Analysis of 1 and 2 Particle Quantum Systems

using Geometric Algebra

Rachel Parker1 and Chris Doran2

Astrophysics Group, Cavendish Laboratory, Madingley Road,

Cambridge CB3 0HE, UK.
arXiv:quant-ph/0106055v1 11 Jun 2001

Abstract
When two or more subsystems of a quantum system interact with each
other they can become entangled. In this case the individual subsystems
can no longer be described as pure quantum states. For systems with
only 2 subsystems this entanglement can be described using the Schmidt
decomposition. This selects a preferred orthonormal basis for express-
ing the wavefunction and gives a measure of the degree of entanglement
present in the system. The extension of this to the more general case
of n subsystems is not yet known. We present a review of this process
using the standard representation and apply this method to the geometric
algebra representation. This latter form has the advantage of suggesting
a generalisation to n subsystems.

1 Introduction
Quantum entanglement in 2-particle systems is currently well understood (for
a useful review, including an extensive list of references, see [1]). But the quan-
tum behaviour of many-particle systems is more complicated and less well un-
derstood, and it is these systems that are of interest experimentally. The main
limitation to the theoretical understanding of such systems is that the techniques
which have been developed to analyse 2-particle systems do not easily gener-
alise. Geometric algebra has the advantage that the number of particles being
analysed dictates the size of the space but otherwise does not alter the analy-
sis used. In this way, results developed in simple cases (such as the 2-particle
system) can be more easily generalised to the n-particle case.
In this paper we focus on 2-state quantum systems in the cases of one and two
particles. We start by reviewing the standard matrix-based approach to single
and two-particle pure states. We describe the Schmidt Decomposition, which
provides a measure of the degree of entanglement present in a given system.
We then introduce the density matrix to describe both pure and mixed states
in a unified manner. We next turn to an analysis of the same systems using
the multiparticle spacetime algebra framework developed by Doran, Lasenby
and Gull [2, 3, 4]. As a simple application we review the properties of the spin
singlet state, frequently encountered in discussions of the Bell inequalities and
EPR-type experiments [5].

1 e-mail: [email protected]
2 e-mail: [email protected], http://www.mrao.cam.ac.uk/∼cjld1/

1
2 Single-Particle Pure States
If there is only one particle present in the system then the spinor, |ψi ∈ H, can
always be written in the form

|ψi = c0 |0i + c1 |1i (1)

where |0i and |1i are some pair of orthogonal basis states and c0 and c1 are
complex coefficients. Alternatively, all information about the state of the par-
ticle can be expressed in a polarisation (or spin) vector, P , whose components
are given by

Pi = hσ̂i i = hψ|σ̂i |ψi/hψ|ψi (2)

where the σ̂i are the Pauli matrices

     
0 1 0 −i 1 0
σ̂1 = , σ̂2 = , σ̂3 = . (3)
1 0 i 0 0 −1

It follows from this definition that |P | = 1. In this way the spin state of
the particle can be expressed graphically as a point on the 2-sphere, and any
evolution of the state of the particle can be thought of as a rotation of the
polarisation vector. In many applications this sphere is known as the Bloch
sphere.

3 2-Particle Systems
Suppose that two particles are described by states belonging to individual
Hilbert spaces H1 and H2 . The joint Hilbert space for the interacting system
is H1 ⊗ H2 , consisting of complex superpositions of tensor products of states in
the individual spaces. A basis for H1 ⊗ H2 is constructed by taking the tensor
products of the basis vectors for H1 and H2 . Therefore any pure state of the
composite system, |ψi ∈ H1 ⊗ H2 , can be expressed as
X X
|ψi = ci,j |ii ⊗ |ji ≡ ci,j |i, ji. (4)
i,j=0,1 i,j=0,1

If more than one of the ci,j ’s are non-zero then each subsystem is no longer in
a pure state and the system is entangled. In order to quantify the degree of
entanglement we re-express equation (4) in the form

|ψi = cos(α/2)|0′ , 0′ i + sin(α/2)|1′ , 1′ i, (5)

which can always be done via a suitable change of basis. Here, |0′ i and |1′ i
are orthonormal vectors in H1 and H2 and 0 ≤ α ≤ π/2 (so that cos(α/2) >
sin(α/2)). Then |0′ , 0′ i can be thought of as the separable state ‘closest’ to |ψi
and α as the degree of entanglement present in the system. The procedure by
which this basis is constructed is the Schmidt Decomposition, which we now
describe.

2
3.1 Schmidt Decomposition
An arbitrary wavefunction |ψi can be rewritten as a sum of two state vectors
through a transformation of the basis vectors in the following way [6]. Let |ui
and |vi be unit vectors of the first and second particles respectively. Define M
by
M = hu, v|ψi. (6)
|M |2 is nonnegative and bounded so it attains its maximum, |M1 |2 , for some
|u1 i and |v1 i. The choice of |u1 i and |v1 i is not unique since they can at most be
determined only up to phase and it is possible that there are other degeneracies
as well.
Let |u′ i be any state of the first particle which is orthogonal to |u1 i, and let
ǫ be an arbitrarily small number. Then
2
| |u1 i + ǫ|u′ i| = 1 + O(ǫ2 ) (7)
so that up to order ǫ2 , |u1 i + ǫ|u′ i is a unit vector. We then find that
hu1 + ǫu′ , v1 |ψi = M1 + ǫhu′ , v1 |ψi (8)
so
2
|hu1 + ǫu′ , v1 |ψi| = |M1 |2 + 2Re(M1 ǫhu′ , v1 |ψi) + O(ǫ2 ). (9)
But |u1 i was chosen so that the scalar product in Eq. (6) is a maximum. There-
fore we must have that
|M1 |2 + 2Re(M1 ǫhu′ , v1 |ψi) ≤ |M1 |2 ,
=⇒ Re(M1 ǫhu′ , v1 |ψi) ≤ 0. (10)
The choice of phase of |v1 i is arbitrary, however, so to ensure that Eq. (10) is
satisfied we must have that
u⊥
hu′ , v1 |ψi = 0, ∀ u′ ∈ H1 1 = {u′ ∈ H1 |hu1 |u′ i = 0}. (11)
Similarly, we can show that the same restriction applies to the second particle
so that
v⊥
hu1 , v ′ |ψi = 0, ∀ v ′ ∈ H21 = {v ′ ∈ H2 |hv1 |v ′ i = 0}. (12)
If we now define a new wavefunction
|ψ ′ i = |ψi − M1 |u1 , v1 i (13)
then |ψ ′ i also satisfies Eqs. (11) and (12) and has the additional property that
hu1 , v1 |ψ ′ i = 0. (14)
u⊥ v1⊥
From this it follows that ψ ′ ∈ H1 ⊗ H2 and we can repeat the above process
1

u⊥ v⊥
on ψ ′ . Importantly, the dimension of H1 1 ⊗ H21 is smaller than the dimension
of H1 ⊗ H2 so this process must terminate. We finally obtain
X
|ψi = Mi |ui , vi i, (15)
i

3
where the sum is over the smaller dimension of H1 and H2 and {|ui i} and {|vi i}
are orthonormal sets.
For the case where each subsystem has dimensionality 2 we find that

|ψi = M1 |u1 , v1 i + M2 |u2 , v2 i. (16)

The phases of M1 and M2 can be absorbed into |u1 , v1 i and |u2 , v2 i so we can
set them to be real. This decomposition can be written explicitly as
−iφ1 /2
cos(θ2 /2)e−iφ2 /2
    
1/2 iχ iτ /2 cos(θ1 /2)e
|ψi =ρ e cos(α/2)e ⊗
sin(θ1 /2)eiφ1 /2 sin(θ2 /2)eiφ2 /2
sin(θ1 /2)e−iφ1 /2 sin(θ2 /2)e−iφ2 /2
   
−iτ /2
+ sin(α/2)e ⊗ . (17)
− cos(θ1 /2)eiφ1 /2 − cos(θ2 /2)eiφ2 /2

This is the Schmidt decomposition for a bipartite 2-state system. In writing this
2
we have satisfied the condition that cosα ≥ sinα since otherwise |hu2 , v2 |ψi| >
2
|hu1 , v1 |ψi| which contradicts our choice of |u1 , v1 i.

3.2 The Density Matrix

If we want to calculate expectation values for one particle only and the state
of the other particle is unknown then clearly we cannot write down the full
wavefunction. We are forced instead to turn to the density operator, ρ̂, defined
by
X
ρ̂ = |ψihψ| = ci,j c∗k,l |iihk| ⊗ |jihl| (18)
i,j,k,l

In terms of the density operator the expectation value of any observable Q̂ is

given by

hQ̂i = tr(ρ̂Q̂). (19)

The density operator for each particle is given by

X X
ρ̂1 = tr2 ρ̂ = hj|ρ̂|ji ρ̂2 = tr1 ρ̂ = hi|ρ̂|ii (20)
j i

so that the expectation value for the ith particle can be calculated by

hQii = tr(ρ̂i Q). (21)

For systems entangled with an (unknown) environment the density matrix rep-
resents our ultimate state of knowledge of the system. This has important
consequences for the interpretation of quantum mechanics. For a recent review
of these ideas, see Paz & Zurek [7].

4 Geometric Algebra
Geometric algebra (GA) is essentially Clifford algebra with added geometric
content. Since Clifford algebras are a fundamental part of the treatment of 2-
state quantum systems (through the description of quantum spin), we expect

4
that formulating the theory in a GA framework should bring added geometric
insight. This idea was first explored by Hestenes in a series of papers dating
back to the sixties [8, 9, 10]. We start by reviewing the treatment of single-
particle systems. These are described within the GA of 3D space, denoted G3 .
As an orthonormal basis for this we take

1, {σk }, {Iσk }, I = σ1 σ2 σ3 . (22)

The reverse operation (which flips signs of bivectors and trivectors) is denoted
with a tilde, and angle brackets hM ik are used to project onto the grade-k part
of M . For the projection onto the scalar part we simply write hM i. For an
introduction into the geometric algebra of 3D space see [11, 12].

4.1 Single-Particle Systems

The simplest example of a 2-state system is provided by quantum spin. Spin
states can be represented as complex 2-component vectors known as spinors.
These can be given a more natural encoding within G3 by defining a linear
one-to-one map between the state (as a complex vector) and a multivector con-
structed from even-grade terms (scalars and bivectors). The simplest such map-
ping is defined by [2, 3]
 
a0 + ia3
|ψi = ↔ ψ = a0 + ak Iσk , (23)
−a2 + ia1

so that the basis elements |0i and |1i map as

|0i ↔ 1 and |1i ↔ −Iσ2 . (24)

In this way ψ sits inside the space spanned by {1, Iσk }, (k = 1, 2, 3). It follows
that

ψ ψ̃ = (a0 + ak Iσk )(a0 − ak Iσk ) = a20 + a21 + a22 + a23 = ψ̃ψ ≡ ρ (25)

where ρ is the scalar magnitude of the state vector. The multivector ψ can then
be written as

ψ = ρ1/2 R. (26)

R is then an even, normalised multivector in 3-dimensions and so is a rotor —

a generator of rotations.
The action of the Pauli matrices of Eq. (3) is given by

σ̂k |ψi ↔ σk ψσ3 = −Iσk ψIσ3 . (27)

It follows that multiplication by i is represented by

i|ψi = σ̂1 σ̂2 σ̂3 |ψi ↔ ψIσ3 . (28)

To construct observables we define the inner product of two spinors, ψ and φ

by

hψ|φi ↔ (ψ, φ)s = hφψ̃i − hφIσ3 ψ̃ii. (29)

5
As we will see shortly, this definition generalises simply to multiparticle systems.
From Eq. (29) the probability density is
hψ|ψi ↔ (ψ, ψ)s = hψ ψ̃i − hψIσ3 ψ̃iIσ3 . (30)

But ψIσ3 ψ̃ reverses to give minus itself, so it contains no scalar part. This
leaves
hψ|ψi ↔ (ψ, ψ)s = hψ ψ̃i. (31)

For the 1-particle case ψ ψ̃ is purely a scalar and is equal to ρ. In the more general
case of n-particles we cannot assume that ψ ψ̃ is purely scalar and so Eq. (31),
suitably normalised, provides the most general definition for the probability
density.
The other observable we can construct is the expectation for the spin in the
k-direction. This is given by
hψ|σk |ψi/hψ|ψi ↔ ρ−1 (ψ, −Iσk ψIσ3 )s
=ρ−1 h−Iσk ψIσ3 ψ̃i − ρ−1 h−Iσk ψIσ3 Iσ3 ψ̃iIσ3
= − ρ−1 Iσk · hψIσ3 ψ̃i2 − ρ−1 hIσk ψ ψ̃iIσ3 . (32)

Since ψ ψ̃ is a scalar hIσk ψ ψ̃i = 0. Also, ψIσ3 ψ̃ reverses to give minus itself and
has even grade, so is a pure bivector (again in the multiparticle space we cannot
make this assumption). Therefore, using Eq. (26) we can define the polarisation
bivector by
P = hρ−1 ψIσ3 ψ̃i2 = hRIσ3 R̃i2 (33)
so that
Pk = hψ|σk |ψi/hψ|ψi = −Iσk · hRIσ3 R̃i2 = −Iσk · P. (34)
In this way the spin of the particle can be thought of as a rotation of the
Iσ3 plane, where the rotation is given by the wave function of the particle.
The expectation value for the polarisation in the k-direction is then simply the
component of P in the k-direction. This was Hestenes’ original insight. A
challenge is to extend these ideas to the multiparticle framework.
As an example, consider the wavefunction employed in the Schmidt decom-
position (17)
 
cos(θ/2)e−iφ/2
|ψi = . (35)
sin(θ/2)eiφ/2
In our single particle space this becomes
ψ = cos(θ/2)e−φIσ3 /2 − sin(θ/2)Iσ2 eφIσ3 /2
= e−φIσ3 /2 (cos(θ/2) − sin(θ/2)Iσ2 )
= e−φIσ3 /2 e−θIσ2 /2 (36)
and the polarisation vector P is simply
P = ψIσ3 ψ̃ = e−φIσ3 /2 e−θIσ2 /2 Iσ3 eθIσ2 /2 eφIσ3 /2
= sin(θ) cos(φ)Iσ1 + sin(θ) sin(φ)Iσ2 + cos(θ)Iσ3 (37)

6
4.2 2-Particle Systems
States for 2-particle systems are constructed in the multiparticle spacetime al-
gebra (MSTA), which is built from n-particle relativistic configuration space. A
basis for this space is provided by the vectors {γµa } where the superscript labels
the individual particle space. Vectors from different spaces are orthogonal and
so anticommute. It follows that bivectors from different spaces commute, and
hence the even subalgebra of the MSTA contains the tensor product of a set
of non-relativistic algebras. This is precisely the algebra needed to construct a
multiparticle wavefunction. A basis for a 2-particle wavefunction is provided by
sums and products of the 1-particle basis elements {1, Iσj1 , Iσk2 }, where

σka = γka γ0a , I a = γ0a γ1a γ2a γ3a , (no sum). (38)

Again, the superscript denotes the particle label, and we abbreviate I 1 σk1 to
Iσk1 , etc.
Currently our basis set gives 4 × 4 = 16 real degrees of freedom, whereas
we should have only 8 for a 2-particle state. The solution to this problem
is to demand a consistent meaning for the unit imaginary. In each separate
space multiplication by the imaginary corresponds to right multiplication by
Iσ3 . Since our new space has two such bivectors we require that

ψIσ31 = ψIσ32 , ψ = ψ 21 (1 − Iσ31 Iσ32 ). (39)

We therefore define the 2-particle correlator [2, 3]

E = 12 (1 − Iσ31 Iσ32 ), E 2 = E. (40)

E is a projection operator and reduces the number of degrees of freedom by a

factor of 2. The complex structure in the 2-particle algebra is now defined by
the non-simple bivector J, where

EIσ31 = EIσ32 = 12 (Iσ31 + Iσ32 ) ≡ J. (41)

The two particle spinor |ψ, φi is now mapped to the multivector

|ψ, φi ↔ ψ 1 φ2 E, (42)

where the superscripts again denote which space the multivector inhabits. The
result of the action of the unit imaginary becomes

i|ψ, φi ↔ ψ 1 φ2 EIσ31 = ψ 1 φ2 EIσ32 = ψ 1 φ2 J. (43)

Consistency in this formulation is ensured by the results

J 2 = −E and J = JE = EJ. (44)

The action of the individual Pauli matrices now becomes, for example,

σ̂k ⊗ Iˆ |ψi ↔ −Iσk1 ψJ (45)

where Iˆ is the 2×2 identity matrix. A similar result holds for the second particle
space. The action on the right-hand side keeps us in the space of correlated

7
products of even elements of G3 . The quantum inner product is replaced by the
operation

hψ|φi ↔ (ψ, φ)s = 2hφE ψ̃i − 2hφJ ψ̃ii. (46)

The factor of E in the real part is not strictly necessary as it is always present
in the spinors, but including it does provide a neat symmetry between the real
and imaginary parts. The factor of 2 is included to ensure complete consistency
with the standard quantum inner product. (In the general n-particle case a
factor of 2n−1 is required.)
In Section 3.1 we found that a general 2-particle wavefunction can be written
in the form of Eq. (17). To find the geometric algebra form of this we first define
the spinor

ψ(θ, φ) = e−φIσ3 /2 e−θIσ2 /2 . (47)

We also need a representation of the orthogonal state to this, which is

 
sin(θ/2)e−iφ/2
↔ sin(θ/2)e−Iσ3 φ/2 + cos(θ/2)Iσ2 eIσ3 φ/2
− cos(θ/2)eiφ/2
=ψ(θ, φ)Iσ2 . (48)

It is a straightforward exercise to confirm that this state is orthogonal to ψ(θ, φ),

as required. We can now construct the MSTA version of the Schmidt decompo-
sition. We replace Eq. (17) with

ψ =ρ1/2 cos(α/2)ψ 1 (θ1 , φ1 )ψ 2 (θ2 , φ2 )eJτ /2

+ sin(α/2)ψ 1 (θ1 , φ1 )ψ 2 (θ2 , φ2 )Iσ21 Iσ22 e−Jτ /2 eJχ E
=ρ1/2 ψ 1 (θ1 , φ1 )ψ 2 (θ2 , φ2 )eJτ /2 cos(α/2) + sin(α/2)Iσ21 Iσ22 eJχ E. (49)

If we now define the individual rotors

R = ψ(θ1 , φ1 )eIσ3 τ /4 , S = ψ(θ2 , φ2 )eIσ3 τ /4 , (50)

then the wavefunction ψ can be written compactly as

ψ = ρ1/2 R1 S 2 cos(α/2) + sin(α/2)Iσ21 Iσ22 eJχ E.

(51)

This gives a neat, general form for an arbitrary 2-particle state. In particular,
all reference to the tensor product has been dropped in favour of the somewhat
simpler geometric product. The degrees of freedom are held in an overall mag-
nitude and phase, two separate rotors in the individual particle spaces, and a
single entanglement angle θ. In total this gives 9 degrees of freedom, so one of
them must be redundant. This redundancy is in the single-particle rotors. If
we take

R 7→ ReIσ3 β , S 7→ Se−Iσ3 β (52)

then the overall wavefunction ψ is unchanged. In practice this redundancy is

not a problem, and the form of (51) turns out to be extremely useful.

8
 The GA form of the Schmidt decomposition in (51) is very suggestive of
a more general pattern. To the left we have rotation operators in each of the
individual spaces. In one sense the rotors R1 S 2 can be viewed as representing
the nearest direct product (separable) state. Next comes a term describing the
2-particle entanglement. The generalisation seems fairly clear. For a 3-particle
system we expect to see terms describing the various 2-particle entanglements,
followed by a term for the 3-particle entanglement. Finding precisely the optimal
decomposition along these lines is an open problem, but the GA formalism
has suggested an approach to the general problem of classifying multiparticle
entanglement which has not been tried before.

4.3 2-Particle Observables

We can start to appreciate the utility of the form of (51) by studying the 2-
particle observables. These go as, for example

hψ| σ̂k ⊗ Iˆ |ψi ↔ (ψ, −Iσk1 ψJ)s = −2Iσk1 ·(ψJ ψ̃) (53)

and

hψ| σ̂j ⊗ σ̂k |ψi ↔ (ψ, −Iσj1 Iσk2 ψ)s = −2(Iσj1 Iσk2 )·(ψE ψ̃). (54)

All of the observables one can construct are therefore contained in the multivec-
tors ψE ψ̃ and ψJ ψ̃. This is true in the general n-particle case, and is a major
strength of the MSTA approach.
To study the form of the observables we first simplify slightly and set ρ = 1.
We find that (using E Ẽ = EE = E)

ψE ψ̃ = R1 S 2 cos(α/2) + sin(α/2)Iσ21 Iσ22 E

cos(α/2) + sin(α/2)Iσ21 Iσ22 R̃1 S̃ 2

= R1 S 2 1 + sin(α)Iσ21 Iσ22 E R̃1 S̃ 2

(55)

Substituting in the form of E from Eq. (40) gives

 
ψE ψ̃ = 21 R1 S 2 1 − Iσ31 Iσ32 + sin(α)(Iσ21 Iσ22 − Iσ11 Iσ12 ) R̃1 S̃ 2 . (56)

To make this result clearer we introduce the notation

Ak = RIσk R̃, Bk = SIσk S̃ (57)

so that

ψE ψ̃ = 21 (1 − A13 B32 ) + 1
2 sin(α)(A12 B22 − A11 B12 ). (58)

From this we see that

hψE ψ̃i = 12 . (59)

This factor of one-half is absorbed by the factor of 2 in the definition of the

quantum inner product (46) and shows that the state is correctly normalised
to 1. The 4-vector part of the observable is more interesting, as it contains
combinations of A1 , A2 , B1 , B2 , none of which are accessible to measurement

9
in the single-particle case (as they are not phase invariant). This is one place
where differences between classical and quantum models of spin start to emerge.
The second observable to form from the 2-particle state ψ is ψJ ψ̃, which is
given by

ψJ ψ̃ = R1 S 2 cos(α/2) + sin(α/2)Iσ21 Iσ22 J

cos(α/2) + sin(α/2)Iσ21 Iσ22 R̃1 S̃ 2

= 21 R1 S 2 cos2 (α/2) − sin2 (α/2) (Iσ31 + Iσ32 )R̃1 S̃ 2

= 1
2 cos(α)(A13 + B32 ) (60)

This result extends the definition of the polarisation bivector to multiparticle

systems. An immediate consequence of this definition is that the lengths of the
bivectors are no longer fixed, but instead depend on the entanglement.

4.4 The Density Matrix

The density matrix for a normalised 2-particle pure state can be expanded in
terms of products of Pauli matrices as
1 ˆ ˆ
I ⊗ I + ak σ̂k ⊗ Iˆ + bk Iˆ ⊗ σ̂k + cjk σ̂j ⊗ σ̂k .


ρ̂ = |ψihψ| = (61)
4
The various coefficients are found by forming, for example

ak = hψ| σ̂k ⊗ Iˆ |ψi = −2Iσk1 ·(ψJ ψ̃) (62)

It follows that all of the degrees of freedom present in the density matrix are
contained in the multivector observables ψE ψ̃ and ψJ ψ̃. For mixed states we
simply add the weighted values of these observables formed from the pure states.
This picture is quite general and works for any number of particles. One small
complication is that the terms in ψJ ψ̃ are anti-Hermitian, whereas the density
matrix is Hermitian. One way round this is to correlate all of the pseudoscalars
together and map all bivectors back to their dual vectors [13] . One can often
ignore this feature, however, and work directly with the observables ψE ψ̃ and
ψJ ψ̃.
An advantage of this way of encoding the density matrix is that the partial
trace operation to form the reduced density matrix simply consists of throwing
away any terms in the observables coming from spaces where the state is un-
known. For example, taking the 2-particle entangled state (51) and tracing out
the degrees of freedom in space 2 just leaves

ρ̂ = 21 (1 + Pk σ̂k ), Pk = (−Iσk )· cos(α)RIσ3 R̃ .

(63)

This shows that the effect of the entanglement is to reduce the expectation value
for the polarisation from 1 to cos(α), but leave the direction of polarisation
unchanged. For 2-particle pure states we also see that the polarisation vector
is the same length for both particles, so each particle is effected equally by any
entanglement which is present. For higher particle number or mixed states the
effects of entanglement are more complicated, though the formula

Pk = −2n−1 (Iσka )·(ψJ ψ̃) (64)

10
holds whenever we form the reduced density matrix for particle a from a larger,
entangled state.
Our simple 2-particle system exhibits one of the basic results of quantum
theory. When a system entangles with a second, unknown system (usually
the environment), the state of the system of interest can no longer be known
for certain and we are forced to adopt a density matrix viewpoint. That is,
entanglement with the environment leads to decoherence and loss of information.
A useful application of the preceding is to the overlap probability for the
inner product of two states. Given two normalised states we have

P (ψ, φ) = |hψ|φi|2 = tr(ρ̂ψ ρ̂φ ). (65)

The degrees of freedom in the density matrices are contained in ψE ψ̃ and ψJ ψ̃,
with equivalent expressions for φ. One can then show that the probability is
given by the compact expression

P (ψ, φ) = h(ψE ψ̃)(φE φ̃)i − h(ψJ ψ̃)(φJ φ̃)i. (66)

This formula holds in the n-particle case as well, except for the presence of
an additional factor of 2n−2 to give the correct normalisation. This compact
expression is a unique feature of the MSTA approach.
As a check on the preceding, suppose we have two separable states

ψ = R1 S 2 E, φ = U 1V 2E (67)

with

ψE ψ̃ = 21 (1 − A1 B 2 ), φE φ̃ = 12 (1 − C 1 D2 ). (68)

We find that

P (ψ, φ) = 14 h(1 − A1 B 2 )(1 − C 1 D2 ) − (A1 + B 2 )(C 1 + D2 )i

= 14 (1 + A·C B ·D − A·C − B ·D)
= 21 (1 − A·C) 21 (1 − B ·D) (69)

which shows that the probability is the product of the separate single-particle
probabilities. If one of the states is entangled this result no longer holds.

4.5 Example – The Singlet State

As a simple example of the some of the preceding ideas, consider the spin-singlet
state
1 1
|ψi = √ (|01i − |10i) ↔ ψ = √ (Iσ21 − Iσ22 )E. (70)
2 2
This state is maximally entangled (α = π/2), and isotropic. Forming the two
observables we find that

ψE ψ̃ = 12 (1 + Iσk1 Iσk2 ) (71)

and

ψJ ψ̃ = 0. (72)

11
It follows that the reduced density matrix for either particle space is simply
one-half of the identity matrix, and so all directions are equally likely. If we
align our measuring apparatus along some given axis and measure the state of
particle one then both up and down have equal probabilities of one-half.
Suppose now that we construct a joint measurement on the singlet state.
We can model this as the overlap probability between ψ and the separable state

φ = R1 S 2 E. (73)

Denoting the spin directions by

RIσ3 R̃ = P, SIσ3 S̃ = Q, (74)

we find that, from (66)

P (ψ, φ) = h 12 (1 − P 1 Q2 ) 12 (1 + Iσk1 Iσk2 )i

= 14 (1 − P ·(Iσk ) Q·(Iσk ))
= 14 (1 − cos θ) (75)

where θ is the angle between the spin bivectors P and Q. So, for example, the
probability that both measurements result in the particles having the same spin
(θ = 0) is zero, as expected. Similarly, if the measuring devices are aligned, the
probability that particle one is up and particle two is down is one-half, whereas
if there was no entanglement we should get the product of the separate single
particle measurements (resulting in 1/4).
It is instructive to see how all of strange quantum entanglement results for
the singlet state are contained in the 4-vector part of the observables. This
reveals some of the complex geometry associated with multiparticle quantum
mechanics. And this is only for 2-particle systems! Most proposals for quantum
computers have in mind a far greater number of entangled qubits. We hope that
this paper has demonstrated some of the potential power of geometric algebra
for helping to navigate through these large Hilbert spaces.

5 Acknowledgements
Rachel Parker is supported by the Cambridge Commonwealth Trust and the
Kerry Packer Scholarship Foundation. Chris Doran is supported by the EPSRC.
The authors thank Anthony Lasenby and Timothy Havel for helpful discussions.

References
[1] S.J. Lomonaco. An entangled tale of quantum entanglement. quant-
ph/0101120.
[2] C.J.L. Doran, A.N. Lasenby, and S.F. Gull. States and operators in the
spacetime algebra. Found. Phys., 23(9):1239, 1993.
[3] C.J.L Doran, A.N. Lasenby, S.F. Gull, S.S. Somaroo, and A.D. Challinor.
Spacetime algebra and electron physics. Adv. Imag. & Elect. Phys., 95:271,
1996.

12
 [4] S.S. Somaroo, A.N. Lasenby, and C.J.L. Doran. Geometric algebra and
the causal approach to multiparticle quantum mechanics. J. Math. Phys.,
40(7):3327–3340, 1999.
[5] J.S. Bell. Speakable and unspeakable in quantum mechanics. Cambridge
University Press, 1987.
[6] A. Ekert and P.L. Knight. Entangled quantum systems and the Schmidt
decomposition. Am. J. Phys., 63(5):415, 1995.
[7] J.P. Paz and W.H. Zurek. Environment-induced decoherence and the tran-
sition from quantum to classical. quant-ph/0010011.
[8] D. Hestenes. Real spinor fields. J. Math. Phys., 8(4):798, 1967.
[9] D. Hestenes. Vectors, spinors, and complex numbers in classical and quan-
tum physics. Am. J. Phys., 39:1013, 1971.
[10] D. Hestenes and R. Gurtler. Consistency in the formulation of the Dirac,
Pauli and Schrödinger theories. J. Math. Phys., 16(3):573, 1975.
[11] D. Hestenes. New Foundations for Classical Mechanics (Second Edition).
Kluwer Academic Publishers, Dordrecht, 1999.
[12] C.J.L. Doran and A.N. Lasenby. Physical applications of geometric alge-
bra. Cambridge University Lecture Course. Lecture notes available from
http://www.mrao.cam.ac.uk/∼clifford/ptIIIcourse.
[13] T.F. Havel and C.J.L. Doran. Geometric algebra in quantum information
processing. quant-ph/0004031, AMS Contemporary Math series, to appear.

13