Bloch sphere model for two-qubit pure states

Chu-Ryang Wie1

State University of New York at Buffalo, Department of Electrical Engineering, 230B Davis Hall, Buffalo,
NY 14260
(Dated: March 28, 2014)

The two-qubit pure state is explicitly parameterized by three unit 2-spheres and a phase factor.
For separable states, two of the three unit spheres are the Bloch spheres of each qubit with
coordinates (A,A) and (B,B). The third sphere parameterizes the degree and phase of
concurrence, an entanglement measure. This sphere may be considered a ‘variable’ complex
imaginary unit t where the stereographic projection maps the qubit-A Bloch sphere to a
complex plane with this variable imaginary unit. This Bloch sphere model gives a consistent
description of the two-qubit pure states for both separable and entangled states. We argue that
the third sphere (entanglement sphere) parameterizes the nonlocal properties, entanglement and
a nonlocal relative phase, while the local relative phases are parameterized by the azimuth
angles, A and B, of the two quasi-Bloch spheres. On these three unit spheres and a phase
factor (B), the two-qubit pure states and unitary gates can be geometrically represented.
Accomplished by means of Hopf fibration, the complex amplitudes (, , , ) of a two-qubit
pure state and the Bloch sphere parameters are related by a single quaternionic relation:
 θ 
cos A 
   j   2    
    cos B  sin B ekB j  e k B
    j   sin θ A etA   2 2 
 
 2 

described by a unit 7-sphere S7 for pure states and by a

four-dimensional special unitary group SU(4) for a mixed
I. INTRODUCTION state density matrix and two-qubit operators. There have
been various attempts to parameterize the two qubit state
The single qubit Bloch sphere provides a useful means of
space, including an explicit parameterization of SU(4) [4],
visualizing the state and a neat description of many single
and construction of special unitary group on two qubit
qubit operations, enabling physical intuitions and serving
Hilbert space using geometric algebra of a six-
as an excellent testbed for ideas about quantum
dimensional real Euclidean vector space [5]. Havel and
computation and information [1]. However there is no
Doran presented, as a special case to their discussion of
simple generalization of Bloch sphere for multiple qubits,
SU(4) mixed state model, an explicit parameterization of
not even for two qubits. Much that is weird and
a two-qubit pure state with seven angle parameters which
wonderful about quantum mechanics can be appreciated
was written as a convex combination of tensor products of
by considering the properties of the quantum states of two
one-qubit states [5]. Mosseri and Dandoloff applied
qubits [2]. Any multi-qubit quantum logic gates can be
Hopf fibration and attempted to develop a Bloch sphere
built out of a two-qubit operation and a number of single-
model for two-qubit pure state [6,7]. They proposed a 3-
bit operations [1-3]. All of the above points give a strong
dimensional ball with equal-concurrence concentric
reason to develop a practical Bloch sphere-like tool for
spherical shells as a possible model, which obviously has
the two-qubit states and gates. In spite of a considerable
far too few degrees of freedom to specify the two-qubit
amount of efforts spent to develop a feasible two-qubit
states. However, their approach with Hof fibration has a
Bloch sphere model [4-7], there still is no Bloch sphere
close connection to our work in this paper.
model that is simple enough to serve as useful a tool as
the single qubit Bloch sphere does for state representation In this paper, we limit to the two-qubit pure states and
and manipulation. attempt to find a Bloch sphere model that is simple and
practical. We parameterize the S7 two-qubit pure state
For complete description, two qubit states require seven
space by 7 angle parameters which are grouped into the
parameters for pure states and fifteen parameters for
two Bloch coordinates (A, A) and (B, B), the
general mixed states. The two qubit state space is usually
entanglement and its relative phase (, ), and a phase

1
Electronic address: [email protected]
 2

angle (B). These seven angle parameters are related to Throughout this paper, k will be assumed to be the
the two-qubit quantum state amplitudes, , ,  and  via imaginary unit of ordinary complex numbers. A
a single equation. The rest of this article is organized as quaternion forms non-commutative algebra and may be
follows: In section II, we briefly review the Hopf map written in terms of two complex numbers as in Eqs.(2)
with the non-commutative quaternion algebra. The Hopf and (3), or in terms of four real numbers, a, b, c and d as
fibration maps the S7 two-qubit pure state space to an S4
base space. This is done by mapping to each point of the q  a  bi  cj  dk , a, b, c, d  (5)
S4 base space the entire S3 fiber space. This Hopf map
enables us to model a quantum state in the S7 space by a A pure imaginary unit quaternion squares to -1 like the
product of a function in the S4 base space and a function three imaginary units in Eq.(4) and thus, a pure unit
in the S3 fiber space. In section III, we define the three quaternion t provides the same relation between an
unit 2-spheres and a phase factor to serve as the Bloch exponential function and trigonometric (sine and cosine)
sphere model for two-qubit pure states. In section IV, we functions as the complex imaginary unit does, and it can
present a streamlined procedure for, given the four be parameterized in terms of two real numbers. That is,
complex state amplitudes (, ,  and ), how to find the
seven angle parameters of the Bloch spheres, and vice t 2  1; et  cos  tsin ;
. (6)
versa. In this section we also show some concrete t  i sin cos  j sin sin  k cos  ;  ,  ,  
examples of Bloch coordinates of the maximally
entangled Bell states and several separable states, and
The normalization condition
examples of the two-qubit quantum gates, CNOT, CZ and
SWAP gates. In section V, we discuss the entanglement
q0 A  q1 A  q0 B  q1B          1 (7)
2 2 2 2 2 2 2 2
sphere parameters (, ), phase angle parameter (B), and
the mixed-state single-qubit Bloch ball after partial trace
and its relation to the quasi-Bloch sphere. In section VI, identifies the state space of a two-qubit pure state as a unit
8
we give a summary conclusion. 7-sphere, S7, which is embedded in .

The Hopf map from the two-qubit state space S7 to the

4
base S4 is a composition of two maps: first from S7 to
II. THE HOPF FIBRATION OF TWO-QUBIT
(+), then to S4 by an inverse stereographic projection [7,
STATES AND PARAMETERIZATION
8]. An explicit treatment of conversion from the complex
For two-qubit pure states,
amplitudes,  and , to the quaternion ‘amplitudes’
q0 and q1 is useful in relating the target space of the Hopf
ΨAB   00   01   10   11 (1)
map to each qubit space (and thus identifying the Bloch
parameters of each space). We do this as follows:
where,
A. In the basis of qubit-A
 ,  ,  ,  and         1,
2 2 2 2

Ψ AB  0 A   0  B 
 1 B  1 A  0  B
 1 B 
our Bloch sphere representation starts by introducing
quaternion ‘amplitudes’ by defining them as follows:
    j   q0 A  (8)
ΨA  0    j   1 A    j     
q0 A     j, q1A     j     j   q1 A 
A
(2)

or where, after the Hopf fibration, qubit-A is identified with

the base space S4 and the qubit-B with the fiber space S3.
q0 B     j, q1B     j (3) The above conversion process keeps the qubit-A basis
vectors intact while identifying the qubit-B basis vectors
where, the quaternion imaginary units i, j and k satisfy the with the quaternion units: 0  1, 1  j . This
B B
following multiplication rules and identities:
correspondence is applied again when recovering the
qubit-B state at the end of calculation. Let us define a
i 2  j 2  k 2  ijk  1, i  jk  kj , density matrix-like quantity as follows:
(4)
j  ki  ik , k  ij   ji.
 3

q     j 
 A  Ψ A Ψ A   0 A   q0 A , q1 A   B  ΨB ΨB        j,    j 
 q1 A    j
(9) (12)
0  j
   j    B      
     j,   j 
,
 j 0 
   j 
  2   2    
B   
where the over-bar indicates the quaternion or complex      2   2 
conjugation which reverses the sign of imaginary units:  
i  i, j   j and k  k . As will be shown below,
Where B is the reduced density matrix of qubit-B:
 B  trA  ΨAB ΨAB  .
all local parameters of the qubit-B are cancelled out in It is interesting to note that
this dyad Ψ A Ψ A ; and the only parameters remaining
whether we choose qubit-A basis or qubit-B basis, the
in  A are the Bloch coordinates of the base qubit (qubit- resulting quasi density matrix  consists of the reduced
A) and two nonlocal parameters (entanglement and density matrix of the base qubit plus an identical term
phase). These four parameters (of the Hopf base) are representing the concurrence. Also the quasi-density
determined from the products of the complex amplitudes matrix  possesses the pure-state density matrix
, ,  and . From Eq.(9) we obtain properties:  2   , tr(  )=1, and det(  )=0. A two-
qubit SWAP operation will send the quasi state Ψ A to
0  j
 A   A        , where
j 0 Ψ B , or the quasi-density matrix  A to  B , and vice
  2   2     versa. The trajectory of a SWAP gate on the Bloch
A    (10) spheres will be shown later.
     2   2 
 
C. Hopf fibration
Here, 2    is the concurrence, an entanglement
Before we parameterize the bipartite pure state space S7,
measure [9], and A is the reduced density matrix for we briefly review the Hopf fibration. The Hopf fibration
qubit-A given by  A  trB  Ψ AB Ψ AB  . It is worth consists of a composition of two maps: the Hopf map h1
which maps S7 to 4 , followed by an inverse
noting that the off-diagonal element of A represents the 4
stereographic projection h2 which maps to
coherence of qubit-A after qubit-B is traced out, and 4
S \(1,0,0,0,0).
Eq.(10) shows that the other off-diagonal element of  A
represents the entanglement of the composite state. q 
h1 :  0   q0 q11  Q  Q1  Q2i  Q3 j  Q4 k
B. In the basis of qubit-B  q1  4 (13)
S7
We write the bipartite state in the qubit-B basis by 2 2
q0 , q1 , Q  ; q0  q1  1; Qi  ; i  1, 2,3, 4
identifying the qubit-A basis vectors with the quaternion
units 1 and j as follows:
h2 :  Q1 , Q2 , Q3 , Q4    x0 , x1 , x2 , x3 , x4 

Ψ AB   0 A
 1 A  0 B 
  0 A
 1 A  1 B
4
S4 (14)
4

    j   q0 B  (11) x 2
i  1, xi  ,
ΨB     j  0 B      j  1 B    
i 0

    j   q1B 
The stereographic projection gives
 x x x x 
where, after the Hopf fibration, qubit-B will be in the base q0 q11  Q   Q1 , Q2 , Q3 , Q4    1 , 2 , 3 , 4  (15)
space S4 and qubit-A in the fiber space S3.  1  x0 1  x0 1  x0 1  x0 
which maps S 4 \ 1,0,0,0,0  to 4
, or vice versa. This is
The quasi density matrix is
illustrated in Fig.1.
 4

h3. Its fiber S1 is a unit circle which can be represented by

a unit complex number: ek.

D. Parameterization of the S4 Hopf base and the

S3 Hopf fiber

The above discussion of Hopf fibration of the S7 two-

qubit pure state space suggests a parameterization in
terms of two groups of parameters: the S4 base space
parameters and the S3 fiber space parameters. We first
parameterize the S4 Hopf base as follows:

Figure 1 Stereographic projection from S4 to 4 where S4

the coordinates of S4 are related to the coordinates of 4
through Eq.(15).
S7  θ  3
 cos S
 q0  2
Therefore, a pair of quaternions, q0 and q1, are mapped to   qf (19)
the coordinate (x0, x1, x2, x3, x4) on the 4-sphere S4 which  q1   sin θ et 
 
is embedded in 5 with the North pole excluded. This  2 
S3 t  i sin cos  j sin sin  k cos 
Hopf fibration reads h : S 7  S 4 where h  h2 h1 . An , (20)
 t xi  t y j  t z k ;  ,  ,  ,  
inspection shows that all the pairs of quaternions (q0, q1),
which differ only by a right multiplication of an arbitrary
unit quaternion qf, will map to the same point Q in 4 . where qf is the unit quaternion, representing the S3 Hopf

 
fiber. By substituting Eqs.(19) and (20) into Eq.(15), we
q1 q0 (q1q f ) q0 q f find
Q 2
 2
where
q1 q1q f
2
x0  cos , x1  sin cos , x2  sin sin sin cos 
qf  qf q f  q f qf 1 (16)
c cos  bt x , x3  sin sin sin sin  c sin  bt y ,
Therefore,
x4  sin sin cos   bcos   bt z ,
(21)
 q0 q f  b  sin sin , c  sin sin sin  b sin
h1 :    Q for any unit quaternion qf (17)
 q1 q f  Here, the S4 Hopf base is parameterized in terms of two
This unit quaternion qf is the S3 fiber in the Hopf map
angles  and  and a pure-imaginary unit quaternion t
S3
h : S  S 4 . It is worth repeating here that the Hopf map
7 which is parameterized by another two angles  and .
5
h maps an entire subspace S3 of S7 to a single point of the The Cartesian coordinate of S4 embedded in is (x0, x1,
base space, the 4-sphere S4. The unit quaternion qf is x2, x3, x4), as shown in Fig.1 and given in Eq.(21) in terms
S1 of these four angle parameters. We may regard the pure-
itself fibrated by another Hopf map h3: S 3  S 2 [8], imaginary unit quaternion t as a ‘variable’ complex
which maps an entire circle S1, a subspace of S3, to a imaginary unit for the complex plane onto which a 2-
single point on S2. sphere S2 with (, ) spherical coordinate is mapped by a
stereographic projection. Hence, the Hopf base S4
S3 S2
consists of a unit 2-sphere S2 (parameters: , ) sitting
h3 : q f  q f kq f (18 ) atop another S2 (parameters: , ). The latter S2
represents the ‘variable’ imaginary unit t. Also, a pure-
where k is an imaginary unit. The image of the map h3, imaginary quaternion can be viewed as a vector in a 3-
q f kq f , is another pure-imaginary unit quaternion which dimensional space with the orthogonal axes defined by
imaginary units i, j, k. The angles,  and , define the
is obtained by rotating k around an axis of rotation given Bloch coordinates of the qubit in the base space. The
by the imaginary part of qf by an angle of rotation, 2cos- pure unit quaternion t embodies the nonlocal properties
1
(Re(qf)), given by the real part of qf [8]. A pure unit such as concurrence and its phase. This will be discussed
quaternion can be parameterized by a unit 2-sphere S2 and
further in the next section. The interpretation of the S4
will form the base (or the image space) of the Hopf map
 5

base as being a 2-sphere S2 sitting on top of another 2- The Hopf base (S4) coordinates (x1, x2, x3, x4, x0) in R5 are
sphere S2 is also suggested by the following relation: separated into two S2 Cartesian coordinates, (x1, b, x0) and
  (tx, ty, tz), according to:
x0 2  x12  x2 2  x32  x4 2  x0 2  x12  b 2  t x 2  t y 2  t z 2   1 x12 + x22 + x32 + x42+ x02 = x12 + b2(tx2+ ty2+ tz2) + x02 = 1
  For the S4 Hopf base, the first S2 is qubit-A Bloch sphere
S4 S2  S2  with x1-b-x0 axis, and the second S2 parameterizes the
(22) global parameter t with the degree of entanglement and its
This equation suggests that the S4 Hopf base can be phase on the tx-ty-tz (or equivalently, i-j-k) axis. The
parameterized in terms of two independent S2 spheres. qubit-B Bloch sphere is identical to a single qubit Bloch
The first S2 is in the x1-b-x0 coordinate frame (the base sphere coordinate as it arises from the Hopf fiber, a unit
qubit Bloch sphere), and the second S2 is in the tx-ty-tz (or quaternion qB represented by 3-sphere S3 which is further
i-j-k) coordinate frame (of the nonlocal parameter t). mapped by h3 to the S2 base (Bloch sphere, coordinates B
and B) and the S1 fiber (a phase factor, exp(kB)). See
The unit quaternion qf representing the Hopf fiber enters Eqs.(23) and (24). Therefore, the S7 pure two-qubit states
Eq.(19) through a right multiplication to the base are represented by three unit 2-spheres (S2) and one unit
quaternions, and because the quaternion multiplication is circle S1.
non-commutative, this relative position is important. The
unit quaternion qf is parameterized as
Qubit-A quasi-Bloch Sphere S2 (A, A):
S2
S1 x1  sin AcosA , b  sin A sinA , x0  cos A .
S3
f f k f k f
q f  (cos  sin e j) e ;  f , f ,  f  (23) For general, entangled states, the qubit-A quasi-Bloch
2 2 coordinate is (A, A). For separable states, the qubit-A
According to the Hopf map defined in Eq.(18), a given Bloch coordinate is also (A, A). [c=bsin; x2=c cos;
point (f, f) on the base space S2 has a preimage which is x3=c sin; x4=b cos.]
a great circle on the 3-sphere S3, and this great circle is
parameterized by a unit complex number exp(kf). This Entanglement Sphere S2 (, ): tx=sincos,
great circle is also called the Hopf circle and is the 1- ty=sinsin, tz=cos; t=txi+tyj+tzk.
sphere S1 of Eq.(23). In order to correctly parameterize
For separable states (zero concurrence, c=bsin=0),
the great circles of S3, the Hopf fiber S1, exp(kf), must either the qubit-A Bloch coordinate is at the great circle in
appear at the right end of the expression in Eq(23), and the x0x1-plane (i.e., b=0), or the entanglement sphere
once again, this relative position is important because
coordinate is at its North pole (=0, t=k) or at South pole
quaternion multiplication is not commutative. The Hopf
k
(=, t=-k). The phase factor influences .
fiber e f not only provides the global phase of the
bipartite state but it affects both the relative phase of Qubit-B quasi-Bloch Sphere S2 (B, B):
qubit-B state and the azimuth angle  of t. This will be xB  sin B cosB , yB  sin B sinB , zB  cos B
shown explicitly below. For general states, the qubit-B quasi-Bloch coordinate is
k
In summary, the angle parameters are: Hopf base S4 (, , (B, B). The phase factor e B influences B. But (B,
, ) and Hopf fiber S3 (f, f, f) which is further mapped B) is used in plotting the Bloch sphere coordinate in this
to S2 base (f, f) and S1 fiber (f). paper.

k B
III. TWO-QUBIT BLOCH SPHERE MODEL The Phase S1: e
In the remainder of this paper we discuss parameterization k
S1( e B ) is the Hopf fiber of qB and it must appear at the
with respect to the qubit-A basis. The Hopf map, right-most end of the expression for quaternion qB, as
S3
h : S 7  S 4 and the parameterization Eq.(19), and the shown in Eq.(25) below. This phase factor may be
S1 considered the global phase only after properly adjusting
Hopf map h3: S 3  S 2 and the parameterization Eq.(23), (i.e., shifting by 2B) the azimuths  and B.
are combined as    
qB   cos B  sin B ekB j  e k B (25)
 θ   2 2 
cos A 
   j   2    
ΨA      cos B  sin B ekB j  ek B
    j θ
  sin A etA   2 2  k
  In summary, the three S2 spheres and a S1 unit circle e B
 2  provide a complete description of the S7 two-qubit pure
(24) state. This is shown in Fig.2.
 6

Figure 2: Three S2 spheres and one S1 phase factor ( ek ). The single qubit states |A> and |B> do not represent a
B

physical state but are only shown here as a placeholder for the coordinate. The entanglement sphere parameterizes the
pure unit quaternion t, where  affects the degree of concurrence and  is its phase angle (see Eq.(28)). The Bloch
coordinates and the state amplitudes are related by Eq.(24). Phase factor ek affects both global phase and relative
B

phase. For nonzero B, the two phase angles need to be translated -2B and BB-2B and then, the factor ek can B

be regarded as the global phase factor. The concurrence is given by c = sinAsinAsin = b sin.

1  1 x x1  x4 k 
There are four important points to be mentioned here: 1) A   0
 (29)
2  x1  x4 k 1  x0 
The azimuth  of the entanglement sphere is the phase
angle of concurrence. 2) For B≠0, the phase factor
influences the azimuth angles of both the qubit-B sphere Here, A is the reduced density matrix of qubit-A, and |c|
and the entanglement sphere, and therefore care must be is the concurrence [9]. Consider, for example, these
taken before it can be discarded as the global phase factor. maximally entangled states (MES) with a varying relative
k

phase: 00  e 11 and 01  e 10 for   0, 2  . All

k
3) The qubit-A hemispheric points (b) and the
entanglement sphere antipodal points (t) are identified 2 2
according to (b, t) = (-b, -t). This is a two-fold ambiguity. of their Bloch coordinates are the same as when =0
And 4) this two-qubit Bloch sphere model, Eq.(24), can (which are two of the Bell states, see section IV) except
not be used to find the Bloch coordinates for the separable for the angle :  = /2 +  for the former and =3/2 + 
states of the form ΨAB  1  ψB . Instead, this state for the latter. Hence,  parameterizes the relative phase of
A
may be represented as two single-qubit states. Also, if the the entangled states. We will discuss this further in
qubit-B quasi-Bloch coordinate is B=, then the angles section V.
B and B are interchangeable. These points are discussed
k
in more detail below. (2) The phase factor e B is the global phase only if it is
located at the left-most end of expression. This can be
(1) Azimuth  of entanglement sphere is the phase angle seen by rewriting Eq.(24) as below:
of concurrence.    j 
Writing the parameter t as ΨA   
   j 
t  cos k  sin ek i (26) (30)
 θA 
it follows that  cos    k   2 
  cos B  sin B e  B B  j 
2
 
1 k     0  j  e k B 
A  A  c e  2    (27)  sin θ A e  k B etA e k B   2 2 
2 j 0  
 2 
where, by comparing to Eq.(10), we find It is clear that the Hopf circle S1( ek ) is not simply the
B

 
global phase, but it is also part of the relative phase of
k   
 2     , c  concurrence,
 2 qubit-B. It also similarly affects the azimuth of the
ce (28) entanglement sphere. A unit quaternion is a spinor and a
c  sin A sin A sin  rotation operator [8, 10]. For example, e k B tek B is
another pure unit quaternion which is t rotated clockwise
and around the k-axis by angle 2B:
 7

e k B tek B 
(31) A. Parameter determination procedure
i sin cos   2 B   j sin sin   2 B   k cos 
Hence, for B≠0, the azimuth angles of the entanglement It is straightforward to obtain the amplitudes , and 
sphere () and qubit-B sphere (B) need be rotated from the seven angle parameters (we used imaginary unit
clockwise by 2B before ignoring the phase factor ek B as i in the formulas of Eq.(34)):
the global phase.
A B A B
  -2B; and B B-2B (32)   cos cos ei B ;   cos sin ei (B  B ) ;
Note that this same phase factor must be factored out of 2 2 2 2
the state amplitudes  for Eqs.(24) and (30) to  B 
remain consistent after the global phase factor is A 
 cos A  i sin Acos   cos   (34)
  sin  2 i B
e ;
discarded. 2   B i B  
i sin A sin  sin e
 2 
(3) The b hemispheric points of qubit-A sphere and the
  
t antipodal points of the entanglement sphere are
   cos A  i sin Acos   sin B  
identified according to: (b,t)=(-b,-t); or (b,-t)=(-b,t). 2 i   
  sin A  e B B
This introduces a two-fold ambiguity in the combined 2   B i B  
i sin A sin  cos e
coordinates of the qubit-A sphere and the entanglement  2 
sphere. This is because only the bt product appears in the
base parameters: In these expressions, it is clear that factoring out the
1  1  x0 x1  bt  and global phase requires the azimuth angles  and B to
A =  
2  x1  bt 1  x0  be shifted accordingly: -2B and BB-2B. Note
here that since no quaternion is involved in Eq.(34), the
1  1  x0  imaginary unit k was replaced by i. We present in this
ΨA    qB (33)
2 1  x0   x1  bt  section a streamlined procedure on how to obtain the
Bloch coordinates (the seven angles) from the complex
The Bloch sphere coordinates with the same bt product
amplitudes , and .
(while all other coordinates are being the same)
correspond to the same state. On a different note, Eq.(33)
can be used as a shortcut to Eq.(24) when determining the The four angle parameters (A, A, , ) and their
Cartesian coordinates (x0, x1, x2, x3, x4) of the Hopf base
Bloch sphere coordinates. The column vector in the quasi
state can be taken straight from the first column of the are determined from the quasi density matrix which is
quasi density matrix. One can determine the values of x0, given below in Eq.(39). Once the base parameters have
x1, b and t by comparing  A of Eq.(33) with Eq.(10), and been found, the three parameters (B, B, B) of the Hopf
fiber qB are found from the quasi state Ψ A , Eqs.(35) and
noting that t is a pure-imaginary unit quaternion. A
detailed step-by-step parameter determination procedure (25). From Eqs.(24) and (25),
of the seven angle parameters is given in the next section  θA
cos
(section IV).    j   2
ΨA    qB (35)
   j   θ A tA 
(4) This Bloch sphere parameterization of Eq.(24) is not  sin e 
valid for a separable state given by ΨAB  1  ψB .  2 
A
the quasi density matrix is
For this state, qubit-A is at the South pole, x0= -1 or A=.
In this case, the qubit-B can be plotted on its Bloch sphere  θA 
 cos 2   θ θ 
using the single qubit state ψ B . This separable state  A  ΨA ΨA    qB q B  cos A , sin A e tA 
θ
 sin A etA   2 2 
appears to be the only exception to our model, Eq.(24).  
Another minor issue occurs when the qubit-B is at the  2 
  I  nA  σ  t  
South pole: zB= -1 or B=. In this case, the terms, ek B
1 (36)
2
and e k B , are interchangeable in qB with the appropriate
Where,
change of sign. The final state, however, is consistent
after factoring out the global phase factor.

IV. PROCEDURE TO OBTAIN (A, A), (, ), (B, B),

B FROM , AND EXAMPLES
 8

qB q B  1, σ  t    x , y  t  , z  ,  θA 
step-4)  cos 2     j 
nA   sin Acos A , sin A sin A , cos A    x1 , b, x0  ΨA   qB    qB   B ,B , B
 sin θ A etA     j 
 
t  sin cos i  sin sin j  cos k ;  2 
θA , θB ,   0,   ;A ,B , B ,  0, 2π  (37) Solving for qB, we get

θA θ
The Pauli matrices are defined as qB  cos    j   sin A etA    j 
0 1  0 t   1 0  (38)
2 2
x    , y  t     , z   
 1 0   t 0   0 1 The qubit-B parameters are found by equating this qB to
The Hopf fiber qB (parameters: B, B, B) is completely
eliminated from  A . Hence,  A depends only on t and the    
qB   cos B  sin B ekB j  ek B  B ,B , B
qubit-A parameters. Here, t is the imaginary unit in the  2 2 
Pauli matrix y(t). The quasi-density matrix is written
in full as Through these four steps, the seven angle parameters are
1  1  x0 x1  bt  obtained from the four complex amplitudes of the state.
A    At this point, if B0, then 2B may be subtracted from 
2  x1  bt 1  x0 
(39) and B (where they become the new  and B) while the
1 1  x0  x1  x4 k    x3  x2 k  j  k
global phase factor e B is factored out of the complex
 
2   x1  x4 k    x3  x2 k  j 1  x0  amplitudes . Then the global phase can be reset


 
2 2
         j  (that is, set B=0) or ignored. All these four steps can be
 done quickly using the shortcut Eq.(33).
 
        j
   
2 2 

B. Examples of two-qubit states on Bloch sphere
where,
x0  cos A , x1  sin Acos A , b  sin A sin A , c  b sin , 1. Maximally entangled states, Bell states
x2  c cos , x3  c sin , x4  bcos  , bt  x2i  x3 j  x4k As a first example, we plot the four maximally entangled
states (MES), namely the Bell states defined as:
All angle parameters of the S4 Hopf base, (A, A) and (,
00  11 01  10
), and the equivalent Cartesian coordinates (x0, x1, x2, x3,  00  ,  01  ,
x4), are obtained. The procedure is outlined step by step 2 2 (40)
below: 00  11 01  10
10  , 11 
2 2
1. S Hopf base parameters:(A,A),(,)(x0, x1, x2, x3, x4)
4
In the qubit-A basis, converting qubit-B basis states to
quaternion units, we get the following quasi states:
x0  cos A          A
2 2 2 2
step-1) 1 1 1  j
 00    ,  01   ,
A 1  x0  1  x0 2  j 2 1 (41)
sin A  1  x0 2 , cos  , sin A 
2 2 2 2 1  1  1  j
10    , 11 
step-2) 
x1  kx4  2     A ,   2  j
 
2  1
x1  sin Acos A  A  two possiblevalues  The quasi density matrices are:

 b  sin A sin A 11  j 1 1 j

00    , 01   ,
x4  bcos    c  b sin 2 j 1  2 j 1 (42)
1 1 j 1 1  j
10    , 11   
step-3) x3  kx2  2      2  j 1 2 j 1 
x2  c cos , x3  c sin  
For this example we use the shortcut Eq.(33). From the
t  i sin cos  j sin sin  k cos  density matrix we find:
00 : x0=0, x1=0, b=1, t=j
2. S3 Hopf fiber parameters: (B, B) and B
Once the base parameters are determined, the parameters, 01 : x0=0, x1=0, b=1, t=-j
(B, B) and B, are obtained from the quasi state Ψ A . 10 : x0=0, x1=0, b=1, t=-j
 9

11 : x0=0, x1=0, b=1, t=j 10 : qB=1  B=0, B=0

From the quasi state in Eq.(33), equating it to Eq.(41), we 11 : qB=j  B=, B=0, B=0
find qB, which we then equate to Eq.(25), to find
The Bloch coordinates and Bloch spheres are summarized
 00 : qB=1  B=0, B=0
in the Table I and Table II, respectively.
 01 : qB=j  B=, B=0, B=0

Table I The Bloch coordinates of maximally entangled Bell states. Since (b, t) and (-b, -t) give the same state, the
coordinate for |cd> alternate to the table is cd = (x1, b, x0) & (tx, ty, tz): 00 = (0,-1,0) & (0,-1,0); 01 = (0,-1,0) & (0, 1,
0); 10 = (0,-1,0) & (0, 1, 0); 11 = (0,-1,0) & (0, -1, 0).

Bloch sphere | 00> | 01> | 10> | 10>

Coordinates
(x0, b, x1) (0, 1, 0) (0, 1, 0) (0, 1, 0) (0, 1, 0)
(tx, ty, tz) (0, 1, 0) (0, -1, 0) (0, -1, 0) (0, 1, 0)
(xB, yB, zB) (0, 0, 1) (0, 0, -1) (0, 0, 1) (0, 0, -1)
B 0 0 0 0

Table II Bell State Bloch Spheres. For all cases, B=0. The first three Bloch sphere rows are for the Bell states. In the
first two rows, filled (red) and hollow (blue) dots are the alternate coordinates (b,t) and (-b,-t), respectively. The last row
is the qubit-A Bloch sphere obtained by applying CNOT gate to the Bell states. The CNOT gate did not affect the
qubit-B Bloch coordinate.

| 00> | 01> | 10> | 11>

Qubit-A Bloch
Sphere

Entanglement
Sphere

Qubit-B Sphere
before and after
CNOT

CNOT CNOT| 00> CNOT| 01> CNOT| 10> CNOT| 11>

Qubit-A
after
CNOT
 10

Applying the CNOT gate to the Bell state, a product state  θA   θA 

results. The qubit-B state remained unchanged, while  cos 2   cos 2  (43)
qubit-A coordinate changed to x1= 1, b=0 (which gives   
zero concurrence, c = b sin = 0). The entanglement  sin θ A etA   sin θ A e kA 
   
sphere can be ignored for any state with b=0. The product  2   2 
state can be easily written down from the Bloch S4 S2

coordinates. For the entangled states, the states can be In this case, the quasi state
recovered from the Bloch coordinates using Eq.(24).  θ 
 cos A 
   k  2  (44)
  cos B  sin B e  B B  j 
Ψ A  e k B  2
 sin θ A e kA   2 2 
2. Separable states  
 2 
For separable states the concurrence is zero, which is corresponds to the composite state:
satisfied if b=sinAsinA=0 or if sin=0: the former  θ    
 cos A   cos B  (45)
condition (b=0) is satisfied if the qubit-A is on the great Ψ AB  ek  2 2
B
   ψA  ψB
circle on the x0x1-plane (b=0). In this case, the  θ A k    B k   2  
 sin e   sin e 
A B B

entanglement sphere is undefined because t appears in the  2   2 

quasi-state and quasi-density matrix only as a product k
The global phase factor e B may be ignored (i.e., set to
with b (see Eq.(33)). In the second case (sin=0), the unity) after the azimuth of qubit-B Bloch sphere has been
entanglement sphere coordinate is either at the North pole translated: BB-2B. It is clear that when t = k (or –k),
(=0) or at the South pole (=). In both cases, since the the entire coordinate of qubit-A Bloch sphere is valid for
concurrence is zero (c=0), the angle  is undefined (see the single qubit state. The separable state Bloch spheres
step-3). From the two-fold ambiguity in (b, t), choosing are shown in Fig.3. Some specific examples are shown in
t=k (i.e., =0, North pole), the qubit-A Bloch sphere the Table III.
coordinate agrees with the single qubit state of the
original product state.

For a separable state where t equals k, the Hopf base S4 is

reduced to S2:

Figure 3: The separable state Bloch spheres of Eq.(45). The entanglement sphere is necessary when the qubit-A Bloch
coordinate is at b≠0, and in which case, the entanglement sphere coordinate must be either the North pole (filled gray
dot) or the South pole (open gray dot).

Table III Separable state examples: ΨAB  ψA  ψB . In all cases, B = 0. In the coordinates of qubit-A sphere and
entanglement sphere, the two pairs of corresponding shape (or color) coordinates give the same bt product in Eq.(33).
The North pole (filled red dot) of entanglement sphere (i.e., t = k) gives qubit-A coordinate which agrees with the single-
qubit state |A>.
 11

Qubit-A Sphere Entanglement Sphere Qubit-B Sphere

C. Trajectory of CNOT, CZ and SWAP gates Then, the phase is fixed at /2, =/2, and the rotation
angle increases from =0 to  = around a given axis
We plot the Bloch sphere trajectory of two-qubit gates. (marked ‘2’ in the figure).
Let us consider a controlled-unitary, C(U), gate in two
steps. Here, U is a single-qubit unitary which can be 
written as '1  ek Rn  0   e k ,  0  ;
2 (47)


'2 '  e 2 Rn    kRn   ,  0  
k
k nσ
U  ek Rn    ek e 2 (46)

We consider the trajectories of CNOT, CZ and SWAP

where  is a phase angle, Rn() is the rotation operator
gates following these two steps.
that rotates a single-qubit state by an angle  around the
axis n, and  is the Pauli matrix,  = (x,y(k),z). The 1. CNOT and CZ gates
Pauli X and Z gates can be written as
00  11
 Let us take 10  as an example. The CNOT
X  e Rx   tobeused inCNOT and SWAP
k
2 2
gate and CZ gate trajectories are shown in Figures 4 and
 5, respectively. For the CZ gate in Figure 5, the entire
Z  e 2 Rz   tobeused inCZ
k
process amounts to a change in the nonlocal phase angle 
by  (from =3/2 to /2 counterclockwise).
We take a two step process: First, the phase angle 
increases from 0 to /2, = 0/2 (marked ‘1’ in the
figure) while the rotation angle is fixed at zero, =0.
 12

Figure 4: Evolution of CNOT gate on the Bell state |10>: CNOT   0  1  0 . The alternate (b, t) =(-b,-t)
10
2
trajectory is not shown. B=0. See Eq.(47).

Figure 5: Evolution of the CZ gate on the Bell state |10>: CZ| 10>=|00>. The alternate (b, t) =(-b,-t) is not shown.
B=0. See Eq.(47).

2. SWAP gate figure, followed by an X-rotation   0   which is

marked ‘2’ in the figure. As an example, in Fig.6 we
We consider the SWAP operation as show the SWAP gate operating on a separable state
01  k 11 . The resulting state is another separable
           Ψ 
AB
         2
         
SWAP     X      e 2 Rx        e k Rx     
k

        state 10  k 11 .
      

  0
        2,
  0 2
          

Similar to the CNOT and CZ gates above, we first let the

phase evolve   0   which is marked as ‘1’ in the
2

Figure 6: The action of a SWAP gate on 00  k 01  0  0  k 1 : 00  k 01 00  k 10 0 k 1

SWAP   0 .
2 22 2 2
The hollow red circle in the entanglement sphere is the initial state at =/2, =0. B=0.
 13

V. DISCUSSION quasi-Bloch sphere would have its B coordinate depend

The two quasi-Bloch spheres seem to be a natural choice on 1-2. The same argument applies to the a-c and b-d
for their name and coordinates because for separable pairs, leading us to expect the qubit-A Bloch coordinate
states, these spheres reduce to the respective single qubit A to depend on 1+2. Moreover, the a-d and the b-c
Bloch sphere (with the azimuth of qubit-B Bloch sphere pairs are entangled pairs with a relative phase factor
adjusted to B-2B). The phase angle B is not simply the exp(2k1) and exp(2k2), respectively. In section III, it
global phase, but it also affects the azimuth angles B and was mentioned for the two MES states that  = 21+/2
. After properly adjusting these two azimuth angles (that for a=1/2=d (b=0=c), and =22+ 3/2 for b=1/2=c
is, after subtracting 2B from each and replacing the (a=0=d). Another example may be a=b=d=1/3, c=0,
azimuth with these new values, Eq.(32)), the phase factor which is a partially entangled state (concurrence = 2/3).
exp(kB) can be considered a pure global phase factor and For this state we obtain =21+/2, B=1-2, and cosA
ignored. = (1/2)cos(1+2). For a separable state example, we
find B=1-2 for a=1/2=b (c=0=d); A=1+2 for
The entanglement sphere and its coordinates deserve b=1/2=d (a=0=c); and A=1+2 and B=1-2 for
more discussion. This sphere seems to capture the a=b=c=d=1/2. In all cases, any non-zero global phase
nonlocal parameters. The maximally entangled states
angle B had been factored out to the left end of
exist on the equator of this sphere (= /2) together with
expression as in Eq.(30) with the angles  and B properly
the possible two coordinates b=±1 (East and West) on the
shifted. These examples suggest that  is a ‘nonlocal’
equator of qubit-A Bloch sphere. If  = 0 or , then the
relative phase angle of the composite state while A and
state is separable and the Bloch spheres may be
considered the single-qubit Bloch sphere of each qubit. B are the ‘local’ relative phase angles. However, this
A product state can also happen if the qubit-A coordinate interpretation needs more investigation.
is on the great circle corresponding to b=0 (prime
meridian of the qubit-A sphere). Let us discuss the single-qubit Bloch sphere (or ball) of
qubit-A after a partial trace over qubit-B. Once qubit-B is
traced out, the quasi-state vector Ψ A is no longer a well
The azimuth  of the entanglement sphere deserves much
more discussion. This appears to parameterize a defined quantity. The partial trace over qubit-B maps the
‘nonlocal relative phase’ angle of the composite state. As two-quit density matrix AB=|AB><AB| to the reduced
was mentioned in section III (see Eqs.(27) and (28) and density matrix A. Similarly, let us define the partial trace
discussion there) this angle is the phase angle of over qubit-B as a map on the quasi-density matrix  A ,
concurrence. It was mentioned that this angle is equal to
the relative phase angle of MES states offset by some trB:  A A.
fixed constant. Furthermore, the azimuth angles A and 1  1  x0 x1  bt 
B seem to parameterize the ‘local’ relative phase angle in trB :  A   
2  x1  bt 1  x0  (49)
the sense that they capture the relative phase angle
between the ‘separable pairs’ of the basis components in 1  1  x0 x1  x4 k 
A   
the composite state. To be more specific, let us consider 2  x1  x4 k 1  x0 
the following state:
Let us note first that  A has the properties of a pure-state
k
ΨAB  e ae  k1
00  be  k2
01  ce k2
10  de k1
11  density matrix:  A 2 =  A , tr(  A )=1, and det(  A )=0.
(48) Two things are immediately clear in this definition:
(1) The partial trace over qubit-B is equivalent to
where a, b, c, d are a non-negative real number, projecting the unit quaternion t onto the k-axis in the
normalized; and , 1 and 2 are real numbers. This is a entanglement sphere: bt  btzk = bcosk = x4k (see
state with a definite phase factor for the concurrence Fig.7(b)). And
(where, = (ad-bc)ek2) while the component basis (2)  A indicates that the qubit-A quasi-Bloch sphere is the
states can vary with a relative phase from each other. Riemann sphere with x1-b-x0 axes. This unit sphere
Suppose we consider a pair of component states at a time. intersects at its equator a complex plane which has x1-real
The a-b pair is for a product state with |0> for qubit-A axis (x) and b-imaginary axis (y) with imaginary unit t,
with qubit-B having a relative phase factor exp[k(1-2)] and a point is given by z=x+ty on this complex plane.
between its two basis states. For the c-d pair, qubit-A is Then, in terms of the stereographic projection, the point
in |1> and qubit-B has a relative phase factor exp[-k(1- (A, A) on the Bloch sphere is projected to a point z on
2)] between its two basis states. Therefore, the qubit-B
 14

the complex plane given by z  x1  bt  cot  A et (see

A

1  x0 2 VI. SUMMARY
Fig.7(a)). One the other hand, after tracing out qubit-B, A A two-qubit pure-state Bloch sphere model was proposed
indicates that the Bloch ball coordinate axes for the and some examples of two-qubit states and gates were
mixed-state single qubit-A are x1-x4-x0 where the complex presented. The model consists of three unit 2-spheres
plane has x1-real axis and x4-imaginary axis with (Bloch spheres of each qubit and a nonlocal,
imaginary unit k. Then, the qubit-A Bloch coordinate is entanglement sphere) and a phase factor. This model can
that of a Bloch ball given by x02+x12+x42=1-c2 where c is consistently represent both separable and entangled two-
the concurrence of the two-qubit state before taking the qubit pure states except one notable exception,
partial trace. In this single-qubit mixed-state Bloch ball, a ΨAB  1  ψB , in which case, the states can be
A
constant concurrence corresponds to a sphere of radius represented directly, without following the procedure
(1-c2). Using the angular parameters in this paper, the presented here. The model is presented in terms of qubit-
Bloch coordinate would be (sinAcosA, sinAsinAcos, A basis. This three sphere model seems to be able to
cosA). For a fixed angle , this gives a flattened sphere, represent all two qubit pure states and two qubit gates
flattened along the imaginary-axis by factor |cos| relative geometrically, enabling visualization and intuition. The
to the unit 2-sphere. On the x0x1-plane, it has a unit main result is given by Eq.(24), which has a shortcut in
radius (see Fig.7(c)). These relationships are illustrated in Eq.(33). A detailed step-by-step procedure of parameter
Figure 7. Finally, Figure 8 summarizes the maximally determination was given as the step-1 through step-4 in
entangled states and the separable states on the qubit-A section-IV.
sphere and the entanglement sphere.

Figure 7 An illustration of Eq.(49). (a) The quasi-density matrix A and quasi-Bloch sphere of qubit-A. (b) The
complex planes. Partial trace over qubit-B projects the t imaginary-axis (b-axis) to the k imaginary-axis (x4-axis),
shrinking Bloch sphere(ball) in this direction by a factor |cos|. The k imaginary-axis (c-axis) is the concurrence axis
(where k=ek(-/2)j , see Eq.(27)). (c) After tracing out qubit-B, the reduced density matrix A is plotted for a constant 
(at =/4) and for a constant concurrence (at c=1/2).
 15

Figure 8 The maximally entangled states and the separable states are summarized on the qubit-A sphere and the
entanglement sphere. This summary is consistent with the examples given in the Tables I, II and III.

[1] M.A.Nielsen and I.L.Chuang “Quantum Computation Quantum information and computation; arXiv:quant-
and Quantum Information”, Cambridge University ph/0403136
Press, 2000 [6] R Mosseri and R Dandoloff “Geometry of entangled
[2] John Preskill, Physics/Computer Science 219 Lecture states, Bloch spheres and Hopf fibrations”, J. Phys.
Notes, California Institute of Technology, A: Math. Gen. 34, 10243–10252 (2001); arXiv:quant-
http://theory.caltech.edu/~preskill/ph219/#lecture ph/0108137
[3] A. Barenco, C. H. Bennett, R. Cleve, D. P. [7] R. Mosseri “Two-Qubit and Three-Qubit Geometry
DiVincenzo, N. Margolus, P. Shor, T. Sleator, J. A. and Hopf Fibrations” , Topology in Condensed
Smolin, and H. Weinfurter “Elementary gates for Matter, Springer Series in Solid-State Sciences
quantum computation ”, Phys. Rev. A 52, 3457 Vol.150, 187-203(2006); arXiv:quant-ph/0310053
(1995) [8] D.W.Lyons, “An Elementary Introduction to the Hopf
[4] T. Tilma, M.S. Byrd, E.C.G. Sudarshan, “A Fibration”, Math. Mag. 72(2), 87-98 (2003)
Parametrization of Bipartite Systems Based on SU(4) [9] W.K.Wootters, “Entanglement of Formation of an
Euler Angles” J. Phys. A: Math. Gen. 35, 10445- arbitrary state of two qubits”, Phys.Rev.Lett. 80(10),
10466 (2002); arXiv:math-ph/0202002v7. 2245 (1998); S.Hill and W.K.Wooters,
[5] T. F. Havel, C. J. L. Doran , “A Bloch-Sphere-Type “Entanglement of a Pair of Quantum Bits”,
Model for Two Qubits in the Geometric Algebra of a Phys.Rev.Lett. 78, 5022 (1997)
6-D Euclidean Vector Space”, PROC. SPIE THE [10] D. Hestenes “Vectors, Spinors, and Complex
INTERNATIONAL SOCIETY FOR OPTICAL Numbers in Classical and Quantum Physics” Am. J.
ENGINEERING Vol.5436, 93-106 (2004), 2nd Conf. Phys., Vol. 39/9, 1013-1027 (1971)