GRAVITY IN NONCOMMUTATIVE GEOMETRY

CHRIS GEORGE

Introduction
The traditional arena of geometry and topology is a set of points with some
particular structure that, for want of a better name, we call a space. Thus, for
instance, one studies curves and surfaces as subsets of an ambient Euclidean space.
It was recognized early on, however, that even such a fundamental geometrical
object as an elliptic curve is best studied not as a set of points (a torus) but rather
by examining functions on this set, specifically the doubly periodic meromorphic
functions. Weierstrass opened up a new approach to geometry by studying directly
the collection of complex functions that satisfy an algebraic addition theorem, and
derived the point set as a consequence. In probability theory, the set of outcomes
of an experiment forms a measure space, and one may regard events as subsets of
outcomes; but most of the information is obtained from “random variables”, i.e.,
measurable functions on the space of outcomes.
In noncommutative geometry, under the in influence of quantum physics, this
general idea of replacing sets of points by classes of functions is taken further. In
many cases the set is completely determined by an algebra of functions, so one
forgets about the set and obtains all information from the functions alone. Also,
in many geometrical situations the associated set is very pathological, and a direct
examination yields no useful information. The set of orbits of a group action, such
as the rotation of a circle by multiples of an irrational angle, is of this type. In such
cases, when we examine the matter from the algebraic point of view, we often obtain
a perfectly good operator algebra that holds the information we need; however, this
algebra is generally not commutative.
The Gelfand-Naimark theorem gives us a complete equivalence between the cat-
egory of locally compact Hausdorff spaces with continuous maps and the category
of commutative C ∗ -algebras with *-homomorphisms. In a famous paper [9] that
has become a cornerstone of noncommutative geometry, Gelfand and Naimark in
1943 characterized the involutive algebras of operators by just dropping commuta-
tivity from the most natural axiomatization for the algebra of continuous functions
on a locally compact Hausdorff space. Thus, a noncommutative C ∗ -algebra will
be viewed as the algebra of continuous functions on some ’virtual noncommutative
space. The algebra-topology duality can be neatly summed up with the following
dictionary, from [15],

1
2 CHRIS GEORGE

TOPOLOGY ALGEBRA
Locally compact space C ∗ -algebra
Compact space Unital C ∗ -algebra
Compactification Unitization
Continuous proper map ∗-homomorphism
Homeomorphism Automorphism
Open subset Ideal
Closed subset Quotient algebra
Second countable Separable
Measure Positive functional
In order to go further however, particularly with regard to physical examples, we
need to move beyond this noncommutative topology, and into some sort of dif-
ferential structure. This will be the focus of this paper. This new calculus can
be described simply in the following dictionary. For an involutive algebra A, we
construct a spectral triple (A, H, D), where H is the Hilbert space on which A is
realized, and D is a selfadjoint unitary on H. The dictionary then will be [4],

CLASSICAL NONCOMMUTATIVE
Complex function Operator on H
Real function Selfadjoint operator on H
Infinitesimal Compact operator
Differential df Commutator da = [D, a]
Integral Dixmier trace trω
The first section of this paper will be devoted to developing this noncommuta-
tive differential geometry. In the second section, we will apply these new tools to
construct a gravity model through noncommutative geometry.

1. The Spectral Calculus

In this section we will introduce the basic notions of the noncommutative gen-
eralization of the usual calculus on manifolds.

1.1. Infinitesimals. Recall that for any T ∈ K(H), we have the polar decomposi-
tion T = U |T |, where |T |= (T ∗ T )1/2 . Then, the eigenvalues of |T |, with multiplic-
ity, are called the characteristic values of T . These characteristic values, denoted by
{µn (T )}n∈N , are enumerated in decreasing order. So, µ0 (T ) = kT k, the operator
norm of T , and µn (T ) → 0 as n → ∞.
Because compact operators are, in a sense, ‘small’, they play the role of infinites-
imals in Connes’ theory. The rate of decay of {µn (T )} as n → ∞ tells us the size
of the infinitesimal T ∈ K(H), as set out below:

Definition 1.1.1. For any α ∈ R+ , the infinitesimals of order α are all T ∈ K(H)
such that

(1.1.1) µn (T ) = O(n−α ), as n→∞

(1.1.2) ie. ∃ C < ∞ such that µn (T ) ≤ Cn−α , ∀ n ≥ 1

 GRAVITY IN NONCOMMUTATIVE GEOMETRY 3

Now, for any two compact operators T1 and T2 , we have the submultiplicative
property[13],
(1.1.3) µn+m (T1 T2 ) ≤ µn (T1 )µm (T2 ),
implying that the orders of infinitesimals are well-behaved,
(1.1.4) Ti of order αi ⇒ T1 T2 of order α1 + α2
And lastly, infinitesimals of order α form a (two-sided) ideal in B(H), since for any
B ∈ B(H) and T ∈ K(H) [13],
(1.1.5) µn (T B), µn (BT ) ≤ kBkµn (T )

1.2. The Dixmier Trace. In the noncommutative approach, as in the ordinary

calculus, one looks for an ‘integral’ neglecting infinitesimals of order > 1. The
Dixmier trace, constructed below, will be exactly what we are looking for.
For a positive compact operator T , we have
∞
X
(1.2.1) tr T = µn (T )
0

Now, in general, infinitesimals of order 1 are not trace class, since the only bound
we have on the characteristic values is µn ≤ C n1 , for some positive constant C. But,
from this we see that the usual trace (1.2.1), is at most logarithmically divergent
for (positive infinitesimals order 1:
N
X −1
(1.2.2) µn (T ) ≤ C logN
0

We wil use the Dixmier trace as a way to simply extract this coefficient C of
logarithmic divergence. It is rather interesting that this coefficient behaves like a
trace[8].
The ideal of compact operators which are infinitesimals of order 1 will be denoted
by L(1,∞) . For any positive T ∈ L(1,∞) , the natural thing to do is take the limit of
the cut-off sums,
N −1
1 X
(1.2.3) lim µn (T )
N →∞ logN
0

However, problems with linearity and convergence arise in this definition. Now for
any T ∈ K(H) consider the sums,
N
X −1
σN (T )
(1.2.4) σN (T ) = µn (T ) , γN (T ) =
0
logN

Then, we have[4],
(1.2.5) σN (T1 + T2 ) ≤ σN (T1 ) + σN (T2 ), ∀ T1 , T2 ,

(1.2.6) σ2N (T1 + T2 ) ≥ σN (T1 ) + σN (T2 ), ∀ T1 , T2 > 0

4 CHRIS GEORGE

As a result, for any two positive operators T1 and T2 ,

log2
(1.2.7) γN (T1 + T2 ) ≤ γN (T1 ) + γN (T2 ) ≤ γ2N (T1 + T2 )(1 + )
logN
Thus, convergence would imply linearity, but the sequence {γN } (though bounded)
does not, in general, converge. However in most examples of physical interest, {γN }
is, in fact, convergent. So, (1.2.3) certainly yields a positive linear functional, since
L(1,∞) is generated by its positive elements. Thus, we will define the Dixmier trace,
trω , by
N −1
1 X
(1.2.8) trω (T ) = lim µn (T )
N →∞ logN
0

Also, from (1.1.5), for any T ∈ L(1,∞) ,

(1.2.9) trω (BT ) = trω (T B), ∀ B ∈ B(H)
And, any infinitesimal T of order greater than 1 satisfies
1
(1.2.10) µn (T ) = o( ), i.e. nµn (T ) → 0 as n → ∞
n
So the corresponding sequence γN (T ) converges to zero, showing that the Dixmier
trace vanishes on infinitesimals of order > 1. Thus, the Dixmier trace is, in fact, a
trace whose domain is L(1,∞) , and neglects all infinitesimals of order > 1.

1.3. Spectral Triples. Here we will introduce the main concept used by Connes
to develop the analogue of differential calculus for noncommutative algebras.
Definition 1.3.1. A spectral triple (A, H, D) is given by an involutive algebra A
of bounded operators on the Hilbert space H, together with a self-adjoint operator
D on H (called the Dirac operator) satisfying:
• The resolvent (D − λ)−1 , λ ∈ C\R, is a compact operator on H ;
• The commutator [D, a] = Da − aD ∈ B(H), ∀ a ∈ A
These conditions imply that the collection {λn } of eigenvalues of D forms a
discrete subset of R [4]. In addition, (D − λ)−1 being compact has characteristic
values µn ((D − λ)−1 ) → 0, and thus |λn | = µn (|D|) → ∞.
Also, consider the derivation δ on B(H) defined by
(1.3.1) δ(T ) = [|D|, T ].
Given this derivation δ, the subalgebra Ak ⊂ A, with k ≥ 2, consists os all elements
a
T ∈ A such that both a, and ∞ [D, a] are in the domain of δ k−1 . The elements of
Ak are said to be of class C .
Now consider a closed n-dimensional Riemannian spin manifold (M, g). The
corresponding spectral triple (A, H, D) is called the canonical triple over M . In
this example we have,
• A = F(M ), the algebra of smooth, complex-valued functions on M .
• H = L2 (M, S), the space of square integrable sections of the irreducible
spinor bundle over M , with the natural scalar product.
• D is the Dirac operator associated with the Levi-Cevita connection of the
metric g.
 GRAVITY IN NONCOMMUTATIVE GEOMETRY 5

Recall that taking the norm closure of an involutive algebra A, yields a C ∗ -

algebra A. Leaving the details to [10], we state the following proposition without
proof.
Proposition 1.3.2. For the canonical triple (A, H, D), we have
• M is the structure space of the C ∗ -algebra A of continuous functions on M
(This, of course, is simply the Gelfand-Naimark theorem).
• The geodesic distance between any two points p, q ∈ M is given by
(1.3.2) d(p, q) = sup {|f (p) − f (q)|; k[D, f ]k ≤ 1}, ∀p, q ∈ M.
f ∈A

• The Riemannian measure on M is given by

Z
(1.3.3) f = C trω (f |D|−n ), ∀ f ∈ A,
M
(n−[n/2]−1) n/2
Where C = 2 π nΓ(n/2).
In accordance with this proposition we can define noncommutative analogues of
(1.3.2), and (1.3.3). The distance function on the state space S(A) is given by
(1.3.4) d(φ, ξ) = sup {|φ(a) − xi(a)|; k[D, a]k ≤ 1}, ∀φ, ξ ∈ S(A).
a∈A

To define the analogue of the measure integral, we first need the notion of the
dimension of a spectral triple.
Definition 1.3.3. A spectral triple (A, H, D) is of dimension n > 0 if |D|−n is an
infinitesimal of order 1.
Now, given an n-dimensional spectral triple, for any a ∈ A its integral is defined
by
Z
1
(1.3.5) a= trω a|D|−n ,
V
where the constant V is determined by the characteristic values of |D|−n . More
precisely, µj (|D|−n ) ≤ Vj for j → ∞.

1.4. Universal Differential Forms. For a unital L involutive algebra A over C,

the universal differential algebra of forms ΩA = p Ωp A is a graded algebra which
we will construct below. In degree 0, we have Ω0 A = A. The space of one-forms,
Ω1 A, is generated as a left A-module by symbols of degree δa, for a ∈ A, satisfying
the relations
(1.4.1) δ(ab) = (δa)b + aδb

(1.4.2) δ(αa + βb) = αδa + βδb

The Leibniz rule (1.4.1) automatically gives
δ1 = δ(1 · 1) = (δ1) · 1 + 1 · (δ1) = 2(δ1) =⇒ δ1 = 0 =⇒ δC = 0.
So, a generic ω ∈ Ω1 A is just a finite linear combination
X
(1.4.3) ω= ai δbi , f or some ai , bi ∈ A
f inite
6 CHRIS GEORGE

We can also give Ω1 A a right A-module structure by setting

³X ´ X X X
(1.4.4) ai δbi c = ai (δbi )c = ai δ(bi c) − ai bi δc

where we have used the Leibniz rule to obtain the final equality.
The map
δ : A → Ω1 A
can be considered a derivation of A with values in the bimodule Ω1 A. Then, the
pair (δ, Ω1 A) is characterized by the universal property[4].
In higher degrees, the space Ωp A is given by
(1.4.5) Ωp A = Ω 1 1
| AΩ A
1
{z· · · Ω A}
p−times

where the product of any two one-forms is defined simply by ‘juxtaposition’,

(1.4.6) (a0 δa1 )(b0 δb1 ) = a0 (δa1 )b0 δb1 ) = a0 δ(a1 b0 )δb1 − a0 a1 δb0 δb1 , ∀ai , bi ∈ A.
p
Thus, Ω A consists of finite linear combinations of monomials of the form
(1.4.7) ω = a0 δa1 δa2 · · · δap , ai ∈ A.
So, the product of a p-form and a q-form yields a (p + q)-form which is, again,
defined by juxtaposition, and rearranging the result using the Leibniz rule,
(a0 δa1 · · · δap )(ap+1 δap+2 · · · δap+q )
= a0 δa1 · · · (δap )ap+1 δap+2 · · · δap+q
= (−1)p a0 a1 δa2 · · · δap+q
p
X
(1.4.8) + (−1)p−i a0 δa1 · · · δai−1 δ(ai ai+1 )δai+2 · · · δap+q
i=1
1 p
As with Ω A, Ω A is a left A-module by construction. It can also be endowed with
a right A-module structure in the same manner, using the Liebniz rule,
(a0 δa1 · · · δap )b = a0 δa1 · · · (δap )b
= (−1)p a0 a1 δa2 · · · δap δb
p
X
+ (−1)p−i a0 δa1 · · · δai−1 δ(ai ai+1 )δai+2 · · · δap δb
i=1

(1.4.9) +a0 δa1 · · · δap−1 (δap b), ∀ai , b ∈ A

Now extending δ so that for any p we have a map
δ : Ωp → Ωp+1
defined by the equation
(1.4.10) δ(a0 δa1 · · · δap ) = δa0 δa1 · · · δap .
Then we have the relations
• δ 2 = 0,
• δ(ω1 ω2 ) = δ(ω1 )ω2 + (−1)p ω1 δω2 , ∀ω1 ∈ Ωp A, ω2 ∈ ΩA.
 GRAVITY IN NONCOMMUTATIVE GEOMETRY 7

And lastly, we define an involution on ΩA by

(1.4.11) (δa)∗ = −δa∗ , ∀a ∈ A

(a0 δa1 · · · δap )∗ = (δap )∗ · · · (δa1 )∗ a∗0

= a∗p δa∗p−1 · · · δa∗0

p−1
X
(1.4.12) + (−1)p+i δa∗p · · · δ(a∗i+1 a∗i ) · · · δa∗0 .
i=0

Thus, (ΩA, δ) is a graded differential involutive algebra, once again characterized

by the universal property[4].

1.5. Connes’ Differential Forms. Given a spectral triple (A, H, D), using a rep-
resentation of the universal algebra ΩA on B(H), we will construct the exterior
algebra of forms. The map π : ΩA → B(H) defined by
(1.5.1) π(a0 δa1 · · · δap ) = a0 [D, a1 ] · · · [D, ap ], ai ∈ A
is clearly a homomorphism, since both δ and [d, ·] are derivations on A. Also, it is
a *-homomorphism, because [D, a]∗ = −[D, a∗ ].
The natural thing do next would be to define π(Ω) as the space of forms. In
general, however, π(Ω) = 0 does not imply π(δω) = 0. In order to proceed in
constructing a true differential algebra, we will need to dispose of these so called
junk forms.
Proposition 1.5.1. Let J0 = ⊕p J0p be the graded (two-sided) ideal of ΩA given by
(1.5.2) J0p = {ω ∈ Ωp A, π(ω) = 0}.
Then, J = J0 + δJ0 is a graded (two-sided) ideal of ΩA.
Proof. J is obviously graded, and the property δ 2 = 0 implies it is differential. Let
ω ∈ J. Then, ω = ω1 +δω2 , for some ω1 ∈ J0p³ and ω2 ∈ J0p−1 . Then
´ for any η ∈ ΩA,
we have ωη = ω1 η + δ(ω2 η) + (−1)p ω2 δη = ω1 η + (−1)p ω2 δη + δ(ω2 η) ∈ J. In a
similar way, we find ηω ∈ J. ¤

Definition 1.5.2. The graded differential algebra of Connes’ forms over A is de-
fined by
(1.5.3) ΩD A = ΩA/J ' π(ΩA)/π(δJ0 ),
with the space of p-forms given by
(1.5.4) ΩpD A = Ωp A/J p .
Now, since J is a differential ideal, the exterior differential δ defines a differential
on ΩD A, d : ΩpD A → Ωp+1
D , defined by

(1.5.5) d[ω] = [δω] ' [π(δω]

with ω ∈ Ωp A and [ω] the corresponding equivalence class in ΩpD A.
8 CHRIS GEORGE

1.6. Scalar Product of Forms. While we showed that the Dixmier trace was
indeed a trace state on A, we need to extend the result to all of π(ΩA). This is
not possible, in general, and some conditions need be imposed on the algebra A.
Recall that the subalgebra A2 ⊂ A is generated by all elements a ∈ A such that
both a and [D, a] are in the domain of the derivation δ defined in (1.3.1). It has
been shown in [3] that provided A2 = A (or A2 is a sufficiently large subalgebra),
the Dixmier trace is indeed a trace state on π(ΩA). As a result, the following three
traces coincide, and will be taken as the definition of an inner product on π(Ωp A),
hT1 , T2 ip = trω (T1∗ T2 |D|−n )
= trω (T1∗ |D|−n T2 )

(1.6.1) = trω (T2 |D|−n T1∗ ), ∀T1 , T2 ∈ π(Ωp A).

And forms of different degrees are defined to be orthogonal.
Denote the completion of π(Ωp A) by H̃p . Let Pp be the orthogonal projection of
H̃p , with respect to the inner product (1.6.1), which projects onto the orthogonal
complement of π(δ(J0 ∩ Ωp−1 A)).

1.7. Universal Connections on Modules. In this section, we will develop the

notion of a connection on a (finite projective), with respect to the universal calculus
ΩA. Since ΩA is the prototype for any calculus on A, by a connection we really
mean a universal connection, but this adjective will be dropped when there is no
confusion.
Definition 1.7.1. A (universal) connection on the right A-module E is a complex-
linear map
(1.7.1) ∇ : E ⊗A Ωp A → E ⊗A Ωp+1 A,
defined for any p ≥ 0, satisfying the Leibniz rule
(1.7.2) ∇(ωρ) = (∇ω)ρ + (−1)p ωδρ, ∀ω ∈ E ⊗A Ωp A, ρ ∈ ΩA.
Thus, a connection is completely determined by its restriction ∇ : E → E ⊗A Ω1 A,
which satisfies
(1.7.3) ∇(ηa) = (∇ω)a + η ⊗A δa, a ∈ A,
and then extends by using the Leibniz rule (1.7.2).
Proposition 1.7.2. The map
(1.7.4) ∇2 = ∇ ◦ ∇ : E ⊗A Ωp A → E ⊗A Ωp+2 A,
is ΩA-linear.
Proof. From the Leibniz rule 1.7.2, we have
³ ´
∇2 (ωρ) = ∇ (∇ω)ρ + (−1)p ωδρ

= (∇2 ω)ρ + (−1)p+1 (∇ω)δρ + (−1)p (∇ω)δρ + ωδ 2 ρ

(1.7.5) = (∇2 ω)ρ

¤
 GRAVITY IN NONCOMMUTATIVE GEOMETRY 9

Definition 1.7.3. The restriction θ, of ∇2 to E,

(1.7.6) θ : E → E ⊗A Ω2 A,
is the curvature of the connection.
The curvature is A-linear and satisfies
(1.7.7) ∇2 (η ⊗A ρ) = θ(η)ρ, ∀η ∈ E, ρ ∈ ΩA.
Proposition 1.7.4. The curvature θ satisfies the Bianchi identity,
(1.7.8) [∇, θ] = 0.
Proof. Since θ : E → Ω2 A, the map [∇, θ] makes sense. And,
(1.7.9) [∇, θ] = ∇ ◦ ∇2 − ∇2 ◦ ∇ = ∇3 − ∇3 = 0.
¤

The next proposition, for which the proof can be found in [7], takes care of any
concerns over the existence of connections.
Proposition 1.7.5. There exists a connection on a right module if and only if it
is projective.
And, of course, we will only be considering finite projective modules in physical
examples.

2. Gravity Models
While there have been a number of different approaches to constructing gravity
models in noncommutative geometry (see, for instance [5], [11, 12]), we will fol-
low the approach developed in [1, 2] for which our discussion of connections leads
directly into.

2.1. Connections Revisited. Given a spectral triple (A, H, D) with the associ-
ated differential calculus (ΩD A, d), then by Serre-Swan [14], the space Ω1D A is the
noncommutative analogue of the space of sections of the cotangent bundle. It is
naturally a right A-module, and we will assume it is projective of finite type, as
well. In order to develop ‘noncommutative Riemannian geometry’ here, we will
need an analogue of a metric on Ω1D A. There is a canonical Hermitian structure
on Ω1D A which is uniquely determined by the triple (A, H, D) given by,
(2.1.1) hα, βiD = P0 (α∗ β) ∈ A, α, β ∈ Ω1D A,
where P0 is the orthogonal projector ont A determined by the scalar product (1.6.1.
We will also assume that if (Ω1D A)0 is the dual module, the Riemannian structure
defines a right-module isomorphism,
(2.1.2) Ω1D A → (Ω1D A)0 , α 7→ hα, ·i.
We can now define a linear connection which is formally the same as that in the
remarks surrounding (1.7.3), taking E = Ω1D A.
10 CHRIS GEORGE

Definition 2.1.1. A linear connection on Ω1D A is a complex-linear map

(2.1.3) ∇ : Ω1D → Ω1D ⊗A Ω1D A,
satisfying the Leibniz rule
(2.1.4) ∇(αa) = (∇α)a + αda, ∀α ∈ Ω1D A, a ∈ A
Definition 2.1.2. The Riemannian curvature of ∇ is the A-linear map R∇ ,
(2.1.5) R∇ = ∇2 : Ω1D → Ω1D A ⊕A Ω1D A.
We call the connection ∇ metric if it is compatible with the Riemannian structure
h·, ·iD :
(2.1.6) −h∇α, βiD + hα, ∇βiD = dhα, βiD , ∀α, β ∈ Ω1D A.
Definition 2.1.3. The torsion of the connection ∇ is the map
T∇ : Ω1D A → Ω2D A,

(2.1.7) T∇ = d − m ◦ ∇,
where m is just the multiplication operator, m(α ⊗A β) = αβ.
It is easy to check that T∇ is a tensor, and for an ordinary manifold with a
linear connection, the above definition gives the cotangent (dual) space version of
the usual definition of torsion.
Definition 2.1.4. A connection ∇ on Ω1D A is a Levi-Cevita connection if it is
metric, and its torsion vanishes.
Unlike in ordinary differential geoemtry, Levi-Cevita connections do not neces-
sarily exist, or may not be unique for a given spectral triple. Now, for simplicity
we will take Ω1D A to be a free module with orthonormal basis {E A , A = 1, · · · , N }
with respect to the Riemannian structure h·, ·iD ,
(2.1.8) hE A , E B iD = η AB = diag(δAB , · · · , δAB ), A, B = 1, · · · , N
As we noted earlier, a connection ∇ on Ω1D A is completely determined by its
action on 1-forms ΩB 1
A ∈ ΩD A defined by,

(2.1.9) ∇E A = E B ⊗A ΩA
B , A = 1, · · · , N

Then the components of torsion T A ∈ Ω2D A and the curvature RA

B
∈ Ω2D A are
given by
T∇ (E A ) = T A

(2.1.10) R∇ (E A ) = E B ⊗A RA
B
, A = 1, · · · , N
Then, by using (2.1.5), and (2.1.7) we obtain the Cartan structure equations,
T A = dE A − E B ΩA
B , A = 1, · · · , N

B
(2.1.11) RA = dΩB C B
A + ΩA ΩC , A, B = 1, · · · , N.

Now, the metric condition reads,

(2.1.12) −ΩA∗
C η
AB
+ η AC ΩB
C = 0.
 GRAVITY IN NONCOMMUTATIVE GEOMETRY 11

Since metricity and vanishing torsion don’t necessarily fix the connection uniquely,
sometimes additional constraints are imposed by requiring that the connection is
Hermitian on 1-forms,
(2.1.13) ΩB B∗
A = ΩA

The components of torsion and curvature transform in the expected way under
a change orthonormal basis. For another orthonormal basis {Ẽ A , A = 1, · · · , N } of
Ω1D A, the relationship between the two bases is given by,
(2.1.14) Ẽ A = E B (M −1 )A A B A
B , E = Ẽ MB

with
(2.1.15) MAC (M −1 )B
C = (M
−1 C
)A MCB = δAB
ie. the matrix M = (MAB ) is invertible with inverse M −1 = ((M −1 )B
A ). The
requirement that the new basis be orthonormal gives,
η AB = hE A , E B iD
= hẼ C MCA , Ẽ D MD
B
iD
= (MCA )∗ hẼ C , Ẽ D iD MD
B

(2.1.16) = (MCA )∗ η CD MD
B
.
And so we get,
(2.1.17) (M −1 )B
A =η
AD
(MCD )η CB ,
that is, M −1 = M ∗ , and thus M is unitary.
It is easy to then find the components of curvature and torsion after an orthonormal
change of basis,
B
Ω˜A = MAC ΩDC (M
−1 B
)D + MAC d(M −1 )B
C,
B
R̃A = MAC RC
D
(M −1 )B
D,

(2.1.18) T̃ A = T B (M −1 )A
B.

Now, let {²A , A = 1, · · · , N } be the dual basis of {E A }. That is, ²A ∈ (Ω1D )0 , and
(2.1.19) ²A (E B ) = δAB
Then, by the isomorphism (2.1.2), for each ²A , there is an ²̂A ∈ Ω1D A given by,
(2.1.20) ²A = h²̂A , αiD , ∀α ∈ Ω1D A, A = 1, · · · , N
And thus,
(2.1.21) ²̂A = E B η AB , A = 1, · · · , N.
So, under an orthonormal change of basis, they transform as
(2.1.22) ²̂˜A = ²̂B (M B )∗ , A = 1, · · · , N.
A

Now, we are equipped to define the final noncommutative analogues needed for
differential geometry.
∇
Definition 2.1.5. The Ricci 1-forms, RA of a connection ∇ are given by
∇ B
(2.1.23) RA = P1 (RA (²̂B )∗ ) ∈ Ω1D A
12 CHRIS GEORGE

Definition 2.1.6. The scalar curvature, r∇ , of a connection ∇ is defined by

(2.1.24) r∇ = P0 (E A P1 (RA
B
²̂B )∗ )
where, again, the projectors P0 and P1 are those determined from the scalar product
of forms (1.6.1).
One can easily check that the scalar curvature does not depend on the choice of
orthonormal basis.
At long last, we have developed all the tools necessary for a noncommutative
differential geometry upon which gravity may based. The culmination of this work
is the noncommutative analogue of the Einstein-Hilbert action, defined by [2]
(2.1.25) IHE (∇) = trω r∇ |D|−n .

Conclusion
In this paper we have developed a noncommutative analogue of Riemannian
geometry. For a manifold M , with its canonical spectral triple, the classical and
noncommutative geometries agree. While this is interesting in and of its own right,
there are far greater implications. In [6], Connes and Lott modelled space-time
by the so-called Connes-Lott space M × Y , the product of a four-dimensional spin
manifold M with a discrete internal space Y consisting of two points. Using the
simply the noncommutative geometry of M × Y , Connes and Lott derived a La-
grangian reproducing the Standard Model. Applying the techniques developed in
this paper to the Connes-Lott space, as in [2], gives us an exciting, alternative
method for bringing about the unification of gravity with quantum mechanics.
 GRAVITY IN NONCOMMUTATIVE GEOMETRY 13

References
[1] A.H. Chamseddine, J. Fröhlich, Some Elements of Connes’ Non-Commutative Geometry,
And Space-Time Geometry, hep-th/9307012.
[2] A.H. Chamseddine, J. Fröhlich, O. Grandjean, The Gravitational Sector in the Connes-Lott
Formulation of the Standard Model, hep-th/9503093, J.Math.Phys. 36 (1995) 6255-6275.
[3] F. Cipriani, D. Guido, S. Scarlatti, A Remark on Trace Properties of K-Cycles, J. Oper.
Theory 35 (1996) 179-189.
[4] A. Connes, Noncommutative Geometry (Academic Press, 1994).
[5] A. Connes, Noncommutative Geometry and Reality, J. Math. Phys. 36 (1995) 6194-6231.
[6] A. Connes and J. Lott, Particle models and noncommutative geometry, Nucl. Phys. B (Proc.
Suppl.) 18 (1990), 29-47.
[7] J. Cuntz, D. Quillen, Algebra extension and nonsingularity, J. Amer. Math. Soc.
textbf8 (1995) 251-289.
[8] J. Dixmier, Existence de traces non normal, C.R. Acad. Sci. Paris, Ser. A-B 262 (1966)
A1107-A1108
[9] I. M. Gelfand and M. A. Naimark, On the embedding of normed rings into the ring of
operators in Hilbert space, Mat. Sbornik 12 (1943), 197-213.
[10] G. Landi, An Introduction to Noncommutative Spaces and Their Geometries, Lecture Notes
in Physics. New Series M, Monographs No. 51, Springer-Verlag, Heidelberg, 1997.
[11] G. Landi, C. Rovelli, General Relativity in Terms of Dirac Eigenvalues, Phys. Rev. Lett. 78
(1997) 3051-3054.
[12] G. Landi, C. Rovelli, Kinematics and Dynamics of General Relativity with Noncommutative
Geometry, in preperation.
[13] B. Simon, Trace Ideals and Their Applications, London Mathematical Society Lecture Notes
35 (Cambridge University Press, 1979).
[14] R. G. Swan, Vector bundles and projective modules, Trans. Amer. Math. Soc. 105 (1962),
264-277.
[15] N. E. Wegge-Olsen, K-theory and C ∗ -algebras: a Friendly Approach (Oxford Univ. Press,
1993).