《GESDA:2024迈向具有积极可持续性影响的再生量子计算白皮书(英文版)(72页).pdf》由会员分享,可在线阅读,更多相关《GESDA:2024迈向具有积极可持续性影响的再生量子计算白皮书(英文版)(72页).pdf(72页珍藏版)》请在三个皮匠报告上搜索。
1、1Towards Regenerative Quantum Computing with proven positive sustainability impact White paper Towards Regenerative Quantum Computing with proven positive sustainability impact04EDITORIALGoal of the documentCHAPTER 01The general context:IT,scientific computing,and sustainability challenges1.1 Inform
2、ation Technologies role in help meet SDG1.2 Classical supercomputers footprint1.3 Addressing SDG challenges and lowering the energy footprint with quantum computersCHAPTER 02The Blaise Pascal quantum challenge approach for assessing the quantum sustainability advantage2.1 What is a use case with imp
3、act?2.2 A lifecycle approach to assess the energy/carbon footprint of quantum vs.classical computing2.3 Energy/carbon footprint of classical HPC centersCHAPTER 03Lessons drawn from the Blaise Pascal Quantum Challenge:use-cases and maturityCHAPTER 04Energy/carbon footprint of PASQAL technology4.1 PAS
4、QAL neutral atom quantum processing unit energy cost4.2 PASQAL quantum processing unit carbon footprint and benchmark with classical HPCCHAPTER 05Perspectives on sustainable quantum advantage by the research and institutional communities5.1 A Perspective by the quantum energy research community5.2 F
5、ostering key sustainable use cases:the approach of the Open Quantum Institute061018263035CONTENTS05CHAPTER 06The perspective of key High-Performance Computing centers on quantum contribution to decarbonizing HPC6.1 A French perspective by GENCI:France hybrid HPC quantum initiative(HQI)5.2 Jlich pers
6、pective on HPC integrating quantum computing and energy advantage CHAPTER 07The perspective of a key industrial end-user:EDF,the leading French energy utility CONCLUSION An agenda to move forward in search of sustainable quantum advantage REFERENCES APPENDIX The Blaise Pascal Regenerative Quantum Ch
7、allenge1.The hackathon scoring process as an effort to standardize sustainable quantum advantage 2.Overview of the spirit of the regenerative quantum challenge425256586406EDITORIAL07Quantum computing promises to exploit quantum superposition and entanglement to efficiently address complex problems t
8、hat are usually inaccessible or significantly resource-consuming to classical computing.This gain in performance can manifest itself in various ways,such as acceleration in terms of execution speed,enhancement in terms of the accuracy and the quality of the solution.However,one may look at this incr
9、eased efficiency from a completely different angle:the energy/sustainability advantage perspective.Indeed,for specific use cases,the gain is expected to emerge in the form of lesser energy requirements(hence,lesser carbon emissions)for existing computational tasks or for solving previously intractab
10、le tasks.This point of view is a less traveled road for now,as we do not know yet which of the many potential quantum technologies and use cases would bring a true energy/carbon emission advantage.Therefore,the goal of this document is to establish a joint statement of interest and raise awareness i
11、n the academic,research and development,industrial,and end-user communities to call for action,to encourage experimentation and investigation of this emerging technology,its impact,and its potential use cases from an energy footprint perspective.PASQAL co-founded with Blaise Pascal Advisors and a co
12、nsortium of partners,including Genci/HQI,QEI,EDF R&D and Jlich Forschungszentrum,an international initiative to seed proofs for the impact of quantum computing as well as to foster that the industry takes its share in complying with the United Nations sustainable development goals.This initiative in
13、cluded a hackathon launched in the summer of 2023the Blaise Pascal regenerative quantum challengefollowed by the present whitepaper and numerous associated actions meant to promote meaningful and sustainable uses of quantum computing.We pursue a deliberate attempt to bring forward positive contribut
14、ions in the context of a sometimes controversial debate on the impact of technology,including but not restricted to climate change.These endeavors require clear criteria,accountable figures of life-cycle analyses of a Quantum Processing Unit(QPU),and a sound comparative basis to classical computing.
15、GOAL OF THE DOCUMENT08Documenting the hackathon experience as an avant-garde experiment to investigate the potential of sustainable quantum computing.THE PRESENT DOCUMENT AIMS AT :Setting reference elements for a life-cycle and use-case-based approach of sustainable quantum advantage.Preparing futur
16、e standards for regenerative and impactful quantum computing using neutral atoms technology.Outlining later research and communications agenda.0910CHAPITRE 0111THE GENERAL CONTEXT :IT,SCIENTIFIC COMPUTING AND SUSTAINABILITYCHALLENGESIn 2017,The United Nations General Assembly adopted the 2030 Agenda
17、 for Sustainable Development Goals(SDG).This agenda is the result of many decades of global efforts to promote human rights and sustainable development for all countries and to find common ground to create an effective plan of action to strengthen universal peace and freedom and eradicate poverty in
18、 all its forms,balanced in three main dimensions:economic,social,and environmental(UN Resolution,2017).Comprised of 17 goals,each plan of action has a specific target that should be achieved by 2030,such as zero hunger,achieving affordable healthcare and medicine for all,limiting global warming to 1
19、.5 C above pre-industrial levels,and reducing the greenhouse emissions in 43%,aiming at net zero by 2050 (Dpi campaign,2017).But to be successful and accomplish these goals requires the attention and compromise of all actors:society,governments,and enterprises.We are in 2023,the clock is ticking and
20、 we are still facing enormous challenges worldwide,while some improvements suffered a setback due to the COVID-19 pandemic.In fact,the UN SDG 2023 report states that“slow and uneven progress on poverty reduction may leave hundreds of millions in extreme poverty by 2030.”(UN-DESA,2023).Here,we extrac
21、t some figures from the UNs latest report as examples that show the dimension of humanitys sustainability challenges.SDG2 target:zero hunger.In 2022,between 691 and 783 million people faced hunger worldwide(UN News,2023b,July 13),and around 2.3 billion people suffered from moderately or severely foo
22、d insecurity(WHO,2022,July 6).SDG3 target:access to safe,effective,quality,and affordable essential healthcare,medicines,and vaccines for all.Although effective HIV treatment has cut global AIDS-related deaths by 52%since 2010 and at least one neglected tropical disease has been eliminated in 47 cou
23、ntries.AIDS,tuberculosis,and malaria still have epidemic proportions in developing countries.Around 2 billion people globally have no access Written by Alexandra de Castro,science and technology communicator at PASQAL 12to essential medicines,particularly in lower-and middle-income countries(Chattu
24、et al.,2023).SDG13 target:zero emissions and limit the global warming to 1.5 C:The impact of climate change is increasing in intensity and frequency.Since the last decade,most regions have experienced more and more extreme weather events than in the past such as floods,wildfires,and storms.The globa
25、l sea level rise is accelerating:it has more than doubled,from 1.4 millimeters per year throughout most of the twentieth century to 3.6 millimeters per year from 20062015.In 2022,the global average sea level set a new record high101.2 mm above 1993 levels,24 centimeters since 1880.Currently,2.15 bil
26、lion people live in the near-coastal zone and 898 million in the low-elevation coastal zone globally(Reimann et al.,2023).Being SDG 13 only one in 17 goals,climate action is key to help improve in all other sustainability targets.Climate change can undo the progress made over the years.For these rea
27、sons,society,governments,and enterprises must ensure limiting global temperature rise to well below 2C,and according to the US National and Atmospheric Administration we are facing the warmest years in the historical record since 2010(Climate Change:Global Temperature,2023b).During the hackathon The
28、 Blaise Pascal regenerative quantum challengelaunched this year in October,PASQAL proposed six sustainability challenges related to ten SDGs that have the best chance Fig.1.United Nations Sustainable Development Goals(SDG)-Source https:/sdgs.un.org/goals13to be addressed using neutral atom technolog
29、y,effectively reducing the carbon footprint in comparison with classical methods.They are:Sustainable Agriculture(included in zero hunger SDG2).Drug Discovery(included in access to safe,effective,quality,and affordable essential healthcare,medicines,and vaccines for all SDG3).Smart cities(included i
30、n clean water and sanitation-SDG 6&11).Smart grids and Affordable/Clean Energy(SDG 7).Sustainable transport,industry and circular economy.(SDG 9&12).Environment,Climate,and Biodiversity(SDG 13,14&15).1.1 Information Technologies role in help meet SDGThe UN also recognizes the crucial role that Infor
31、mation Technologies(IT)play in achieving SD goals.For instance,digital documents with digital signatures to enhance the good use of natural resources or online meetings to decrease the transportation carbon footprint.Within IT is essential the access and smart use of supercomputers,also called data
32、centers or high performance computers(HPC).Scientists have been using powerful supercomputers to store and process data related to the sustainability challenges.For example,Addressing food security(SDG2)through effective models in food supply optimization(Angarita-Zapata et al.,2021).These computati
33、onal approaches hold enormous potential to help manage supply chains more efficiently and sustainably.For healthcare and medicines security,related to SDG3,we have that computer-aid drug discovery has significantly reduced the time and expenditure of drug development(Ece,2023).Although these methods
34、 face some limitations,in some cases,computational techniques can provide a better view of the structure and molecular dynamics of the biological system than experimental setups(DArcangelo et al.,2023).Addressing climate impact and targets(SDG13),scientists have been collecting enormous amounts of d
35、ata from many sources,such as atmospheric,geochemistry,and ecology,and simulating the intricate patterns and modeling the future implications of climate change.However,the data collection,storage,and process have a drawback:supercomputing centers consume vast amounts of energy!Since the last century
36、,high-performance computers have been growing,offering more precise and faster calculations,and today,they accommodate thousands of processor cores requiring entire buildings with costly cooling systems.Rethinking energy consumption and reducing greenhouse emissions has become an essential mission f
37、or enterprises offering computational and data storage services.However,despite the efforts to lower energy consumption it is still extremely high.Therefore,real,transformative strategies and actions are crucial,beyond mere plans and promises.141.2 Classical supercomputers footprint In 2020,IT repre
38、sented 11%of the global electricity consumption(Puebla et al.,2020).An analysis by Huawei Technologies shows that it can increase to 51%in 2050(Andrae&Edler,2015),that this electricity usage could contribute up to 23%of the globally released greenhouse gas emissions in 2030,including,but not limited
39、,to data center computation and storage(also called HPC centers or supercomputers).Supercomputers can consume the same amount of electricity as a town.For instance,the Frontier supercomputer,fabricated by Hewlett Packard and hosted at the DoE Oak Ridge Laboratory in Tennessee,USA,uses 504 MWh on ave
40、rage daily,summing up the energy consumed by around 17 thousand average homes in the U.S.daily(EIA,2023).And this is only one data center.Regardless of these numbers being too high,they are actually a success in efficiency(Koomey,2023).Digital engineers have created smaller and more efficient transi
41、stors,improved the circuits,the software,and the power-management schemes.But despite these tremendous improvements,the workload has also increased so that more enterprises need more and larger data centers with an annual energy consumption growth of 20-40%(Data centres&networks,2023).1.3 Addressing
42、 SDG challenges and lowering the energy footprint with quantum computersQuantum computing is rapidly emerging to play a key role in the next generation of high performance computing to address complex problems,inaccessible to traditional devices(see Figure 1 by Ezratty,2023).They are called inaccess
43、ible or intractable because their computation time increases exponentially with their size(Ezratty,2023).The good news is that most of these problems,that have industrial and scientific relevance,exhibit a natural encoding onto quantum physics,which is the basics of quantum computing.Common examples
44、 are optimization problems,such as food chain supply optimization(FSO),which encompasses all activities,organizations,actors,technologies,information,resources,a nd services involved in producing agri-food products for consumer markets.The upstream and downstream of FSO comprises sectors from the su
45、pply of agricultural inputs,such as seeds,fertilizers,feed,medicine,or equipment,to production,post-harvest handling,processing,transportation,marketing,distribution,and retailing(Angarita-Zapata et al.,2021).The Food Supply Optimization represents a complex challenge in finding the balance between
46、productive efficiency and sustainability of food supply systems.Other typical examples of intractable problems are chemistry simulations at the molecular level,which are particularly crucial in healthcare,such as drug design(Varsamopoulos et al.,2022)or toxicity prediction(Albrecht et al.,2023).15Th
47、e promise of quantum technologies to tackle such intractable problems in a human scale time period is called quantum advantage(Ezratty,2023).With its promising ability to solve complex problems,quantum computing can provide innovative solutions to address the most pressing humanity and environmental
48、 challenges.By combining quantum computing with artificial intelligence and big data analytics,quantum technologies can exploit potential synergies to accelerate environmental innovation.This will identify integrated solutions,such as smart grids,sustainable agricultural practices,and circular econo
49、mic mechanisms,to maximize positive environmental impact.Will quantum advantage arrive with energy consumption advantage?We still dont know,but what we can say today is that current quantum computers electricity usage is orders of magnitude much less than any supercomputer,counting all the different
50、 quantum architectures available.Being superconducting qubits the most expensive architecture,they only consume about 25 kW(Ezratty,2023).That amounts to 600 kWh daily,a thousand times less than the Frontier supercomputer.Much less is the consumption of neutral atom quantum devices,such as PASQALs,w
51、hich amount up to 3 kW.But again,no current quantum computing can address all the problems that Frontiers is able to tackle.Although quantum computers have proven superiority over classical computers in tackling particular scientific problems(See Scholl et al.,2021;Chen et al.,2023),they are not yet
52、 ready to solve real-world complex problems,since current quantum computers are noisy,and fault-tolerant quantum computers wont be available for a while.However,because some quantum algorithms are designed to successfully work within the so-called Noisy Intermediate Scale Quantum(NISQ)era,certain ar
53、chitectures,such as PASQAL neutral atom devices hold the potential to tackle many industry-level use cases before the fault tolerance era.In this short term scenario,there PROBLEM SIZEExtremelyunreasonable time like the age of the UniverseTheoretical quantumcomputing speedupCLASSICAL computing(now a
54、nd soon)QUANTUM computing(some day)Solving intractable/exponential problems in reasonable timeReasonable human time depending on the use caseCOMPUTING TIME13,8 billion yearsFig.2.Illustrating the quantum advantage-(CC)Olivier Ezratty,202316is room for real energy quantum advantage.Another point to b
55、e considered is that quantum technologies are not meant to replace classical CPUs or GPUs but to be in synergy with them.Therefore,the hope is that quantum computers effectively lower the carbon footprint for hybrid workflows.There are still many questions that need to be addressed,for example,if th
56、is potential energy advantage of quantum computing will hold for all kinds of algorithms and applications?Is the fabrication process green enough for computers in both cases,classical supercomputers and quantum computers?Scientists and engineers need supercomputers for data collection,storage,and co
57、mputations to create more knowledge,manufacture better products,and solve the most pressing problems of our times,such as food security,drugs discovery and accessibility,and the climate crisis.But to tackle the sustainability problems,enterprises offering computing services,whether classical or quan
58、tum,should assume the compromise to lower their carbon footprint,and if possible to reach zero emission,while helping humanity to reach our global sustainable development goals.17CHAPITRE 021819THE BLAISE PASCAL QUANTUM CHALLENGE APPROACH FOR ASSESSING THE QUANTUM SUSTAINABILITY ADVANTAGETo celebrat
59、e the 400th birth anniversary of the renowned French mathematician,engineer,philosopher,physicist,and philanthropic entrepreneur Blaise Pascal,we have launched the hackathon“Blaise Pascal Regenerative Quantum challenge”to underscore the benefits of quantum computing for sustainable development.This
60、endeavor aims to serve in an accountable manner the UN SDGs outlined in the previous section,as well as ensuring a more sustainable energy use of computing resources.To achieve these objectives,we called for SDG-impact use cases evaluating the classical/quantum computing ratio of power consumption.H
61、owever,for a meaningful evaluation,first,we need to consider the distinction inherited from Green IT vs.IT for Green:Sustainable QPUs(Green quantum IT):an approach whereby quantum computing can solve tasks traditionally performed by classical HPC with a significantly lower energy/carbon footprint.QP
62、U for sustainability(quantum IT for Green):an approach whereby quantum computing can solve tasks that are core to applications with a positive impact on sustainability,such as better battery storage,lower car consumption,or food supply optimization),hard or impossible to tackle through classical HPC
63、.For the purpose of the hackathon,however,use cases without SDG impact1 are meaningless,as well as those impactful use cases whereby quantum computing has no present(or at least future)advantage over classical computing.A better distinction would be between:Relative sustainability advantage of QPU:U
64、se cases for sustainability whereby quantum computing shows a better energy/carbon footprint over classical HPC.Written by Etienne de Rocquigny,CEO Blaise Pascal Advisors,cofounder of the Blaise Pascal Regenerative Quantum challenge 1 Hereinafter,.Impact will stand for impact on SDGs,as subsumed in
65、for instance impact finance,impact funds,etc.20 Absolute sustainability advantage of QPU:Use cases for sustainability that are impractical(impossible or far too costly)to tackle with classical HPC.Under this criteria,we proposed two main requirements to evaluate the projects contenders of the hackat
66、hon:It must address a use case with sustainability impact,clearly framed in at least one of a selection of SDG challenges(see Appendix for details).The projects must provide evidence showing that their solution has a relative footprint quantum computing advantage over classical computing that is hig
67、her than 1,ideally infinite(a use case intractable with classical methods).Those requirements filter out projects judged irrelevant enough at the early stages and rank them until the final phase of the evaluation and selection process.The proposed solutions quantum feasibility/technical credibility
68、is considered an important part of the score:projects should lead to code that can run on existing quantum technology.2.1 What is a use case with impact?The Blaise Pascal regenerative quantum challenge is meant to promote meaningful and sustainable uses of quantum computing,as a deliberate attempt t
69、o bring forward positive contributions including but not restricted to climate change.As advocated in a number of papers(for example in Berger et al.,2021),impactful use cases involve potentially:Quantum simulation of classically-intractable quantum chemistry for drug design or material science appl
70、ied to innovations in solar panels and battery design.Quantum optimization:solving hard(incl.NP-complex)problems that are involved in operations research and graph theory that can be applied to transportation or energy network layout.Quantum based hybrid solvers for nonlinear partial differential eq
71、uations,or machine learning used in environmental technology,climate modeling,or resource allocation.It is helpful to link to existing international standards on common good objectives and ethics of tech,particularly the European,UN,and UNESCO pieces(cf infra).However,honoring Blaise Pascals philoso
72、phy and entrepreneurial spirit,somewhat skeptical of the perfection of law and justice,but fully committed to exploring the public s attempt to contribute to the common good,it was suggested to take a rather open-minded attitude than to strictly follow a list.Hence,the criteria was open to a variety
73、 of statements of purpose,freely argued by the competing teams.The teams also had to discuss how they eschewed 21potential negative impacts or even excluded use cases.Within the context of the present hackathon,the following general categories were outlined:Drug Discovery Sustainable Transport,Indus
74、try,and Circular Economy Smart Cities Smart Grids,and Affordable,Clean Energy Environment,Climate,and Biodiversity Sustainable AgricultureBeyond the choice of the category,the expected positive impact contribution must avoid too general proposals.For instance,energy-efficient AI is not enough to qua
75、lify for a fully-positive impact if nothing is said about the applications of such AI.In this sense,the teams should provide:A.A statement of purposeTeams had to argue in their own terms the purpose of the use case they chose:how does it illustrate a contribution of quantum computing to the sustaina
76、ble development challenges,such as transport,energy,or health within the UNESCO recommendation on the ethics of AI or the UN sustainable development goals.B.Trustworthiness/limitation of negative impactsTeams had also to check how the proposed use case builds a trustworthy use of computing as define
77、d for instance by,but not limited to,the European upcoming AI-Act or the UNESCO recommendation on the ethics of AI.They may have had to discuss the mitigation of potential negative impacts,such as lack of intelligibility/transparency,unfairness or endangering of vulnerable people,addictiveness of IT
78、 uses,lack of consent or breach of autonomy of human beings,lack of responsibility or human oversight,unwanted safety or security risks,and disrespect of privacy.Use cases prohibited by the European upcoming AI-Act were excluded from the contest:subliminal or purposefully manipulative techniques,exp
79、loiting peoples vulnerabilities,social scoring,remote biometric identification systems.Military applications are also excluded.2.2 A lifecycle approach to assess the energy/carbon footprint of quantum vs.classical computingIt is crucial to go beyond the pure“core run”comparison between CPU and QPU i
80、n terms of number of cycles or power requirements.For instance,the list of low energy data centers Green500 distinguishes the highest energy efficiency in Gflops/W(currently led by the Henri system at the Flatiron Institute in New York City,United States,with an energy efficiency of 65.40 Gflops/W).
81、However,the entire energy/carbon load of fabrication,as well as the sustainability goal of the use cases running on them,are not mentioned.Equally,the emerging trend in AI to compute the CO2 emission of machine learning training 22relies mostly of direct power consumption converted according to the
82、carbon content of the server location(see for instance https:/mlco2.github.io/impact/#compute)A life-cycle analysis(LCA)with due overheads associated with the fabrication of the computing machines(generally considered as amounting to more than half of the footprint of IT,depending also on the carbon
83、 content of electricity),as well as the entire cycle(pre,post,hybrid architecture)must be subject to constant review and regulation.See for instance the French 2021 REEN Act2,or MITs LCA methodology adopted by DELL3.Fabricating a supercomputer(be it classical or quantum)indeed involves a complex pro
84、cess with numerous components and considerations,each contributing to an overall carbon footprint:Design:The design of a supercomputer involves a large amount of engineering and computational cost,as well as hardware prototyping,which add to the carbon emissions during the manufacturing period.Manuf
85、acturing Materials:The production of a supercomputer involves a wide range of materials,including metals,semiconductors,lasers,cameras,and plastics.The extraction,processing,and transportation of these materials contribute to carbon emissions.For instance,producing semiconductors,which are at the co
86、re of supercomputer components,is energy-intensive.Energy Consumption:Fabricating and assembling a supercomputer requires significant energy,especially in manufacturing semiconductor components,plus optoelectronics for quantum.The electricity used in clean rooms,where microchips are fabricated,often
87、 comes from fossil fuels or energy-intensive processes,which can substantially increase the carbon footprint.Transportation:Supercomputer components are manufactured in different locations worldwide,and they are often transported across long distances.The shipping of components,including server rack
88、s,cooling systems,and networking equipment,contributes to emissions due to transportation fuels.Manufacturing Processes:The manufacturing process itself involves various stages,including etching,lithography,and assembly.These processes can release greenhouse gasses,depending on the energy sources us
89、ed and the efficiency of the manufacturing facilities.Cooling Systems:Supercomputers generate a significant amount of heat,and efficient cooling systems are essential.The energy used for cooling can be substantial and may involve refrigerants that have a high global warming potential.End-of-Life Dis
90、posal:The carbon footprint also includes considering the disposal of supercomputers at the end of their life cycle.Proper recycling and disposal methods can mitigate environmental impact.In brief,a comprehensive analysis should consider the entire life-cycle assessment of a supercomputer,from the ex
91、traction of raw materials to manufacturing,operation,and disposal.This includes both:23 The direct emissions associated with the hardware life-cycle.The indirect emissions from the energy required to operate the supercomputer.The entire computation cycle should be considered,given that many HPC task
92、s include hybrid architectures involving CPU+GPU and QPU in the quantum cases,as well as pre-and post-analysis often involving ordinary computers,cloud servers,databases,and IT networks.Sustainability is here considered,then,as an energy balance,putting apart the issue of heterogeneous carbonation o
93、f energy according to places,as well as other sustainability issues,such as biodiversity,and pollution,less amenable to scoring the differential impact of quantum computing.Ratios on PASQALs machines were provided to the candidates in order to have a basis of comparison:A figure for the kgCO2-equiva
94、lent per hour of use of the PASQAL quantum stack in a run.A figure for the kgCO2-equivalent per hour of use of the PASQAL quantum stack accounting for the life-cycle footprint of the hardware.Corresponding figures for at least one reference HPC classical computer,including reference to number of cod
95、es and memory size.2.3 Energy/carbon footprint of classical HPC centersComputers,in general,have a carbon footprint typically coming from:Building the computer.Memory(SSD solid-state drives)now weighs the highest,followed by chips within CPU/GPU Running programs direct power consumption by the CPU/G
96、PU complemented by cooling power for large machines.Though sometimes the heat extracted is partially reused,its carbon footprint remains high(as heat use is of lesser carbon offset that the carbonation of electricity supplied)HPC supercomputers vary largely in terms of size,performance,and power nee
97、ds,and we could not find any life-cycle analysis of carbon footprint.As a starting point,here are a few orders of magnitude4 from the greenest according to Green500 score(limited to energy efficiency only)to the quickest in flops :The winner of Green500,Henri,with ar.8200 cores consumes 44kW of powe
98、r for 2,9 Pflops/s.Hard to figure precisely the tCO2eq.fabrication emissions,around 624 tCO2eq.This is around 200 times more power 2 https:/www.vie-publique.fr/loi/278056-loi-15-novembre2021-reen-reduire-empreinte-environnementale-du-numerique3 MIT developed a life-cycle analysis methodology for IT
99、products http:/msl.mit.edu/projects/paia/main.html,adopted for instance by a Dell whitepaper evidencing 80%of the carbon footprint relevant to manufacturing.4 Those orders of magnitude are derived from ratios on the published number of CPUs,GPUs,the amount of RAM,SSD and the power consumption24and 3
100、00 times more CO2 emissions than a standard GPU server(200W to run and some 2t CO2 eq.to manufacture).The winner of TOP500,Frontier with millions of cores consumes 22 MW of electricity power for 1194 Pflops/s.Manufacturing carbon emissions could be around 160 000 tons of CO2.Joliot-Curie Rome,benchm
101、ark machine for GENCI with 7Pflops/s involved around 2,2kt CO2eq of carbon footprint for manufacturing as well as 1.5MW of computing power,meaning with French carbon content 0,1tCO2/kwh as much in run emissions as manufacturing emissions in order of magnitude.25CHAPITRE 032627LESSONS DRAWN FROM THE
102、BLAISE PASCAL QUANTUM CHALLENGE:USE CASES AND MATURITYFor a neutral atom quantum computer,such as PASQALs,problems or methods connected to graphs are the best fit due to its relatively native implementability.This is due to the fact that the atomic register of cold neutral atoms can be naturally int
103、erpreted as a graph-like structure from a mathematical point of view;indeed by identifying individual trapped atoms as vertices and truncated electric dipole-dipole(van der Waals)forces in between them as edges(Thabet et al.,2023).Graphs(or more generally speaking,networks)can appear in many shapes
104、or forms for a large variety of problems,and the candidates managed to exploit this particular feature in various ways.The complexity of the Hackathon came from the fact that the participants were required to identify and precise a fitting problem and propose a solution that is implementable in the
105、NISQ-regime.Problematics of the sustainable transport,smart-grid and smart-city management sectors give rise to many complex network-based optimization tasks,and,as most graph-based optimization problems are connected to combinatorial optimization problems,this yields to large-scale,complex challeng
106、es that rarely have exact solutions.Scheduling or planning type problems,resource allocation related issues,object placement and control type problems are typical examples in this field,and most of them are addressed with heuristic methods,providing a time-efficient but only approximate solution.Due
107、 to the heuristic nature of existing classical solutions,there is room for improvement in the quality of the obtainable solution,which gives relevance to quantum-enhanced optimization.However,as the participants tackling problems of this domain were quick to discover,industrially relevant problem si
108、zes far surpass the encoding capabilities of currently existing QPUs.It is indeed a question of the number of qubits after all.Typically,a telecommunications network of varying frequencies tackles a couple thousand nodes simultaneously for a current classical simulation,just as a modeled electricity
109、,gas,or water Written by Krisztian BENYO,PASQAL team and chief mentor of the Blaise Pascal Regenerative Quantum challenge 28distribution system contains 2000-5000 junctions,nodes,and entry and exit points.Since the network structure has to be captured on the quantum register,amplitude-based efficien
110、t quantum information encoding methods are not applicable.Moreover,because of the many constraints on such systems,ancilla qubits are also necessary,making a typical problem encoding many-to-one in terms of qubits-to-node ratio.Designing intelligent embedding methods that focus more on the structure
111、 of the constraints is a valid strategy;however,it is only relevant to certain types of problem classes(planning and scheduling typically).Quantum chemistry simulation-related problems face similar issues since describing molecular dynamics requires working on the orbital level(so on the scale of th
112、e electrons)rather than on the atomic level.Proteins that the pharmaceutical industry deals with typically have more than a thousand atoms(the smallest known protein,the TRP-cage,consists of 154 atoms),implying that the corresponding amino acid chains would need more than 10.000 orbitals for an actu
113、al quantum embedding.This makes ab initio chemistry-related applications infeasible to the current generation of QPUs,requiring quantum-enhanced applications to divert their gaze towards more structural approaches,focusing on the geometry of molecules rather than their molecular dynamics.Such method
114、s,combined with classical Machine Learning techniques(Henry,et al.,2021),can lead to exploitable algorithms in the domain.The Hackathon has inspired candidates to investigate in various directions in the drug discovery field,leading to original embedding techniques transforming a molecular dynamics-
115、related problem into an adapted optimization problem instead.Exploiting the binding interaction graph in a ligand-receptor interaction,such as the Molecular Docking project(3rd place in the Hackathon),or even encoding crystalline structures on unit-disk graphs,has led to very promising exploratory p
116、rojects that,given some time for quantum technologies to mature to a larger scale,will lead to exploitable applications in the near future.A special mention needs to be made for the NeutroGen project(2nd place in the Hackathon)for taking a more theoretical approach and reinforcing the data pre-proce
117、ssing method for hybrid algorithms in general.As they have reasoned,neutral atom quantum computing provides a promising framework for addressing a wide variety of data-driven sampling problems that can be adapted to combinatorial optimization that can leverage any previously-found solutions from a c
118、onventional optimizer.Incoming data can be automatically translated to high-quality neutral atom positions via techniques from spectral graph theory.The resulting atom embedding gives rise to generated data with similar correlation patterns as the original data using current-day adiabatic protocols.
119、29Interestingly enough,the most promising use cases from the Hackathon have arisen from the renewable,clean energy sector.Not only did the previously mentioned NeutroGen team successfully apply their data-driven approach to benchmark a wind farm optimal placement solution,but the winning NAREF team(
120、1st place in the Hackathon)have managed to propose a novel method based on quantum reservoir computing to describe a complex system of renewable energy resource management.A“reservoir”is itself a physical system that exhibits complex behavior.As a metaphor,its mathematical dynamics are used to repli
121、cate the behavior of underlying time series.It is an alternative to deep recurrent neural networks,linear regression being the only training step implicated,and is done after all data has been fed successively through the reservoir,therefore has a very low training cost.As it was demonstrated by the
122、 team,a free arrangement of neutral atoms makes reservoir computing particularly well suited for analog frameworks such as PASQALs machines.CHAPITRE 043031ENERGY/CARBON FOOTPRINT OF PASQAL TECHNOLOGY4.1 PASQAL neutral atom quantum processing unit energy costIn neutral atom quantum computing architec
123、ture,a sort of vapor of atoms,typically rubidium or strontium,is created inside an ultra-high vacuum chamber.These atoms are trapped and manipulated within the vacuum chamber using highly focused lasers,so-called optical tweezers,to create 2D and 3D arbitrary shapes.The quantum information is encode
124、d by manipulating the atoms electronic states.Each qubit is represented by a two-level energy state in an atom,usually a ground state and a Rydberg state.The current generation of PASQAL devices uses around 100 87Rb(rubidium)atoms for computations.Rubidium is an easy-to-mine alkali metal that benefi
125、ts from well-established laser technology that brings its electrons to various energy levels,including the Rydberg states.In the Rydberg state,atoms physically grow and get polarized,facilitating van der Waals interactions and entanglement.The fact that in PASQAL devices,each qubit is created out of
126、 a single atom makes this technology less prone to errors since the qubits are identical,not fabricated but provided by nature(see Henriet et al.,2020 for technical details).Various components are needed to set up and complete the calculations in a neutral atom processing unit and each of them sum u
127、p to the energy consumption of the device.Since current PASQAL devices function at room temperature,no power consuming refrigeration component needs to be considered.In the worst case scenario the total Fresnels energy consumption amounts to 2-3 kW.4.2 PASQAL quantum processing unit carbon footprint
128、 and benchmark with classical HPC In this subsection we use the carbon dioxide equivalent(CO2eq)standard unit,a standard metric unit defined to compare emissions from various greenhouse gasses on the basis of their global-warming potential(GWP),to evaluate PASQAL QPUs carbon footprint and compare wi
129、th classical architectures.Written by Alexandra de Castro,science and technology communicator at PASQAL,Etienne de Rocquigny,Blaise Pascal Advisors and Matthieu Courtire 500ppm advisory and PASQALs hardware team.32At the present stage a starting estimate for PASQALs Fresnel would involve:In order of
130、 magnitude,a few tens of tons of CO2eq are related to manufacturing the typical hardware equipment(including lasers,control,and opto-electronics).In order of magnitude,a few kW of electricity are used to run.In September 2023,the hardware team estimated approximately 3kW,meaning that,under favorable
131、 low-carbon French electricity,a few tens of tons of CO2eq would be released over an operational lifetime of 7 years.The following table summarizes the benchmark basis between HPC(Joliot Curie supercomputer of GENCI at TGCC/CEA)and PASQALs technology used for the hackathon.Key remaining sources of u
132、ncertainty for later research:The scalability factor:Those orders of magnitude hold for the current 100+qubits QPU and up to the next generation until 1000 qubits(Rubi).Preliminary PASQAL studies suggest a likely slow increase in power requirements up to 10 000 qubits,to be confirmed.The complete ti
133、me needed to perform a quantum computational task benchmarked with the equivalent classical run,including preparation,repetition of noisy cycles,etc.The present-day clock rate is around 1Hz and it is estimated to increase up to 100Hz with good likelihood by 24 months with no increase of power consum
134、ption.The effective ratio of use including technical uptime+saturation of use.Quantum computing is still noisy and a universal fault-tolerant QPU will need to implement error corrections.To implement error corrections we need to encode quantum information into larger Hilbert spaces,which physically
135、might translate into adding more qubits.We will also need to check on the use of classical devices involved in the error correction process(Auffves,2022).The power cost related to the classical CPU/GPU needed to prepare and/or coprocess the computation.In a follow-up white paper,we will provide a co
136、mplete analysis benchmarking quantum/classical power consumption associated with two SDG relevant use cases that are currently being implemented on Fresnel and are candidates to reach quantum advantage.We will analyze the implementation of the same problems addressed with different state-of-the-art
137、classical numerical techniques used for quantum simulations benchmarked with hybrid quantum/classical workflows and discuss the energy consumption and its impact on the environment.33Reference computationJoliot-Curie Rome 12 Pflops/sFresnel under 140 QubitsRubi500 to 1000 QubitsBasicGPU serverImpact
138、 unit in kgCO2eq./uCPU(units)4 5 4 58484020GPU(units)128320RAM(TB)5730,1283 600SDD(PB)00,01551 000HDD(PB)50,153 750Total hardware manufacturing(tCO2 eq)2 17625252Total emissions for HW manufacturing,transport and disposal over lifetimeConservative lifetime hours amortizing hardware emissions49 05620
139、 44020 44028 032Reference number of running hours over lifetime taken as the amortizing basis for the hardware emissions per runEquivalent manufacturing emissions(kgCO2 eq/run hour)441,21,20,07 Nominal Power requirement(kW)1 4363100,2Overhead provision for run power equiv( cooling,maintenance,etc.)1
140、,043,53,51,25Carbonation of electricity(kgCO2 eq/MWh)85858585French electricity is taken as referenceEquivalent run emissions(kgCO2 eq/run hour)1270,93,00,02Total emissions(kgCO2 eq/run hour)1712,14,20,09CHAPITRE 053435 Written by Stphane Requena,CTO and Sabine Mehr,Chief Quantum Officer of GENCI.Th
141、e Quantum Energy Initiative(QEI)argues that quantum technologies will only be scalable in a world of finite resources if we make them as energy-efficient as possible.The Quantum Energy Initiative is a worldwide network of researchers created last year(2022)to encourage and support research on the en
142、ergy costs of quantum technologies.As founders of this initiative,we believe that we need high-quality research to minimize the costs of quantum technologies,in parallel to the ongoing worldwide research to maximize their benefits.If not,we risk arriving at a situation where a quantum technology wil
143、l outperform an existing technology,but at a cost and environmental footprint that is too large to be acceptable for many applications.To avoid this undesirable outcome,it is crucial that researchers develop quantitative methods to minimize energy consumption(and other resource consumptions)in manne
144、rs that do not impact the benefits brought by quantum technology.This requires a system-level view,sometimes called the full-stack view of quantum technology,in which the full-stack is made of layers that contain everything in the quantum technology from the quantum devices up to the end-user.In thi
145、s full-stack picture,there is a layer containing the quantum devices(qubits,etc),another layer containing quantum software(algorithms and protocols),and another layer containing all the enabling technologies(cryogenics,lasers,control electronics,etc).We argue that there are often complicated interre
146、lations between different layers.This makes the energy optimization of the full stack a challenging research problem,requiring more than just the optimization of each layer individually.As each layer is typically the domain of a different discipline(Software being quantum information science.Quantum
147、 hardware being quantum physics.Enabling technologies being engineering),such research is necessarily interdisciplinary and PERSPECTIVES ON SUSTAINABLE QUANTUM ADVANTAGE BY THE RESEARCH AND INSTITUTIONAL COMMUNITIES5.1 A Perspective by the quantum energy research community36requires the development
148、of a common language(and common tools)between disciplines.As an example,consider a noisy quantum computer with two potential methods of achieving the desired precision for a given calculation with that quantum computer.The first method would be to cool the qubits further to reduce the thermal noise,
149、while the second method consists of performing more quantum error correction or mitigation.Both of these come with an energy cost;adding more cooling requires more cryogenics or more powerful laser-cooling(consuming more electrical power),and adding more error correction requires more complicated co
150、rrection protocols(also consuming more electrical power).A quantitative analysis of the costs and benefits of these methods (with costs being power consumption and benefits being improved precision)is required to find the amount of cooling versus error correction that achieves the minimal power cons
151、umption under the constraint of the desired precision.However,we emphasize that this is only one example.This optimum will depend crucially on the type of qubits,the type of cryogenics,and the type of error correction.Improvements in the energy efficiency of either of them will require a re-evaluati
152、on of the optimum,so the methods used to determine the optimum should be robust enough for constant re-use.Each implementation of each quantum technology(quantum computing and simulation,quantum communication,and quantum sensing)will bring up other similar examples,each of which requires optimizatio
153、n.It is thus crucial to share methodologies and results as openly as possible.DOMAINS OF QUANTUM TECHNOLOGIES Quantum Computing Quantum Simulation Quantum Communication Quantum Sensingquantum,algorithms,quantum,error correction,etcqubit enginneering single photon control,etcquantum batteries catalys
154、ts,etcheat,work&entropy in quantum devices,etcQUANTUM THERMODYNAMICS COMMUNITYQUANTUM MATERIALS COMMUNITYMinimizing Energy&Resources in Quantum TechFig.3.The context of quantum energy advantage-Source:Quantum Energy Initiative 2022ENABLING TECHNOLOGIE COMMUNITIESQUANTUM INFORMATION COMMUNITYQUANTUM
155、DEVICES COMMUNITYcryogenics,electronics,lasers,etc37 Written by Catherine Lefebvre,Senior Advisor OQI at GESDA and VP Global Policy and Partnerships at PASQAL;and Marieke Hood,Executive Director Impact Translator at GESDA.The Open Quantum Institute(OQI)is a multilateral governance initiative that pr
156、omotes quantum computing for the benefit of humanity.The launch of OQI was announced in October 2023,in partnership with the Geneva Science and Diplomacy Anticipator(GESDA)Foundation,the Swiss Federal Department of Foreign Affairs(FDFA),CERN and UBS.The OQI was designed and incubated over the period
157、 of 2020-2023 by GESDA an independent non-profit foundation under Swiss law and a private-public partnership with the Swiss and Geneva authorities created in 2019 to strengthen the impact and innovation capacity of the international community through science and diplomacy anticipation and with the p
158、articipation of 130 partners from all over the world,including PASQAL.CERN will host the OQI during a three-year pilot implementation phase from 2024-2026,while GESDA will remain involved,ensuring the further growth of the organization.This ambitious mission of the OQI in pioneering quantum computin
159、g for the benefit of all is structured around four core objectives:Accelerating applications for humanity:Realizing the full potential of quantum computing by accelerating the use cases geared towards achieving the UNs Sustainable Development Goals(SDGs),thanks to the combined forces of researchers
160、and developers,entrepreneurs,the United Nations,and large NGOs.Access for all:Providing global,inclusive,and equitable access to a pool of public and private quantum computers and simulators available via the cloud.Advancing Capacity Building:Developing educational tools to enable everyone around th
161、e world to contribute to the development of quantum computing and make the most of the technology.There is also the open research question of whether quantum technologies are able to consume less resources while performing tasks that conventional technologies can already do,a so-called quantum energ
162、y advantage.This differs from better-known quantum advantages,such as quantum computational advantage,which correspond to the perspective that quantum technology could perform a task faster than existing technology.We at the Quantum Energy Initiative ask whether a quantum energy advantage may become
163、 the motivating factor for a significant proportion of use cases of quantum technologies.However,it is clear that this will only occur if there are serious research efforts to minimize the energy consumption of all aspects of each quantum technology.5.2 Fostering key sustainable use cases:the approa
164、ch of the Open Quantum Institute38 Activating multilateral governance for the SDGs:Providing a neutral forum to help shape multilateral governance of quantum computing for the SDGs.Today,quantum computing is still in its early stage of development and computational resources remain limited.For the f
165、ew applications that can be implemented on current quantum computers,the focus is on applications presenting an immediately graspable commercial or geostrategic advantage,sponsored by organizations with the means to bet on a return-on-investment in the distant future.As a result,too few resources-in
166、 terms of computing and expertise-have been allocated to investigating how quantum computing could be harnessed to achieve the SDGs.Understanding the importance of accelerating the implementation of the SDGs,the OQI has mobilized stakeholders to participate in re-balancing the focus and resources to
167、wards applications beneficial to the SDGs and global challenges.In the past two years,OQI teams from academia,industry,NGOs and International Organizations worked collaboratively to further explore use cases related to Food(SDG2),Health(SDG3)and Climate Change(SDG13).Below are some examples of the S
168、DG use cases(please refer the OQI White Papers 2022 and 2023 for more details):A.Carbon Reduction(SDG 13):Quantum Computing simulation to reduce carbon dioxide(CO2)in the atmosphere by improving catalysis process responsible for the fixation of carbon on the surface of materials by a team of experts
169、 from Swiss Federal Institute of Technology Zrich,Ecole Polytechnique Fdrale Lausanne and United Nations Framework Convention on Climate Change.The window is rapidly narrowing to achieve the goal set by the Paris Agreement of limiting global temperature increase to well below 2 degrees Celsius,while
170、 pursuing efforts to limit the increase to 1.5 degrees Celsius.Carbon Dioxide Removal could play a key role in achieving and sustaining net negative greenhouse gas emission in the long-term.One of such key challenges is to find effective solutions to store and valorize the CO2 once captured.One appr
171、oach to recycling carbon dioxide is to sequester it and then to transform it into other compounds,such as formic acid,methane,methanol,ethanol,or ethene.This chemical transformation can be accelerated by the interaction of carbon dioxide molecules with copper surfaces,through the process so-called h
172、eterogeneous catalysis(Nitopi et al.,2019).Although the catalysis process is relatively well adopted,it is still unclear how these chemical reactions are accomplished locally on a copper surface.Quantum computing is a natural tool to simulate chemical systems(Von Burg et al.,2021)to collect sufficie
173、ntly accurate energy estimates.This is key to deriving a correct mechanism of the chemical processes that will support efficient carbon dioxide reduction solutions.3940B.Mitigating Antimicrobial Resistance(SDG 3):Addressing global public health challenges,by developing a quantum computing solution t
174、o improve current AI models,predict more quickly and accurately patterns of resistance and identifying new chemical compounds with low resistance on more targeted bacteria by a team of experts University of Copenhagen,Alphanosos and Global Antibiotic Research&Development PartnershipWHO declared anti
175、microbial resistance(AMR)as one of the top ten threats to global public health(WHO-GLASS,2021;EClinicalMedicine,2021).Furthermore,the World Bank estimates that if AMR is unchecked,by 2030,an additional 24.1 million people could be forced into extreme poverty(World Bank,2017).Despite the significant
176、economic,environmental,and societal costs,the development of new antimicrobials has not kept pace with the emergence of resistance,leaving healthcare providers with fewer options for treating infections.No new classes of antibiotics have been discovered in the past decades.An approach to AMR researc
177、h is to identify new drugs that have more targeted action to specific diseases and that are efficient against any known resistance mechanism.Machine learning methods are used to find new and efficient combination of compounds and prediction of patterns of resistance(Alphanosos,2021).A quantum machin
178、e learning(QML)algorithm(Cerezo et al.,2022)could replace the standard machine learning algorithms to lead to improved mixes that can be tested experimentally,resulting in an iterative quantum computing/biological experiment workflow with faster convergence and higher-qualitative final samples.C.Sus
179、tainable Food Systems&Global Food Security(SDG 2):Improving sustainability of global food systems by making them more resilient to climate change through a quantum optimization solution to produce more nutritious food locally in less land,and by lowering costs and emissions of food transport-by two
180、teams of experts from Ecole Polytechnique Fdrale Lausanne,National Institute for Theoretical and Computational Sciences and Global Alliance for Improved Nutrition;and from ForeQast,Ernst&Young and University of Oxford.Conflict and insecurity,economic shocks,and extreme weather events are the main dr
181、ivers of acute food insecurity.To achieve SDG 2,the Food and Agriculture Organization(FAO)affirms that a better understanding of food systems requires a local-to-global perspective.Food systems are complex networks of activities,actors,resources,and environments encompassing the production,processin
182、g,distribution,consumption,and disposal of food products.Comprehending these networks is arduous as they are constantly evolving,and are interconnected with broader social,economic and policy systems.41Current projections indicate a necessary increase of food production per hectare by almost 60 perc
183、ent by 2050 to meet the needs of the projected global population of 10 billion(GESDA Radar Breakthrough,2023).As populations continue to grow,innovative solutions are much needed to devise sustainable agricultural practices that provide for more affordable nutritious diets,while respecting planetary
184、 limits.Furthermore,climate change has a very strong impact on food systems,exacerbating hunger and malnutrition issues,especially in regions that are already vulnerable.It triggers extreme weather events,the emergence and spread of pests and diseases,and it compromises the adaptation of traditional
185、ly grown foods,resulting in reduced crop yields(IPCC:Mbow C.,Rosenzweig,C.,Barioni,L.G.,et al.,2019).On the other hand,food systems also contribute significantly to climate change.While the production of food is a large contributor,greenhouse gas emissions associated with food transport also keep in
186、creasing(Pradhan,2022).Accurate modeling of food systems will help provide the basis for improving our ability to produce nutritious food sustainably and foster resilience and respond to unforeseeable food systems disruptions due to climate change.This is a class of optimization problems known as mi
187、xed-integer linear programming.Classical algorithms to find approximate solutions(heuristics)exist.However,this class of problems is in general computationally hard,making large-scale optimization challenging with classical computers and algorithms,as they would require exceedingly long computationa
188、l time to reach a good approximate solution.Quantum algorithms hold promise to significantly enhance the quality of the solution to these optimization problems and represent therefore an ideal use case.The OQI is building a large repository of quantum computing for the SDGs with the objective of ins
189、piring greater participation in this ambitious endeavor to impact humanity.Throughout the process of developing use cases to their implementation on quantum computing devices,the OQI values adopting a responsible approach to quantum computing.This encompasses assessing both societal and environmenta
190、l impacts(including energy/carbon footprint).In doing so,the OQI collaborates with experts to assess such impacts,anticipate negative externalities,and prioritize use cases.CHAPITRE 064243THE PERSPECTIVE OF KEY HIGH-PERFORMANCE COMPUTING CENTERS ON QUANTUM CONTRIBUTION TO DECARBONIZING HPC Written b
191、y Stphane Requena,CTO and Sabine Mehr,Chief Quantum Officer de GENCI of GENCI Created in 2007 by the French Ministry of Higher Education and Research,CEA,CNRS,France Universits and Inria,GENCI(Grand Equipement National de Calcul)has the mission to provide French and European researchers(from academi
192、a and industry)access to leading-edge supercomputing facilities and services for their workloads in numerical simulation,high performance analytics and artificial intelligence.To date,GENCI is giving free access to its facilities to more than 1500 projects per year based on Open Research(with the pu
193、blication of the results at the end of the grand period).Such projects encompass a wide range of scientific and industrial domains,including climate research,material sciences,life sciences and biology,renewable energies,use of artificial intelligence at scale,fundamental sciences,astrophysics,or se
194、ismology,to name a few.In order to address such a variety of heterogeneous needs GENCI is deploying 3 different and complementary HPC architectures,representing a cumulated performance of more than 130 PFlops,hosted and operated by 3 national centers :TGCC for CEA,IDRIS for CNRS,and CINES for France
195、 Universits.In order to continue this dynamics GENCI is also engaged into 2 majors actions :As the Hosting Entity together with CEA(as Hosting Site)and SURF(The Netherlands,as a partner)of the Jules Verne consortium for the co-funding and the hosting/operation of the 2nd Exascale system in Europe ow
196、ned by EuroHPC.Such a system will provide end of 2025 HPC/AI resources as well as exploratory quantum computing accelerators for European researchers;French(HQI)and European(HPCQS,EuroQCS-France)hybrid HPC/quantum computing initiatives.6.1 A French perspective by GENCI:France hybrid HPC quantum init
197、iative(HQI)44HQI(France Hybrid HPC-Quantum Initiative)stems from the French National Quantum Strategy,announced in January 2021 by Emmanuel Macron,the French President,and other initiatives related to quantum communications,sensors,and enabling technologies.It is led by CEA,GENCI,and Inria to tackle
198、 the hybridization between traditional HPC and Quantum computing and simulation technologies.Its a threefold project consisting in:A HPC-QCS platform exhibits various QPU technologies(to date,a device based on neutral atoms from PASQAL and soon another owned by EuroHPC based on photonics),coupled wi
199、th GENCI Joliot Curie supercomputer hosted at CEA-TGCC and later with the future EuroHPC exascale system hosted by France.The underlying idea is to allow French and European researchers to assess various QPU architectures,even in the NISQ noisy era,in order to find(or not)a possible match between th
200、ese hybrid HPC+QC architectures and their algorithms and develop/prepare their workloads using scalable emulators(EVIDEN QaptivaTM,PASQAL Pulser or Quandela Perceval)and real physical systems.This coupling leverages other advanced features of the EVIDEN QaptivaTM solution.This will allow a seamless
201、integration with the HPC resource management,as well as the provision of a portable programming environment,alongside with full-stack approaches provided by QPU vendors.A broad academic and industrial research program,targeting the actual HPC-QCS coupling from a holistic point of view,spanning from
202、system integration up to hybrid end-user applications,as well as more exploratory topics such as noise characterization and mitigation,or leveraging quantum links to scale up.A dynamic dissemination(including hackathons)and end-user support program,to ensure French and European researchers from Acad
203、emia and Industry can assess and benefit from these new Community platformsEventsInternational relationshipsHands-on trainingApplications support team(HLST)use casesMaisons du QuantiqueFIRST SERVICES AVAILABLE SINCE THE END OF2022 STAY TUNED!Hosted atPilot design and implementationApplicationsExplor
204、ationHQI FRANCE HYBRID HPC QUANTUM INITIATIVEScope:2021-2028PERCEVALPULSEREmulationPhotonic QPUEuroQCS-FranceQPUShared software stackHPC72.3MFig.4.Overall view of the French HQI initiative45hybrid technologies.As an example at date HQI with the Paris Region,Le Lab Quantique and Teratec supported mor
205、e than 10 proofs of concept of hybrid quantum computing exploration by industrial users(large groups,SMEs)in the field of optimisation,quantum chemistry,energy,CFD or finance/insurance to name a few.To allow the coverage of a maximum of QPU architectures in Europe,share best practices,improve users
206、engagement,and foster joint R&I actions,HQI is fully integrated inside European initiatives including HPCQS(aiming to couple 2 analogue QPUs from PASQAL to supercomputers in France(GENCI/CEA)and Germany(FZJ),see Fig.5 and to build a common software stack),and EuroQCS-France(aiming to couple a photon
207、ic QPU to Joliot Curie),and collaborating with other EuroHPC JU Hosting Entities through a common integration program.While these actions will engage users in the assessment of the potential of hybrid HPC/quantum computing for accelerating some workloads(in quantum chemistry,optimisation or quantum
208、machine learning)leading to possibly demonstrating quantum advantage in performance in the future,there is also another kind of quantum advantage that could come earlier in the field of power consumption.Today,typical power consumption of multi petascale supercomputers is in the range of 1 to 2 MW w
209、hile next generation systems(so called Exascale)will be in the range of 15 to 20 MW(the#1 top500 supercomputer Frontier owned by DoE at ORNL is requiring 22.7MW for 1.2 Exflops HPL sustained performance).Power consumption is nowadays representing the biggest expenditure in operational expenses of HP
210、C infrastructures and energy efficiency in a systemic approach(from single components,to infrastructure,middleware and end user applications)is a key challenge to address.In that field,a lot of actions are already implemented by HPC centers/agencies in moving to GPUs(more efficient in term of perfor
211、mance per Watt but requiring code porting efforts),using hot water direct liquid(DLC)cooling techniques,profiling/monitoring end users applications for adapting on-the-fly the level of HPC resources,reusing fatal heat for cooling external building or using low-carbon or decarbonized energy sources.T
212、he use of hybrid HPC/quantum computing i.e QPUs coupled as accelerators to other classical compute nodes(CPU,GPU),if successful,could lead first for some specific workloads to an energy quantum advantage by perhaps running algorithms 10 x slower at the beginning but requiring 100 to 1000 x less ener
213、gy.Its important to mention that some scientific fields like quantum chemistry(applied either for materials science,medecine/health or biology)are one of the biggest consumers of HPC cycles of current supercomputers(to 1/3 of the yearly cycles).46Fig.5.Integration of the PASQAL simulators with the J
214、URECA DC and the Joliot Curie supercomputers at FZJ and GENCI/CEA,respectively,as part of the HPCQS project.Quantum Flagship terminology:quantum simulators and quantum annealers are viewed as analogue versions of(digital)quantum computers.The hybrid HPC/quantum computing journey has just started and
215、 its important to be cautious but initiatives like HQI,HPCQS,the ones started by EuroHPC and Member States in the field of hybrid quantum computing as well as QEI(Quantum Energy Initiative)could pave the path to concretely reach in some cases energy and performance quantum advantage before the end o
216、f the decade.6.2 Jlich perspective on HPC integrating quantum computing and energy advantage Written by Kristel Michielsen,Jlich Supercomputing Centre(JSC),Forschungszentrum Jlich(FZJ),Germany Although the Jlich Supercomputing Centre(JSC)first became known by this name in 2007,its predecessor has a
217、long history that goes back to 1961.In autumn 1961,the Central Institute for Applied Mathematics(ZAM)was founded at Forschungszentrum Jlich(FZJ)as a combination of a mathematical institute and a computing center.In 1984,the institute entered the field of supercomputing with the installation of the C
218、ray X-MP supercomputer,which at that time was the fastest supercomputer in Europe.ZAM played a pivotal role in founding the first German national supercomputing center(HLRZ)in 1987.The Cray supercomputer complex installed in 1996 was another milestone in 100+qubit PASQAL quantum computers*Fig.5.Inte
219、gration of the PASQAL computers with the JURECA DC and the Joliot Curie supercomputers at FZJ and GENCI/CEA,respectively,as part of the HPCQS project*referred to as quantum simulators according to quantum flagship terminology47ZAMs history:for the first time,a supercomputer at FZJ was among the top
220、10 in the TOP500 list of the fastest supercomputers worldwide.After 2004,the institute expanded to a leading supercomputing center not only in Germany but also in Europe.In 2007,JSC joined forces with the High Performance Computing Centre Stuttgart(HLRS)and the Leibniz Supercomputing Centre(LRZ)in G
221、arching near Munich to form the Gauss Centre for Supercomputing(GCS),which unites the three most powerful German computing centers under one roof.In 2010,GCS was one of the four founding members of the European HPC infrastructure PRACE Partnership for Advanced Computing in Europe.Since 2004,in an in
222、creasingly competitive research landscape,funds could be secured to continue to procure and install leadership-class supercomputers.In 2004,the massively parallel supercomputer IBM p690 cluster JUMP was installed and JSC started working with IBM on the IBM Blue Gene series of supercomputers.Utilizin
223、g the effect that using lower frequencies for computing cores leads to quadratically decreasing energy consumption,these machines used a very large number of nodes,but the CPU in each node was relatively slow.The energy efficiency increases when the frequency is lowered,and the communication network
224、 and main memory can be utilized far more effectively.This paradigm shift in frequency marked the JSCs transition to highly scalable systems.In this way,factors of up to over three in energy efficiency could be achieved compared to all other known technologies.The Blue Gene systems have frequently t
225、opped the Green500 ranking of the most energy-efficient supercomputers.In 2006,JSC installed its first IBM Blue Gene system called JUBL,an IBM BlueGene/L supercomputer.Its successor JUGENE,based on IBMs BlueGene/P architecture,went into operation in 2008.In 2009,two supercomputers JUROPA,built by Su
226、n and successor of JUMP,and HPC-FF,built by Bull,were installed.Both computers could be connected for specific tasks and together they achieved a performance of 274.8 teraflop/s.It was the first computer to be co-designed by JSC and realized by a multinational collaboration of several companies.With
227、 an upgrade of JUGENE in 2009,JSC became the first computing center having petaflop capability and in 2010 the first center delivering computing time to the European PRACE community.Since then,Tier-0 supercomputing resources and services including leadership-class supercomputing systems and state-of
228、-the-art operation and high-level support services are provided to national and international users.To keep pace with the rapidly growing demand for supercomputing resources and the fast development of new hardware the 1 petaflop/s IBM Blue Gene/P system JUGENE was replaced by the next-generation 5.
229、9 petaflop/s IBM Blue Gene/Q system JUQUEEN in 2012.JUQUEEN was the first HPC system in Europe to pass the 5 petaflop/s barrier.With the succession of various IBM Blue Gene systems,JSC has been operating one of the most energy-efficient computers in the world for some time.48In the summer of 2015,th
230、e JURECA system,the successor of JUROPA,went into operation.At first,it only consisted of the JURECA cluster module,serving as a general-purpose supercomputing resource.In 2017,the system was augmented with a many-core processor based booster module to enable highly scalable applications to leverage
231、 the system more efficiently.Both modules were tightly integrated and operated as a single system with a modular system architecture(MSA)(see Fig.XX),following the modular supercomputing paradigm,pioneered by JSC in the context of the DEEP series of EU-funded projects.The JURECA cluster module,desig
232、ned by JSC and T-Platforms,had a peak performance of 1.7 petaflop/s.The JURECA booster module,designed by JSC and Intel,had a peak performance of 5 petaflop/s.JURECA was the first modular supercomputer to be included on the Top500 list.In 2020,the JURECA cluster module was replaced by JURECA DC,a fl
233、exible Data Centric(DC)module supplied by Atos to process large volumes of data.In its current configuration JURECA DC has a peak performance of 18.5 petaflop/s.In 2022,the JURECA booster module was decommissioned.In spring 2018,JUQUEEN was shut down and its successor JUWELS,built by Atos,was instal
234、led.JUWELS is constructed as a modular supercomputer(see Fig.6),an architectural paradigm developed at JSC over the past years(Suarez et al.,2019),(Suarez et al.,2022).JUWELS consists of multiple,architecturally diverse but fully integrated,Tier-0 modules designed for specific simulation,data scienc
235、e and AI tasks.The first module,replacing JUQUEEN,is a versatile cluster architecture based on commodity multi-core CPUs.The JUWELS cluster module has a peak performance of 10.6(CPU)+1.7(GPU)petaflop/s consuming 1.2 megawatts of power.For the second module JSC has focused heavily on energy-efficient
236、 GPUs as the main elements of parallel computing,so that the system receives a large performance boost with the lowest possible energy consumption.This GPU-based booster module was installed in 2020.The JUWELS booster module has a peak performance of 73 petaflop/s consuming about 1.1 megawatts of po
237、wer.The JUWELS booster module was the most energy-efficient system of the hundred most powerful computers in the world when it was introduced.The combined cluster and booster modules have a peak performance of 85 petaflop/s,which made JUWELS the most powerful supercomputer in Europe.In 2024,JSC will
238、 install and operate the first European exascale computer JUPITER,also based on the MSA.The JUPITER booster module will comprise close to 24,000 Nvidia GH200 chips.Being the worlds most powerful AI system,JUPITER can deliver 93 exaflop/s of peak performance for AI training 45x more than JUWELS boost
239、er and 1 exaflop/s for HPC applications,while consuming only 18.2 megawatts of power.In normal operation,the power consumption is expected to be about 12 megawatts.To maintain its position at the forefront in the field of high-performance computing JSC pursues active research and technological devel
240、opment.49For this purpose,several cooperations with leading hardware vendors,software companies and system integrators were established.This research effort is complemented by JSCs participation in different national and international research projects.In addition,at JSC there is a deep involvement
241、in research and development on key components for advanced computer technologies such as quantum computing and neuromorphic computing.The search is on for a technology that offers more computing power at a comparably lower energy cost.The advent of MSAs with for example quantum computing architectur
242、es might be a first step to reach this goal as it will enable hybrid quantum-HPC applications at reduced energy costs,but it will not solve the fundamental problems of digital HPC.The quantum computing strategy of JSC(Lippert et al.,2022)consists of four pillars.The first pillar is the modelling and
243、 emulation of quantum computing devices.The JSC began this work back in 2004.The activities in this pillar consist of developing software for validating designs of quantum processors and researching performance expectations for quantum algorithms.To this end,the emulator JUQCS(Jlich Universal Quantu
244、m Computer Simulator)emulator was developed(De Raedt et al.,2007)(De Raedt et al.,2019).With JUQCS several world records have been achieved in emulating the largest quantum computer,a 48-qubit quantum computer,on an HPC system.JUQCS was also used in a collaboration with Google and others to benchmar
245、k the Google quantum processor Sycamore and to demonstrate quantum supremacy(Arute et al.,2019),marking the moment when a quantum computer outperforms state-of-the-art conventional computers for a given task for the first time.In the second pillar,the provision strategy,JSC has been installing and o
246、perating a quantum annealer(since 2021),an analog quantum computer with superconducting qubits,from the Canadian company D-Wave Systems.The hosting of a second analog quantum computer is planned for mid-2024 as part of the European project HPCQS,namely a quantum simulator with neutral atom qubits fr
247、om the French start-up PASQAL.Also in 2024,the first digital quantum computer will be installed,an ion trap quantum computer from the German start-up eleQtron,as part of a project funded by Nort-Rhine Westphalia(NRW).This complements the remote access that JSC will provide to the digital quantum com
248、puters with superconducting qubits that are being developed in the German and European research projects Qsolid and OpenSuperQPlus.In pillar three,HPC systems and quantum computers are integrated.The MSA(see Fig.6)of the HPC systems at JSC is ideal for integrating quantum computing functions into an
249、 HPC workflow.After a preparatory phase that started in 2015,JSC founded the Jlich UNified Infrastructure for Quantum computing(JUNIQ)(see Fig.7)in the context of the fourth pillar of its quantum computing strategy.JUNIQ is jointly funded by the Government(BMBF),NRW and 50the HELMHOLTZ Association.J
250、UNIQ launched user operation in 2019.JUNIQs mission is to provide German and European science and industry with cloud-access to various types of quantum computers and to run software emulators of ideal and real quantum computers on JSCs Exascale-class supercomputers.Right from its start,JUNIQ is at
251、the vanguard of deeply integrating HPC,and quantum computers and simulators at system level with lowest latency and highest data throughput taking advantage of the MSA.Today,JUNIQ can already integrate its quantum computing systems into JSCs modular HPC environment.With its unified development platf
252、orm,JUNIQ will include NISQ systems like OpenSuperQplus/QSolid(2024)or eleQtron(2024),quantum simulators like PASQAL/HPCQS(2024),and quantum annealers like the D-Wave Advantage system(2022).All these systems have a power consumption of less than 25 kilowatts,much less than a conventional supercomput
253、er.JUNIQ offers a continuously running call for projects(peer reviewed)together with pertinent support through its Simulation Lab for Quantum Computing to realize hybrid quantum-HPC applications on the integrated systems.Fig.6.JSCs Modular Supercomputer ArchitectureDate StorageModuleModular Supercom
254、puterCPU ClusterModuleGPU BoosterModuleDate AnalyticsModuleQuantumModuleNeuromorphicModuleJSC51Fig.7.Emulators and quantum computers hosted and operated by JUNIQ*EuroHPC formulation for an analog quantum computerEMULATORSDIGITAL QuantumcomputersJUQCSJlich QuantumComputer SimulatorQAPTIVAQuantumAppli
255、cation PlatformQuantumCOMPUTERQuantumCOMPUTERModular SupercomputerJSCQuantum ANNEALERQuantum SIMULATOR*ANALOG QuantumcomputersIn conclusion,it is clear that the highest efficiency gains can be achieved by reducing primary energy consumption(hardware technology solutions).Further reductions can be ob
256、tained by efficient computer room infrastructures and green coding(software solutions).COMPUTER*referred to as quantum simulators according to quantum flagship terminologyCHAPITRE 075253THE PERSPECTIVE OF A KEY INDUSTRIAL END-USER:EDF,THE LEADING FRENCH ENERGY UTILITYFor a major player in the energy
257、 sector,such as EDF,the use of High Performance Computing(HPC)resources should constantly grow.In the early 1980s,EDFs research and development department started to carry out model implementations and simulations in silico on a Cray,a state-of-the-art supercomputer in terms of performance at the ti
258、me.Since then,EDF has been replacing one of the two supercomputers every two and a half years,with technologies consistently ranking in the top 100 in the world at the time of the purchase.EDFs simulations were initially focused on physics and have been evolving towards diverse applications,such as
259、machine learning and solving combinatorial problems.Moreover,the energy transition makes most of the computational problems that energy companies face even harder to solve.For this reason,EDF turned its attention to quantum computing as it has the potential to become not only an ally at every stage
260、of the EDF Groups value chain but also a potential asset to accompany the transition to sustainable energies.Indeed,quantum computing holds the potential to help tackle crucial problems for EDF,such as supporting the electrification of vehicles,forecasting production and demand,and simulating the ag
261、ing of equipment.The first example concerns the smart charging of vehicles.It typically encompasses a range of issues,from the routing of electric vehicles(EVs)to what is known as Vehicles to Grid(V2G).The first case aims to direct EVs to the least congested charging stations,ensuring that priority
262、vehicles,such as ambulances,are charged before others.In the second case,the aim is to use the mobile batteries of EVs to relieve the grid during peak periods.In both cases,complex combinatorial problems arise:exact solutions become difficult beyond a few hundred EVs,and even approximate methods str
263、uggle beyond a few thousand vehicles.When considering fleets of several million EVs,quantum computing emerges as a potential solution;see for example,our paper in collaboration with PASQAL and Loria(Daylac et al.,2021).Written by Joseph Mikael,Head of Quantum Computing,EDF R&D Another example might
264、be one of the historic challenges for utilities:forecasting production and demand.Until recently,electricity production mainly relied on decisions made by the operator,resulting in relatively low volatility.On the consumption side,fluctuations were also comparatively less volatile.Today,we witness i
265、ncreased consumption volatility due to the electrification of usages and renewable energy sources that further complicate production forecasting.A gust of wind,for instance,can lead to a significant variability in production.The techniques employed to address these new phenomena need to be updated,a
266、nd again,quantum computing might be an asset to handle them(see outputs of the Blaise Pascal Quantum Challenge).The third example refers to the microscopic simulation of the aging of materials,another subject linked to the energy transition,where quantum physics may have a role to play.Extending the
267、 operation of certain facilities and studying the aging of static batteries,nuclear internals,or photovoltaic panels all rely on atomic-scale simulations of the behavior of materials.Today,these simulations are carried out conventionally using a set of assumptions we would like to eliminate.Doing th
268、ese computations more precisely would allow us to identify swelling in the structures,understand how batteries age with intensive cycles of charging/discharging periods,and have a detailed understanding of the long-term performance of photovoltaic farms.Together with PASQAL,IOGS,and EVIDEN,EDF has m
269、ade an effort in this direction(Michel et al.,2022).For EDF,the advantage of quantum computing can be associated with the quality and precision of a solution(Machine Learning,optimization),the size of the problems that are doable(optimization),be able to avoid assumptions(material simulation),and,la
270、st but not least,it might be the energetical cost of running an algorithm.At EDF,one of the two HPCs has a power of 1.4MW,whereas Frontier has 22MW.Reducing this C02 bill might be key if we want this technology to be adopted.5455AN AGENDA TO MOVE FORWARD IN SEARCH OF SUSTAINABLE QUANTUM ADVANTAGEAt
271、the end of this whitepaper,the following conclusions can be drawn:The potential of quantum computing to contribute to sustainability could be significant.We have seen that researchers can implement use cases specified in the international standard of United Nations Sustainable Development Goals on a
272、 quantum processor.These could include computing for drug discovery(included in SDG3,instrumental in accessing essential healthcare,medicines,and vaccines),renewable energies(included in SDG13,essential for climate change mitigation and biodiversity),smart cities and transport systems(included in SD
273、G 6&11,necessary to sanitation and circular economy).Challenges associated with sustainable transport,smart grid,or smart city management are associated with optimization tasks that yield large-scale,complex calculations that rarely have exact solutions within classical methods.These combinatorial o
274、ptimization problems that are best tackled using graph methods and molecular chemistry simulations are natively implementable on the PASQAL quantum processor and effectively addressed in synergy with artificial intelligence as exemplified by the Hackathon winner.Establishing a demonstration of a cle
275、ar energy/carbon emission advantage with quantum/classical HPC benchmarks is a challenging ongoing research.However,it may be at hand before quantum computing delivers a fault tolerance performance advantage.It will CONCLUSION56be essential to provide a solid analysis benchmarking quantum/classical
276、power consumption associated with relevant use cases and an estimation of the life-cycle analysis accounting for all emissions associated with design,manufacturing,maintenance,and end-of-life beyond the pure electrical power to run the algorithms on the CPU,GPU or QPU.Current PASQAL neutral atoms te
277、chnology may be better placed than many other quantum technologies to deliver decarbonizing,due to low energy requirements to run the algorithms in hybrid workflows and limited manufacturing emissions concerning QPU and GPUs.The issue of scalability will be further investigated.PASQAL will issue a c
278、omplementary publication with a solid analysis benchmarking quantum with classical computing associated with two relevant use cases.The following vital stakes remain for further analysis and discussion:1.Broaden the committed community by working with partners with similar values as the present whit
279、e paper advocates,particularly with the Quantum Energy Initiative and Open Quantum Institute as facilitators.2.Foster sustainable quantum meetings/events,such as PASQALs presence at COP28,workshops organized by the Quantum Energy Initiative(QEI 2023),Australias International Conference on Quantum En
280、ergy(ICQE 2023),or the Blaise Pascal ReGenerative Quantum Challenge hackathon.3.Build on current frameworks outlined in this manifesto to assess hardware-linked emissions and promote eco-design practices for hardware efficiency concerning energy consumption.4.Propose use cases with the most favorabl
281、e sustainability advantage.5.Deeply analyze the quantum-classical computing performances to ascertain quantum energy advantage.6.Investigate the computational life-cycle from the materials used to construct the computers to their end-of-life disposal.From a more general perspective,comparing quantum
282、 and classical computing can be virtuous,leading to improvements in both,or synergy between them.However,we encourage those involved to consider use-cases related to sustainability,and to aim for the most energy-efficient and sustainable computing possible.This would only be possible with capacity-b
283、uilding,including a new generation of talented professionals and scientists :in that respect also,maybe the quest for sustainability could become an attractive motivation.5758REFERENCES5901.(Albrecht et al.,2023)Albrecht,B.,Dalyac,C.,Leclerc,L.,Ortiz-Gutirrez,L.,Thabet,S.,DArcangelo,M.,Cline,J.R.K.,
284、Elfving,V.E.,Lassablire,L.,Silvrio,H.,Ximenez,B.,Henry,L.,Signoles,A.,&Henriet,L.(2023).Quantum feature maps for graph machine learning on a neutral atom quantum processor.Physical Review,107(4).https:/doi.org/10.1103/physreva.107.042615.02.(Alphanosos,2021)Alphanosos.(2021,July 20).Home-Alphanosos.
285、https:/ Global Electricity Usage of Communication Technology:Trends to 2030.Challenges,6(1),117157.https:/doi.org/10.3390/challe601011704.(Angarita-Zapata et al.,2021)Angarita-Zapata,J.S.,Alonso-Vicario,A.,Masegosa,A.D.,&Legarda,J.(2021).A Taxonomy of Food Supply Chain Problems from a Computational
286、Intelligence Perspective.Sensors,21(20),6910.https:/doi.org/10.3390/s2120691005.(Arute et al.,2019)F.Arute et al.(2019).Quantum supremacy using a programmable superconducting processor,Nature 574,505-510.06.(Auffves,2022)Auffves,A.(2022).Quantum technologies need a quantum energy initiative.PRX Quan
287、tum,3(2).https:/doi.org/10.1103/prxquantum.3.020101.07.(Berger,at al.,2021)Berger,C.et al.(2021).Quantum technologies for climate change:Preliminary assessment.Available in:2107.05362.pdf(arxiv.org).08.(Bilan Carbone Du Numrique :Impacts,Consquences Et Solutions,2023)Bilan carbone du numrique :impac
288、ts,consquences et solutions.(2023,July 28).Climate Consulting.https:/ et al.,2022)Bobier,J.,Gerbert,P.,Burchardt,J.,&Gourvitch,A.(2022,September 15).A quantum advantage in fighting climate change.BCG Global.https:/ et al.,2022)Cerezo,M.,Verdon,G.,Huang,H.,Cincio,.,&Coles,P.J.(2022).Challenges and op
289、portunities in quantum machine learning.Nature Computational Science,2(9),567576.https:/doi.org/10.1038/s43588-022-00311-3.11.(Chattu et al.,2023)Chattu,V.K.,Singh,B.P.,Pattanshetty,S.,&Reddy,K.S.(2023).Access to medicines through global health diplomacy.Health Promotion Perspectives,13(1),4046.http
290、s:/doi.org/10.34172/hpp.2023.0512.(Chen et al.,2023)Chen,C.,Bornet,G.,Bintz,M.et al.Continuous Symmetry Breaking in a Two-dimensional Rydberg Array.Nature(2023).https:/doi.org/10.1038/s41586-023-05859-2.13.(Cigref,2020)Cigref.(2020,November 27).Publication Sobrit numrique :une dmarche dentreprise re
291、sponsable.Cigref.https:/www.cigref.fr/publication-sobriete-numerique-une-demarche-d-entreprise-responsable.14.(Climate Change:Global Temperature,2023b)Climate change:global temperature.(2023b,January 18).NOAA Climate.gov.https:/www.climate.gov/news-features/understanding-climate/climate-change-globa
292、l-temperature.15.(CodeCarbon,2020)CodeCarbon 2.3.2 documentation:Methodology.(2020).GitHub.Retrieved November REFERENCES6030,2023,from https:/mlco2.github.io/codecarbon/index.html.(DArcangelo et al.,2023)DArcangelo,M.,Loco,D.,Fresnel team,et al.(2023).Leveraging Analog Quantum Computing with Neutral
293、 Atoms for Solvent Configuration Prediction in Drug Discovery.Preprint available here:https:/arxiv.org/abs/2309.12129.16.(Data centres&networks,2023)Data centres&networks-IEA.(n.d.-b).(2023)IEA.https:/www.iea.org/energy-system/buildings/data-centres-and-data-transmission-network.17.(De Raedt et al.,
294、2007)K.De Raedt,K.Michielsen,H.De Raedt,B.Trieu,G.Arnold,M.Richter,Th.Lippert,H.Watanabe,N.Ito.(2007).Massive Parallel Quantum Computer Simulator,Comp.Phys.Comm.176,121136 18.(De Raedt et al.,2019)H.De Raedt,F.Jin,D.Willsch,M.Nocon,N.Yoshioka,N.Ito,S.Yuan,K.Michielsen.(2019).Massively parallel quant
295、um computer simulator,eleven years later,Comp.Phys.Comm.237,47-61.19.(Dhar,2020)Dhar,P.(2020).The carbon impact of artificial intelligence.Nature Machine Intelligence,2(8),423425.https:/doi.org/10.1038/s42256-020-0219-9.20.(Dpi campaigns,2017)Dpi campaigns.(2017).Take Action for the Sustainable Deve
296、lopment Goals-United Nations Sustainable Development.United Nations Sustainable Development.https:/www.un.org/sustainabledevelopment/sustainable-development-goals/21.(Ece,2023)Ece,A.(2023).Computer-aided drug design.BMC Chemistry,17(1).https:/doi.org/10.1186/s13065-023-00939-w.22.(EClinicalMedicine,
297、2021)EClinicalMedicine.(2021).Antimicrobial resistance:a top ten global public health threat.EClinicalMedicine,41,101221.https:/doi.org/10.1016/j.eclinm.2021.10122123.(EIA,2023)Frequently asked questions(FAQs)-U.S.Energy Information Administration(EIA).(October,2023).https:/www.eia.gov/tools/faqs/fa
298、q.php?id=97&t=324.(Ezratty,2023)Ezratty,O.(Ed.).(2023,January 23).Manifesto archive Quantum Energy Initiative.Quantum Energy Initiative.Retrieved November 30,2023,from https:/quantum-energy-initiative.org/manifesto-archive/25.(Ezratty,2023)Ezratty,O.(2023).Understanding Quantum Technologies.6th edit
299、ion.Understanding Quantum Technologies 2023().26.(Fellous-Asiani,2022)Fellous-Asiani,M.(2022,October 26).Could energy efficiency be quantum computers greatest strength yet?PhysOrg.Retrieved November 30,2023,from https:/phys.org/news/2022-10-energy-efficiency-quantum-greatest-strength.html27.(Fellous
300、-Asiani et al.,2022)Fellous-Asiani,M.,Chai,J.H.,Thonnart,Y.,Ng,H.K.,Whitney,R.S.,&Auffves,A.(2022).Optimizing resource efficiencies for scalable full-stack quantum computers.https:/doi.org/10.1103/prxquantum.4.040319.https:/arxiv.org/pdf/2209.05469.pdf 28.(Ganascia,2022)Ganascia,J.-G.(2022,September
301、 22).Et si lavantage quantique tait nergtique?larecherche.fr.Retrieved November 30,2023,from Et si lavantage quantique tait nergtique?|l29.(GESDA Radar Breakthrough,2023)GESDA.(2023).GESDA Radar Breakthrough,https:/radar.gesda.global/sub-topics/resilient-farming6130.(Green500|TOP500,2023)Green500|TO
302、P500.(2023,November 14).Top 500.Retrieved November 29,2023,from https:/top500.org/lists/green500/Green500 :ranking supercomputers according to run efficiency ratio(GFlops/W)-https:/top500.org/lists/green500/31.(Heikkil,2022)Heikkil,M.(2022,November 15).Were getting a better idea of AIs true carbon f
303、ootprint.MIT Technology Review.https:/ et al.,2020)Henriet,L.,Beguin,L.,Signoles,A.,Lahaye,T.,Browaeys,A.,Reymond,G.O.,&Jurczak,C.(2020).Quantum computing with neutral atoms.Quantum,4,327.https:/doi.org/10.22331/q-2020-09-21-327(https:/doi.org/10.22331/q-2020-09-21-327)33.(Henry,et al.,2021)Henry,L.
304、,Thabet,S.,Dalyac,C.,&Henriet,L.(2021).Quantum evolution kernel:Machine learning on graphs with programmable arrays of qubits.Physical Review,104(3).https:/doi.org/10.1103/physreva.104.032416.34.(IPCC:Mbow C.,Rosenzweig,C.,Barioni,L.G.,et al.,2019)IPCC:Mbow C.,Rosenzweig,C.,Barioni,L.G.,et al.,2019,
305、Food Security.In:Climate Change and Land:an IPCC special report on climate change,desertification,land degradation,sustainable land management,food security,and greenhouse gas fluxes in terrestrial ecosystems https:/www.ipcc.ch/srccl/chapter/chapter-5/35.(Koomey,2023)Koomey,J.(2023,March 29).Moores
306、Law Might Be Slowing Down,But Not Energy Efficiency.IEEE Spectrum.https:/spectrum.ieee.org/moores-law-might-be-slowing-down-but-not-energy-efficiency36.(Lacoste A.et al,2019)Lacoste A.et al.(2019).Quantifying the Carbon Emissions of Machine Learning.https:/arxiv.org/abs/1910.09700 37.(Lannelongue et
307、 al.,2021)Lannelongue,L.,Grealey,J.,&Inouye,M.(2021).Green Algorithms:Quantifying the carbon footprint of computation.Advanced Science,8(12).https:/doi.org/10.1002/advs.202100707.38.(Lippert et al.,2022)Thomas Lippert and Kristel Michielsen.(2022).“Perspectives of Quantum Computing at the Jlich Supe
308、rcomputing Centre”,NIC Symposium 2022,Publication Series of the John von Neumann Institute for Computing(NIC)NIC Series 51,3 23.39.(Lorenzini,2021)Lorenzini,R.(2021,November 25).Digital&environment:How to evaluate server manufacturing footprint,beyond greenhouse gas emissions?|Boavizta.Boavizta.http
309、s:/boavizta.org/en/blog/empreinte-de-la-fabrication-d-un-serveur.40.(Luccioni,2023)Luccioni,A.S.(2023).Estimating the carbon footprint of BLOOM,a 176B parameter language model.http:/jmlr.org/papers/v24/23-0069.html41.(Michel et al.,2023)Michel,A.,Grijalva,S.,Henriet,L.,Domain,C.,&Browaeys,A.(2023).B
310、lueprint for a digital-analog variational quantum eigensolver using Rydberg atom arrays.Physical Review,107(4).https:/doi.org/10.1103/physreva.107.04260242.(MIT Materials Systems Laboratory,2019)MIT Materials Systems Laboratory.(2019).Intended Uses and Limitations of the PAIA Model.PAIA Research App
311、roach|Materials Systems Laboratory.Retrieved December 1,2023,from https:/p1-ofp.static.pub/ShareResource/social_responsibility/PAIA_Intended_Use/PAIA_Intended_Use.pdf.43.(Nitopi et al.,2019)Nitopi,S.,Bertheussen,E.,Scott.,S.,et al.(2019).Progress and Perspectives of Electrochemical CO2 62Reduction o
312、n Copper in Aqueous Electrolyte.Chemical Reviews,119(12),76107672.https:/doi.org/10.1021/acs.chemrev.8b0070544.(Numrique|Impact CO2,2023)Numrique|Impact CO2.(2023).Impact CO2.Retrieved November 29,2023,from https:/impactco2.fr/numerique.Elementary ADEME impact calculus https:/impactco2.fr/numerique4
313、5.(Portella and Tissier,2023)Portella,Y.and Tissier,J.(April 21,2023).Study on PASQAL machines-Energetic consumption of a neutral atom-based quantum computer:a comparative and predictive full-stack analysis.46.(Pradhan,2022)Pradhan,P.(2022).Food transport emissions matter.Nature Food,3(6),406407.htt
314、ps:/doi.org/10.1038/s43016-022-00524-947.(Product Carbon Footprints|Dell USA,2023)Product Carbon Footprints|Dell USA.(2023,November).Dell Technologies.Retrieved November 30,2023,from https:/ et al.,2020)Puebla,J.,Kim,J.,Kondou,K.,&Otani,Y.(2020).Spintronic devices for energy-efficient data storage a
315、nd energy harvesting.Communications Materials,1(1).https:/doi.org/10.1038/s43246-020-0022-5.49.(Reimann et al.,2023)Reimann,L.,Vafeidis,A.T.,&Honsel,L.E.(2023).Population development as a driver of coastal risk:Current trends and future pathways.Cambridge Prisms Coastal Futures,1.https:/doi.org/10.1
316、017/cft.2023.350.(Serret et al.,2020)Serret,M.F.,Marchand,B.,&Ayral,T.(2020).Solving optimization problems with Rydberg analog quantum computers:Realistic requirements for quantum advantage using noisy simulation and classical benchmarks.Physical Review,102(5).https:/doi.org/10.1103/physreva.102.052
317、617.51.(Scholl et al.,2021)Scholl,P.,Schuler,M.,Williams,H.J.,Eberharter,A.A.,Barredo,D.,Schymik,K.,Lienhard,V.,Henry,L.,Lang,T.C.,Lahaye,T.,Luchli,A.M.,&Browaeys,A.(2021).Quantum simulation of 2D antiferromagnets with hundreds of Rydberg atoms.Nature,595(7866),233238.https:/doi.org/10.1038/s41586-0
318、21-03585-1.52.(Suarez et al.,2019)E.Suarez,N.Eicker and T.Lippert,.(2019).Modular Supercomputing Architecture:from idea to production,Chapter 9 in Contemporary High Performance Computing:from Petascale toward Exascale,vol.3,J.S.Vetter,Ed.,CRC Press,pp.223-251.53.(Suarez et al.,2022)Suarez,E.,Eicker,
319、N.,Moschny,T.,Pickartz,S.,Clauss,C.,Plugaru,V.,Herten,A.,Michielsen,K.,&Lippert,T.(2022).Modular Supercomputing Architecture.Zenodo.https:/doi.org/10.5281/zenodo.650839454.(Thabet et al.,2023)Thabet,S.,Fouilland,R.,Djellabi,M.,Sokolov,I.,Kasture,S.,Henry,L-P.,and Henriet,L.(2023).Enhancing Graph Neu
320、ral Networks with Quantum Computed Encodings.Available at:https:/arxiv.org/abs/2310.2051955.(UN-DESA,2023)United Nations publication issued by the Department of Economic and Social Affairs(DESA).(2023).The Sustainable Development Goals Report 2023:Towards a Rescue Plan for People and Planet(J.Ross,E
321、d.;Special Edition)Online.United Nations.ISBN:978-92-1-101460-0.The-Sustainable-Development-Goals-Report-2023.pdf(un.org).56.(UN Resolution,2017)UN Resolution adopted by the General Assembly on 6 July 2017.(July,2017).Work of the Statistical 63Commission pertaining to the 2030 Agenda for Sustainable
322、 Development.Seventy-first session.A/RES/71/313.A/RES/71/313-E-A/RES/71/313-Desktop(archive.org).Visited on 24/11/2023.57.(Varsamopoulos et al.,2022)Varsamopoulos,S.,et al,(2022).Quantum Extremal Learning.Preprint available:2205.02807 Quantum Extremal Learning(arxiv.org)58.Von Burg,V.,Low,G.H.,Hner,
323、T.,Steiger,D.S.,Reiher,M.,Roetteler,M.,&Troyer,M.(2021).Quantum computing enhanced computational catalysis.Physical Review Research,3(3).https:/doi.org/10.1103/physrevresearch.3.033055.59.(WHO,2022,July 6)World Health Organization:WHO.(2022,July 6).UN Report:Global hunger numbers rose to as many as
324、828 million in 2021.World Health Organization.Retrieved November 25,2023,from https:/www.who.int/news/item/06-07-2022-un-report-global-hunger-numbers-rose-to-as-many-as-828-million-in-2021.60.(WHO-Division,2021)WHO-Division,A.R.(2021,June 9).Global Antimicrobial Resistance and Use Surveillance Syste
325、m(GLASS)Report:2021.https:/www.who.int/publications/i/item/978924002733661.(World Bank,2017)World Bank,2017,Drug-Resistant Infections:A Threat to Our Economic Future.,Washington,DC:World Bank.License:Creative Commons Attribution CC BY 3.0 IGO,https:/documents1.worldbank.org/curated/en/32331149339699
326、3758/pdf/final-report.pdf(accessed 15 September 2023).62.(Wu et al.,2021)Wu,C-J.,et al.(October 30,2021).Sustainable AI:Environmental Implications,Challenges and Opportunities.Published in Conference on Machine-Computer Science.2111.00364.pdf(arxiv.org).APPENDIX6465THE BLAISE PASCAL REGENERATIVE QUA
327、NTUM CHALLENGE The hackathon scoring process as an effort to standardize sustainable quantum advantageThe hackathon candidates were scored according to the following processFirst stage involves filtering on the use case for sustainability assessment Dates and Process :Submission by October 17th,By O
328、ctober 20th validated maximum 50 projects that can advance Deliverable:text or slideshow detailing the use case in a sustainability statement of purpose,methodology and general computational approach(including relevance to the challenge&common good,potential,and quantum project relevance)Scoring is
329、built with as following subscores:Exclusion criterion for some use cases:those prohibited by the upcoming European AI-act as well as military applications Sustainability of the statement of purpose Inspiration can be found in the UN sustainable development goals,or the,or the Corporate Dreams(climat
330、e/circular economy/health/fulfilling jobs),Each project should contribute at least to one of the 6 following domains :sustainable transport,industry and circular economy(SDG9 et 12;corp dream“circular”)smart city(incl clean water and sanitation-SDG 6 et 11),smart grids and affordable/clean energy(SD
331、G 7),Environment,climate and biodiversity(SDG 13,14 et 15;corp dream“climate”);drug discovery(SDG3;corp dream“health”)sustainable agriculture(incl.Zero hunger-SDG2),66 Important notice:any of those challenge being open to using quantum computation,including energy-efficient quantum AI Argue against
332、potential adverse effects of the use-case,such as listed of UNESCO recommendation on the ethics of AISecond stage is rated on feasibility/credibility assessment of the proposed quantum program Dates and Process:Submission by October 25th,By October 27th validated maximum 20 projects can advance Deli
333、verable:description of the algorithm with arguments suggesting implementability on 2023-2024 PASQAL machines and later scalability analog(with potentially partial addressability functionalities)Scoring was based on Quality of the argument why the proposed method solves the claimed use-case,Quality of the proposed technical framework as well as feasibility as opposed to present technical capacities