NANOPHOTONIC QUANTUM MEMORY

Abstract

Systems and methods are disclosed for making a quantum network node. A plurality of scoring function F values are calculated for an array of at least two photonic crystal cavity unit cells, each having a lattice constant a and a hole having a length Hx and a width Hy. A value of a, a value of Hx, and a value of Hy are selected for which a scoring function value is at a maximum. A waveguide region and the array of at least two photonic crystal cavity unit cells based on the selected values are formed on a substrate. At least one ion between a first photonic crystal cavity unit cell and a second photonic crystal cavity unit cell are implanted and annealed into a quantum defect. A coplanar microwave waveguide is formed on the substrate in proximity to the array of at least two photonic crystal cavity unit cells.

Description

TECHNICAL FIELD

The invention relates to quantum memory, and more particularly to quantum network nodes and quantum repeaters.

COPYRIGHT NOTICE

This disclosure may contain material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure as it appears in the U.S. Patent and Trademark Office patent file or records, but otherwise reserves any and all copyright rights.

BACKGROUND

The realization of quantum networks is one of the central challenges in quantum science and engineering with applicability to long-distance communication, non-local sensing and metrology, and distributed quantum computing. Practical realizations of such networks require individual nodes with the ability to process and store quantum information in multi-qubit registers with long coherence times, and to efficiently interface these registers with optical photons. Cavity quantum electrodynamics (QED) is an approach to enhance interactions between atomic quantum memories and photons. Trapped atoms in optical cavities are one type of cavity QED platforms for quantum processing and have demonstrated gates between atoms and photons as well as interactions between multiple qubits mediated by the optical cavity.

SUMMARY

Systems and methods are disclosed for making a quantum network node. In some embodiments, the methods include calculating a plurality of scoring function F values for an array of at least two photonic crystal cavity unit cells, each photonic crystal cavity unit cell having a lattice constant a and a hole having a length H_x and a width H_y, wherein the scoring function comprises:

F=min(Q,Q_cutoff)/(Q_cutoff×V_mode)

wherein Q is a cavity quality factor, Q_cutoff is an estimated maximum realizable Q, and V_mode is a cavity mode volume; selecting a value of a, a value of H_x, and a value of H_y for which the scoring function value meets a scoring function value criteria; forming, on a substrate, a waveguide region and the array of at least two photonic crystal cavity unit cells based on the selected value a, the selected value H_x, and the selected value H_y; implanting at least one ion between a hole of a first photonic crystal cavity unit cell and a second photonic crystal cavity unit cell; annealing the at least one implanted ion into at least one quantum defect; and forming a coplanar microwave waveguide on the substrate in proximity to the array of at least two photonic crystal cavity unit cells.

In some embodiments, the scoring function value criteria include one or more of a maximum scoring function value of the plurality of scoring function F values, a threshold value exceeded by at least one of the plurality of scoring function F values, and a maximum scoring function value of the plurality of scoring function F values after a predetermined number of iterations calculating scoring function F values.

In some embodiments, the methods can include tapering an input end of the waveguide region; connecting the tapered input end of the waveguide region to a tapered optical fiber to optically couple the optical fiber to the array of at least two photonic crystal cavity unit cells.

In some embodiments, the methods can include coupling the tapered optical fiber to at least one probing light source and to at least one single photon detector.

In some embodiments, the substrate includes a diamond substrate. In some embodiments, the implanted ion is a silicon ion and wherein the at least one quantum defect is a silicon-vacancy color center. In some embodiments, Q_cutoff is not greater than 5×10⁵.

In some embodiments, the methods include forming a mask, after the forming the array of at least two photonic crystal cavity unit cells, on the array of at least two photonic crystal cavity unit cells; etching, with an ion beam, the substrate, at two or more angles, to form a tapered column between the hole of a first photonic crystal cavity unit cell and the second photonic crystal cavity unit cell.

In some embodiments, the methods include forming a mask on the array of at least two photonic crystal cavity unit cells; forming at least one aperture in the mask at a location between the hole of a first photonic crystal cavity unit cell and the second photonic crystal cavity unit cell; and implanting the at least one ion through the at least one aperture in the mask, wherein ions are not implanted through the mask.

In some embodiments, the methods include comprising mounting the substrate in a refrigeration unit, wherein the refrigeration unit is configured to cool the substrate to less than 100 mK such that the spin coherence time T₂ of the at least one quantum defect is extended.

In some embodiments, the systems can include a substrate; an array of at least two photonic crystal cavity unit cells on the substrate, wherein each photonic crystal cavity unit cell has a lattice constant a and a hole having a length H_x and a width H_y, wherein a value of a, a value of H_x, and a value of H_y are selected so that a scoring function F value meets a scoring function value criteria, and wherein the scoring function comprises:

F=min(Q,Q_cutoff)/(Q_cutoff×V_mode)

wherein Q is a cavity quality factor, Q_cutoff is an estimated maximum realizable Q, and V_mode is a cavity mode volume; at least one quantum defect in the substrate between a first photonic crystal cavity unit cell in the array of at least two photonic crystal cavity unit cells and a second photonic crystal cavity unit cell in the array of at least two photonic crystal cavity unit cells; and a coplanar microwave waveguide disposed on the substrate in proximity to the array of at least two photonic crystal cavity unit cell.

In some embodiments, the systems include at least one probing light source and at least one single photon detector.

In some embodiments, methods of operating the quantum network device include receiving, with the at least one single photon detector, at least two photons; and in response to the receipt of two photons, measuring the state of the quantum defect using the probing light source. In some embodiments, the methods include a Bell-state measurement.

In some embodiments, methods of encoding of quantum information using the quantum network device, include: for n time-bin qubits comprising n+1 optical pulses, applying phase control with a phase modulator to each optical pulse, wherein each time-bin qubit stores quantum information in a relative amplitude and phase between a pair of neighboring optical pulses among the n+1 optical pulses; guiding the n+1 optical pulses to the at least one quantum defect; alternating, with each pulse, coherent microwave control of the quantum defect; and interfering, with a time-delay interferometer, each pulse with a previous optical pulse, wherein the time-delay interferometer delays the previous optical pulse by the time between the pulse and the previous optical pulse.

These and other capabilities of the disclosed subject matter will be more fully understood after a review of the following figures, detailed description, and claims. It is to be understood that the phraseology and terminology employed herein are for the purpose of description and should not be regarded as limiting.

BRIEF DESCRIPTION OF THE FIGURES

For a more complete understanding of various embodiments of the disclosed subject matter, reference is now made to the following descriptions taken in connection with the accompanying drawings, in which:

FIG. 1 shows schematics and a scanning electron micrograph of a SiV-nanophotonic quantum register, according to some embodiments of the present disclosure;

FIG. 2 shows schematics of level structure, spin-dependent reflection spectrum, and spin coherent time of SiV spin-cavity system, according to some embodiments of the present disclosure;

FIG. 3 shows schematics for heralded photo storage and a schematic for spin-photon storage fidelity, according to some embodiments of the present disclosure;

FIG. 4 shows schematics of an SiV coupled to nearby ¹³C nuclear spins, Ramsey measurement, spin echo on ¹³C, and reconstructed amplitudes for a CNOT gate transfer matrix, according to some embodiments of the present disclosure;

FIG. 5 shows schematics of a quantum network and a quantum network node based on the SiV, according to some embodiments of the present disclosure;

FIG. 6 shows schematics of a nanofabrication process used to produce devices and scanning electron micrographs corresponding to steps in the fabrication procedure, according to some embodiments of the present disclosure;

FIG. 7 shows schematics of the example set-up and fiber network used to probe devices, according to some embodiments of the present disclosure;

FIG. 8 shows schematics of SiV level diagram, qubit frequency for different strained emitters, optical splitting, histogram of MW transition frequency for two different emitters, histogram of optical transition frequency for two different emitters, and simultaneous measurement of f_↑↓ and f_↑↑, according to some embodiments of the present disclosure;

FIG. 9 shows a SiV-cavity reflection spectrum at several detunings, spin-dependent reflection for large SiV-cavity detuning, spin-dependent reflection near resonance, and a probe at the frequency of maximum contrast, according to some embodiments of the present disclosure;

FIG. 10 shows example schematics for microwave control, microwave-induced heating of devices, modeled temperature at the SiV, effect of microwave heating on SiV coherence time, and Hahn-echo for even lower Rabi frequencies, according to some embodiments of the present disclosure;

FIG. 11 shows schematics of T2 scaling for two different SiVs, DEER ESR on SiV 2, DEER Echo on SiV 2, and dynamical decoupling on SiV 1, according to some embodiments of the present disclosure;

FIG. 12 shows schematics of example sequences for generating and verifying spin-photon entanglement, spin-photon correlations measured in the Z-Z basis, spin-photon correlations measured in the X-X basis, and preparation of second spin-photon Bell state, according to some embodiments of the present disclosure;

FIG. 13 shows schematics of XY8-2 spin echo, trajectory of ¹³C on the Bloch sphere during a maximally entangling gate, maximally entangling gates, tuning up an initialization gate, nuclear Ramsey measurement, electron-nuclear correlations measured in the ZZ-basis, and electron-nuclear correlations measured in the XX-basis, according to some embodiments of the present disclosure;

FIG. 14 shows a schematic of RF Rabi oscillations, and SiV coherence in the presence of an RF drive, according to some embodiments of the present disclosure;

FIG. 15 shows schematics of unit cell of a photonic crystal cavity, electric field intensity profile of the TE mode inside the cavity, a photonic crystal design, and a plot of a, Hx, and Hy for the cavity, according to some embodiments of the present disclosure;

FIG. 16 shows plots of spectral diffusion of SiV 2 and SiV 1 in nanostructures, according to some embodiments of the present disclosure;

FIG. 17 shows schematics of original initialization sequence, simplified initialization sequence, simulated performance of the initialization gate, and simulated performance, according to some embodiments of the present disclosure;

FIG. 18 shows schematics of memory-enhanced quantum communication, according to some embodiments of the present disclosure;

FIG. 19 shows schematics of memory-assisted implementations of Charlie's measurement device, reflection spectrum of memory node, histogram of detected photon numbers, a schematic of spin-photon quantum logic operation, and characterization of resulting spin-photon correlations in the ZZ and XX bases, according to some embodiments of the present disclosure;

FIG. 20 shows schematics of asynchronous Bell-state measurements using quantum memory, according to some embodiments of the present disclosure;

FIG. 21 shows a plot of performance of memory-assisted quantum communication, according to some embodiments of the present disclosure;

FIG. 22 shows schematics of an example system set-up, according to some embodiments of the present disclosure;

FIG. 23 shows plots of characterization of device cooperativity, according to some embodiments of the present disclosure;

FIG. 24 shows plots of microwave characterization of spin-coherence properties, according to some embodiments of the present disclosure;

FIG. 25 shows schematics of measurement on a single time-bin qubit in Z and X bases, according to some embodiments of the present disclosure; and

FIG. 26 shows plots of performance of memory-device versus of channel loss, according to some embodiments of the present disclosure.

DESCRIPTION

Quantum networks are interconnected networks of quantum nodes that can store and/or process information. Quantum networks can be used in a plethora of different technologies, including secure communication, enhanced metrology, and distributed quantum computing (e.g., for processing and/or storage). In such networks, it's useful for to implement network nodes which perform quantum processing on a small register of interconnected qubits with long coherence times. In some examples, distant nodes are connected by efficiently interfacing qubits with optical photons that can be coupled into an optical fiber.

One strategy for engineering an efficient, coherent optical interface is that of cavity quantum electro-dynamics (QED) based on the interaction of photons and photonic cavities. Such cavity QEDs can be designed to enhance the interactions between atomic quantum memories and photons. Nanophotonic cavity QED systems are particularly appealing, as the tight confinement of light inside optical nanostructures enables strong, high-bandwidth qubit-photon interactions. In practice, nanophotonic devices also have a number of technological advantages over macroscopic optical cavities, as they can be fabricated en masse and interfaced with on-chip electronics and photonics, making them suitable for scaling up to large-scale networks. While strong interactions between single qubits and optical photons have been demonstrated in a number of cavity QED platforms, existing implementations lack many other properties that are useful for applications as a quantum network node. For example, simultaneously achieving high-fidelity, coherent control of multiple long-lived qubits inside of a photonic structure has remained challenging. Embodiments of the present disclosure describe systems, methods of manufacturing, and methods of operation that can combine one or more of these properties in a single system and/or method.

Quantum networks can be implemented using functional nodes consisting of stationary registers with the capability of high-fidelity quantum processing and storage, which efficiently interface with photons propagating in an optical fiber. Quantum defects can be used in nodes for quantum network applications. In some embodiments, quantum defects can include silicon-vacancy color-center in diamond (SiV). The SiV is an optically active point defect in the diamond lattice. Its D_3d inversion symmetry results in a vanishing permanent electric dipole moment of the ground and excited states, rendering the transition insensitive to electric field noise typically present in nanostructures. SiV centers in nanostructures can display strong interactions with single photons and that SiV centers at temperatures below 100 mK (achievable in dilution refrigerators) can exhibit long coherence times.

Embodiments of the present disclosure describes aspects of such nodes using quantum defects such as a diamond nanocavity with an embedded silicon-vacancy (SiV) color center and a proximal nuclear spin. For example, embodiments of the present disclosure provide systems, methods of manufacturing, and methods of operation that provide for efficient SiV-cavity coupling (with cooperativity C>30) providing a nearly deterministic interface between photons and the electron spin memory, featuring coherence times exceeding 1 ms.

Employing coherent microwave control, embodiments of the present disclosure demonstrate storage of heralded single photons (e.g., pairs of single photons generated in highly correlated states from using a single high-energy photon) in the long-lived spin memory as well as a universal control over a cavity-coupled two-qubit register consisting of a SiV and a proximal ¹³C nuclear spin with nearly second-long coherence time. In some embodiments, the coherence time can be in the range of 100 μs to 1 s. In some embodiments, the coherence time can be in the range of 500 μs to 1 s. In some embodiments, the coherence time can be in the range of 1 ms to 1 s. In some embodiments, the coherence time can be in the range of 2 ms to 1 s. Such long-lived coherence times have applicability to implementations in, for example, quantum repeaters (e.g., a chain of entangled quantum nodes that can be used, for example, for long-distance communications).

Embodiments of the present disclosure also describe memory-enhanced quantum communication. In some embodiments, one can use a single solid-state spin memory integrated in a nanophotonic diamond resonator to implement asynchronous photonic Bell-state measurements. Such solid-state spin memories can be used as quantum repeaters and in large-scale quantum networks.

Embodiments of the present disclosure describe a quantum network node for applications in spin memory and memory-enhanced quantum communication. In some embodiments, the quantum network node includes an array of at least two photonic crystal cavity unit cells, each photonic crystal cavity unit cell having a lattice constant a and a hole having a length H_x and a width H_y. One can calculate a plurality of scoring function F values. In some embodiments, the scoring function includes:

F=min(Q,Q_cutoff)/(Q_cutoff×V_mode)

wherein Q is a cavity quality factor based on a, Hx and Hy, Q_cutoff is an estimated maximum realizable Q, and V_mode is a cavity mode volume. A value of a, a value of H_x, and a value of H_y can be chosen so that the scoring function value meets a scoring function value criteria. In some embodiments, a waveguide region and the array of at least two photonic crystal cavity unit cells can be fabricated on a substrate based on the selected value a, the selected value H_x, and the selected value H_y. In some embodiments, at least one ion can be implanted between a hole of a first photonic crystal cavity unit cell and a second photonic crystal cavity unit cell, and annealed into at least one quantum defect. In some embodiments, a coplanar microwave waveguide can be formed on the substrate in proximity to the array of at least two photonic crystal cavity unit cells.

1. Quantum Network Nodes Based on Diamond Qubits

1.1. Integrated Nanophotonic Quantum Register Based on Quantum Defects, Such as Silicon-Vacancy Spins in Diamond

Nanophotonic cavity QED systems with solid-state emitters are suitable for use as quantum nodes as they can be interfaced with on-chip electronic control and photonic routing, making them suitable for integration into large-scale networks. One example property useful for a quantum node is that the node can interface with nanophotonic structures that can be implemented on a chip. For example, self-assembled quantum dots in GaAs can be efficiently interfaced with nanophotonic structures, enabling a fast, on-chip spin-photon interface. Another example property useful for a quantum node is long-lived coherence times (which enables use over longer periods of time). For example, nitrogen-vacancy color centers (NVs) in diamond can be used as multi-qubit quantum processors with coherence times approaching one minute and can be used to implement quantum error correction and teleportation. However, functional nodes that possess long coherence times and the ability to interface with nanophotonic structures on a single chip have not yet been realized. For example, quantum memory times in quantum dots are limited to a few μs by the dense bath of surrounding nuclear spins. Conversely, an efficient nanophotonic interface to NVs remains elusive, in part due to the degradation of their optical properties inside nanostructures arising from electrical noise induced by fabrication.

Embodiments of the present disclosure describe an integrated network node that possesses multiple properties useful for implementation in a scalable quantum network. This can be achieved, for example, by coupling a negatively charged silicon-vacancy color-center (SiV) to a diamond nanophotonic cavity and a nearby nuclear spin, illustrated schematically in the embodiment of FIG. 1 panel (a). In some embodiments, such nodes possess long-lived coherence times as well as the ability to efficiently couple the quantum node to photonic structures for implementations on chips that can be more easily manufactured and used outside of laboratory settings.

FIG. 1 panel (a) shows schematics of a SiV-nanophotonic quantum register 100, according to some embodiments of the present disclosure. A diamond nanostructure 102 with nanocavities 104, embedded SiV centers 106 and ancillary ¹³C nuclei 108 are coupled via a waveguide 110 and an optical fiber 116 to a fiber network 112. Optical excitations (e.g., photons) 114 (within the device but shown as Gaussian distributions on the surface thereof) can be introduced to the diamond nanostructure 102 via an optical fiber 116. Spins are controlled by the on-chip microwave coplanar waveguide (CPW) 110 at 0.1 K. FIG. 1 panel (b) shows scanning electron micrograph of several completed SiV-nanophotonic quantum registers 100, according to some embodiments of the present disclosure. The diamond nanostructure 102 is coupled with gold CPW 110 which is designed to localize microwave fields around the cavity center.

Nanophotonic quantum registers are quantum nodes that can receive information in the form of photons, store that information for long periods of time using a quantum state, and then permit readout of that state using photons. During storage, energy, such as microwaves can be used to manipulate spins in the quantum nodes to enhance the quantum memory time.

As shown in FIG. 1, the SiV 106 is an optically active (i.e., responsive to photons) point defect in the diamond lattice. Its D_3d inversion symmetry results in a vanishing electric dipole moment of the ground and excited states, rendering optical transitions insensitive to electric field noise typically present in nanofabricated structures. The interactions between SiVs and optical photons can be enhanced by incorporating them into nanocavities, which are coupled to on-chip waveguides. Itinerant photons in a fiber network are adiabatically transferred to this waveguide, allowing for the collection of reflected photons with efficiencies exceeding 90%. After an initial optical characterization of the devices, a shorted, gold coplanar waveguide (CPW) 110 is deposited in close proximity to a small subset of cavities [FIG. 1 panel (b), inset]. This enables coherent microwave manipulation of the SiV ground state spin in a cryogenic environment (T<0.1 K), where phonon-mediated dephasing and relaxation processes are mitigated.

In some embodiments, the disclosed devices can provide at least three useful functions of a quantum network node: (i) an efficient spin-photon interface, (ii) a long-lived quantum memory, and (iii) access to multiple interacting qubits.

FIG. 2. panel (a) shows a schematic of energy level structure of SiV spin-cavity system, according to some embodiments of the present disclosure. The presence of SiV 106 in the nanostructure 102 can cause changes to its optical properties. The SiV optical transition 202 at 737 nm is coupled to the nanocavity 102 with detuning Δ. Spin conserving transitions of SiV 106 are split by an external magnetic field (B_ext) 204, at an angle α with respect to the SiV symmetry axis 206 (dotted line). This produces two different pairs of states |↑, |↑′210 and |↓, |↓′208. The transitions between |↑, |↑′ and |↓, |↓′, respectively allows for interacting with different transitions with photons. Photons are reflected by the cavity when the SiV is in state |↑. Microwave fields at frequency f_↑↓ coherently drive the qubit states between |↓ to |↑. Such driving allows for selective interactions with photons and efficient readout of the spin states.

In some embodiments, high-fidelity spin-photon interaction can help with the fast readout of the spin state and further help with the nuclear spin control. FIG. 2 panel (b) shows spin-dependent reflection spectrum for B_ext=0.19T, α≈π/2 at Δ=0.25κ, according to some embodiments of the present disclosure. Probing at the point of maximum contrast (f_Q) results in high-fidelity spin-photon correlations and single-shot readout (inset, F=0.92). FIG. 2 panel (c) shows SiV spin coherence time T₂(N=64)>1.5 ms with dynamical decoupling, according to some embodiments of the present disclosure. The inset of FIG. 2 shows fast microwave Rabi driving of the SiV spin.

In some embodiments, the efficient spin-photon interface can be enabled by SiV coupling to a diamond nanophotonic cavity. The coupling can change energy spectrum of the cavity and therefore facilitate efficient readout of the spin state. As shown in FIG. 2 panel (a), the presence of an SiV 106 modulates the bare nanocavity reflection spectrum (i.e., in the absence of SiV). The strength of this modulation can be parametrized by the cavity cooperativity C=4g²/(κγ)˜38 (with the single photon Rabi frequency, cavity, and atomic energy decay rate {g, κ, γ}=2π×{5.6, 33, 0.1} GHz). For C>1, a high-contrast modulation is expected for a small detuning (Δ) between the cavity and the SiV resonance near 737 nm. An external field B_ext lifts the degeneracy of the SiV spin-½ sub-levels, creating spin-dependent reflection: photons at the frequency of maximum contrast (f_Q) are reflected from the cavity when the SiV is in a specific spin state ([FIG. 2 panel (a)], |↑). Spin readout of the SiV can also be performed with B_ext parallel to the SiV symmetry axis, where the spin-conserving transitions are highly cycling. The high collection efficiency into a tapered fiber allows for fast single-shot readout of the SiV even in a misaligned field [FIG. 2 panel (b)], which is to facilitate the nuclear spin control described below. A readout fidelity of F=0.92 in 13 μs is observed even when a few (˜10) photons are scattered.

In order for a quantum node to act as a quantum memory, it is useful to store the quantum information for a long period of time. In some embodiments, the SiV spin in a nanocavity can be a suitable quantum memory. Microwave pulses at f_↑↓=6.7 GHz can coherently manipulate the SiV spin qubit. The resulting Rabi oscillations, which can be driven in excess of 80 MHz while maintaining acceptable sample temperatures, are shown in the inset of FIG. 2 panel (c). These rotations can be used to probe the coherence properties of the spin via dynamical decoupling sequences [FIG. 2 panel (c)]. The coherence time of the SiV inside the nanocavity is measured to be T₂>1.5 ms and scaled with the number of decoupling pulses as T₂∝N^2/3. The coherence scaling observed here differs from that observed in bulk diamond and is similar to NVs near surfaces. This suggests that SiV memory in nanostructures can be limited by an electron spin bath, for example residing near the surface of the nanostructure or resulting from implantation-induced damage.

In some embodiments, the efficient spin-photon interface and control over the SiV spin state as discussed above can be combined to demonstrate heralded storage of photonic qubit states in the spin-memory, a key feature of a network node. FIG. 3 panel (a) shows a schematic for heralded photon storage, according to some embodiments of the present disclosure. After photonic qubit is reflected off the cavity, an X measurement on the photon heralds successful state transfer which is stored for 2T=20 μs. FIG. 3 panel (b) shows a schematic for spin-photon storage fidelity, according to some embodiments of the present disclosure. The state |±=|↓±|↑ can be mapped onto the SiV, with average fidelity =87(6)%.

FIG. 3 panel (a) outlines an example scheme where photonic qubits are prepared using time-bin encoding and mapped onto the SiV spin. In the example of FIG. 3, panel (a), the SiV is first initialized into a superposition state |→∝|↑+|↓ by optical pumping followed by a microwave

$\frac{\pi }{2}-\mathrm{pulse}.$

A pair of weak coherent pulses separated by δt=30 ns at frequency f_Q are then sent to the cavity. The single photon sub-space corresponds to an incoming qubit state |Ψi∝β_e|e+β_l|l, where |e (|l) denotes the presence of a photon in the early (late) time-bin. As a photon can be reflected from the device if the SiV is in state |↑ [FIG. 2 panel (a)], particular components of the initial product state

$f\left(x\right)=a_0+{\sum }_{n=1}^{\infty }\left(a_n\cos \frac{n\pi x}{L}+b_n\sin \frac{n\pi x}{L}\right)$

can be effectively “carved out”. The SiV spin can be inverted with a π-pulse between the arrival of the two time bins at the cavity, such that a photon detection event indicates that the final state has no |e ↑ or |l ↓ component. This can leave the system in the final spin-photon entangled state |Ψf∝β_e|e↓+βl|l ↑. The reflected photon can enter a time-delay interferometer, where one arm passes through a delay line of length δt, allowing the two time-bins to interfere and erase which-time-bin information. As can be seen by expressing the final state in the corresponding photon basis:

|ψ_f∝|+(β_e|↓+β_l|↑+|−(β_e|↓−β_l|↑,  (1)

a detection event on either the ‘+’ or ‘−’ arm of the interferometer can represent a measurement in the X-basis (|± ∝|e{circumflex over ( )}±|l{circumflex over ( )}), effectively teleporting the initial photonic state onto the electron (up to a known local rotation). The generation of the entangled state |ψ_f for input states |ψ_i=|± can be verified by measuring spin-photon correlations and use it to extract a teleportation fidelity of 0.92(6).

After detection of the heralding photon, the teleported photonic states (initially prepared in {|± or |−}) can be stored in spin memory for 20 μs by applying an additional decoupling π-pulse on the SiV spin. The overall fidelity of teleportation and storage is F=0.87(6) after corrected for readout errors [FIG. 3 panel (b)]. The quantum storage time can be extended by additional decoupling sequences [FIG. 2 panel (c)], enabling entanglement distribution up to a T₂-limited range of 500 km.

In order to extend this range and to enable more generic quantum communication protocols, a two-qubit register can be demonstrated based on the cavity coupled SiV electronic spin and a nearby ¹³C nuclear memory. The ¹³C isotope of carbon is a spin-½ nucleus which has ˜1% natural abundance in diamond and is known to exhibit exceptional coherence times. While direct radio-frequency manipulation of nuclear spins can be impractical due to heating concerns, control over ¹³C spins can be achieved by adapting electron mediated techniques developed for Nitrogen vacancy (NV) centers. The physical principle of the SiV-¹³C interaction is depicted in FIG. 4 panel (a).

FIG. 4 panel (a) shows a schematic of an SiV coupled to nearby ¹³C nuclear spins, according to some embodiments of the present disclosure. Vectors 402 (404) are conditional fields when the SiV is in state |↑(|↓). FIG. 4 panel (b) shows a plot of XY8-2 spin-echo, according to some embodiments of the present disclosure. (Top) envelope for spin-echo shows a T₂ (N=16)=603 μs. XY8-2 at early times (Center) exhibits collapses in signal due to interaction with nuclear spins. Single ¹³C cannot be identified at early times (406), but separate from the bath at long times (408). FIG. 4 panel (c) shows Ramsey measurement on the ¹³C nuclear spin, according to some embodiments of the present disclosure. The nuclear spin precesses at a different Larmor frequency depending on whether the SiV is prepared in |↑ or |↓. Coherent oscillations persist for T*₂>2 ms. (d) Spin echo on ¹³C, revealing T₂>0.2 s. FIG. 4 panel (e) shows a schematic for reconstructed amplitudes for a CNOT gate transfer matrix, according to some embodiments of the present disclosure.

The SiV generates a spin-dependent magnetic field B_SiV at the position of the ¹³C, which is located a few lattice sites away. This is described by a hyperfine interaction Hamiltonian:

$\begin{array}{cc}{\hat{H}}_{HF}=\hslash A_{\left|\right|}\frac{{\hat{S}}_z}{2}\frac{{\hat{I}}_z}{2}+\hslash A_{\perp }\frac{{\hat{S}}_z}{2}\frac{{\hat{I}}_x}{2}& \left(2\right)\end{array}$

where Ŝ_z,x(I_z,x) are the Pauli operators for the electron (nuclear) spin, and A_∥,⊥ are the coupling parameters related to the parallel and perpendicular components of B_SiV with respect to the bias field B_ext. Hyperfine interactions manifest themselves in spin-echo measurements as periodic resonances, shown in FIG. 4 panel (b) for an XY8-2 decoupling sequence π/2−(τ−π−τ)¹⁶−η/2, where τ is the free evolution time. The coherence envelope for this sequence is T₂(N=16)=603 μs [FIG. 4 panel (b), upper panel].

For weakly coupled ¹³C (A_⊥<<ω_l, and A_∥<<ω_l, as used in this letter), the positions of the resonances

$\begin{array}{cc}{\tau }_k\approx \frac{2k+1}{2{\omega }_l}\left(1-\frac{1}{2}{\left(\frac{A_{\perp }}{2{\omega }_l}\right)}^2\right),& \left(3\right)\end{array}$

here ω_l is the larmor frequency of a bare ¹³C, are insensitive to specific ¹³C hyperfine parameters at first order, rendering them indistinguishable at early times (τ_k<<4 μs, [FIG. 4 panel (b), 406]). Individual ¹³C can be isolated at longer times and are used to engineer gates between a single ¹³C and the SiV [FIG. 4 panel (b), 408]. The fundamental two-qubit gate associated with such interaction is a conditional ±π/2 rotation of the ¹³C-spin around the X axis

$\left(R_x^{{\pm }^{\pi /2}}\right),$

which is a maximally entangling gate. Together with unconditional rotations of the nuclear spin (which are also generated via dynamical decoupling sequences), and MW rotations on the SiV, these sequences form a universal set of gates for the register.

The ¹³C can be characterized via Ramsey spectroscopy [FIG. 4 panel (c)]. The nuclear spin can be initialized and read out via the optically addressable SiV spin by transferring population between the SiV and ¹³C. Depending on the SiV state before the Ramsey sequence, oscillations of the nuclear spin can be observed at its eigenfrequencies ω_↑,←²=(ω_l±A_∥/2)²+(A_⊥/2)²), allowing one to determine the hyperfine parameters {ω_l, A_∥, A_⊥}=2π{2.0, 0.70, −0.35} MHz. This coherence persists for T*₂>2 ms, and can be further extended to T₂>0.2s by applying a single dynamical decoupling π-pulse on the nucleus, demonstrating the exceptional memory of the ¹³C nuclear spin [FIG. 4 panel (d)].

In some embodiments, the two-qubit register can be benchmarked by demonstrating an SiV-controlled X-gate (CNOT) on the ¹³C-spin by combining a

$R_x^{{\pm }^{\pi /2}}$

with an unconditional nuclear π/2 rotation. This gate results in a spin flip of the ¹³C if the SiV spin is in the state |↓ [FIG. 4 panel (e)]. One can use this gate to prepare a Bell state by initializing the register in |↓↓, and applying a π/2-rotation gate on the SiV spin followed by a CNOT gate. Correlation measurements can yield a concurrence of C=0.22(9) corresponding to a Bell state fidelity of F=0.59(4) after correcting for readout errors.

Embodiments of the present disclosure can provide for a nanophotonic quantum network node combining multiple ingredients in a single physical system. Both spin-photon and spin-spin experiments can be performed in the same device under identical conditions (cavity detuning and bias field), thereby providing simultaneous demonstration of all example properties of a network node.

In some embodiments, the main limitation on the fidelities can be related to the specific ¹³C in the proximity of the SiV, which can require an unfavorable alignment of the external magnetic field in order to isolate a single ¹³C. For example, the fidelity of two-qubit gates can be limited by residual coupling to bath nuclei, SiV decoherence during the gate operations, and under/over-rotations of the nuclear spin arising from the granularity of spin-echo sequences. To reduce these errors, fine-tuned adaptive pulse sequences can be used to enhance sensitivity to specific nearby ¹³C, and tailor the rotation angle and axis of rotation. Alternatively, replacing gold with superconducting microwave coplanar waveguides will significantly reduce ohmic heating, and allow direct radio-frequency control of nuclear spins. These improvements could also enable the realization of a deterministic two-qubit register based on ²⁹SiV, which contains both electronic and nuclear spins in a single defect.

The fidelity of the heralded photon storage can be limited primarily by single shot readout and imperfect critical coupling of the cavity. The improvements of the nuclear spin control mentioned above can allow for working in an external magnetic field aligned to the SiV axis, which can improve readout fidelity from F˜0.90 (reported here) to 0.99. The impedance mismatch of the cavity used in this example can also give rise to residual reflections which are not entangled with the SiV. Over-coupled cavities can enable the use of a SiV spin-dependent phase flip for reflected photons, improving both the fidelity and success probability of spin-photon interactions.

In some embodiments, in conjunction with controlling emitter inhomogeneity via electromechanical tuning, these techniques can allow for chip-scale fabrication of quantum network nodes, facilitating the realization of scalable quantum repeater architectures. The ability to store quantum information in highly coherent ¹³C nuclei, as well as the opportunity to extend these results to other group-IV color-centers, may open up the possibility of operating such nodes at temperatures >1 K. Finally, the efficient quantum network node demonstrated by embodiments of the present disclosure could enable generation of multi-dimensional cluster states of many photons, which could facilitate realization of novel, ultra-fast one-way quantum communication architectures.

1.2. Nanophotonic Device Fabrication

1.2.1. Device Design

FIG. 5 panel (a) shows a schematic of a quantum network, according to some embodiments of the present disclosure. Nodes consisting of several qubits can be coupled together via an optical interface. FIG. 5 panel (b) shows a schematic of a quantum network node based on the SiV, according to some embodiments of the present disclosure. SiV centers and ancilla ¹³C can be incorporated into a nanophotonic device and addressed with a coupled fiber and microwave coplanar waveguide. A diamond nanostructure 102 with embedded SiV centers 106 and ancillary ¹³C nuclei 108 are coupled via a waveguide 110 to a fiber network 112. Optical excitations (e.g., photons) 114 can be introduced to the diamond nanostructure 102 via an optical fiber 116. Spins are controlled by the on-chip microwave coplanar waveguide (CPW) 110 at 0.1 K.

The example devices in the present disclosure can integrate nanophotonic cavities, implanted quantum defects such as SiV centers, and microwave coplanar waveguides onto a single diamond chip.

Typically, high-quality photonic crystal resonators are fabricated from 2-D membranes, which tightly confine light due to total internal reflection off of material boundaries. Difficulties in growing high-purity, single-crystal diamond films on non-diamond substrates are one of the key challenges to fabricating such resonators in diamond. As a result, nanophotonic diamond structures can be etched out of bulk diamond, which can involve nontraditional etching techniques. For example, two methods have emerged for creating freestanding diamond nanostructures: isotropic undercutting and angled ion-beam etching (IBE). In some embodiments, the latter technique can be used, resulting in freestanding, triangular-cross-section waveguides.

Preliminary design of the nanophotonic structures can be optimized to maximize atom-photon interaction while maintaining high waveguide coupling. To take advantage of the scalable nature of nanofabrication, these optimized devices can be patterned in sets of roughly 100 with slightly modified fabrication parameters. The overall scale of all photonic crystal cavity parameters can be varied between different devices on the same diamond chip to compensate for fabrication errors (which lead to unexpected variations in the resonator frequency and quality-factor). Due to these errors, roughly one in six cavities can be suitable for SiV applications. In some embodiments, hundreds of devices are made in a single fabrication run, ensuring that every run yields many usable devices.

The diamond waveguide region (as opposed to the photonic crystal cavity region) can have two distinguishing features. First, thin support structures are placed periodically along the waveguide and are used to suspend the structures above the substrate. These supports are portions of the waveguide which are adiabatically tapered to be ˜30% wider than the rest of the waveguide, and take longer to etch away during the angled etch process. By terminating the etch after normal waveguide regions are fully etched through, these wide sections become ˜10 nm thick supports which tether the waveguide structures to the substrate while minimizing scattered loss from guided modes. Second, one end of the waveguide structure is adiabatically tapered into free space. These tapers are formed by a linear taper of the waveguide down to less than 50 nm wide over a 10 μm length. This tapered region can be coupled to a similarly tapered optical fiber, allowing structures to efficiently interface with a fiber network. This tapered end of the waveguide can be the most fragile portion of the structure and can break after repeated fiber coupling attempts. This can be what limits the total measurement lifetime of a device.

The number of devices (and thus the relative yield of the fabrication process) can be limited by the maximum packing density on the diamond chip. This can be primarily limited by the need to accommodate 10 μm wide microwave coplanar waveguides (CPWs) between devices, which are patterned directly onto the diamond surface to efficiently control SiV spins using microwaves. Simulations of prospective design geometries can ensure that the CPW is impedance matched with the 50Ω feed lines, which minimizes scattered power from the waveguides. Tapers in the CPW near the center of the cavity regions concentrate current and increase the amplitude of the microwave field near the SiVs, and CPWs are terminated with a short in provide a magnetic field maximum along the device.

1.2.2. Device Fabrication

FIG. 6 panel (a) shows a schematic of the nanofabrication process used to produce devices, according to some embodiments of the present disclosure. In step I, titanium-HSQ mask 602 is patterned using EBL. In step II, pattern is transferred onto diamond 604 using top down O₂ RIE. In step III, angled IBE is used to separate structures from substrate. In step IV, devices are covered in PMMA 606 and implantation apertures 608 are formed using EBL. Device are then cleaned, implanted, and annealed. In step V, PMMA is used in a liftoff procedure to pattern gold microwave strip lines 610. In step VI, final devices are cleaned and prepared for use. FIG. 6 panel (b) shows scanning electron micrographs corresponding to steps II, III, and VI in the fabrication procedure, according to some embodiments of the present disclosure.

In some embodiments, during the fabrication processes of the diamond structures, the angled etch is conducted not with a Faraday cage loaded inside a reactive ion etching chamber, but instead with an IBE. The Faraday cage technique offered the benefit of simplicity and accessibility in that it uses the reactive ion etching chamber in question was large enough to accommodate the cage structure—but suffered from large fluctuations in etch rate across the surface of the sample, as well as between different fabrication runs, due to imperfections in the Faraday cage mesh. These irregularities could be partially compensated for by repeatedly repositioning and rotating the cage with respect to sample during the etch, but this process proved to be laborious and imprecise. Instead, IBE offers collimated beams of ions several cm in diameter, leading to almost uniform etch rates across the several mm diamond chip. This technique allowed for consistent fabrication of cavities with Q>10⁴, V<0.6[λ/(n=2.4)]³, and resonances within ˜10 nm of SiV optical frequencies.

Once the diamond cavities are fabricated [FIG. 6 panel (a) steps I-III)], SiV centers can be incorporated. In some embodiments, to ensure the best possible atom-photon interaction rate, SiVs can be positioned at the cavity mode maximum. In some embodiments, this can involve implantation accuracy of better than 50 nm in all 3 dimensions due to the small mode volume (˜0.5[λ/(n=2.4)]³) of the cavities used. In the past, implantation of silicon ions (which form SiV centers following a high-temperature anneal) was done using focused ion-beam implantation, but, in some embodiments, this technique can require specialized tools and lacked accuracy for maximally efficient mode coupling. Instead, the present disclosure describes methods that adapt the standard masked implantation technique and use commercial foundries for ion implantation.

In some embodiments, for the implantation process, one can repeatedly spin and bake MMA EL11 and PMMA C4 (Microchem®) to cover the nanophotonic cavities completely with polymer resist. A conductive surface layer of Espacer (Showa Denko®) can then be spin-coated. An E-beam lithography (EBL) tool can then align with large markers underneath the polymer layer, allowing it to expose an area surrounding smaller, high-resolution alignment markers on the diamond. The exposed regions are developed in a 1:3 mixture of MIBK:IPA. Espacer is again spin-coated, and a second EBL write can be done, aligned to the high-resolution markers. Based on these alignment markers, holes of less than 65 nm diameter (limited by the resolution of PMMA resist) are patterned onto the center of the photonic crystal cavity which, after subsequent development, act as narrow apertures to the diamond surface [FIG. 6 panel (a) step IV)]. The rest of the diamond surface is still covered in sufficiently thick PMMA to prevent ions from reaching masked portions of the device. Diamonds are then sent to a commercial foundry (Innovion®) where they are implanted with silicon ions at the appropriate energy and dose [FIG. 6 panel (b)]. Annealing in a UHV vacuum furnace (Kurt-Lesker®) at ˜1400 K can convert these implanted ions into SiV centers.

CPWs can be fabricated using a liftoff process similar to that used to create masked implantation windows. The most notable difference is an additional oxygen plasma descum after development to remove PMMA residue from the surface. Following development, a 10 nm titanium film can serve as an adhesion layer for a 250 nm thick gold CPW [FIG. 6 panel (a) step V)]. Liftoff is performed in heated Remover PG (Microchem®) [FIG. 6 panel (a) step VI)]. The metal thicknesses used here are chosen to improve adhesion of the gold, as well as prevent absorption of cavity photons by the metallic CPW. It is observed that the cavity quality factor significantly degrades with gold films with a thickness greater than 300 nm. Due to ohmic heating, which can degrade the coherence properties of SiV spins, the length of the CPW is constrained to address a maximum of roughly 6 devices.

Additional improvements in diamond device performance can be implemented by improvements of the fabrication technology. For example, device quality factors can be limited by deviations in device cross section caused by imperfect selectivity of the HSQ hard mask to oxygen etching. Replacing this mask with a sufficiently smooth metal mask could result in improved etch selectivity and device performance. Isotropic undercut etching could also lead to improved control over device cross sections and facilitate more sophisticated device geometries at the cost of reduced control over isotropically etched surface roughness. Various techniques exist for the formation of smaller implantation apertures, but these techniques are difficult to use in conjunction with implantation into completed nanophotonic devices. Finally, the use of superconducting strip lines could reduce heating, which can enable the CPW to address all devices on the diamond chip and allow for faster driving of SiV spin and nuclear transitions.

1.3. Example Setup

FIG. 7 panel (a) shows an example setup schematic, according to some embodiments of the present disclosure. Devices 702 are mounted in the bore of a SC magnet 704 inside of a dilution refrigerator and imaged with wide-field imaging 706 and piezo steppers 708. Devices are addressed with a tapered optical fiber 710 positioned using a second set of piezo steppers 712. Cavities are tuned using nitrogen 714. FIG. 7 panel (b) shows a schematic of a fiber network used to probe devices, according to some embodiments of the present disclosure. Excitation light is monitored 716 and sent to the device. Collected light is monitored 718 and filtered 720 then sent to one or several SPCMs 722. N.C. indicates no connection. 90:10 fiber beam splitter and 50:50 fiber beam splitter are labeled as 724 and 726, respectively.

In some embodiments, processes described herein can be performed in a photonic-probe setup inside of a dilution refrigerator (DR, Blue-Fors® BF-LD250) [FIG. 7 panel (a)]. The diamond substrate is mounted to a gold-plated copper sample holder via indium soldering below the mixing chamber in the bore of a (6, 1, 1) T superconducting vector magnet (American Magnetics Inc.) anchored to the 4 K stage. A thermal link between the device and the mixing chamber plate is provided by gold-plated copper bars, as well as oxygen-free copper braids (Copper Braid® Products), which can ensure maximal thermal conductivity between the mixing chamber plate and the sample, which reaches a base temperature of roughly 60 mK. Single nanophotonic device can be addressed via a tapered optical fiber, which can be coupled in-situ with collection efficiencies exceeding 90%. The tapered fiber is mounted to a 3-axis piezo stepper (ANPx101, ANPz101), and imaged in free-space by an 8f wide-field scanning confocal microscope which focuses onto a cryo-compatible objective (Attocube® LT-APO-VISIR). This setup allows for coupling to several cavities during a single cooldown.

In some embodiments, once coupled, the cavity resonance can be red-shifted via nitrogen gas condensation. A copper tube is weakly thermalized with the 4 K plate of the DR and can be heated above 80 K in order to flow N₂ gas onto the devices. This gas condenses onto the photonic crystal, modifying its refractive index and red-shifting the cavity resonance. When the copper tube is not heated, it thermalizes to 4K, reducing the blackbody load on the sample and preventing undesired gas from leaking into the vacuum chamber.

In some embodiments, after red-tuning all devices in this way, each cavity can be individually blue-tuned by illuminating the device with a ˜100 μW broadband laser via the tapered fiber, locally heating the device and evaporating nitrogen. This laser-tuning can be performed very slowly to set the cavity resonance with a few GHz. The cavity tuning range exceeds 10 nm without significantly degrading the cavity quality factor, and is remarkably stable inside the DR, with no observable drift over several months of measurements.

In previous work, SiVs were probed in transmission via the free-space confocal microscope focused onto a notch opposing the tapered fiber. Mechanical vibrations arising from the DR pulse tube (˜1 μm pointing error at the sample position) result in significant fluctuations in power and polarization of the coupled light. The present disclosure demonstrates a fully integrated solution by utilizing the same tapered fiber to both probe the device and collect reflected photons. This approach can stabilize the excitation path and improve the efficiency of the atom-photon interface, allowing for deterministic interactions with single itinerant photons. High-contrast reflection measurements are enabled by the high-cooperativity, critically coupled atom-cavity system. Resonant light is sent via the fiber network [FIG. 7 panel (b)] and reflected off of the target device. A small fraction (˜10%) of this signal can be picked off and used to monitor the wide-band reflection spectrum on a spectrometer (Horiba® iHR-550) as well as calibrate the coupling efficiency to the nanocavity. The remaining reflection can then be routed either directly to a single-photon counting module (SPCM, Excelitas® SPCM-NIR), or into a time-delay interferometer for use in spin-photon applications. Due to this high-efficiency fiber-coupled network, overall collection efficiencies of ˜40% can be observed, limited by the quantum efficiency of the APDs.

1.4. Example Strain Regimes for SiV Spin-Photon Applications

Similar to other solid-state emitters, the SiV can be sensitive to local inhomogeneity in the host crystal. In the case of the SiV, which has D_3d symmetry, the dominant perturbation is crystal strain. This section describes the effects of strain on the SiV spin and optical properties, and how they can enable efficient microwave and optical control of SiV centers inside nanostructures.

1.4.1. SiV Hamiltonian in the Presence of Strain and Spin-Orbit Coupling

The SiV electronic structure is comprised of spin-orbit eigenstates split by spin-orbit interactions. Optical transitions connect the ground state manifold (LB, UB) and excited state manifold (LB^I, UB^I) [FIG. 4(a)]. In a DR, phonon absorption LB→UB (and LB′→UB′) is suppressed, resulting in thermal polarization into LB.

Without wishing to be bound by theory, one can consider the ground state SiV Hamiltonian with spin-orbit and strain interactions, in the combined orbital and spin basis {|e_y↑, |e_y ↓, |e_x↑, |e_x↓}

$\begin{array}{cc}{\mathscr{H}}_{SiV}={\mathscr{H}}_{SO}+{\mathscr{H}}_{strain}& \left(4\right)\\ =\left(\begin{array}{cccc}\alpha -\beta & 0& \gamma -i\lambda & 0\\ 0& \alpha -\beta & 0& \gamma +i\lambda \\ \gamma +i\lambda & 0& \alpha +\beta & 0\\ 0& \gamma -i\lambda & 0& \alpha +\beta \end{array}\right)& \left(5\right)\end{array}$

where α corresponds to axial stain, β and γ correspond to transverse strain, and λ is the strength of spin-orbit interaction. Diagonalizing this reveals the orbital character of the lower branch:

$\begin{array}{cc}LB\propto \{\begin{array}{c}|e_x\uparrow \rangle -\frac{1+\sqrt{1+(\gamma /\beta {)}^2+{\left(\lambda /\beta \right)}^2}}{\frac{\gamma }{\beta }-\frac{i\lambda }{\beta }}|e_{y\uparrow \rangle }\\ |e_x\downarrow \rangle -\frac{1-\sqrt{1+(\gamma /\beta {)}^2+{\left(\lambda /\beta \right)}^2}}{\frac{\gamma }{\beta }-i\lambda /\beta }|e_{y\downarrow \rangle }\end{array}& \left(6\right)\end{array}$

One can identify these electronic levels in the context of the SiV as a spin-photon interface.

FIG. 8 panel (a) shows a schematic SiV level diagram, according some embodiments of the present disclosure. Optical transitions f_↑↑I,f_↑↑I˜737 nm can be coupled to a nanophotonic cavity with mean detuning Δ. Microwaves at frequency f_↑↓ drive rotations in the lower branch (LB). FIG. 8 panel (b) shows a plot of qubit frequency f_↑↓ differently strained emitters, according to some embodiments of the present disclosure. Modeled splitting for ground state g-factors g_gs1=1.99, g_gs2=1.89, g_gs3=1.65 (solid lines) based on independent measurements of Δ_gs. FIG. 8 panel (b) inset shows a plot of angle dependence of f_↑↓ at fixed field B_ext=0.19T. Solid lines are predictions using the same model parameters. FIG. 8 panel (c) shows a plot of optical splitting for f_↑↑I−f_↓↓I, according to some embodiments of the present disclosure. Fits extract excite state g-factors g_es1=1.97, g_es2=1.83, g_es3=1.62 (solid lines). FIG. 8 panel (c) inset shows a plot of angle dependence of f_↑↑I−f_↓↓I at fixed field B_ext=0.1T. FIG. 8 panel (d) shows a histogram of MW transition frequency for two different emitters, according to some embodiments of the present disclosure. FIG. 8 panel (e) shows a histogram of optical transition frequency for two different emitters, according to some embodiments of the present disclosure. FIG. 8 panel (f) shows simultaneous measurement of f_↑↓ and f_↑↑I revealing correlations between optical and microwave spectral diffusion for emitter 2, according to some embodiments of the present disclosure.

1.4.2. Effects of Strain on the SiV Qubit States

Without wishing to be bound by theory, in the limit of zero crystal strain, the orbital factors simplify to the canonical form

$\begin{array}{cc}LB=\{\begin{array}{c}|e_{+}\downarrow \rangle \\ |e_{-}\uparrow \end{array}& \left(7\right)\end{array}$

In this example regime, the spin-qubit has orthogonal electronic orbital and spin components. As result, in some non-limiting embodiments, one would need to simultaneously drive an orbital and spin flip to manipulate the qubit, which is forbidden for direct microwave driving alone. Thus, in the low strain regime, two-photon optical transitions between the qubit states in a misaligned external field, can be used to realize a SiV spin qubit.

Without wishing to be bound by theory, in the high strain limit (√{square root over (β²+γ²>>λ)}), these orbitals become

$\begin{array}{cc}LB=\{\begin{array}{c}(\cos \left(\theta /2\right)|e_x\rangle -\sin \left(\theta /2\right)|e_y\rangle )\otimes |\downarrow \rangle \\ (\cos \left(\theta /2\right)|e_x\rangle -\sin \left(\theta /2\right)|e_y\rangle )\otimes |\uparrow \rangle \end{array}& \left(8\right)\end{array}$

where tan

$\left(\theta \right)=\frac{\beta }{\gamma }.$

In this regime, the ground state orbital components are identical, and the qubit states can be described by the electronic spin degree of freedom. As such, the magnetic dipole transition between the qubit states is now allowed and can be efficiently driven with microwaves.

In addition to determining the efficiency of qubit transitions, the spin-orbit nature of the SiV qubit states also determines its susceptibility to external fields. In an externally applied magnetic field, LB splits due to magnetic moments associated with both spin and orbital angular momenta. This splitting is parameterized by an effective g-tensor which, for a fixed angle between the external field and the SiV symmetry axis, can be simplified to an effective g-factor: μg B_ext/h=f_↑↓. In the limit of large strain, the orbital component of the two LB wavefunctions converge, and g trends towards that of a free electron (g=2). As a result, the qubit states behave akin to a free electron in the high strain regime, and there is no dependence of g on external field angle or small perturbations in crystal strain.

While it is difficult to probe β or γ directly, they relate to the energy difference between UB and LB via Δ_gs=2√{square root over (β_gs²+γ_gs²+λ_gs²)} [FIG. 8 panel (a)]. From this, one can extract √{square root over (β²+γ²)}, given the know value of λ_gs=46 GHz. Numerically diagonalizing the SiV Hamiltonian using the extracted values for β and γ closely matches the measured ground state splitting, both as a function of applied field magnitude and angle [FIG. 8 panel (b)].

1.4.3. Effects of Strain on the SiV Spin-Photon Interface

In some embodiments, strain also plays a crucial role in determining the optical interface to the SiV spin qubit. The treatment shown above can be repeated for the excited states, with the caveat that the parameters β, γ, and λ are different in the excited state manifold as compared to the ground state manifold. These differences can give rise to a different g-factor in the excited state (g_es). If the strain is much larger than both λ_gs=46 GHz and λ_es=255 GHz, then g_gs≈g_es≈2. In this case, the two spin-cycling transition frequencies f_↑↑I and f_↓↓I are identical, and the spin-selective optical transitions can be the dipole-forbidden spin-flipping transitions f_↑↓I and f_↑↓I.

Under more moderate strain, the difference δg=|g_es−g_gs| splits the degenerate optical transitions f_↑↑I and f_↓↓I, making them spin-selective as well. Due to differences in the anisotropic g-tensor in the ground and excited states, δg depends on the orientation of the magnetic field as well and is minimized in the case of a <111>-aligned field [FIG. 8 panel (c), inset]. In such an external field aligned with the SiV symmetry axis, optical transitions become highly spin-conserving, allowing many photons to scatter without altering the SiV spin state. This high cyclicity enables high-fidelity single-shot readout of the spin state, even without high collection efficiencies. This makes working with the spin-cycling transitions highly desirable, at the expense of a reduced ability to resolve spin-selective transitions for a given field magnitude. The need to resolve individual transitions suggests a nonlimiting, example strain regime where √{square root over (β_gs²+γ_gs²>>λ_gs²)}, where MW driving is efficient, while √{square root over (β_gs²+γ_gs²≤λ_gs²)}, where one can independently address f_↑↑I and f_↓↓I [FIG. 8 panel (c)].

1.4.4. Effects of Strain on SiV Stability

Despite the SiV's symmetry-protected optical transitions, spectral diffusion of the SiV has been observed in many example use cases (but still much smaller compared to emitters without inversion symmetry, for example, nitrogen-vacancy centers). While the exact nature of this diffusion has not been studied in depth, it is often attributed to the second-order Stark effect or strain fluctuations, both of which affect the energies of SiV orbital wavefunctions. In some embodiments, one can also observe significant fluctuations of the spin qubit frequency.

In some embodiments, for an appropriately low static strain value, fluctuating strain can give rise to fluctuations in the g-tensor of the ground state, causing spectral diffusion of the qubit frequency f_↑↓ [FIG. 8 panel (d)]. Since g_gs asymptotically approaches 2 as the static strain increases, the qubit susceptibility to this fluctuating strain is reduced in the case of highly strained SiV centers, resulting in a more stable qubit.

In some embodiments, while spectral diffusion of the optical transition should not saturate in the same way as diffusion of the microwave transition, qualitatively different spectral diffusion properties are observed for different emitters [FIG. 8 panel (e) and FIG. 16]. SiV 1 (Δ_gs=500 GHz) displays slow drift of the optical line which is stable to <100 MHz over many minutes. In some embodiments, one does not observe significant fluctuations (>500 kHz) of the microwave transition for this SiV. On the other hand, SiV 2 (Δ_gs=140 GHz) drifts over a wider range, and also exhibits abrupt jumps between several discrete frequencies.

In some embodiments, one can simultaneously record the optical transition and qubit frequency for SiV 2 and observe correlations between the two frequencies [FIG. 8 panel (f)], indicating that they could arise from the same environmental perturbation. In some embodiments, one can calculate the qubit and optical transition frequencies using the strain Hamiltonian (Eqn. 5) and find that both correlations and absolute amplitudes of spectral diffusion can simultaneously be explained by strain fluctuations on the order of 1% (˜10⁻⁷ strain).

In some embodiments, one can rely on static strain, which can result from damage induced by ion implantation and nanofabrication and select for spectrally stable SiVs with appropriate strain profiles. This is characterized by first measuring Δ_gs in zero magnetic field at 4 K by exciting the optical transition LB→LB^I and measuring emission from the LB^I→UB on a spectrometer. One can use this to screen for SiVs with Δ_gs>100 GHz which can provide efficient MW driving of the spin qubit. One can further apply a static external magnetic field and measure spectral stability properties as well as f_↑↑I−f_↓↓I to provide a good spin-photon interface. In some embodiments, 10 candidate emitters are measured, and 4 are found to satisfy all of the example criteria for spin-photon applications.

1.5. Regimes of Cavity-QED for SiV Spin-Photon Interfaces

Efficient spin-photon interactions can be enabled by incorporating SiV centers into nanophotonic cavities. This section describes SiV-cavity measurements in several regimes of cavity QED, and comment on example uses for spin-photon applications.

1.5.1. Spectroscopy of Cavity Coupled SiVs

FIG. 9 panel (a) shows a SiV-cavity reflection spectrum at several detunings, according to some embodiments of the present disclosure. The bare cavity spectrum (902) is modulated by the presence of the SiV. When the atom cavity detuning is small (904, 906), high-contrast, broad features are the result of Purcell enhanced SiV transitions. Far from the cavity resonance (908), interaction results in narrow SiV-assisted transmission channels. FIG. 9 panel (b) shows a plot of spin-dependent reflection for large SiV-cavity detuning Δ≈−3κ, B_ext=0.35 T, according to some embodiments of the present disclosure. In this regime, SiV spin states can be individually addressed. FIG. 9 panel (c) is a plot showing that probing either transmission dip results in high-fidelity single-shot readout in an aligned field (F=0.97, threshold on detecting 13 photons), according to some embodiments of the present disclosure. FIG. 9 panel (d) shows a plot of spin-dependent reflection near resonance Δ≈0.5κ, B_ext=0.19 T, according to some embodiments of the present disclosure. Dispersive line shapes allow for distinguishable reflection spectra from both SiV spin states. FIG. 9 panel (e) is a plot showing that a probe at the frequency of maximum contrast (f_Q) can determine the spin state in a single shot in a misaligned field (F=0.92, threshold on detecting >1 photon), according to some embodiments of the present disclosure.

In some embodiments, one can measure the spectrum of the atom-cavity system at different atom-cavity detunings in order to characterize the device and extract key cavity QED parameters [FIG. 9 panel (a)]. Without wishing to be bound by theory, the reflection spectrum of a two-level system coupled to a cavity can be modeled by solving the frequency response of the standard Jaynes-Cummings Hamiltonian using input-output formalism for a cavity near critical coupling:

$\begin{array}{cc}ℛ\left(\omega \right)=1-{(\frac{2{\kappa }_l}{i\left(\omega -{\omega }_c\right)+{\kappa }_{\mathrm{tot}}+g^2/\left(i\left(\omega -{\omega }_a\right)+\gamma \right)}}^2& \left(9\right)\end{array}$

where κ_l is the decay rate from the incoupling mirror, κtot is the cavity linewidth, ω_c(ω_a) is the cavity (atom) resonance frequency, g is the single-photon Rabi frequency, and γ is the bare atomic linewidth. Without wishing to be bound by theory, interactions between the SiV optical transition and the nanophotonic cavity result in two main effects. First, the SiV center can modulate the reflection spectrum of the bare cavity, as seen in the curves of FIG. 9 panel (a). Second, the coupling to the cavity can broaden the linewidth of the SiV based on the Purcell effect:

$\begin{array}{cc}\Gamma \approx \gamma +4_g^2/\kappa \frac{1}{1+4{\left({\omega }_c-{\omega }_a\right)}^2/{\kappa }^2}& \left(10\right)\end{array}$

When the cavity is far detuned from the atomic transition |ω_c ω_a|Δ>κ [FIG. 9 panel (a), 908], Purcell enhancement is negligible and the cavity and atomic linewidths κ, γ=2π×{33, 0.1} GHz are estimated. When the cavity is on resonance with the atom (Δ=0), one can fit Eqn. (9) using previously estimated values of κ and γ to extract g=2π 5.6 GHz. Together, these measurements can determine the atom-cavity cooperativity C=4g²/κγ=38. In some embodiments, interactions between the SiV and single photons becomes deterministic when C>1.

In some embodiments, one can make use of spectrally resolved spin conserving optical transitions (f_↑↑I, f_↓↓I) to build a spin-photon interface using the SiV. In some embodiments, one can make this criterion more explicit: f_↑↑I and f_↓↓I can be resolved when |f_↑↑I−f_↓↓I≥Γ.

1.5.2. Cavity QED in the Detuned Regime

In the detuned regime (Δ>κ), Γ≈γ, and narrow atom-like transitions are easily resolved under most magnetic field configurations, including when the field is aligned with the SiV symmetry axis [FIG. 9 panel (b)]. In this case, optical transitions are highly spin-conserving, and many photons can be collected allowing for high-fidelity single-shot readout of the SiV spin state (F=0.97) [FIG. 9 panel (c)]. Rapid, high-fidelity, nondestructive single-shot readout can enable projective-readout based initialization: after a single measurement of the SiV spin state, the probability of a measurement-induced spin flip is low, effectively initializing the spin into a known state.

While this regime is useful for characterizing the system, the maximum fidelity of spin-photon entanglement based on reflection amplitude can be limited. As seen in FIG. 9 panel (b), the contrast in the reflection signal between an SiV in |↑ vs. |↓ is 80%, implying that in 20% of cases, a photon is reflected from the cavity independent of the spin state of the SiV, resulting in errors. It is noted that the residual 20% of reflection can be compensated by embedding the cavity inside an interferometer at the expense of additional technical stabilization challenges, discussed below.

1.5.3. Cavity QED Near Resonance

Tuning the cavity onto the atomic resonance (Δ≈0) dramatically improves the reflection contrast [FIG. 9 panel (a) 904]. Here, nearly full contrast of the reflection spectrum is observed due to the presence of the SiV. Unfortunately, this is associated with a broadened atomic linewidth (Γ=γ(1+C)˜4 GHz). While it is, in principle, still possible to split the atomic lines by going to higher magnetic fields, there are several technical considerations which make this impractical. Large magnetic fields (|B_ext|>0.5 T) correspond to large qubit frequencies (f_↑↓), which can induce spontaneous qubit decay due to phonon emission (|↑|↓>), as well as increased local heating of the device from microwave dissipation, both of which reduce the SiV spin coherence time rendering it ineffective as a quantum memory.

At intermediate detunings (0<Δ<κ), the SiV resonance is located on the cavity slope and results in high-contrast, spin-dependent Fano line shapes which exhibit sharp features smaller than Γ [FIG. 9 panel (a) 906]. By working at a B_ext, for example where the peak of one spin transition is overlapped by the valley of the other, the best features of the resonant and far-detuned regimes are recovered [FIG. 9 panel (e)]. Probing the system at the point of maximum contrast (f_Q≈(|f_↑↑I−f_↓↓I)/2, contrast >90%) enables single-shot readout of the SiV spin state for an arbitrary field orientation, even when transitions are not cycling [FIG. 9 panel (f)].

This demonstrates an optical regime of cavity QED where high-contrast readout can be achieved while spin-dependent transitions is maintained. In this regime, one can still expect residual reflections of about 10%, which end up limiting spin-photon entanglement fidelity. This infidelity arises because the cavity is not perfectly critically coupled (κ_t=κ_tot/2), and can in principle be solved by engineering devices that are more critically coupled. Alternatively, this problem can be addressed for any cavity by interfering the signal with a coherent reference to cancel unwanted reflections. In this case, in some embodiments, one would have to embed the cavity in one arm of a stabilized interferometer. This is quite challenging, as it involves stabilizing ˜10 m long interferometer arms, part of which lie inside the DR (and experience strong vibrations from the pulse-tube cryocooler).

A fundamental issue with critically coupled cavities is that not all of the incident light is reflected from the device. If the spin is not initialized in the highly reflecting state, photons are transmitted and not recaptured into the fiber network. Switching to over-coupled (single-sided) cavities, where all photons are reflected with a spin-dependent phase, could improve both the fidelity and efficiency of spin-photon entanglement. Once again, however, measurement of this phase can sometimes require embedding the cavity inside of a stabilized interferometer. As such, the un-compensated reflection amplitude-based scheme employed here is the most technically simple approach to engineering spin-photon interactions.

1.6. Microwave Spin Control

While the optical interface described in previous sections enables high-fidelity initialization and readout of the SiV spin qubit, direct microwave driving can be the most straightforward path towards coherent single-qubit rotations. Typically, microwave manipulation of electron spins involves application of significant microwave power. This presents a challenge, as some SiV spins need to be kept at local temperatures below 500 mK in order to avoid heating-related dephasing. In some embodiments, one can implement coherent microwave control of SiV centers inside nanostructures at temperatures below 500 mK.

1.6.1. Generating Microwave Single-Qubit Gates

FIG. 10 panel (a) is a schematic of a setup for microwave control, according to some embodiments of the present disclosure. The amplitude and phase of a CW microwave source 1002 are modulated via a microwave switch and IQ mixer controlled externally by an AWG 1004. A CW radio frequency source 1006 is controlled using a digital delay generator 1008. Both signals are amplified by 30 dB amplifiers 1010 before entering the DR. 0 dB cryo-attenuators 1012 thermalize coax cables at each DR stage, ultimately mounted to a PCB 1014 on the sample stage and delivered to the devices. FIG. 10 panel (b) shows a schematic depicting microwave-induced heating of devices, according to some embodiments of the present disclosure. FIG. 10 panel (c) is a plot showing modeled temperature at the SiV from a dynamical decoupling sequence, according to some embodiments of the present disclosure. At long τ, device cools down between each decoupling pulse, resulting in low temperatures. At short τ, devices are insufficiently cooled, resulting in a higher max temperature (T_max). FIG. 10 panel (d) is a plot showing effects of microwave heating on SiV coherence time, according to some embodiments of the present disclosure. The top panel shows that at high Rabi frequencies, SiV coherence is temporarily reduced for small T. The bottom panel shows the local temperature (T_max) at the SiV calculated by taking the maximum value of the plots in panel (c). FIG. 10 panel (e) is a plot of Hahn-echo for even lower Rabi frequencies, showing coherence times that scale with microwave power, according to some embodiments of the present disclosure.

The SiV spin can be coherently controlled using amplitude and phase-controlled microwave pulses generated by a Hittite® signal generator (HMC-T2220). A target pulse sequence can be loaded onto an arbitrary waveform generator (Tektronix® AWG 7122B), which uses a digital channel to control a fast, high-extinction MW-switch (Custom Microwave Components, CMCS0947A-C2), and the analog channels adjust the amplitude and phase via an IQ-mixer (Marki®, MMIQ-0416LSM). The resulting pulse train is subsequently amplified (Minicircuits®, ZVE-3 W-183+) to roughly 3 W of power and sent via a coaxial cable into the dilution refrigerator. At each cryogenic flange, a 0 dB attenuator is used to thermalize the inner and outer conductors of the coaxial line while minimizing microwave dissipation. The signal is then launched into a coplanar waveguide on a custom-built circuit board (Rogers4003C, Bay Area Circuits®) so it can be wire-bonded directly to the diamond chip [FIG. 10 panel (c)]. The qubit frequency (f_↑↓) can be measured by its optically detected magnetic resonance spectrum (ODMR). In some embodiments, ODMR can be observed from 2 GHz to 20 GHz (corresponding to fields from 0.1 T to 0.7 T), implying that microwave control of SiV centers in this configuration is possible at a wide variety of external field magnitudes. This allows the freedom of tuning the field to optimize other constraints, such as for resolving spin transitions and identifying ancillary nuclear spins.

Once the qubit frequency has been determined for a given field, single-qubit gates are tuned up by measuring Rabi oscillations. The frequency of these oscillations scales with the applied microwave power Ω_R˜√{square root over (P)} and determines the single-qubit gate times. One can perform π-pulses R_ϕ^π in under 12 ns, corresponding to a Rabi frequency exceeding 80 MHz. This coherent control is used to implement pulse-error correcting dynamical decoupling sequences, either CPMG-N sequences of the form R_x^π/2−(τ−R_y^π−τ)^N−R_x^π/2=x−(Y)^N−x or XY8-N sequences of the form x−(XY XYY XYX)^N−x. Sweeping the inter-pulse delay τ measures the coherence time T₂ of the SiV.

1.6.2. Effects of Microwave Heating on Coherence

In some embodiments, thermally induced T₁ relaxation can dramatically reduce SiV coherence times. Without wishing to be bound by theory, to explain this phenomenon, one can model the nanobeam as a 1D beam weakly coupled at two anchor points to a uniform thermal bath [FIG. 10 panel (b)]. Initially, the beam is at the steady-state base temperature of the DR. A MW pulse instantaneously heats the bath, and the beam re-thermalizes on a timescale τ_th set by the thermal conduction of diamond and the beam geometry. Once the pulse ends, this heat is extracted from the beam on a similar timescale. Without wishing to be bound by theory, by solving the time-dependent 1-D heat equation, one can find that the change in temperature at the SiV caused by a single pulse (starting at time t₀) scales as ∝(e^{−(t−t0)/τth}−e^{−9(t−t0)/τth}). One can take the sum over N such pulses to model the effects of heating from a dynamical-decoupling sequence of size N.

In some embodiments, at early times (τ<τ_th), the SiV does not see the effects of heating by the MW line, and coherence is high. Similarly, at long times (τ>>τ_th) a small amount of heat is able to enter the nanostructure and slightly raise the local temperature, but this heat can be dissipated before the next pulse arrives [FIG. 10 panel (c), 30 μs]. At intermediate timescales however, a situation can arise where the nanobeam has not fully dissipated the heat from one MW pulse before the second one arrives [FIG. 10 panel (c), 10 μs]. One can plot the maximum temperature as seen by the SiV as a function of pulse spacing [FIG. 10 panel (d), lower panel], and observe a spike in local temperature for a specific inter-pulse spacing τ, which depends on τ_th. Dynamical-decoupling sequences using high Rabi frequency pulses reveal a collapse in coherence at a similar time [FIG. 10 panel (d), upper panel]. This collapse disappears at lower Rabi frequencies, suggesting that it is associated with heating-related dephasing. One can fit this collapse to a model where the coherence time T₂ depends on temperature and extract the rate of heating τ_th=70 μs.

Typically, faster π-pulses improve measured spin coherence by minimizing finite-pulse effects and detuning errors. Unfortunately, as seen above, faster pulses can involve higher MW powers which cause heating-related decoherence in the system. One can measure Hahn-echo at lower MW powers [FIG. 10 panel (e)] and find MW heating limits T₂ even at Ω_R˜10 MHz. For applications where long coherence is important, such as electron-nuclear gates, one can operate at a Rabi frequency, for example at ΩR=2π 10 MHz where nuclear gates are as fast as possible while maintaining coherence for the entire gate duration. For applications such as spin-photon entangling gates where fast gates are used, one can operate at higher Rabi frequencies ΩR=2π×80 MHz at the cost of reduced coherence times.

In some embodiments, heating related effects could be mitigated by using superconducting microwave waveguides. This approach can also enable the fabrication of a single, long superconducting waveguide that could simultaneously address all devices on a single chip. However, it is still an open question whether or not superconducting waveguides with appropriate critical temperature, current, and field properties can be fabricated around diamond nanostructures.

1.7. Example Noise Bath of SIVs in Nanostructures

FIG. 11 panel (a) is a plot showing T2 scaling for two different SiVs, according to some embodiments of the present disclosure. SiV 2 exhibits no scaling with number of pulses (T_2,SiV2=30 μs). FIG. 11 panel (b) is plot of DEER ESR on SiV 2, according to some embodiments of the present disclosure. Vertical dashed line is the expected frequency of a g=2 spin based on the ability to determine the applied external field (typically to within 10%). FIG. 11 panel (c) is a plot of DEER Echo on SiV 2. T_2,DEER=10 μs, according to some embodiments of the present disclosure. FIG. 11 panel (d) is a plot of dynamical decoupling on SiV 1, according to some embodiments of the present disclosure. Data points are T₂ measurements used in SiV 1, and solid lines are a noise model consisting of two Lorentzian noise baths.

At low temperatures, the coherence time of SiV centers can drastically depend on the surrounding spin bath, which can differ from emitter to emitter. As an example, it is noted that the T₂ of two different SiV centers in different nanostructures scales differently with the number of applied decoupling pulses [FIG. 11 panel (a)]. In some embodiments, the coherence time of SiV 2 does not scale with the number of applied pulses, while the coherence time of SiV 1 does scale as T₂(N)˜N^2/3. Notably, both scaling can be different as compared to what was previously measured in bulk diamond: T₂ (N)˜N¹. In some embodiments, one can probe the spin bath of these two SiVs in nanostructures to identify example explanations for the above observations.

1.7.1. Double Electron-Electron Resonance Spectroscopy of SiVs in Nanostructures

In order to identify the poor coherence of SiV 2, one can perform double electron-electron resonance (DEER) spectroscopy to probe the spin bath surrounding this SiV. In some embodiments, one can perform a Hahn-echo sequence on the SiV, and sweep the frequency of a second microwave pulse (taking the RF path in FIG. 10 panel (a)), contemporaneous with the echoing SiV π-pulse [FIG. 11 panel (b), upper panel]. If this second pulse is resonant with a spin bath coupled to the SiV, the bath can flip simultaneously with the SiV, leading to increased sensitivity to noise from the bath [FIG. 11 panel (b), lower panel]. In some embodiments, one can observe a significant reduction of coherence at a frequency consistent with that of a free-electron spin bath (g_path=2) (resonance expected at 12(1) GHz).

In some embodiments, one can repeat a standard Hahn-echo sequence where a it-pulse resonant with this bath is applied simultaneously with the SiV echo pulse (DEER echo). The coherence time measured in DEER echo is significantly shorter than for standard spin-echo, indicating that coupling to this spin bath is a significant source of decoherence for this SiV. Without wish to be bound by theory, one possible explanation for the particularly severe bath surrounding this SiV is a thin layer of alumina (Al₂O₃) deposited via atomic layer deposition on this device in order to tune cavities closer to the SiV transition frequency. The amorphous oxide layer—or its interface with the diamond crystal—can be host to a large number of charge traps, all located within ˜50 nm of this SiV. In some embodiments, one cannot measure this device without alumina layer due to difficulties to gas-tune the nanophotonic cavity close enough to the SiV resonance.

These observations are further corroborated by DEER measurements in SiV 1, where the alumina layer was not used (N₂ was used to tune this cavity). In some embodiments, one can observe longer coherence times which scale T₂(N) N^2/3, as well as no significant signatures from g_path=2 spins using DEER spectroscopy. Without wish to be bound by theory, one can fit this scaling to a model consisting of two weakly-coupled spin baths [FIG. 11 panel (d)], and extract bath parameters b₁=5 kHz, τ₁=1 μs, b₂=180 kHz, τ₂=1 ms, where b corresponds to the strength of the noise bath, and τ corresponds to the correlation time of the noise.

In some embodiments, one can find that the b₂ term (for example due to bulk impurities) is the dominant contribution towards decoherence in the system. Removing this term from the model results in coherence times up to a factor of 1000 times larger than measured values. Higher-temperature or in situ annealing can mitigate this source of decoherence in some embodiments by eliminating paramagnetic defects such as vacancy clusters. Additionally, by accompanying Si implantation with electron irradiation, SiV centers could be created more efficiently, and with reduced lattice damage. Finally, working with isotopically purified diamond samples with very few ¹³C, a spin-½ isotope of carbon, could also result in a reduced spin bath.

1.8. Spin-Photon Entanglement

The previous sections characterize the SiV as an efficient spin-photon interface and a quantum memory with long-lived coherence. In some embodiments, one can combine these two properties to demonstrate entanglement between a spin qubit and a photonic qubit. The mechanism for generating entanglement between photons and the SiV can be seen in FIG. 9 panels (b) and (d). Depending on the spin state of the SiV, photons at the probe frequency are either reflected from the cavity and detected or are transmitted and lost.

FIG. 12 panel (a) is a schematic showing an example sequence for generating and verifying spin-photon entanglement, according to some embodiments of the present disclosure. In some embodiments, a time-bin encoded qubit can be utilized. A time-bin encoded qubit can refer to a particular type of photonic qubit in which information is stored in the arrival time of the photon. A time-bin encoded qubit is reflected by the cavity, and both the SiV and the photonic qubits are measured in the Z and X bases. FIG. 12 panel (b) is a plot of spin-photon correlations measured in the Z-Z basis, according to some embodiments of the present disclosure. Light (dark) bars are before (after) correcting for known readout error associated with single-shot readout of the SiV. FIG. 12 panel (c) is a plot of spin-photon correlations measured in the X-X basis, according to some embodiments of the present disclosure. Bell-state preparation fidelity of F≥0.89(3) and a concurrence C≥0.72(7). FIG. 12 panel (d) is a plot showing preparation of second spin-photon Bell state, according to some embodiments of the present disclosure. Changing the phase of the incoming photonic qubit prepares a Bell-state with inverted statistics in the X basis.

1.8.1. Generating Time-Bin Qubits

In some embodiments, time-bin encoding can be chosen for photonic qubits. One straightforward possibility is to use the Fock state of the photon. However, it can be challenging to perform rotations on a Fock state, and photon loss results in an error in the computational basis. Another possibility is to use the polarization degree of freedom. While the SiV spin-photon interface is not polarization selective (both spin states couple to photons of the same polarization), one could consider polarization based spin-photon entangling schemes already demonstrated in nanophotonic systems. However, this can involve embedding the nanostructure inside of a stabilized interferometer, which has a number of challenges. In addition, it can involve careful fabrication of over-coupled, single-sided cavities (unlike the critically coupled diamond nanocavities). As such, time-bin encoding can be a natural choice given the critically coupled SiV-cavity interface described here.

These qubits are generated by passing a weak coherent laser though a cascaded AOM, amplitude-EOM, and phase-EOM. The time-bins are shaped by an AWG-generated pulse on the amplitude-EOM and chosen to be much narrower than the delay δt between time bins. One can choose to prepare arbitrary initial photonic states by using the phase-EOM to imprint an optional phase shift to the second bin of the photonic qubit. In some embodiments, one can use a laser with Poissonian photon number statistics and set the average photon number <n_ph>=0.008<<1 using the AOM to avoid events where two photons are incident on the cavity.

Using this encoding, measurements in a rotated basis (X-basis) can become straightforward. One can send the time-bin qubit into an actively stabilized, unbalanced, fiber-based, Mach-Zender interferometer, where one arm passes through a delay line of time δt. With 25% probability, |e> enters the long arm of the interferometer and |l> enters the short arm, and the two time bins interfere at the output. Depending on the relative phase between the two bins, this will be detected on one of the two arms of the interferometer output [FIG. 7 panel (b)], corresponding to a measurement in the X basis of |±>.

1.8.2. Spin-Photon Bell States

One can prepare and verify the generation of maximally entangled Bell states between the SiV and a photonic qubit using the example sequence depicted in FIG. 12 panel (a). First, the SiV is initialized into a superposition state |→=1/√{square root over (2)}(|↑+|↓). Then photons at frequency f_Q are sent to the cavity, corresponding to an incoming photon state |+=1/√{square root over (2)}(|e+|l), conditioned on the eventual detection of one photon during the example run. Before any interactions, this state can be written as an equal superposition: Ψ₀=|→⊗|+=½(|e↑+|e↓+|l↑+|l↓). The first time bin is reflected from the cavity if the SiV is in state |↑>, effectively carving out |e↓>, in reflection. A π-pulse on the SiV transforms the resulting state to Ψ₁=1/√{square root over (3)}(|e↓+|l↓+|1↑). Finally, reflection of the late time-bin off of the cavity carves out the state |l↓>, leaving a final entangled state Ψ₂=1/√{square root over (2)}(|e↓+|l↑). To characterize the resulting state, one can perform tomography on both qubits in the Z and X bases [FIG. 12 panel (a)].

In order to enable high-bandwidth operation and reduce the requirements for laser and interferometric stabilization in generating and measuring time-bin qubits, it can be generally beneficial to set δt as small as possible. The minimum δt is determined by two factors. First, it's useful for each pulse to be broad enough in the time-domain (narrow enough in the frequency domain) so that it does not distort upon reflection off of the device. From FIG. 9 panel (d), the reflection spectrum is roughly constant over a ˜100 MHz range, implying that nanosecond pulses are sufficient. The second consideration is that a microwave π-pulse can be placed between the two pulses. In some embodiments, one can drive fast (12 ns) π-pulses. As such, one can set δt=30 ns and use 5 ns optical pulses to satisfy these criteria.

1.8.3. Spin-Photon Entanglement Measurements

In some embodiments, for Z-basis measurements, photons reflected from the cavity are sent directly to a SPCM and the time-of-arrival of the time-bin qubit is recorded. Afterwards, the SiV is read out in the Z-basis. Single-shot readout is calibrated via a separate measurement where the two spin-states are prepared via optical pumping and readout, and the fidelity of correctly determining the |↑>(|↓>) state is F_↑=0.85 (F_↓=0.84), limited by the large 0 component of the geometric distribution which governs photon statistics for spin-flip systems. In other words, since one can work in a misaligned field in this example, the probability of a spin flip is high, making it somewhat likely to measure 0 photons regardless of initial spin state. Even before accounting for this known error, one can observe clear correlations between the photonic and spin qubits [FIG. 12 panel (b), light-shading]. Error bars for these correlation histograms (and the following fidelity calculations) are estimated by statistical bootstrapping, where the scattered photon histograms (post-selected on the detection of |e or |l are randomly sampled in many trials, and the variance of that ensemble is extracted.

In some embodiments, measurements in the X-basis are performed similarly. The photon is measured through an interferometer as described above, where now the detector path information is recorded for the overlapping time-bin. After a R_y^π/2 pulse on the SiV, the scattered photon histograms again reveal significant correlations between the ‘+’ and ‘−’ detectors and the SiV spin state [FIG. 12 panel (c)]. By adding a π-phase between the early and late time bins, one can prepare an orthogonal Bell state. Measured correlations of this state are flipped in the X-basis [FIG. 12 panel (d)].

In some embodiments, measurements of this Bell state in the Z- and X-bases are used to estimate a lower bound on the fidelity: F=Ψ⁺|ρ|Ψ⁺≥0.70 (3) (F≥0.89(3) after correcting for readout errors). The resulting entangled state can be quantified by its concurrence C≥0.42(6) (C≥0.79(7) after correcting for readout errors). This high-fidelity entangled state between a photonic qubit and a quantum memory is a fundamental resource for quantum communication and quantum computing schemes, and can be used, for example, to demonstrate heralded storage of a photonic qubit into memory.

1.9. Control of SIV-¹³C Register

While demonstrations of a quantum node with a single qubit is useful for some protocol, nodes with several interacting qubits enable a wider range of applications, including quantum repeaters. In some embodiments, one can introduce additional qubits based on ¹³C naturally occurring in diamond.

1.9.1. Coupling Between the SiV and Several ¹³C

FIG. 13 panel (a) is a plot of XY8-2 spin echo sequence reveals coupling to nuclear spins, according to some embodiments of the present disclosure. The left panel shows collapses <S_x>=0 at short times indicate coupling to many nuclei. The right panel shows collapses <S_x>≠0 at long times indicate conditional gates on a single nuclear spin. FIG. 13 panel (b) is a schematic showing trajectory of ¹³C on the Bloch sphere during a maximally entangling gate, according to some embodiments of the present disclosure. Lines 1302 (1304) correspond to the SiV initially prepared in state |↑>(|↓>); transitions from solid to dashed lines represent flips of the SiV electronic spin during the gate. FIG. 13 panel (c) is a schematic showing maximally entangling gates of the form

$R_{\overrightarrow{n\uparrow },\overrightarrow{n\downarrow }}^{\varnothing }$

are used to initialize and readout the two-qubit register, according to some embodiments of the present disclosure. FIG. 13 panel (d) is a schematic showing tuning up an initialization gate, according to some embodiments of the present disclosure. Inter-pulse spacing T for Init and Read gates are swept to maximize polarization. Solid line is the modeled pulse sequence using the hyperfine parameters extracted from panel (a). FIG. 13 panel (e) is a schematic showing nuclear Ramsey measurement, according to some embodiments of the present disclosure. Driving the ¹³C using composite gates on the SiV reveals T₂*=2.2 ms. Orange points in the inset are coherent oscillations of the Ramsey signal due to hyperfine coupling to the SiV. FIG. 13 panel (f) is a schematic showing electron-nuclear correlations measured in the ZZ-basis, according to some embodiments of the present disclosure. Light (dark) bars are before (after) correcting for known errors associated with reading out the SiV and ¹³C. FIG. 13 panel (g) is a schematic showing electron-nuclear correlations measured in the XX-basis. It is estimated a Bell state preparation fidelity of F≥0.59(4) and a concurrence C≥0.22(9).

In some embodiments, one can observe collapses in the echo signal corresponding to entanglement with nearby nuclear spins [FIG. 13 panel (a)]. As the diamond used in some exemplary embodiments has 1% ¹³C, one can typically observe several such nuclei, with all of their resonances overlapping due to their second-order sensitivity to hyperfine coupling parameters. Consequently, during a spin echo sequence the SiV entangles with many nuclei, quickly losing coherence and resulting in a collapse to <S_z>=0 [FIG. 13 panel (a), left side]. If single ¹³C can be addressed however, this entanglement results in coherent population transfer and echo collapses which can, in some cases, completely flip the SiV spin state (<S_z>=±1). This entanglement forms the basis for quantum gates [FIG. 13 panel (a), right side]. These gates can be tuned by changing the alignment of B_ext with respect to the hyperfine coupling tensor, or by using different timings. In some embodiments, as a result of the complicated nuclear bath for this device, a majority of field orientations and amplitudes show collapses to <S_z>=0. The highest fidelity nuclear gates demonstrated here are based on echo resonances with the largest contrast which, crucially, were not commensurate with an aligned field. Thus, in this device, single ¹³C could be isolated at the cost of lower SSR fidelity.

1.9.2. Initializing the Nuclear Spin

Once a single nuclear spin is identified, resonances in spin-echo form the building block for quantum gates. For example, a complete flip of the SiV is the result of the nuclear spin rotating by π conditionally around the axes±X (R_±x,SiV-C^π), depending on the state of the SiV. One can vary the rotation angle of this pulse by choosing different spacings T between pulses [FIG. 13 panel (a)], or by using different numbers of π-pulses. In some embodiments, one can find a maximally entangling gate (R_±x,SiV-C^π/2) by applying N=8 π-pulses separated by 2τ=2×2.859 μs. This can be visualized on the Bloch sphere in FIG. 13 panel (b), where the state of the SiV (1302 or 1304) induces different rotations of the ¹³C.

In some embodiments, a similarly constructed entangling gate

$R_{\overrightarrow{n\uparrow },\overrightarrow{n\downarrow }}^{\varnothing },$

can be used to coherently map population from the SiV onto the nuclear spin or map population from the nuclear spin onto the SiV [FIG. 13 panel (c)]. The fidelity of these gates is estimated by polarizing the SiV, mapping the population onto the ¹³C, and waiting for T>>T₂* (allowing coherence to decay) before mapping the population back and reading out [FIG. 13 panel (d)]. In some embodiments, it is found that one can recover 80% of the population in this way, giving us an estimated initialization and readout fidelity of F=0.9.

Based on the contrast of resonances in spin-echo (also 0.9), this can be limited by entanglement with other nearby ¹³C for this emitter, as well as other choices for τ and N. In some embodiments, coupling to other ¹³C results in population leaking out of the two-qubit register, and can be improved by increasing sensitivity to single ¹³C, or by looking for a different emitter with a different ¹³C distribution. The misaligned external field further results in slight misalignment of the nuclear rotation axis and angle of rotation and can be improved by employing adapted control sequences to correct for these errors.

1.9.3. Microwave Control of Nuclear Spins

As demonstrated above, control of the ¹³C via composite pulse sequences on the SiV is also possible. A maximally entangling gate has already been demonstrated and used to initialize the ¹³C, so in order to build a universal set of gates, in some embodiments all one requires are unconditional single-qubit rotations. This can be done where unconditional nuclear rotations occur in spin-echo sequences when the inter-pulse spacing τ is halfway between two collapses. For the following gates, one can use an unconditional π/2-pulse composed of 8 π-pulses separated by τ=0.731 μs.

FIG. 14 panel (a) is a schematic showing RF Rabi oscillations, according to some embodiments of the present disclosure. In some embodiments, applying an RF tone can directly drive nuclear rotations of a coupled ¹³C. FIG. 14 panel (b) is a plot showing SiV coherence in the presence of an RF drive, according to some embodiments of the present disclosure. As the strength of the RF drive is increased, local heating from the CPW reduces the SiV T₂.

In some embodiments, one can use this gate to probe the coherence time T₂* of the ¹³C. After mapping population onto the nuclear spin, the SiV is re-initiated, and then used to perform unconditional π/2-rotations on the ¹³C [FIG. 13 panel (d)]. Oscillations in the signal demonstrate Larmor precession of the nucleus at a frequency determined by a combination of the external field as well as ¹³C-specific hyperfine interactions, which are seen as the data points in FIG. 13 panel (d). The envelope is calculated by fitting the oscillations and extracting their amplitude. The decay of this envelope T₂*=2.2 ms shows that the ¹³C has an exceptional quantum memory, even in the absence of any dynamical decoupling.

In some embodiments, one can characterize the fidelity of the conditional and unconditional nuclear gates by generating and reading out Bell states between the SiV and ¹³C. First, one can initialize the 2-qubit register into one of the 4 eigen-states: {|↑e↑N|↑e↓|↓e≡N|↓e↓N}, then perform a π/2-pulse on the electron to prepare a superposition state. Afterward, a CNOT gate, comprised of an unconditional π/2 pulse followed by a maximally entangling gate, prepares one of the Bell states |Ψ_±|Φ_± depending on the initial state [FIG. 13 panels (e) and (f)]. in some embodiments, one can get an error corrected fidelity of F≥0.59(4) and C≥0.22(9), primarily limited by the inability to initialize the ¹³C.

1.9.4. Radio-Frequency Driving of Nuclear Spins

The previous section demonstrated a CNOT gate between SiV and ¹³C using composite MW pulses. This approach can have several drawbacks. First, the gate fidelity is limited by difficulties to finely tune the rotation angle of the maximally entangled gate which cannot be done in a continuous fashion [FIG. 13 panel (a)]. Second, in some embodiments, this gate requires a specific number of MW pulses and delays between them, making the gate duration (˜50 μs in this work) comparable to the SiV coherence time. Finally, this scheme relies on a second order splitting of individual ¹³C resonances to resolve individual ones; residual coupling to additional ¹³C limits the fidelity for a pulse sequence of given total length.

In some embodiments, direct RF control can provide be a simple way to make a fast and high-fidelity CNOT gate since it doesn't always require a single RF π-pulse on a nuclear spin transition. Furthermore, since the nuclear spin transition frequencies depend on the hyperfine coupling to leading order, these pulses could have higher ¹³C selectivity and in some embodiments shorter gate duration.

In some embodiments, one can use the RF port inside the DR to apply RF pulses resonant with nuclear spin transitions. FIG. 13 panel (a) shows RF Rabi oscillations of the nuclear spin. Since the ¹³C gyromagnetic ratio is about 3 orders of magnitude smaller compared to the SiV spin, RF driving is much less efficient than MW one and can use much more power. To identify characteristics of local heating of the SiV one can measured the SiV spin coherence contrast in spin-echo sequence right after applying off-resonant RF pulse of 100 μs at different power (calibrated via RF rabi oscillations) [FIG. 13 panel (b)]. Unfortunately, Even modest Rabi frequencies (Ω_RF˜1 kHz) result in 20% loss in SiV coherence. Replacing the gold CPW used in this work by superconducting ones may solve heating issue and make RF driving practically useful.

The SiV center in diamond has rapidly become a leading candidate to serve as the building block of a quantum network. Embodiments of the present disclosure describes the underlying technical procedures and example parameter regimes for utilizing the SiV-nanocavity system as a quantum network node. For example, the present disclosure describes the effect of static and dynamic strain on the properties of the SiV spin qubit and its optical interface, with direct application to quantum networking applications. The present disclosure demonstrates techniques for coherently controlling and interfacing SiV spin qubits inside of nanophotonic structures at millikelvin temperatures to optical photons. The present disclosure identifies and coherently controls auxiliary nuclear spins, forming a nanophotonic two-qubit register.

Embodiments of the present disclosure illustrates the path towards the realization of a first-generation quantum repeater based on SiV centers inside diamond nanodevices. In some embodiments, a key ingredient enabling large-scale applications involving several solid-state SiV-nanocavity nodes can be the incorporation of strain tuning onto each device. Precise tuning of both the static and dynamic strain can overcome the limitations of inhomogeneous broadening and spectral diffusion and enable scalable fabrication of quantum repeater nodes.

1.10. Nanophotonic Cavity Design

In some embodiments, one can simulate and optimize the nanophotonic structures to maximize atom-photon interactions while maintaining high waveguide coupling, which can provide good collection efficiency for the devices. For example, this may involve optimizing the device quality-factor to mode volume ratio, the relative rates of scattering into waveguide modes, and the size and shape of the optical mode. Each of these quantities are considered in a three-step simulation process (FDTD, Lumerical®). In some embodiments, one can first perform a coarse parameter sweep over all possible unit cells which define the photonic crystal geometry and identify families of bandgap-generating structures. These structures are the starting point for a gradient ascent optimization procedure, which results in generating high quality-factor, low mode volume resonators. Finally, the generated designs are modified to provide efficient resonator-waveguide coupling.

FIG. 15 panel (a) is a schematic showing a unit cell of a photonic crystal cavity (bounded by black lines), according to some embodiments of the present disclosure. H_z and H_y define the size and aspect ratio of the hole, a determines the lattice constant, and w sets the waveguide width. FIG. 15 panel (b) is a plot showing electric field intensity profile of the TE mode inside the cavity, indicating strong confinement of the optical mode inside the waveguide, according to some embodiments of the present disclosure. FIG. 15 panel (c) is a schematic of photonic crystal design, according to some embodiments of the present disclosure. Shaded region 1504 is the bandgap generating structure, shaded region 1502 represents the cavity structure. FIG. 15 panel (d) is a plot of a, H_x, and H_y for the cavity shown in panel (c), showing cubic taper which defines the cavity region, according to some embodiments of the present disclosure. All sizes are shown in fractions of a_nominal, the unperturbed lattice constant.

Optimization can begin by exploring the full parameter space of TE-like bandgap generating structures within the waveguide geometry. For hole-based cavities [FIG. 15 panel (a)], this sweep covers a 5-dimensional parameter space: The lattice constant of the unit cell (a), the hole size and aspect ratio (e.g., length/width H_x and H_y), the device etch angle (θ) and the waveguide width (w). Due to the size of this parameter space, one can start by performing a low-resolution sweep over all parameters, with each potential design simulated by a single unit cell with the following boundary conditions: 4 perfectly matched layer (PML) boundary conditions in the transverse directions and 2 Bloch boundary conditions in the waveguide directions. The band structure of candidate geometries is determined by sweeping the effective k-vector of the Bloch boundary condition and identifying allowed modes. Using this technique, families of similar structures with large bandgaps near the SiV transition frequency are chosen for further simulation. In some embodiments, each candidate photonic crystal is also inspected for the position of its optical mode maximum, to ensure that it has first-order modes concentrated in the center of the diamond, where SiVs will be incorporated [FIG. 15 panel (b)].

The second step is to simulate the full photonic crystal cavity design, focused in the regions of parameter space identified in step one. This is done by selecting a fixed θ, as well as a total number of unit cells that define the structure, then modifying the bandgap of the photonic crystal with a defect region to form a cavity mode. Without wish to be bound by theory, one can define this defect using a cubic tapering of one (or several) possible parameters:

A(x)=1−d_max|2x³−3x²+1|  (11)

Where A is the relative scale of the target parameter(s) at a distance χ from the cavity center, and d_max is the defect depth parameter. Photonic Crystal cavities with multi-parameter defects are difficult to reliably fabricate, therefore, devices used in this work have cavity defect geometries defined by variations in the lattice constant. Without wish to be bound by theory, the cavity generated by this defect is scored by simulating the optical spectrum and mode profile and computing the scoring function F:

F=min(Q,Q_cutoff)/(Q_cutoff×V_mode)  (12)

Where Q is the cavity quality-factor depending on the dimensions of the cavity (i.e., a dimensionless product of the cavity frequency times the “ringdown” time). In some embodiments, the Q can be calculated in simulations using the ringdown time of the cavity, i.e. by inserting light into the cavity, and measuring how long it takes for the light to leave the cavity. In experiments, it can be measured using the linewidth of the cavity spectrum (see e.g. FIG. 9 panel (a), “bare cavity” curve). This value can be proportional to the inverse of the ringdown time. In some embodiments, Q_cutoff=5×10⁵ is an estimated maximum realizable Q based on fabrication constraints, and V_mode is the cavity mode volume. Based on this criteria, one can employ a gradient ascent process over all cavity design parameters (except θ and the total number of unit cells) until F is maximized, or a maximum number of iterations has occurred. Due to the complexity and size of the parameter space, a single iteration of this gradient ascent may not always find the structure. Instead, several candidates from each family of designs found in step one are explored, with the best moving on to the final step of the simulation process. These surviving candidates can again be checked, for example to ensure confinement of the optical mode in the center of the cavity structure and to ensure that the structures fall within the tolerances of the fabrication process.

The final step in the simulation process is to modify the optimized designs to maximize resonator-waveguide coupling. This can be done by removing unit cells from the input port of the device, which decreases the overall quality-factor of the devices in exchange for better waveguide damping of the optical field. Devices are once again simulated and analyzed for the fraction of light leaving the resonator through the waveguide compared to the fraction scattering into free space. The number of unit cells on the input port is then optimized for this ratio, with simulations indicating that more than 95% of light is collected into the waveguide. In practice, fabrication defects increase the free space scattering rate, placing resonators close to the critically coupled regime. Finally, the waveguide coupling fraction is increased by appending a quadratic taper to both ends of the devices such that the optical mode is transferred adiabatically from the photonic crystal region into the diamond waveguide. This process produces the final cavity structure used for fabrication [FIG. 15 panel (c)].

1.11. Strain-Induced Frequency Fluctuations

In some embodiments, one can calculate changes the SiV spin-qubit frequency and optical transition frequency arising from strain fluctuations. Without wish to be bound by theory one can start with the Hamiltonian for SiV in an external magnetic field B_z aligned along the SiV symmetry axis:

$\begin{array}{cc}H=-\lambda (\underset{\underset{\mathrm{spin}-\mathrm{orbit}}{︸}}{\begin{array}{cccc}0& 0& i& 0\\ 0& 0& 0& -i\\ -i& 0& 0& 0\\ 0& i& 0& 0\end{array}})+\left(\underset{\underset{\mathrm{strain}}{︸}}{\begin{array}{cccc}\alpha -\beta & 0& \gamma & 0\\ 0& \alpha -\beta & 0& \gamma \\ \gamma & 0& \beta & 0\\ 0& \gamma & 0& \beta \end{array}}\right)+B_z\left(\underset{\underset{\mathrm{orbital}\mathrm{Zeeman}}{︸}}{\begin{array}{cccc}0& 0& i& 0\\ 0& 0& 0& i\\ -i& 0& 0& 0\\ 0& -i& 0& 0\end{array}}\right)+\frac{B_z}{2}\left(\underset{\underset{\mathrm{spin}\mathrm{Zeeman}}{︸}}{\begin{array}{cccc}1& 0& 0& 0\\ 0& -1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& -1\end{array}}\right)& \left(13\right)\end{array}$

where λ is a spin-orbit coupling constant, γ_L=μB and γS=2μB are Landé g-factors of the orbital and spin degrees of freedom (μ_B the Bohr magneton), q=0.1 is a Ham reduction factor of the orbital momentum, and α, β, γ are local strain parameters which can be different for the ground and excited sates.

Without wish to be bound by theory, as measuring the exact strain parameters is challenging one can assume one non-zero component in this tensor (ϵ_zx) in order to simplify the calculations. In this case, strain parameters are:

β=f_g(e)ϵzx,  (14)

α=γ=0  (15)

where f_g(e)=1.7 10⁶ (3.4 106) GHz/strain for the ground (excited) state and the GS splitting is:

Δ_GS=2√{square root over (λ_g²)}+β²,  (16)

where λg 25 GHz is the SO-constant for the ground state. Next, one can solve this Hamiltonian and determine how the qubit frequency changes as a function of relative strain fluctuations (ξ):

$\begin{array}{cc}\Delta f_{MW}=\frac{2{\left(f_{g^{ϵ}zx}\right)}^2{\lambda }_g{B}_{\mathrm{zq}\gamma L}}{{\left({\left(f_{g^{ϵ}zx}\right)}^2+{\lambda }_g^2\right)}^{3/2}}\xi .& \left(17\right)\end{array}$

The corresponding change in the optical frequency is:

$\begin{array}{cc}\Delta f_{\mathrm{optic}al}=\left(\frac{{\left(f_{g^{ϵ}zx}\right)}^2}{\sqrt{{\left(f_{g^{ϵ}zx}\right)}^2+{\lambda }_g^2}}-\frac{{\left(f_{g^{ϵ}zx}\right)}^2}{\sqrt{{\left(f_{g^{ϵ}zx}\right)}^2+{\lambda }_e^2}}\right)\xi ,& \left(18\right)\end{array}$

where λe≈125 GHz is the SO-constant for the excited state.

In some embodiments, for SiV 2 one can get Δ_GS=140 GHz and find ϵ_zx=3.8 10⁻⁵. With ξ=1% strain fluctuations (corresponding to 10⁻⁷ strain), frequencies change by Δf_MW≈4 MHz and Δf_optical≈−300 MHz. This quantitatively agrees with the data presented in [FIG. 8 panel (f)].

1.12. Mitigating Spectral Diffusion

In some embodiments, in order to couple SiV centers to a quantum network, electronic transitions can be stabilized with respect to a probe laser. Such spectral diffusion can be a challenge for solid-state quantum systems. In the case of the SiV center, spectral diffusion can be seen explicitly in FIG. 16 panel (a), where the optical transition frequency can either drift slowly (central region) or undergo large spectral jumps. As this diffusion can be larger than the SiV linewidth, any given instance of an example use could have the probe laser completely detuned from the atomic transition, resulting in failed operation.

There are several possible solutions to mitigate this spectrum diffusion. First, exploiting a high-cooperativity interface, one can Purcell-broaden the optical linewidth to exceed the spectral diffusion. Second, a high collection efficiency can be used to read out the optical position faster than the spectral diffusion. The frequency can then be probabilistically stabilized by applying a short laser pulse at 520 nm which dramatically speeds up the timescale of spectral diffusion [FIG. 16 panel (b)]. Alternatively, this signal could be used to actively stabilize the line using strain-tuning. From the observations in FIG. 6 panel (f), this technique should mitigate spectral diffusion of both the optical and spin transitions. Strain tuning also offers the capability to control the DC strain value, which has effects on qubit properties as discussed previously, and enables tuning multiple SiV centers to a common network operation frequency. As such, this tunability is desirable for quantum networking technologies based on SiV centers.

The severity of spectral diffusion is different for different emitters however, and this control is not always required. For SiV 1, the main SiV used in the following sections, one can find almost no spectral diffusion, with optical transitions stable over many minutes [FIG. 16 panel (c)]. This is an example configuration, as some applications can be implemented without any need to verify the optical line position.

FIG. 16 panel (a) is a plot of spectral diffusion of SiV 2, according to some embodiments of the present disclosure. In some embodiments, one can observe slow spectral wandering as well as spectral jumps. FIG. 16 panel (b) shows applying a short green repumping pulse before every measurement significantly speeds up the timescale for spectral diffusion, according to some embodiments of the present disclosure. FIG. 16 panel (c) is a plot of spectral diffusion of SiV 1 in nanostructures, according to some embodiments of the present disclosure. Line is stable to below 100 MHz over many minutes. Scale bar indicates normalized SiV reflection signal.

1.13. Model for SiV Decoherence

The scaling of T₂ (N) N^2/3 is identical to that found for nitrogen-vacancy centers, where it is assumed that T₂ is limited by a fluctuating electron spin bath Motivated by DEER measurements with SiV 2, one can estimate the noise bath observed by SiV 1.

Without wish to be bound by theory, the measured coherence decay is modeled by:

S _z =Exp(−∫dw S(ω)F_N(t,ω)),  (19)

where S(ω) is the noise power-spectrum of the bath, and F_N (t, ω)=2 sin(ωt/2)(1 sec(ωt/2N))²/ω² is filter function for a dynamical-decoupling sequence with an even number of pulses. One can fit successive T₂ echo curves to the functional form A+Be^−(t/T2)β, with A, B being free parameters associated with photon count rates, and β=3 providing the best fit to the data. This value of β implies a decoherence bath with a Lorentzian noise power-spectrum, S(ω, b, τ)=b2τ/π1/(1+ω²τ²), where b is a parameter corresponding to the strength of the noise bath, and τ is a parameter corresponding to the correlation time of the noise.

Empirically, no one set of noise parameters faithfully reproduces the data for all measured echo sequences. Adding a second source of dephasing {tilde over (S)}=S(ω, b₁, τ₁)+S(ω, b₂, τ₂), gives reasonable agreement with the data using parameters b₁=5 kHz, τ₁=1 μs, b₂=180 kHz, τ₂=1 ms [FIG. 11 panel (d)]. The two drastically different set of noise parameters for each of the sources can help illuminate the source of noise in the devices.

As explained in the previous section, one candidate for this decoherence can be a bath of free electrons arising from improper surface termination or local damage caused during nanofabrication, which are known to have correlation times in the μs range. The SiV studied in this analysis is approximately equidistant from three surfaces: the two nearest holes which define the nanophotonic cavity, and the top surface of the nanobeam, all of which are approximately 50 nm away. One can estimate a density of σ_surf=0.067 spins/nm2 using:

$\begin{array}{cc}{b}_1{=}_{\gamma \mathrm{SiV}}\langle B_{\mathrm{surf}}\text{?}=\frac{g^2{\mu }^2{B}^{{\mu }_0}}{\stackrel{\_}{h}}\frac{1}{4\pi \sum d_i^2}\sqrt{\frac{\pi }{4{\sigma }_{\mathrm{surf}}}}\text{?}\text{indicates text missing or illegible when filed}& \left(20\right)\end{array}$

where b₁ is the measured strength of the noise bath, g is the electron gyromagnetic ratio, and d_i are the distances to the nearest surfaces. This observation is consistent with surface spin densities measured using NVs.

The longer correlation time for the second noise term suggests a different bath, possibly arising from free electron spins inside the bulk diamond. Vacancy clusters, which can persist under annealing even at 1200 C, are known to possess g=2 electron spins, and are one possible candidate for this noise bath. Integrating over din Eqn. 20, one can estimate the density of spins that can be used to achieve the measured b₂. One can estimate ρ_bulk˜0.53 spins per nm³, which corresponds to a doping of 3 ppm. Interestingly, this is nearly identical to the local concentration of silicon incorporated during implantation (most of which is not successfully converted into negatively charged SiV) and could imply implantation-related damage as a possible source of these impurities.

Another possible explanation for this slower bath could be coupling to nuclear spins in the environment. The diamond used in this example has a natural abundance of ¹³C, a spin-½ isotope, in concentrations of approximately 1.1%. Replacing μ_B→μ_N in the term for (B) gives an estimated nuclear spin density of ρ(bulk,N)=0.6%, a factor of two different than the expected nuclear spin density.

1.14. Concurrence and Fidelity Calculations

1.14.1. Spin-Photon Concurrence and Fidelity Calculations

From correlations in the Z- and X-bases, one can estimate a lower bound for the entanglement in the system. Without wish to be bound by theory, the density matrix of the system conditioned on the detection of one photon can be described as:

$\begin{array}{cc}\rho zz=1/2\left(\begin{array}{cccc}{p}_{e\uparrow }& 0& 0& 0\\ 0& p_{e\downarrow }& c_{e\downarrow ,l\uparrow }& 0\\ 0& c_{e\downarrow ,l\uparrow }^{†}& p_{l\uparrow }& 0\\ 0& 0& 0& p_{l\downarrow }\end{array}\right)& \left(21\right)\end{array}$

where p_ij are the probabilities of measuring a photon in state i, and the spin in state j. c_e↓, l↑ represents entanglement between p_e↑ and p_l↓. One can set all other coherence terms to zero, as they represent negligibly small errors in the system (for example, c_e↑,e↓>0 can imply that the SiV was not initialized properly at the start of the measurement). Without wish to be bound by theory, one can quantify the degree of entanglement in the system by its concurrence , which is 0 for separable states, and 1 for a maximally entangled state:

$\begin{array}{cc}C=\mathrm{Max}\left(0,{\lambda }_0^{1/2}-\sum _{i=1}^N{\lambda }_i^{1/2}\right)& \left(22\right)\end{array}$

where λ_i are the eigenvalues of the matrix pzz·(σ_y·pzz·σ_y^†), and σ_y is the standard Pauli matrix acting on each qubit basis separately (σ_y=σ_y,ph⊗σ_y,e1). While this can be solved exactly, the resulting equation is complicated. Without wish to be bound by theory, taking the first-order terms, this can be simplified to put a lower bound on the concurrence:

C≥2(|c_e↓,l↑|=√{square root over (p_e↑p_l↓)})  (23)

One can measure p directly in the Z basis, and estimate |c_e↓,l↑| by performing measurements in the X basis. A π/2-rotation on both the photon and spin qubits rotates:

e →1/√{square root over (2)}(|e+|l),|l→1/√{square root over (2)}(|e−|l)

|↓→1/√{square root over (2)}(|↓+|↑),|↑→1/√{square root over (2)}(|↓−|↑)

After this transformation, the signal contrast directly measures c_e↓,l↑:

2c_e↓,l↑=p_−,←+p_+,→−p_−→−p_+←⇒C≥0.42(6)  (24)

Similarly, the fidelity of the entangled state (post-selected on the detection of a photon) can be computed by the overlap with the target Bell state:

F= Ψ ⁺|_pzz|Ψ⁺=(p_e↑+p_l↓+2C_e↓,l↑)2≥0.70(3)  (25)

1.14.2. Correcting for Readout Infidelity

Errors arising from single-shot readout incorrectly assign the spin state, results in lower-contrast histograms for spin-photon correlations. One can correct for readout errors using a transfer matrix formalism. The measured spin-photon correlations p_ij are related to the ‘true’ populations P_ij via:

$\begin{array}{cc}\left(\begin{array}{c}{p}_{e\downarrow }\\ p_{e\uparrow }\\ p_{l\downarrow }\\ p_{l\uparrow }\end{array}\right)=(\begin{array}{cccc}{F}_{\uparrow }& 1-F_{\uparrow }& 0& 0\\ 1-F_{\downarrow }& F_{\uparrow }& 0& 0\\ 0& 0& F_{\downarrow }& 1-F_{\uparrow }\\ 0& 0& 1-F_{\downarrow }& F_{\uparrow }\end{array})\left(\begin{array}{c}{p}_{e\downarrow }\\ p_{e\uparrow }\\ p_{l\downarrow }\\ p_{l\uparrow }\end{array}\right)& \left(26\right)\end{array}$

with F_↓, F_↑ defined above. After this correction, an identical analysis is performed to calculate the error-corrected histograms [FIG. 12 panels (b)-(d) dark-shading]. One can find an error-corrected concurrence C≥0.79(7) and fidelity F≥0.89(3).

Where F↓,_e≈F↑,_e=0.85 and F↓,_N≈F↑,_N=0.72. Following this analysis, one can obtain an error-corrected concurrence of C≥0.22(9).

1.14.3. Electron-Nuclear Concurrence and Fidelity Calculations

For spin-spin Bell states, in contrast to the spin-photon analysis, one can no longer set any of the off-diagonal terms of the density matrix [Eqn. 21] to zero due to the limited (˜90%) nuclear initialization fidelity. It is noted that neglecting these off-diagonal terms can decrease the estimated entanglement in the system, thus the concurrence can still be written as:

C≥2(|c_↓↑|−√{square root over (p_↑↑p_↓↓)})  (27)

where the first subscript is the electron spin state, and the second is the nuclear state. One can estimate c↓↑ again by using the measured populations in an orthogonal basis. In this case, off-diagonal terms add a correction:

2c_↓↑+2c_↑↓=p_←←+p_→→−p_←→−p_→←  (28)

In order for the density matrix to be properly normalized, c_↑↓≤√{square root over (p_↑↑p_↓↓)}, giving us the final concurrence:

C≥p _←← +p _→→ −p _←→ −p _→←−4√{square root over (p_↑↑p_↓↓)}  (29)

Additionally, both electron readout error as well as ¹³C mapping infidelity can misreport the true spin state. As such, the new transfer matrix to correct for this error is:

$\begin{array}{cc}(\begin{array}{cccc}{F}_{\downarrow ,e}{F}_{\downarrow ,N}& F_{\downarrow ,e}\left(1-F_{\uparrow ,N}\right)& \left(1-F_{\uparrow ,e}\right)F_{\downarrow ,N}& \left(1-F_{\uparrow ,e}\right)\left(1-F_{\uparrow ,N}\right)\\ F_{\downarrow ,e}\left(1-F_{\downarrow ,N}\right)& F_{\downarrow ,e}{F}_{\uparrow ,N}& \left(1-F_{\uparrow ,e}\right)\left(1-F_{\downarrow ,N}\right)& \left(1-F_{\uparrow ,e}\right)F_{\uparrow ,N}\\ \left(1-F_{\downarrow ,e}\right)F_{\downarrow ,N}& \left(1-F_{\downarrow ,e}\right)\left(1-F_{\uparrow ,N}\right)& F_{\uparrow ,e}{F}_{\downarrow ,N}& F_{\uparrow ,e}\left(1-F_{\uparrow ,N}\right)\\ \left(1-F_{\downarrow ,e}\right)(1-F_{\downarrow ,N}& \left(1-F_{\downarrow ,e}\right)F_{\uparrow ,N}& F_{\uparrow ,e}\left(1-F_{\downarrow ,N}\right)& F_{\uparrow ,e}{F}_{\uparrow ,N}\end{array})& \left(30\right)\end{array}$

1.14.4. Electron-Nuclear CNOT Gate

One can further characterize the CNOT gate itself as a universal quantum gate. Due to the relatively poor read-out fidelity (see above), one may not do this by performing quantum state tomography. Instead, one can estimate entries in the CNOT matrix using measurements in the Z-basis. As a control measurement, one can first initialize the two qubits in all possible configurations and read out, averaged over many trials. Next, one can initialize the qubits, perform a CNOT gate, and read out, again averaged over many trials, normalized by the control data. Any reduction in contrast after normalization is attributed to the opposite spin state, establishing a system of equations for determining the CNOT matrix. One can solve this system of equations, marginalizing over free parameters to determine an MLE estimate for the CNOT transfer matrix, as seen in reference.

1.14.5. Nuclear Initialization and Readout

FIG. 17 panel (a) is a schematic showing original initialization sequence _z,C^π/2 rotation, according to some embodiments of the present disclosure. FIG. 17 panel (b) is a schematic showing simplified initialization sequence used in this work, according to some embodiments of the present disclosure. FIG. 17 panel (c) is a plot showing simulated performance of the initialization gate from panel (b) using 8 π-pulses per each nuclear gate, the initial state being |↑↑> (1702) and |↑↓> (1704), according to some embodiments of the present disclosure. The resonances are narrow compared to panel (d) due to applying effectively twice more π-pulses. FIG. 17 panel (d) is a plot showing simulated performance of _±,SiV-C^π/2 gate for 8 π-pulses for SiV-¹³C register initialized in |↑↑> (1706) and |↑↓> (1708), according to some embodiments of the present disclosure.

Initialization (and readout) of the ¹³C spin can be done by mapping population between the SiV spin and the ¹³C. It is noted that Z and X gates are possible with dynamical-decoupling based nuclear gates, thus a natural choice for initialization are gates comprised of both _±x,SiV-C^π/2 and _z,SiV-C^π/2, as shown in FIG. 17 panel (a). It is noted here that it should be possible to combine the effects of _x and _z rotations in a single gate, which can shorten and simplify the total initialization gate, in some embodiments. Without wish to be bound by theory, one proposed sequence uses the following entangling gate:

$\begin{array}{cc}{ℛ}_{\overrightarrow{n\uparrow },\overrightarrow{n\downarrow }}^{\varphi }=(\begin{array}{cccc}\left(1-i\right)/2& i/\sqrt{2}& 0& 0\\ i/\sqrt{2}& \left(1-i\right)/2& 0& 0\\ 0& 0& \left(1-i\right)/2& -i/\sqrt{2}\\ 0& 0& -i/\sqrt{2}& \left(1-i\right)/2\end{array})=\left(\begin{array}{cc}{ℛ}_{\ominus =\pi /4}^{\pi /2}{ℛ}_z^{\pi /2}& 0\\ 0& {ℛ}_{\ominus =\pi /4}^{-\pi /2}{ℛ}_z^{\pi /2}\end{array}\right)& \left(31\right)\end{array}$

which corresponds to a rotation on the angle ϕ=2χ/3 around the axes n_↑,↓={±√{square root over (2)}, 0, 1}/√{square root over (3)}. Without wish to be bound by theory, the matrix of entire initialization gate [FIG. 17 panel (b)] built from this gate can then be:

$\begin{array}{cc}\mathrm{Init}=(\begin{array}{cccc}0& 0& -\left(1-i\right)/2& -1/\sqrt{2}\\ i/\sqrt{2}& -\left(1-i\right)/2& 0& 0\\ 0& 0& -\left(1-i\right)/2& -i/\sqrt{2}\\ 1/\sqrt{2}& \left(1-i\right)/2& 0& 0\end{array})& \left(32\right)\end{array}$

Which results in an initiated ¹³C spin.

To demonstrate this, one can numerically simulate a MW pulse sequence using the exact coupling parameters of the ¹³C and 8 π-pulses for each _n↑,n↓^ϕ gate. FIG. 17 panel (c) shows that regardless of the initial state, the ¹³C ends up in state |↓> (given that the SiV was initialized in |↑>). As expected, the timing of this gate (τ_init=2.857 μs) is noticeably different from the timing of the _±x,SiV-C^π/2 gate (τ_π/2=2.851 μs), which occurs at spin-echo resonances [FIG. 17 panel (d)].

The rotation matrix for this sequence at τ=τ_init (with the SiV initialized in |↑>) is:

$\begin{array}{cc}{ℛ}_{n_{\uparrow }}^{\varphi }\left(\begin{array}{cc}0.55+0.51i& 0+0.65i\\ 0.65i& 0.55-0.52i\end{array}\right)& \left(33\right)\end{array}$

corresponding to a rotation angle ϕ=0.63π around the axis n_↑={0.78, 0, 0.62}, very close to theoretical result. Since the example fidelities for both initialization gates [FIG. 17 panels (a) and (b)] are similar, one can use sequence (b) to make errors.

2. Memory-Enhanced Quantum Communication

The ability to communicate quantum information over long distances is of central importance in quantum science and engineering. For example, it enables secure quantum key distribution (QKD) relying on fundamental physical principles that prohibit the “cloning” of unknown quantum states. The range of QKD can be limited by photon losses and cannot be extended using straightforward measure-and-repeat strategies without compromising its unconditional security. Alternatively, quantum repeaters, which utilize intermediate quantum memory nodes and error correction techniques, can extend the range of quantum channels.

Efficient, long-lived quantum memory nodes are expected to play a role in extending the range of quantum communication, as they enable asynchronous quantum logic operations, such as Bell-state measurements (BSM), between optical photons. For example, the BSM is crucial to MDI-QKD, which is a specific implementation of quantum cryptography illustrated in FIG. 18 panel (a). Two remote communicating parties, Alice and Bob, try to agree on a key that is secure against potential eavesdroppers. They each send a randomly chosen photonic qubit {|±x, |±y} encoded in one of two conjugate bases (X or Y) across a lossy channel to an untrusted central node (Charlie), who is asked to perform a BSM and report the result over an authenticated public channel. After a number of iterations, Alice and Bob publicly reveal their choice of bases to obtain a sifted key from the cases when they used a compatible basis. A provably secure key can subsequently be extracted provided the BSM error rate is low enough. While MDI-QKD can be implemented with just linear optics and single photon detectors, the BSM in this “direct-transmission” approach can be successful when photons from Alice and Bob arrive simultaneously. Thus, when Alice and Bob are separated by a lossy fiber with a total transmission probability p_A→B<<1, Charlie measures photon coincidences with probability also limited by p_A→B, leading to a fundamental bound on the maximum possible secret key rate of R_max=p_A→B/2 bits per channel use for an unbiased basis choice. While linear optical techniques to circumvent this bound are now being actively explored, they offer limited improvement and cannot be scaled beyond a single intermediate node. Alternatively, this bound can be broken using a quantum memory node at Charlie's location. In this approach, illustrated in FIG. 18 panel (b), the state of Alice's photon is efficiently stored in the heralded memory while awaiting receipt of Bob's photon over the lossy channel. Once the second photon arrives, a BSM between Alice's and Bob's qubits yields a secret key rate that for an ideal memory scales as R_s ∝√{square root over (p_A→B)}, which can lead to substantial improvement over direct transmission in some embodiments. Beyond this specific protocol, memory-based asynchronous Bell-state measurements are central for the realization of scalable quantum repeaters with multiple intermediate nodes.

2.1. Demonstration of Memory-Enhanced Quantum Communication

Embodiments of the present disclosure describe an example realization of memory-enhanced quantum communication. In some embodiments, one can use a single solid-state spin memory integrated in a nanophotonic diamond resonator to implement asynchronous photonic Bell-state measurements. This enables a four-fold increase in the secret key rate of measurement device independent (MDI)-QKD over the loss-equivalent direct-transmission method while operating at megahertz clock rates. The results represent a significant step towards practical quantum repeaters and large-scale quantum networks.

Embodiments of the present disclosure describe the operation of such a quantum memory node, enabling MDI-QKD at rates that exceed those of an ideal system based on linear optics. In some embodiments, the realization is based on a single silicon-vacancy (SiV) color-center integrated inside a diamond nanophotonic cavity [FIG. 19 panel (a)]. Its key figure-of-merit, the cooperativity C, describes the ratio of the interaction rate with individual cavity photons compared to all dissipation rates. A low mode volume (0.5(λ/n)³), high quality factor (2×10⁴), and nanoscale positioning of SiV centers enable an exceptional C=105±11. Cavity photons are critically coupled to a waveguide and adiabatically transferred into a single-mode optical fiber that is routed to superconducting nanowire single-photon detectors, yielding a full system detection efficiency of about 85%. The device is placed inside a dilution refrigerator, resulting in electronic spin quantum memory time T₂>0.2 ms.

The operating principle of the SiV-Cavity based spin-photon interface is illustrated in FIG. 19 which shows heralded spin-photon gate, according to some embodiments of the present disclosure. Panel (a) is a schematic of memory-assisted implementation of Charlie's measurement device. Weak pulses derived from a single laser simulate incoming photons from Alice and Bob. Reflected photons are detected in a heralding setup (dashed box). Panel (b) is a reflection spectrum of memory node, showing spin-dependent device reflectivity. Panel (c) is a histogram of detected photon numbers during a 30 μs laser pulse, enabling single-shot readout based on a threshold of 7 photons. The inset shows electron spin quantum jumps under weak illumination. Panel (d) is a schematic of spin-photon quantum logic operation used to generate and verify spin-photon entangled state. Panel (e) is a plot showing characterization of resulting spin-photon correlations in the ZZ and XX bases. Dashed bars show ideal values. Panel (f) is a plot showing measured spin-photon entanglement fidelity as a function of n, the average incident photon number during each initialization of the memory.

Spin dependent modulation of the cavity reflection at incident probe frequency f₀ [FIG. 19 panel (b)] results in the direct observation of electron spin quantum jumps [FIG. 19 panel (c), inset], enabling nondestructive single-shot readout of the spin state [FIG. 19 panel (c)] in 30 μs with fidelity F=0.9998_−0.002^+0.002. Coherent control of the SiV spin qubit (f_Q≈12 GHz) is accomplished using microwave fields delivered via an on-chip gold coplanar waveguide. In some embodiments, one can utilize both optical readout and microwave control to perform projective feedback-based initialization of the SiV spin into the |↓ state with a fidelity of F=0.998±0.001. Spin-dependent cavity reflection also enables quantum logic operations between an incoming photonic time-bin qubit and the spin memory. In some embodiments, one can characterize this by using the protocol illustrated in FIG. 19 panel (d) to generate the spin-photon entangled state a (|e↑+|l↓)/√{square root over (2)} conditioned on successful reflection of an incoming single photon with overall heralding efficiency η=0.0423±0.004. Here, |e and |l|l) denote the presence of a photon in an early or late time-bin separated by δt=142 ns respectively. One can characterize the entangled state by performing measurements in the joint spin-photon ZZ and XX bases [FIG. 19 panel (e)], implementing local operations on the reflected photonic qubit with a time-delay interferometer [FIG. 19 panel (a), dashed box]. By lowering the average number of photons n incident on the device during the SiV memory time, one can reduce the possibility that an additional photon reaches the cavity without being subsequently detected, enabling high spin-photon gate fidelities for small n [FIG. 19 panel (f)]. For n=0.002 one can measure a lower bound on the fidelity of the spin-photon entangled state of F≥0.994±0.008, primarily limited by residual reflections from the |↓ state.

FIG. 20 is a schematic showing asynchronous Bell-state measurements using quantum memory, according to some embodiments of the present disclosure. Panel (a) shows an example sequence with N=6 photonic qubits sent in a single memory time. Microwave π pulses (2002) are interleaved with incoming optical pulses. Photons have fixed amplitude (2004) and qubits are defined by the relative phases between subsequent pulses (2006). Panel (b) shows a schematic of Bloch sphere representation of input photonic time-bin qubits used for characterization. Panel (c) shows characterization of asynchronous BSM. Conditional probabilities for Alice and Bob to have sent input states (i, j) given a particular parity outcome for input states in the X (top) and Y (bottom) bases. Panel (d) shows Bell test using the CHSH inequality. Conditioned on the BSM outcome, the average correlation between input photons is plotted for each pair of bases used. Shaded backgrounds denote the expected parity.

This spin-photon logic gate can be directly used to herald the storage of an incoming photonic qubit by interferometrically measuring the reflected photon in the X basis. To implement memory-assisted MDI-QKD, one can extend this protocol to accommodate a total of N photonic qubit time-bins within a single initialization of the memory [FIG. 20 panel (a)]. Each individual time-bin qubit is encoded in the relative amplitudes and phases of a pair of neighboring pulses separated by δt. Detection of a reflected photon heralds the arrival of the photonic qubit formed by the two interfering pulses without revealing its state. Two such heralding events, combined with subsequent spin-state readout in the X basis, constitute a successful BSM on the incident photons. This can be understood without loss of generality by restricting input photonic states to be encoded in the relative phase ϕ between neighboring pulses with equal amplitude: (|e+e^iϕ|l)/√{square root over (2)} [FIG. 20 panel (b)]. Detection of the first reflected photon in the X basis teleports its quantum state onto the spin, resulting in the state (|↑+m₁e^iϕi|↓)/√{square root over (2)}, where m₁=±1 depending on which detector registers the photon. Detection of a second photon at a later time within the electron spin T₂ results in the spin state (|↑+m₁m₂e^i(ϕ1+ϕ2)|↓)/√{square root over (2)}. The phase of this spin state can depend on the sum of the incoming phases and the product of their detection outcomes, but not the individual phases themselves. As a result, if the photons were sent with phases that meet the condition ϕ₁+ϕ₂ϵ{0, π}, a final measurement of the spin in the X basis (m₃=±1) completes an asynchronous photon-photon BSM, distinguishing two of the four Bell-states based on the total parity m₁m₂m₃=±1.

This approach can be directly applied to generate a secure key within the MDI-QKD protocol illustrated in FIG. 18 panel (a). One can analyze the system performance by characterizing the overall quantum-bit error rate (QBER) for N=124 photonic qubits per memory initialization. One can use several random bit strings of incoming photons from {|±x,|±y} and observe strong correlations between the resulting BSM outcome and the initial combination of input qubits for both bases [FIG. 20 panel (c)]. Using this method, one can estimate the average QBER to be E=0.116±0.002 for all combinations of random bit strings measured, significantly below the limit of E_i=0.146 providing security against individual attacks. This value is affected by technical imperfections in the preparation of random strings of photonic qubits. One can find specific periodic patterns of photonic qubits to be less prone to these effects, resulting in a QBER as low as E=0.097±0.006, which falls within the threshold for unconditional security of E_u=0.110 with a confidence level of 0.986. One can further verify security by testing the Bell-CHSH inequality using input states from four different bases, each separated by an angle of 45°. One can find that the correlations between input photons [FIG. 20 panel (d)] violate the Bell-CHSH inequality S_±<2, observing S₊=2.21±0.04 and S₋=2.19±0.04 for positive and negative BSM parity results respectively. This result demonstrates that this device can be used for fundamentally secure quantum communication.

FIG. 21 shows the performance of memory-assisted quantum communication with a log-log plot of key rate in bits per channel use versus effective channel transmission (p_A→B=n, where n is the average number of photons incident on the measurement device per photonic qubit). Without being bound by theory, line 2102 shows theoretical maximum for equivalent direct transmission MDI-QKD applications. Open circles 2104 show example measured sifted key rate (line 2106 is the expected rate). In some embodiments, to facilitate operation of the memory, n=nN≈0.02 is kept constant. From left to right, points correspond to N={60, 124, 248, 504}. Filled circles 2108 show secure key rates R_S using memory device. Vertical error bars are given by the 68% confidence interval and horizontal error bars represent the standard deviation of the systematic power fluctuations.

Finally, one can benchmark the performance of memory-assisted QKD. For each example, one can model an effective channel loss by considering the mean photon number n incident on the device per photonic qubit. Assuming that Alice and Bob emit roughly one photon per qubit, this yields an effective channel transmission probability p_A→B=n, resulting in the maximal secret key rate R_max per channel use for direct transmission MDI-QKD, given by the red line in FIG. 21. Without wish to be bound by theory, this may be a theoretical upper bound on linear optics based MDI-QKD, assuming ideal sources and detectors and balanced basis choices. The measured sifted key rates of the memory-based device are plotted as open circles 2104 in FIG. 21. Due to the high overall heralding efficiency and the large number of photonic qubits per memory time (up to N=504), the memory-assisted sifted key rate exceeds the capability of direct-transmission MDI-QKD by a factor of 78.4±0.7 at an effective channel loss of about 88 dB.

In practice, errors introduced by the quantum memory node could leak information to the environment, reducing the security of the sifted key. The fraction of secure bits r_s that can be extracted from a sifted key with finite QBER using conventional error correction and privacy amplification techniques rapidly diminishes as the QBER approaches E_i=0.147. For each value of the effective channel loss, one can estimate the QBER and use it to compute r_s, enabling extraction of distilled secure key rates R_S, plotted in black in FIG. 21. Even after error-correction, one can find that the memory-assisted secret key rate outperforms the ideal limit for the corresponding direct-transmission implementation of MDI-QKD by a factor of up to R_S/R_max=4.1±0.5(±0.1 systematic uncertainty, for N=124). In some embodiments, one can further find that this rate also exceeds the fundamental bound on repeaterless communication R_S≤1.44p_A→B with a statistical confidence level of 99.2% (_−0.3%^+0.2% systematic uncertainty). Despite overhead time associated with operating the quantum memory node (T_R in FIG. 17 panel (b)), the performance of the memory assisted BSM (for N=248) enables MDI-QKD that is competitive with an ideal unassisted system running at a 4 MHz average clock rate.

Embodiments of the present disclosure demonstrate the viability of memory-enhanced quantum communication and represent a crucial step towards realizing functional quantum repeaters. Several other technical aspects can be combined with embodiments of the present disclosure for quantum communication. In an example, this protocol can be implemented using truly independent, distant communicating parties. In another example, frequency conversion from telecommunications wavelengths, as well as low-loss optical elements used for routing photons to and from the memory node, can be incorporated. In another example, rapid generation of provably secure keys can be realized using an implementation of decoy-state protocols, biased bases, and finite-key analyses, all compatible with the present approach. The present approach is well-suited for such deployments. Embodiments of the present disclosure do not require phase stabilization of long-distance links and operates efficiently in the relevant regime of p_A→B≈70 dB, corresponding to about 350 km of telecommunications fiber. Additionally, a single device can be used at the center of a star (e.g., multi-spoked) network topology, enabling quantum communication between several parties beyond the metropolitan scale. Furthermore, the present approach can be extended along several directions. The use of long-lived ¹³C nuclear spin qubits can eliminate the need to operate at low total n and can provide longer storage times, which can enable hundred-fold enhancement of BSM success rates, in some embodiments. Recently implemented strain-tuning capabilities should allow for operation of many quantum nodes at a common network frequency. Unlike linear optics-based alternatives, the approach presented here can be extended to implement the full repeater protocol, enabling a polynomial scaling of the communication rate with distance. Finally, the demonstrated multi-photon gate operations can also be adapted to engineer large cluster-states of entangled photons, which can be utilized for rapid quantum communication. Implementation of these techniques could enable the realization and applications of scalable quantum networks beyond QKD, ranging from non-local quantum metrology to modular quantum computing architectures.

2.2. Example Setup

In some embodiments, measurements can be performed in a dilution refrigerator (DR, BlueFors® BF-LD250) with a base temperature of 20 mK. The DR is equipped with a superconducting vector magnet (American Magnets Inc. 6-1-1 T), a home-built free-space wide-field microscope with a cryogenic objective (Attocube® LT-APO-VISIR), piezo positioners (Attocube® ANPx101 and ANPx311 series), and fiber and MW feedthroughs. Tuning of the nanocavity resonance is performed using a gas condensation technique. The SiV-Cavity system is optically interrogated through the fiber network without any free-space optics. The operating temperature of the memory node during the BSM measurements was 100-300 mK.

2.2.1. Example Implementation of Asynchronous BSM

FIG. 22 is an example setup schematic, according to some embodiments of the present disclosure. Panel (a) shows control flow of experiment. Opt (MW) AWG 2202 is a Tektronix® AWG7122B 5 GS/s (Tektronix® AWG70001a 50 GS/s) arbitrary waveform generator used to generate photonic qubits (microwave control signals). All signals are recorded on a time-tagger 2204 (TT, PicoQuant® HydraHarp 400). Panel (b) shows a schematic of a fiber network used to deliver photons to and collect photons from the memory device, including elements for polarization control and diagnostic measurements of coupling efficiencies. Panel (c) shows a schematic of preparation of optical fields. The desired phase relation between lock and qubit paths is facilitated by modulating AOMs using phase-locked RF sources with a precise 1.8 MHz frequency shift between them.

An asynchronous BSM [FIG. 20 panel (a)] relies on (1) precise timing of the arrival of optical pulses (corresponding to photonic qubits from Alice and Bob) with microwave control pulses on the quantum memory and (2) interferometrically stable rotations on reflected time-bin qubits for successful heralding. In order to accomplish (1), all equipment used for generation of microwave and optical fields is synchronized by a single device (National Instruments® HSDIO, FIG. 22 panel (a)) with programming described in TABLES 1 and 2.

In order to accomplish (2), one can use a single, narrow linewidth (<50 kHz) Ti:Sapphire laser (M Squared® SolsTiS-2000-PSX-XF, FIG. 22 panel (b)) both for generating photonic qubits and locking the time-delay interferometer (TDI) used to herald their arrival. In the example, photonic qubits are reflected from the device, sent into the TDI, and detected on superconducting nanowire single photon detectors (SNSPD, Photon Spot™) All detected photons are processed digitally on a field-programmable gate array 2206 (FPGA, FIG. 22 panel (a)), and the arrival times of these heralding signals are recorded on a time-tagger 2204 (TT, FIG. 22 panel (a)), and constitute one bit of information of the BSM (m₁ or m₂). At the end of the example application, a 30 μs pulse from the readout path is reflected off the device, and photons are counted in order to determine the spin state (m₃) depending on the threshold shown in FIG. 19 panel (c).

To minimize thermal drift of the TDI, it is mounted to a thermally weighted aluminum breadboard, placed in a polyurethane foam-lined and sand filled briefcase, and secured with glue to provide passive stability on the minute timescale. One can halt the process and actively lock the interferometer to the sensitive Y-quadrature every 200 ms by changing the length of the roughly 28 m long (142 ns) delay line with a cylindrical piezo. In order to use the TDI for X-measurements of the reflected qubits, one can apply a frequency shift of 1.8 MHz using the qubit AOM, which is ¼ of the free-spectral range of the TDI. Since the nanophotonic cavity, the TDI, and the SNSPDs are all polarization sensitive, one can use various fiber-based polarization controllers [FIG. 22 panel (b)]. All fibers in the network are covered with aluminum foil to prevent thermal polarization drifts. This results in an interference visibility of the TDI of >99% that is stable for several days without any intervention with lab temperature and humidity variations of ±1° C. and ±5% respectively.

In some embodiments, in order to achieve high-fidelity operations one can ensure that the laser frequency (which is not locked) is resonant with the SiV frequency f₀ (which is subject to the spectral diffusion). To do that one can implement a so-called preselection procedure, described in TABLES 1 and 2 and FIG. 22 panel (a). First, the SiV spin state is initialized by performing a projective measurement and applying microwave feedback. During each projective readout, the reflected counts are compared with two thresholds: a “readout” threshold of 7 photons (used to record m₃), and a “status” threshold of 3 photons. The status trigger is used to prevent the process from running in cases when the laser is no longer on resonance with f₀, or if the SiV has ionized to an optically inactive charge state. The duty cycle of the status trigger is externally monitored and is used to temporarily abort the process and run an automated re-lock procedure that locates and sets the laser to the new frequency f₀, reinitailizing the SiV charge state with a 520 nm laser pulse. This protocol enables fully automated operation at high fidelities (low QBER) for several days without human intervention.

2.2.2. Calibration of Fiber Network

The schematic of the fiber-network used to deliver optical pulses to and collect reflected photons from the nanophotonic memory device is shown in FIG. 22 panel (b). Photons are routed through the lossy (1%) port of a 99:1 fiber beam-splitter (FBS) to the nanophotonic device. In some embodiments, it is noted that for practical implementation of memory-assisted quantum communication, an efficient optical switch or circulator should be used instead. In this example, since one focuses on benchmarking the performance of the memory device itself, the loss introduced by this beam-splitter is incorporated into the estimated channel loss. Reflected photons are collected and routed back through the efficient (99%) port of the FBS and are sent to the TDI in the heralding setup.

TABLE 1
High-level example sequence.
Step	Process	Duration	Proceed to
1	Lock time-delay	200	ms	2
	interferometer
2	Readout SiV	30	μs	If status LOW:
				4, else: 3
3	Apply microwave π pulse	32	ns	2
4	Run main example script	~200	ms	1

As shown in TABLE 1, this sequence is programmed into the HSDIO and uses feedback from the status trigger sent from the FPGA [see FIG. 22 panel (a)]. Main example sequence is described in TABLE 2. External software is also used to monitor the status trigger. If it is HI for ≥2 s, the software activates an automatic re-lock procedure which compensates for spectral diffusion and ionization of the SiV center.

TABLE 2
Main example sequence for memory-enhanced quantum
communication
Step	Process	Duration	Proceed to
1	Run sequence in FIG. 3a for a	10-20 μs	2
	given N
2	Readout SiV + report readout	30 μs	If status LOW:
	to TT		1, else: 3
3	Apply microwave π pulse	32 ns	4
4	Readout SiV	30 μs	If status LOW:
			3, else: 1

As shown in TABLE 2, this script is followed until step 1 is run a total of 4000 times, and then terminates and returns to step 1 of TABLE 1. The longest step is the readout step, which is limited by the fact that one can operate at a photon detection rate of 1 MHz to avoid saturation of the SNSPDs.

The outputs of the TDI are sent back into the dilution refrigerator and directly coupled to superconducting nanowire single-photon detectors (SNSPDs, PhotonSpot®), which are mounted at the 1K stage and are coated with dielectrics to optimize detection efficiency exactly at 737 nm. To estimate the quantum efficiency (QE) of the detectors one can compare the performance of the SNSPDs to the specifications of calibrated conventional avalanche photodiodes single-photon counters (Laser Components™ COUNT-10C-FC). The estimated QEs of the SNSPDs with this method are as close to unity as one can verify. Additionally, one can measure <1% reflection from the fiber-SNSPD interface, which typically is the dominant contribution to the reduction of QE in these devices. Thus, one can assume the lower bound of the QE of the SNSPDs to be η_QE=0.99 for the rest of this section. Of course, this estimation is subject to additional systematic errors. However, the actual QE of these detectors can be a common factor (and thus drop out) in a comparison between any two physical quantum communication systems.

The total heralding efficiency η of the memory node is an useful parameter since it directly affects the performance of the BSM for quantum communication applications. In some embodiments, one can use 2 different approaches to estimate the overall heralding efficiency n. One can first measure the most dominant loss, which arises from the average reflectivity of the critically coupled nanophotonic cavity [FIG. 19 panel (b)]. While the |↑ state is highly reflecting (94.4%), the |↓ state reflects 4.1% of incident photons, leading to an average device reflectivity of η_sp=0.493.

In method (1), one can compare the input power photodiode M1 with that of photodiode MC. This estimates a lower-bound on the tapered-fiber diamond waveguide coupling efficiency of η_c=0.930±0.017. This error bar arises from uncertainty due to photodiode noise and does not include systematic photodiode calibration uncertainty. However, in some embodiments, it is noted that if the tapered fiber is replaced by a silver-coated fiber-based retroreflector, this calibration technique extracts a coupling efficiency of η_c^cal≈0.98, which is consistent with the expected reflectivity from such a retroreflector. In some embodiments, one can independently calibrate the efficiency through the 99:1 fiber beam-splitter and the TDI to be η_f=0.934. In some embodiments, one can obtain a first estimate on the overall heralding efficiency η=η_spη_cη_fη_QE=0.425±0.008.

In method (2), during the example, one can compare the reflected counts from the highly-reflecting (|↑) spin-state measured on the SNSPDs with the counts on an avalanche photodiode single photon counting module [M2 in FIG. 22 panel (b)] which has a calibrated efficiency of ≈0.7 relative to the SNSPDs. From this measurement, one can estimate an overall efficiency of fiber-diamond coupling, as well as transmission through all relevant splices and beam-splitters of η_cη_f=0.864±0.010. This error bar arises from shot noise on the single photon detectors. Overall, this gives us a consistent estimate of η=η_spη_cη_fη_QE=0.422±0.005.

In some embodiments, one can use an average value of the heralding efficiency inferred from the two calibration techniques: η=0.423±0.004. Methods (1) and (2), which each have independent systematic uncertainties associated with imperfect photodetector calibrations, are consistent to within a small residual systematic uncertainty, which is noted in the text where appropriate. It may be noted that this heralding efficiency is consistent with the scaling of spin decoherence with the number of photons at the cavity n. An example of this effect is shown in the red point in FIG. 24 panel (e).

2.3. Characterization of the Nanophotonic Quantum Memory.

FIG. 23 shows characterization of device cooperativity, according to some embodiments of the present disclosure. Panel (a) shows cavity reflection spectrum far-detuned (2302) and on resonance (2304) with SiV center. Solid line 2306 is a fit to a Lorentzian, enabling extraction of linewidth κ=21.8 GHz. Solid line 2308 is a fit to a model used to determine the single-photon Rabi frequency g=8.38±0.05 GHz and shows the onset of a normal mode splitting. Panel (b) shows measurement of SiV linewidth far detuned (Δ_c=248 GHz) from cavity resonance. Solid line 2310 is a fit to a Lorentzian, enabling extraction of natural linewidth γ=0.123 GHz.

FIG. 24 shows microwave characterization of spin-coherence properties, according to some embodiments of the present disclosure. Panel (a) shows ODMR spectrum of the qubit transition at ˜12 GHz split by coupling to a nearby ¹³C. Panel (b) shows Rabi oscillations showing π time of 30 ns. A π time of 32 ns is used for some embodiments in the present disclosure. Panel (c) shows XY8-1 dynamical decoupling signal (unnormalized) as a function of total time T, showing coherence lasting on the several hundred μs timescale. Panel (d) shows XY8-8 dynamical decoupling signal (normalized) revealing region of high fidelity at relevant value of 2τ=142 ns. Panel (e) shows fidelity of spin state after dynamical decoupling sequence with varying number of π pulses (Na), blue points. Red point (diamond) is under illumination with n=0.02.

A spectrum of the SiV-Cavity system at large detuning (248 GHz) allows one to measure the cavity linewidth κ=21.6±1.3 GHz, (FIG. 23 panel (a), blue curve) and natural SiV linewidth γ=0.123±0.010 GHz (FIG. 23 panel (a), red curve). One can find spectral diffusion of the SiV optical frequency to be much smaller than γ on minute timescales with an excitation photon flux of less than 1 MHz. Next, one can estimate the single-photon Rabi frequency, g, using the cavity reflection spectrum for zero atom-cavity detuning, shown as 2304 in FIG. 23 panel (a). For a resonant atom-cavity system probed in reflection from a single port with cavity-waveguide coupling κ_wg, the cavity reflection coefficient as a function of probe detuning Δ_c is given by

$\begin{array}{cc}r\left({\Delta }_c\right)=\frac{i{\Delta }_c+\frac{g^2}{i{\Delta }_c+\frac{\gamma }{2}}-{\kappa }_{wg}+\frac{\kappa }{2}}{i{\Delta }_c+\frac{g^2}{i{\Delta }_c+\frac{\gamma }{2}}+\frac{\kappa }{2}}.& \left(34\right)\end{array}$

By fitting |rΔ_c|² μsing known values of κ and γ, one can obtain the solid curve 2308 in FIG. 23 panel (a) which corresponds to a single-photon Rabi frequency g=8.38±0.05 GHz, yielding the estimated cooperativity

$C=\frac{4g^2}{\kappa \gamma }=105\pm 11.$

In some embodiments. one can use resonant MW pulses delivered via an on-chip coplanar waveguide (CWG) to coherently control the quantum memory. First, one can measure the spectrum of the spin-qubit transition by applying a weak, 10 μs-long microwave pulse of variable frequency, observing the optically detected magnetic resonance (ODMR) spectrum presented in FIG. 24 panel (a). It is noted that the spin-qubit transition is split by the presence of a nearby ¹³C. While coherent control techniques can be employed to utilize the ¹³C as an additional qubit, one does not control or initialize it in this example. Instead, one can drive the electron spin with strong microwave pulses at a frequency f_Q such that both ¹³C-state-specific transitions are addressed equally. This also mitigates slow spectral diffusion of the microwave transition of 100 kHz.

After fixing the MW frequency at f_Q one can vary the length of this drive pulse [τ_R in FIG. 24 panel (b)] and observe full-contrast Rabi oscillations. One can choose a π time of 32 ns in some embodiments in present disclosure, which is an example compromise of two factors: (1) it is sufficiently fast such that one can temporally multiplex between 2 and 4 time-bin qubits around each microwave π pulse and (2) it is sufficiently weak to minimize heating related effects from high microwave currents in resistive gold CWG.

In some embodiments, with known π time one can measure the coherence time of the SiV spin qubit under an XY8-1 dynamical decoupling sequence to exceed 200 μs [FIG. 24 panel (c)]. In some embodiments, one can use decoupling sequences with more π pulses. As an example, FIG. 24 panel (d) shows the population in the |↑ state after XY8-8 decoupling sequence (total N_π=64 π pulses) as a function of τ, half of the inter-pulse spacing. For some BSM applications, this inter-pulse spacing, 2τ, is fixed and is matched to the time-bin interval δt. While at some times (e.g. τ=64.5 ns) there is a loss of coherence due to entanglement with the nearby ¹³C, at 2τ=142 ns it is decoupled from this ¹³C and can maintain a high degree of spin coherence. Thus, one can choose the time-bin spacing to be 142 ns. The spin coherence at 2τ=142 ns is plotted as a function N_π in FIG. 24 panel (d), and decreases for large N_π, primarily due to heating related effects.

2.4. Description of Asynchronous Bell State Measurement

Due to the critical coupling of the nanocavity, the memory node reflects photons when the SiV spin is in the state |↑. The resulting correlations between the spin and the reflected photons can still be used to realize a BSM between two asynchronously arriving photonic time-bin qubits using an adaptation of the well-known proposal of Duan and Kimble for entangling a pair of photons incident on an atom-cavity system. As a result of the critical coupling, one may have access to two of the four Bell states at any time, with the inaccessible Bell states corresponding to photons being transmitted through the cavity (and thus lost from the detection path). Depending on whether there was an even or odd number of π-pulses on the spin between the arrival of the two heralded photons, one can distinguish either the {|Φ_±} or {|Ψ_±} states (defined below). For the sake of simplicity, one can first describe the BSM for the case when the early time bin of Alice's and Bob's qubits both arrive after an even number of microwave π pulses after its initialization. Thereafter one can generalize this result and describe the practical consequences for the MDI-QKD protocol.

The sequence begins with a π/ 2 microwave pulse, preparing the spin in the state |ψ_i=(|↑+|↓)/√{square root over (2)}. In the absence of a photon at the device, the subsequent microwave π-pulses, which follow an XY8-N type pattern, decouple the spin from the environment and at the end of the sequence should preserve the spin state |ψ_i. However, reflection of Alice's photonic qubit |A=(|e+e^iϕ1|l)/√{square root over (2)} from the device results in the entangled spin-photon state |ψ_A=(|↑e+e^iϕ1|↓l))/√{square root over (2)}. Without wish to be bound by theory, the full system is in the state

$\begin{array}{cc}{\psi }_A\rangle \frac{+x\rangle \left(\uparrow \rangle +e^{i_{{\varphi }_1}}\downarrow \rangle \right)+-x\rangle \left(\uparrow \rangle -e^{i_{{\varphi }_1}}\downarrow \rangle \right)}{2}.& \left(35\right)\end{array}$

Regardless of the input photon state, there is equal probability to measure the reflected photon to be |±x. Thus, measuring the photon in X basis (through the TDI) does not reveal the initial photon state. After this measurement, the initial state of the photon IA) is teleported onto the spin: |ψ_m1=(|↑+m₁^iϕ1|↓)√{square root over (2)}, where m₁=±1 denotes the detection outcome of the TDI. The quantum state of Alice's photon is now stored in the spin state, which is preserved by the dynamical decoupling sequence.

Reflection of the second photon |B=(|e+e^iϕ2|l)√{square root over (2)} from Bob results in the spin-photon state |ψ_m1,B=(|↑e+m₁e^iϕ1+ϕ2|↓l)√{square root over (2)}. This state now has a phase that depends on the initial states of both photons, enabling the photon-photon BSM measurements described below. Without wish to be bound by theory, rewriting Bob's reflected photon in the X basis, the full system is in the state

|ψ_m1,B={|+x(|↑+m₁e^i(ϕ1+ϕ2)|↓)+|−x(|↑−m₁e^i(ϕ1+ϕ2)|↓)}2.  (36)

The second measurement result m₂ once again contains no information about the initial state |B. In yet heralds the final spin state |m_m1,m2)=(|↑+m₁m₂e^i(ϕ1+ϕ2)|↓) as described throughout the present disclosure. When this state lies along the X axis of the Bloch sphere (ϕ₁+ϕ₂={0, ϕ}), the final result of the X basis measurement on the spin state m₃ has a deterministic outcome, dictated by all values of the parameters {ϕ₁, ϕ₂} (known to Alice and Bob) and {m₁, m₂} (which are known to Charlie, but are completely random). Conversely, all information available to Charlie {m₁, m₂, m₃} contains information on the correlation between the photonic qubits, not on their individual states. The resulting truth table for different input states is given in TABLE 3. For all input states, there is equal probability of measuring ±1 for each individual measurement m_i. However, the overall parity of the three measurements {m₁m₂m₃} depends on whether or not the input photons were the same, or opposite, for inputs |A, |Bϵ|±x or |±y.

One can now address the fact that the BSM distinguishes either between {|Φ_±} or {|Ψ_±} if there was an even or odd number of microwave π pulses between incoming photons respectively. This effect arises because each π pulse in the dynamical decoupling sequence toggles an effective frame change: Y↔−Y. Without wish to be bound by theory, the impact on this frame change on the BSM can be seen by writing the pairs of Bell states (|Φ_±=(|ee±|ll)/√{square root over (2)} and |Ψ_±=(|el±|le)/√{square root over (2)}) in the X and Y bases, where one can have

|Φ_±^(X)=(|+x|±x+|∓x|−x)/√{square root over (2)}  (37)

|Φ_±^(Y)=(|+y|±y+|±y|−y)/√{square root over (2)}  (38)

|Ψ_±^(X)=(|+x|±x+|∓x|−x)/√{square root over (2)}  (39)

|Ψ_±^(Y)=i(|+y|±y−|∓y|−y)/√{square root over (2)}  (40)

For X basis inputs, as seen by Eqns. 37 and 39, switching between {|Φ_±} and {|Ψ_±} measurements does not affect the inferred correlation between input photons. For Y basis inputs however, this does result in an effective bit flip in the correlation outcome (see Eqn. 38 and 40). In practice, Alice and Bob can keep track of each Y photon sent and apply a bit flip accordingly, as long as they have the appropriate timing information about MW pulses applied by Charlie. If Charlie does not give them the appropriate information, this will result in an increased QBER which can be detected.

As a final remark, this scheme also works for pairs of photons that are not both in the X or Y basis but still satisfy the condition ϕ₁+ϕ₂=0. For example, |a and |b from FIG. 20 panel (b) satisfy this condition. In this case, adequate correlations can still be inferred about the input photons, although they were sent in different bases.

2.4.1. Test of Bell-CHSH Inequality

In order to perform a test of the Bell-CHSH inequality, one can send input photons equally distributed from all states {|±x, |±y, |±a, |±b} [ FIG. 20 panel (b)]. One can select for cases where two heralding events arise from input photons {A, B}=±1 that are either 45° or 135° apart from one another. Conditioned on the parity outcome of the BSM (±1), the Bell-CHSH inequality bounds the correlations between input photons as

S _± =| A·B − A≠B − A·B − A·B |≤2,  (41)

where the subscripts denote the bases the photons were sent in. The values of each individual term in Eqn. 41, denoted as “input correlations,” are plotted in FIG. 3d for positive and negative parity outcomes.

FIG. 25 shows measurements on a single time-bin qubit in Z and X bases, according to some embodiments of the present disclosure. Panel (a) shows an example of optical pulses sent for example in the example described in FIG. 19 panel (d). Panel (b) shows a time trace of detected photons on+detector when pulses shown in panel (a) are sent directly into the TDI. The first and last peaks correspond to late and early photons taking the long and short paths of the TDI, which enable measurements in the Z basis {|e, |l}. The central bin corresponds to the late and early components overlapping and interfering constructively to come out of the +port, equivalent to a measurement of the time bin qubit in the |+x state. A detection event in this same timing window on the—detector (not shown) can constitute a |−x measurement.

TABLE 3
Truth table of asynchronous BSM protocol
	Alice	Bob	Parity	Bell state
	|+ x	|+ x	+1	|Φ+
	|+ x	|− x	−1	|Φ−
	|− x	|+ x	−1	|Φ−
	|− x	|− x	+1	|Φ+
	|+ y	|+ y	−1	|Φ−
	|+ y	|− y	+1	|Φ+
	|− y	|+ y	+1	|Φ+
	|− x	|− y	−1	|Φ−

TABLE 3 shows the parity (and BSM outcome) for each set of valid input states from Alice and Bob. In the case of Y basis inputs, Alice and Bob adjust the sign of their input state depending on whether it was commensurate with an even or odd numbered free-precession interval, based on timing information provided by Charlie.

2.5. Analysis of Quantum Communication Applications

2.5.1. Estimation of QBER

In order to achieve the lowest QBER, one can routinely monitor the status trigger of the pre-selection routine and adjust the TDI. Additionally, one can keep track of the timing when the TDI piezo voltage rails. In some embodiments, this guarantees that the SiV is resonant with the photonic qubits and that the TDI performs high-fidelity measurements in X basis. This is implemented in software with a response time of 100 ms.

For each example, one can estimate the QBER averaged over all relevant basis combinations. This is equivalent to the QBER when the random bit string has all bases occurring with the same probability, (an unbiased and independent basis choice by Alice and Bob). One may first note that the QBER for positive and negative parity announcements are not independent. One can illustrate this for the example, that Alice and Bob send photons in the X basis. One can denote the probability P that Alice sent qubit |ψ, Bob sent qubit |ξ and the outcome of Charlie's parity measurement is m_C, conditioned on the detection of a coincidence, as P(ψ_A ∩ξ_B ∩m_C). Without wish to be bound by theory, one can find for balanced inputs P(+X_A ∩−X_B)=P(−X_A ∩+X_B) that P(E_XX|+_C)=P(E_XX|−_C) with E_XX denoting the occurrence of a bit error in the sifted key of Alice and Bob. One thus find for the posterior probability L for the average QBER for XX coincidences

L(P(E_XX))=L(P(−_C|+X_A∩+X_B))*L(P(+_C|+X_A∩−X_B))*L(P(+_C|−X_A∩+X_B))*L(P(−_C|−X_A∩+X_B)).  (42)

Note that this expression is independent of the actual distribution of P(ψ_A ∩ξ_B). Here, the posterior probability L(P(+_C|+X_A ∩−X_B)) is based on the a binomial likelihood function

$P\left(N_{m_C\cap {\psi }_A\cap {\xi }_B}|N_{{\psi }_A\cap {\xi }_B,}L\right),$

where N_C denotes the number of occurrences with condition C. Finally the posterior probability of the unbiased QBER is L(P(E))=L(P(E_XX))*(P(E_YY)). All values presented in the text and figures are maximum likelihood values with bounds given by the confidence interval of ±34.1% integrated posterior probability. Confidence levels towards a specific bound (for example, unconditional security) are given by the integrated posterior probability up to the bound.

FIG. 26 shows performance of memory-device versus of channel loss, according to some embodiments of the present disclosure. Panel (a) shows enhancement of memory-based approach compared to direct transmission approach, keeping N=124 fixed and varying n in order to vary the effective channel transmission probability p_A→B. At high p_A→B (larger n), r_s approaches 0 due to increased QBER arising from undetected scattering of a third photon. Panel (b) shows: (left) a plot of QBER for same sweep of n shown in panel (a); and (right) a plot of QBER while sweeping N in order to vary loss. These points correspond to the same data shown in FIG. 21. At lower p_A→B (larger N), microwave-induced heating-related dephasing leads to increased QBER.

To get the ratio of the distilled secret key rate with respect to the sifted key rate by (ideal) error correction and privacy amplification, one can use the bounds given by difference in information by Alice and Bob with respect to a potential eavesdropper who performs individual attacks [6]: r_s=I(A, B)−I(A/B, E)^max. One can use the full posterior probability distribution of QBER (which accounts for statistical and systematic uncertainty in the estimate) to compute the error bar on r_s, and correspondingly, the error bars on the extracted secret key rates plotted in FIG. 21.

2.5.2. Example Parameters for Asynchronous Bell State Measurements

One can minimize the extracted QBER for the asynchronous BSM to optimize the performance of the memory node. The first major factor contributing to QBER is the scattering of a third photon that is not detected, due to the finite heralding efficiency η=0.423±0.04. This is shown in FIG. 19 panel (f), where the fidelity of the spin-photon entangled state diminishes for n≥0.02. At the same time, one can work at the maximum possible n in order to maximize the data rate to get enough statistics to extract QBER (and in the quantum communication setting, efficiently generate a key).

To increase the key generation rate per channel use, one can also fit many photonic qubits within each initialization of the memory. In practice, there are 2 physical constraints: (1) the bandwidth of the SiV-photon interface and (2) the coherence time of the memory. One can find that one can satisfy (1) at a bandwidth of roughly 50 MHz with no measurable infidelity. For shorter optical pulses (<10 ns), the spin-photon gate fidelity is reduced. In principle, the SiV-photon bandwidth can be increased by reducing the atom-cavity detuning (here ˜60 GHz) at the expense of having to operate at higher magnetic fields where microwave qubit manipulation is not as convenient.

Even with just an XY8-1 decoupling sequence (number of π pulses N_π=8, the coherence time of the SiV is longer than 200 μs [FIG. 24 panel (c)] and can be prolonged to the millisecond range with longer pulse sequences. Unfortunately, to satisfy the bandwidth criteria (1) and to drive both hyperfine transitions [FIG. 24 panel (a)], one can use short (32 ns long π pulses), which cause additional decoherence from ohmic heating already at N_π=64 [FIG. 24 panel (e)]. Due to this one can limit the pulse sequences to a maximum N_π=128, and use up to ≈20 μs of the memory time. One solution is to switch to superconducting microwave delivery. Alternatively, one can use a larger value of τ to allow the device to cool down in between subsequent pulses at the expense of having to stabilize a TDI of larger δt. Working at larger δt also enables temporal multiplexing by fitting multiple time-bin qubits per free-precession interval. In fact, with 2τ=142 ns, even given constraint (1) and the finite π time, one can already fit up to 4 optical pulses per free-precession window, enabling a total number of photonic qubits of up to N=504 for N_π=128.

In benchmarking the asynchronous BSM for quantum communication, one can optimize the parameters n and N to maximize the enhancement over the direct transmission approach, which is a combination of both increasing N and reducing the QBER, since a large QBER results in a small secret key fraction r_s. As described throughout embodiments of the present disclosure, the effective loss can be associated with n, which is the average number of photons per photonic qubit arriving at the device, and is given straightforwardly by n=n/N. The most straightforward way to sweep the loss is to keep the example sequence the same (fixed N) and vary the overall power, which changes n. The results of such a sweep are shown in FIG. 26 panels (a) and (b). For larger n(corresponding to lower effective channel losses), the errors associated with scattering an additional photon reduce the performance of the memory device.

TABLE 4
Quantum-memory-based advantage
	per channel	per channel	per channel	per channel
	occupancy	occupancy	use	use
X:Y basis bias	50:50	99:1	50:50	99:1
Secure key	1.19−0.14+0.14	2.33−0.28+0.28	2.37−0.28+0.28	4.66−0.55+0.56
rate R[10−7]
R/Rmax(X:Y)	2.06−0.25+0.25	2.06−0.25+0.25	4.13−0.49+0.50	4.13−0.49+0.50
R/(1.44pA→B)	0.71−0.08+0.09	1.40−0.17+0.17	1.43−0.17+0.17	2.80−0.33+0.34
1—confidence		1.1−0.3+0.4 ×	8−2+3 ×	1.3−0.3+0.5 ×
level		10−2	10−3	10−7

TABLE 4 shows secret key rates with the asynchronous BSM device and comparison to ideal direct communication implementations, based on the performance of the network node for N=124 and n˜0.02. Distillable key rates for E=0.110±0.004 for unbiased and biased basis choice are expressed in a per-channel-occupancy and per-channel-use normalization. Enhancement is calculated versus the linear optics MDI-QKD limit (R_max(50:50)=p_A→B/ 2 for unbiased bases, R_max(99:1)=0.98p_A→B with biased bases) and versus the fundamental repeaterless channel capacity (1.44p_A→B). Confidence levels for surpassing the latter bound are given in the final row.

Due to these considerations, one can work at roughly n≤0.02 for some examples shown in FIGS. 20 and 21, below which the performance does not improve significantly. At this value, one can obtain BSM successes at a rate of roughly 0.1 Hz. By fixing n and increasing N, one can maintain a tolerable BSM success rate while increasing the effective channel loss. Eventually, as demonstrated in FIG. 26 panel (c) and in the high-loss data point in FIG. 21, effects associated with microwave heating result in errors that again diminish the performance of the memory node for large N. In some embodiments, one can conclude that the performance of the node occurs for n·0.02 and N≈124, corresponding to an effective channel loss of 69 dB between Alice and Bob, which is equivalent to roughly 350 km of telecommunications fiber.

In some embodiments, one can also find that the QBER and thus the performance of the communication link is limited by imperfect preparation of photonic qubits. Photonic qubits are defined by sending arbitrary phase patterns generated by the optical AWG to a phase modulator. For an example of such a pattern, see the blue curve in FIG. 20 panel (a). One can use an imperfect pulse amplifier with finite bandwidth (0.025-700 MHz), and find that the DC component of these waveforms can result in error in photonic qubit preparation on the few % level. By using a tailored waveform of phases with smaller (or vanishing) DC component, one can reduce these errors. One can perform such an example application during the test of the Bell-CHSH inequality. One can find that by evaluating BSM correlations from |±a and |+b inputs during this measurement, one can estimate a QBER of 0.097±0.006.

Finally, one can obtain the effective clock-rate of the communication link by measuring the total number of photonic qubits sent over the course of an entire run time. In practice, one can record the number of channel uses, determined by the number of sync triggers recorded [see FIG. 22 panel (a)] as well as the number of qubits per sync trigger (N). One can then divide this number by the total run time from start to finish (˜1-2 days for example runs), including all experimental downtime used to stabilize the interferometer, readout and initialize the SiV, and compensate for spectral diffusion and ionization. For N=248, one can extract a clock rate of 1.2 MHz. As the secret key rate in this configuration exceeds the conventional limit of p/2 by a factor of 3.8±1.1, it is competitive with a standard MDI-QKD system operating at 4.5_−1.2^+1.3 MHz clock rate.

2.5.3. Performance of Memory-Assisted MDI-QKD

A single optical link can provide many channels, for example, by making use of different frequency, polarization, or temporal modes. To account for this, when comparing different systems, data rates can be defined on a per-channel-use basis. In an MDI-QKD setting, full usage of the communication channel between Alice and Bob means that both links from Alice and Bob to Charlie are in use simultaneously. For an example asynchronous sequential measurement, typically half of the channel is used at a time, for example from Alice to Charlie or Bob to Charlie. The other half can in principle be used for a different task when not in use. For example, the unused part of the channel could be routed to a secondary asynchronous BSM device. In the example, one can additionally define as a second normalization the rate per channel “occupancy”, which accounts for the fact that half the channel is used at any given time. The rate per channel occupancy is therefore half the rate per full channel use. For comparison, one can typically operate at 1.2% channel use and 2.4% channel occupancy.

To characterize the performance of the asynchronous Bell state measurement device, one can operate it in the regime determined above (N=124, n≤0.02. Without wish to be bound by theory, it is noted that the enhancement in the sifted key rate over direct transmission MDI-QKD is given by

$\begin{array}{cc}\frac{R}{R_{\max }}={\eta }^2\frac{\left(N_{\pi }-1\right)\left(N_{\pi }-2\right)N_{sub}}{2N_{\pi }}& \left(43\right)\end{array}$

and is independent of n for a fixed number of microwave pulses N_π and optical pulses per microwave pulse N_sub and thus fixed N=N_πN_sub. For low n, three photon events become negligible and therefore QBER saturates, such that the enhancement in the secret key rate saturates as well [FIG. 26 panel (a)]. One can therefore combine all data sets with fixed N=124 below n≤0.02 to characterize the average QBER of 0.116±0.002 [FIG. 20 panel (c)]. The key rates cited in in some examples in the present disclosure relate to a data set in this series (n≈0.02), with a QBER of 0.110±0.004. A summary of key rates calculated on a per-channel use and per-channel occupancy basis, as well as comparisons of performance to ideal MDI-QKD and repeaterless bounds are given in TABLE 4.

Furthermore, one can extrapolate the performance of the memory node to include biased input bases from Alice and Bob. This technique enables a reduction of channel uses where Alice and Bob send photons in different bases, but is still compatible with secure key distribution, allowing for enhanced secret key rates by at most a factor of 2. The extrapolated performance of the node for a bias of 99:1 is also displayed in TABLE 4, as well as comparisons to the relevant bounds. It is noted that basis biasing does not affect the performance when comparing to the equivalent MDI-QKD example, which is limited by p_A→B/2 in the unbiased case and p_A→B in the biased case. However, using biased input bases does make the performance of the memory-assisted approach more competitive with the fixed repeaterless bound of 1.44p_A→B.

While the invention has been particularly shown and described with reference to specific preferred embodiments, it should be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims

1. A method of making a quantum network node comprising: calculating a plurality of scoring function F values for an array of at least two photonic crystal cavity unit cells, each photonic crystal cavity unit cell having a lattice constant a and a hole having a length Hx and a width Hy, wherein the scoring function comprises: F=min(Q,Qcutoff)/(Qcutoff×Vmode)

2. The method of claim 1, wherein the scoring function value criteria comprises one or more of a maximum scoring function value of the plurality of scoring function F values, a threshold value exceeded by at least one of the plurality of scoring function F values, and a maximum scoring function value of the plurality of scoring function F values after a predetermined number of iterations calculating scoring function F values.

3. The method of claim 1, further comprising: tapering an input end of the waveguide region;connecting the tapered input end of the waveguide region to a tapered optical fiber to optically couple the optical fiber to the array of at least two photonic crystal cavity unit cells.

4. The method of claim 3, further comprising coupling the tapered optical fiber to at least one probing light source and to at least one single photon detector.

5. The method of claim 1, wherein the substrate comprises a diamond substrate.

6. The method of claim 5, wherein the implanted ion is a silicon ion and wherein the at least one quantum defect is a silicon-vacancy color center.

7. The method of claim 1, wherein Qcutoff is not greater than 5×105.

8. The method of claim 1, further comprising: forming a mask, after the forming the array of at least two photonic crystal cavity unit cells, on the array of at least two photonic crystal cavity unit cells;etching, with an ion beam, the substrate, at two or more angles, to form a tapered column between the hole of a first photonic crystal cavity unit cell and the second photonic crystal cavity unit cell.

9. The method of claim 1, wherein the implanting the at least one ion further comprises: forming a mask on the array of at least two photonic crystal cavity unit cells;forming at least one aperture in the mask at a location between the hole of a first photonic crystal cavity unit cell and the second photonic crystal cavity unit cell; andimplanting the at least one ion through the at least one aperture in the mask, wherein ions are not implanted through the mask.

10. The method of claim 1, further comprising mounting the substrate in a refrigeration unit, wherein the refrigeration unit is configured to cool the substrate to less than 100 mK such that the spin coherence time T2 of the at least one quantum defect is extended.

11. A quantum network device comprising: a substrate;an array of at least two photonic crystal cavity unit cells on the substrate, wherein each photonic crystal cavity unit cell has a lattice constant a and a hole having a length Hx and a width Hy, wherein a value of a, a value of Hx, and a value of Hy are selected so that a scoring function F value meets a scoring function value criteria, and wherein the scoring function comprises: F=min(Q,Qcutoff)/(Qcutoff×Vmode)

12. The device of claim 11, wherein the scoring function value criteria comprises one or more of a maximum scoring function value of the plurality of scoring function F values, a threshold value exceeded by at least one of the plurality of scoring function F values, and a maximum scoring function value of the plurality of scoring function F values after a predetermined number of iterations calculating scoring function F values.

13. The device of claim 11, wherein the coplanar microwave waveguide comprises a tapered input end, and wherein the tapered input end is connected to a tapered optical fiber to optically couple the optical fiber to the array of at least two photonic crystal cavity unit cells.

14. The device of claim 11, further comprising at least one probing light source and at least one single photon detector.

15. The device of claim 11, wherein the substrate comprises a diamond substrate.

16. The device of claim 11, wherein the quantum defect is a silicon-vacancy color center.

17. The device of claim 11, further comprising a refrigeration unit, wherein the refrigeration unit is configured to cool the substrate to less than 100 mK such that a spin coherence time T2 of the at least one quantum defect is extended.

18. A method of operating the quantum network device of claim 14, the method comprising: receiving, with the at least one single photon detector, at least two photons; andin response to the receipt of two photons, measuring the state of the quantum defect using the probing light source.

19. The method of claim 18, wherein the receiving, with the at least one single photon detector, at least two photons and the measuring the state of the quantum defect with the laser comprises a Bell-state measurement.

20. A method of encoding of quantum information using the quantum network device of claim 11, comprising: for n time-bin qubits comprising n+1 optical pulses, applying phase control with a phase modulator to each optical pulse, wherein each time-bin qubit stores quantum information in a relative amplitude and phase between a pair of neighboring optical pulses among the n+1 optical pulses;guiding the n+1 optical pulses to the at least one quantum defect;alternating, with each pulse, coherent microwave control of the quantum defect; andinterfering, with a time-delay interferometer, each pulse with a previous optical pulse, wherein the time-delay interferometer delays the previous optical pulse by the time between the pulse and the previous optical pulse.