]> Keio University

5322 Endo, Fujisawa Kanagawa 252-0882 JP rdv@sfc.wide.ad.jp

Keio University

5322 Endo, Fujisawa Kanagawa 252-0882 JP whit3z@sfc.wide.ad.jp

Keio University

5322 Endo, Fujisawa Kanagawa 252-0882 JP amin@sfc.wide.ad.jp

IRTF QIRG network architecture quantum repeater This quantum network architecture defines a set of planes providing different views of the network, supporting different responsibilities and modes of operation; a set of device, node and link types; some network topologies, deployment scenarios and their relationship to applications; and key design decisions as a result of corresponding requirements. About This Document The latest revision of this draft can be found at . Status information for this document may be found at . Source for this draft and an issue tracker can be found at .

Introduction This document introduces the key architectural decisions, classical and quantum communication systems, and main node types for several classes of quantum networks. We define the verb to architect as: within a set of environmental constraints, using a set of building blocks, design a system that satisfies a need, elegantly and economically. We use the noun architecture as: the set of blocks or subsystems, their roles and their interfaces and their overall arrangement, that defines the system. This architecture defines the overall structure, and is connected to a specific implementation as an example. For a description of the key concepts in quantum networks and additional references, see and the book Quantum Communications . For more background and discussion of the design choices in this architecture, see the Ph.D. dissertation of Naphan Benchasattabuse .

Requirements Language The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here.

Goals and Non-Goals of this Document This section describes goals and non-goals for this document itself, rather than the technical goals and requirements for a network.

Goals

• To define the principal concepts in a quantum network, principally for quantum multicomputer interconnects but also data center and wide-area networks where possible.
• To enumerate some of the key architectural decisions for this architecture.
• To describe how device, link and node types are defined.
• To provide a guide to other documents.

Non-Goals

• Specification of physical links (including photonic wave packets)
• Detailed definition of application-layer protocols and software interfaces such as sockets
• Internetworking

Relationship to Documents by QIRG and Other Organizations Other organizations, including national laboratories and standards development organizations, are developing documents describing quantum networks and quantum computing technology. These are mostly pre-standardization documents, not yet on any formal standardization track. To the extent possible, this document conforms to their terminology. However, as this document describes a specific quantum network architecture, it does not attempt to conform to specific design decisions made in other contexts. See an August 2025 Science Policy Forum for additional discussion of some standardization efforts and their value. Some of these are listed here for reference:

• ETSI
  • Industry Specification Group (ISG) on Quantum Key Distribution (QKD)
• IEEE Standards Association
  • IEEE Standards & Projects for Quantum Technologies
  • Standardization Roadmap on Quantum Applications
• ISO
  • IEC/ISO JTC 3 Quantum technologies
• ITU-T
  • ITU-T Focus Group on Quantum Information Technology for Networks (FG-QIT4N)
  • Y.3800 series on quantum key distribution networks
• National Institute of Standards and Technology (NIST)
  • Single-Photon Sources and Detectors Dictionary
• Quantum Internet Research Group (QIRG) (part of IRTF)

Prerequisite Knowledge This document assumes basic knowledge of the underlying technology and goals of quantum communications. The following list of topics may help readers who are not yet familiar with the concepts.

• Linear algebra
• Quantum information basics
  • Dirac ket notation
  • von Neumann density matrix notation
  • Superposition
  • Entanglement
  • Interference
  • Unitary operation
  • Measurement
  • Decoherence
  • No-cloning theorem
  • Clifford group
  • Basics of quantum error correction (QEC)
• Classical Internet-family networking
• Quantum networking
  • Teleportation
  • Purification
  • Entanglement swapping
  • Quantum key distribution , ,
  • "Generations" of quantum repeaters ,
  • (Repeater graph states may be helpful, but are not used in the current architecture , ,, )

Because terms such as fidelity have varying definitions, they will be defined in this set of documents . Readers needing additional background are referred to:

• Van Meter, Quantum Networking
• Hajdusek and Van Meter, Quantum Communications

Terminology In this document, we use the abbreviations and other related technical terms listed in the following table:

Term	Description
BSA	Bell state analyzer, generally optical and incorporating one or more beamsplitters and either two or four single-photon detectors
CRQC	cryptographically relevant quantum computer
device	manipulates photons in some fashion; a component of a node
FASQ	fault-tolerant application-scale quantum
fidelity	measures how close a quantum state is to the state we have tried to create. Varies between 0 and 1, with unit fidelity indicating the actual state is the same as the desired state. It expresses the probability that the state will behave exactly the same as our desired state. (adapted from RFC 9340)
FTQC	fault-tolerant quantum computer
group switch	In the Q-Fly architecture, the set of devices that connect the end nodes to the pool of BSAs, and the group to other groups
NISQ	near-term intermediate-scale quantum
node	a self-contained subsystem with a clear boundary that is visible to other such nodes as a single entity on one or more planes
optical mode	roughly, a place and time that a photon might be, which may be occupied by some, all or none of the amplitude of a single photon. More accurately, it is a field distribution obtained from solving Maxwell's equations given some boundary conditions. An optical mode is an eigensolution of the wave equation given the physical properties of the waveguide (dimensions, refractive index). It is a particular field distribution that can propagate through the waveguide. The state of a single photon can be expressed as a coherent superposition of such optical modes. For optics engineers, a 'mode' (or optical mode to be more precise) refers to a specific spatiotemporal distribution of electromagnetic field fluctuation. For quantum network engineers, especially those that would most likely be reading RFCs, two optical detectors are detecting different optical modes if one detector can be activated by light without activating the other. In that case, it is not wrong to say that the two detectors are detecting two independent optical modes.
path	the set of nodes and links that a connection passes through, including the end points
plane (architectural)	a view of the entire network for a particular purpose, such as management or data transmission
plane (optical)	a defined plane in physical space through which a photon passes; also referred to as a surface
QBER	quantum bit error rate
QNIC	quantum network interface card; in practice, the physical interface to a link plus the set of quantum memories under the control of the network stack.
QPU	quantum processing unit
RuleSet	a set of SDN-inspired, event-driven, short, real-time or near-real time, near-determininistic programs executed at nodes along a path to build application-requested entangled states
SNSPD	Superconducting Nanowire Single Photon Detector
multiqubit Pauli operators	tensor product of Pauli operators acting on two or more qubits
switch point	a 2x2 junction that can be either X (cross) or = (straight)
switch device	a single integrated, fiber- or free space-connected (physical) component, comprising one or more switch points
teledata	application execution via teleporting data from node to node, then executing gates locally. May be done remotely, mediated by Bell pairs.
telegate	application execution via remote gates (as defined by Eisert et al. ). May be done remotely, mediated by Bell pairs.
time bin	a method of qubit encoding using photons; compare to window and time slot
time slot	a fully specified time and place where a photon may be detected; compare to window and time bin
time window	the time interval where a detector is enabled; compare to time bin and time slot

Applications of Networks The requirements for a network are determined by the application workload. This section orients architectural decisions by briefly introducing applications and the communication patterns they exhibit, which is a key factor determining the suitability of particular architectures.

Types of Applications Applications fall into two large categories: inherently distributed applications, or subdivision of monolithic applications for distributed execution, as in supercomputing applications running on multicomputer architectures. The inherently distributed applications can be further categorized into cryptographic functions, distributed computation, and distributed sensing. For a discussion of some inherently distributed applications of quantum networks, see . This network architecture supports the applications listed in that RFC, though not all networks will support all applications.

Entangled States Consumption Patterns The text in this section is adapted and extended from unpublished text in . Distinct from the classification of quantum repeater generations by Muralidharan et al. , one can categorize distributed quantum systems by how applications interface with the network; specifically, the timing at which network interface qubits are freed after attempting entangled state generation. In the early days of quantum information research, Bennett et al. recognized that the component qubits of an entangled state may be held at different times in different locations, termed time-separated Bell pairs. Based on this principle, we can describe the timeline of information availability between two nodes. This model assumes entangled states are requested dynamically during execution, rather than pre-caching entangled states　 for immediate consumption.

The Entanglement Information Timeline The lifecycle of an entanglement request, from initiation to full state knowledge, follows three distinct stages:

1. Attempt: The entangled states are requested. For memory-based links, this corresponds to the quantum memory emitting a photon and transmitting it through the link.
2. Heralded: The entangled states are physically established, but the specific states are unknown. The node has received confirmations that photons arrived at the BSA and the BSM succeeded, but the Pauli frame information is not yet available. Without this information, an entangled state will be fully mixed, with a fidelity of 0.25 for a two-qubit state.
3. Correct: The classical message regarding the Pauli frame arrives. The node now knows the exact entangled state created and can apply corrections (or software frame updates) to align with the expected state.

Classification of Consumption Patterns Based on the timeline above, we classify Bell pair consumption into three classes . These classes are defined by whether the application must block execution while waiting for information at the Heralded or Correct stages. Note that the term "blocking" here refers to the blocking versus non-blocking execution models, similar to kernel-level I/O blocking or the event-driven programming paradigm, and is distinct from the concept of blocking in network switches.

Class	Wait for Heralding?	Wait for Pauli Frame?	Application Behavior
Unentangled State Tolerant (B)	No	No	Non-Blocking: The application consumes the qubit immediately, handling failures or corrections in post-processing.
Reactive Correction (C)	Yes*	No	Partially Blocking: The application waits only for confirmation of existence (heralding), then proceeds by tracking errors in software.
Deterministic (T)	Yes	Yes	Strictly Blocking: The application blocks until the state is fully verified and corrected.

*Note for Class C: If the link is fully deterministic (guaranteeing success), the application does not need to pause for heralding and become non-blocking.

Class Selection and Resource Implications It is important to note that these classes are determined by the nodes, not the network. The application nodes assess their own capabilities (number of buffer memories, gate and qubit error rate) and the network's performance (fidelity, rate, success probability) to select the appropriate operating class. The selection of a class significantly impacts network resource utilization. For example, if a node possesses sufficient buffer memory, it may elect to operate in Class B from the perspective of the network interface. By moving the entangled state from the communication qubit to storage immediately (or measuring it immediately), the node frees up the network interface to service other requests. This reduces the workload on the network and increases the repetition rate, even if the application logic itself eventually requires the data for a Class T operation.

Unentangled State Tolerant (B Class) Applications The defining characteristic of B Class applications is the ability to consume the (potential) entangled states immediately without waiting (fully non-blocking). The application effectively takes over the burden of validation from the network. Execution Flow: The application does not stop. It measures or stores the qubit immediately. If the entanglement attempt failed (information received later), the data is discarded or treated as an erasure error. Examples:

• Classical Correlation: Applications like Quantum Key Distribution (QKD) protocols (e.g., E91) or link fidelity estimation. The application filters out failed attempts during classical post-processing.
• Fault-Tolerant Operations: Certain remote quantum error correction schemes, such as remote lattice surgery , , , of the surface code. If the probability of creating entanglement is high enough, unsuccessful attempts can be treated as depolarizing errors, which the logical code can tolerate without stalling the pipeline.

Reactive Correction (C Class) Applications In C Class applications, the system requires confirmation that a link exists, but does not wait for the state details. The application proceeds by assuming a specific Bell state and managing deviations via software tracking. Execution Flow: The application pauses briefly to ensure the Bell pair is Heralded. Once confirmed, it executes immediately. It does not wait for the Correct stage (Pauli frame); instead, it uses a "Pauli Frame Tracker" to propagate the necessary corrections through the circuit virtually. If the network link is deterministic, the specific wait for heralding is removed, as the node assumes success by default. Examples: Clifford circuit execution, distributed Pauli-based computation with time-optimal scheme , and state teleportation.

Deterministic (T Class) Applications T Class applications impose the strictest timing constraints, requiring the entangled state to idle the longest before being consumed. Execution Flow: The application completely stalls. It must wait until the network provides both the confirmation of creation (Heralded) and the specific state information (Correct). Execution only resumes once the exact state is known or corrected. Examples: Execution of circuits involving non-Clifford gates. While non-Clifford operations can theoretically be corrected post hoc (similar to Class C), doing so often requires consuming additional entangled states to fix the error. Since consuming extra resources is more costly than waiting, these operations default to Class T to ensure the state is correct before proceeding.

Architectural Concepts This section introduces concepts in classical and quantum computer and network architecture that may be unfamiliar, or that have a specific role in this network architecture.

Quantum Devices A quantum device stores, carries, measures or computes on quantum variables. For the purposes of this architecture, a quantum device (often shortened to just "device" in this specification) is under software control; i.e. passive elements such as optical fibers or beam splitters are not classified as quantum devices. Control of devices is usually done with respect to some physical characteristic of the device itself, rather than dealing with the abstract notion of qubits. A controllable wave plate, for example, may be adjusted in units of degrees of rotation of the plate. An example of a software package that provides such functionality is PnPQ .

Quantum Nodes A quantum node comprises one or more quantum devices, and serves as a single locus of control for network protocols. The classes of nodes are described in detail in . A logical node or composite node provides a single communication and control point for the services of a particular node type, but may comprise a set of physical devices rather than a single device, and may be physically packaged in more than one box.

Quantum Links A quantum link provides entanglement suitable for one of the application classes described in . Links are described in .

Photonic Synchronization Domains A photonic synchronization domain (PSD) is the range of devices and channels (fiber or free space) over which photons must be controlled with high precision in order to effect e.g. photonic entanglement swapping . The primary concern of a PSD is getting photons to arrive at beamsplitters "simultaneously", with sufficient overlap in their wavepackets, as specified in . Note that the most common operational mode uses only pairs of photons, one from each of two nodes, not multi-photon operations; the definition of PSD does not imply that all channels in a network must be synchronized with each other and used in a single operation.

Direct and Indirect Multicomputer Architectures In multicomputer architectures, a direct architecture features links that go directly from computational node to computational node. Hypercubes, meshes and toruses are typically direct architectures. An indirect architecture interposes one or more switches between computational nodes. Fat trees, Clos and Benes networks, and the various -fly topologies are generally indirect . The distinction is somewhat artificial in that direct architectures sometimes incorporate a small switch (sometimes called a "router") inside the node, in which case the choice of term depends on where you draw the boundary of the node, and because computational nodes can be configured to act only as routers within the network, modeling an indirect architecture using direct hardware.

Detector-centric Architecture Several detectors may be packaged as a single subsystem for purposes such as cooling, power, control and time stamping. An architecture built around a shared pool of detectors is a detector-centric architecture. This structure facilities entanglement distribution in a quantum system interconnect, data center network or some forms of local area network. A detector-centric architecture is an indirect architecture if the pool of detectors is behind a switch or network of switches. If the detectors are arranged as Bell State Analyzers (BSAs), we can call the architecture BSA-centric. This will be the common case.

Source-centric Architecture Many of the detector-centric network topologies can be inverted, such that the pool of detectors is replaced by a pool of entangled photon pair sources. The network then is used to distribute entanglement, with the photons measured, absorbed into memories, interfered with locally generated photons, or otherwise utilized at or near the end nodes. Such a network was proposed by Drost et al. .

Qubit-Level v. Device-Level Access Conventionally, entangled quantum networks are described in terms of unitary and non-unitary actions on qubits, just as the Internet protocols are defined in terms of bits and bytes (octets) rather than voltages, signal times and optical signal strengths. This qubit-level access is natural and allows for technology-independent, high-level protocol development. However, as described above, quantum nodes are built of quantum devices, and at times it is preferable to use the hardware's device-level access. Such an access mode is inevitably hardware-dependent, requiring special-purpose software, limiting network scalability and code portability. Cases where device-level access is useful include:

• as a building block for higher-level protocols;
• during hardware development;
• during system installation or maintenance at a customer site;
• in link-level monitoring protocols; and
• when a multicomputer compiler and/or distributed-computation middleware wants to optimize the system for themselves, incorporating higher-level operations into its own application and using lower-level methods to access the system components on the wire.

Quantum Optical Building Blocks This section informally describes the physical building blocks and concepts used in the physical layer of a quantum network. These definitions will use equations defined in .

Qubits This network architecture uses only qubits, which may have two states identified as 0 and 1. Quantum information systems using qutrits, qudits, qunats or continuous variable (c.v.) quantum states are beyond the scope of the current set of specifications. Qubits (defined in RFC 9340) must conform to a sufficient subset of the DiVincenzo criteria .

Photons, Wave Packets and Optical Modes Loosely, a photon is a quantum of light. More detailed analysis of a photon requires understanding wave packets and optical modes. An optical mode is a well-defined state of a physical link, specified by path, time window, frequency, polarization, or similar parameters, such that the receiver can be configured to monitor the channel and determine whether that state is occupied by at least one photon or is empty. The mode exists regardless of whether a photon is present; a photon is an excitation of the mode, not the mode itself. (Perhaps a good analogy is the constellation of states in a quadrature amplitude-modulated (QAM) signal, combined with whatever other multiplexing or signal encoding is used, such as temporal, spatial, or wavelength. The constellation is the full state space, and a mode is the equivalent of one of those states, which may be empty or occupied.) A wave packet is the non-zero amplitude of the electrical and magnetic fields, and generally corresponds to a single excitation of the electrical field, i.e. a single photon. The length of the wave packet is determined by the physics of the photon emission process, possibly modified by dispersion in the channel, and in turn determines the minimum length of a time slot in the link design. In quantum networking, link capacity and state must be described in terms of modes, not photons; photons merely occupy modes, while empty modes correspond to vacuum states that are still physically and operationally meaningful. In idealized models, distinct modes correspond to orthogonal solutions of the field equations, ensuring perfect distinguishability. As an example, a quantum optical link may define one mode per time slot or detection window. During each window, the receiver monitors the mode. A detection event indicates that the mode was occupied by at least one photon; the absence of a detection indicates that the same mode was empty. Both outcomes correspond to distinct physical states of the link.

Photonic Qubits Photons are used in networks to carry qubits. There are several possible on-the-wire photonic qubit representations, e.g.

• Polarization (e.g. horizontal v. vertical)
• Time bin (e.g. early v. late)
• Which path (e.g. passing through a particular set of lenses or fibers)
• Energy level (wavelength)

These representations and more about the concept of an optical mode are discussed in the "standard photon" and documents, and are provided here only for informational purposes.

Memories The physical implementation of memories is beyond the scope of this document. For architectural purposes, it matters only that they may hold quantum data (qubits) that may be under the management of the network software and protocols. They may be entangled with photons or with other memory qubits. Generally, the creation or entanglement of a state in memory is imperfect, and the memory has a finite lifetime. This process of accumulated errors is decoherence. The entanglement of a memory qubit with a photon is a technology-dependent process. To create fiber-compatible and optical equipment-compatible photons, wavelength conversion via transduction may be necessary. In 2025, transduction is a low-probability process, and hurts fidelity as well.

Photon Sources Photons may be emitted by sources of many types . Single photons may come from attenuated lasers, or be emitted by a variety of quantum devices, such as quantum dots, or by individual atoms. Photons may be unentangled, entangled with other photons, or entangled with quantum memories.

Unentangled Single Photons Unentangled photons exhibit quantum properties. They can carry information in any of the characteristics listed above, and may be put into a superposition of multiple basis states for e.g. quantum key distribution purposes. In this document, unentangled individual photons are not used.

Entangled Photon Pairs Pairs of photons entangled with each other can be made via a variety of physical processes. Devices that make such pairs can be components of nodes such as the Entangled Photon Pair Source (EPPS), described in a separate document.

Memory-Emitted Photons Photons emitted by quantum memories, such as single atoms, may remain entangled to the memory, if the memory was in a superposition of basis states.

Detectors Detectors may be either single-photon detectors, which click when one or more photons hit the detector, or number resolving detectors, which can distinguish between one, two, or more photons hitting the detector within the same time window . In this document, detectors may be assumed to be single-photon detectors.

A Sketch of the System Model As noted in the 2022 roadmap for quantum interconnects , entangled quantum network technology can be deployed in a variety of scenarios with different requirements and assumptions. A full description of each of these is delegated to other documents, but a brief description here will help to orient discussions of design points in order to justify certain decisions.

Multicomputer The first deployment of production-level, distant quantum entanglement is likely to be in a quantum multicomputer , based on the same principles as classical distributed-memory supercomputers from the Caltech Cosmic Cube to Fugaku . Multicomputer deployments will likely involve computational nodes, optical switches, Bell state analyzers, and possibly entangled photon pair sources (all defined below). Quantum repeaters with memory are less likely to be deployed in multicomputers, though one such architecture has been proposed. Because the current technology roadmaps favor this type of deployment, where network design choices are in conflict or unclear, multicomputer designs are given priority over wide-area networks in this set of specifications. A multicomputer may be a noisy, intermediate-scale quantum (NISQ) system without quantum error correction , or may be a fault-tolerant system. Fault-tolerant systems are variously described as fault-tolerant quantum computers (FTQC), fault-tolerant application-scale quantum (FASQ) systems, or cryptographically relevant quantum computers (CRQC). FTQCs require the network to generate many more entangled states with tight timing requirements to allow distributed quantum error correction or the creation of fully error-corrected entangled states for application use. The program structural model is expected to be much like the classical supercomputing Message Passing Interface (MPI) , in the sense that nodes cooperate in a tightly coupled fashion, exchanging data and controlling program execution, with each node executing a portion of a single, large computation. One specific implementation of Quantum MPI has, in fact, been proposed , but here we mean this in the broader sense. General hardware environment:

• A multicomputer consists of a set of quantum nodes that are connected via quantum optical channels. The system may be either a direct (point-to-point, end node-to-end node) or an indirect (switched) design.
• Every quantum node has a set of quantum devices and a classical controller.
• End nodes are generally computational nodes, but measurement-only, sensor, and specialized memory storage nodes may be included.
• End nodes are capable of running application programs as well as executing the minimum operations to build end-to-end entangled states.
• End nodes may be built using multi-level hardware interconnects; when gates between physically distant qubits are mediated by first creating shared entangled states, the creation of those entangled states is the responsibility of the network subsystem.
• Multicomputer interconnects are closed systems, with no need for cryptographic security mechanisms.

Many aspects of compilation and job execution are beyond the scope of this set of specifications, but some points will affect the network node definitions and interfaces, so it is important to present them here to establish a basis for design decisions:

• The general purpose of a multicomputer system is to execute application programs that exceed the capabilities of a single quantum node. The subdivision of the application into smaller quantum programs and the assignment of those sub-programs to nodes in the network is beyond the scope of this specification, and may be either automatically done by the compiler or manually done by the programmer.
• Programs are centrally compiled and distributed to nodes. (Note that, technically, this is different from classical MPI, where a command is sent to worker nodes, but the mapping of that command to an executable program and versioning and distribution of programs are outside the scope of MPI. It is, however, a convenient assumption and common practice to ensure the same program is executed at all worker nodes.)
• Execution is coordinated by a job controller: every node executes the same program, but with different parameters (e.g., which portion of the problem to work on).
• Applications consist of both classical computation and quantum computation; within a node, the classical portion of the program delegates certain computational tasks to the quantum processor (sometimes called a QPU), similar to a classical coprocessor such as a GPU.
• Classical data related to quantum operations (principally measurement results and event notifications that trigger further actions) is sent peer-to-peer, not back to the centralized controller, during execution.
• Within the runtime system, the interface between the (portion of the) application running at each node's classical controller is analogous to the interface between an application and the MPI messaging system or a socket in an ordinary Internet application. This quantum socket is an active area of research and is not defined here.
• In principle, all nodes are running the same program, distributed to all nodes. Since the compiled application circuits are often parameter- or input-dependent as well, separate nodes may have separate instances of the application. This may result in a small additional burden on the execution management system.
• In principle, the application and the communication system are separately compiled and managed. However, in practice the RuleSet ( and ) may be compiled as part of the application by using a library of network functions. (As with classical parallel program runtime systems, the boundary between the application program, supplied libraries, and the kernel itself (if any) is implementation-dependent.)
• Compiling the network communication into the application program eliminates the need for separate program and RuleSet distribution protocols. However, the event messages that are part of the architecturally defined RuleSet operation are sent and received as usual, such that the behavior of the node is the same regardless of such implementation choices.
• Compilation and execution may achieve application goals via teledata, telegate or measurement of multiqubit Pauli operators transparently; the network is unaware of this distinction. Management of application-level variables and their movement from node to node, if any, is the responsibility of the compiler and is beyond the scope of this specification.
• The resources in a multicomputer may be partitioned to run multiple jobs , but this is beyond the scope of the current specifications.

The realities of quantum hardware result in a few important differences from classical multicomputers:

• Multicomputer interconnects may be either optically switched or composed entirely of point-to-point links. In switched networks, depending on optical hardware as well as photonic qubit representation, reconfiguring the switch or switches may be an extremely high-latency operation.
• Application execution at end nodes proceeds in phases tied to specific communication patterns. Even with a fixed sequence of application-level operations, the execution time of a phase can be variable because entanglement generation is probabilistic. Local gates can also be probabilistic under some circumstances, but for many node hardware types, fast local gate execution means the impact of variable gate execution time will be small.
• In switched systems, reconfiguration of switches is centrally coordinated between phases. Thus, while classical data moves directly node-to-node during autonomous phase execution, advancing from phase to phase in quantum multicomputer program execution must be done at the direction of the job controller.
• Because execution is centrally coordinated, the network system is not required to provide multiplexing. (This is a substantial difference from data center networks, QLANs and QWANs.)
• The compiler may include multihop communication within a phase using hop-by-hop teleportation without reconfiguring any switches; this is beyond the scope of this architecture.
• Execution of the quantum portion of the node program generally involves hard real-time actions, both unconditional and conditioned on prior quantum measurement results. This generally requires compilation of the quantum program to very low-level actions to be executed by FPGAs or ASICs.
• Systems may be partially or completely emulated, decoupling development of different subsystems. e.g., an EPPS plus a MEAS together can emulate a COMP node that emits single photons.
• Systems may be noisy, intermediate-scale quantum (NISQ); near-term, small-scale fault tolerant; or fault-tolerant, application-scale quantum (FASQ).
• Quantum error correction is above the level of these specifications, but may involve distributed lattice surgery , ,.

Data Center Network (QDCN) The hardware for quantum data center networks will be almost identical to multicomputers; the primary differences are in the workload, scheduling, programming model and assumptions of trust. Distributed control and protocols for multiplexing and connection setup may be necessary.

Local-Area Network (QLAN) A QLAN will be deployed within a building or across a campus. It may connect quantum computers (including but not necessarily multicomputers) and sensors. A sensor, for example, may be connected to a computer with substantial memory for e.g. shadow tomography, learning from few measurements and related protocols. A QLAN will have a less regular topology than a multicomputer or QDCN. Distance, latency, fidelity, and success probability will all vary on a per-link basis. Distributed control and protocols for multiplexing and connection setup are necessary .

Wide-Area Network (QWAN) Wide-area networks may involve client-server or peer-to-peer communication. One particular scenario of interest is a measurement-only (MEAS) client end node connecting to a centralized, supercomputer-scale quantum computer (perhaps, but not necessarily, a multicomputer) for the purposes of executing blind quantum computation , . QWAN client-server communication very likely will suffer from the "incast" problem of excessive traffic concentrating near certain nodes . Management of this problem is beyond the scope of this document.

Requirements and Goals for the Architecture This section documents the requirements for all networks adhering to this architecture.

Normative Requirements

• The architecture must support an open development process .
• The architecture must support the creation of robust, extensible, maintainable protocols in keeping with best current practice .

General Requirements

Functional Goals The following are goals of this architecture. Note that while the architectural decisions support all of these goals, individual implementations may not support every item in the list, i.e. a single-PSD implementation may not support memory-based entanglement swapping or may be insufficient for wide-area networking.

• The architecture is defined only for qubits. Qutrits, qudits, qunats and continuous-variable systems are out of scope of this architecture, except where physical or link layers present such physical variables as qubits.
• The architecture defines support for pairwise Bell pair creation between nodes with one or more of the B, C or T timing classes in . (The architecture does not define multipartite states as a service of the network; applications that require them are responsible for building them from Bell pairs.)
• The architecture is independent of physical implementation of memories, photonic data representations, etc.
• The architecture supports deployments ranging from multicomputer to wide area networks.
• The architecture supports multiple photonic synchronization domains, as either point-to-point or optically switched paths. The architecture must support some form of buffering between PSDs.
• The architecture supports entanglement swapping. (Note that single PSD deployments may not need entanglement swapping.)
• The architecture supports methods of assessing and managing errors, including photon loss and decoherence.
• The architecture aims to support evolution of single-photon, unentangled, single-purpose quantum key distribution networks (perhaps with trusted relay nodes) to fully entangled, multipurpose networks .

Interface Requirements

• Supports one or more applications, such as the ones in , with APIs consistent with the B, C, or T classes.
• The network must enable applications to match quantum states at each end of the Bell pair by name.

Physical Requirements Physical requirements such as distance, wavelength, vibration, power, etc. will be case-dependent.

Environmental Requirements Physical requirements such as distance, wavelength, vibration, power, etc. will be case-dependent.

Network Management Requirements Network management tasks largely revolve around installation, maintenance and fault isolation and correction. Links and nodes may also be enabled or disabled through automated or manual subsystems.

Configuration Management The architecture must support the use of both manual and automated network configuration tools to suit different installations and development processes. Configuration will include manual tasks such as physically setting up equipment, as described in . Configuration also includes setting up software: manually configuring node devices that do not autoconfigure; selecting, turning on, and configuring the quantum-related software services for the node; and configuring the underlying classical control operating system. Integration of these processes into existing or new network management tools is beyond the scope of this document.

Fault Management Quantum networks likely will be less stable than their classical counterparts for the foreseeable future. Hardware will require extremely low-level, hardware-dependent monitoring and adjustment. The architecture must:

• support isolation of hardware and software faults, and
• support monitoring and reporting of fidelity.

Integration of these processes into existing or new network management tools is beyond the scope of this document.

Online Operations Link and node monitoring tools, either automated or manual, can enable or disable links and nodes. For example, a link that falls below a certain fidelity or entanglement generation rate may be taken out of service for maintenance. Such an action must notify the routing subsystem, which in turn will relay the information via routing protocols. The behavior of connections passing through nodes or links whose status changes to disabled is currently undefined. Integration of these processes into existing or new network management tools is beyond the scope of this document.

Architecture Overview This section describes key design decisions made to achieve the goals outlined in . Details are provided in the following sections.

Deterministic Classical Control of Quantum States This network architecture is entirely classically controlled. Its task is to generate shared quantum states for applications residing at separate nodes. While many quantum events are inherently probabilistic, and loss of photons is also inherently probabilistic, this architecture does not use multipartite quantum states for e.g. routing of two-party requests. (An extension of this network architecture may support generation of multi-partite state for applications at a later date.) This control requires extensive classical communication, which is carried over a variety of channels , according to the requirements for various tasks .

Nodes, Links and Operational Views As with typical networks, this architecture comprises nodes () and links (), which will connect two or more nodes. While the end-to-end goal is the delivery of qubit-based entanglement, for a variety of purposes lower-level access to hardware can be desirable, so nodes may present a qubit-level view (incorporated in the data plane) or a device-level view (incorporated in the device plane).

Communication Service (Substantial portions of this section are adapted from Naphan Benchasattabuse's Ph.D. thesis, which in turn is adapted from earlier papers by Van Meter et al. and others.) The design of a quantum network must begin with a clear definition of its fundamental services --- what quantum states or capabilities the network is expected to provide to end users. These decisions determine the complexity of the protocols at the network layer and the applications that run above it. A minimalist design treats Bell pairs as the primary network-level service. Bell pairs serve as the smallest unit of entanglement and the foundation for nearly all quantum communication protocols. Restricting the service to Bell pair distribution simplifies the network's responsibilities. However, this approach shifts complexity to the applications, which must construct multipartite or fault-tolerant states themselves and manage the coordination overhead that entails. At the other end of the spectrum, networks may offer richer services such as multipartite entangled states , , , or fault-tolerant state teleportation. While applications can, in theory, synthesize these states from Bell pairs, direct network-level support may offer efficiency gains and reduce sensitivity to noise by internalizing complex procedures like direct graph state generations or supporting the delivery of error-correcting code encoded logical qubits. More concretely, this architecture supports the B, C, and T application classes described in , requiring both Bell pairs and accompanying timestamps for some applications.

Bell Pairs as the Core Network Service In our architecture, we adopt Bell pair distribution as the core network service, as it allows for a well-scoped, foundational architectural framework. Importantly, the semantics of Bell pair distribution are not merely those of passive delivery. Even in this basic model, distributed quantum computation occurs along the path via entanglement swapping, possibly combined with purification at intermediate repeater nodes. A proper service definition must account for this processing, as it directly affects fidelity, latency, and trust assumptions in the network.

Timing as a Critical Auxiliary Service In addition to quantum state delivery, timing information is often a critical part of the service. Applications in distributed quantum sensing and clock synchronization , , , , , require precise knowledge of when entanglement was established or when measurement events occurred. Hence, high-precision timestamps may need to be bundled into the service interface offered by the network.

Stateful Connections Along a Pre-established Path Connections in both 1G and 2G networks require entanglement swapping at the physical at logical level, respectively. Managing entanglement swapping to avoid the leapfrog problem and misrouted Bell pairs requires active, coordinated control of operations on qubits at multiple nodes in the network. In this architecture, this task is accomplished by establishing long-lived connections along a path selected at connection setup time. Each such connection consumes a small amount of classical memory at each repeater or router. Corresponding quantum resource management (buffering, multiplexing, etc.) is described below. Two connection setup mechanisms are supported: a two-pass mechanism suited for multi-hop communication, especially across technological or administrative boundaries, and a direct-control mechanism, best suited for data center-scale and locally-orchestrated communication. Authentication, authorization and accounting (AAA) and security concerns, including denial of service attacks, are beyond the scope of the present architecture.

Four-Layer Protocol Stack The protocol stack for creating end-to-end entangled states consists of four layers: physical, link, network, and application. The task of dealing with errors and ensuring reliable E2E creation, often handled in a transport layer that sits above the network layer, is subsumed into the network layer in this architecture.

RuleSet-based Network Layer The creation of sequences of end-to-end entangled states, roughly equivalent to TCP, is the responsibility of RuleSets, inspired by software-defined networking (SDN) and Active Networks . A RuleSet can also be viewed as something like a Berkeley Packet Filter (BPF) : it's a small program that handled actions that the application could do, but the application can't achieve the low, reliable latency to do it.

Five Planes, or Views, of the Network A plane is a view of the network, presented by the elements of the network for a particular purpose. This architecture provides five planes: device, quantum, data, control, and management, all described in .

Architectural Planes All nodes in the network will provide one or more of the following classes of interfaces, also referred to as planes. A plane provides a view of the network. Nodes that do not provide an interface are not visible on the corresponding plane.

• Device: Control and monitoring of hardware devices such as switches, wave plates, detectors, etc.
• Quantum: The quantum signals and in-channel, hardware-dependent, real-time classical signals for timing and synchronization of photon wave packets. A node with a quantum plane incorporates one or more quantum devices. The quantum devices may be local or remote.
• Data: Classical data plane for exchange of messages about in-progress quantum communication sessions. Generally, the data plane consists of event notifications and measurement results related to RuleSet-driven connections or testing sessions. A data plane must be accompanied by a control plane.
• Control: Classical control plane for establishing and managing connections and testing sessions. Routing and multiplexing messages are included in the control plane.
• Management: Network management for configuring the devices and monitoring operation. Managing services provided by nodes, security, addressing, etc., and monitoring health of links, collecting statistics on traffic through the node, any alerts such as security, etc.

Device This view is primarily for low-level development and maintenance and as a building block for higher-level protocols and software, as discussed in . No interpretation of the classical data as qubits or controls as qubit unitary or non-unitary operations is provided here. Software packages providing this level of access to the devices include PnPQ and tdc_toolkit . Generally, network-level access to this functionality will be via a form of RPC or streaming data.

Quantum The quantum signals and in-channel, hardware-dependent, real-time classical signals for timing and synchronization of photon wave packets. A node with a quantum plane incorporates one or more quantum devices. The quantum devices may be local or remote. The quantum plane functionality executes the functions described as "Interferometric Stabilization" and "Wave Packet Overlap" and is subject to the constraints in "Detector Timing Windows" in the Timing Regimes document .

Data When operating in qubit mode, the Data Plane consists of classical messages that convey events for RuleSet operation and the communication ports and software that generate and consume such messages. Data plane functions may share data with management plane functions as part of the link management process, e.g. using disti-mation . If this is done, security and privacy concerns must be addressed. The control plane functionality executes the functions described as "Pre-configured Event-driven Tasks" in the Timing Regimes document .

Control The classical control plane is responsible for establishing and managing connections and testing sessions. Routing and multiplexing messages are included in the control plane. Control plane functions may share data with management plane functions, e.g. sharing parameter adjustment values and timings. If this is done, security and privacy concerns must be addressed. The control plane functionality executes the functions described as "Measurement basis selection", "Optical switch control" and some tasks in "Host-side Application-level Tasks" in the Timing Regimes document .

Management While connection-specific changes to configuration, such as switching and necessary, immediate changes to e.g. polarization and optical delay may appear as control plane functions, slow-rate monitoring and adjustment of parameters such as timing or polarization due to drift in temperature, voltage or other parameters is the responsibility of the management plane. The management plane may receive useful data on the health and fidelity of links as a result of device plane, data plane and control plane operations. The management plane functionality executes the functions described as "Background Tasks" in the Timing Regimes document .

Protocol Layers This document describes a network architecture. Network designs are often described in terms of a layered protocol stack such as the 7-layer OSI model, although the Internet can be more accurately described as a three-, four-, or five-layer model, depending in part on whether a distinction is made between the physical and link layers and in part on how the application layer is subdivided. In a layered network architecture, each layer processes a prepended header to complete its task. As messages move down the stack (from application toward actual transmission), they may be subdivided into smaller units, and additional layer-specific headers are prepended. On receipt, headers are removed as the message moves up the stack, and boundaries between messages may be altered. A complete network architecture consists of much more than the layers processing individual packets; many of the critical supporting protocols around naming, routing, security, network management, etc. utilize messages carried using the same protocol stack designed for application data. In a quantum network, this layering is less clear. Classical network protocol layering is often presumed to include the sequential addition of headers while preparing a packet for transmission and processing proceeds down the protocol stack, with the process inverted on packet reception. In quantum networks, the qubits themselves are not processed in this stacked fashion, and the related classical information typically involves separate messages with separate end points as hop-by-hop entanglement is extended to end-to-end entanglement through the actions of nodes all along the path. Nevertheless, the notion of layers of responsibility or separation of concerns is valid,with different needs at the link, network, and application. (More to be added here.)

Protocol Layers

Physical Layer The physical layer incorporates the quantum channel itself, whether free space or waveguide. Specifcation of the physical layer is out of scope for this document, but the physical-layer specifications must include:

• signal transmission characteristics of the channel
• how nodes physically connect to the channel
• choice of photonic qubit
• wavelength
• wave packet envelope shape and duration
• timing/trial rate
• how link synchronization is achieved
• certain characteristics of detection subsystem

Additional details on how to specify a physical layer will be presented in documents on the "standard photon" and "photon train", which are currently being drafted.

Link Layer The primary service provided by the link layer is heralded, named entanglement across a quantum channel with defined endpoints , , , . The link layer sits below the network layer and uses physical-layer quantum phenomena plus classical messaging to provide its service. In this architecture, the link service provides only bipartite entanglement. The link layer architecture will support B, C and T class applications, though not all links are required to support all classes. In support of these classes, the entangled states may be either still active and available for further use by the network layer, or one or both of the qubits may already have been measured, with the measurement basis and results and possibly a high-precision timestamp recorded and reported. See p. 36 of for a description of the link service. They refer to K-type (keep), M-type (measure) and R-type (remote) entanglement requests. The link layer may deliver entangled states to the network layer either singly or as an ordered, tagged batch. The link may operate entirely on demand, or as an always-on, free-running service where requests from the network layer are serviced by using recently created entanglement. The network layer can request states at one end where the status of the qubit at the other end is unknown, in either heralding or Pauli frame. This is necessary to support B class applications. For multidrop links, the link layer must coordinate with multiplexing and switch control as part of the control plane . The link may consist of several network nodes. Besides the two endpoints, switches, BSAs, and EPPSes may be involved. Although these support nodes are part of the network architecture, their presence is not visible beyond the boundaries of the link or PSD. In this architecture, fidelity is the responsibility of the network layer; the link layer always provides raw Bell pairs of base fidelity. Thus, fidelity is part of neither the request to nor response from the link layer. During connection setup, information about link fidelity is acquired from the link management and used by the Responder to plan the connection. Thus, link monitoring and link fidelity management are the responsibility of the link architecture but are not part of the link layer protocol.

• "message" format
• addressing
• multiplexing for multi-drop links

Network Layer: RuleSets The network layer is responsible for both single- and multi-hop purification, and for entanglement swapping. The network layer subsumes the functions commonly divided into network and transport in many architectures. In this architecture, the network layer is a lightweight, restricted (not Turing complete) distributed computation platform known as RuleSets. RuleSets are distributed to nodes during the connection setup process . RuleSets themselves are a form of state, and they maintain internal state during the life of the connection. The syntax and semantics of RuleSets are described in .

Application Layer Definition of the application layer protocols is beyond the scope of this document.

Nodes and Node Types (Substantial portions of this section are adapted from Naphan Benchasattabuse's Ph.D. thesis, which in turn is adapted from earlier papers by Van Meter et al. and others.) The architecture of a quantum network is defined by its constituent nodes and their specialized functions. In our architecture, we classify nodes into three main groups: end nodes, for application interaction; repeater and router nodes, for extending entanglement and path management; and support nodes, for auxiliary operational tasks. The qNode specification (a document to be published) will provide additional details on the common roles and responsibilities of all quantum network nodes, and serves as the equivalent of the Internet hosts requirements RFCs , . Each node type is further defined in a detailed specification in a separate document.

End Nodes End nodes represent hosts that wish to execute a quantum application such as quantum key distribution, secret sharing and blind quantum computation , . The technological maturity required of an end node heavily depends on the desired application. There are four major kinds of end nodes: A measurement (MEAS) node is the most basic type of quantum end node, designed primarily for protocols that do not require quantum state storage. Its core capability is to receive individual photons and perform measurements on them in at least two different bases. Lacking quantum memory, MEAS nodes are well-suited for applications like quantum key distribution (QKD) or as simple terminals in certain forms of secure delegated computation protocols ,, typically interacting with the network in a synchronous manner where measurement results directly yield classical data. A sensor (SNSR) node is a specialized end node designed to utilize entangled states, often shared with distant parties, for high-precision measurements of physical quantities, for clock synchronization tasks, or for distributed sensing tasks , , , , . These nodes typically feature limited quantum memory to hold working qubits (e.g., one half of an entangled pair) and possess specific quantum processing capabilities tailored for sensing protocols. Such capabilities include performing joint measurements, like Bell State Measurements (BSMs), between their stored qubits and photons that have interacted with the environment , . While an SNSR node's internal processing is specialized, certain sensing applications may also necessitate high-rate entanglement generation from the network to achieve desired performance. For SNSR nodes, precise timing information is almost invariably a critical component of the service they provide or require, and their operation typically culminates in outputting classical data that corresponds to the sensed phenomenon. A store (STOR) node is a specialized end node whose primary function is to serve as a high-fidelity quantum data repository. Its core capabilities are the long-term storage of quantum states---often prepared and teleported from other locations---and the ability to teleport these states out on demand. While a STOR node does not require a universal gate set, it must support certain gates (e.g., Clifford gates) for active quantum error correction to preserve the stored quantum data. This includes using quantum error-correcting codes to protect against decoherence, along with mitigation strategies for correlated errors from events such as cosmic ray strikes , , , , , . In a network context, STOR nodes may function as data servers, enabling asynchronous applications where valuable states are prepared and stored for later retrieval. A computational (COMP) node represents a full-fledged quantum processing endpoint within the network. Equipped with quantum memory and additional algorithmic qubits, it can store, manipulate, and perform complex computations on quantum states received from the network or generated locally. COMP nodes support a wide range of advanced quantum network applications, including distributed quantum algorithms, more general forms of blind quantum computation, and potentially fault-tolerant quantum computing, often requiring asynchronous interfaces to coordinate their local quantum workloads with network operations , , , , , , , , , , , .

Support Nodes An entangled photon pair source (EPPS) is a device dedicated to generating pairs of entangled photons, commonly through processes like Spontaneous Parametric Down-Conversion (SPDC). These entangled photons are then typically distributed over quantum channels to be captured or measured at link endpoints, forming the initial resource for entanglement-based protocols. EPPS nodes can be deployed in various scenarios, including terrestrial fiber links or free-space satellite-to-ground communication , , , . A Bell state analyzer (BSA) is a crucial component for performing projective measurements on two incoming photons, ideally projecting their combined state into one of the four Bell states. BSAs are fundamental for realizing photonic entanglement swapping, a primary process that creates link-level entanglement, used particularly to convert memory-photon entanglement into memory-memory entanglement between distant quantum memories. The efficiency and complexity of a BSA depend on the optical implementation and the specific photonic qubit encoding used. A Repeater Graph State Source (RGSS) is a specialized source that generates multipartite entangled photonic states, specifically tailored for all-photonic (memory-less) quantum repeater architectures , ,, . An RGSS typically distributes segments of the generated repeater graph state to adjacent network nodes, where subsequent measurements on these photonic qubits are performed to establish long-distance entanglement without relying on quantum memories. An advanced Bell state analyzer (ABSA) represents a more sophisticated version of a BSA, particularly required in advanced all-photonic repeater protocols based on repeater graph states, ,, . Unlike basic BSAs, an ABSA must be capable of performing measurements on single or multiple photons in dynamically selectable bases. The choice of measurement basis often depends on the outcomes of prior measurements within the network and the specific structure of the repeater graph state being utilized, implying more complex hardware and real-time classical control logic. An optical switch (OSW) is a device that can passively route photons from input optical fibers or paths to different output paths without performing measurements on them . OSWs, which can be based on technologies like nanomechanical systems or nanophotonic circuits, can be integrated as components within other node types (e.g., routers or complex end nodes) or can function as standalone elements in the network to dynamically reconfigure optical pathways. An OSW node contains at least one switch device, which comprises at least once switch point.

Repeater Nodes A first-generation repeater (REP1) is a network node with two quantum interfaces, whose main task is to extend entanglement. Its primary operations include generating link-level Bell pairs with its neighbors, performing entanglement swapping to connect these segments, and managing errors through heralded entanglement purification on physical qubits along the connection path. A second-generation repeater (REP2) also focuses on entanglement swapping to bridge distances but generally requires a larger quantum memory capacity and higher fidelity local quantum operations than a REP1. It utilizes quantum error correction (QEC) to manage operational errors in conjunction with heralded link-level entanglement generation. Instead of, or in addition to, purifying link-level physical Bell pairs, a REP2 node operates on logical qubits, where quantum information is encoded across multiple physical qubits to protect it from errors. This approach inherently demands more sophisticated hardware and advanced computational capabilities for encoding, decoding, and error correction routines during swapping. A quantum router (RTR) is a more complex and versatile node, possessing all capabilities of a quantum repeater and typically featuring three or more quantum network interfaces, enabling it to make sophisticated path selection decisions in complex topologies. Architecturally, an RTR may consist of multiple line cards and a quantum backplane, allowing it to run a full suite of network operation protocols. Beyond basic repeating functions like entanglement swapping, an RTR can govern network borders potentially interfacing between different repeater generations or technologies, participate in generating multipartite entangled states if the network provides such a service , and may act as a Responder in connection setups by rewriting or generating new RuleSets for different network domains.

Composite Nodes A node may also aggregate the functions of more than one node, in a form known as a composite node or composite logical node. A common form of composite node is a switching BSA. When multiple devices of the same type are controlled by a single controller, they may be presented either as a node with multiple devices or as a composite node where each device is in turn represented as a node. Presenting as a single node is preferred.

Links

The Link Service A link provides Bell pairs across a single PSD. Each Bell pair is named via an identifier. This service may be either real time or batched. Generally, the physical link entanglement generation mechanism is probabilistic but heralded. Although it is possible to use repeated trials to present a near-deterministic service , in this architecture we choose instead to expose the asynchronous creation of Bell pairs with timestamps. (Does software control of buffer qubits appear in the architecture, or is it purely an implementation detail?)

Link/Network Layer Interface The link/network interface is best described as a software interface, executed separately at each link endpoint. The network-->link request:

• addresses of nodes
• single/fixed number/stream
• disposition: keep or measure, basis

The link-->network response:

• addresses of nodes
• length of array of entangled states
• array of entangled states
  • identifier for entangled state
  • (timestamp separate from identifier? needed or not?)
  • flags
    • confirmed/pending confirmation
    • measured/live
  • if measured, basis
  • if measured, result

Link Structures

Photonic Path Description The optical path over which photons flow from source to detector in the process of creating a Bell pair can be described using terminology that names the node types in the path; the direction of flow of photons can be inferred. This path description can be applied to point-to-point or switched links within a single PSD. Device type single-letter abbreviations and their corresponding node type:

• M: memory (COMP or STOR)
• S: source (of entangled photons) (EPPS)
• I: interference (i.e., Bell state measurement) (BSA)
• D: detector (i.e. a measurement node) (MEAS)
• X: switch (OSW)

Examples of path descriptions that may commonly appear:

• DSD: detector-source-detector
• MIM: memory-interference-memory
• MSM: memory-source-memory
• MM: memory-memory
• DSISD: detector-source-inteference-source-detector

At any point in a path except the ends, a photon may pass through one or more switches. In a switched architecture, for example, photons may pass through paths such as:

• DXSXD
• MXIXM
• MXXIXM

(Question: Does this notation also need to represent frequency conversion?)

Point-to-point Point-to-point links may be either fiber-based or free space. A link encompassing the path of one or more photons may be partially fiber and partially free space. The photonic path description notation will include no 'X's.

Switched A system built around a pool of detectors, particularly organized as BSAs, utilizing switched MIM links can also be characterized as a detector-centric architecture. For pseudocode for switching (routing) certain types of devices, see Koyama et al. . The photonic path description notation will include one or more 'X's.

Multidrop or Bus A multidrop link, or a bus, is a shared physical channel to which more than two stations may be attached.

Link Design Choices and Operational Settings Beyond the top-level architecture of the link, defining placement of subsystems (MIM, MSM, or MM links), and choice of photonic qubit (polarization, time bin, etc.), designers make many choices that affect performance, robustness, and interoperability. Many of these choices will remain implicit, but others will be reported or managed through link management software (below). Links depend upon other parameters of the physical channel and equipment, including:

• entanglement generation trial rate,
• driving laser power (e.g. for SPDC-based EPPS),
• duration and spacing of photon detection windows,
• detector amplifier settings, and
• other hardware-specific parameters.

Depending on the implementation, some of these parameters can be modified under software control, some can be monitored and reported, and some are only available to software if manually added to a configuration database. Note that this task is far more complicated in a switched or multidrop configuration than in a point-to-point configuration.

Installing and Bootstrapping a Link During installation and/or bootstrapping of a link (and monitoring during operation), qubit- and technology-appropriate reference frames must be established. For example, the sender and receiver, or sender and both receivers in the case of an EPPS-based link, must agree on "horizontal" and "vertical" when using polarization qubits or "early" and "late" when using time bin qubits. A time reference may also be required for timestamping events such as photon detection time . For computational and cryptographic applications, these reference frames are entirely arbitrary across a link or PSD; in fact, it only matters that the link participants can agree on two orthogonal states usable as a qubit, and can correctly agree upon matching of events such as photon detection. For sensing applications (including clock synchronization), depending on the application, these reference frames may need to be tied to physical ground truth or a standard reference. Whether the link reference frames are absolute or relative should be reported in link management information.

Spontaneous Changes to Link Conditions During ordinary operations, the physical parameters of the channel may change either gradually or suddenly. Gradual changes may be due to e.g. temperature changes over a 24-hour cycle. Sudden changes may be due to e.g. physical contact with or jostling of optical fibers or other vibration- and position-sensitive optical equipment. Detection of such changes is the responsibility of link management services (below). Note that this task is far more complicated in a switched or multidrop configuration than in a point-to-point configuration. Depending on operational needs and the severity of such changes, necessary adjustments may be conducted without suspending link operation or may require the link to be declared down long enough for recalibration to be conducted. In extreme cases, manual intervention may be required.

Link and Channel Management The first task of link management is naturally configuration of the link. During operation, the ongoing tasks include management of fidelity, trial rate, detection parameters, etc. Supplies information about link fidelity and rate to other subsystems such as routing and connection setup.

Connections No task involving quantum communication ever involves a single qubit or single entangled state. The connection provides the framework for managing the creation of an order set of entangled states to be consumed by applications. A connection is stateful at the end nodes. Nodes involved in the creation of end-to-end entanglement for those end nodes will be connection aware, meaning that they can identify resources and messages and carry out communication tasks necessary for a specific connection, but may not have substantial amounts of state that is dynamically updated on a per-action basis; any actions for nodes in this class must be idempotent or known to occur only once. Some or all nodes may be fully stateful, tracking the disposition of specific, named quantum states. Connections may be created using either a fully-distributed protocol or a centralized mechanism. In either case, qNodes involved in the connection receive RuleSets that are created by a single controller to coordinate local operations to build the end-to-end entangled states requested by an application. Connections are unaware of the shared use of resources and of other connections. Multiplexing is the responsibility of a separate subsystem, though connection setup should be done with awareness of the availability of unavailability of resources at involved nodes.

Application Layer

APIs for Network Service ("Quantum Sockets") The API used by classical software to interface with the quantum depends on which class of timing dependency pattern (B, C, or T) is to be supported.

Resource Management: Multiplexing and Routing Both link usage time slots and memory can be shared among multiple connections and therefore must be actively managed via a multiplexing system. This task is particularly challenging in switched networks. The discovery of the physical topology can be based on link state protocols, such as direct adaptation of OSPF or IS-IS . These protocols use a unitless link cost. It is known that, at least for certain physical systems, selecting a link cost corresponding to "seconds per Bell pair of a given fidelity" produces good correlation between low path cost and high throughput , . A number of approaches have been proposed .

Classical Communication A quantum network depends on classical communication; indeed, almost all of the behavior is governed by, initiated by, or managed and reported via classical messages and signals. These messages and signals have several key roles, described in the following subsections. See the Timing Regimes document for an outline of the physics driving these requirements. Classical communication takes three major forms:

• direct physical signal
• request for an action, formulated as a remote procedure call (RPC)
• a stream of events

Device Plane Device plane classical communications will be via RPC or streaming data service, as appropriate. Details are beyond the scope of this document.

Quantum Plane The quantum plane includes some classical signals to support tasks described in . These signals are generally hard real time, used for demarking per-qubit time bins and possibly the start and end of longer trains of qubits or photons for switching purposes, triggering emission of photons, etc. These signals may also be non-coding, involved in link management rather than creation of qubits. Not all quantum link designs will require such classical signals. Due to the hard real time nature of communications at this level, these signals are not carried over TCP/IP and indeed may be analog events rather than digitized signals. Links should be designed with soft-failure behavior such that synchronization and full link operation can be recovered without manual intervention after the loss of individual signal events or short outages. Where appropriate, the status of such classical communication should be reported to or made available to management tools as part of link health monitoring. The details at this level are generally link-specific and are beyond the scope of this architecture document.

Data Plane Classical communication within the data plane sits above the link, and concerns the creation of end-to-end entangled states for established connections, as described in .

• connection-level control of switching of photons
• soft real-time event notification for RuleSets

Messages in the data plane are generally soft real time and assumed to be reliable.

Control Plane Within the control plane, all communication is classical, supporting the establishment of connections and management of resources as described in .

• connection-level control of switching of photons
• soft real-time event notification for RuleSets

Management plane: Node and link management Within the management plane, all communication is classical, supporting the network monitoring and management of links and ordinary network node management, as described in .

• reporting of parameters fixed by device physics
• reporting and setting of parameters selectable by node configuration
• reporting and adjustment of slowly-changing parameters (such as polarization drift)
• typical network node management tasks such as software updates

Mechanisms

• RPC for some tasks
  • especially device-mode control
  • qRPC wrapper for classical RPC mechanisms
• message broker or event broker for other tasks
  • especially qubit-mode RuleSet notifications

Naming and Addressing

State naming and management in RuleSets (Incorporates naming material from Naphan's thesis. Probably needs updating to account for testbed realities.) Managing and agreeing upon the specific qubits for joint operations like entanglement purification or swapping among collaborating nodes are critical responsibilities handled by RuleSets. While it is possible to draw inspiration from the IP architecture by employing network-wide unique naming for qubits, RuleSets have local execution contexts, making global naming for individual qubits unnecessary. Instead, RuleSets use local logical names for qubits. At the lowest level, physical qubits within a QNIC can be addressed by the local node using a tuple like <QNICAddress,QubitIndex>. However, this level of detail is not, and should not be, visible to the RuleSet logic. Within a RuleSet instance on a given node, the RuleSet mechanism identifies a quantum resource primarily by its tag --- a label that categorizes the resource into a specific pool corresponding to its current role or stage in the protocol. To ensure an unambiguous and absolute ordering of multiple resources that may share the same tag (e.g., several Bell pairs awaiting purification), each resource, upon being allocated to a tag, is automatically assigned a sequence number by the local RuleSet engine. The combination <tag,seq no.> thus serves as a distinct and locally unique key within that RuleSet instance for referencing a resource and associating it with relevant metadata, such as its tracked fidelity. Actions specified within Rules typically reference these resources based on their tag and their relative order within that tag, for instance, by operating on the resource with the smallest sequence number (i.e., the oldest available in that pool). This structured internal naming is vital for deterministic local operations and for preventing local operational mismatches that could lead to problems like the leapfrogging issue if not handled carefully at a higher protocol design level. The RuleSet execution mechanism itself, therefore, only requires and manages this local <tag,seq no.> system for resource identification within a single node's RuleSet instance; it does not impose any inherent restrictions or requirements on how shared entangled states are identified between nodes. It is recommended that a strategy be used wherein the RuleSet creator (e.g., the Responder during connection setup) devises RuleSets such that the <tag,seq no.> is kept consistent across the RuleSets of all nodes involved in a particular joint operation on that resource. The responsibility thus lies with the RuleSet author to ensure that RuleSets are designed---potentially using this consistent naming strategy---to correctly manage shared resources and achieve the desired end-to-end protocol behavior. This minimal local naming scheme provided by the RuleSet engine, combined with judicious design of the RuleSets themselves, is sufficient for implementing a wide range of complex quantum network protocols while allowing creativity with minimal restrictions on RuleSet creator parts.

Example Networks While a full taxonomy of networks is neither desirable nor possible here, we present a few network examples using point-to-point links or switched architectures. In this section, the topology is briefly described, followed by analysis of the path characteristics of the shortest path and network diameter.

Fully Connected Point-to-Point Small networks may use nodes with multiple interfaces and provide a direct link between each pair of nodes, making a direct network. Assuming sufficient node-internal capabilities, such a network provides lowest loss (giving the fastest individual links) and the highest aggregate bandwidth, and is inherently nonblocking. However, with n-1 links per node and O(n^2) links in the network, its scalability is poor. It will always provide the highest performance, but not necessarily highest price/performance ratio. If links are midpoint interference-based, an issue in design is packaging of the BSAs and detectors. PSD: Each link is a separate PSD. All paths: MIM

Single Optical Switch A number of the early quantum multicomputer proposals assumed a single, large optical switch. Such a network is indirect. PSD: Although the entire network is a single PSD, each pairing is adjusted independently. All paths: MXIXM

Q-Fly Multicomputer An indirect interconnect. A multi-group, BSA-centric architecture. All nodes are part of the same PSD. The Q-Fly architecture is described in Sakuma et al. . PSD: Although the entire network is a single PSD, each pairing is adjusted independently. For DPFD topologies:

• Shortest paths
  • intra-group: MIXM
  • inter-group (network diameter): MXXIM
• Longer paths:
  • non-shortest path: MXXIXM or longer

For DPHD topologies:

• Shortest paths
  • intra-group: MXIXM
  • inter-group (network diameter): MXXIXM
• Longer paths:
  • non-shortest path: MXXXIXM or longer

For SPHD topologies:

• Shortest paths
  • intra-group: MXIXM
  • inter-group (network diameter): MXXIXM
• Longer paths:
  • non-shortest path: MXXXIXM or longer

Optically Switched Fat Tree An indirect interconnect. Several parameters are needed to describe the full topology of a fat tree, which can also be called a k-ary n-tree:

• k is the switch radix
• n is the tree depth

This simplest description assumes homogeneous hardware, where all switches have the same number of ports and all links are the same bandwidth. Leiserson's original fat tree proposed single links of increasing bandwidth at higher levels of the tree, giving the network its name; this approach provides no redundancy or path diversity, and achieving higher transfer rates is impractical in some technologies, including quantum. Consequently, most fat tree deployments use multiple links to several switches at higher levels of the tree, in a configuration that is also know as a folded Clos network. PSD: Although the entire network is a single PSD, each pairing is adjusted independently.

Repeater Fat Tree The repeater fat tree is described in . PSD: Each link is a separate PSD.

2-D Grid Multicomputer A direct interconnect. A 2-D grid of nodes, where nodes with memory and certain computational capabilities (canonically COMP nodes) have up to four interfaces connecting to neighboring nodes. Each node must act as a memory buffer and repeater to enable communication between non-neighboring nodes. PSD: Each link is a separate PSD.

Ring A direct interconnect. All nodes in a ring have exactly two neighbors. Each node must act as a memory buffer and repeater to enable communication between non-neighboring nodes. A ring is described in (something from Simon's group). PSD: Each link is a separate PSD.

QLAN A quantum local area network will have:

• irregular topology, possibly of heterogeneous link types
• distributed multiplexing
• distributed routing

PSD: If the QLAN does not include optical switches, but uses repeaters, each link is a separate PSD. If the QLAN includes optical switches, the PSDs may have irregular boundaries.

Security Considerations Quantum multicomputer systems are assumed to be constructed as isolated, centrally controlled systems with no need for confidentiality, integrity, and availability (the "CIA triad") assurance via cryptographic methods. Security considerations for other network types are an open topic of study and as such are not yet ready for specification and standardization . Similarly, the relationship to AAA (authentication, authorization and accounting) is an open topic for research.

References Normative References In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings. Informative References This RFC is an official specification for the Internet community. It incorporates by reference, amends, corrects, and supplements the primary protocol standards documents relating to hosts. [STANDARDS-TRACK] This RFC is an official specification for the Internet community. It incorporates by reference, amends, corrects, and supplements the primary protocol standards documents relating to hosts. [STANDARDS-TRACK] The Internet and its architecture have grown in evolutionary fashion from modest beginnings, rather than from a Grand Plan. While this process of evolution is one of the main reasons for the technology's success, it nevertheless seems useful to record a snapshot of the current principles of the Internet architecture. This is intended for general guidance and general interest, and is in no way intended to be a formal or invariant reference model. This memo provides information for the Internet community. This memo does not specify an Internet standard of any kind. The Internet community has specified a large number of protocols to date, and these protocols have achieved varying degrees of success. Based on case studies, this document attempts to ascertain factors that contribute to or hinder a protocol's success. It is hoped that these observations can serve as guidance for future protocol work. This memo provides information for the Internet community. This RFC attempts to document various aspects of the IP service model and how it has evolved over time. In particular, it attempts to document the properties of the IP layer as they are seen by upper- layer protocols and applications, especially properties that were (and, at times, still are) incorrectly perceived to exist as well as properties that would cause problems if changed. The discussion of these properties is organized around evaluating a set of claims, or misconceptions. Finally, this document provides some guidance to protocol designers and implementers. This document is not an Internet Standards Track specification; it is published for informational purposes. On 29 August 2012, the leaders of the IEEE Standards Association, the IAB, the IETF, the Internet Society, and the W3C signed a statement affirming the importance of a jointly developed set of principles establishing a modern paradigm for global, open standards. These principles have become known as the "OpenStand" principles. This document contains the text of the affirmation that was signed. This document is not an Internet Standards Track specification; it is published for informational purposes. The vision of a quantum internet is to enhance existing Internet technology by enabling quantum communication between any two points on Earth. To achieve this goal, a quantum network stack should be built from the ground up to account for the fundamentally new properties of quantum entanglement. The first quantum entanglement networks have been realised, but there is no practical proposal for how to organise, utilise, and manage such networks. In this document, we attempt to lay down the framework and introduce some basic architectural principles for a quantum internet. This is intended for general guidance and general interest. It is also intended to provide a foundation for discussion between physicists and network specialists. This document is a product of the Quantum Internet Research Group (QIRG). The main goal of the networking standards process is to enable the long-term interoperability of protocols. This document describes active protocol maintenance, a means to accomplish that goal. By evolving specifications and implementations, it is possible to reduce ambiguity over time and create a healthy ecosystem. The robustness principle, often phrased as "be conservative in what you send, and liberal in what you accept", has long guided the design and implementation of Internet protocols. However, it has been interpreted in a variety of ways. While some interpretations help ensure the health of the Internet, others can negatively affect interoperability over time. When a protocol is actively maintained, protocol designers and implementers can avoid these pitfalls. The Quantum Internet has the potential to improve application functionality by incorporating quantum information technology into the infrastructure of the overall Internet. This document provides an overview of some applications expected to be used on the Quantum Internet and categorizes them. Some general requirements for the Quantum Internet are also discussed. The intent of this document is to describe a framework for applications and to describe a few selected application scenarios for the Quantum Internet. This document is a product of the Quantum Internet Research Group (QIRG). QuTech, Delft University of Technology QuTech, Delft University of Technology QuTech, Delft University of Technology In a classical network the link layer is responsible for transferring a datagram between two nodes that are connected by a single link, possibly including switches. In a quantum network however, the link layer will need to provide a robust entanglement generation service between two quantum nodes which are connected by a quantum link. This service can be used by higher layers to produce entanglement between distant nodes or to perform other operations such as qubit transmission, without full knowledge of the underlying hardware and its parameters. This draft defines what can be expected from the service provided by a link layer for a Quantum Network and defines an interface between higher layers and the link layer. Keio University Keio University Entangling quantum networks build on new physical mechanisms to distribute quantum entanglement among a set of nodes over a set of links. To design a complete network protocol stack with proper division of responsibilities into layers, hardware and protocol engineers must share an understanding of those physical mechanisms and use a common vocabulary. This document bridges the abstract concepts described in [RFC9340] and the underlying physics to engineering concerns such as timing constraints on arrival of photons and exchange of supporting classical messages. The equations presented here will serve as reference points for architectural decisions in future documents, allowing future documents to deal directly in code without complex mathematics. Application-layer developers will not need the low-level physics presented here. Juniper Networks, Inc. Juniper Networks, Inc. QuTech Redbit Networks QuTech This document describes the use of link-state routing protocols on classical links in Quantum Networks. It contains proposals for additions to the IS-IS and OSPF protocols in order for them to transport relevant information for a Quantum Network, specifically, for the creation and manipulation of entangled pairs. The document will describe some of the necessary attributes and some suggestions of how this information may be used. No Schrodinger's cats were harmed in the creation of this document. University of Pennsylvania School of Engineering and Applied Science University of Pennsylvania School of Engineering and Applied Science University of Pennsylvania School of Engineering and Applied Science University of Pennsylvania School of Engineering and Applied Science University of Pennsylvania School of Engineering and Applied Science This document specifies the Quantum Datagram protocol a lightweight transport protocol designed to operate over UDP in IP optical environments. QDCP (formerly QFCP) enables the transmission of control- plane parameters required for transporting quantum information and associated optical configurations, including polarization stabilization, timestamp alignment, ROADM port selection, and spectral parameters. The protocol uses a Type-Length- Value (TLV) structure to support versioning and extensibility and is prototyped for the transport of third-order nonlinear generated quantum information on IP optical infrastructure. This work is motivated by recent demonstrations of a classical-decisive quantum internet using integrated photonics. National Institute of Standards and Technology National Institute of Standards and Technology, Gaithersburg, MD, USA Aliro Technologies, Inc., Brighton, MA, USA National Institute of Standards and Technology, Gaithersburg, MD, USA National Institute of Standards and Technology, Gaithersburg, MD, USA Institute of Electrical and Electronics Engineers (IEEE) Centre for Intellectual Property and Information Law, University of Cambridge, Cambridge, UK.; Centre for Law, Medicine, and Life Sciences, University of Cambridge, Cambridge, UK. School of Social Sciences and Technology, Technical University of Munich, Munich, Germany. Harvard Law School, Harvard University, Cambridge, MA, USA. Stanford Center for Responsible Quantum Technologies, Stanford University, Stanford, CA, USA. American Association for the Advancement of Science (AAAS) IEEE SPIE Keio University Argonne National Laboratory (ANL), Argonne, IL (United States); Univ. of Chicago, IL (United States) Univ. of Chicago, IL (United States) HRL Laboratories, LLC, Malibu, CA (United States) Univ. of Chicago, IL (United States) Univ. of Illinois at Urbana-Champaign, IL (United States) Argonne National Laboratory (ANL), Argonne, IL (United States) Stanford Univ., CA (United States) Univ. of Wisconsin, Madison, WI (United States) Univ. of Chicago, IL (United States) Cornell Univ., Ithaca, NY (United States) International Business Machines Corp., Armonk, NY (United States) Univ. of Illinois at Urbana-Champaign, IL (United States) Argonne National Laboratory (ANL), Argonne, IL (United States); Univ. of Chicago, IL (United States) Argonne National Laboratory (ANL), Argonne, IL (United States) Stanford Univ., CA (United States) Univ. of California, Santa Barbara, CA (United States) Univ. of Chicago, IL (United States) Univ. of Chicago, IL (United States) Stanford Univ., CA (United States) Univ. of Wisconsin, Madison, WI (United States) Univ. of Illinois at Urbana-Champaign, IL (United States) HRL Laboratories, LLC, Malibu, CA (United States) Boeing Aviation and Aerospace Component Manufacturing, Arlington, VA (United States) International Business Machines Corp., Armonk, NY (United States) Univ. of Illinois at Urbana-Champaign, IL (United States) Intel Labs, Hillsboro, OR (United States) Univ. of Wisconsin, Madison, WI (United States) Keio University, Tokyo (Japan) Quantum Opus, Plymouth, MI (United States) IBM, Yorktown Heights, NY (United States). Thomas J. Watson Research Center Univ. of Wisconsin, Madison, WI (United States) Stanford Univ., CA (United States) Univ. of Chicago, IL (United States) Argonne National Laboratory (ANL), Argonne, IL (United States) Argonne National Laboratory (ANL), Argonne, IL (United States) JPMorgan Chase, New York, NY (United States) Verizon, Richardson, TX (United States) Univ. of Wisconsin, Madison, WI (United States) Univ. of Chicago, IL (United States) Office of Scientific and Technical Information (OSTI) NTT Basic Research Laboratories, NTT Corporation, 3-1 Morinosato-Wakamiya, Atsugi, Kanagawa 243-0198, Japan and NTT Research Center for Theoretical Quantum Information, NTT Corporation, 3-1 Morinosato-Wakamiya, Atsugi, Kanagawa 243-0198, Japan Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA QuTech, Delft University of Technology, Lorentzweg 1, 2628 CJ Delft, The Netherlands and Networked Quantum Devices Unit, Okinawa Institute of Science and Technology Graduate University, Okinawa, Japan Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA and Quandela SAS, 10 Boulevard Thomas Gobert, 91120 Palaiseau, France Pritzker School of Molecular Engineering, The University of Chicago, Chicago, Illinois 60637, USA Quantum Bridge Technologies, Inc., 100 College Street, Toronto, Ontario M5G 1L5, Canada, Department of Physics, University of Hong Kong, Pokfulam, Hong Kong, and Center for Quantum Information and Quantum Control, Department of Physics and Department of Electrical and Computer Engineering, University of Toronto, Toronto, Ontario M5S 3G4, Canada Department of Physics, University of Toronto, Toronto, Ontario, Canada American Physical Society (APS) Springer Science and Business Media LLC Elsevier BV American Physical Society (APS) Keio University American Physical Society (APS) American Physical Society (APS) IEEE Instituto Superior Técnico, Universidade de Lisboa, Portugal; Physics of Information and Quantum Technologies Group, Centro de Física e Engenharia de Materiais Avançados (CeFEMA), Portugal; PQI – Portuguese Quantum Institute, Portugal Instituto de Telecomunicações, Portugal Instituto de Telecomunicações, Portugal; ISCTE - Instituto Universitário de Lisboa, Portugal Instituto Superior Técnico, Universidade de Lisboa, Portugal; Physics of Information and Quantum Technologies Group, Centro de Física e Engenharia de Materiais Avançados (CeFEMA), Portugal; PQI – Portuguese Quantum Institute, Portugal Verein zur Forderung des Open Access Publizierens in den Quantenwissenschaften American Physical Society (APS) Springer Science and Business Media LLC RAD Lab EECS Dept UC Berkeley, Berkeley, CA, USA RAD Lab EECS Dept UC Berkeley, Berkeley, CA, USA RAD Lab EECS Dept UC Berkeley, Berkeley, CA, USA RAD Lab EECS Dept UC Berkeley, Berkeley, CA, USA Intel Labs Berkeley, Berkeley, CA, USA ACM arXiv Springer Berlin Heidelberg Springer Science and Business Media LLC American Physical Society (APS) arXiv arXiv arXiv Springer Science and Business Media LLC Technical University of Delft Optica Publishing Group IOP Publishing American Physical Society (APS) American Physical Society (APS) American Physical Society (APS) Department of Computer Science, Stony Brook University, Stony Brook, NY, USA Argonne National Laboratory, Lemont, IL, USA Department of Computer Science, Stony Brook University, Stony Brook, NY, USA Department of Computer Science, Stony Brook University, Stony Brook, NY, USA Institute of Electrical and Electronics Engineers (IEEE) Elsevier BV IEEE Sorbonne Universit&#x00E9;,CNRS,Paris,France,F-75005 Mercari R4D, Mercari, Inc.,Japan Graduate School of Media and Governance,Kanagawa,Japan Graduate School of Media and Governance,Kanagawa,Japan Keio University,Faculty of Environment and Information Studies,Kanagawa,Japan Sorbonne Universit&#x00E9;,CNRS,Paris,France,F-75005 IEEE Springer Science and Business Media LLC American Physical Society (APS) arXiv Springer Science and Business Media LLC American Physical Society (APS) Microsoft Quantum, Switzerland Microsoft Quantum, Switzerland ETH Zürich, Switzerland Microsoft Quantum ACM Springer Science and Business Media LLC arXiv American Physical Society (APS) American Physical Society (APS) IOS Press Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA Sorbonne Université, CNRS, LIP6, F-75005 Paris, France American Physical Society (APS) Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA Department of Physics, Virginia Tech, Blacksburg, Virginia 24061, USA Verein zur Forderung des Open Access Publizierens in den Quantenwissenschaften IOP Publishing American Physical Society (APS) Springer Science and Business Media LLC Springer Science and Business Media LLC Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea Department of Electrical and Computer Engineering, Seoul National University, Seoul, Republic of Korea ACM Keio University,Quantum Computing Center,Kanagawa,Japan College of Agriculture and Life Science, Cornell University,Department of Information Science,New York,United States Graduate School of Media and Governance, Keio University Shonan Fujisawa Campus,Kanagawa,Japan Graduate School of Media and Governance, Keio University Shonan Fujisawa Campus,Kanagawa,Japan Graduate School of Media and Governance, Keio University Shonan Fujisawa Campus,Kanagawa,Japan Graduate School of Media and Governance, Keio University Shonan Fujisawa Campus,Kanagawa,Japan Graduate School of Media and Governance, Keio University Shonan Fujisawa Campus,Kanagawa,Japan Keio University,Quantum Computing Center,Kanagawa,Japan IEEE arXiv Springer Science and Business Media LLC Dahlem Center for Complex Quantum Systems, Freie Universität Berlin, Arnimallee 14, 14195 Berlin, Germany Verein zur Forderung des Open Access Publizierens in den Quantenwissenschaften Society for Industrial & Applied Mathematics (SIAM) Springer Science and Business Media LLC American Physical Society (APS) arXiv arXiv USENIX'93: Proceedings of the USENIX Winter 1993 Conference Proceedings on USENIX Winter 1993 Conference Proceedings American Physical Society (APS) Emerging Media Initiative, Kanazawa University,Kanazawa,Japan Graduate School of Engineering, The University of Tokyo,Department of Applied Physics,Tokyo,Japan Graduate School of Engineering, Science Osaka University,Osaka,Japan Mercari R4D, Mercari, Inc.,Tokyo,Japan Emerging Media Initiative, Kanazawa University,Kanazawa,Japan Graduate School of Media and Governance, Keio University,Fujisawa,Japan Keio University,Faculty of Environment and Information Studies,Fujisawa,Japan Mercari R4D, Mercari, Inc.,Tokyo,Japan IEEE American Physical Society (APS) Springer Science and Business Media LLC Zenodo Institute for Quantum Information and Matter and Walter Burke Institute for Theoretical Physics, California Institute of Technology, Pasadena CA 91125, USA Verein zur Forderung des Open Access Publizierens in den Quantenwissenschaften arXiv American Physical Society (APS) Springer Science and Business Media LLC Keio University arXiv arXiv Graduate School of Media and Governance, Keio University,Japan Graduate School of Media and Governance, Keio University,Japan Graduate School of Media and Governance, Keio University,Japan IEEE Institute of Electrical and Electronics Engineers (IEEE) Keio University arXiv California Institute of Technology, Pasadena Association for Computing Machinery (ACM) arXiv arXiv Nu Quantum Ltd.,Cambridge,United Kingdom Nu Quantum Ltd.,Cambridge,United Kingdom Nu Quantum Ltd.,Cambridge,United Kingdom Nu Quantum Ltd.,Cambridge,United Kingdom Nu Quantum Ltd.,Cambridge,United Kingdom IEEE Shahid Beheshti University Shahid Beheshti University Keio University IOP Publishing Zenodo Telemedia, Networks and Systems Group, MIT Telemedia, Networks and Systems Group, MIT Association for Computing Machinery (ACM) Keio University and CREST-JST, Japan Hewlett-Packard Laboratories, Bristol, United Kingdom National Institute of Informatics, Tokyo, Japan Keio University and CREST-JST, Japan Association for Computing Machinery (ACM) arXiv Springer Science and Business Media LLC Keio University Shonan Fujisawa Campus,Graduate School of Media and Governance,Kanagawa,Japan Keio University Shonan Fujisawa Campus,Graduate School of Media and Governance,Kanagawa,Japan Keio University Shonan Fujisawa Campus,Graduate School of Media and Governance,Kanagawa,Japan Keio University Shonan Fujisawa Campus,Graduate School of Media and Governance,Kanagawa,Japan current address: softbank,Wide Project Keio University Shonan Fujisawa Campus,Graduate School of Media and Governance,Kanagawa,Japan Keio University,Graduate School of Science and Technology,Kanagawa,Japan Mercari, Inc.,Tokyo,Japan Keio University Shonan Fujisawa Campus,Graduate School of Media and Governance,Kanagawa,Japan IEEE Wiley Institute of Electrical and Electronics Engineers (IEEE) Keio University Keio University Near-term quantum networks will grow to form a Noisy, Intermediate- Scale Quantum Internet (NISQI). Connection setup will require adapting behavior along the path to the noise levels of individual elements. In this proposal, path creation is triggered by an application at the Initiator, information is accumulated node-by-node on an outbound pass in a series of QCap (quantum capability) blocks, then the RuleSets are created at the Responder. RuleSets are installed at the individual nodes on the return pass. This document describes the architecture of connection setup in a network. Details of the RuleSets and QCaps, addressing architecture, link protocols, routing, resource allocation (multiplexing), extension of this setup procedure to an internetwork, and extension to multiparty communications are beyond the scope of this document. Springer Science and Business Media LLC QuTech, Delft University of Technology, Post Office Box 5046, 2600 GA Delft, Netherlands. QuTech, Delft University of Technology, Post Office Box 5046, 2600 GA Delft, Netherlands. QuTech, Delft University of Technology, Post Office Box 5046, 2600 GA Delft, Netherlands.; Kavli Institute of Nanoscience, Delft University of Technology, Post Office Box 5046, 2600 GA Delft, Netherlands. American Association for the Advancement of Science (AAAS) University of Chicago University of Chicago University of Chicago University of Chicago University of Chicago University of Chicago University of Chicago University of Chicago University of Chicago; University of Chicago University of Chicago University of Chicago; University of Chicago University of Chicago University of Chicago University of Chicago University of Chicago; University of Chicago University of Chicago; Rensselaer Polytechnic Institute University of Chicago; Argonne National Laboratory American Physical Society (APS) Pritzker School of Molecular Engineering, The University of Chicago, Chicago, Illinois 60637, USA Pritzker School of Molecular Engineering, The University of Chicago, Chicago, Illinois 60637, USA Pritzker School of Molecular Engineering, The University of Chicago, Chicago, Illinois 60637, USA Pritzker School of Molecular Engineering, The University of Chicago, Chicago, Illinois 60637, USA Graduate School of Media and Governance, Keio University, 5322 Endo, Fujisawa 252-0882, Japan Faculty of Environment and Information Studies, Keio University, 5322 Endo, Fujisawa 252-0882, Japan Pritzker School of Molecular Engineering, The University of Chicago, Chicago, Illinois 60637, USA; Center for Molecular Engineering and Material Science Division, Argonne National Laboratory, Lemont, Illinois 60439, USA Pritzker School of Molecular Engineering, The University of Chicago, Chicago, Illinois 60637, USA American Physical Society (APS) Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China.; Key Laboratory of Space Active Opto-Electronic Technology, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Key Laboratory of Space Active Opto-Electronic Technology, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, China. Key Laboratory of Space Active Opto-Electronic Technology, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, China. Key Laboratory of Space Active Opto-Electronic Technology, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, China. Key Laboratory of Optical Engineering, Institute of Optics and Electronics, Chinese Academy of Sciences, Chengdu 610209, China. Key Laboratory of Optical Engineering, Institute of Optics and Electronics, Chinese Academy of Sciences, Chengdu 610209, China. Shanghai Engineering Center for Microsatellites, Shanghai 201203, China. Key Laboratory of Space Object and Debris Observation, Purple Mountain Observatory, Chinese Academy of Sciences, Nanjing 210008, China. Xinjiang Astronomical Observatory, Chinese Academy of Sciences, Urumqi 830011, China. Yunnan Observatories, Chinese Academy of Sciences, Kunming 650011, China. Shanghai Engineering Center for Microsatellites, Shanghai 201203, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China.; Key Laboratory of Space Active Opto-Electronic Technology, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China.; Key Laboratory of Space Active Opto-Electronic Technology, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, China. Department of Modern Physics and Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, China.; Chinese Academy of Sciences (CAS) Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics, University of Science and Technology of China, Shanghai 201315, China. American Association for the Advancement of Science (AAAS) arXiv Lawrence Berkeley National Laboratory University of California,Berkeley University of California,Berkeley Lawrence Berkeley National Laboratory California Institute of Technology University of California,Berkeley University of California,Berkeley Lawrence Berkeley National Laboratory California Institute of Technology University of California,Berkeley University of California,Berkeley California Institute of Technology Lawrence Berkeley National Laboratory Lawrence Berkeley National Laboratory California Institute of Technology University of California,Berkeley University of California,Berkeley IEEE American Physical Society (APS)

Contributors Keio University

5322 Endo, Fujisawa Kanagawa 252-0882 JP

Michal was involved in document development and technical discussions from the beginning. Keio University

5322 Endo, Fujisawa Kanagawa 252-0882 JP

Andrew leads the software team implementing much of the network, and provides feedback on system structure. Keio University

5322 Endo, Fujisawa Kanagawa 252-0882 JP

Contributed presentations and clarity; reviewed this document; authoring related specifications. Keio University

5322 Endo, Fujisawa Kanagawa 252-0882 JP

Contributed presentations and clarity; reviewed this document and related documents; technical and managerial leadership. National Institute for Informatics

2-1-2 Hitotsubashi, Chiyoda-ku Tokyo 101-8430 JP

Technical and managerial discussions.