Witten index of BMN matrix quantum mechanics

Chi-Ming Chang

Yau Mathematical Sciences Center (YMSC), Tsinghua University, Beijing 100084, China
arXiv:2404.18442v1 [hep-th] 29 Apr 2024

Beijing Institute of Mathematical Sciences and Applications (BIMSA)

Beijing 101408, China

[email protected]

Abstract

We compute the Witten index of the Berenstein-Maldacena-Nastase matrix quan-

tum mechanics, which counts the number of ground states as well as the difference
between the numbers of bosonic and fermionic BPS states with nonzero spins. The
Witten index sets a lower bound on the entropy, which exhibits an N 2 growth that
predicts the existence of BPS black holes in M-theory, asymptotic to the plane wave
geometry. We also discuss a relation between the Witten index in the infinite N limit
and the superconformal index of the Aharony-Bergman-Jafferis-Maldacena theory.
Contents
1 Introduction 1

2 BMN index 3
2.1 BMN matrix quantum mechanics . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Witten index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Computation at weak coupling . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.4 Trivial vacuum sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5 Irreducible vacuum sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 Discussion 12

A Small black holes in AdS5 14

B Review of ABJM index 16

B.1 N = 8 superconformal index . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
B.2 Integral formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1 Introduction
The Berenstein-Maldacena-Nastase (BMN) matrix quantum mechanics [1] is a supersym-
metric gauged quantum mechanics with sixteen supercharges. It is a mass deformation of
the Banks-Fischler-Shenker-Susskind (BFSS) matrix quantum mechanics [2], and converges
to the BFSS matrix quantum mechanics when the dimensionless effective gauge coupling g
(normalized by the mass scale) approaches infinity.
The BMN matrix quantum mechanics participates in a very rich array of dualities. First,
by the BFSS conjecture [2], the BMN matrix quantum mechanics is dual to M-theory on
the uncompactified plane wave background [3], when the rank N of the matrices approaches
infinite with the gauge coupling g fixed. A stronger version of the conjecture states that,
at finite N and g, the BMN matrix quantum mechanics is dual to the Discrete Lightcone
Quantization (DLCQ) of M-theory on the plane wave background with N unit of light cone
momentum [4–6]. Finally, in the ’t Hooft limit (λ = g 2 N fixed and N → ∞), the BMN matrix
quantum mechanics is dual to M-theory on asymptotically plane wave backgrounds [7,8], for

1
instance, at large ’t Hooft coupling (λ ≫ 1), the vacuum states are dual to the Lin-Lunin-
Maldacena geometries [9, 10], and the deconfined phase at high temperature is dual to black
holes [11].
Witten index is a very powerful tool in the study of supersymmetric quantum mechanics
[12]. It is defined as a trace over the Hilbert space, weighted by the Boltzmann factor and
an additional insertion of the fermion parity operator. The trace is equal to the number of
bosonic ground states minus the number of fermionic ground states because there are equal
numbers of bosonic and fermionic states at each non-zero energy level of the Hamiltonian
due to supersymmetry. Hence, Witten index is independent of the temperature and any
deformation of the Hamiltonian as long as the deformation is gentle enough that does not
change the Hilbert space. When the theory has global symmetries that commute with the
Hamiltonian, the Witten index could be further refined by incorporating flavor fugacities
and receiving contributions from not only ground states but also BPS states with non-zero
symmetry charges.
The Witten index of the BMN matrix quantum mechanics can be computed in the
weak coupling (large mass) limit due to the protection from supersymmetry. In this limit,
the theory divides into superselection sectors associated with each supersymmetric vacuum,
that contains a set of free bosonic and fermionic harmonic oscillators [13]. The spectrum
in each superselection sector is thereby generated by acting on the supersymmetric vacuum
state with arbitrary numbers of creation operators of the oscillators. The result could be
assembled into a matrix integral formula for the Witten index (see Section 2.3), that is
similar to the formula for the superconformal index of the N = 4 super-Yang-Mills (SYM)
theory [14]. The result is different from the Witten index of the BFSS matrix quantum
mechanics [15, 16], as the Hilbert space at infinite coupling (zero mass) is different from the
one at finite coupling due to the flat directions in the classical potential.
The Witten index restricted in different superselection sectors exhibits very different
behaviors in the large N limit. We will focus on the sectors associated with the trivial vacuum
(the maximally redicuble vacuum) and the irreducible vacuum. In the trivial vacuum sector,
the matrix integral can be evaluated directly as a power series of the fugacities at finite N ,
or using a saddle point approximation in the large N and small chemical potentials limit.
Both results exhibit an N 2 growth in the entropy. This strongly indicates the existence of
BPS black holes, analogous to the non-BPS black holes [11], in the bulk dual of the BMN
matrix quantum mechanics (see Section 2.4). In the irreducible vacuum sector, BMN matrix
quantum mechanics describes a single M2-brane with N units of lightcone momentum along
the M-theory circle [1]. On the other hand, a system of coinciding M2-branes is described
by the Aharony-Bergman-Jafferis-Maldacena (ABJM) theory [17]. We find that the Witten
index in the irreducible vacuum sector in the large N limit is equal to the superconformal

2
index of the U(1)1 × U(1)−1 ABJM theory [18] up to a divergent factor (see Section 2.5).

2 BMN index

2.1 BMN matrix quantum mechanics

The Hamiltonian of the BMN matrix quantum mechanics is [19]

" 9 9
#
1X I 2 1 X I J2 1 T I I
H = R Tr (P ) − 6 [X , X ] − 3 Ψ γ [X , Ψ]
2 I=1 4ℓP I,J=1 2ℓP
" 3 9
# (2.1)
R  µ 2 X  µ 2 X µ 2µ
+ Tr (X i )2 + (X m )2 + i ΨT γ 123 Ψ + i ϵijk X i X j X k ,
2 3R i=1 6R m=4 4R 3Rℓ3P

where X I for I = 1, · · · , 9 are nine N ×N bosonic matrices, Ψ with a spinor index supressed
denotes sixteen N × N fermionic matrices, and P I are the conjugate momenta of X I . X i
and X m for i = 1, 2, 3 and m = 4, · · · , 9 are the first three and last six components of X I .
We will focus on the U(N ) gauge group, and the gauge symmetry acts on the Hermitian
matrices X I , Ψ, and P I as conjugation by unitary matrices. The parameters R, ℓP and
1/µ have dimension of length. ℓP could be obsorbed by field and parameter redefinitions:
X I = ℓP X
e I and R = ℓ2 R,
P
e so that Xe I are dimensionless and 1/Re has dimension of length.
Therefore, there is only a single dimensionless coupling
R3
g2 = . (2.2)
µ3 ℓ6P

At infinite coupling g → ∞ (µ = 0), the BMN matrix quantum mechanics becomes the
Banks-Fischler-Shenker-Susskind (BFSS) matrix quantum mechanics [2], with the Hamilto-
nian given by the first line of (2.1) that preserves SO(9) symmetry. X I , P I are in the vector
representation and Ψ is in the spinor representation. The potential terms on the second line
of (2.1) breaks the SO(9) to SO(3) × SO(6).

2.2 Witten index

The BMN matrix quantum mechanics has SU(2|4) supersymmetry, which contains the
SO(3)×SO(6) symmtry as the maximal bosonic subgroup. The SU(2|4) supersymmetry con-
tains sixteen supercharges Qm m † ∼
α and (Qα ) , where α = ± is the spinor index of SU(2) = SO(3).
The supercharges have the anti-commutator [13]
n † µ 2µ α m
{Qm m α
α , (Qβ ) } = 2δn δβ H + ϵijk (σ k )βα δnm M ij − δ R , (2.3)
3 3 β n
3
where M ij and Rnm are the rotation generators of SO(3) and SU(4), respectively. Let us pick
a supercharge Q := Q4− . The anti-commutator of Q with its Hermitian conjugate Q† is

2µ 12 2µ 4 2µ 12 µ
2∆ := {Q, Q† } = 2H − M − R4 = 2H − M − (M 45 + M 67 + M 89 ) , (2.4)
3 3 3 3
where we expand R44 in terms of the Cartan generators M 45 , M 67 , M 89 of SO(6), which
correspond to the rotations along the three orthogonal two-planes in R6 . It is convenient to
assemble the 5 Cartan generators of SU(2|4) into a vector as
 
12H 12 45 67 89
, 4M , 2M , 2M , 2M . (2.5)
µ

The supercharge Q has the Cartan charges (1, −2, 1, 1, 1).

Let us consider the thermal partition function
12 −∆ 45 −∆ 67 −∆ 89
Z = Tr Ω , Ω := e−β∆−2ωM 1M 2M 3M
. (2.6)

The Boltzmann factor Ω anti-commutes with the supercharge Q if the chemical potentials
satisfy the linear relation

∆1 + ∆2 + ∆3 − 2ω = 2πi mod 4πi . (2.7)

The Witten index is defined by

h 12 45 12 67 12 89
i
I = Z (2.7) = Tr (−1)F e−β∆−∆1 (M +M )−∆2 (M +M )−∆3 (M +M ) , (2.8)

where we identify 2M 12 with the fermion number F beause the fermion fields Ψ’s have half-
integer M 12 -eigenvalues. The Witten index I is independent of the inverse temperature β,
because the states with nonzero ∆-eigenvalues are all paired up by the action of the super-
charge Q and their contributions to the Witten index cancel. Since all the explicit coupling
constant g dependence is inside the Hamiltonian, the Witten index is also independent of
g away from the infinite coupling (g = ∞) point where the dimension of the Hilbert space
becomes uncountable due to continuums in the spectrum.

2.3 Computation at weak coupling

Since the Witten index is independent of the coupling constant g, it can be computed for
BMN matrix quantum mechanics in the weak coupling (small g) limit, or equivalently, in the
large µ limit. We will begin by reviewing the computation of the weak coupling spectrum
in [13], and then use the results to compute the Witten index.

4
 In the large µ limit, the bosonic potential becomes very steep, and we could expand
the Hamiltonian about the minima of the potential (classical supersymmetric vacua). The
bosonic potential can be written in a manifestly positive form as
 
R µ i −3 ijk j k
2 1 m n 2 1 m i 2
 µ 2
m 2
V = Tr X + ℓP iϵ X X + 2 (i[X , X ]) + 2 (i[X , X ]) + (X ) .
2 3R 2ℓP ℓP 6R
(2.9)
The classical supersymmetric vacua are given by [1]
µℓ3P i
Xm = 0 , Xi = J , (2.10)
3R
where J i is a N -dimensional matrix representation of SU(2), i.e. [J i , J j ] = iϵijk J k . Any
N -dimensional representation of SU(2) can be written as a direct sum of irreducible repre-
sentations, and is labeled by the partition of the integer N ,
K
X
N= nk Nk . (2.11)
k=1

More explicitly, the (k, l)-th block of J i is given by

i
Jkl = δkl Inl ⊗ Jli , (2.12)

where Jli is the Nl -dimensional irreducible representation, and Inl is a nl ×nl identity matrix.
In the trivial vacuum (K = 1, n1 = N , and N1 = 1), the U(N ) gauge symmetry is preserved.
In the nontrivial vacua, the U(N ) gauge symmetry is broken to

U(n1 ) × · · · × U(nK ) . (2.13)

The classical vacua correspond to degenerate ground states, which are protected quantum
mechanically and have vanishing energy at finite coupling g > 0 [20–22].
The expansion about the classical vacua was studied in [13]. In the small g limit, the
Hamiltonian reduces to a free quadratic term plus higher order interaction terms, which are
suppressed by powers of g relative to the quadratic term. Hence, the theory divides into
superselection sectors associated with each classical vacuum and labeled by the partition
(2.11) of N . Each sector contains a single ground state, and excited states given by acting
creation operators of a set of bosonic and fermionic harmonic oscillators. In the (k, l)-th
block, the bosonic oscillators are

(αkl )jm , (βkl )jm , (xakl )jm , (2.14)

and fermionic oscillators are

(χIkl )jm , (ηI,kl )jm , (2.15)

5
which are all nk ×nl matrices, and transform in the bi-fundamental representation of U(nk )×
U(nl ). The ranges of j for each oscillator are given in Table 2 in [13]. For example, the matrix
a
Xkl has the expansion
1
2
(Nk +Nl )−1 j
X X Nk Nl
a
Xkl = (xakl )jm ⊗ Yjm , (2.16)
j= 12 |Nk −Nl | m=−j

Nk Nl
where Yjm is a Nk × Nl matrix as a spin-j representation in the tensor product of the
Nk −1
and spin- Nl2−1 representations.



spin- 2
We focus on BPS excited states, which are given by acting creation operators of the BPS
letters, the oscillators satisfying the BPS condition ∆ = 0. The BPS letters are listed in
Table 1.

letter charges index

β (4j, 4j, 0, 0, 0) e−2jω
(4j + 2, 4j, 2, 0, 0) e−2jω−∆1
x (4j + 2, 4j, 0, 2, 0) e−2jω−∆2
(4j + 2, 4j, 0, 0, 2) e−2jω−∆3
∆1 ∆ ∆
− 22 − 23
χ (4j + 3, 4j, 1, 1, 1) e−2jω− 2
∆1 ∆2 ∆3
(4j + 1, 4j, 1, 1, −1) e−2jω− 2 − 2 + 2
∆1 ∆2 ∆3
η (4j + 1, 4j, 1, −1, 1) e−2jω− 2 + 2 − 2
∆1 ∆2 ∆3
(4j + 1, 4j, −1, 1, 1) e−2jω+ 2 − 2 − 2

Table 1: The BPS letters and their charges (in the vector form (2.5)) and indices.

The single-letter partition function for the BPS letters is

1
2
(Nk +Nl )−1
X
zkl (ω, ∆i ) = e−2jω (e−∆1 + e−∆2 + e−∆3 )
j= 12 |Nk −Nl |
1 1
2
(Nk +Nl ) 2
(Nk +Nl )− 32
X X ∆1 ∆ ∆
− 22 − 23
+ e−2jω + e−2jω− 2 (2.17)
j= 21 |Nk −Nl |+1 j= 12 |Nk −Nl |− 21
1
2
(Nk +Nl )− 21
X ∆1 ∆ ∆ ∆1 ∆ ∆ ∆1 ∆ ∆
− 22 + 23 + 22 − 23 − 22 − 23
+ e−2jω (e− 2 + e− 2 +e 2 ).
j= 12 |Nk −Nl |+ 21

6
Imposing the relation (2.7) among the chemical potentials, we find the single-letter index

ιkl (∆i ) = zkl (ω, ∆i ) (2.7)

1
(Nk +Nl )−1
2 X (2.18)
= (−1)2j+1 e−j(∆1 +∆2 +∆3 ) (1 − e−∆1 )(1 − e−∆2 )(1 − e−∆3 ) + δNk ,Nl .
j= 21 |Nk −Nl |

The multi-letter index is given by the plethystic exponential of the single-letter index ιkl (∆i ).
The Witten index in the superselection sector is given by further imposing the gauge invari-
ance under the unbroken subgroup (2.13) by matrix integrals,
Z YK
" ∞ K #
XX 1 †m
InBMN
i ;Ni
= [dUk ] exp ιkl (m∆i )Tr Uk Tr Ulm , (2.19)
k=1 m=1 k,l=1
m

where Uk is a nk × nk unitary matrix. The total Witten index is given by

X
I BMN = InBMN
i ;Ni (2.20)
ni ,Ni ∈Z>0 , (2.11)

2.4 Trivial vacuum sector

Let us consider the Witten index in the trivial vacuum sector given by K = 1, n1 = N ,
N1 = 1. The integral formula (2.19) reduces to
( ∞ )
Z X 1 − (1 − e−m∆1 )(1 − e−m∆2 )(1 − e−m∆3 )
INBMN
;1 = [dU ] exp Tr U †m Tr U m . (2.21)
m=1
m

The same index INBMN ;1 was studied in [23, 24] but with the interpretation as the su-
perconformal index of the N = 4 SYM truncated to the BMN sector, that contains only
fundamental fields invariant under the chiral SU(2)R rotation. It was argued in [24], follow-
ing a similar analysis in [25,26], that in the large N limit with small and fixed ∆i , the matrix
integral (2.21) has a saddle point with the eigenvalue distribution
3
ρ(α) = (π 2 − α2 ) for α ∈ (−π, π) . (2.22)
4π 3
The saddle point contributes to the index as

3N 2
log INBMN
;1 = − ∆1 ∆2 ∆3 . (2.23)
2π 2

7
Let us consider the Legendre transform of log INBMN
;1 given by the extremization

BMN 12 45 67 89
SN ;1 (M , M , M , M )
= ext log INBMN 12
+ M 45 ) + ∆2 (M 12 + M 67 ) + ∆3 (M 12 + M 89 ) ,
 
;1 + ∆1 (M
∆i
r (2.24)
2(M 12 + M 45 )(M 12 + M 67 )(M 12 + M 89 )
= 2π ,
3N 2
that is valid in the large N limit with ϵi := (M 12 + M 2i+2,2i+3 )/N 2 (for i = 1, 2, 3) held fixed
BMN
and small ϵi ≪ 1. The function SN ;1 sets a lower bound on the entropy of the BPS states;
BMN
however, for simplicity, we will also refer to SN ;1 as the entropy. We find that the entropy
BMN √ 2 2
SN ;1 ∼ ϵ1 ϵ2 ϵ3 N exhibits N scaling, which implies that the theory is in a deconfined
phase and should correspond to a BPS black hole in the bulk dual.
A very subtle point of the above analysis is that the contribution from the non-trivial
saddle (2.22) is negative (2.23), which is smaller than the contribution from the trivial saddle,
corresponding to a confined phase, with a constant eigenvalue distribution
1
ρ(α) = . (2.25)
2π
However, the entropy (2.24) contributed from this saddle is of order N 2 , much larger than the
entropy from the trivial saddle. Hence, the deconfined phase is subdominant to the confined
phase in the grand canonical ensemble, but dominant in the microcanonical ensemble. The
deconfined phase has a negative specific heat C (or more precisely the susceptibility) given
by1
d2 log INBMN
;1 9N 2 3
C = ∆2 = − ∆ < 0, (2.26)
d∆2 π2
where we specialize chemical potentials as ∆ := ∆1 = ∆2 = ∆3 . These behaviors are similar
to the small black holes in AdS5 [26], which we review in Appendix A.
An important question is whether the N 2 scaling of the entropy persists at finite ϵi , and
if so, whether the deconfined phase also dominates in the grand canonical ensemble. We
provide positive evidence for the N 2 scaling from evaluating the BMN index (2.21) as an
expansion of the fugacities to high powers. In the microcanonical ensemble, the entropy is
defined as the log of the coefficients in the expansion of the index
X
SBMN (j) = log |dBMN
j | , I BMN
N ;1 = dBMN
j tj , t2 = e−∆1 = e−∆2 = e−∆3 , (2.27)
j

where the quantum number j is the following linear combination of angular momenta

j = 6M 12 + 2(M 45 + M 67 + M 89 ) . (2.28)
1
I thank Sunjin Choi for a discussion on this point.

8
For inspecting the the growth of the degeneracy dBMN
j in the large N limit with j/N 2 fixed,
we define
sBMN j/N 2 = N −2 log |dBMN


j |. (2.29)
Let us set j = N 2 ,2 and plot log |dBMN
j | and sBMN against N 2 in Figure 1. The BMN index
looks to achieve convergence when N ≳ 6, as

log |dBMN
j | ∼ 0.21 × N 2 . (2.30)

For comparison, we perform the same analysis for the superconformal index of the N = 4
SYM, which is given by the matrix integral [14]
 −m∆1 −m∆2 )(1−e−m∆3 )

Z X ∞
1 − (1−e (1−e)(1−e
−mω1 )(1−e−mω2 )

IN =4 = [dU ] exp Tr U †m Tr U m , (2.31)

m=1
m 

with the constraint ∆1 + ∆2 + ∆3 − ω1 − ω2 = 2πi. We expand the index as

X
IN =4 = dN
j t , t2 = e−∆1 = e−∆2 = e−∆3 , t3 = e−ω1 = e−ω2 .
=4 j
(2.32)
j

Again, we consider the large N limit with j/N 2 fixed and define

sN =4 (j/N 2 ) = N −2 log |dN

j
=4
|. (2.33)

The entropy sN =4 (j/N 2 ) can be computed exactly in the grand canonical ensemble by a
saddle point approximation [27–29]. When j = N 2 , we have sN =4 (1) = 0.357. The plots of
log |dN
j
=4
| and sN =4 against N 2 are given in Figure 1. We can see that sN =4 converges to
the asymptotic value already at N ∼ 5.

0.5

20 0.4

15 0.3

10 0.2

5 0.1

0 0.0
2 4 6 8 10 0 2 4 6 8 10

Figure 1: The log(|dBMN

j |), sBMN (blue) and log(|dN j
=4
|), sN =4 (orange) at j = N 2 with
increasing N . The green line is the asymptotic value for sN =4 .
2
There is a small subtlety that the states in the BMN sector all have even j. Therefore, when N is odd,
we consider the average 21 (log |dBMN BMN 2
j−1 | + log |dj+1 |) for j = N .

9
2.5 Irreducible vacuum sector

Let us consider the Witten index in the sector given by K = 1, n1 = 1, N1 = N corresponding

to an irreducible representation of SU(2). The integral formula (2.19) reduces to
( ∞ )
X 1  −m∆1 −m∆2 −m∆3 −mN (∆1 +∆2 +∆3 )
BMN (1 − e )(1 − e )(1 − e )(1 − e )
I1;N = exp 1− .
m=1
m 1 − e−m(∆1 +∆2 +∆3 )
(2.34)
In this sector, the BMN matrix model describes a single M2-brane carrying N units of
lightcone momentum along the M-theory circle and wrapping an S2 in the plane wave ge-
ometry [1]. On the other hand, the ABJM theory with U(1)1 × U(1)−1 gauge group in
radial quantization also describes a single M2-brane wrapping an S2 . Hence, we expect some
relations between the BMN index and the superconformal index of ABJM theory [18].
In Appendix B, we review the N = 8 ABJM index. It depends on five chemical potentials
associated with one spin and four R-charges of the superconformal symmetry OSp(4|8), and
with one linear between the chemical potentials. However, only the subgroup SU(2|4) ⊂
OSp(4|8) is manifest in the BMN matrix quantum mechanics, whose Witten index depends
on four chemical potentials with one linear relation. Hence, to compare the BMN index and
the ABJM index, one needs to first figure out a relation between the chemical potentials.

Relation between the chemical potentials

Under the embedding SU(2|4) ,→ OSp(4|8), the supercharges in SU(2|4) are related to the
supercharges in OSp(8|4) by
r r
n µ 2n−1 2n n † µ 2n−1,α
Qα = (Qα + iQα ) , (Qα ) = (S − iS2n,α ) , (2.35)
3 3

where n = 1, · · · , 4 and QIα and SIα are the supercharges and the conformal supercharges
in OSp(8|4). Plugging this into (B.2), we find the anti-commutator

n † 2µ m β 2µ α m
{Qm m β
α , (Qβ ) } = 2δn δα H − δ J − δ R , (2.36)
3 n α 3 β n
where Rnm are the generators of SU(4) ⊂ SO(8). This precisely matches with the anti-
commutator (2.3) in SU(2|4), with the identification Jαβ = − 12 ϵijk (σ i )βα M jk . The Cartan

10
generators in SU(2|4) and osp(8|4) are related by
3 1
H = D − (M12 + M34 + M56 + M78 ) , M 12 = J−− ,
µ 4
1 1
R11 = (3M12 − M34 − M56 − M78 ) , R22 = (−M12 + 3M34 − M56 − M78 ) ,
4 4
3 1 12 34 56 78 1
R3 = (−M − M + 3M − M ) , R4 = (−M12 − M34 − M56 + 3M78 ) ,
4
4 4
(2.37)
where M are the SO(8) generators. The Cartan generaotrs M , M , M of SO(6) ∼
IJ 45 67 89
=
SU(4) are related to the Cartan generators of SO(8) by
1
M 45 = (M12 − M34 − M56 + M78 ) ,
2
1
M 67 = (−M12 + M34 − M56 + M78 ) , (2.38)
2
1
M 89 = (−M12 − M34 + M56 + M78 ) .
2
Substituting the relations (2.37) and (2.38) into the formula (2.8) of the Witten index, we
find the following relation between the fugacities in (2.8) and (B.6),

∆1 − ∆2 − ∆3 −∆1 + ∆2 − ∆3
γ1 = , γ2 = ,
2 2 (2.39)
−∆1 − ∆2 + ∆3 ∆1 + ∆2 + ∆3
γ3 = , γ4 = .
2 2

Matching the indices in the infinite N limit

Now, we are ready to compare the BMN index and the ABJM index. The U(1)1 × U(1)−1
ABJM index is given by the integral,
Z π
ABJM
X −1n
|n−ñ| dαdα̃ i(nα−ñα̃)
I1,1 = y3 x
2
2
e
n,ñ∈Z −π (2π)
" ∞
# (2.40)
XX 1 m|n−ñ|
× exp x fs (xm , y1m , y2m )e−smi(α−α̃) ,
s=± m=1
m

where the functions f± (x, y1 , y2 ) are given in (B.11). The fugacities x, y1 , y2 , y3 are defined
in (B.9). The integral and summation can be easily performed after the change of the
integration variables,
α + α̃
α− = α − α̃ , α+ = . (2.41)
2

11
The result is
∞
1 e−mγ4
X 
−m
X
ABJM (s1 γ1 +s2 γ2 +s3 γ3 +s4 γ4 )
I1,1 = exp s4 e2 , (2.42)
m=1
m e−2mγ4 − 1 si =±
s1 s2 s3 s4 =−1

Substituting the relation (2.39) between the fugacities into the above formula, we find
( ∞ )
X 1  −m∆1 −m∆2 −m∆3
ABJM (1 − e )(1 − e )(1 − e )
I1,1 = exp 2− −m(∆ +∆ +∆ )
. (2.43)
m=1
m 1−e 1 2 3

which agrees with the BMN index (2.34) in the N → ∞ limit up to an divergent factor.
The divergent in the ABJM index is due to the specialization of the chemical potentials
(2.39) from four γi ’s to three ∆i ’s. A potential explanation of the divergent factor in the
relation between the BMN and ABJM indices is that the M2-brane theory on the sphere has
infinitely many degenerate vacua that are not distinguished by the three chemical potentials,
∆i . However, the BMN quantum mechanics only includes one of these vacua that is chosen
by the value of N . Hence, in this sense the infinity is due to the sum over N , which in
the ABJM theory would correspond to tracing over different states with the corresponding
chemical potential being set to zero.3
The full ABJM index, with all four chemical potentials γi turned on, could potentially
be recovered by performing a sum over the BMN indices with different N weighted by an
additional chemical potential. We leave this for future work.

3 Discussion
In this paper, we compute the Witten index for the BMN matrix quantum mechanics, which
counts (with signs) the number of the BPS states. The computation relies on the non-
renormalization of the Witten index and is performed at weak coupling, where the theory is
divided into superselection sectors associated with each supersymmetric vacuum. The result
is a sum of contributions from all superselection sectors, with each term being a matrix
integral. In the trivial vacuum sector, we evaluate the matrix integral, and show that it
contributes an N 2 growth to the entropy. In the irreducible vacuum sector, we find a novel
relation between the BMN index and the ABJM index.
Let us discuss some open problems/future directions:

• The N 2 growth of the entropy in the trivial vacuum sector indicates that there should
exist BPS black holes in M-theory, asymptotic to the plane wave geometry. However,
3
I think Juan Maldacena for suggesting this explanation.

12
 no solution of this sort in the eleven-dimensional supergravity is currently known.
Different BPS black hole geometries, with different horizon topologies, could potentially
dominate in different regions of the chemical potentials. It is important to extend the
saddle point analysis of the Witten index in the trivial vacuum sector to finite ϵi and to
other vacuum sectors, as it could shed light on the phase structure of these BPS black
holes. The phase structure of non-BPS black holes at finite temperature were explored
in [30–36] on the matrix quantum mechanics side by lattice simulations, and [11] on
the gravity side.

• The relation between the BMN and ABJM indices can be extended to the case of
multiple M2-branes. On the BMN side, this involves the vacua corresponding to direct
sums of multiple copies N -dimensional irreducible SU(2) representations. Understand-
ing this generalization may require finding a precise relation between the configurations
of M2-branes in these two setups, in particular, the relation between the N of the BMN
quantum mechanics and a U(1) charge in the ABJM theory.

• Supersymmetric localization of the BMN matrix quantum mechanics was developed

in [37], and used in the study of the 1/4-BPS sector [38, 39] and the transverse M5-
branes [40, 41]. The same techniques may be used to obtain a path integral derivation
of the matrix integral formulae (2.19) with (2.18) for the Witten index.

• Following [14, 42, 43] and more recently [44, 45, 23, 46, 24, 47], we could study the Q-
cohomology of the BMN matrix quantum mechanics.4 The Q-cohomology classes
correspond one-to-one with the BPS states, and the Euler characteristic of the Q-
cohomology equals the Witten index. The Q-cohomology can be straightforwardly
computed at weak coupling, although there is no proof or argument that the Q-
cohomology for BMN matrix quantum mechanics remains invariant when varying the
coupling constant g.

Acknowledgments
I am grateful to Sunjin Choi, Shota Komatsu, Ying-Hsuan Lin, and Jorge E. Santos for
inspiring discussions, and especially to Juan Maldacena for introducing me to this problem
and generously sharing many of his important insights. CC is partly supported by the
National Key R&D Program of China (NO. 2020YFA0713000). This research was supported
in part by grant NSF PHY-2309135 to the Kavli Institute for Theoretical Physics (KITP). We
4
The supercharge Q is defined above (2.4).

13
thank the hospitality of the KITP Program “What is String Theory? Weaving Perspectives
Together” and Conference “Spacetime and String Theory”.

A Small black holes in AdS5

In this appendix, we review the discussions about small AdS black holes in [26]. The BPS
black hole in AdS5 has a partition function given by the Euclidean gravity path integral as
N 2 ∆1 ∆2 ∆3
log Z = , ∆1 + ∆2 + ∆3 − ω1 − ω2 = 2πi . (A.1)
2 ω1 ω2
Let us consider the specialization to only one fugacity x = e−β as

∆i = 2β + 2πi , ωi = 3β + 2πi . (A.2)

Since the thermal AdS phase has a free energy of order 1, the black hole phase is in the
region  2
N (2β + 2πi)3

Re (log Z) = Re > 0. (A.3)
2 (3β + 2πi)2
We plot this region in blue on the |x|-ϕ plane in Figure 2, where x = e−β = |x|eiϕ .
The large/small black holes are in the region of positive/negative specific heat C (sus-
ceptibility) given by
2
 
2 d log Z
C = Re βR , (A.4)
dβR2
where βR = Re β = − log |x|. We plot the region of positive specific heat in orange in
Figure 2.
The entropy of the black hole is given by extremizing the function

S(β; Q, J) = log Z + Q1 ∆1 + Q2 ∆2 + Q3 ∆3 + J1 ω1 + J2 ω2 Qi =Q, Ji =J, (A.2)

2
N (2β + 2πi) 3 (A.5)
= + 3Q(2β + 2πi) + 2J(3β + 2πi) .
2 (3β + 2πi)2
This function has three extrema. We pick the one that gives a positive real part of S.
Imposing the condition that the imaginary part of S vanishes, we find
N2 2
 2 
3 N
Im S = 0 ⇒ Q + J = + 3Q (3Q2 − N 2 J) , (A.6)
2 2
which admits one solution with Q, J > 0. The entropy at the chosen extremum is
p
S(Q) = 2π 3Q2 − N 2 J , (A.7)

14
 3.0

2.5

2.0

1.5

1.0

0.5

0.0
0.0 0.2 0.4 0.6 0.8 1.0

Figure 2: The blue and orange regions are given by (A.3) and C > 0 for C given in (A.4).
The red curve is given by the function β(Q) for Q varying from 0 to ∞.

which is always positive for Q, J > 0 satisfying the condition (A.6).

Let us write the chosen extremum as β(Q), after substituting J using (A.6). We plot β(Q)
on the |x|-ϕ plane as the red curve in Figure 2 for Q varing from 0 to ∞. The small charge
limit Q → 0 is at (|x|, ϕ) = (1, π), and the large charge limit Q → ∞ is at (|x|, ϕ) = (1, 2π
3
).
There is an intersection between the curve β(Q) and the phase boundary at

N2 √
Q∗ = (1 + 33) . (A.8)
24
We see that when increasing Q from 0 to ∞, the red curve is first inside the white region
(log Z < 0 and c < 0) corresponding to a small black hole phase that is subdominant to
thermal AdS in the grand canonical ensemble. Then, the red curve enters the orange region
(log Z < 0 and c > 0) corresponding to a large black hole phase that is also subdominant. Fi-
nally, the red curve enters the orange and blue region (log Z > 0 and c > 0) corresponding to
a large and dominant black hole phase. While the small black holes are always subdominant,
they still carry macroscopic entropy (A.7).

15
B Review of ABJM index

B.1 N = 8 superconformal index

The N = 8 ABJM theory has superconformal symmetry OSp(8|4). Let us denote the
supercharges and the conformal supercharges in OSp(8|4) by

QIα , SIα , (B.1)

where I = 1, · · · , 8 is the vector index of SO(8) and α = ± is the spinor index of SU(2) ⊂
Sp(4). We have the anti-commutator

{QIα , SJβ } = δ IJ δαβ D − δ IJ Jβα − iδβα MIJ , (B.2)

where D is the dilatation charge, Jβα are the generators of SU(2), and MIJ are the generators
of SO(8). Let us define
1
∆ := {Q7− + iQ8− , S7− − iS8− } = D − J − M78 , (B.3)
2
where J := J−− . Let us consider the partition function
12 −γ M34 −γ M56 −γ M78
Z = Tr Ω , Ω := e−β∆−2ωJ−γ1 M 2 3 4
. (B.4)

The supercharge Q7− + iQ8− anti-commutes with the Boltzmann factor Ω if

γ4 − ω = 2πi mod 4πi . (B.5)

We define the superconformal index as

h 12 34 56 78
i
I = Z (B.5) = Tr (−1)F e−β∆−γ1 M −γ2 M −γ3 M −γ4 (2J+M ) , (B.6)

where we identify 2J with the fermion number F , since all the fermionic states have half-
integer spin J. By the standard arguments, the superconformal index I is independent of
the inverse temperature β.

16
B.2 Integral formula

The superconformal index for the U(N )k × U(N )−k ABJM theory was computed in [18],56

X − k PN ni Z π Y N
ABJM 1 dαi dα̃i ik(ni αi −ñi α̃i )
IN,k = 2
y3 2 i=1
2
e
(N !) n ,ñ ∈Z −π i=1 (2π)
"i i ∞
# (B.8)
X 1
× exp −ωϵ0 + f (xm , y1m , y2m , eimαi , eimα̃i ) ,
m=1
m

where the fugacities x and yi are

x = e−γ4 , y1 = e−γ1 , y2 = e−γ2 , y3 = e−γ3 . (B.9)

The ϵ0 and f are X X

ϵ0 = |ni − ñj | − (|ni − nj | + |ñi − ñj |) , (B.10)
i,j i<j

and
X X
f (x, y1 , y2 , eiαi , eiα̃i ) = − x|ni −nj | e−i(αi −αj ) + f+ (x, y1 , y2 )x|ni −ñj | e−i(αi −α̃j )
i̸=j i,j
X X
− x|ñi −ñj | e−i(α̃i −α̃j ) + f− (x, y1 , y2 )x|ñi −nj | e−i(α̃i −αj ) ,
i̸=j
 r
i,j
   (B.11)
√
r
1 1 y1 y2 3 1
f+ (x, y1 , y2 ) = x 2 + −x 2 y1 y2 + √ ,
1 − x2 y2 y1 y1 y2
f− (x, y1 , y2 ) = f+ (x, y1 , y2−1 ) .

References
[1] D. E. Berenstein, J. M. Maldacena, and H. S. Nastase, Strings in flat space and pp
waves from N=4 superYang-Mills, JHEP 04 (2002) 013, [hep-th/0202021].

[2] T. Banks, W. Fischler, S. H. Shenker, and L. Susskind, M theory as a matrix model: A

Conjecture, Phys. Rev. D 55 (1997) 5112–5128, [hep-th/9610043].
5
In a flux sector labeled by {n1 , · · · , nN }, the U(N ) gauge group is broken to G{ni } = U(N1 )×U(N2 )×· · · .
We have the formula X 1 1 X
= , (B.7)
|G{ni } | N!
{ni } ni ∈Z

where |G{ni } | is the order of the Weyl group of G{ni } .

6
Note that our formula differs by the formula (3.11) in [18] by y3 → y3−1 .

17
 [3] J. Kowalski-Glikman, Vacuum States in Supersymmetric Kaluza-Klein Theory, Phys.
Lett. B 134 (1984) 194–196.

[4] L. Susskind, Another conjecture about M(atrix) theory, hep-th/9704080.

[5] A. Sen, D0-branes on T**n and matrix theory, Adv. Theor. Math. Phys. 2 (1998)
51–59, [hep-th/9709220].

[6] N. Seiberg, Why is the matrix model correct?, Phys. Rev. Lett. 79 (1997) 3577–3580,
[hep-th/9710009].

[7] N. Itzhaki, J. M. Maldacena, J. Sonnenschein, and S. Yankielowicz, Supergravity and

the large N limit of theories with sixteen supercharges, Phys. Rev. D 58 (1998) 046004,
[hep-th/9802042].

[8] J. Polchinski, M theory and the light cone, Prog. Theor. Phys. Suppl. 134 (1999)
158–170, [hep-th/9903165].

[9] H. Lin, O. Lunin, and J. M. Maldacena, Bubbling AdS space and 1/2 BPS geometries,
JHEP 10 (2004) 025, [hep-th/0409174].

[10] H. Lin and J. M. Maldacena, Fivebranes from gauge theory, Phys. Rev. D 74 (2006)
084014, [hep-th/0509235].

[11] M. S. Costa, L. Greenspan, J. Penedones, and J. Santos, Thermodynamics of the BMN

matrix model at strong coupling, JHEP 03 (2015) 069, [arXiv:1411.5541].

[12] E. Witten, Constraints on Supersymmetry Breaking, Nucl. Phys. B 202 (1982) 253.

[13] K. Dasgupta, M. M. Sheikh-Jabbari, and M. Van Raamsdonk, Matrix perturbation

theory for M theory on a PP wave, JHEP 05 (2002) 056, [hep-th/0205185].

[14] J. Kinney, J. M. Maldacena, S. Minwalla, and S. Raju, An Index for 4 dimensional

super conformal theories, Commun. Math. Phys. 275 (2007) 209–254,
[hep-th/0510251].

[15] P. Yi, Witten index and threshold bound states of D-branes, Nucl. Phys. B 505 (1997)
307–318, [hep-th/9704098].

[16] S. Sethi and M. Stern, D-brane bound states redux, Commun. Math. Phys. 194 (1998)
675–705, [hep-th/9705046].

[17] O. Aharony, O. Bergman, D. L. Jafferis, and J. Maldacena, N=6 superconformal

Chern-Simons-matter theories, M2-branes and their gravity duals, JHEP 10 (2008)
091, [arXiv:0806.1218].

18
[18] S. Kim, The Complete superconformal index for N=6 Chern-Simons theory, Nucl.
Phys. B 821 (2009) 241–284, [arXiv:0903.4172]. [Erratum: Nucl.Phys.B 864, 884
(2012)].

[19] J. M. Maldacena, M. M. Sheikh-Jabbari, and M. Van Raamsdonk, Transverse

five-branes in matrix theory, JHEP 01 (2003) 038, [hep-th/0211139].

[20] N. Kim and J. Plefka, On the spectrum of PP wave matrix theory, Nucl. Phys. B 643
(2002) 31–48, [hep-th/0207034].

[21] K. Dasgupta, M. M. Sheikh-Jabbari, and M. Van Raamsdonk, Protected multiplets of

M theory on a plane wave, JHEP 09 (2002) 021, [hep-th/0207050].

[22] N. Kim and J.-H. Park, Superalgebra for M theory on a pp wave, Phys. Rev. D 66
(2002) 106007, [hep-th/0207061].

[23] S. Choi, S. Kim, E. Lee, S. Lee, and J. Park, Towards quantum black hole microstates,
JHEP 11 (2023) 175, [arXiv:2304.10155].

[24] J. Choi, S. Choi, S. Kim, J. Lee, and S. Lee, Finite N black hole cohomologies,
arXiv:2312.16443.

[25] C. Copetti, A. Grassi, Z. Komargodski, and L. Tizzano, Delayed deconfinement and

the Hawking-Page transition, JHEP 04 (2022) 132, [arXiv:2008.04950].

[26] S. Choi, S. Jeong, and S. Kim, The Yang-Mills duals of small AdS black holes,
arXiv:2103.01401.

[27] A. Cabo-Bizet, D. Cassani, D. Martelli, and S. Murthy, Microscopic origin of the

Bekenstein-Hawking entropy of supersymmetric AdS5 black holes, JHEP 10 (2019)
062, [arXiv:1810.11442].

[28] S. Choi, J. Kim, S. Kim, and J. Nahmgoong, Large AdS black holes from QFT,
arXiv:1810.12067.

[29] F. Benini and E. Milan, Black Holes in 4D N =4 Super-Yang-Mills Field Theory,

Phys. Rev. X 10 (2020), no. 2 021037, [arXiv:1812.09613].

[30] S. Catterall and G. van Anders, First Results from Lattice Simulation of the PWMM,
JHEP 09 (2010) 088, [arXiv:1003.4952].

[31] Y. Asano, V. G. Filev, S. Kováčik, and D. O’Connor, The non-perturbative phase

diagram of the BMN matrix model, JHEP 07 (2018) 152, [arXiv:1805.05314].

19
[32] D. Schaich, R. G. Jha, and A. Joseph, Thermal phase structure of a supersymmetric
matrix model, PoS LATTICE2019 (2020) 069, [arXiv:2003.01298].

[33] Monte Carlo String/M-theory (MCSMC), MCSMC Collaboration,

G. Bergner, N. Bodendorfer, M. Hanada, S. Pateloudis, E. Rinaldi, A. Schäfer,
P. Vranas, and H. Watanabe, Confinement/deconfinement transition in the D0-brane
matrix model — A signature of M-theory?, JHEP 05 (2022) 096, [arXiv:2110.01312].

[34] D. Schaich, R. G. Jha, and A. Joseph, Thermal phase structure of dimensionally

reduced super-Yang–Mills, PoS LATTICE2021 (2022) 187, [arXiv:2201.03097].

[35] S. Pateloudis, G. Bergner, N. Bodendorfer, M. Hanada, E. Rinaldi, and A. Schäfer,

Nonperturbative test of the Maldacena-Milekhin conjecture for the BMN matrix model,
JHEP 08 (2022) 178, [arXiv:2205.06098].

[36] Monte Carlo String/M-theory (MCSMC) Collaboration, S. Pateloudis,

G. Bergner, M. Hanada, E. Rinaldi, A. Schäfer, P. Vranas, H. Watanabe, and
N. Bodendorfer, Precision test of gauge/gravity duality in D0-brane matrix model at
low temperature, JHEP 03 (2023) 071, [arXiv:2210.04881].

[37] Y. Asano, G. Ishiki, T. Okada, and S. Shimasaki, Exact results for perturbative
partition functions of theories with SU(2—4) symmetry, JHEP 02 (2013) 148,
[arXiv:1211.0364].

[38] Y. Asano, G. Ishiki, T. Okada, and S. Shimasaki, Emergent bubbling geometries in the
plane wave matrix model, JHEP 05 (2014) 075, [arXiv:1401.5079].

[39] Y. Asano, G. Ishiki, and S. Shimasaki, Emergent bubbling geometries in gauge theories
with SU(2—4) symmetry, JHEP 09 (2014) 137, [arXiv:1406.1337].

[40] Y. Asano, G. Ishiki, S. Shimasaki, and S. Terashima, Spherical transverse M5-branes

in matrix theory, Phys. Rev. D 96 (2017), no. 12 126003, [arXiv:1701.07140].

[41] Y. Asano, G. Ishiki, S. Shimasaki, and S. Terashima, Spherical transverse M5-branes

from the plane wave matrix model, JHEP 02 (2018) 076, [arXiv:1711.07681].

[42] L. Grant, P. A. Grassi, S. Kim, and S. Minwalla, Comments on 1/16 BPS Quantum
States and Classical Configurations, JHEP 05 (2008) 049, [arXiv:0803.4183].

[43] C.-M. Chang and X. Yin, 1/16 BPS states in N = 4 super-Yang-Mills theory, Phys.
Rev. D 88 (2013), no. 10 106005, [arXiv:1305.6314].

[44] C.-M. Chang and Y.-H. Lin, Words to describe a black hole, JHEP 02 (2023) 109,
[arXiv:2209.06728].

20
[45] S. Choi, S. Kim, E. Lee, and J. Park, The shape of non-graviton operators for SU (2),
arXiv:2209.12696.

[46] K. Budzik, D. Gaiotto, J. Kulp, B. R. Williams, J. Wu, and M. Yu, Semi-Chiral

Operators in 4d N = 1 Gauge Theories, arXiv:2306.01039.

[47] C.-M. Chang and Y.-H. Lin, Holographic covering and the fortuity of black holes,
arXiv:2402.10129.

21