A lattice bosonic model as a quantum theory of gravity

Zheng-Cheng Gu

and Xiao-Gang Wen

Department of Physics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139

Center for Advanced Study, Tsinghua University, Beijing, China, 100084

(Dated: June 22, 2006)

A local quantum bosonic model on a lattice is constructed whose low energy excitations are
gravitons described by linearized Einstein action. Thus the bosonic model is a quantum theory
of gravity, at least at the linear level. We nd that the compactication and the discretization of
metric tenor are crucial in obtaining a quantum theory of gravity.
Seven wonders of our universe: Our world has
many mysteries and wonders. At the most fundamental
level, there are may be seven deep mysteries and/or won-
ders in our universe: (1) Identical particles. (2) Gauge
interactions.
13
(3) Fermi statistics.
4,5
(4) Tiny masses of
fermions ( 10
20
of the Planck mass).
6,7,12
(5) Chiral
fermions.
8,9
(6) Lorentz invariance.
10
(7) Gravity.
11
Here
we would like to question where those wonderful and mys-
terious properties come from. We wish to have a single
unied understanding of all of the above mysteries. Or
more precisely, we wish that we can start from a single
model to obtain all of the above wonderful properties.
Recent progresses showed that, starting from a sin-
gle origin a local bosonic model (which is also called
spin model), the rst four properties in the above list
emerge at low energies.
1215
Thus we may say that the
local bosonic model provides a unied origin for iden-
tical particles, gauge interactions, Fermi statistics and
near masslessness of the fermions. Gauge interaction and
Fermi statistics are unied under this point of view of
emergence. However, three more mysteries remain to be
understood. In this paper, we will show that it is possible
to construct a local bosonic model from which gravitons
emerge. On one hand, such a model can be viewed as
a theory of quantum gravity. It solves a long standing
problem of putting quantum mechanics and gravity to-
gether. On the other hand, the model provides a design
of a condensed matter system which has an emergent
quantum gravity at low energies. This allows us to study
some quantum eects of gravity (a simulated one) in a
laboratory.
The belief that all the wonderful phenomena of our
universe (such as gauge interaction, Fermi statistics, and
gravitons) emerge from a lowly local bosonic model is
called locality principle.
12
Using local bosonic model as
a underlying structure to understand the deep myster-
ies of our universe represent a departure from the tra-
ditional approach of gaining a better understanding by
dividing things in to smaller parts. In the traditional
approach, we assume that the division cannot continue
forever and view everything as made of some simple indi-
visible building blocks the elementary particles. How-
ever, the traditional approach may represent a wrong di-
rection. For example, phonons behave just like any other
elementary particles at low energies. But if we look at
phonons closely, we do not see smaller parts that form a
phonon. We see the atoms that form the entire lattice.
The phonons are not formed by those atoms, the phonons
are simply collective motions of those atoms. This makes
us to wonder that photons, electrons, gravitons, etc , may
also be emergent phenomena just like phonons. They
may not be the building blocks of everything. They may
be collective motions of a deeper underlying structure.
This paper uses a particular emergence approach: we
try to obtain everything from a local bosonic model. The
detail form of the bosonic model is not important. The
important issue is how the bosons (or the spins) are or-
ganized in the ground state. It is shown that if bosons
organize into a string-net condensed state, then photons,
electrons and quarks can emerge naturally as collective
motions of the bosons.
1215
In this paper, we will nd an
organization of bosons such that the collective motions
of bosons lead to gravitons.
Other emergence approaches were developed in super-
string theory
16
in last 10 years, as demonstrated by the
duality relations among various superstring models and
matrix models.
17
The anti-de Sitter-space/conformal-
eld-theory duality even shows how space-time and grav-
ity emerge from a gauge theory.
18
The rule of game: There are many dierent ap-
proaches to quantum gravity
2022
based on dierent prin-
ciples. Some approaches, such as loop quantum gravity,
23
stress the gauge structure from the dieomorphism of
the space-time. Other approaches, such as superstring
theory,
16,17
stress the renormalizability of the theory. In
this paper, we follow a dierent rule of game by stressing
niteness and locality.
Our rule of game is encoded in the following working
denition of quantum gravity. Quantum gravity is
(a) a quantum theory.
(b) Its Hilbert space has a nite dimension.
(c) Its Hamiltonian is a sum of local operators.
(d) The gapless helicity 2 excitations are the only low
energy excitations.
(e) The helicity 2 excitations have a linear dispersion.
(f) The gravitons
35
can interact in the way consistent
with experimental observations.
The condition (b) implies that the quantum gravity
considered here has a nite cut-o. So the renormaliz-
ability is not an issue here. The condition (c) is a locality
condition. It implies two additional things: (1) the total
Hilbert space is a direct product of local Hilbert spaces
2
H
tot
=
n
H
n
, (2) the local operators are dened as op-
erators that act within each local Hilbert space H
n
or
nite products of local operators. The conditions (a
c) actually dene a local bosonic model. Certainly, any
quantum spin models satisfy (a c). It is the conditions
(df) that make the theory to look like gravity.
The condition (d) is very important. It is very easy
to construct a quantum model that contains helicity 2
gapless excitations, such as the theory described by the
following Lagrangian L =
1
2

0
h
ij

0
h
ij

1
2

k
h
ij

k
h
ij
.
Such a theory is not a theory of gravity since it also
contain helicity 0, 1 gapless excitations.
Superstring theory
16,17
satises the conditions (a), (e)
and (f), but in general not (d) due to the presence of
dilatons (massless scaler particles). The superstring the-
ory (or more precisely, the superstring eld theory) also
does not satisfy the condition (b) since the cut-o is not
explicitly implemented. The spin network
24
or the quan-
tum computing
25
approach to quantum gravity satises
the condition (a,b) or (ac). But the properties (df) are
remain to be shown. The induced gravity from superuid
3
He discussed in Ref. 26 does not satisfy the condition (d)
due to the presence of gapless density mode. In Ref. 27,
it is proposed that gravitons may emerge as edge exci-
tations of a quantum Hall state in 4 spatial dimensions.
Again the condition (d) is not satised due to the pres-
ence of innite massless helicity 1 ,2, 3, modes.
In Ref. 28, a very interesting bosonic model is con-
structed which contains gapless helicity 0 and 2 exci-
tations with quadratic dispersions. The model satises
(ac), but not (d,e). In this paper, we will x the two
problems and construct a local bosonic model that satis-
es the conditions (a e) and possibly (f). To quadratic
order, the low energy eective theory of our model is the
linearized Einstein gravity. The key step in our approach
is to discretize and compactify the metric tensor.
Review of emergence of U(1) gauge theory: Our
model for emergent quantum gravity is closely related
to the rotor model that produces emergent U(1) gauge
theory.
19,30
So we will rst discuss the emergence of U(1)
gauge theory to explain the key steps in our argument in
a simpler setting.
To describe the rotor model, We introduce an angular
variable a
ij
a
ij
+ 2 and the corresponding angular
momentum c
ij
for each link of a cubic lattice. Here i
labels the sites of the cubic lattice and a
ij
and c
ij
satisfy
a
ij
= a
ji
and c
ij
= c
ji
. The phase space Lagrangian
for physical degrees of freedom a
ij
and c
ij
is given by
19
L =

ij
c
ij
a
ij

J
2

ij
c
2
ij
+ g

ijkl
cos B
ijkl

U
2

i
Q
2
i
B
ijkl
= a
ij
+ a
jk
+ a
kl
+ a
li
, Q
i
=

j next to i
c
ij
(1)
where

i
sums over all sites,

ij
over all links, and

ijkl
over all square faces of the cubic lattice. We note
that after quantization, c
ij
are quantized as integers.
To obtain the low energy dynamics of the above ro-
tor model, let us assume that the uctuation of c
ij
are
large and treat c
ij
as a continuous quantity. We also
assume that the uctuations of a
ij
are small and expand
(1) to the quadratic order of a
ij
. Then we take the con-
tinuum limit by introducing the continuous elds (c
i
, a
i
)
and identifying c
ij
=
_
j
i
dx
i
c
i
and a
ij
=
_
j
i
dx
i
a
i
. Here
we have assumed that the lattice constant a = 1. The
resulting continuum eective theory is given by
L = c
i

0
a
i

1
2
J(c
i
)
2

1
2
g(B
i
)
2

1
2
U(
i
c
i
)
2
, (2)
where B
i
=
ijk

j
a
k
. We nd that the rotor model has
three low lying modes. Two of them are helicity 1
modes with a linear dispersion
k


gJ[k[ and the
third mode is the helicity 0 mode with zero frequency

k
= 0. We know that a U(1) gauge theory only have
two helicity 1 modes at low energies. Thus the key to
understand the emergence U(1) gauge theory is to un-
derstand how the helicity 0 mode obtain an energy gap.
To understand why helicity 0 mode is gapped, let us
consider the quantum uctuations of c
i
and a
i
. We note
that the longitudinal mode and the transverse modes sep-
arate. Introduce c
i
= c
||
+c

and a
i
= a
||
+a

, we nd
that the dynamics of the transverse mode is described
by L

= a

0
c

J
2
c
2

g
2

i
a

i
a

. At the lattice
scale x 1, the quantum uctuations of c

and a

are
given by c

_
g
J
and a

_
J
g
. We see that the as-
sumptions that we used to derive the continuum limit are
valid when J g. In this limit we can trust the result
from the continuum eective theory and conclude that
the transverse modes (or the helicity 1 modes) have a
linear gapless dispersion.
The longitudinal mode is described by (f(x), (x))
with a
i
=
i
f and =
i
c
i
. Its dynamics is determined
by L
||
=
0
f
J
2
(
2
)
1
2
U
2
. At the lattice scale,
the quantum uctuations of and f are given by = 0
and f = . We see that a positive U and J will make
the uctuations of f much bigger than the compacti-
cation size 2 and the uctuations of much less then
the discreteness of c
i
which is 1. In this limit, the result
from the classical equation of motion cannot be trusted.
In fact the weak quantum uctuations in the discrete
variable and the strong quantum uctuations in the
compact variable f indicate that the corresponding mode
is gapped after the quantization. Since has weak uctu-
ations which is less than the discreteness of , the ground
state is basically given by = 0. A low lying excitation
is then given by = 0 everywhere except in a unit cell
where = 1. Such an excitation have an energy of or-
der U. The gapping of helicity 0 mode is conrmed by
more careful calculations.
19
From those calculations, we
nd that the weak uctuations of lead to a constraint
=
i
c
i
= 0 and the strong uctuations of f lead to
a gauge transformation a
i
a
i
+
i
f. The Lagrangian
(2) equipped with the above constraint and the gauge
transformation becomes the Lagrangian of a U(1) gauge
3
theory. We will use this kind of argument to argue the
emergence of gravitons and the linearized Einstein action.
The emergence of quantum gravity: First, let us
describe a model that will have emergent gravitons at
low energies. The model has six variables
xx
(i),
yy
(i),

zz
(i), L
xx
(i), L
yy
(i), and L
zz
(i) on each vertex of a
cubic lattice. The model also has two variables on each
square face of the cubic lattice. For example, on the
square centered at i+
x
2
+
y
2
, the two variables are
xy
(i+
x
2
+
y
2
) =
yx
(i +
x
2
+
y
2
) and L
xy
(i +
x
2
+
y
2
) = L
yx
(i +
x
2
+
y
2
). The bosonic model is described by the following
phase space Lagrangian
L =

i,ab=xy,yz,zx
L
ab
(i +
a
2
+
b
2
)
0

ab
(i +
a
2
+
b
2
)
+

i,a=x,y,z
L
aa
(i)
0

aa
(i) H
U
H
J
H
g
(3)
where
H
U
=n
G
U
1

a=x,y,z
1 cos[2Q(i, i +a)/n
G
]
+n
G
U
2

i
1 cos[(i)]
H
J
=n
G
J

i,a=x,y,z
1 cos[2L
aa
(i)/n
G
]
+2n
G
J

ab=xy,yz,zx
1 cos[2L
ab
(i +
a
2
+
b
2
)/n
G
]

1
2
n
G
J

i
1 cos[2

a=x,y,z
L
aa
(i)/n
G
]
H
g
=
n
G
g
4

i,a=x,y,z
1 cos[
a
a
(i)]
+
n
G
g
4

i,ab=xy,yz,zx
sin[
a
b
(i)] sin[
b
a
(i)] (4)
Here
i
j
(i), (i) and Q(i, i + x) are dened as
x
x
(i) =

zx
(i + y +
z
2
+
x
2
) +
zx
(i +
z
2
+
x
2
)
xy
(i + z +
x
2
+
y
2
)
xy
(i +
x
2
+
y
2
),
x
y
(i) =
yz
(i +
z
2
+
y
2
)
yz
(i +
z
2

y
2
) + 2
yy
(i + z) + 2
yy
(i),
x
z
(i) =
2
zz
(i +y) 2
zz
(i) +
yz
(i +
y
2
+
z
2
) +
yz
(i +
y
2

z
2
),
(i) =

a=x,y,z

b=x,y,z
[
bb
(i + a) +
bb
(i a) +
2
bb
(i)]

a=x,y,z
[
aa
(i + a) +
aa
(i a) + 2
aa
(i)]

ab=xy,yz,zx
[
ab
(i +
a
2
+
b
2
) +
ab
(i
a
2
+
b
2
) +
ab
(i +
a
2

b
2
) +
ab
(i
a
2

b
2
)], and Q(i, i +x) = L
xx
(i +x) +
L
xx
(i) +L
yx
(i +
x
2
+
y
2
) +L
yx
(i +
x
2

y
2
) +L
zx
(i +
x
2
+
z
2
) +L
zx
(i +
x
2

z
2
). Other components are obtained by
cycling xyz to yzx and zxy.
Note that both
ab
and its canonical conjugate L
ab
are
compactied:
ab

ab
+ 2, L
ab
L
ab
+ n
G
. Hence
after quantization they are both discretized and n
G
is
an integer. Due to the compactication, only W
ab
L
=
e
2i L
ab
/n
G
, W
ab

= e
i
ab
and their products are physical
operators. For a x ab and i, W
ab
L
(i) and W
ab

(i) satisfy
the algebra
W
ab
L
(i)W
ab

(i) = e
2i /n
G
W
ab

(i)W
ab
L
(i) (5)
Such an algebra has only one n
G
dimensional represen-
tation. This n
G
dimensional representation becomes our
local Hilbert space H
i,ab
. The total Hilbert space of our
model (3) is given by H =
i,ab
H
i,ab
after quantization.
In other words, there are n
3
G
states on each vertex and
n
G
states on each square face of the cubic lattice. Note
that the Hamiltonian
H = H
U
+ H
J
+ H
g
(6)
is a function of the physical operators W
ab
L
, W
ab

and
their hermitian conjugates. So the quantum model de-
ned through the Hamiltonian (6) and the algebra (5)
is a bosonic model whose local Hilbert spaces have nite
dimensions.
Next, we would like to understand low energy excita-
tions of the quantum bosonic model (6) in large n
G
limit.
We rst assume that the uctuations of
ij
2L
ij
/n
G
and
ij
are much bigger then 1/n
G
(the discreteness of

ij
and
ij
) so that we can treat
ij
and
ij
as contin-
uous variables. We also assume that the uctuations of

ij
and
ij
are much smaller then 1 so that we can treat

ij
and
ij
as small variables. Under those assumptions,
we can use semiclassical approach to understand the low
energy dynamics of the quantum bosonic model (6).
Expanding the Lagrangian (3) to quadratic order in
ij
and
ij
, we can nd the dispersions of collective modes
of the bosonic model. There are total of six collective
modes. We nd four of them have zero frequency for all
k, and two modes have a linear dispersion relation near
k = (, , ). Near k = (, , ), the dynamics of the
six modes are described by the following continuum eld
theory:
L = n
G
_

ij

ij

J
2
_
(
ij
)
2

(
ii
)
2
2
_

g
2

ij
R
ij

U
1
2
(
i

ij
)
2

U
2
2
(R
ii
)
2
_
(7)
where R
ij
=
imk

jln

nk
and we dene the contin-
uum eld
ab
(x) as (1)
i 1
2

ab
(i +
a
2
+
b
2
) for a ,= b and
as (1)
i

ab
(i) for a = b. The helicity 2 modes have
a linear dispersion relation
k


gJ[k[. We nd that
for large n
G
, the quantum uctuations of the helicity 2
modes is of order
ij
,
ij

_
1/n
G
(assuming U
1,2
, J
and g are of the same order). So the uctuations of
ij
and
ij
satisfy 1/n
G

ij
,
ij
1 and the semiclassi-
cal approximation is valid for the helicity 2 modes. In
this case, the result
k

gJ[k[ can be trusted.

The helicity 1 modes and one of the helicity 0 mode
are described by
ij
=
i

j
+
j

i
and
i
=
j

ji
. Their
frequency
k
= 0. For such modes, the Hamiltonian
only contains
i
. Thus the quantum uctuations satis-
es
i
1/n
G
and
i
1. So the semiclassical ap-
proximation is not valid and the result
k
= 0 cannot be
trusted. Using the similar argument used in emergence
of U(1) gauge bosons, we conclude that those modes are
gapped. The strong uctuations
ij
=
i

j
+
j

i
1
4
and the weak uctuations
i
1/n
G
lead to gauge trans-
formations and the constraints

ij

ij
+
i

j
+
j

i
,
j

ji
= 0 (8)
The second helicity 0 mode is described by
ij
=
(
ij

j
) and = (
ij

j
)
ij
. Its frequency is
again
k
= 0. The Hamiltonian for such a mode contains
only . So the quantum uctuations satises 1 and
1/n
G
. The second helicity 0 mode is also gapped.
The strong uctuations
ij
= (
ij

2

i

j
) 1 and
the weak uctuations = (
ij

j
)
ij
1/n
G
lead
to a gauge transformation and a constraint

ij

ij
+(
ij

j
), (
ij

j
)
ij
= 0 (9)
The Lagrangian (7) equipped with the gauge trans-
formations and the constraints (8,9) is nothing but the
linearized Einstein Lagrangian of gravity, where
ij

g
ij

ij
represents the uctuations of the metric tenor
g
ij
around the at space. So the linearized Einstein grav-
ity emerge from the quantum model (6) in the large n
G
limit. The local bosonic model (6) can be viewed as a
quantum theory of gravity.
We have seen that the gapping of the helicity 0 mode in
the rotor model (1) leads to an emergence of U(1) gauge
structure at low energies. The emergence of a gauge
structure also represents a new kind of order quantum
order
12,31
in the ground state. In Ref. 32, it was shown
that the emergent U(1) gauge invariance, and hence the
quantum order, is robust against any local perturbations
of the rotor model. Thus the gaplessness of the emer-
gent photon is protected by the quantum order.
33
Sim-
ilarly, the gapping of the two helicity 0 modes and the
helicity 1 modes in the bosonic model (6) leads to an
emergent gauge invariance of the linearized coordinate
transformation. This indicates that the ground state of
the bosonic model contains a new kind of quantum order
that is dierent from those associated with emergent or-
dinary gauge invariances of internal degrees of freedom.
We expect such an emergent linearized dieomorphism
invariance to be robust against any local perturbation of
the bosonic model. Thus the gaplessness of the emergent
gravitons is protected by the quantum order.
The emergent gravitons in the model (3) naturally in-
teract but the interaction may be dierent from that de-
scribed by the higher order non-linear terms in Einstein
gravity. However, those higher order terms are irrelevant
at low energies. Thus it may be possible to generate
those higher order terms by ne tuning the lattice model
(3), such as modifying the Hamiltonian (H
J
and H
g
),
the constraints (H
U
), as well as the Berrys phase term
in (3). So it may be possible that local bosonic models
can generate proper non-linear terms to satisfy (f).
29
Our result appears to contradict with the Weinberg-
Witten theorem
34
which states that in all theories with
a Lorentz-covariant energy-momentum tensor, compos-
ite as well as elementary massless particles with helicity
h > 1 are forbidden. However, the energy-momentum
tensor in our model is not invariant under the linearized
dieomorphism (although the action is invariant). This
may be the reason why emergent gravitons are possible
in our model.
We would like to thank J. Polchinski and E. Witten for
their very helpful comments. This research is supported
by NSF grant No. DMR-0433632 and ARO grant No.
W911NF-05-1-0474.
1
H. Weyl, Space, Time, Matter (Dover, 1952).
2
W. Pauli, Rev. Mod. Phys. 13, 203 (1941).
3
C. N. Yang and R. L. Mills, Phys. Rev. 96, 191 (1954).
4
E. Fermi, Z. Phys. 36, 902 (1926).
5
P. A. M. Dirac, Proc. Roy. Soc. A112, 661 (1926).
6
D. J. Gross and F. Wilczek, Phys. Rev. Lett. 30, 1343
(1973).
7
H. D. Politzer, Phys. Rev. Lett. 30, 1346 (1973).
8
T. Lee and C. Yang, Phys. Rev. 104, 254 (1956).
9
C. Wu et al., Phys. Rev. 105, 1413 (1957).
10
A. Einstein, Annalen der Physik 17, 891 (1905).
11
A. Einstein, Annalen der Physik 49, 769 (1916).
12
X.-G. Wen, Phys. Rev. D 68, 065003 (2003).
13
M. A. Levin and X.-G. Wen, Rev. Mod. Phys. 77, 871
(2005).
14
M. Levin and X.-G. Wen, hep-th/0507118.
15
X.-G. Wen, cond-mat/0508020.
16
M. B. Green, J. H. Schwarz, and E. Witten, Superstring
Theory (Cambridge University Press, 1988).
17
J. Polchinski, String Theory (Cambridge University Press,
1998).
18
O. Aharony etal, Phys. Rept. 323, 183 (2000).
19
X.-G. Wen, Quantum Field Theory of Many-Body Systems
From the Origin of Sound to an Origin of Light and
Electrons (Oxford Univ. Press, Oxford, 2004).
20
T. Thiemann, gr-qc/0210094.
21
S. Carlip, Rept. Prog. Phys. 64, 885 (2001).
22
P. Peldan, Class. Quant. Grav. 11, 1087 (1994).
23
L. Smolin, hep-th/0408048.
24
C. Rovelli and L. Smolin, Phys.Rev. D52, 5743 (1995).
25
L. Smolin, hep-th/0507235.
26
G. Volovik, J. Low. Temp. Phys. 113, 667 (1998).
27
S.-C. Zhang and J. Hu, Science 294, 823 (2001).
28
C. Xu, cond-mat/0602443.
29
Z.-C. Gu and X.-G. Wen, to appear.
30
O. I. Motrunich and T. Senthil, Phys. Rev. Lett. 89,
277004 (2002).
31
X.-G. Wen, Phys. Rev. B 65, 165113 (2002).
32
M. B. Hastings and X.-G. Wen, Phys. Rev B 72, 045141
(2005).
33
X.-G. Wen, Phys. Rev. Lett. 88, 11602 (2002).
34
S. Weinberg and E. Witten, Phys. Lett. B 96, 59 (1980).
35
In this paper, we dene gravitons as linearly dispersing
gapless helicity 2 excitations.