Spin-Based Quantum Computing with Silicon MOS Quantum Dots by Michael Allan Fogarty
A thesis in fulfilment of the requirements for the degree of Doctor of Philosophy
School of Electrical Engineering and Telecommunications Faculty of Engineering
November 2018 PLEASE TYPE THE UNIVERSITY OF NEW SOUTH WALES Thesis/Dissertation Sheet
Surname or Family name: Fogarty
First name: Michael Other name/s: Allan
Abbreviation for degree as given in the University calendar: PhD
School: Electrical Engineering and Telecommunications Faculty: Engineering
Title: Spin-Based Quantum Computing with Silicon-MOS Quantum Dots
Abstract 350 words maximum: (PLEASE TYPE)
This thesis describes advancements in the silicon metal-oxide-semiconductor (SiMOS) quantum dot qubit platform. Recent experimental realisations of coherent single qubit and two qubit operations calls for the demonstration of a fully integrated SiMOS platform, showing a scalable implementation of all necessary elements for quantum computation operating together. Integration of qubit initialisation, control and readout via methods considered robust against scaling qubit array dimensions is the first fundamental step towards realising fully error corrected qubit registers useful for quantum computation.
This thesis commences by analysing the result of standard experiments for benchmarking performance and integrity of a singular SiMOS qubit. This experiment, known as Randomised Benchmarking, yields uncharacteristic results for the SiMOS platform when compared to experiment theory and other results in literature. This is addressed through extending the experiment to take low-frequency environmental fluctuations into account. Resulting analysis indicates that fidelities for the majority of qubit operation is considered above the threshold for fault-tolerant quantum computation.
Having demonstrated high quality qubit control using SiMOS quantum dots, the capability of preparing and measuring qubits via methods robust against scaling are detailed. Recent studies have detailed how the singlet-triplet basis can be utilised for the initialisation and readout of qubits within a scaled quantum register. These techniques are discussed and experimentally demonstrated in a SiMOS quantum dot device alongside single qubit addressability through electron spin resonance. Electrostatic control over the Heisenburg exchange coupling between two adjacent dots produces a two qubit SWAP operation realised in this device. These results together demonstrate, for the first time within a single silicon qubit device, the initialisation and readout of qubit pairs by scalable methods integrated with single qubit addressability and two-qubit logical operations.
Furthermore, having experimentally realised integration of fundamental building blocks of a scaled quantum register, what follows is a discussion of how this platform can be scaled into demonstrations of a fully error corrected logical qubit. Experimental state-of-the-art in SiMOS technology is discussed in the context of this logical qubit protocol, which employs quantum dots as both data qubits and singlet-triplet ancillas.
Declaration relating to disposition of project thesis/dissertation
I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.
rtation Abstracts International (this is applicable to doctoral
/ / rt .... .lf, .?. ��.Date ......
The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and re uire the a roval of the Dean of Graduate Research.
FOR OFFICE USE ONLY Date of completion of requirements for Awa"rd: i Copyright Statement
I hereby grant the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all proprietary rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation. I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstract International (this is applicable to doctoral theses only). I have either used no substantial portions of copyright material in my thesis or I have obtained permission to use copyright material; where permission has not been granted I have applied/will apply for a partial restriction of the digital copy of my thesis or dissertation.
Date ,1_�/:.?/�r...... ····•· .
ii Authenticity Statement
I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis. No emendation of content has occurred and if there are any minor variations in formatting, they are the result of the conversion to digital format.
Date ,1_�/:.?/�r...... ····•· .
iii Abstract
This thesis describes advancements in the silicon metal-oxide-semiconductor (SiMOS) quantum dot qubit platform. Recent experimental realisations of coherent single qubit and two qubit operations calls for the demonstration of a fully integrated SiMOS platform, showing a scalable implementation of all necessary elements for quantum computation operating together. Integration of qubit initialisation, control and readout via methods considered robust against scaling qubit array dimensions is the first fundamental step towards realising fully error corrected qubit registers useful for quantum computation. This thesis commences by analysing the result of standard experiments for benchmarking performance and integrity of a single SiMOS qubit. This experiment, known as randomised benchmarking, yields uncharacteristic re- sults for the SiMOS platform when compared to experiment theory and other results in literature. This is addressed through extending the experiment to take low-frequency environmental fluctuations into account. Resulting anal- ysis indicates that fidelities for the majority of qubit operation is considered above the threshold for fault-tolerant quantum computation. Having demonstrated high quality qubit control using SiMOS quantum dots, the capability of preparing and measuring qubits via methods robust against scaling are detailed. Recent studies have detailed how the singlet- triplet basis can be utilised for the initialisation and readout of qubits within a scaled quantum register. These techniques are discussed and experimen- tally demonstrated in a SiMOS quantum dot device alongside single qubit addressability through electron spin resonance. Electrostatic control over the Heisenburg exchange coupling between two adjacent dots produces a two-qubit SWAP operation realised in this device. These results together demonstrate, for the first time within a single silicon qubit device, the ini- tialisation and readout of qubit pairs by scalable methods integrated with single qubit addressability and two-qubit logical operations.
iv ABSTRACT
Finally, having experimentally realised integration of fundamental build- ing blocks of a scaled quantum register, what follows is a discussion of how this platform can be scaled into demonstrations of a fully error corrected log- ical qubit. Experimental state-of-the-art in SiMOS technology is discussed in the context of this logical qubit protocol, which employs constituent quantum dots as both data qubits (the carriers and propagators of quantum informa- tion) or singlet-triplet ancilla qubits for error detection.
v Acknowledgement
The work presented within this dissertation was primarily supervised by Sci- entia Professor Andrew Dzurak, with co-supervisors Dr. Menno Veldhorst and Professor Andrea Morello. Without the support I received from these three great researchers, this manuscript would not exist.
First and foremost, I would like to thank my primary supervisor Andrew Dzurak. The continual support and mentorship he has provided throughout my studies has been invaluable. The passion and enthusiasm Andrew shows for his research is truly inspiring, an attitude which continually fuels a pro- ductive and interesting workplace. He has facilitated many opportunities for me to present my research, form new and fruitful collaborations and develop as a young researcher. It has been a privilege and an honour to be a part of his research team. My co-supervisor, Professor Andrea Morello, was first person to introduce me to the concepts of quantum computing. Taking his course on quantum devices was a pivotal moment, and the advice he has pro- vided to me throughout my early career has been invaluable. The first year of my studies was co-supervised by Dr. Menno Veldhorst, who coached me through my first hands-on experience with an actual qubit device. Many of the skills I learned under his guidance during this time facilitated the success of my experiments in later years of my studies. I would like to acknowledge the past and present members of the Dzurak lab for their support, in particular the post-doctoral research team including Dr. Henry Yang, Dr. Kok Wai Chan, Dr. Tuomo Tanttu, Dr. Bas Hensen and Dr. Jason Hwang, fellow students Wister Huang, Ruichen Zhao, Yuxin Sun, Anderson West, Ross Leon. Thank you for sharing your knowledge and expertise, and for making the years a thoroughly enjoyable experience. A special mention to Bas, who went above and beyond for me throughout the past year, giving me advice on both journal articles and this dissertation.
vi ACKNOWLEDGEMENT
Also to Anderson, who was always willing to bounce ideas around or lend a helping hand. I would like to thank the CQC2T and ANFF staff and students for providing a welcoming and collaborative research environment, with a special mention to Karen Jury for her friendship and support. I would also like to acknowledge my collaborators Professor Stephen Bartlett, Associate Professor Steven Flammia, Dr. Robin Harper, Dr. Thaddeus Ladd, Dr. Cody Jones, Dr. Mark Gyure, Dr. Arne Laucht and Dr. Dimi Culcer. Their support and advice has been invaluable. Beyond the research environment, I would like to extend a huge debt of gratitude to those who kept me sane throughout the past few years. I will always cherish the memories of pub trivia and games night with my friends. Thank you to my parents, Allan and Judy, who have been ever supportive of my decisions. Thank you to my siblings, Rebecca, Jack and Emma, whom always believed in me. Without the ongoing support of my family, I would not have been able to achieve this goal. Last, but certainly not least, I would like to thank my partner, Grace. Not only was she happy to help by proof-reading, but her patience, motivation, love and support during this time saw me through to the end. Without her I surely would not have had the strength.
vii Research Contributions Peer reviewed journal articles
M. A. Fogarty, M. Veldhorst, R. Harper, C. H. Yang, S. D. Bartlett, S. T. Flammia, and A. S. Dzurak, “Non-exponential Fidelity Decay in Randomized Benchmarking with Low- Frequency Noise”, Physical Review A, 92 (2), 2015. The contribution statement for this work can be found in Appendix A. J. C. C. Hwang, C. H. Yang, M. Veldhorst, N. Hendrickx, M. A. Fogarty,W.Huang,F.E.Hudson,A.MorelloandA.S.Dzurak, “Impact of g-factors and valleys on spin qubits in a silicon double quantum dot”, Physical Review B, 96 (4), 2017. The contributions of M.A.F to this work involved preliminary characteri- sation of device behaviour, modelling via TCAD and input on the results interpretation and manuscript. C. Jones, M. A. Fogarty,A.Morello,M.F.Gyure,A.S.Dzurak,and T. D. Ladd, “A logical qubit in a linear array of semiconductor quantum dots”. Physical Review X, 8 (2), 021058. The contribution statement for this work can be found in Appendix C. M. A. Fogarty,K.W.Chan,B.Hensen,W.Huang,T.Tanttu,C.H.Yang, A.Laucht,M.Veldhorst,F.E.Hudson,K.M.Itoh,D.Culcer,T.D.Ladd, A. Morello, A. S. Dzurak, “Integrated silicon qubit platform with single-spin addressability, exchange control and robust single-shot singlet-triplet readout”. Nature communica- tions, 9 (1), 4370. The contribution statement for this work can be found in Appendix B.
viii RESEARCH CONTRIBUTIONS
Articles in submission
T. Tanttu, B. Hensen, K. W. Chan, C. H. Yang, W. Huang, M. A. Foga- rty, F. Hudson, K. Itoh, D. Culcer, A. Laucht, A. Morello, and A S. Dzurak “Controlling spin-orbit interactions in silicon quantum dots using magnetic field direction”. arXiv:1807.10415 The contributions of M.A.F to this work involved preliminary characterisa- tion of device behaviour. W. Huang, C. H. Yang, K. W. Chan, T. Tanttu, B. Hensen, R. C. C. Leon, M. A. Fogarty, J. C. C. Hwang, F. E. Hudson, K. M. Itoh, A. Morello, A. Laucht, and A. S. Dzurak “Fidelity benchmarks for two-qubit gates in silicon”. arXiv:1805.05027 The contributions of M.A.F to this work involved input on discussing device behaviour, results interpretation and manuscript preparation.
ix RESEARCH CONTRIBUTIONS
Presentations: Oral
Presenting Author: M. A. Fogarty
“High-fidelity single-shot readout and exchange control of spin qubits in silicon-MOS quantum dots” International Conference on Nanoscience and Nanotechnology, Jan 29-Feb 02, 2018, Wollongong, NSW, Aus- tralia.
“Silicon Quantum Computing” Postgraduate research symposium, Sep 26-29, 2017, UNSW Sydney, Australia.
“Integrated silicon qubit platform with single-spin addressability, ex- change control and robust single-shot singlet-triplet readout” Interna- tional workshop on silicon quantum electronics, Aug 18-21, 2017, Hills- boro, Oregon, USA.
“Nonexponential fidelity decay in randomized benchmarking with low- frequency noise”. 13th Conference on Optoelectronics and Microelec- tronic Materials and Devices, Dec 12-14, Sydney, NSW, Australia.
“Nonexponential fidelity decay in randomized benchmarking with low- frequency noise”. International Conference on Nanoscience and Nan- otechnology, Feb 07-11, Canberra, ACT, Australia.
“Nonexponential fidelity decay in randomized benchmarking with low- frequency noise”. CQC2T Internal Seminar, Nov 12, 2015, UNSW Sydney, Australia.
“Nonexponential fidelity decay in randomized benchmarking with low- frequency noise”. International workshop on silicon quantum electron- ics, Aug 03-04, Takamatsu, Japan.
Presentations: Poster
Presenting Author: M. A. Fogarty
“Integrated silicon qubit platform with single-spin addressability, ex- change control and robust single-shot singlet-triplet readout” Spin Qubit 3, Nov 06-10, 2017 Sydney, NSW, Australia.
x RESEARCH CONTRIBUTIONS
“Integrated silicon qubit platform with single-spin addressability, ex- change control and robust single-shot singlet-triplet readout” Labora- tory for Physical Science: quantum computing progress review, Aug 14-16, 2017 La Jolla, California, USA.
“Single qubit randomized benchmarking and tomography in Silicon- MOS quantum dots” Laboratory for Physical Science: quantum com- puting progress review, Jul 17-20, 2016 Alexandria, Virginia, USA.
“Randomized benchmarking in the presence of semiconductor qubit noise sources” CQC2T Annual Workshop, 2015.
“Simulation and characterization of SiMOS qubits for quantum compu- tation” UNSW honors thesis poster competition, 2014 (1st prize win- ner).
“Randomized benchmarking in the presence of semiconductor qubit noise sources” Laboratory for Physical Science: quantum computing progress review, 2014. Washington DC, USA.
“Randomized benchmarking in the presence of semiconductor qubit noise sources” International workshop on silicon quantum electronics, 2014. Albuquerque, New Mexico, USA.
Awards
Australian Postgraduate Award (2015-2018). Australian Postgraduate Awards provide a living allowance for high- quality research higher degree students. They are awarded on academic merit and research experience and/or potential.
ANFF-UNSW publication award (2016). Awarded for “Nonexponential fidelity decay in randomized benchmark- ing with low-frequency noise”, the ANFF-NSW publication award is presented in recognition of high-quality research outcomes that have been enabled by ANFF-NSW facilities.
xi Contents
Abstract iv
Acknowledgement vi
Research Contributions viii
Table of Contents xii
List of Figures xviii
List of Tables xxi
Introduction 1
1 Background theory and qubit fundamentals 12 1.1Quantumtwolevelsystemsandthequantum-bit...... 14 1.1.1 Twolevelsystems...... 14 1.1.2 Paulimatrices...... 15 1.1.3 Density matrix and the Bloch representation ...... 16 1.1.4 Anti-crossings...... 18 1.2 Postulates of quantum mechanics and measuringaquantumsystem...... 21 1.2.1 FirstPostulate...... 21 1.2.2 SecondPostulate...... 22 1.2.3 ThirdPostulate...... 22 1.2.4 FourthPostulate...... 23 1.2.5 FifthPostulate...... 24 1.2.6 SixthPostulate...... 25 1.3Anelectronwithinanexternalmagneticfield...... 26
xii CONTENTS
1.3.1 Larmor precession ...... 26 1.3.2 Therotatingframe...... 29 1.4 Operations on a single qubit ...... 30 1.4.1 TheRabiformula...... 30 1.4.2 Anti-crossingdynamics...... 35 1.5Noiseandquantumcoherence...... 41 1.5.1 Qubit noise ...... 41 1.5.2 Qubit characteristic times ...... 44 1.5.3 Noisy qubit channels ...... 51 1.6 Experiments for benchmarking qubit performance...... 56 1.6.1 Tomographic experiments ...... 57 1.6.2 RandomisedBenchmarking...... 59 1.7 Systems of two qubits ...... 64 1.7.1 Expansion from the single-spin basis to a two-spin basis 64 1.7.2 Two qubit Hamiltonian in the spin basis ...... 66 1.7.3 Systems of two coupled electrons: ...... 67 1.7.4 Two qubits in the singlet-triplet basis ...... 69 1.7.5 Coupling energies and the singlet-triplet qubit ..... 72 1.7.6 Two-qubit logical operations ...... 73 1.8 Introduction to logical qubits and fault-tolerant quantum com- putation...... 82 1.8.1 Logical qubits and encoding schemes ...... 83 1.8.2 Quantumerrordetection...... 85 1.8.3 Stabiliser codes ...... 88 1.8.4 Fault tolerant quantum computation ...... 95
2 The Silicon-MOS quantum dot platform 100 2.1 Accumulation of electrons at a semiconductor/oxide interface . 101 2.2Quantumdots...... 102 2.3Thesingleelectrontransistor...... 105 2.3.1 CoulombBlockade...... 105 2.3.2 TheSETasachargesensor...... 107 2.4 Dots as qubits ...... 108 2.4.1 Statereadoutmethods...... 110 2.4.2 State initialisation methods ...... 116 2.4.3 Qubit excited states ...... 119 2.4.4 Spin-orbit coupling ...... 122
xiii CONTENTS
2.5 SiMOS quantum dot qubit architectures ...... 128 2.5.1 Device architecture 1: Single-qubit experiments ....128 2.5.2 Device architecture 2: Singlet-triplet experiments . . . 132 2.5.3 Devicearchitecture3:Improvedconceptdevice....135
3 Randomised benchmarking for SiMOS quantum dot single- qubit fidelities 138 3.1 The Randomized Benchmarking experiment on a SiMOS qubit ...... 140 3.1.1 ExperimentalrealisationoftheCliffordgates...... 140 3.1.2 Result:Non-exponentialdecays...... 140 3.2 Randomized Benchmarking in the presence of low frequency noise ...... 142 3.2.1 Non-Markovian noise in a quantum dot qubit .....145 3.2.2 Modelling noise sources ...... 146 3.3 Extending Randomised Benchmarking ...... 148 3.3.1 Eliminating the constant for a single-qubit randomised benchmarkingmodel...... 148 3.3.2 Re-analysisofexperimentaldata...... 151 3.3.3 Interpretingthetwofidelitymodel...... 156 3.4Discussionofresults...... 159
4 Enhanced qubit readout in the singlet-triplet basis via metastable state latching 162 4.1 Physical description ...... 163 4.2Experimentsfordetection...... 165 4.3Improvementstoreadout...... 170 4.4Discussionofresults...... 173
5 Anti-crossing dynamics of SH and T− in a SiMOS quantum double-dot 175 5.1 Coupling between the singlet and polarised triplets ...... 176 5.1.1 Physicalorigins...... 176 5.2 Experimental protocols probing and utilising the SH /T− coupling...... 178 5.3 Detection and measurement of Δ(θ): Single passage Landau-Zener excitations...... 180 5.3.1 Single passage experiment details ...... 181
xiv CONTENTS
5.3.2 Single passage experiment results and discussion ....182 5.4 Further mapping of the two-qubit Hamiltonian ...... 184 5.4.1 Landau-Zener-St¨uckelbergInterferometry...... 185 5.4.2 Spin-funnel experiment ...... 187 5.5Discussionofresults...... 188
6 Integrated platform for SiMOS quantum dot qubits 190 6.1 From the singlet-triplet basis to the computational basis: Increasing applied magnetic field ...... 191 6.2 Single spin addressability via electron spin resonance techniques194 6.2.1 Spectrum...... 194 6.2.2 Experiment...... 195 6.3 Two-qubit exchange based SWAP operations...... 198 6.3.1 Experimentfordetection...... 198 6.4 Modelling exchange coupling between two quantum dots . . . 202 6.4.1 Modelling tunnel coupling ...... 202 6.4.2 Exchangefittingfromtwo-qubitdatasets...... 202 6.5 Fidelity concerns of two qubit operations ...... 205 6.5.1 Preparation/Initialisation...... 205 6.5.2 Ramping related errors ...... 205 6.5.3 Controlerrors...... 206 6.5.4 Simulations of error processes ...... 207 6.6Discussionofresults...... 212
7 Towards logical qubits and scaled quantum systems 214 7.1 An extensible logical qubit ...... 215 7.2 A linear array of SiMOS quantum dot qubits ...... 216 7.2.1 Fundamental control operations for single qubits ....217 7.2.2 Exchange control in a linear array ...... 220 7.2.3 Simulated qubit performance ...... 223 7.2.4 Tick-tockcontrolprotocol...... 230 7.3 A logical qubit in one dimension ...... 234 7.4Discussionofresults...... 238
8 Discussions, conclusions and future work 241
xv CONTENTS
Appendices 263
A Acknowledgements: Part 1 264
B Acknowledgements: Part 2 266
C Acknowledgements: Part 3 268
D Lanau-Zener Transitions 270 D.1Wavefunctioninthediabaticpicture...... 270 D.2 Non-adiabatic transitions: the solution when t →∞...... 272
E No-cloning theorem for quantum information 276
F The SiMOS fabrication process 278 F.1Microfabricationstages...... 278 F.2Nanofabricationstages...... 279
G Experimental equipment and protocols 283 G.1Devicepackaging...... 283 G.2 Cryogenic equipment and measurements ...... 285 G.2.1 Liquid helium dewar ...... 285 G.2.2 Dilution refrigerator ...... 287 G.3Electronichardware...... 288
H Configuration regimes of Architecture 3 291 H.1 Single-reservoir configuration ...... 292 H.2 Dual-reservoir configuration ...... 293
I Additional device characterisation 295 I.1 Deviceleverarms...... 295 I.2 Excitedstates...... 296
J Chronological experimental pathway for |Δ(θ)| 298
z K Fitting Bos from the spin funnel experiment 300
L Additional spin-funnel data 304
xvi CONTENTS
M Instruction set and design rules for linear nearest-neighbour error correction 306
N Encoding and concatenation schemes for linear nearest-neighbour error correction 309
O Simulated logical qubit performance of a linear array of SiMOS quantum dots 312
P Experimental pathway to a logical qubit 316 P.1ParityMeasurements...... 316 P.2Correctingonetypeoferror...... 317 P.3Correctingasingleerrorofanytype...... 319 P.4 Demonstrating a LNN logical qubit ...... 320
Bibliography 322
xvii List of Figures
1.1Blochsphereforatwolevelsystem...... 17 1.2Anti-crossingbetweentwostates...... 19 1.3 Single electron qubit Bloch sphere showing two axis control . . 35 1.4 Operation protocol and evolution Bloch spheres for dephasing andechoexperiments...... 47 1.5 Operation protocol for CP, CPMG and CDD sequences .... 50 1.6Bit-flipchannel...... 52 1.7Phase-flipandthebit-phase-flipchannels...... 53 1.8Depolarisingchannel...... 55 1.9 Operation protocol for the randomised benchmarking experi- ment...... 60 1.10 Bloch sphere representation of an error process ...... 61 1.11 Anti-crossing between the (1,1) states and the (0,2) singlet state. 70 1.12 Singlet-triplet qubit Bloch sphere...... 72 1.13 Control-based two qubit operations...... 74 1.14 Two qubit SWAP operation family circuit symbols and SWAP operationontheBlochsphere...... 76 1.15 Two qubit operations composed from the CZ gate and single qubit unitary operations...... 79 1.16 Two qubit operations composed from the SWAP-class gates and single qubit unitary operations...... 80 1.17 Encoding circuit for the logical qubit ...... 85
2.1 SiMOS quantum dot structural cross-sections and accompa- nyingschematicofelectrochemicalpotentials...... 103 2.2SETdeviceimageandoperation...... 106 2.3SETdeviceasasensor...... 107 2.4 Elzerman readout protocol for electron spin-to-charge conversion111
xviii LIST OF FIGURES
2.5Paulispinblockadefromkeyliteraturesources...... 115 2.6LiftingofPaulispinblockade...... 118 2.7 Breaking of valley degeneracy in silicon based nano-devices . . 120 2.8OrbitalstructureinaSiMOSquantumdot...... 121 2.9 Spin-orbit coupling in momentum space ...... 123 2.10 Spin-orbit coupling due to cyclotron motion of the electron . . 124 2.11SiMOSdevicearchitectureforCh.3...... 129 2.12SiMOSdevicearchitectureforCh.4-Ch.6...... 133 2.13 Concept SiMOS device architecture for improved performance 136
3.1 Non-exponential average fidelity from the randomised bench- markingexperimentduetolowfrequencynoise...... 143 3.2Modelfitstonon-exponentialfidelitydecays...... 152 3.3Two-frequencymodelfittingtobenchmarkingdata...... 158
4.1 Pictorial representation of the device cross section through the active region and the latching process...... 164 4.2 Three level pulse for spin blockade detection ...... 166 4.3 Extended study on latching ...... 168 4.4 Readout comparison between standard and latched Pauli spin- blockade...... 171
5.1 Pulse sequences for Δ(θ) coupling experiments ...... 179 5.2ResultsforsinglepassageLandau-Zenerexperiments.....183 5.3 Results for double passage Landau-Zener-St¨ukelberg experiment185 5.4 Results for the spin funnel experiment ...... 187
6.1 Energy diagram of the two-qubit charge-state anti-crossing il- lustratingthecomputationalbasis...... 192 6.2 Theory for spectral response of ESR for two electron spins under increasing exchange coupling ...... 195 6.3 Experimental observation of individual addressability of two spin qubits in a double quantum-dot ...... 196 6.4 Experiment design and results for coherent exchange driven oscillations ...... 200 6.5 Effective exchange with detuning as accumulated from multi- ple two-qubit experimental results ...... 203 6.6 Numerical simulation of exchange oscillations under realistic noise...... 208
xix LIST OF FIGURES
7.1 Proposed architectures for expansion into a logical qubit. . . . 217 7.2 Distribution and tunable range for g-factors in a linear array. . 219 7.3 Two-qubit exchange coupling via tilt control...... 221 7.4 Two-qubit exchange control via tunnel coupling modulation. . 222 7.5 Actively decoupled CZ gate...... 226 7.6 CZ simulation pulse shape characteristics...... 227 7.7ResultsofCZfidelitysimulations...... 228 7.8 Tick-tock control protocol for a linear array...... 230 7.9 Control sequence for parity-measurement experiment...... 233 7.10 Code concatenation using tiled representation...... 235 7.11 Simulated logical error rates for the two-qubit repetition code. 237 7.12 Experimental pathway to a logical qubit in quantum dots. . . 238
G.1RFenclosureusedtosupportthedevice...... 284 G.2 Liquid helium dewar device dipping stick schematic ...... 286 G.3 Hardware connection diagram for dilution refrigerator .....288
H.1 Multiple configurations of Architecture 3 ...... 292
I.1 Double dot chemical potential and valley splitting ...... 296 I.2 Energydiagramforthefirstexcitedstate...... 297
J.1 Results for single passage Landau-Zener experiments compar- ingvoltageramprates...... 299
K.1 Fitting spin funnel data ...... 302
L.1 Original spin-funnel data ...... 305
M.1 Standard-LNN instructions and their circuit-diagram symbols. 307
N.1 Encoding scheme for LNN blocks...... 310
O.1 Simulated logical error rates for the four-qubit code...... 314
P.1 Tile formalism for the three qubit repetition code...... 318
xx List of Tables
1.1TruthtableforCNOToperation...... 75 1.2 Unitaries in the stabiliser formalism ...... 91
3.1 Implementations of the single qubit Clifford set ...... 141 3.2AkaikeInformationCriterion...... 155 3.3Gatefidelityestimatesforeachofthedatasets...... 159
6.1 Key noise coupling mechanisms for singlet-triplet Hamiltonian 209
G.1 Room temperature electronics used to support dilution refrig- eratormeasurements(Table1)...... 289 G.2 Room temperature electronics used to support dilution refrig- eratormeasurements(Table2)...... 290
K.1 Fitting parameters from spin funnel data ...... 303
xxi Introduction
Quantum computers
The full realisation of error corrected quantum computation is recognised as one of today's greatest scientific and technological challenges, resulting in a rapidly growing multi-disciplinary research area which explores diverse physical systems including photons, trapped ions and nuclei, superconducting circuits and semiconductor devices. The theorised quantum computer behaves vastly different from its classi- cal counterpart through its fundamental operation. Rather than manipulat- ing a singular bit of classical information represented by either a 0 or a 1, the quantum computer encodes quantum information onto the wavefunction of a quantum two-level system, forming a quantum bit or “qubit”. Through utili- sation of a quantum mechanical property known as entanglement,or“spooky action at a distance” as it is often referred to colloquially, the quantum com- puter is capable of exploiting the full complexity of a many-particle quantum wavefunction to solve a particular set of computational problems [1].
1 INTRODUCTION
Applications of a quantum computer
Such a machine is capable of executing its own set of algorithms, with promi- nent examples outlined in Ref. [2]. These include Grover's algorithm, which is capable of performing an optimal search for specific entries in an unordered dataset [3, 4] and Simon's algorithm for determining the exclusive-or (XOR) mask over which a given black-box function is invariant [5] (corresponding to the hidden Abelian subgroup problem, which includes factoring integers into primes and calculating discrete logarithms [6]). Perhaps the most influential or iconic algorithm known today is indeed that of Shor [7], whom theorised an algorithm capable of prime factorisation of integers in near polynomial time. This example became an iconic demonstration of how a quantum com- puter would be capable of out-performing a classical computer the execution of some specific tasks, while also placing many common encryption methods at risk. The tantalising implications of being able to efficiently produce solutions to the aforementioned problems, among others, has since lead to the rapidly expanding fields of quantum computation, quantum cryptography, quantum secure networks and quantum information processing. Today, quantum com- puters are theorised to have vast and revolutionary impact on many areas in science, technology, intelligence and commerce. Specific examples stem from cryptography [8] artificial intelligence [9], and dynamic systems with a large number of input variables such as the weather or the stock market [10]. Methods of solving various logistics-based problems derived from the trav-
2 INTRODUCTION elling salesman problem [11, 12] also stand to benefit, providing efficient solutions to freight distribution, PCB drilling, wiring, X-ray crystallography sample positioning optimisation etc. Perhaps the most intriguing and impactful application of the quantum computer is, in fact, the purposes for which it was originally theorised. The original proposal by Feynman [13] was to use quantum computers for the ef- ficient simulation of quantum systems and quantum mechanics itself. While this gives the ability to simulate the very properties which allows the com- puter to function, it results in the chance to explore many other physical pro- cesses such as particle physics [14] and chemical reactions [15, 16, 17]. The latter example is of particular revolutionary importance as it would make rapid prototyping of new pharmaceutical products conceivable by adding ef- ficient simulation stages before expensive compound synthesis is required. Further, with the advantages of more efficient DNA sequencing (again, a problem related to the travelling salesman), patient-tailored treatment pro- grams can also be designed, leading to improvements in healthcare and indi- vidual quality-of-life within a community.
Approaches to quantum computation
Quantum computation has been approached using a wide range of physi- cal implementations, including chip-scale waveguide quantum circuits [18, 19], trapped ions [20, 21] and neutral atoms in an optical lattice [22, 23], liquid-state nuclear magnetic resonance [24, 25, 26, 27], superconducting res-
3 INTRODUCTION onators [28, 29, 30, 31], topological qubits [32, 33], optically addressable colour centres in diamond [34, 35], and a variety of qubit species constructed using solid-state materials based on donors and quantum-dots. Some of these species draw advantage by being optically active materials, such as self- assembled quantum dots in InGaAs/GaAs heterostructures [36] and Silicon- Carbide [37], producing qubits which naturally couple to optical resonators. Other species such electrostatically defined quantum dots [38] in GaAs [39] and Silicon (both Si/SiGe [40] heterostructures and Silicon based Metal- Oxide-Semiconductor dots [41, 42, 43]) as well as singular dopants [44, 45] or dopant clusters [46] draw upon the ability to define large arrays of densely packed qubits assembled on a solid-state host.
Dot-based architectures and scaling
For electrostatically defined quantum dots in semiconductors, the earliest proposal for a quantum computer is that of Loss and DiVincenzo [38], which details arrays of dots occupied by a singular electron whose spin state forms an individual qubit. Execution of preliminary gates forming quantum algo- rithms would be achieved by altering voltages on the lithographically defined gates to either isolate or couple adjacent qubits via the (relatively) short range electron-electron Heisenburg exchange interaction. Based on these fundamental qubit design and interaction concepts, sev- eral architecture schemes have been derived in order to address the issue of scaling these systems into quantum registers useful for quantum compu-
4 INTRODUCTION tation. Experimental proposals have been put forward in order to move such qubit architectures from the “test-bench” laboratory demonstrations of single-qubit and two-qubit control operations, to full implementations of er- ror corrected logical qubits [47]. Further, array-based quantum computer architectures are considered advantageous for achieving fault-tolerant oper- ation leading to many proposed architectures based on qubit arrays in one dimension [47, 48], while modern algorithms such as the Surface Code [49] have resulted in several 2D qubit array designs capable of achieving such a goal [50, 51, 52, 53].
CMOS industry compatibility
Modern classical semiconductor-based computational technologies are cen- tred around the CMOS industry, with silicon as the semiconductor used in these devices. As this industry is very mature in its design methods and fab- rication processes, it is apparent that compatible semiconductor-based quan- tum computing platforms stand to benefit from the available technologies of the global CMOS industry. Several material platforms used to approach constructing a quantum computer intend to exploit the state-of-the-art in semiconductor manufacturing in order to undertake rapid prototyping and to push quantum computers towards the commercial sector. These include the Silicon-Metal-Oxide-Semiconductor (SiMOS) metal-gate [41] and polysilicon- gate [43] devices, as well as the Silicon-On-Insulator (SOI) nanowire tech- nologies [54]. Recent developments have seen the SOI technology being
5 INTRODUCTION used in commercial semiconductor foundries to produce quantum dot based qubits [55, 56], and have even been integrated with a field effect transistor (FET) [57] used for switchable readout of quantum dots.
Thesis aims and scope
This thesis undertakes various studies using the SiMOS quantum-dot qubit architecture [41]. Although these devices have not yet been produced via a commercial semiconductor foundry, the material components of this platform are compatible with the CMOS industry. The studies presented here yield strong evidence of the ability for this platform to be scaled into a large-scale quantum register, fully fledged with high single-qubit control fidelity [58], capability to individually address single qubits within an array [59, 60] and qubit readout via scalable methods [60]. Several of these requirements have been realised on a singular qubit device demonstrating the SiMOS platforms capability to fully integrate all qubit initialization, readout and control in a scalable manner [60]. Further, experimental pathways have been instigated for the study of individual requirements for quantum error correction building up to the demonstration of a fully error correctable logical qubit in a 1D array of quantum dots [47]. The extensive scope of this dissertation is broken down into the following chapters:
6 INTRODUCTION
Chapter 1: Background theory and qubit fundamentals
This chapter introduces fundamental physical concepts within the field of quantum computing. Building from the ground up, this chapter explains the basics of qubits and how the quantum information they represent can be ma- nipulated in a controlled manner. This then moves on to discuss the theory behind manipulating quantum systems through dynamic potentials and ap- plied fields within the quantum dot devices, and how this can influence the qubit states via Landau-Zener or Rabi processes. Qubit noise and effects upon qubit coherence is also discussed, leading into experiments which can assess qubit coherence and fidelity. As scaled systems often employ ancillary qubit states for initialization and/or readout of data qubit states, the ex- pansion into two-electron systems is also introduced and discussed at length. Finally, as many of the experimental results presented in this dissertation are accompanied by simulation work, a brief discussion of the simulation of qubit systems is also included.
Chapter 2: The Silicon-MOS quantum dot platform
This section introduces detailed concepts of the SiMOS quantum dot qubit platform studied. Basics of the platform are discussed including how the quantum dots are electrostatically defined, and how they can be used to pro- duce sensitive electrometers as well as qubits. For quantum dots as qubits, state initialization and readout methods are discussed, as well as qubit ex- cited states. Further, a number of fundamental concepts specific to the ma-
7 INTRODUCTION terials system used are reviewed as many are later integrated as a part of experimental processes.
Chapter 3: Randomised benchmarking for SiMOS quantum dot single-qubit fidelities
The first results chapter presented in this dissertation is that of the single- qubit randomized benchmarking (RB) results as presented in Ref. [58]. Here, the assumptions put in place for the RB experiment are challenged as en- vironmental noise sources within the SiMOS system are non-Markovian in nature. The implications of this are discussed, while the analysis of data pro- duced by the RB experiment is extended to be more tolerant of such noise sources. Using the novel analysis techniques presented, the single-qubit sys- tem analysed is shown to be capable of operating at error rates compatible with fault-tolerant operation of a scaled quantum register.
Chapter 4: Enhanced qubit readout in the singlet-triplet basis via metastable state latching
This second results chapter illustrates the first step towards demonstrating a fully integrated platform for saleable quantum computation using the SiMOS quantum dot qubit platform as presented in Ref. [60]. This chapter details how measurement signals from readout which exploits Pauli spin blockade of a system of two electrons and two quantum dots can be enhanced through a state latching protocol. This chapter presents the physical description of
8 INTRODUCTION the enhancement protocol, as well as analysing the sensitivity enhancement of the sensors to state readout. This enhancement to readout sensitivity is exploited for the remainder of the results chapters affiliated with Ref. [60].
Chapter 5: Anti-crossing dynamics of SH and T− in a SiMOS quan- tum double-dot
The third results chapter details the analysis of the interactions between the hybridising singlet state SH and polarised triplets (results presented for the
T− triplet state) following results as discussed in Ref. [60]. The origins of such an interaction are discussed in context within current studies presented in literature, as well as discussions of how such an interaction is to be measured via exploiting coherent Landau-Zener transitions over the anti-crossing. Not only is the magnitude of this interaction measured for the experiments in- volved in this dissertation, but the interaction is further utilised to probe other components of the two-electron interaction Hamiltonian (namely, the Heisenburg exchange energy J). Knowledge of the magnitude of this inter- action term and anti-crossing location in parameter Hamiltonian space are vital for the success of experiments presented in later chapters.
Chapter 6: Integrated platform for SiMOS quantum dot qubits
This fourth results chapter presents the most important advancements to the SiMOS platform presented in Ref. [60]. Utilising the results discussed in the previous two chapters, a set of experiments which demonstrate a fully inte-
9 INTRODUCTION grated platform for quantum computation using SiMOS quantum dot qubits. This chapter details how state transformations between singlet-triplet basis states, where the electrons are initialised and measured, are mapped semi- adiabatically to prepare computational basis states receptive to quantum control algorithms. These algorithms are achieved through individually ad- dressing constituent electrons within the system, as demonstrated experi- mentally through electron spin resonance (ESR) techniques. Alongside the requirement for single-qubit control is the capability to perform two-qubit entangling gates between adjacent quantum dots. This is often achieved through mediation of the Heisenburg exchange interaction between two elec- trons. This chapter details experiments which not only demonstrate the ability to measure the Heisenburg exchange interaction in parameter space, but to fully control the exchange to produce one of the first instances of a two-qubit SWAP operation in the SiMOS system. Finally, this chapter concludes with a discussion on the sources of error mechanisms within the complete exchange based protocol, including state preparation, measurement and control errors based on two prominent noise sources within the SiMOS system.
Chapter 7: Towards logical qubits and scaled quantum systems
This final results chapter is dedicated to the discussion of the experimental state of the art for the SiMOS quantum dot qubit system, and the experi- mental roadmap towards realising a fully error corrected logical qubit in a
10 INTRODUCTION linear array of quantum dots following Ref. [47]. The chapter discusses how the current SiMOS test-bench devices presented in earlier parts of this the- sis can be scaled to system sizes capable of demonstrating such intermediate experiments. This includes discussions on how elementary operations will be executed on scaled qubit systems, as well as the presentation of qubit per- formance simulations under these protocols based on realistic noise sources. This work forms a feasibility study for the use of state-of-the-art SiMOS quantum dot qubits to demonstrate elements of an extensible logical qubit. This work is extended further, describing encoding schemes and execution of stabiliser codes on a linear array, the performance of which are also simu- lated to demonstrate thresholds for error correction. These elements are the main focus of Ref. [47], and so are also summarised in this chapter.
Chapter 8: Discussion, conclusions and future work
This chapter takes a broader view of the experimental scope presented in this dissertation. The results of these experiments are discussed and compared with parallel studies produced from other qubit platforms. The impact of these results on the field to date, as well as potential avenues for future work are also highlighted.
11 Chapter 1
Background theory and qubit fundamentals
Just as a single binary bit of information is to a classical computer, the quantum-bit, or qubit is the fundamental element of a quantum computer. Any quantum two level system, or larger physical systems which can be trun- cated into a quantum two level system, can be used to form a qubit. There- fore, there are many diverse platforms upon which qubits can be studied, offering a vast array of methods to control, measure or probe qubit physics. In fact, five essential criteria have been identified [61] which any system must be able to satisfy in order to viably create a quantum computer.
12 CHAPTER 1
The five essential criteria which a qubit platform is required to satisfy are known as the DiVincenzo criteria, and are summarised as the following:
1. The platform produces a scalable physical system with well charac- terised qubits
2. The ability to reliably initialise the qubits in a known state
3. Long coherence of qubit states
4. A universal set of qubit operations
5. Ability to measure the qubit state.
Using these criteria as a framework, this chapter discusses the elements which are required form a qubit, and how qubit characteristics are used to represent and manipulate quantum information. Some specific examples will be given for electron spins in the SiMOS platform, however the core requirements to form a qubit extend to all platforms and quantum computing architectures. The chapter concludes by introducing fundamental concepts required for the expansion of qubit systems into fully scaled, error corrected quantum processors, as well as the conditions necessary for attaining fault tolerant quantum computation.
13 CHAPTER 1
1.1 Quantum two level systems and the quantum-
bit
This section introduces how a qubit is represented as a two level system, using basic concepts of linear algebra. Characteristics of a single qubit are discussed, including common concepts such as the qubit wavefunction and density matrix, visualisation on the Bloch-sphere, energy representation, cou- plings and dynamics.
1.1.1 Two level systems
A two level system is a quantum mechanical system where the system wave- function |ψ can be represented by linear combinations of two discrete and orthogonal eigenvectors |φ1 and |φ2: ⎛ ⎞ ⎜α⎟ |ψ = α |φ1 + β |φ2 = ⎝ ⎠ , (1.1) β where α, β ∈ C, the set of complex numbers, and |α2 + β2| =1asgiven by wavevector normalisation. The Hamiltonian, H, of such a system is a 2 × 2 Hermitian operator corresponding to the total energy of the system, with eigenenergies giving the energies of the two discrete basis eigenstates
{|φ1 , |φ2}. The evolution of the wavevector with respect to time can be given by the
14 CHAPTER 1
Schr¨odinger equation ∂ H |ψ(t) = i |ψ(t) , (1.2) ∂t for a time-independent Hamiltonian. This can be solved to yield the evolution of the wavefunction after time t = t0 → tf as
|ψ(tf ) = U |ψ(t0) , (1.3) where the time evolution operator
−iHt U =exp . (1.4)
1.1.2 Pauli matrices
Electron spin can be represented as a quantum two level system which exists in the vector space of 2 × 2 Hermitian matrices. In order to describe the dynamics of the spin as represented in this form, a set of matrices which form a basis for this vector space are the Pauli matrices:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎜10⎟ ⎜01⎟ ⎜0 −i⎟ ⎜10⎟ σ0 = ⎝ ⎠ ,σxˆ = ⎝ ⎠ ,σyˆ = ⎝ ⎠ ,σzˆ = ⎝ ⎠ , 01 10 i 0 0 −1 (1.5) where σ0 = 1 is the two dimensional identity matrix.
15 CHAPTER 1
1.1.3 Density matrix and the Bloch representation
Another convenient notation used to express the quantum two level system is by means of a density matrix given by ρ. This has the following relation to the wavevector:
ρ = Pj |φjφj| , (1.6) j
P ≥ P × where j 0and j j = 1. The result is a 2 2 Hermitian operator interpreted as describing the statistical ensemble of states in a quantum sys- tem [62], being in state |φj with probability Pj. Thus, it is clear that taking the trace over the density matrix yields
Tr(ρ)=1. (1.7)
Further, ρ2 = ρ if, and only if, the state is pure (i.e. ρ = |ψψ|). It follows that the trace over ρ2, also known as the purity,isgivenby
Tr(ρ2) ≤ 1, (1.8) which is equal to 1 only for a pure state. From what is presented in Sec. 1.1.2, this density matrix can be written as an expansion of a 3D vector over the
16 CHAPTER 1 orthonormal basis given by the Pauli matrices
1 ρ = (1 + Rxσxˆ + Ryσyˆ + Rzσzˆ) (1.9) 2 1 = 1 + R· σ , (1.10) 2
with R =(Rx, Ry, Rz), known as the Bloch vector,and σ =(σxˆ,σyˆ,σzˆ). From the purity relationship
|R| 2 +1 Tr(ρ2)= ≤ 1, (1.11) 2 which leads to important conclusions:
1. The length of the Bloch vector cannot exceed unity.
2. Any point where |R|2 = 1 corresponds to a pure state.
z
R Ɇ y ɔ x
Figure 1.1: Bloch sphere representation of the density matrix in 3D space. The state of the quantum mechanical systems is represented by Bloch vector R .
17 CHAPTER 1
The first conclusion tells us that a natural visualisation for the density matrix is as a unity sphere in spherical coordinates (1,ϑ,ϕ), known in this context as the Bloch sphere. From the second conclusion, the surface of this sphere represents a manifold of pure states. Such a sphere is visualised in Fig. 1.1.
1.1.4 Anti-crossings
In this section the system describing Hamiltonian is introduced on a more basic level. Consider a two level represented by the two states |φ1 and |φ2, which have energies given by E1 = −ε and E2 = ε respectively. The Hamiltonian for such a system can be described using the Pauli matrices introduced in Sec. 1.1.2:
H0 = εσzˆ, (1.12) which has eigenbasis representation:
H0 |φ1 = ε |φ1 , (1.13)
H0 |φ2 = −ε |φ2 ,
where eigenenergies E1,2 = ±ε.Forε = 0, these two states become degener- ate. Therefore, the term ε describes the detuning of the system away from that degeneracy point. If we now add a coupling term Δ between these two
18 CHAPTER 1 states, we result in the Hamiltonian described by:
HQ = εσzˆ +Δσxˆ. (1.14)
The associated energy diagram is pictorially illustrated in Fig. 1.2. For the
ۄɔ1| ۄɔ2|
2ȟ Energy
ۄɔ2| ۄɔ1| 0 Detuning Ԫ
Figure 1.2: Energy diagram of the two states |φ1 and |φ2. These two states form an anti-crossing due to a coupling mechanism Δ. The parameter ε represents a detuning of the system eigenenergies away from this avoided crossing point. above Hamiltonian, which is now a coupled two level system, the new eigen- basis can be represented as a combination of the eigenbasis of the decoupled
Hamiltonian H0 of Eq. (1.12):
HQ |ψe = Ee |ψe , (1.15)
HQ |ψg = Eg |ψg .
The high energy (excited) state |ψe and low energy (ground) state |ψg are
19 CHAPTER 1 represented as: θ θ |ψ =sin |φ +cos |φ , e 2 2 2 1 √ (1.16) 2 2 Ee = ε +Δ , and θ θ |ψ =cos |φ −sin |φ , g 2 2 2 1 √ (1.17) 2 2 Eg = − ε +Δ , −1 Δ for θ =tan ε . The Hamiltonian in Eq. (1.14) is often known as the qubit Hamiltonian, and has the ability to access rotations around at least two orthogonal axes (here,x ˆ andz ˆ are involved). This demonstrates the capability to control the state of the system over the entire Hilbert space, thus producing a controllable quantum-bit, or qubit. It is observed by the commutation relation
[σ1,σ2]=2i123σ3, (1.18)
where 123 is the Levi-Civita symbol, that operations involving σyˆ can be constructed through a combination of operations from the remaining two orthogonal Pauli matrices.
20 CHAPTER 1
1.2 Postulates of quantum mechanics and
measuring a quantum system
So far, the discussion of the qubit has been limited to a basic description of how a two level system can be represented by the wavefunction and accom- panying Hamiltonian. Here, the postulates of quantum mechanics are introduced, alongside the process and effects of measuring a quantum system. The discussion below is adapted from that of Ref. [63], which should be referred to for additional details. The postulates are described below and, where relevant, linked to more in-depth discussions within the chapter. The key postulated which pertain to the discussion of the measurement of quantum systems are the Third, Fourth and Fifth, however the complete discussion of the postulates as presented in Ref. [63] is given here for additional context and completeness.
1.2.1 First Postulate
At a time t, the state of a quantum system is given by |ψ(t) which is in a suitable vector space E. This is the topic of discussion in several of the subsections of Sec. 1.1. In brief, it is stated in Eq. (1.1).
21 CHAPTER 1
1.2.2 Second Postulate
Every measurable physical quantity A is described by an observable operator A, acting within E. This is again the topic of discussion in several of the subsections of Sec. 1.1, however it does introduce the concept of system observables.The Hermitian operator A is said to be an observable if its orthonormal set of eigenvectors forms a basis in E. For example, SZ = σzˆ/2 is an observable for the basis defined in Eq. (1.12) due to the following relationship:
S |φ = |φ , Z 1 2 1 (1.19) S |φ = − |φ . Z 2 2 2
1.2.3 Third Postulate
The only possible result of the measurement of a physical quantity A is the one of the eigenvalues of the observable A. This postulate is simply a statement of the system quantisation. The system Hamiltonian H is a description of the energy within the quantum mechanical system, therefore, the only energies possible are those described by the eigenvalues of H. Since A is Hermitian, the measurement of A must be a real value, and if the spectrum of A is discrete, results observed when measuring A are quantised.
22 CHAPTER 1
1.2.4 Fourth Postulate
When the physical quantity A is measured on a system in a normalised state
|ψ, the probability P(an) of obtaining the non-degenerate eigenvalue an of the corresponding observable A is:
2 P(an)=|un|ψ| , (1.20)
where |un is the normalised eigenvector of A associated with eigenvalue an. This introduces the probabilistic nature of measuring a quantum me- chanical system. This concept is discussed in the context of a more concrete example in Sec. 1.4, when discussing the Rabi formula. See Ref. [63] for details on systems involving degenerate energy levels. In the most generic terms, consider an arbitrary spin direction in spherical coordinates (1,ϑ,ϕ) giving the state at time t of
ϑ −iϕ ϑ iϕ |ψ(t) =cos exp |φ +sin exp |φ . (1.21) 2 2 1 2 2 2
When this system is measured, the probability that the system energy will be /2 (corresponding to |φ1)isgivenby: − 2 2 ϑ iϕ P = |φ1|ψ| = cos exp (1.22) 2 2 2 ϑ =cos2 . (1.23) 2
23 CHAPTER 1
It is clear that the phase angle ϕ, a value orthogonal to thez ˆ-axis from which the observable A, does not affect the projected measurement outcome. Further, the expectation value A of the measured result can be found as
A = anP(an) (1.24) n = ψ|A|ψ . (1.25)
Note: The fact that the measurement of quantum systems is probabilistic in nature requires the build-up of an adequate level of measurement statistics when determining the results of experiments performed on such systems.
1.2.5 Fifth Postulate
If the measurement of the physical property A on the system in the state |ψ gives a result an, the state of the system immediately after the measurement is the normalised projection of the state |ψ onto the eigensubspace associated with an. That is; for measurement at time t, the state immediately after the measurement is P |ψ ψ(t+) = n . (1.26) ψ|Pn|ψ
Here, Pn is a projector term given by
Pn = |unun| . (1.27)
24 CHAPTER 1
This implies that the state immediately after a measurement event is an eigenvector of A, with eigenvalue an.
Note: This fundamental aspect of the measurement of a quantum mechan- ical system is also utilised in the initialisation of quantum states before un- dertaking experiments. This concept is discussed further in later chapters.
1.2.6 Sixth Postulate
The time evolution of the state vector |ψ(t) is governed by the Schr¨odinger equation: ∂ i |ψ(t) = H(t) |ψ(t) , (1.28) ∂t where H(t) is the observable associated with the total energy of the system. This is discussed extensively in Sec. 1.1 and Sec. 1.4.
25 CHAPTER 1
1.3 An electron within an external magnetic
field
This section details a textbook example of a quantum mechanical two level system: the spin state of an electron. The section aims to conceptualise the spin, and draws parallels between the quantum mechanical system and an analogy based on classical mechanics. The electron as an elementary quantum two-level system is introduced, and the response of an electron to the externally applied magnetic field is discussed in the context of producing two non-degenerate eigenenergies. Im- portant implications of the of the electron mechanics in this applied field with respect to performing qubit experiments are also discussed.
1.3.1 Larmor precession
The spin of singular electron is a quantum two level system with quantized spin quantum number SZ = ±/2. Hence, the spin of the electron is often represented via the Pauli matrices given in Eq. (1.5) as
S = σ . (1.29) Z 2 zˆ
A classical analogue of spin is a charge q with mass m with an intrinsic an- gular momentum L . Under classical electromagnetism, this system produces
26 CHAPTER 1 a magnetic dipole q M = L. (1.30) 2m
When a magnetic dipole is placed in an external magnetic field B0, the total energy of such a system can be represented as
E = −M · B0. (1.31)
For the spin of an electron in an external magnetic field, this classical analogue gives rise to the basis electron states having an energies given by
| | qe E = gS · B0. (1.32) 2me
where qe and me are the charge and mass of the electron, and g is the electron gyromagnetic ratio [63]. This energy separation forming two discrete electron states due to an externally applied magnetic field is known as the Zeeman energy, which can be instead expressed in terms of the Bohr magneton μB as
gμ E = B S · B . (1.33) 0
Treating this quantity classically once again suggests that the electron can experience a torque Γ, if S and B0 are not parallel. That is;
gμ dS Γ= B S × B = , (1.34) 0 dt
27 CHAPTER 1 which suggests a precession frequency in the time evolution of a non-parallel spin state around the vector direction of the global magnetic field B0. To treat this spin as a quantum mechanical object, we construct the Hamiltonian representing the total energy of the system:
gμ B H = B 0 σ . (1.35) 2 zˆ
Here we have assumed B0 = |B0|·zˆ, which produces a Hamiltonian in the same form of H0 in Eq. (1.12). As an illustrative example, we prepare a state which is perpendicular to the applied field B0, one which is a superposition of the two basis states of the Hamiltonian H: 1 |ψ0 = √ (|φ1 + |φ2) , (1.36) 2 where the projection in thex ˆ direction for this state is given by
S = ψ |σ |ψ = . (1.37) x 2 0 xˆ 0 2
The time evolution of this state under the Hamiltonian H is given by the Schr¨odinger equation Eq. (1.2), resulting in the evolution in time given as
1 gμBB0 |ψ(t) = √ |φ1 exp −i t + |φ2 , (1.38) 2
28 CHAPTER 1 whereby the projection along thex ˆ direction evolves with time as
gμ B S = cos B 0 t . (1.39) x 2
Therefore, it is observed that the orthogonal component of the spin vector to an applied magnetic field B0 undergoes a precession in time with a frequency of gμ B ω = B 0 . (1.40) 0
This physical effect is known as Lamor precession.
1.3.2 The rotating frame
As described in Sec. 1.3.1, a spin in an external magnetic field undergoes a precession at a natural frequency dictated by the applied magnetic field. As the time dynamics of off-axial components become complicated due to the time dependence of the spin, it is a common simplification to make a transformation of the system Hamiltonian into a basis which also rotates at the natural frequency of ω0. From this frame of reference, called the rotating frame, an unperturbed spin vector remains stationary. In order to transform into the rotating frame, the phase accumulation of the eigenenergies of the system described in Eq. (1.1) must be taken into
29 CHAPTER 1 account. This implies a transformation matrix
|ψr = R |ψ ⎛ ⎞ ω0 (1.41) ⎜exp i 2 t 0 ⎟ R = ⎝ ⎠ − ω0 0expi 2 t
where |ψr is the wavefunction in the rotating frame.
1.4 Operations on a single qubit
This section introduces a number of mechanisms which can be utilised in order to manipulate the qubit state. These mechanisms give a certain degree of control over the quantum information represented by the qubit, and can be used to form fundamental control operations for a quantum computer.
1.4.1 The Rabi formula
Consider once again the two level system which is described by the qubit
Hamiltonian HQ of Eq. (1.14). Choose the initial condition such that |ψ0 begins in the unperturbed ground state |φ1. This state can be instead represented in the basis states of the qubit Hamiltonian HQ as
|ψ0 = |φ1 (1.42) θ θ =cos |ψ −sin |ψ . (1.43) 2 g 2 e
30 CHAPTER 1
The time evolution of this state is then
θ iE t θ iE t |ψ(t) =cos exp − g |ψ −sin exp − e |ψ . (1.44) 2 g 2 e
The time evolution can result in the population of the |φ2 excited state. This can be observed from the state projection
1 θ iE t iE t φ |ψ(t) = sin exp − g − exp − e , (1.45) 2 2 2 and, using the fact that
√ θ Δ 2 2 sin = √ ,Ee = −Eg = Δ + ε , (1.46) 2 Δ2 + ε2 the probability of finding the time evolved state in the opposite basis state becomes √ iΔ Δ2 + ε2 φ2|ψ(t) = √ sin t , (1.47) Δ2 + ε2 where the measurable projection into the opposite basis state is √ Δ2 Δ2 + ε2 |φ |ψ(t)|= sin2 t . (1.48) 2 Δ2 + ε2
This is known as the Rabi formula [63].
31 CHAPTER 1
For spins in the rotating frame: Electron spin resonance
As described in Sec. 1.3.2, an electron in an external magnetic field will undergo a Lamor precession. However, the time dependence of in-plane (ˆx andy ˆ) components of the spin vector can be mitigated through a change of basis to the rotating frame (See Sec. 1.3.2). Therefore, in order to reproduce the qubit Hamiltonian HQ within the rotating frame, one must be able to produce a coupling term Δ which is also time dependent in the same manner. Such a coupling can be produced via an in-plane, rotating magnetic field
B1 = B1(ˆx cos(ωt)+ˆy sin(ωt)), (1.49) which rotates with driving frequency ω. When applying such a driving field to the spin, the system Hamiltonian in the standard laboratory frame becomes
ω ω H(t)= 0 σ + 1 (cos(ωt)σ + sin(ωt)σ ), (1.50) 2 zˆ 2 xˆ yˆ where the time evolution becomes
∂ |ψ(t) i = H(t) |ψ(t) . (1.51) ∂t
32 CHAPTER 1
Under the transformation R into the rotating frame, this becomes
∂(R |ψ(t)) i = H(t)(R |ψ(t)), (1.52) ∂t ∂ |ψ (t) i r = H |ψ (t) , (1.53) ∂t r r
where Hr is the time independent Hamiltonian
H = (Ωσ − δωσ + ω σ ), (1.54) r 2 zˆ zˆ 1 xˆ
where δω = ω − ω0 which represents the difference between the Lamor fre- quency and the frequency of the driving field B1,andΩ=ω + ω0 is a “fast” rotating term. Close to the resonance condition, where δω Ω, the ef- fect of the fast term Ω averages to almost zero, leading to the rotating wave approximation: H ≈ (−δωσ + ω σ ), (1.55) r 2 zˆ 1 xˆ with motion dominated by the slowly rotating term. Substituting into the Rabi formula of Eq. (1.48), a driven spin system exhibits the following time dependent probability based on ω2 ω2 + δω2 |φ |ψ(t)|= 1 sin2 1 t . (1.56) 2 2 2 ω1 + δω
33 CHAPTER 1
Driving field phase considerations
The overall phase of the driving field is important as it determines which axis in the rotating frame the electron spin rotates around. We define the following states which are aligned in the positivex ˆ andy ˆ directions
1 |ψx = √ (|φ1 + |φ2) , (1.57) 2 1 |ψy = √ (|φ1 + i |φ2) . (1.58) 2
From the derivations above, it is clear to see that under driving field
B1 = B1(ˆx cos(ωt + ϕ)+ˆy sin(ωt + ϕ)), (1.59) which includes a global phase term ϕ, the respective Rabi formulae become: ω2 ω2 + δω2 |ψ |ψ(t)|= 1 cos2 1 t cos2(ϕ), (1.60) x 2 2 ω1 + δω ω2 ω2 + δω2 |ψ |ψ(t)|= 1 cos2 1 t sin2(ϕ). (1.61) y 2 2 ω1 + δω
Therefore, the choice in phase ϕ is crucial when attempting coherent con- trol experiments as it represents an offset in phase with respect to the global Lamor precession from an arbitrary instance in time (such as the commence- ment of the experiment). Further, as depicted in Fig. 1.3, using phase mod- ulation with a ϕ = π/2 allows for control and manipulation of the state over two orthogonal axes in the rotating frame, and thus complete control over
34 CHAPTER 1
ۄb) |՛ ۄa)|՛
ۄ߰y| ۄ߰y|
ۄ߰x| ۄ߰x|
ۄ՝| ۄ՝|
Figure 1.3: Two axis control for a single qubit constructed from a single electron spin. Based on modulation of the phase ϕ of the driving field B1 thespincanberotateda) around the x-axis for ϕ =0,andb) around the x-axis for ϕ = π/2 the qubit state.
1.4.2 Anti-crossing dynamics
When two states are coupled in the manner presented in Section 1.1.4, via coupling energy Δ, the ground state of the system |ψg becomes dependent upon the coupling Δ and detuning energy ε. That is; as the system is detuned from one side of the anti-crossing to the other, the ground state of the system
|ψg(ε) transfers from one of the two basis states (|φ1 or |φ2 in Fig. 1.2) to the other. Consider the case when transferring from a region where detuning
ε 0and|ψg |φ1, to a region where detuning is ε 0, and |ψg |φ2. From the adiabatic theorem, the state of the system |ψ(t) tracks the ground state |ψg(θ) when the transferral ε 0 → ε 0 is performed infinitely slowly. When the adiabatic condition is not held however, the state |ψ(t)
35 CHAPTER 1 can become a mixture (or a superposition, see Sec. 1.1.1) with the excited state of the anti-crossing |ψ(t) = α1 |ψe(θ) + α2 |ψg(θ). The following sections, together with Appendix. D, detail the mechanisms under which such a superposition is produced.
Landau-Zener processes
The following is a brief overview of the Landau-Zener transition and the parameters which govern the process. A full derivation of the process is given in Appendix D which closely follows the solution in Ref. [64]. Consider again, the quantum two level system represented by the wave- function |ψ |φ1 which is initially tuned such that ε 0. The system can transfer the state from one side of the anti-crossing to the other by modulat- ing ε. The rate at which this is done is identified as the quantity ν,givenby the rate of change in the eigenenergies of the uncoupled system Hamiltonian
H0 of Eq. (1.12) d(E − E ) ν = 1 2 . (1.62) dt
Here, the interaction quantity Δ which gives rise to the qubit Hamiltonian HQ in Eq. (1.14) is considered to be the perturbation under which the interaction takes place [65, 66]. Based on the state energy velocity ν, there is a certain probability that the final state is instead the excited state |ψe, as given by the Landau-Zener
36 CHAPTER 1 excitation probability [65, 66, 67]:
−2π|Δ|2 P =exp (1.63) LZ ν
By controlling ν during the passage through the anti-crossing, the adiabatic- ity of the passage is modulated to produce a certain transition probability into the excited state. For the case where ν Δ, the passage is adiabatic with respect to the perturbation Δ, and it can be seen that PLZ → 0. Conversely, the diabatic limit is the case where ν Δ, where PLZ → 1. For experiments where the coupling energy Δ is a fixed quantity, control over ν can not only be used to probe the magnitude of this coupling energy (as explored in Chapter. 5), but can also be utilised to coherently produce a superposition state of an arbitrary two level system of the form HQ. The creation of such a superposition state can be described in the tran- sition matrix formalism [68, 69, 70, 71], where the transition at the level crossing is given by matrix ⎛ ⎞ √ √ ⎜ PLZ 1 − PLZ exp(iϕS)⎟ M = ⎝ √ √ ⎠ (1.64) − 1 − PLZ exp(−iϕS) PLZ
for the case where |φ1 crosses |φ2 from the low energy side. For a passage through the other direction, the equivalent transition matrix is simply the transpose of M. The phase factor ϕS from Eq. (1.64) is termed the Stokes
37 CHAPTER 1 phase,givenby
π |Δ|2 |Δ|2 |Δ|2 ϕ = +argΓ 1 − i + ln − 1 . S 4 |ν| |ν| |ν|
Here, Γ(...) is the gamma function, which has asymptotic limits [71] ⎧ ⎪ | |2 2 ⎨ Δ | |2 | | |Δ| C | | , Δ ν arg Γ 1 − i ν ⎪ | |2 | |2 |ν| ⎩ − π − Δ Δ − | |2 | | 4 |ν| ln |ν| 1 , Δ ν where C is the Euler constant.
Landau-Zener-St¨ukelberg processes
The transition matrix given in Eq. (1.64) is dependent upon the direction of passage through the anti-crossing, and it can be observed that for multiple passages back and forth through this anti-crossing, the Stokes phase factor ϕS is accumulative. It is therefore more convenient to describe each transition by a diabatic unitary evolution matrix N,givenby ⎛ ⎞ √ √ ⎜ 1 − PLZ exp(−iϕ S) − PLZ ⎟ N = ⎝ √ √ ⎠ (1.65) PLZ 1 − PLZ exp(iϕ S) as derived in detail in Ref. [71]. We have a modification to the Stokes phase parameter ϕ S = ϕS − π/2 due to the evolution. Landau-Zener-St¨ukelberg (LZS) processes describe the passage through the anti-crossing followed by a dwelling time τD and second, identical passage
38 CHAPTER 1 back through the anti-crossing. The time evolution during the dwell time simply results in the unitary ⎛ ⎞ ⎜ 10⎟ U = ⎝ ⎠ , (1.66) − − δE 0 exp( i τD) representing the accumulation of phase based on the energy separation of the two eigenenergies at the dwelling position in ε detuning (recall Fig. 1.2). Therefore, the evolution of this system under this regime becomes
|ψf = N · U · N |ψi , (1.67)
for the case where the initial state is the low energy state |ψi = |ψg |φ1 (i.e. the case for a negative detuning, where |ε| |Δ| in Fig. 1.2). After passing through the anti-crossing to/from a location |ε| |Δ|, which is positive in detuning ε, the probability of measuring the excited state |ψe =
|φ2 is given by
2 2 Pe = |φ2|ψf | =4PLZ (1 − PLZ )sin (ΦSt), (1.68)
where ΦSt is the St¨uckelberg phase [65, 72]
δE · τ ϕ + ϕ Φ = D + S1 S2 , (1.69) St 2 2
th where ϕ Sn is the Stokes phase for the n passage.
39 CHAPTER 1
Due to the dependence of the excited state probability on the energy separation δE within the time evolution, the LZS process is a useful inter- ferometric tool for Hamiltonian characterisation [73, 60] and benchmarking the effect of decoherence due to the environment [74, 75].
Landau-Zener-St¨ukelberg-Majorana interferometry
The theory presented in the previous section can be extended for strongly driven systems, giving Landau-Zener-St¨ucelberg-Majorana (LZSM) interfer- ometry. An extensive discussion of these experiments is given in Ref. [71] and the citations therein. A brief introduction is as follows: The LZSM experi- ments are performed on systems that are continuously driven in time using input signal ε(t)=A cos(2πft). This is a simple extension to the above theory which studies the dynamics of n passages through the anti-crossing, where the state is simply modelled as
n |ψf =(UR · N · UL · N) |ψi . (1.70)
Here, |ψi is the ground state on the right side of the anti-crossing and
UL/UR represent the phase evolution on the left/right side of the anti-crossing respectively.
40 CHAPTER 1
1.5 Noise and quantum coherence
This section discusses the concept of a qubit which is coupled to a noisy environment, and how these noise sources can affect the overall coherence of a qubit and the integrity of the quantum information it represents.
1.5.1 Qubit noise
An ideal qubit Hamiltonian HQ is expressed in Eq. (1.14). For this sim- ple, time independent Hamiltonian, the dynamics are well understood and described by detuning ε and coupling Δ terms. Noise sources in the qubit en- vironment are effects which result in time variance of these two parameters, such that the Hamiltonian is instead described as:
HQ(t)=ε(t)σzˆ +Δ(t)σxˆ, (1.71) where the time dependent components of ε(t) and Δ(t) are the variation terms δε(t)andδΔ(t) respectively, such that
ε(t)=ε + δε(t), (1.72)
Δ(t)=Δ+δΔ(t). (1.73)
The postulates of quantum mechanics as presented in Sec. 1.2 have fun- damental implications of how environmental noise fluctuations impact qubit performance. From the Schr¨odinger equation, it is clear that qubit evolution
41 CHAPTER 1 in a noisy environment leads to the temporal integration over those noise sources. One of the implications of the fourth postulate (Sec. 1.2.4) is that experimental results will often require measurement statistics to probabilisti- cally determine the final state. The integration over noise spectra will result in this final state being subject to a probability envelope when comparing shot-to-shot experiments. This often leads to concepts such as the visibility of qubit readout. These effects must be taken into account when determining an adequate level of measurement statistics. What follows is a brief discus- sion of common examples for each noise type in the context of semiconductor qubits. This discussion highlights the fact that the time dependence in the qubit Hamiltonian is intrinsically tied to fluctuations of environmental elec- tric and magnetic fields.
Environmental sources of δε(t)
As discussed in Sec. 1.4.1 for single electrons driven by a resonant magnetic field, the dynamics are described by the Rabi formula. For this system, the qubit energy separation ε is represented as the difference between the driv- ing frequency ω and Larmor frequency ω0. This fundamental frequency of the spin is defined by the externally applied magnetic field B0,andthere- fore fluctuations in this field can result in δε(t) during electron resonance. Fluctuations of this variety can, for example, be produced by the ensemble effect of a nuclear spin bath in a wide range of semiconductor materials. The detuning parameter ε is also often given in the context of charge based anti-
42 CHAPTER 1 crossings produced by tunnel coupling between two quantum dots [76]. Here the detuning is a function of the electrostatic potentials used to define the quantum wells, often manipulated in semiconductor quantum dots through voltages applied to gate electrodes. Variation in these potentials can arise due to phenomena such as volatile charge or spin fluctuations due to mate- rial defects in oxides and at interfaces (often giving the classic 1/f α noise spectra) [77], or as voltage noise capacitively coupled from surrounding gate electrodes.
Environmental sources of δΔ(t)
For electrons driven by a resonant magnetic field, as described by the Rabi formula, the qubit energy coupling term Δ is described by
gμ |B | Δ=ω = B 1 (1.74) 1
and therefore any fluctuations in magnitude |B1(t)| of the driving field will result in a δΔ(t) term. This effect has been discussed in Ref. [78] with the potential source being thermal Johnson-Nyquist radiation, coupling to an electron spin qubit through the on-chip microwave antenna [79]. For the Landau-Zener crossings discussed in Sec. 1.4.2, the coupling term Δ is instead described by a number of different phenomena. Common phys- ical phenomena such as the tunnel coupling, t0, between two dots can give rise to such effects [80] as well as coupling between the two-electron singlet
43 CHAPTER 1
|S and polarised triplets |T± due to spin-orbit coupling [81, 82, 83, 84, 85]. Both of these physical phenomena are also intrinsically tied to the variation in electrostatic potentials which define the qubit, similar to δε(t). Another means of creating noise of this type is the Overhauser effect between the qubit and a fluctuating nuclear spin environment [86, 87, 88].
1.5.2 Qubit characteristic times
The lifetime and integrity of quantum states intrinsic to the qubit can be represented through two characteristic time-frames known as T1 and T2. These characteristics are adopted from the field of nuclear magnetic reso- nance (NMR) and are a method through which the qubit interaction with the noisy environment can be assessed.
T1 processes
Noisy fluctuations can perturb the qubit operation, adding and subtracting energy from the system. Eventually, any quantum system is brought to an equilibrium with the environment, and the timescale over which the qubit can persist out of equilibrium is referred to as the T1 time. In nuclear physics, this timescale is known as spin-lattice relaxation time, and is a measure of the mechanisms under which the spin vector within a static magnetic field reaches thermodynamic equilibrium with its surroundings (i.e. the “lattice”). For a singular electron, this is the relaxation from an excited state into the ground eigenstate as determined by an externally applied magnetic field, and results
44 CHAPTER 1 in a change in the energy of the qubit (thus, termed a “relaxation” process). Experimentally, the time can be gauged by preparing some superposition state |ψ = α |↑ + β |↓, and monitoring relaxation of this state into the spin down ground state |↓. The characteristic T1 time is extracted from fitting the excitation probability to the exponential form:
2 P↑(t)=α exp(−t/T1). (1.75)
T2 processes and related measures
A qubit interacting with the environment can also undergo an interaction where energy is absorbed or lost in the form of a phase shift. The result is the loss in coherence in the plane orthogonal to the magnetisation axis. This process is given by a timescale called T2. From NMR theory, the T2 time is the mechanism by which the transverse component of the magnetisation vector exponentially decays towards an equilibrium value. Therefore, it is often associated as a measure of the NMR signal strength. The characteristic time is observed due to the interactions of an ensemble of spins dephasing from each other through sampling local magnetic field inhomogeneities. This results in the respective accumulated phases of each spin in the ensemble to deviate from the expected value.
In the context of quantum computers, the value of T2 is used as an initial characterisation of a qubit, as the elements of a quantum computer must operate on timescales much faster than T2 in order to execute algorithms
45 CHAPTER 1 in a fault-tolerant manner. This necessity for fast drive with respect to
T2 is particularly evident in the context of the work presented in Ref. [89], which describes the power-law decay of the Rabi formula in the presence of decoherence as
P A ↑(tMW)= α cos (2πfRabitMW + ϕ) (1.76) tMW
where tMW is the microwave burst duration. As microwave power (i.e. |B1|) decreases, the susceptibility to the influences of dephasing increases. This is due to slower Rabi oscillations resulting in a longer sample over the environ- ment for each rotation period.
The actual value of T2 for a singular qubit is a difficult value to attain experimentally, as the act of taking an ensemble of measurement statistics tends to obscure the information gathered. As a result, there are several experiments which are employed to probe qubit coherence, resulting in a family of T2 measurements as described below.
∗ Pure dephasing T2 ∗ The pure dephasing time T2 represents trial-to-trial variations in qubit res- onant frequency, leading to an apparent damping of wave interference in an ensemble of measurements [1]. Fitting this damping rate yields the time
∗ constant T2 , which also is also representative of the coherence of the Lamor oscillations of a spin [90]. The standard experiment used to gauge this de-
46 CHAPTER 1
a) X Dwell Time ɒ X 2 2
ۄ՛| ۄ՛| ۄ՛| ۄ՛|
y y y y x x x x
ۄ՝| ۄ՝| ۄ՝| ۄ՝|
b) X X Dwell Time ɒȀʹ X Dwell Time ɒȀʹ 2 2
ۄ՛| ۄ՛| ۄ՛| ۄ՛| ۄ՛| ۄ՛|
y y y y y y x x x x x x
ۄ՝| ۄ՝| ۄ՝| ۄ՝| ۄ՝| ۄ՝|
∗ Figure 1.4: a) Ramsey experimental protocol yielding T2 . For each stage of the protocol an ensemble of qubit measurements is shown on Bloch spheres in the rotating frame. As a function of τ, a distribution of dephased states is projected onto thez ˆ-axis for probabilistic quantum measurement. b) H Hahn echo experimental protocol yielding T2 as a function of decaying spin- up probability P↑(τ). As illustrated by accompanying Bloch spheres, the X refocusing pulse applied at τ/2 reverses the dephasing due to fluctuations with timescales greater than τ. cay timeframe is the Ramsey interference experiment, which is described in Fig. 1.4a. For a qubit initial state of |↓, the first X /2 pulse (π/2 rotation around thex ˆ-axis) rotates the spin vector to they ˆ-axis. The qubit then pre- cesses about thez ˆ-axis for a time τ. Finally, the second X /2 pulse rotates they ˆ component of the state back to thez ˆ-axis. If the driving field is ex- actly on-resonance with the Larmor frequency, the qubit is stationary in the
47 CHAPTER 1 rotating frame. However, off resonant driving is detuned in frequency by δω, and bothx ˆ andz ˆ components of the final state beat as a function of δω and τ. This beating produces an oscillatory pattern which is characteristically known as Ramsey fringes. The decay time of these oscillations in turn yields
∗ the T2 dephasing time.
H Hahn echo: T2 The pure dephasing time is subject to all noise sources coupling into the qubit system, however, a single experimental trial may retain its phase co-
∗ herence for much longer than T2 . Certain environmental fluctuations which cause the pure dephasing to be shorter than T2 can be removed by applying additional refocusing pulses such as those shown in Fig 1.4b and Fig 1.5. The simplest of such a sequence is the Hahn echo [91] depicted in Fig 1.4b. The additional X (full π) rotation in the centre of the sequence decouples the ef- fect of scalar spin-spin couplings and spatial variations of the static magnetic field along thez ˆ-axis [91, 90], provided the variations are constant through- out the experiment. This results in the spin being in-phase by the end of the second time interval τ/2, and the resulting echo signal decays as a func-
H tion of τ. The decay time constant is given by T2 , and is a measure of the intrinsic T2 time.
48 CHAPTER 1
CP CP decoupling: T2 The Hahn echo can only decouple fluctuations which occur on a timescale greater than τ. However, this limit can be reduced to decouple fluctuations which occur on the order of τ/n. This is achieved through applying a train of n refocusing X pulses as shown in Fig 1.5a. This is known as the Carr-
CP Purcell sequence [92] and produces the characteristic decay time T2 , which
H extends further towards T2 when compared to T2 for increasing refocusing pulses n. However, this pulse train is limited through the susceptibility to accumulate fixed over- or under-rotation errors in the control pulse.
CPMG CPMG decoupling: T2 This limitation of the CP experiment was addressed by shifting the phase of the refocusing pulses to an orthogonal axis [93] (i.e. Y pulses rather than X ) as shown in Fig. 1.5b. Thus, pulse length errors are compensated on even-numbered echoes. Due to this additional robustness against control
CPMG errors, the T2 time is perhaps one of the best estimates for the intrinsic
T2 coherence time of the qubit.
Dynamical Decoupling A core assumption for the previously presented noise decoupling algorithms is the application of periodic, zero-width π-pulses. These decoupling tech- niques have been extended to preservation of an arbitrary quantum state
49 CHAPTER 1
a) ɒ ɒ X x X ʹ X ɒȀ ʹ 2 2
b) ɒ ɒ X x X ʹ Y ɒȀ ʹ 2 2
c) Y ɒȀX ɒȀY ɒȀX ɒȀ
Y p1 X p1 Y p1 X p1
Y p2 X p2 Y p2 X p2
ɒ ɒ X X ʹ Y p X p Y p X p ʹ 2 Ǧͳ Ǧͳ Ǧͳ Ǧͳ 2
CP Figure 1.5: a) CP experimental protocol yielding T2 , refocussing fluctua- tions which occur down to timescales of τ/n. b) CPMG experimental proto- CPMG col yielding T2 . Refocussing down to τ/n is maintained, however the se- quence is robust against control errors in the form of over- or under-rotations c) Protocol for Concatenated Dynamical Decoupling illustrating the tempo- ral recursive structure which aims to correct for errors at different levels of resolution.
50 CHAPTER 1
(i.e. a quantum memory) through processes known as Dynamical Decoupling (DD). One example of particular note in literature is that of Concatenated Dynamical Decoupling (CDD). It has been shown in Ref. [94] that (CDD) sequences can be made to correct for errors at different levels of resolution through temporal recursive structure as shown in Fig. 1.5c. The sequence type can prevent error build-up often observed in standard decoupling ex- periments. Such CDD pulses are often discussed in the context of quantum memory [94] or as a means of preserving the qubit state during idling tasks for times as close to the limitation of the qubit T2 coherence time as possi- ble [95].
1.5.3 Noisy qubit channels
It is often convenient to represent the application of a noisy operation on a quantum bit as the passage of that qubit though a noisy quantum channel. The channel is often represented by a state transformation given by a linear map E, with operational elements {Ej} such that
E ≡ † (ρ) EjρEj . (1.77) j
Summarised below are common channels used to represent noisy quantum processes or errors.
51 CHAPTER 1
Bit-flip channel
The bit-flip channel represents operations ρ →EX (ρ), which makes the trans- formation between |↓ ↔ |↑ states with an error probability p as represented in Fig. 1.6a.
a) b) z
1-p ۄͳ| ۄͳ| p y p x ۄp |Ͳ-1 ۄͲ|
Figure 1.6: a) Representation of the bit-flip channel EX where bit-flip errors occur with probability p. b) Deformed surface illustrating the passage of all pure states through the bit-flip channel. Thex ˆ components of each state are preserved, while they ˆ andz ˆ components are reduced by 1 − 2p
The operational elements of this channel are given by
√ E0 = p 1,E1 = 1 − pσxˆ. (1.78)
52 CHAPTER 1
Phase-flip channel
The phase flip channel is represented by the operation ρ →EZ (ρ) where the operational elements of EZ are given by
√ E0 = p 1,E1 = 1 − pσzˆ. (1.79)
The effect of this channel is illustrated in Fig. 1.7a where thez ˆ components a) b) z z
y y x x
Figure 1.7: Surfaces on the Bloch sphere representing all pure states passing through the a) EZ phase-flip channel and b) EY bit-phase-flip channel. of each state are preserved, while thex ˆ andy ˆ components are reduced by a factor of 1 − 2p.
53 CHAPTER 1
Bit-phase-flip channel
The bit-phase-flip channel EY is illustrated in Fig. 1.7c and is the combination of both a bit-flip and phase-flip error occurring on the system simultaneously. The operational elements are
√ E0 = p 1,E1 = 1 − pσyˆ. (1.80)
Depolarising channel
The depolarising channel ρ →E(ρ) represents a channel which replaces the input state ρ with a completely mixed (or depolarised) state 1/2 with prob- ability p. The channel is described by
p1 E(ρ)= +(1− p)ρ. (1.81) 2
The effect on the set of pure states is illustrated in Fig. 1.8. Using the fact that 1 1 = (ρ + σ ρσ + σ ρσ + σ ρσ ) , (1.82) 2 4 xˆ xˆ yˆ yˆ zˆ zˆ the depolarising channel can be re-written as
3p p E(ρ)= 1 − ρ + (σ ρσ + σ ρσ + σ ρσ ) . (1.83) 4 4 xˆ xˆ yˆ yˆ zˆ zˆ
54 CHAPTER 1
z
y x
Figure 1.8: Surfaces on the Bloch sphere representing all pure states passing through the E depolarising channel.
Therefore, the operational elements of the channel are given by
E0 = 1 − 3p/4 1, (1.84) √ E1 = pσxˆ/2, (1.85) √ E2 = pσyˆ/2, (1.86) √ E3 = pσzˆ/2, (1.87) demonstrating that the depolarising channel combines the action of all three previous channels equally together with a combined error probability of p. This depolarising channel is an important tool used to model qubit behaviour in the presence of a noisy environment, and will be re-visited again in the discussion of the Randomised Benchmarking experiment.
55 CHAPTER 1
1.6 Experiments for benchmarking qubit
performance
In this section two different families of experiments which aim to bench- mark qubit performance are introduced. The tomographic family (including state and process tomography) are considered very useful for analysing the performance of qubit systems with the capacity to assist in error diagno- sis. However, these tomographic experiments do not scale efficiently with increasing numbers of qubits, requiring more and more experimental over- head in order to extract all required data. These experiments are also unable to isolate qubit control errors from for state-preparation-and-measurement (SPAM) errors. On the other hand, the Randomised Benchmarking family of experiments is robust against SPAM errors and the experiment has been shown to scale efficiently with increasing numbers of qubits. The experiment is useful in the regard that the extracted information and experimental output is expressly the qubit control fidelity as a singular unit of merit. However, the experiment often suffers from strict assumptions being made on input noise processes and from having arguably less capacity to diagnose system error processes for self-improvement of qubit control fidelity.
56 CHAPTER 1
1.6.1 Tomographic experiments
The two varieties of tomographic experiments attempt to characterise the quantum system by determining elements of the density matrix ρ or an arbi- trary quantum process E. The description of these two experiments presented below follows the summarised discussion of Ref. [90].
State Tomography
In order to determine the overall state of a prepared quantum system, mul- tiple measurements of the same prepared state are performed to determine the individual elements of ρ [96, 97, 98] ⎛ ⎞
⎜ρ11 ρ01⎟ ρ = ⎝ ⎠ . (1.88) ρ10 ρ00
The repeated measurements of the same state are performed with respect to various measurement bases until all elements of ρ can be extracted from a set of linear equations. Measurement in the standard qubit basis of {|↓ , |↑} yields ρ00,withρ11 =1− ρ00. For the off-diagonal elements of ρ, it is often more convenient to make a basis transformation to rotate ρ with respect to the standard measurement basis, rather than performing the measurements in another basis. Therefore, transforming ρ using a 90◦ rotation around thex ˆ-axis and measuring in the standard basis yields Im(ρ10)=−Im(ρ01). Further, a similar measurement after the same rotation around they ˆ-axis
57 CHAPTER 1
will yield Re(ρ10)=Re(ρ01), allowing for the full density matrix ρ to be determined. The expansion into state tomography of multiple qubits results in the production of linear equations from which the density matrix can still be determined, the details of which are summarised in Ref. [90].
Process Tomography
Process tomography is an experiment which builds upon state tomography, but is instead used to determine some process E which acts upon an arbitrary quantum state ρ [99, 100, 101, 102]. The overall procedure is to determine the output state of E for a set of input states which form a basis for the system Hilbert space, and then compute the entire transfer function from the finite set of input-output pairs. The transformation of the arbitrary quantum state ρ by E is the linear map E(ρ) ρ → , (1.89) Tr(E(ρ)) where E(ρ) can be expressed in operator-sum representation [103, 15]
E † (ρ)= AjρAj. (1.90) j
This expansion is not unique, but a fixed set of operators Aj whichforma basis for the set of operators on the state space can always be found such
58 CHAPTER 1 that
E † (ρ)= χpqApρAq. (1.91) p,q
Because Aj are fixed, the Hermitian operator χ completely describes E. Determining χ for a system of n qubits requires a basis of 4n linearly independent density matrices ρk which span the Hilbert space. A set of linear equations of the form presented in Eq. (1.91) can then be constructed from measuring E(ρk) for all k, allowing χpq to be determined.
1.6.2 Randomised Benchmarking
Acknowledgement: This introduction to the Randomised Benchmarking (RB) experiment was largely influenced by the introduction presented in Ref. [58]. The discussions and results presented therein were produced in collaboration with Dr. M. Veld- horst, Dr. R. Harper, Dr. C. H. Yang, Dr. S. D. Bartlett, Dr. S. T. Flammia, and Dr. A. S. Dzurak. The full acknowledgements for this manuscript can be found in Appendix A.
This experiment will be elaborated upon in Chapter. 3, but will be dis- cussed briefly for comparison with the tomographic experiments. The RB experiments have the useful advantage of being robust against these SPAM errors, but pay the price of being subject to a near over-constraining set of
59 CHAPTER 1 assumptions to be able to extract meaningful results. The standard random-
Figure 1.9: Randomised benchmarking consists of applying multiple se- quences of random Clifford gates, a final recovery Clifford gate to ensure that each sequence ends with the qubit in an eigenstate, and reading out the qubit state. In interleaved randomized benchmarking, an additional test Clif- ford gate is inserted in between the random Clifford gates. Figure is sourced from Ref. [58]. ized benchmarking procedure involves subjecting a quantum system to long sequences of randomly sampled Clifford gates followed by an inversion step and a measurement, as depicted in Fig. 1.9. The unitary operations of the Clifford group G are those that map the set of Pauli operators to itself under conjugation. They are a discrete set of gates that exactly reproduce the uni- form average gate fidelity, averaged over the set of all input pure states [104]. An alternate version known as interleaved benchmarking [105] inserts a sys- tematic application of a given gate, such as the H gate shown in Fig. 1.9. The difference from the reference sequence gives information about the spe- cific average gate fidelity of the given gate, rather than the average fidelity additionally averaged over the ensemble of gates.
60 CHAPTER 1
z
y D
CN x CI
Figure 1.10: Bloch sphere representation for the breakdown of a general noisy operation CN into an ideal CI rotation followed by a noise operation D. Figure is sourced from Ref. [58].
Consider a general noise process D, depicted in Fig. 1.10, which repre- sents the deviation of a noisy Clifford gate CN from an ideal unitary Clifford operation CI :
CN = DCI . (1.92)
The above equation uses the formalism of completely positive maps [15], and the multiplication corresponds to composition of maps. The standard ap- proach to randomised benchmarking makes the assumption that D does not depend on the choice of CI or other details such as time, but our simulations and of course real experiments will include such a dependence. The fundamental result of randomised benchmarking [106] is that for suf- ficiently well-behaved noise the observed fidelities only depend on the average
61 CHAPTER 1
error operation ED averaged over the Clifford group G given by
1 −1 ED = C DC , (1.93) |G| I I CI ∈G as well as any SPAM errors present in the system. Furthermore, standard tools from representation theory reduce this average error operation to one that is nearly independent of D, and is characterised by just a single param- eter p. In particular, it is a depolarising channel E with p = p(D) being the polarisation parameter (i.e., the probability of the information remaining un- corrupted as it passes through the channel). For a d-dimensional quantum system, the action of the depolarising channel is given by
1 E(ρ)=pρ +(1− p) , (1.94) d and the polarisation parameter is related to the noisy deviation D by the ¯ average gate fidelity Favg(D) according to [106]