Top Banner
Chien Hsing James Wu David Gottesman Andrew Landahl
58

Chien Hsing James Wu David Gottesman Andrew Landahl.

Dec 19, 2015

Download

Documents

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Chien Hsing James Wu David Gottesman Andrew Landahl.

Chien Hsing James Wu

David Gottesman

Andrew Landahl

Page 2: Chien Hsing James Wu David Gottesman Andrew Landahl.

Outline

• Classical and quantum channels

• Overview of error correction

• Classical linear codes

• Quantum codes

• Conclusions

Page 3: Chien Hsing James Wu David Gottesman Andrew Landahl.
Page 4: Chien Hsing James Wu David Gottesman Andrew Landahl.
Page 5: Chien Hsing James Wu David Gottesman Andrew Landahl.
Page 6: Chien Hsing James Wu David Gottesman Andrew Landahl.

Two types of channels Two types of channels are discussed:are discussed:

1

2

XOR

Standard addition

Page 7: Chien Hsing James Wu David Gottesman Andrew Landahl.

QuantumQuantum Channel Models

Pauli rotations in each qubit

Page 8: Chien Hsing James Wu David Gottesman Andrew Landahl.

Computing Power versus Error Computing Power versus Error ControlControl

Page 9: Chien Hsing James Wu David Gottesman Andrew Landahl.

Basic Concepts in Error Control

Page 10: Chien Hsing James Wu David Gottesman Andrew Landahl.

Error ControlError Control Everywhere

Page 11: Chien Hsing James Wu David Gottesman Andrew Landahl.

History of Classical ErrorCorrection Codes (ECC)

Page 12: Chien Hsing James Wu David Gottesman Andrew Landahl.

Encoding is a mapping Please remember our

hypercube illustration of codes for interpretation

Page 13: Chien Hsing James Wu David Gottesman Andrew Landahl.

Draw yourself hypercube pictures for

these, illustrate our (3,1,1) code from previous lecture

Page 14: Chien Hsing James Wu David Gottesman Andrew Landahl.

(3,1,1)

t=1, correct one error

d=2t+1, t=1, 2t+1=3=d

w=3

n k d

0

1

Page 15: Chien Hsing James Wu David Gottesman Andrew Landahl.

transpose

identity

Page 16: Chien Hsing James Wu David Gottesman Andrew Landahl.

Role of Parity Check Matrix PRole of Parity Check Matrix PExplanation that P returns only error syndrome since it anihilates codewords v

Page 17: Chien Hsing James Wu David Gottesman Andrew Landahl.

Classical Linear Error Control CodesClassical Linear Error Control Codes

Page 18: Chien Hsing James Wu David Gottesman Andrew Landahl.
Page 19: Chien Hsing James Wu David Gottesman Andrew Landahl.

General idea of block linear General idea of block linear codescodes

Matrix vector multiplication

Page 20: Chien Hsing James Wu David Gottesman Andrew Landahl.
Page 21: Chien Hsing James Wu David Gottesman Andrew Landahl.

Galois Field hypercube

Smaller space

generator

We denote it by

Page 22: Chien Hsing James Wu David Gottesman Andrew Landahl.

Big space

distanceSmaller space

n= length of vector

Page 23: Chien Hsing James Wu David Gottesman Andrew Landahl.

Error Error DetectionDetection and and Correction Correction CapabilityCapability

As in general case

3 in our case

1 in our case

Page 24: Chien Hsing James Wu David Gottesman Andrew Landahl.

Detection Capability of Linear Block Codes

If codeword is changed to another codeword it cannot be detected

Page 25: Chien Hsing James Wu David Gottesman Andrew Landahl.

Detection & Correction of (n,k) Linear Block Codes

23-21=6

2 3-1 = 4-1=3

0

1

Page 26: Chien Hsing James Wu David Gottesman Andrew Landahl.

Linear (Linear (nn,,kk)) Cyclic Cyclic Codes over GF(2)GF(2)

Easy hardware to operate on these polynomials

Page 27: Chien Hsing James Wu David Gottesman Andrew Landahl.

Encoding a Encoding a CyclicCyclic Code Code

From slide with general diagram of linear codes

Page 28: Chien Hsing James Wu David Gottesman Andrew Landahl.

Cyclic ShiftsCyclic Shifts in Cyclic Codes

Page 29: Chien Hsing James Wu David Gottesman Andrew Landahl.

Cyclic propertyCyclic property

Thus we can talk about a group

Page 30: Chien Hsing James Wu David Gottesman Andrew Landahl.

Cyclic Group Gc in Code Subspace

Red arrows represent shifts

Page 31: Chien Hsing James Wu David Gottesman Andrew Landahl.

Quantum Quantum Error Error

CorrectionCorrection

Page 32: Chien Hsing James Wu David Gottesman Andrew Landahl.

Outline

• Sources and types of errors

• Differences between classical and quantum error correction

• Quantum error correcting codes

Page 33: Chien Hsing James Wu David Gottesman Andrew Landahl.

Introduction: why quantum error correction?

• Quantum states of superposition (which stores quantum information) extremely fragile.

• Quantum error correction more tricky than classical error correction.

• In the field of quantum computation, what is possible in theory is very far off from what can be implemented.

• Complex quantum computation impossible without the ability to recover from errors

Page 34: Chien Hsing James Wu David Gottesman Andrew Landahl.

What can go wrong?

• Internal:– Initial states on input qubits not prepared properly.– Quantum gates used may not be accurate

• Quantum gates may introduce small errors which will accumulate.

• External:– Dissipation

• A qubit loses energy to the environment.

– Decoherence

Page 35: Chien Hsing James Wu David Gottesman Andrew Landahl.

DecoherenceDecoherence• Decoherence is the loss of quantum

information of a quantum system due to its interaction with the environment.

• Almost impossible to isolate a quantum system from the environment.

• Over time, our quantum system will be entangled with the environment.

Page 36: Chien Hsing James Wu David Gottesman Andrew Landahl.

Detrimental role of environmentDetrimental role of environment

• Information encoded in our quantum system will be encoded instead in the correlations between the quantum system and the environment.

• The environment can be seen as measuring the quantum system, collapsing its superposition state.

• Hence quantum information (encoded in the superposition) is irreversibly lost from the qubit.

Page 37: Chien Hsing James Wu David Gottesman Andrew Landahl.

How to Deal With Decoherence?

1. Design quantum algorithms to finish before decoherence ruins the quantum information.

– Can be difficult as• Decoherence occurs very quickly.• Quantum algorithms may be very complex and

long.

First method to deal with decoherenceFirst method to deal with decoherence

Page 38: Chien Hsing James Wu David Gottesman Andrew Landahl.

Dealing With Decoherence

2. Try to lower the rate at which decoherence occurs.

– Accomplished by using a right combination of:• Quantum particle type• Quantum computer size• Environment

Second method to deal with decoherenceSecond method to deal with decoherence

Page 39: Chien Hsing James Wu David Gottesman Andrew Landahl.

Decoherence times in practiceDecoherence times in practice• Decoherence time refers to the time available

before decoherence ruins quantum information.• Decoherence time is affected by the size of the

system, as well as the environment.

– Decoherence time affected by environmental factors like temperature and amount of surrounding particles in the environment

A ppro xim a te de c o he re nc e tim e (in s e c o nds ) fo r va rio us s ys te m s ize s a nd e nviro nm e nt

System size(cm)

Cosmic Radiation

RoomTemperature

SunlightVacuum

(106 particles/cm3)Air

10 -3 10 -7 10 -14 10 -16 10 -18 10 -35

10 -5 10 15 10 -3 10 -8 10 -10 10 -23

10 -6 10 25 10 5 10 -2 10 -6 10 -19

Page 40: Chien Hsing James Wu David Gottesman Andrew Landahl.

Gate completion timeGate completion time

• Time needed for a quantum gate operation is as important as decoherence time.

• Different types of quantum systems have different decoherence time and per gate operation time.

operation gate quantumper time

timeedecoherenc

edecoherenc before performed becan that operations of noMax

In next time we will compare these coefficients for real technologies

Page 41: Chien Hsing James Wu David Gottesman Andrew Landahl.

Maximum number of operations before decoherence Maximum number of operations before decoherence for various quantum systemsfor various quantum systems

• The better the decoherence time, the slower the quantum gate operations.

Q uantum systemDecoherence

time(sec)

T ime perG ate O peration

(sec)

M ax number ofoperations

beforedecoherence

Electrons from gold atom 10 -8 10 -14 10 6

T rapped indium atoms 10 -1 10 -14 10 13

O ptical microcav ity 10 -5 10 -14 10 9

Electron spin 10 -3 10 -7 10 4

Electron quantum dot 10 -3 10 -6 10 3

Nuclear spin 10 4 10 -3 10 7

D e c o he re nc e tim e ve rs us tim e re quire d fo r a qua ntum g a te o pe ra tio nfo r va rio us qua ntum s ys te m s

Page 42: Chien Hsing James Wu David Gottesman Andrew Landahl.

Dealing With Decoherence and other sources of errors

3. Use Quantum Error correcting codes• Encode qubits (together with extra ancillary

qubits) in a state where subsequent errors can be corrected.

• Allows long algorithms requiring many operations to run, as errors can be corrected after they occur.

Third method to deal with decoherenceThird method to deal with decoherence

Page 43: Chien Hsing James Wu David Gottesman Andrew Landahl.

History of Quantum ErrorQuantum ErrorCorrectionCorrection Codes (QECC)

Page 44: Chien Hsing James Wu David Gottesman Andrew Landahl.

Quantum Error Correcting Codes

Page 45: Chien Hsing James Wu David Gottesman Andrew Landahl.

Quantum Errors

Page 46: Chien Hsing James Wu David Gottesman Andrew Landahl.

General representation of single qubit

Page 47: Chien Hsing James Wu David Gottesman Andrew Landahl.
Page 48: Chien Hsing James Wu David Gottesman Andrew Landahl.

Cloning (copying) operator U does not exist

Assume that such U exists

So we apply it to general superposed state

And we obtain this Which is not what we wanted

But this is still useful. Although not copying , this is a redundancy introducing operator so it may be used for error correcting codes. This was one of main ideas

Page 49: Chien Hsing James Wu David Gottesman Andrew Landahl.

Commuting and Anti-Commuting Commuting and Anti-Commuting Quantum OperatorsQuantum Operators

Commutator of A and B

Anti-commutator of A and B

Page 50: Chien Hsing James Wu David Gottesman Andrew Landahl.

(1-qubit) Pauli Operators

We express Y in terms of X and Z

Page 51: Chien Hsing James Wu David Gottesman Andrew Landahl.

Properties of Pauli Operators

Adjoint operator

commutative

Pauli operators are self-Pauli operators are self-inverses and anti-inverses and anti-commutecommute

anticommutative

Page 52: Chien Hsing James Wu David Gottesman Andrew Landahl.

1-qubit Pauli Group G1

4 * 2 = 8 elements in this group

Pauli operators are a groupPauli operators are a groupPlease remember, this is important

Page 53: Chien Hsing James Wu David Gottesman Andrew Landahl.

Now we extend to group Gn

We model faults in channels by Gn

Page 54: Chien Hsing James Wu David Gottesman Andrew Landahl.

Example: error operator in GG55

Tensor product

This will be our error model from now

Page 55: Chien Hsing James Wu David Gottesman Andrew Landahl.

Quantum network for Quantum network for correcting errorscorrecting errors

00

1s

2s

3eb2eb1eb

• Assume thatbbb

111000

e1e1e1eee 123123

1123 eee }1,0{ie

Input signal with error

Input signal after error correcting

Decoder and corrector

Page 56: Chien Hsing James Wu David Gottesman Andrew Landahl.

Equivalently Equivalently

00

1s

2s

3eb2eb1eb

bbb

1s

2s

Page 57: Chien Hsing James Wu David Gottesman Andrew Landahl.

Perform operations on logical Perform operations on logical bits bits

b H

• e.g. Hadamard gate

bb bbb

2

)1(

bbb2

1

b

Page 58: Chien Hsing James Wu David Gottesman Andrew Landahl.

Quantum Error Correcting Quantum Error Correcting by Peter Shorby Peter Shor

• In 1995, Peter Shor developed an improved procedure using 9 qubits to encode a single qubit of information

• His algorithm was a majority vote type of system that allowed all single qubit errors to be detected and corrected

This was a starting point to great research area, although his paper had many bugs