ORGANIZED BY
JUNE 20TH
2019
French industrial quantum use cases
Henri CALANDRA
Expert in Numerical algorithms and High Performance Computing for Geosciences at Total E&P
Scientific Advisor for Total Corporate Research, France
Quantum Computing @ TOTAL
Henri Calandra, TOTAL E&P
3
Why explore Quantum computing
technology?
EXPLORE QUANTUM COMPUTING TECHNOLOGY
WHAT HAVE WE DONE SO FAR…
CONCLUSION
4
Why explore Quantum computing technology?
Toward more and more complex classical HPC
systems
5
✓ Classical computers have fundamental limits:
▪ Transistor scaling
▪ Energy consumption
✓ HPC systems are likely to become much more heterogeneous and
massively-parallel systems
▪ Parallelism limitations: Adhams’law
✓ End of Moore’s law is expected by around 2025 !!
Compute requirements continue to grow
6
✓ Seismic depth imaging:
✓ compute better, faster
✓ More physics
✓ More data integration
✓ Uncertainty quantification
✓ Reservoir simulation:
✓ Compute faster
✓ Better predictability
✓ Multi real time simulations
✓ Inversion of subsurface models
✓ operations optimization,
✓ cost and risk reduction
✓ More complex targets:
✓ strong geological heterogeneity – several reservoirs
✓ Massive simulations:
✓ history matches; uncertainty Mgt on huge models
✓ New physics for EOR & integration of different processes incl. geomechanics
An there’s more we want to do
7
Machine learning, HPDA
✓ Development of new training set
and algorithms
✓ Classification and sampling of large
dataset
✓ Physics-constrained neural nets…
A
B
Liquid
Vapor
Combinatorial optimization
✓ MINLP (Mixed Integer Non Linear
programming) problems in general
including:
✓ Refinery blending,
✓ Scheduling, production, shipping.
✓ Oil field/reservoir optimization
✓ ….
Computational material science
✓ The ability to accurately model
ground states of fermionic systems
would have significant implications
for many areas of chemistry and
materials science:
✓ Catalysis, Solvents, Lubricants,
batteries…
An there’s more we want to do
8
Machine learning, HPDA
✓ Development of new training set
and algorithms
✓ Classification and sampling of large
dataset
✓ Physics-constrained neural nets…
A
B
Liquid
Vapor
Combinatorial optimization
✓ MNILP problems in general including:
✓ Refinery blending,
✓ Scheduling, production, shipping.
✓ Oil field/reservoir optimization
✓ ….
Computational material science
✓ The ability to accurately model
ground states of fermionic systems
would have significant implications
for many areas of chemistry and
materials science:
✓ Catalysis, Solvents, Lubricants,
batteries…
✓ Compute better, compute faster
✓ Solve intractable problems on classical High Performance Computing
Explore Quantum computing as a disruptive technology ?
9
explore Quantum computing technology
1
0
QUANTUM HARDWARE DEVELOPMENT IS ACCELERATING
✓ A groundbreaking new approach to pave the way for the future of scientific computing
beyond exascale ?
A very challenging technology
✓Quantum Hardware comes in many forms (supra-conductors, ion trap, photonics…)
✓ Still limited to few ten’s of qubits NISQ (Noisy Intermediate scale Quantum device)
From Denis Vion, CEA, 41th Orap Forum
IBM
Google
Rigetti
IONQ
XANADU
OBJECTIVEs
12
✓ Understand Quantum Computers technology evolution
D’WAVE computer
IBM Q
Google: bristlecone Rigetti : 16Q Aspen
✓ Accelerate and build in-house competencies skill set with research
partners and hardware providers ecosystem to develop algorithms for
Total business use cases
✓ Be ready when industrial quantum computers become available and quantum supremacy is
demonstrated.
✓ Quantum computing is a huge paradigm shift and Quantum algorithmics is a brand new science
13
Anticipated Impact ( what we expect)
✓ Compute better, compute faster
✓Open new frontiers in R&D for Chemistry, material science, optimization, machine learning,….
Quantum linear algebra
(qublas), solving ODEs. PDEs,
inverse problems…
Quantum
combinatorial
optimization
Quantum machine
learning
Quantum Chemistry
14
What have we done so Far…
Define an appropriate R&D roadmap and potential
Use CASES
Quantum combinatorial
optimization
Quantum machine learningQuantum Chemistry
NISQ device ~ (102 qubits) 3-5 years (pre)-QEC device ~ (103(-6) qubits) 5+ (++?) years
Quantum linear algebra
(qublas), solving ODEs. PDEs,
inverse problems…
15
Quantum Computing at Total
Q Hardware Q Software Applications
Math libraries
ATOS QLM - emulator
Verification: simulation, benchmarking, testing
Chemistry, Material
Science
Optimization, Machine
Learning
Hybrid (QC-HPC)
Gate-based (IBM, Rigetti,
google…)
Annealer ( D-Wave,….)
Programming models
Hybrid computing ODEs, PDEs, linear algebra,
inverse problem…
Quantum computing global program overview
16
17
academic and industrial collaborations network
Collaboration agreement, purchase of a 30 and 35 qubits QLM
ATOS QLM-30 system, installed September 2018,
ATOS QLM 35 Qubits system upgrade in progress
Quantum Machine learning
for industrial applications.
Ph.D. 07/01/2019-
06/30/2024
Quantum Derivative free
gradient optimization for
inverse problems
Ph.D. 10/01/2019-09/30/2024
Jülich Unified Infrastructure for Quantum computing (JUNIQ)
2 months internship on quantum micro benchmark
Quantum Gibbs sampling on
NISQ devices. 2 years PostDoc,
09/01/2019-08/31/2021
EU-Project (FETFLAG-QUANTUM): PasQuans
Quantum Seismic Imaging, 2 years postdoc,
07/01/2019-06/30/2021
Qualitative Computing for Quantum Computing,
2 years postdoc, 07/01/2019-06/30/2019
Construction of a QBLAS library, Ph.D. 10/01/2019-09/30/2022
18
Get one’s hands dirty in Programming models and
algorithm design✓ Examples of on going dev:
✓ NISQ oriented applications development on ATOS QLM : VQE and QAOA
VQE
QAOA
✓ Understand existing implementations: QISKIT (IBM), Pyquil (Rigetti)….
✓ Understand actual limitations (HW and algorithm) and possible test cases.
✓ « hybrid » programming , mixing quantum speedup procedures and classical ones
conclusion
✓ Quantum computing is a huge paradigm shift and Quantum
algorithmics is a brand new science
19
GROVER ALGORITHM
✓ Quantum computing can provide new opportunities,
opening new frontiers in R&D in many fields of application.
✓ Set up a good academic, partners and industrial network
based on an appropriate research program
✓ Build internal skills to develop algorithms for Total business use case
✓ Quantum Computing for Industrial Applications Project:
✓ 2018: project design, collaboration setup , purchase of a QLM30 qubits
✓ 2019:
✓ 2(3) Total researchers,
✓ 3 Postdocs and 3 PhDs,
✓ Upgrade to a QLM 35 qubits
✓ Develop internal skills on NISQ devices
French industrial quantum use cases: Total

French industrial quantum use cases: Total

  • 1.
    ORGANIZED BY JUNE 20TH 2019 Frenchindustrial quantum use cases Henri CALANDRA Expert in Numerical algorithms and High Performance Computing for Geosciences at Total E&P Scientific Advisor for Total Corporate Research, France
  • 2.
    Quantum Computing @TOTAL Henri Calandra, TOTAL E&P
  • 3.
    3 Why explore Quantumcomputing technology? EXPLORE QUANTUM COMPUTING TECHNOLOGY WHAT HAVE WE DONE SO FAR… CONCLUSION
  • 4.
    4 Why explore Quantumcomputing technology?
  • 5.
    Toward more andmore complex classical HPC systems 5 ✓ Classical computers have fundamental limits: ▪ Transistor scaling ▪ Energy consumption ✓ HPC systems are likely to become much more heterogeneous and massively-parallel systems ▪ Parallelism limitations: Adhams’law ✓ End of Moore’s law is expected by around 2025 !!
  • 6.
    Compute requirements continueto grow 6 ✓ Seismic depth imaging: ✓ compute better, faster ✓ More physics ✓ More data integration ✓ Uncertainty quantification ✓ Reservoir simulation: ✓ Compute faster ✓ Better predictability ✓ Multi real time simulations ✓ Inversion of subsurface models ✓ operations optimization, ✓ cost and risk reduction ✓ More complex targets: ✓ strong geological heterogeneity – several reservoirs ✓ Massive simulations: ✓ history matches; uncertainty Mgt on huge models ✓ New physics for EOR & integration of different processes incl. geomechanics
  • 7.
    An there’s morewe want to do 7 Machine learning, HPDA ✓ Development of new training set and algorithms ✓ Classification and sampling of large dataset ✓ Physics-constrained neural nets… A B Liquid Vapor Combinatorial optimization ✓ MINLP (Mixed Integer Non Linear programming) problems in general including: ✓ Refinery blending, ✓ Scheduling, production, shipping. ✓ Oil field/reservoir optimization ✓ …. Computational material science ✓ The ability to accurately model ground states of fermionic systems would have significant implications for many areas of chemistry and materials science: ✓ Catalysis, Solvents, Lubricants, batteries…
  • 8.
    An there’s morewe want to do 8 Machine learning, HPDA ✓ Development of new training set and algorithms ✓ Classification and sampling of large dataset ✓ Physics-constrained neural nets… A B Liquid Vapor Combinatorial optimization ✓ MNILP problems in general including: ✓ Refinery blending, ✓ Scheduling, production, shipping. ✓ Oil field/reservoir optimization ✓ …. Computational material science ✓ The ability to accurately model ground states of fermionic systems would have significant implications for many areas of chemistry and materials science: ✓ Catalysis, Solvents, Lubricants, batteries… ✓ Compute better, compute faster ✓ Solve intractable problems on classical High Performance Computing Explore Quantum computing as a disruptive technology ?
  • 9.
  • 10.
    1 0 QUANTUM HARDWARE DEVELOPMENTIS ACCELERATING ✓ A groundbreaking new approach to pave the way for the future of scientific computing beyond exascale ?
  • 11.
    A very challengingtechnology ✓Quantum Hardware comes in many forms (supra-conductors, ion trap, photonics…) ✓ Still limited to few ten’s of qubits NISQ (Noisy Intermediate scale Quantum device) From Denis Vion, CEA, 41th Orap Forum IBM Google Rigetti IONQ XANADU
  • 12.
    OBJECTIVEs 12 ✓ Understand QuantumComputers technology evolution D’WAVE computer IBM Q Google: bristlecone Rigetti : 16Q Aspen ✓ Accelerate and build in-house competencies skill set with research partners and hardware providers ecosystem to develop algorithms for Total business use cases ✓ Be ready when industrial quantum computers become available and quantum supremacy is demonstrated. ✓ Quantum computing is a huge paradigm shift and Quantum algorithmics is a brand new science
  • 13.
    13 Anticipated Impact (what we expect) ✓ Compute better, compute faster ✓Open new frontiers in R&D for Chemistry, material science, optimization, machine learning,…. Quantum linear algebra (qublas), solving ODEs. PDEs, inverse problems… Quantum combinatorial optimization Quantum machine learning Quantum Chemistry
  • 14.
    14 What have wedone so Far…
  • 15.
    Define an appropriateR&D roadmap and potential Use CASES Quantum combinatorial optimization Quantum machine learningQuantum Chemistry NISQ device ~ (102 qubits) 3-5 years (pre)-QEC device ~ (103(-6) qubits) 5+ (++?) years Quantum linear algebra (qublas), solving ODEs. PDEs, inverse problems… 15
  • 16.
    Quantum Computing atTotal Q Hardware Q Software Applications Math libraries ATOS QLM - emulator Verification: simulation, benchmarking, testing Chemistry, Material Science Optimization, Machine Learning Hybrid (QC-HPC) Gate-based (IBM, Rigetti, google…) Annealer ( D-Wave,….) Programming models Hybrid computing ODEs, PDEs, linear algebra, inverse problem… Quantum computing global program overview 16
  • 17.
    17 academic and industrialcollaborations network Collaboration agreement, purchase of a 30 and 35 qubits QLM ATOS QLM-30 system, installed September 2018, ATOS QLM 35 Qubits system upgrade in progress Quantum Machine learning for industrial applications. Ph.D. 07/01/2019- 06/30/2024 Quantum Derivative free gradient optimization for inverse problems Ph.D. 10/01/2019-09/30/2024 Jülich Unified Infrastructure for Quantum computing (JUNIQ) 2 months internship on quantum micro benchmark Quantum Gibbs sampling on NISQ devices. 2 years PostDoc, 09/01/2019-08/31/2021 EU-Project (FETFLAG-QUANTUM): PasQuans Quantum Seismic Imaging, 2 years postdoc, 07/01/2019-06/30/2021 Qualitative Computing for Quantum Computing, 2 years postdoc, 07/01/2019-06/30/2019 Construction of a QBLAS library, Ph.D. 10/01/2019-09/30/2022
  • 18.
    18 Get one’s handsdirty in Programming models and algorithm design✓ Examples of on going dev: ✓ NISQ oriented applications development on ATOS QLM : VQE and QAOA VQE QAOA ✓ Understand existing implementations: QISKIT (IBM), Pyquil (Rigetti)…. ✓ Understand actual limitations (HW and algorithm) and possible test cases. ✓ « hybrid » programming , mixing quantum speedup procedures and classical ones
  • 19.
    conclusion ✓ Quantum computingis a huge paradigm shift and Quantum algorithmics is a brand new science 19 GROVER ALGORITHM ✓ Quantum computing can provide new opportunities, opening new frontiers in R&D in many fields of application. ✓ Set up a good academic, partners and industrial network based on an appropriate research program ✓ Build internal skills to develop algorithms for Total business use case ✓ Quantum Computing for Industrial Applications Project: ✓ 2018: project design, collaboration setup , purchase of a QLM30 qubits ✓ 2019: ✓ 2(3) Total researchers, ✓ 3 Postdocs and 3 PhDs, ✓ Upgrade to a QLM 35 qubits ✓ Develop internal skills on NISQ devices