying superconductivity in a top-down approach. The proposal is based on the holographic QCD model constructed by Sakai and Sugimoto [14]. This is a variant of the model proposed by Witten [15] which includes fundamental fermions. Various interesting aspects of holographic QCD models have been explored by several authors. For a partial list of references see [16][17][18][19][20][21][22][23][24][25][26][27][28]. The Sakai-Sugimoto model consists of flavor D8 branes in the D4 brane background. The relevant (3 + 1)D large N QCD resides on the D4 brane which is compactified on S 1 . The D8 branes intersect in the bulk at an angle. In [2] two such D8 branes were used to model the dual holographic superconductor. Intersecting D-branes usually have instabilities manifesting as tachyons in the spectrum. Such a tachyonic instability that arises in the above case of intersecting D8 branes has been proposed to be the dual of the Cooper-pairing instability in the 3 + 1 dimensional theory in [2]. In [1], we have demonstrated a thermal stabilization in a simplified set-up consisting of two D1-branes intersecting at one non-zero angle. This is done by computing a finite temperature mass correction to the tree-level tachyon using the Yang-Mills approximation. This approximation in the case of Dp-branes gives a renormalizable theory only for p ≤ 3.
In this paper we extend the calculations initiated in [1] to the case of intersecting D3 branes. The setup consists of two D3 branes intersecting with one non-zero angle θ in flat space. Similar to such an analogous setup for intersecting D1 branes the spectrum of open strings with the two ends on each brane consists of a Tachyon as the lowest mode. The mass 2 for the Tachyon is given by −θ/(2πα ′ ) [29][30][31][32]. This was rederived in the Yang-Mills approximation (α ′ → 0, θ → 0), with θ/(2πα ′ ) = q fixed in [34][35][36], wherein the end point of Tachyon condensation leading to smoothing out of the brane configuration was studied. Further studies on tachyon condensation in intersecting branes include [37][38][39].
The purpose of the present computation is to analyze the finite temperature corrections to the Tachyon mass which we obtain by computing the 2-point amplitude. Since the finite temperature corrections are positive in sign, there exists a critical temperature T c at which the Tachyon mass is exactly zero. We shall compute this temperature in the limits (as mentioned above) where the Yang-Mills approximation is valid. The calculations in this JHEP04(2017)109 paper follow along the lines of [1]. The intersecting brane configuration consists of giving a non-zero expectation value to one of the scalars namely Φ 3 1 which is of the form qx. Here x denotes a coordinate along the plane of intersection. Due to the linear (in x) nature of the background and in the temporal gauge (A a 0 = 0), the wave-functions that diagonalize the quadratic part of the Lagrangian consist of Hermite Polynomials instead of exponential plane-waves which we have in the usual case. The corresponding momentum modes are discrete and are denoted by n. The wave functions have been derived in [1] for the D1 branes. These functions have been adapted for the present case of D3 branes with some modifications. The additional complication in the present case is the presence of extra momentum degrees of freedom coming from the directions transverse to the plane of intersection. There are two such extra momentum directions relative to the D1 brane case. We call these momentum modes k = (k 2 , k 3 ).
The supersymmetric Yang-Mills theory on the D3 brane is finite. Although supersymmetry is completely broken for the intersecting branes, the ultraviolet property of the theory on the intersecting branes is unaffected. The theory is still finite in the ultraviolet. We shall show analytically the ultraviolet finiteness for all the one-loop two-point amplitudes computed in the paper. Specifically we show that for large n, independent of the value of k, the contributions from the bosons and fermions in the loop cancel. Similar cancellation has also been shown for the large k for the two-point amplitude for the "massless" fields.
The one-loop amplitude for the Tachyon involves fields propagating in the loop that are massless at tree-level. In order to deal with the resulting infrared divergences arising out of these massless fields we first find the one-loop correction to the masses for these fields that are massless at tree-level at finite temperature. These masses are then used to modify the propagators for the "massless" fields. The resulting one-loop Tachyon amplitude is thus infrared finite. The analytic expressions obtained here reduce to those for the intersecting D1 and D2 branes once the extra transverse momenta/momentum is set to zero. The final evaluation of the temperature corrected masses for the "massless" as well as the Tachyon field could not however be performed analytically. We had to resort to numerics due to the complicated nature of the expressions involved. This paper is organized as follows. In the following section 2 we analyze the quadratic part of the Yang-Mills Lagrangian with the background Φ 3 1 = qx. We derive the various wave-functions both for the bosons and fermions and then write out the Lagrangian in terms of momentum modes. The propagators for the various fields can then be read out. In section 3 we compute the one-loop two-point Tachyon amplitude. In the subsection 3.1.1 we analytically show the finiteness of the Tachyon amplitude for large n (ultraviolet) and for all values of the momenta k. The one loop amplitude for the massless fields namely Φ 3 I (I = 1, . . . , 6) and A 3 µ (µ = 1, 2, 3) have been worked out in section 4. These have been done in subsections 4.1, 4.2 for fields Φ 3 1 and Φ 3 2 respectively. The contributions from the bosons and the fermions for fields A 3 µ have been worked out separately in subsections 4.3.1, 4.3.2, 4.3.3 and 4.3.4. In section 5 we show the numerical results for the computation of all the two point amplitudes including transition temperature at which the Tachyon becomes massless. We conclude the paper with some outlook in section 6. Various details have JHEP04(2017)109 been worked out in the appendices. We summarize our notations in appendix A. We list the various parameters, variables used in the paper in table 1 of this appendix. We review the dimensional reduction of D = 10 super Yang-Mills to D = 4 in appendix B. The propagators and vertices for the Tachyon and various massless fields have been worked out in appendices C and E respectively. In appendix D we give details of the one-loop tachyon amplitude incorporating the corrected propagators for the fields with tree-level mass zero. Appendix F contains exact expressions for the two point amplitudes.
2 Tree-level spectrum

Bosons
In the Yang-Mills approximation the intersecting brane configuration with one angle is obtained by putting the background value of Φ 3 1 equal to qx. In this background we first write down the quadratic part of the bosonic action in the temporal gauge A a 0 = 0. Various fields decouple at this quadratic level. We first write down below the various decoupled parts of the resulting from the action (B.11), the quadratic bosonic part of the action can be written as, We have identified z 1 = x. In the following we shall identify z 0 = −iτ and (z 2 , z 3 ) ≡ y.
The eigenfunctions of O 11 B have bee worked out in [1] these are The normalizations are given by The eigenvalues corresponding to ζ n (x) are (2n − 1)q and those corresponding toζ n (x) are all zero. Thus the spectrum in the latter case is completely degenerate. In the non-zero eigenvalue sector we do not have normalizable eigenfunctions corresponding to n = 1. However unlike this sector, in the zero eigenvalue sector we have normalizable eigenfunction for n = 1, which is simply H 1 ( √ qx) but there is no normalizable eigenfunctions for n = 0 in this sector. ζ n (x) andζ n (x) satisfies the orthogonality conditions √ q dxζ † n (x)ζ n ′ (x) = δ n,n ′ √ q dxζ † n (x)ζ n ′ (x) = δ n,n ′ (2.14) They also satisfy, √ q dxζ † n (x)ζ n ′ (x) = 0 for all n and n ′ .
Similarly the eigenfunctions of the operator O ′ B are simply ζ ′ = (φ n (x), A n (x)), and ζ ′ = (φ n (x),Ã n (x)) with eigenvalues (2n − 1)q and 0 respectively. There is thus a two fold degeneracy for this spectrum of the theory.
In the finite temperature analysis that follows we take τ to be periodic with period β which is the inverse of the temperature T . Further defining N = √ q/β, we write down the mode expansion for ζ

JHEP04(2017)109
Using the same analysis as above the second term of the action (2.5) in terms of momentum modes is, 3) yields decoupled eigenvalue equations for the scalar fields. The scalar fields with gauge components (a, b = 1, 2) can be expanded using the basis of harmonic oscillator wavefunctions, N ′ (n)e −qx 2 /2 H n ( √ qx). Thus for example, The scalar fields with the gauge component (a, b = 3) and the gauge fields A 3 i can be expanded using the basis for plane wave as We can thus write down the third term in the action (2.5) in terms of the momentum modes as, Here (I = 2, 3; a = 1, 2), (J = 1, 2, 3; b = 1, 2, ), (i, j = 1, 2, 3) and k 2 = (ω 2 m + k 2 x + k 2 ).

JHEP04(2017)109
The Lagrangian (2.29) can thus be written as where the operators O F andÕ F are given by, where i = 0, 2, 3. The eigenfunctions of the matrix operators have been obtained in [1].
Adopting the same notation as in [1] respectively. Here, The corresponding eigenvalues are −i λ ′ n = −i √ 2nq. We now list some important relations satisfied by the eigenfunctions We can now write down the mode expansions for the fermions where θ i are four component fermions and ω m = (2m+1)π/β with m = 0, 1, 2, · · · . Further The quadratic action in terms of the momentum modes is then

JHEP04(2017)109
The tachyon mass-squared, m 2 eff can be read off from the effective potential for the tachyon namely V |C(0, 0, 0)| 2 . This is given by . We shall thus be interested in computing the two point amplitude for zero external momenta. In this case the various functions appearing in the expressions (3.1) and (3.2) take the following form In the above expressions (i, j) = 2, 3 andk i = k i / √ q,l = l/ √ q.
Using the above forms of the vertices at zero external momenta, we can now write down the contribution to the two-point tachyon amplitude with bosons in the loop. This is given by where, Σ 1 C−C is the contribution with the four point vertices (see figure 1)  Figure 3. Feynman diagrams involving three-point vertices F 6 .
Similarly the contribution with the three point vertices is (see figure 2) The exact expressions are given in the appendix F.1. We now write down the contribution from fermions in the loop. The corresponding Feynman diagram is shown in figure 3.
The factor of 4 in front comes from summing over the contributions from four sets of fermions (see eq. (C.6)).
We will now be interested in evaluating the two point function with the external momenta set to zero. (ω m ′′ =ω m ′′ = k ′′ =k ′′ = n ′′ =ñ ′′ = 0). Evaluating the trace in (3.10) the amplitude reduces to where, Now define the sum over the fermionic frequencies as, Then after performing the sum over the Matsubara frequencies the amplitude can be written as (3.14)

Ultraviolet finiteness
In this section we show that the one-loop amplitude is finite in the UV (large n). Finiteness of the amplitudes in the UV is expected as the underlying N = 4 SYM theory is finite.
In the present case supersymmetry is only broken by the background. Nevertheless we demonstrate this finiteness explicitly by showing the cancellation of contributions from bosons and fermions in the loop in the UV, which in this case is the large n limit. This exercise is particularly useful in order to ensure that the counting of the modes in the loops has been done correctly. In the following we proceed along a slightly alternate route to demonstrate the UV finiteness than that was followed in [1] for the case of intersecting D1 branes. We first note that the factors coming from the three-point vertices are of the form The above asymptotic expression shows that the leading large n contribution from the l integrals come from the regionl 2 = 2n. We will thus, in all the contributions obtained above, replacel 2 by 2n in the propagators. This leaves us with l integrals only over the vertices. We can now perform these integrals over l separately noting that To illustrate this further, consider the integral where the function I(l 2 , ω, |k| 2 ) contains propagators and other polynomials inl 2 in the numerator. According to the above in the large n limit the leading value of the integral (3.17) will be given by 2n I(2n, ω, |k| 2 ). As a first example let us take I(l 2 , ω, |k| 2 ) =l 2 . Performing the integral (3.17) directly and then taking the large n limit we get The large n leading term in the expression (3.18) can also obtained as described above. As a second example let us consider the integral that involves both the sum over n as well as the integral over l (which appears in many of the expressions for the two-point function). A typical such integral is of the form,

JHEP04(2017)109
We first perform the integral over l in (3.19). The result in the large n limit reads (for simplicity we have set q = 1) where we have assumed 2nq ≫ ω 2 m + |k| 2 . Further terms with odd powers of ω m have been dropped. This is exactly as we had proceeded in [1]. Following the other route, we have The leading term in the expressions (3.20) and (3.21) match when we assume 2nq ≫ ω 2 m +|k| 2 . The difference in the sub-leading pieces is expected as we have restricted ourselves to leading contribution as coming froml 2 = 2n in the second approach.
Although this expansions with the assumption 2nq ≫ ω 2 m + |k| 2 is needed to show the agreement of the resulting leading behaviours in the two approaches it turns out that it will not be necessary in the ultimate demonstration of finiteness of the full two-point amplitude at large n. In other words in the following we work with the full expression (3.21) and the assumption that 2nq ≫ ω 2 m + |k| 2 is not invoked. With these observations we proceed to write down the asymptotic contributions from each of the terms ((I) − (XI)).

JHEP04(2017)109
in the expressions of (IV ), (V ) and (V I) and doing the l integral as elaborated above, we get (IV ) = (III) and We now write down the asymptotic forms of the contributions (V II) − (XI) corresponding to the diagrams in figure 2. Here we implement the steps described at the beginning of this section.
Adding all the contributions (I) − (XI) we get (3.31) Now following the above observations it is easy to see that the large n limit of (3.11) is exactly same as (3.31) coming from fermions in the loop but with opposite sign. Thus the one loop amplitude is finite in the UV (large n).
We should note that the finiteness in the large n limit as shown here holds irrespective of the values of ω 2 m and |k| 2 . That is for every fixed value of ω 2 m and |k| 2 the amplitudes are finite in the large n limit. It would similarly be useful to give an analytic proof of the finiteness of the amplitudes for fixed values of ω 2 m and 2nq but large |k|. Unfortunately we are not able to demonstrate this analytically as the closed form expressions (in n) for the integrals (3.3), (3.4), (3.5) that involves product of four Hermite polynomials could not be obtained. Such finiteness of the amplitudes for the massless fields however can be shown both for large n and fixed |k| as well as large |k| and fixed n which we shall see in the following sections.

Infrared issues
Some of the contributions to the two point amplitude (I − XI) diverge in the infrared. These infrared divergences occur from two sources: (i) There are poles of the from of 1/ω 2 m . These poles are artifacts of the A a 0 = 0 gauge and one should use prescriptions to remove them. See for example [40][41][42][43] and references therein. In our case this means that we shall simply drop terms that are proportional to m β √ q (2πm) 2 . (ii) There are also genuine infrared JHEP04(2017)109 divergences due to tree-level "massless" modes propagating in the loop. These modes are the fields with the gauge index a = 3, namely Φ 3 J /Φ 3 J (J = 1, 2, 3) and A 3 i (i = 1, 2, 3). This is because these fields do not couple to the background value of the field Φ 3 1 , and hence the tree-level spectrum for the fields with gauge index equal to 3 is not gapped unlike the fields with gauge indices equal to 1, 2. To ultimately get a finite answer for the one loop amplitude we follow the procedure implemented in [1]. We first find the one-loop correction to the propagators for these fields and then use the corrected propagators to evaluate the one-loop two-point tachyon amplitude. The corrected propagator will involve one-loop corrections to the zero tree-level masses for the fields Φ 3 I , A 3 µ . These corrections will be computed in the next section.
We now give some details on how to incorporate the one-loop masses in the propagators for the tree-level "massless" modes. The quadratic action for the Φ 3 I /Φ 3 I / fields (see eq. 2.28) including the mass terms is written as, For the A 3 i fields writing down the modified propagator requires a bit more work. The one-loop correction breaks the SO(3) in (2.28) to SO (2). As a result the one-loop mass for A 3 1 (m 1 ) field is different from those of A 3 2 and A 3 3 fields which are equal (say m 2 ). Introducing a vector u i ≡ (1, 0, 0), we can write down the quadratic part of the action as . For the one-loop amplitude there are momentum dependent corrections as well. Correspondingly the termÃ 3 j also arises in the one-loop effective action (3.34). The correction coefficients a, b and c are different due to the breaking of SO(3) invariance. In the infrared limit however only the mass terms in (3.34) provide the necessary regulation. Thus for simplicity we shall work with only the non-zero masses m 1 and m 2 . Following these observations the corrected propagator takes the following form (see appendix D for further details) where, the functions A, B, C and D are given by With these modifications we can re-write the corresponding modified expressions in (I)-(XI). For the Φ 3 I fields the mass terms in the propagators in the loop are easy to implement. However for the A 3 i fields some more work is needed. These expressions are listed in appendix D. The numerical results incorporating the above are analyzed in section 5.

Two point amplitude for massless fields
In the following sections we shall compute the one-loop corrections to the masses for the . For each of the two-point amplitudes we also demonstrate the cancellation of contributions from the bosons and fermions in the loop in the large n limit irrespective of the values of ω 2 m and |k| 2 as well as for large |k| at fixed values of ω 2 m and 2nq thereby showing that the amplitudes are UV finite. 3 1 field In this section we compute the two point amplitude for Φ 3 1 field. Using the vertices listed in appendix E.1, we first write down the contributions to the one loop amplitude from bosons in the loop, This contribution consisting of the four-point vertices as shown in figure 4 is given by,

Two point amplitude for Φ
and the contributions from the three-point vertices (see figure 5) is Where For computing the above amplitudes at zero external momentum we list the various exact functions below In the above expressions i = 2, 3 andk i = k i / √ q. We first analyze the contributions to the expressions (4.1) and (4.2) in the large n limit. This will give us their ultraviolet JHEP04(2017)109 behavior. We write down the contributions from each of the terms separately, The full contribution from (4.4) and (4.5) simplifies to where in the last line of the above equation we have kept the zero temperature UV divergent piece.
The exact contribution to the amplitude for all n with bosons in the loop is given in appendix F.2.
We now write down the contribution from fermions in the loop JHEP04 (2017)109 The factor of 4 in front comes from summing over the contributions from all the four sets of fermions (eq. (E.1)).
We will now be interested in evaluating the two point function with the external momenta set to zero. (ω m ′′ =ω m ′′ = k ′′ =k ′′ = 0). At zero external momenta the exact form of this contribution to the two point amplitude is We now perform the sum over the Matsubara frequencies and write down the contributions as where g(λ ′ n , β) is defined in equation (3.13). We can extract the zero temperature contributions from the above equation. To study the UV behaviour of the zero temperature contributions we take the large n limit. In this limit the expression reduces to Thus equations (4.6) and (4.10) show that the UV divergences cancel (for large n) between that contributions from the bosons and fermions in the loop.
Large k. In the following we list the large k (for fixed n) behaviour of the integrands in Σ 1 at zero temperature. So the in the limit k ≫ nq the various expressions reduce to, The above O(1/|k|) contributions from the bosons in the loop in the large k limit adds up to This is exactly the same in magnitude and opposite in sign as the O(1/|k|) term from (4.9). This shows that in this regime of large k and fixed n the integrals converge.

Two point amplitude for Φ 3
I fields There is an unbroken SO(5) invariance in the theory. We thus only consider the amplitude for Φ 3 2 field. Using the vertices listed in the appendix E.2 we first compute the amplitude with bosons in the loop. (4.14) where The vertices at zero external momentum are given by In the above expressions i = 2, 3 andk i = k i / √ q. Using these zero momentum vertices the amplitudes corresponding to the diagrams in figures 7 and 8 are as follows with, and Figure 9. Feynman diagrams involving three-point vertices In the large n limit the above amplitude reduces to We now turn to the fermions. The contribution from fermions in the loop is given by The factor of 2 in front is due to the contributions from the two sets of So the amplitude simplifies as

JHEP04(2017)109
In the above equation we have summed over the Matsubara frequencies. g(λ ′ n , β) is given by eq. (3.13). To show that UV divergence cancellation we take the limit β → ∞. The zero temperature contribution is then This is opposite in sign to the contribution from the bosons in the loop thus showing the finiteness of the one-loop amplitude in the UV.
Large k. As we have done before we list behaviour of the integrands in Σ 1 at zero temperature in the limit k ≫ nq.

Two point amplitude for A 3 µ fields
In this section we compute the two point amplitude for A 3 µ fields. We compute the amplitudes for µ = 1 and µ = 1 in the following two subsections. The contribution consisting of the four-point vertices as shown in figure 10 is given by, Figure 10. Feynman diagrams contributing to Σ 1 and the contributions from the three-point vertices as shown in figure 11 is

JHEP04(2017)109
where Since we are interested in the amplitudes at zero JHEP04(2017)109 external momentum we list the various exact functions at this momentum below In the above expressions i = 2, 3 andk i = k i / √ q. Using the above vertices (eq. (4.33)) we can now write down the contributions from the expressions (4.31) and (4.32) in the large n limit. These are exactly equal to the expressions obtained for the Σ 1 eqs. (4.4) and (4.5). Thus the full contribution Σ 1 in the large n limit is given by eq. (4.6). However the exact form (at all n) for some of the terms of the two point contributions differ from that of the Φ 3 1 amplitude. This is because of the difference in the forms of the vertices given by eqs. This is same as the contribution obtained for Σ 1 where, (I), (II) and (III) are given by eqs. (F.14) and (F.15). The contribution towards the two point amplitude from the three point vertices as shown in figure 11. Using (F.1), this is given by The individual terms are given in appendix F.3. In the large n limit the full expressions sum up to the same value as that of Φ 3 1 given by (4.6).

µ = 2, 3: bosons
We now compute the contribution with bosons in the loop for the two-point function for the A 3 2 . Due to the underlying symmetry the amplitude for µ = 3 is same as the one -25 - for µ = 2. Using the vertices listed in appendix E.3 the contribution consisting of the four-point vertices as shown in figure 12 is given by,

JHEP04(2017)109
and the contributions from the three-point vertices as shown in figure 13 is where

JHEP04(2017)109
In the above expressionsk i = k i / √ q. Using these vertices the various diagrams in figures 12 and 13 reduce to It is now not difficult to show that in the large n limit the full contribution to the two point amplitude reduces to At zero temperature in the limit k ≫ nq the various O(1/|k|) contributions reduce to, which adds up to (4.50)

JHEP04(2017)109
Summing over the Matsubara frequencies we write down various contributions and using the form for g(λ ′ n , β) given by (3.13), we get In the large n limit and at zero temperature (β → ∞) the above contribution has the form This is equal in magnitude to the contribution from bosons in the loop but with opposite sign thus showing that the amplitude is ultraviolet finite. A similar analysis such as the one that was done for the two-point amplitude for the Φ 3 1 field at zero temperature and in the limit k ≫ nq can again be done here. The O(1/|k|) contributions from the bosons and the fermions in the loop can be shown to cancel as done exactly at the end of section 4.1.

µ = 2, 3: fermions
From eq. (4.51) we can now write down the contribution from the fermions in the loop for µ = 2, 3 Now summing over the Matsubara frequencies we arrive at the following expressions where g(λ ′ n , β) is defined in (3.13). From this the zero temperature contribution in the large n limit is Thus confirming the cancellation with the contribution from bosons in the loop. Similarly we can expand the integrand at zero temperature and in the limit k ≫ nq, which gives Comparing with (4.50) we see similar convergence of the full amplitude for large k with nq fixed.

JHEP04(2017)109 5 Numerical results
This section is devoted to study the two-point functions for the massless as well as the Tachyon fields numerically. The analytical expressions for the two-point amplitudes derived in the previous sections can be reduced to the case of intersecting D2 branes by setting one of the momenta transverse to the x direction as zero. However here we restrict ourselves only to the case of intersecting D3 branes. We give the numerical results for the treelevel massless modes and the Tachyon in separate subsections. As explained before the corrections to the masses for the tree-level massless modes are needed to cure the infrared divergences in the Tachyon two-point amplitude. The computations are done for the Yang-Mills coupling, g 2 = 1/100.

Numerical results: tree-level massless modes
In this section we discuss the numerical computation of the two-point functions for the modes that are massless at tree-level, namely the Ψ ≡ (Φ 3 1 , Φ 3 I (I = 2, · · · , 6), The leading terms in the above expansions gives the zero temperature corrections for the intersecting D1 branes. Also note that the leading terms for Φ 3 I and A 3 i are the same as in 1+1 dimensions A 3 i combine with Φ 3 I into a SO(7) multiplet. It has been shown that in the large |k| limit, the contribution to the two-point functions from the bosons and fermions in the loop cancel each other. Thus the amplitude is finite in large |k|. For each of the fields this has been shown towards the end of sec-  the tree-level massless fields, Let us now rescale the momenta as k ′ = βk. Further the term β 2 λ n in the exponent, in the limit β 2 q → 0 can be replaced by a continuous variable, say l ′ 2 . Thus replacing we may write (5.6) as The integrals are independent of β and hence we expect the T 2 behaviour. This assumes that the momentum limits in (5.6) go from zero to infinity. In the numerics however we have to work with finite limits on the momenta. The plots as shown in the figures here are for |k|/ √ q integrated over 0 to 1000 and n is summed over up to 10.

Numerical results: tachyon
The tachyon two-point one-loop amplitude has various massless fields propagating in the loop. As discussed earlier, to regulate the infrared divergences due to the massless fields we incorporate the one-loop corrected masses for the tree-level massless fields. The behaviour of these masses as a function of temperature is computed numerically in the previous section. We now incorporate these masses in the propagators for these fields namely,   The effective mass-squared for the tachyon is defined by Where V |C| 2 is the effective potential for the tachyon. Upto quadratic in the tachyon where g 2 is the Yang-Mills coupling and is the one-loop contribution computed in section 3. Thus, where x 0 is the zero temperature quantum correction and x is independent of temperature. In writing (5.9) we have assumed that T ≫ √ q and have followed the analysis done below eq. (5.6). The ctitical temperature from (5.9) is given by The zero temperature quantum correction for the tachyon mass-squared is found to be m 2 0tach /q = (2g 2 x 0 ) = 0.29. Thus, T c / √ q ∼ 1/g.

JHEP04(2017)109
In the plots 23 and 24, the critical value of √ qβ namely √ qβ c = 0.0277 or the critical temperature Tc √ q = 36.101. This behaviour of the effective mass-squared bears the hallmark of a phase-transition. In the brane picture the mass squared becoming positive indicates the dissolution of the tachyonic mode, by way of which stability is achieved. This is also analogous to the transition from the superconducting BCS phase to the normal phase as studied in [1,2].
Following the scaling arguments similar to that for the massless fields, here too we expect a T 2 behaviour for the tachyon mass for T ≫ √ q. However note that the plot in figure 24 is almost linear in T √ q which is a deviation from the T 2 behaviour. This is an artifact of the numerical computation. The T 2 behaviour and the scalings of the momenta are valid for infinite sums and integrals. In order to perform the numerical calculations, due to technical limitations we had to simplify the computations by summing and integrating over relatively small range for the momenta, namely n, l, k ∼ 0 − 10. For sufficiently large range of momentum sums and integrations, the T 2 behaviour can presumably be reproduced in the numerics.

Discussion and summary
In this paper we have computed the one-loop correction to the tachyon mass for the intersecting D3 branes at finite temperature. The calculation is extension of our previous work [1], wherein a similar computation was done for the case of intersecting D1 branes when the Yang-Mills approximation is valid. The underlying N = 4 Super Yang-Mills for the case of D3 branes makes the theory finite in the ultraviolet. Nevertheless we have explicitly shown that all the one-loop amplitudes computed in the paper are finite in the UV. The intersecting brane configuration breaks the spatial SO(3) invariance and the straight intersecting branes implies that in the Yang-Mills approximation the background values of Φ 3 1 is linear in x. The momentum modes along x direction for the fields that couple to this background are thus discretized and are labeled as n. We have shown the finiteness in the UV for large n irrespective of the values of the momentum |k| in the other directions for all the one-loop amplitudes computed in the paper. Further for the one-loop amplitudes for the tree-level massless fields we have shown that for fixed n the amplitudes are finite for large values of |k|.
The infrared divergence appearing in the one-loop tachyon amplitudes are of two kinds. Ones that appear as artifacts of the A a 0 = 0 gauge can be removed prescriptions discussed in [40][41][42][43]. The genuine infrared divergences appearing from the massless modes in the loop are cured by using the one-loop corrected propagator that incorporates the one-loop masses for the tree-level massless fields. Finally the amplitudes are all finite both in the infrared as well as in the ultraviolet.
It can be seen that the analytic expressions for the amplitudes obtained in the paper reduce to the ones for the D1 branes. Similarly one can easily find the corresponding expressions for the case of intersecting D2 branes as well by setting one of the momentum modes transverse to the x direction to zero. In this paper we have only presented here the numerical results for the case of D3 branes. It is expected that all the masses including that JHEP04(2017)109 of the Tachyon exhibit T 2 behaviour for large T . This however could not be demonstrated for the Tachyon in our numerics due to technical limitations. From the numerics we have found the critical temperature T c when the Tachyon becomes massless. Thus for temperatures above the critical temperature the intersecting configuration is favoured. Although the configuration of intersecting branes here is in flat space the results obtained in this simplified model are consistent with the strong coupling BCS holographic model proposed in [2]. It was shown in [2] that at zero temperature there exist solutions in the Yang-Mills approximation which correspond to breaking of U(1) symmetry thus signaling condensate formation and the brane configuration is smoothed out.
The existence of the critical temperature denotes a phase transition. To analyze the nature of the transition at lower temperatures one needs to analyze the full tachyon potential. This knowledge would in turn shed light on the BCS transition in the dual theory. In the following we conclude with some observations on the effective potential. Here we restrict ourselves to the case of D3 branes. In the Yang-Mills approximation, from the point of view of (2 + 1)D, the modes C(m, n, k) have masses (2n − 1)q. Thus there is an infinite tower of states including the Tachyon with masses proportional to q. In this scenario it might appear that the notion of Tachyon effective action obtained by integrating out massive modes is ill-defined. There have however been extensive studies of Tachyon effective action in open string theory where the effective actions were proposed from the consistency with world-sheet conformal field theory (see [44] for review). Similar techniques should be employed for the construction and further study of effective potential in the present case. For approaches along these directions see [34][35][36][37][38][39]. It would further be interesting to generalize such potential to the case of finite temperature and for nontrivial background geometries.
Alternatively from the point of view of (3 + 1)D the discrete n-modes are momenta for the field ζ(z) (see eq. (2.16)) along the x direction. We can thus construct a (3 + 1)D effective action for the field ζ(z). This in the zero external-momentum limit including n = 0 reduces to the potential in terms of only the C(0, 0, 0) modes. This potential in principle can be computed using perturbative techniques utilized in the paper and would involve computation of N -point amplitudes for the C(0, 0, 0) modes. Compared to the effective potential in the (2 + 1)D theory this would thus give the effective potential when all the modes with positive mass-squared are set to zero i.e. the massive modes only appear in loops. This is exactly what is done in this paper. Here for simplicity we have ignored C(0, n, 0)C(0, n ′ , 0) , amplitudes for n = n ′ . Nevertheless computation of an effective potential only in terms of C(0, 0, 0) for the (3 + 1)D theory should be interesting in its own right. The analysis in this paper is done in the Yang-Mills approximation which is valid so long as a typical energy scale in the problem, E satisfies E ≪ 1/ √ α ′ . We have considered an intersecting configuration such that θ/(2πα ′ ) = q(fixed) with θ, α ′ → 0. We thus define our loop integrals with a cutoff Λ such that √ q ≪ Λ ≪ 1/ √ α ′ . The theory that we are considering is finite in the UV so that when √ q ≪ Λ the amplitudes are independent of the cutoff Λ. This is what has been demonstrated for all the loop amplitudes. The transition temperature at which the tachyon becomes massless

JHEP04(2017)109
(Hagedorn temperature). The massive string modes however become relevant when we consider a finite angle θ. In this case there is no clear separation of the critical temperature scale (T c ) and the Hagedorn scale T H . When θ = π we have a brane-antibrane system. This was analyzed at finite temperature in [45] and further studied in [46].
The holographic model of BCS superconductor studied in [2] required two stacks of intersecting flavor D8 branes so that the instantons on each stack can source an U(1). This U(1) introduces a finite density of baryons in the boundary theory. It is thus desirable that the computations in a simplified setup done here be extended to the case of stacks of intersecting branes. One can realise two intersecting stacks of N f D-branes by considering an U(2N f ) theory which is then broken by a vacuum expectation value of a scalar to U(N f ) × U(N f ). The tachyons in such a non-BPS setup are the strings which stretch from one stack to the other. The 2N 2 f such strings transform as two copies of bifundamental tachyons under U(N f ) × U(N f ). A finite temperature analysis can be done along the same lines as discussed in this paper. It would also be interesting to look into various aspects of this system from the point of view of gauge/gravity duality.
We would like to address these issues in a future work.

A Summary of notations
Matsubara sums [47]. (Also see appendix F of [1]) Sum over bosonic frequencies: Sum over fermionic frequencies:

B Dimensional reduction
Following [48] and [49] we give below the details of the dimensional reduction of D = 10, N = 1 to D = 4. The action for D = 10, N = 1 SYM, 1

JHEP04(2017)109
To proceed with the reduction, the gamma-matrices are first decomposed as, In this representation, The dimensionally reduced action can now be written as, where the Bosonic part if the action is, and the Fermionic part is, The α and β matrices satisfy An explicit representation of these matrices is given by the following We end this section by writing down the explicit forms of the 4D γ-matrices

JHEP04(2017)109
C Propagators and vertices for computation of two point C(m, n, k) amplitude
We also have additional vertices as V 2 2 for the A 2 i (i = 1) fields in place of Φ 2 I /Φ I 2 fields.
We also have additional vertices as V 3 2 for the A 3 i (i = 1) fields in place of Φ 3 I /Φ I 3 fields.

C.2 Fermions
The terms in the action (B.12) that contain C(m, n, k) are where · · · contain other similar terms involving the other three sets (λ 1 2 , λ 2 3 ), (λ 1 3 , λ 2 2 ), (λ 1 4 , λ 2 1 ). Inserting the mode expansions of the fields we can rewrite the action as The required vertices are now identified as JHEP04(2017)109 and similarly there are three other sets of vertices coming from the other terms in the action.

D Modified propagators and tachyon amplitude
As mentioned before for the Φ 3 I fields the mass terms in the propagators in the loop are easy to implement. We thus focus on the A 3 i fields. We first give some details of derivation of the corrected propagator for the A 3 i fields. The one-loop effective action is Here (i, j = 1, 2, 3), k 2 = (ω 2 m + k 2 x + k 2 ) and u i ≡ (1, 0, 0). The inverse of the operator O ij can be written in terms of the tensors δ ij , k i k j , u i k j + u j k i and u i u j . We thus write the inverse as

JHEP04(2017)109
The required vertices are now identified as vertices (There are five massless scalar fields corresponding to Φ 3 2,3 andΦ 3 I , (I = 1, 2, 3). Due to an unbroken SO(5) invariance in the theory the two point amplitudes for all these fields is same. We thus only consider the vertices here for Φ 3 2 .) We first write down the vertices containing only bosonic fields.