Michele Burrello - Fibonacci Anyons and Topological Quantum Computation

Fibonacci Anyons and Topological Quantum
Computation
Michele Burrello
SISSA
Trieste, April 2009
Work done with: Haitan Xu, Giuseppe Mussardo, Xin Wan
Contents
Universal Quantum Computation
Topological Quantum Computation, Anyons and Braid Group
Non Abelian Anyons
Fibonacci Anyons
Quantum Compiling and Icosahedral Hashing
Michele Burrello
Fibonacci Anyons and Topological Quantum Computation
Quantum Computation is considered to be one of the main

challenges for nowadays physics and it is based on the
possibility of storing information in Qubits.
Michele Burrello

Quantum Circuits are unitary operators that act on a Hilbert
space, generated by n qubits, whose states encode the
information we want to process.
Michele Burrello

Quantum Circuits are unitary operators that act on a Hilbert
space, generated by n qubits, whose states encode the
information we want to process.
In order to develop all the possible circuits we must be able to
implement every unitary operator in SU (N ) (with N = 2n ).
Definition
A Universal Quantum Computer is a machine able to realize, at any
accuracy, all the unitary operators in the space SU (N )
Michele Burrello

Every unitary operator in the space of n qubits can be decomposed in
single-qubit gates (elements of U (2)) and Controlled NOT gates:
Single Qubit rotation:
|i U |i ,
U = ei(+~n~/2)
Controlled NOT:
|0i
|0i + |1i
1
0
CN OT =
0
0
|0i
|0i + |1i
d
0
1
0
0
0
0
0
1
0
0
1
0
Michele Burrello
|1i
|0i + |1i
|1i
|0i + |1i
CNOT is the key element in creating

Entanglement
Single-Qubit Gates
U (2) gates can be mapped, up to an unimportant phase, in
SO(3) rotations.
In order to realize, at any given accuracy, all the single-qubit
gates, it is sufficient to have two rotations that generate a dense
infinite subgroup of SU (2)
Example:
1
H=
2

1
1

1
,
1

Fibonacci braidings =
1 = ei 10
ei 10
0
0
7
ei 10
51
2

,
Michele Burrello
T =

ei 8
0
ei 8
2 = ei 10
ei 10

i
ei 10
Topological Quantum Computation and Anyons
Local errors, thermic noise and decoherence are the main

problems in realizing a Quantum Computer
Michele Burrello

Topology
23
9B
23
9C
23
9C
23
9C
23
9D
Time
Michele Burrello
??

Topology
Topological properties are
insensitive to local perturbations!
23
9B
23
9C
23
9C
23
9C
23
9D
Time
Michele Burrello

Topology
23
9B
23
9C
23
9C
23
9C
23
9D
Time
Anyons in 2 + 1 dimensions
Michele Burrello

Topology
23
9B
23
9C
23
9C
23
9C
23
9D
Time
Anyons in 2 + 1 dimensions
Braid Group
Michele Burrello
Elementary Braidings (Generators of the Braid Group)

Counterclockwise: 1
Clockwise:
Michele Burrello

Counterclockwise: 1
Clockwise:
Michele Burrello

Counterclockwise: 1
Clockwise:
Michele Burrello

Counterclockwise: 1
Clockwise:
Michele Burrello
Braid Group
The World Lines in 2 + 1 D of N anyons
describe N -strand Braids.
These trajectories are robust with
respect to local perturbations (Topology
is preserved).
1-1
.....
N-1 N
Michele Burrello
Braid Group
is preserved).
Braids of N strands form an infinite

group generated by:

i , i1
with i = 1, . . . , N
1-1
.....
N-1 N
Michele Burrello
Braid Group
is preserved).

group generated by:

i , i1
with i = 1, . . . , N
1-1
.....
N-1 N
Weaves are braids where only one

anyon moves around the others
Michele Burrello
Braid Group
is preserved).

group generated by:

i , i1
with i = 1, . . . , N
1-1
.....
N-1 N
Weaves are braids where only one

anyon moves around the others
If 2 = 1 we have fermions or bosons (Permutation Group)

Fibonacci anyons: 10 = 1
Michele Burrello
Yang Baxter Braiding
Algebra relations
For non-adjacent operators:
[i , k ] = 0
if
|i k| 2
Yang Baxter Relations:

i i+1 i = i+1 i i+1
Michele Burrello
Algebra relations
[i , k ] = 0
if
|i k| 2

i i+1 i = i+1 i i+1
Michele Burrello
Algebra relations
[i , k ] = 0
if
|i k| 2

i i+1 i = i+1 i i+1
Michele Burrello
Braid Group Representations

Abelian case
The infinite Braid Group can have both one-dimensional and

higher-dimensional representations:
Abelian Anyons correspond to the one dimensional case (U (1)):
i = ei
Abelian Anyons on a torus imply a ground state degeneracy and

so the possibility to store quantum information (Kitaevs Toric
Code)
They do not allow information processing
If = we have fermions, if = 0 we have bosons
Excitations of a Laughlin state of filling factor have =
Michele Burrello

Non-Abelian case
Non-Abelian Anyons correspond to higher dimensional

representations
Michele Burrello

Non-Abelian case

representations
Their braidings, in general, do not commute
We can use in principle the elementray braidings to implement a
Universal Quantum Computation (Fibonacci anyons)
Excitations of Moore-Read and Read-Rezayi states have non
abelian statistics
Michele Burrello

Non-Abelian case

representations
abelian statistics
How can we use Non-Abelian Anyons to encode and process Qubits?
Michele Burrello

Non-Abelian case

representations
abelian statistics
How can we use Non-Abelian Anyons to encode and process Qubits?
What is the physical meaning of the dimension of the representation?
Michele Burrello
Quantum Dimension and Non-Abelian Anyons

Lets consider the simple case of spin
1 1
=0+1
2 2
1
2
(Qubit):
22=13
A particle with spin 1/2 is described by a two-dimensional Hilbert

space
When two of them fuse they give rise to a singlet or to a triplet.
This is a Fusion Rule
Michele Burrello

1 1
=0+1
2 2
1
2
(Qubit):
22=13

space
Non-Abelian anyons are characterized by non trivial fusion rules.
Ising Anyons [M3 ]:
Fibonacci Anyons [M2,5 ]:
=I+
=I+
Michele Burrello

1 1
=0+1
2 2
1
2
(Qubit):
22=13

space
Non-Abelian anyons are characterized by non trivial fusion rules.
Ising Anyons [M3 ]:
Fibonacci Anyons [M2,5 ]:
=I+
=I+
Non abelian anyons are characterized by Hilbert spaces of

dimension D greater than one
A set of N non abelian anyons can be found in DN orthogonal
states; so, they can encode N log2 D qubits
Michele Burrello
Non Abelian Anyons: main ingredients
To describe a non-abelian anyonic model we need a theory

characterized by the following elements:
c
Fusion Rules: Nab
Associativity Rules: Fdabc

xy
c
Braiding Rules: Rab
These rules must have a coherent structure and must obey several
constraints:
Pentagon equation for the F matrices
Hexagon equation for both F and R matrices (Yang Baxter)
Michele Burrello
Fusion Rules
A non abelian anyonic model is defined starting from a finite set of
particles (Superselection Sectors). These particles are linked by the
fusion rules:
X
M
X
c
c
c
c
ab =
Nab
c Va Vb =
Nab
Vab
da db =
Nab
dc
c
c
c
where Nab
= 0, 1 and Vab
= Vc are Hilbert spaces called fusion
spaces: the anyon dimensions are related to these spaces.
Michele Burrello
Fusion Rules
fusion rules:
X
M
X
c
c
c
c
ab =
Nab
c Va Vb =
Nab
Vab
da db =
Nab
dc
c
c
c
where Nab
= 0, 1 and Vab
a is a non-abelian anyon if
P
c
Michele Burrello
c
Naa
2
Fusion Rules
fusion rules:
X
M
X
c
c
c
c
ab =
Nab
c Va Vb =
Nab
Vab
da db =
Nab
dc
c
c
c
where Nab
= 0, 1 and Vab
a is a non-abelian anyon if
P
c
c
Naa
2
c
From the previous relations (Na = Nab
) it follows
Na d~ = da d~
The dimension of anyon a is the Perron Froebenius eigenvalue of
Na and d~ = (da , db , ..., dn ) is an eigenvector.
pP
2
The total quantum dimension of the model is D =
i di
Michele Burrello
Fusion Rules: Ising Anyons

Ising fusion rules (M3 )
= 1 + ,
= ,
Dimensions: d = 1 ,
=1
d = 2
d2 = 1 + 1
0 0 1
In the base (1, , ): N = 0 0 1
1 1 0
The fusion rules N can be rapresented by Brattelli diagrams:
...
=1+
...
=
Michele Burrello
...
1 =
Ising Anyons Chain
1,
1,
1
2
1,
2
3
4
5
...
8
7
The total quantum dimension of the chain depends on Nn and is:

n
D = dn = 2
Michele Burrello
Fusion Rule
CFT Interpretation
Anyonic superselection sectors can be considered as primary
fields in a minimal conformal model.
Fusion rules are linked to OPE:
X
c (z2 )
c
a (z1 )b (z2 ) =
Nab
+
(z1 z2 ) a b c
c
Minimal models can be therefore mapped in anyonic model.
Michele Burrello
Fusion Rule
CFT Interpretation
Anyonic superselection sectors can be considered as primary
fields in a minimal conformal model.
Fusion rules are linked to OPE:
X
c (z2 )
c
a (z1 )b (z2 ) =
Nab
+
(z1 z2 ) a b c
c
Minimal models can be therefore mapped in anyonic model.
Probabilistic Interpretation
Fusion rules and quantum dimensions are linked to the
amplitude of the scattering processes:
P (ab c) =
c
Nab
dc
da db
We can calculate the steady state distribution of an hypothetical

anyonic gas:
d2
pa = a2
D
Michele Burrello
Associativity Rules
F -Matrices
For an anyonic theory to be consistent the fusion rules N must

be associative:
X
X
y
x
d
d
Nab
Nxc
=
Nay
Nbc
x
These relations characterize the fusion process abc d in the

d
d
d
fusion space Vabc
= V(ab)c
= Va(bc)
.
d
The two descriptions of the space Vabc
correspond to different
orthogonal bases
There must be a unitary operator that relates these bases:
Fxy
x
d

abc
Fd
xy
b
y
is this transformation.
Michele Burrello
F Matrices
a
Fxy
b
y
Topologically equivalent to:
b
c
Fxy
x
a
d
a
Michele Burrello
F Matrices
a
Fxy
b
y
Michele Burrello
F Matrices
a
Fxy
b
y
b
c
Fxy
x
a
d
a
Michele Burrello
F Matrices
CFT connection
b
c
Fxy
x
a
d
a
The mapping between minimal unitary models and anyonic

theories allow to calculate F using the Dotsenko - Fadeev technique
for 4-point correlation functions:
h(0)(1)(z)()i = I1 (z) + I2 (z) = 0 I10 (1 z) + 0 I20 (1 z)
Michele Burrello
F Matrices
Probabilistic Interpretation
or1 ?
The probability of fusing b and c in or 1 is defined by the F matrix.

P ( 1) = |(F ) 1 |
P ( ) = |(F ) |
From an analogous scheme it follows that:
1
(Faaaa )11 =
da
Michele Burrello
Pentagon Equation
To be consistent there are two constraint that the F matrices must
satisfy. The first is known as Pentagon Equation.
c
b
Michele Burrello
Braidings Rab
k
A couple of anyons a b can be in a superposition of states Vab
defined by the fusion rules:
a (z1 )b (z2 ) =
Rab
c (z2 )
(z1 z2 )
a +b c
d (z2 )
(z1 z2 )
a +b d
+ ...
The clockwise exchange Rab of a and b is a unitary

operator that does not affect their total charge:
a
0
0
0
Rab
b
0
Rab
0
0
c
Rab = 0
0
R
0
ab
..
.
0
0
0
where:
c
(Rab
) = e2i(a +b c )
The representations of the braid generators i are given by

combinations of F and R as we will see later for Fibonacci anyons.
Michele Burrello
Hexagon Equation
The second contraint that F and R matrices have to satisfy is the
Hexagon Equation which guarantees also YB relations.
F
R
c
a
=
b
Michele Burrello
Fibonacci Anyons
Yang-Lee model
The model is characterized by two sectors:

the Vacuum 1 and the Fibonacci anyon .
Fusion Rules:
=1+
1 =
...
These fusion rules correspond to:

0 1
N =
=
d2 d 1 = 0
1 1
Michele Burrello
...
1+ 5
d =
2
Brattelli diagram
Fibonacci chain
1,
1,
1,
1,
1,
1,
...
Constraint: there cannot be two consecutive vacua 1.
13
13
Michele Burrello
21
Fibonacci F - Matrix
The unitary matrix F can be calculate from the pentagon equation:
F11 = F1 F 1
F11 + F2 = 1
c
b
1,
e
The resulting matrix is:

F =

with = d1
=
Michele Burrello
1,
1 5
2
Encoding Qubits
Our aim is to encode qubits with Fibonacci anyons
To do this we consider a system of 4 anyons whose total charge
is trivial: there are just two possible states (a) and we associate
them to the logic 0 and the logic 1
D
/

/
/

/
1&
Three Fibonacci anyons with total charge equal to 1 are enough

to encode a qubit (b)
With different total charges we obtain non computational states
(that must be avoided)
Michele Burrello
Fibonacci Braidings
To process a single qubit we must find the operators that
defines the braidings.
From the Yang Lee model (or the hexagon equation) one finds
out the R matrix:
4 i

e5
0
R=
2
0
e 5 i
In a Fibonacci chain, to find the representations of s, we need
to make a basis tranformation in order to apply the R - matrix:

a b
Michele Burrello

a b c
Fibonacci Braidings
a
a
e 5 i
0

2
e 5 i
2 = F 1 F =
ei 5
2
ei 5
2
ei 5
a
a
a
a
3 = 1 = R1 =
1 = 3 = R1 =
e 5 i
0

2
e 5 i
Michele Burrello
Single-Qubit Gate Compiling

Bonesteel et al.
To the purpose of Universal Quantum Computation we want to

approximate, at any give accuracy, any single-qubit gate using as
generators the braidings 1 and 2
For Fibonacci anyons the elementary braidings generate an
infinite group, dense in SU (2)
Ising anyons generate in the same way the finite octahedron
group, so they cannot implement a universal computation
= iX 0, 0031
32 22 34 22 34 22 34 22 32
Michele Burrello
Brute Force search

Bonesteel et al.
12
Total weaves:
BN
= 3N
22
2-2
Expected error:
N
=
1
3N/3
1-2
Michele Burrello
Brute Force search

Bonesteel et al.
N=1
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N=2
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N=3
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N=4
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N=5
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N=6
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N=7
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N=8
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N=9
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N = 10
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force search

Bonesteel et al.
N = 11
Total weaves:
BN
= 3N
Expected error:
N
=
Michele Burrello
1
3N/3
Brute Force Search with weaves

Using weaves we can approximate every single-qubit gate choosing
the best braid among the 3L/2 possibilities.
Example:

1 0
Target Gate: Z =
0 1
L=8
Z8
=
0, 31 + 0, 95i
0
Michele Burrello

0
0, 31 0, 95i
8 = 0, 31

Example:

1 0
Target Gate: Z =
0 1
L = 24
Z24
=
0, 0234 0, 9997i
0, 006 + 0, 002i
Michele Burrello
0, 006 + 0, 002i
0, 0234 + 0, 9997i
24 = 0, 024

Example:

1 0
Target Gate: Z =
0 1
L = 32
Z32
=
0.004 0.99997i
0.004 0.003i
Michele Burrello

0.004 0.003i
0.004 + 0.99997i
32 = 0, 007

Example:

1 0
Target Gate: Z =
0 1
L = 44
Z44
=
o 103
i
Michele Burrello
i
o 103

44 = 0, 001
Brute Force Search
Brute Force search allows to find the best weave of a given

length L to approximate every target gate.
The number of possible braids grows exponentially as 3L/2
The accuracy grows exponentially as 3L/6
The time required is exponential in the length!
Calculations become cumbersome for L > 40...
The solution given by the BF approach is the optimal one but it is very
slow to be obtained.
We can search for alternative algorithms (example: Kitaev - Solovay).
Michele Burrello
Icosahedral Quantum Hashing

M.B., Haitan Xu, Giuseppe Mussardo, Xin Wan (2009)
... yet among the better educated Classes it is known that no Circle is really
a Circle, but only a Polygon with a very large number of very small sides
E. A. Abbot, FlatLand
The brute force search is inefficient for long braids because it
samples the whole SU (2) space with almost equal weight
To get a faster algorithm we must enhance the sampling near the
target gate we want to approximate
In this way we can get a much faster algorithm that finds good
approximations for arbitrary SU (2) gates (but in general not the
optimal one)
We will start from the icosahedral group
Michele Burrello
Icosahedral Group
I = {g1 , ..., g60 }
Michele Burrello
Icosahedral Pseudo Group

I is a subgroup of SO(3) and can be mapped in a subgroup of
SU (2)
Every rotation in this subgroup can be compiled with a Brute
force algorithm in braids of length L = 8, 24, 44 : we obtain the
pseudogroup I (L):
I (24)
X = eix /2 I BFL=24 X
=
X
I (L) is a Pseudogroup characterized by errors: gi = gi eii
Michele Burrello

I = {
g1 , ..., g60 }
Michele Burrello

The pseudogroup is not a group and is characterized by errors
depending on the choosen length in the representation
Our algorithm exploits these error to create efficient samplings of
SU (2)
GROUP
PSEUDOGROUP
Closure:
I is not closed:
gi gj = gk
gi gj 6= gk
Identity is given by the right

choice of rotations:
We can span the vicinity of the

identity:
gi1 gi2 ...gin gin+1 = 1
gi1 gi2 ...

gin gin+1 = eiHn
with:
Hn =
n+1
X
i
Michele Burrello
Preprocessor: L = 8
Thanks to these errors with the products of 3 elements of I (8) we
span all SU (2) with 0 = 0, 03:
The preprocessor approximates the target T with

T8,3 = gp1 (8)
gp2 (8)
gp3 (8)
Michele Burrello
Main Processor
We can sample with high precision the vicinity of the identity:
For every n-plet of rotation in I we can find gn+1 such that:
gi1 gi2 ...gin gin+1 = 1
Mapping it into I(24)

we obtain a fine rotation:
= gi gi ...
R
gin gin+1 = eiHn
1
2
where Hn is a random matrix.
from 1 are described according to the Wigner
The distances of R
Dyson (GUE) distribution
If wehave an average error (L) for I(L)

the average distance of
R is n + 1(L)
to correct the result of
The main processor chooses the best R
the preprocessor
Michele Burrello
Distance distribution and random matrices
25
2.5
Brute Force L = 24
Icosahedral L = 8
Icosahedral L = 24
20
2
1.5
10
P(d)
15
0.5
0
0
0.2
0.4
0.6
0.8
1.2
1.4
d (distance/error)
Michele Burrello
Results
Maximal Length = 120
gp1 (8) gp2(8) gp3(8)
gq1 (24)
gq2 (24)
gq3 (24)
gq4 (24)
1500
1200
Average error:
900
P(d)
7, 1 104
600
Time required:
300
less than a second!
0
0
0.0005
0.001
0.0015
d (distance/error)
Michele Burrello
0.002
0.0025
Results
Adding a second main processor with L = 44 we have:

Ltot 296
2, 5 105
t 1s
A BF calculations requires L 60 to have the same precision,

but it would require a very long time...
We can iterate our algorithm to find new pseudogroups.
Iterating p times we have:
Ltot = 5 L0
p
Michele Burrello
60n/3
n+1
p
References
J. Preskill, Lecture notes on topological quantum computation
S. Trebst, A. W. Ludwig et al., A short introduction to Fibonacci

anyon models, arXiv:0902.3275 (2009)
N. E. Bonesteel et al., Topological quantum compiling, Phys. Rev B

75 (2007)
G. Mussardo, X. Wan et al., Topological quantum hashing with

icosahedral group, arXiv:0903.1497 (2009)
Michele Burrello

Michele Burrello - Fibonacci Anyons and Topological Quantum Computation

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Michele Burrello - Fibonacci Anyons and Topological Quantum Computation

Uploaded by

Copyright:

Available Formats

Fibonacci Anyons and Topological Quantum

Trieste, April 2009

Work done with: Haitan Xu, Giuseppe Mussardo, Xin Wan

Universal Quantum Computation

Topological Quantum Computation, Anyons and Braid Group

Non Abelian Anyons

Quantum Compiling and Icosahedral Hashing

Fibonacci Anyons and Topological Quantum Computation

Universal Quantum Computation

Quantum Computation is considered to be one of the main

Fibonacci Anyons and Topological Quantum Computation

Universal Quantum Computation

Quantum Computation is considered to be one of the main

Fibonacci Anyons and Topological Quantum Computation

Universal Quantum Computation

Quantum Computation is considered to be one of the main

Fibonacci Anyons and Topological Quantum Computation

Universal Quantum Computation

CNOT is the key element in creating

Fibonacci Anyons and Topological Quantum Computation

Fibonacci Anyons and Topological Quantum Computation

Topological Quantum Computation and Anyons

Local errors, thermic noise and decoherence are the main

Fibonacci Anyons and Topological Quantum Computation

Topological Quantum Computation and Anyons

Local errors, thermic noise and decoherence are the main

Fibonacci Anyons and Topological Quantum Computation

Topological Quantum Computation and Anyons

Local errors, thermic noise and decoherence are the main

Fibonacci Anyons and Topological Quantum Computation

Topological Quantum Computation and Anyons

Local errors, thermic noise and decoherence are the main

Fibonacci Anyons and Topological Quantum Computation

Topological Quantum Computation and Anyons

Local errors, thermic noise and decoherence are the main

Fibonacci Anyons and Topological Quantum Computation

Elementary Braidings (Generators of the Braid Group)

Fibonacci Anyons and Topological Quantum Computation

Elementary Braidings (Generators of the Braid Group)

Fibonacci Anyons and Topological Quantum Computation

Elementary Braidings (Generators of the Braid Group)

Fibonacci Anyons and Topological Quantum Computation

Elementary Braidings (Generators of the Braid Group)

Fibonacci Anyons and Topological Quantum Computation

Fibonacci Anyons and Topological Quantum Computation

Braids of N strands form an infinite

Fibonacci Anyons and Topological Quantum Computation

Braids of N strands form an infinite

Weaves are braids where only one

Fibonacci Anyons and Topological Quantum Computation

Braids of N strands form an infinite

Weaves are braids where only one

If 2 = 1 we have fermions or bosons (Permutation Group)

Fibonacci Anyons and Topological Quantum Computation

Yang Baxter Braiding

Yang Baxter Relations:

Fibonacci Anyons and Topological Quantum Computation

Yang Baxter Braiding

Yang Baxter Relations:

Fibonacci Anyons and Topological Quantum Computation

Yang Baxter Braiding

Yang Baxter Relations: