Controlled-Phase Quantum Logic Gate

Abstract

A method and circuit QED implementation of a control-phase quantum logic gate UCP(θ)=diag[1, 1, 1, eiθ]. Two qubits Qi, two resonators Ra, Rb and a modulator. Q1 and Q2, each has a frequency ωqi and characterized by σzi. Ra is associated with Q1 and defined by a quantum non-demolition (QND) longitudinal coupling g1zσ1z(â†+â). Rb is integrated into Ra, the QND second longitudinal coupling is defined by Ra as g2zδ2z({circumflex over (b)}†+{circumflex over (b)}) or, when Rb is integrated into Ra, the QND second longitudinal coupling is defined by Ra as g2zσ2z(â†+â) The modulator periodically modulates, at a frequency ωm during a time t, the longitudinal coupling strengths g1z and g2z with respective signals of respective amplitudes {tilde over (g)}1 and {tilde over (g)}2. Selecting a defined value for each of t, g1z and g2z determines θ to specify a quantum logical operation performed by the gate. Q1 and Q2 are decoupled when either one of g1z and g2z is to set to 0.

Description

STATEMENT REGARDING PRIOR DISCLOSURES BY THE INVENTOR OR A JOINT INVENTOR

To the Applicant's knowledge, no public disclosure has been made, by the inventor or joint inventor or by another who obtained the subject matter publicly disclosed directly or indirectly from the inventor or a joint inventor, more than one (1) year before the effective filing date of an invention claimed herein.

TECHNICAL FIELD

The present invention relates to quantum computing and, more particularly, to information unit representation and/or manipulation in quantum computing.

BACKGROUND

Quantum computing is presented as the next computational revolution. Yet, before we get there, different problems need to be resolved. For instance, one needs to reliably store information in the form of a quantum bit (qubit), maintain the information reliably in the qubit and read the stored information reliably and repetitively (i.e., non-destructive readout). Another of the challenges of quantum computing is related to logic treatment of more than one qubit without forcing a defined state (i.e., providing one or more logical gates from different qubits in potentially overlapping states).

The present invention addresses at least partly the need for logic treatment of more than one qubit without forcing a defined state.

SUMMARY

This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

In accordance with a first set of embodiments, a first aspect of the present invention is directed to a circuit quantum electrodynamics (circuit QED) implementation of a control-phase quantum logic gate U_CP(θ)=diag [1,1,1,e^iθ]. The circuit QED implementation comprises two qubits Q_i, two resonators R_a, R_b and a modulator. A first qubit Q₁ and a second qubit Q₂, each has a frequency ω_qi and being characterized by {circumflex over (σ)}_zi. The resonator R_a is associated with the qubit Q₁ and defined by a resonator frequency ω_ra, a resonator electromagnetic field characterized by â^† and â, a longitudinal coupling strength g_1z, with the qubit Q₁, a quantum non-demolition (QND) first longitudinal coupling g_1z{circumflex over (σ)}_1z(â^†+â). The second resonator R_b, when independent from R_a, is associated with the qubit Q₂ and is defined by a longitudinal resonator-resonator coupling g_ab with R_a and R_b is further defined by a second resonator frequency ω_rb, a second resonator electromagnetic field characterized by {circumflex over (b)}^† and {circumflex over (b)}, a second longitudinal coupling strength g_2z with the qubit Q₂, a QND second first longitudinal coupling g_2z{circumflex over (σ)}_2z({circumflex over (b)}^†+{circumflex over (b)}). When the second optional resonator R_b is not independent from R_a and integrated into R_a, then R_a is associated with the qubit Q₂, the longitudinal resonator-resonator coupling g_ab=1, the second resonator electromagnetic field is characterized by â^† and â where {circumflex over (b)}^†=â^† and {circumflex over (b)}=â, the second resonator frequency ω_rb=ω_ra, the second longitudinal coupling strength g_2z is between the qubit Q2 and R_a, the QND second first longitudinal coupling g_2z{circumflex over (σ)}_2z({circumflex over (b)}^†+{circumflex over (b)}) is defined by R_a as g_2z{circumflex over (σ)}_2z(â^†+â). a). The modulator periodically modulates, at a frequency ω_m during a time t, the longitudinal coupling strengths g_1z and g_2z with respective signals of respective amplitudes {tilde over (g)}₁ and {tilde over (g)}₂. Selecting a defined value for each of t, g_1z and g_2z determines θ to specify a quantum logical operation performed by the control-phase quantum logic gate. The qubit Q₁ and the qubit Q2 are decoupled when either one of the defined value of g_1z and the defined value of g_2z is to set to 0.

The circuit QED implementation may further comprise a transmitter for selectively providing a modulator activation signal to the modulator for activating the modulator for the duration t.

The circuit QED implementation may also further comprise a signal injector providing a squeezed input to diminish a which-qubit-state information. The squeezed input may optionally be a single-mode squeezed input or a two-mode squeezed input. The signal injector may rely on broadband squeezed centered at ω_rb and/or ω_ra.

The qubit Q₁ and the qubit Q₂ may optionally be transmons each comprising two Josephson junctions with respectively substantially equivalent capacitive values and the modulator comprises an inductor-capacitor (LC) oscillator, the longitudinal coupling resulting from mutual inductance between the oscillator and the transmons, the oscillator varying a flux Φ₁ in the qubit Q₁ and a flux Φ₂ in the qubit Q₂. A 3-Wave mixing Josephson dipole element may optionally be used to couple the qubit Q₁ and the resonator R_a.

In accordance with the firsy set of embodiments, a second aspect of the present invention is directed to a method for specifying a quantum logical operation performed by a control-phase quantum logic gate U_CP(θ)=diag[1, 1, 1, e^iθ]. The circuit QED implementation comprises (I) two qubit Q_i, where i=1 corresponds to a first qubit Q₁ and i=2 corresponds to a second qubit Q₂, each having a frequency ω_qi and being characterized by {circumflex over (σ)}_zi; (II) a first resonator R_a, associated with the qubit Q₁, defined by a first resonator frequency ω_ra, a first resonator electromagnetic field characterized by â^† and â, a first longitudinal coupling strength g_1z with the qubit Q₁ and a quantum non-demolition (QND) first longitudinal coupling g_1z{circumflex over (σ)}_1z (â^†+â); (III) a second resonator R_b, such that, when the second resonator R_b is independent from R_a, R_b is associated with the qubit Q₂, a longitudinal resonator-resonator coupling g_ab is defined and R_b is further defined by: a second resonator frequency ω_rb, a second resonator electromagnetic field characterized by {circumflex over (b)}^† and {circumflex over (b)}, a second longitudinal coupling strength g_2z with the qubit Q₂, a QND second first longitudinal coupling g_2z{circumflex over (σ)}_2z({circumflex over (b)}^†+{circumflex over (b)}) and (IV), when the second optional resonator R_b is not independent from R_a and integrated into R_a, R_a is associated with the qubit Q₂, the longitudinal resonator-resonator coupling g_ab=1, the second resonator electromagnetic field is characterized by â^† and â where {circumflex over (b)}^†=â^† and {circumflex over (b)}=â, the second resonator frequency ω_rb=ω_ra, the second longitudinal coupling strength g_2z is between the qubit Q₂ and R_a, the QND second first longitudinal coupling g_2z{circumflex over (σ)}_2z({circumflex over (b)}^†+{circumflex over (b)}) is defined by R_a as g_2z{circumflex over (σ)}_2z(â^†+â). The method comprises periodically modulating, at a frequency ω_m during a time t, the longitudinal coupling strengths g_1z and g_2z with respective signals of respective amplitudes {tilde over (g)}₁ and {tilde over (g)}₂, selecting a defined value for each of t, g_1z and g_2z thereby fixing θ to specify the quantum logical operation performed by the control-phase quantum logic gate and setting at least one of the defined value of g_1z and the defined value of g_2z is to 0 to decouple the qubit Q₁ from the qubit Q₂.

Selecting the defined value for each oft, g_1z and g_2z may optionally comprise a selectively providing a modulator activation signal to the modulator for activating the modulator for the duration t.

The method may also further comprise providing a squeezed input to diminish a which-qubit-state information. The squeezed input may be a single-mode squeezed input, a two-mode squeezed input or may rely on broadband squeezed centered at ω_rb and/or ω_ra.

The qubit Q₁ and the qubit Q₂ may optionally be transmons each comprising two Josephson junctions with respectively substantially equivalent capacitive values and modulating may optionally be performed by an inductor-capacitor (LC) oscillator, the longitudinal coupling resulting from mutual inductance between the oscillator and the transmons, the oscillator varying a flux Φ₁ in the qubit Q₁ and a flux Φ₂ in the qubit Q₂. A3-Wave mixing Josephson dipole element may optionally be used to couple the qubit Q₁ and the resonator R_a.

In accordance with a second set of embodiments, a first aspect of the present invention is directed to a circuit quantum electrodynamics (circuit QED) implementation of a quantum information unit (qubit) memory having a qubit frequency ω_a and holding a value {circumflex over (σ)}_z. The circuit QED implementation comprises a resonator, a modulator and a homodyne detector. The resonator is defined by a resonator damping rate κ, a resonator frequency ω_r, a resonator electromagnetic field characterized by â^† and â, a longitudinal coupling strength g_z, an output â_out and a quantum non-demolition (QND) longitudinal coupling g_z{circumflex over (σ)}_z(â^†+â). The modulator periodically modulates the longitudinal coupling strength g_z with a signal of amplitude ĝ_z greater than or equal to the resonator damping rate κ and of frequency ω_m with ω_m±κ resonant with ω_r±a correction factor. The correction factor is smaller than |ω₄/10| and the longitudinal coupling strength g_z varies over time (t) in accordance with g_z(t)=g_z+{tilde over (g)}_z cos (ω_mt) with g_z representing an average value of g_z. The homodyne detector for measuring the value {circumflex over (σ)}_z of the qubit memory from a reading of the output â_out.

Optionally, the correction factor may be between 0 and |ω_r/100|. The homodyne detector may measure the value {circumflex over (σ)}_z of the qubit memory from a phase reading of the output â_out. The signal amplitude {tilde over (g)}_z may be at least three (3) times greater than the resonator damping rate κ or at least ten (10) times greater than the resonator damping rate _N.

The circuit QED implementation may optionally further comprise a signal injector providing a single-mode squeezed input on the resonator such that noise on the phase reading from the output â_out is reduced while noise is left to augment on one or more interrelated characteristics of the output â_out. The average value of g_z, g_z may be 0 and the single-mode squeezed input may be QND.

The qubit memory may be a transmon comprising two Josephson junctions with substantially equivalent capacitive values and the longitudinal modulator may an inductor-capacitor (LC) oscillator with a phase drop δ across a coupling inductance placed between the two Josephson junctions. The longitudinal coupling results from mutual inductance between the oscillator and the transmon and the oscillator may vary a flux Φ_x in the transmon. The transmon may have a flux sweet spot at integer values of a magnetic flux quantum Φ₀, Josephson energy asymmetry of the transmon may be below 0.02 and Φ_x may vary by ±0.05Φ₀ around Φ_x=0.9. A 3-Wave mixing Josephson dipole element may optionally be used to couple the qubit and the resonator. The resonator may further be detuned from the qubit frequency ω_a by |Δ|≥{tilde over (g)}_z. The oscillator inductance may, for instance, be provided by an array of Josephson junctions or by one or more Superconducting Quantum Interference Device (SQUID).

In accordance with the second set of embodiments, a second aspect of the present invention is directed to a method for reading a value {circumflex over (σ)}_z stored in a quantum information unit (qubit) memory having a qubit frequency ω_a, with a resonator defined by a resonator damping rate κ, a resonator frequency ω_r, a resonator electromagnetic field characterized by â^† and â, a longitudinal coupling strength g_z, an output â_out and a quantum non-demolition (QND) longitudinal coupling g_z{circumflex over (σ)}_z(â^†+â). The method comprises, at a modulator, periodically modulating the longitudinal coupling strength g_z with a signal of amplitude {tilde over (g)}_z greater than or equal to the resonator damping rate κ and of frequency ω_m with ω_m±κ resonant with ω_r±a correction factor. The correction factor is smaller than |ω_r/10| and the longitudinal coupling strength g_z varies over time (t) in accordance with g_z(t)=g_z+{tilde over (g)}_z cos (ω_mt) with g_z representing an average value of g_z. The method also comprises, at a homodyne detector, measuring the value {circumflex over (σ)}_z of the qubit memory from a reading of the output â_out.

Optionally, the signal amplitude {tilde over (g)}_z may be at least three (3) times greater than the resonator damping rate κ or at least ten (10) times greater than the resonator damping rate κ.

The method may further comprise, from a signal injector, providing a single-mode squeezed input on the resonator such that noise on the phase reading from the output â_out is reduced while noise is left to augment on one or more interrelated characteristics of the output â_out. The average value of g_z, g_z may be set to 0 and the single-mode squeezed input may be QND.

Optionally, the qubit memory may be a transmon comprising two Josephson junctions with substantially equivalent capacitive values and the longitudinal modulator comprises an inductor-capacitor (LC) oscillator with a phase drop δ across a coupling inductance placed between the two Josephson junctions. The longitudinal coupling results from mutual inductance between the oscillator and the transmon and the oscillator varies a flux Φ_x in the transmon. The transmon may have a flux sweet spot at integer values of a magnetic flux quantum Φ₀, Josephson energy asymmetry of the transmon may be below 0.02 and Φ_x may vary by ±0.05Φ₀ around Φ_x=0.

The method may also further comprise detuning the resonator from the qubit frequency ω_a by |Δ|≥{tilde over (g)}_z. Optionally, the oscillator inductance may be provided by an array of Josephson junctions or by one or more Superconducting Quantum Interference Device (SQUID).

BRIEF DESCRIPTION OF THE DRAWINGS

Further features and exemplary advantages of the present invention will become apparent from the following detailed description, taken in conjunction with the appended drawings, in which:

FIG. 1 is schematic representation of pointer state evolution in a phase space from an initial oscillator vacuum state;

FIG. 2 is graph presenting pointer state separation in a phase space as a function of time;

FIG. 3 is a graphical representation of different signal-to-noise ratios versus integration time different couplings;

FIG. 4 is a graphical representation of SNR versus integration time for different couplings;

FIG. 5 is a graphical representation of the measurement time T required to reach a fidelity of 99.99%;

FIG. 6 is a graphical representation of resonator damping rate κ/2π to reach a fidelity of 99.99% in 50 ns;

FIG. 7 is a logical representation of a circuit QED implementation of the longitudinal coupling modulation;

FIG. 8 is a graphical representation of flux Φ_x/Φ₀ dependence of both g_z and g_x;

FIG. 9 is a graphical representation of transmon frequency versus flux; FIG. 9 is a graphical representation of transmon frequency versus flux;

FIG. 10 is a logical representation of physical characteristics of a transmon qubit coupling longitudinally to a resonator;

FIG. 11 is a close-up logical representation of physical characteristics of the transmon qubit;

FIG. 12 is a logical representation of physical characteristics of a multi-qubit architecture;

FIGS. 13A, 13B, 13C and 13D are phase space representation of pointer states for multiple qubits;

FIGS. 14A, 14B and 14D are schematic illustrations, in rotating frame, of the qubit-state dependent oscillator field in phase space;

FIGS. 15A, 15B and 15C are schematic illustrations of average gate infidelity and gate time;.

FIG. 16 is a logical representation of physical characteristics of longitudinal interaction with two qubits coupled to a common LC oscillator;

FIG. 17 is a logical representation of physical characteristics of longitudinal interaction with two qubits is series; and

FIGS. 18A and 18B are tensor representations of a CZ gate supermatrix.

DETAILED DESCRIPTION

In one set of embodiments, a controlled-phase gate based on longitudinal coupling is disclosed. The controlled-phase gate is obtained by simultaneously modulating the longitudinal coupling strength of two qubits to a common resonator or two coupled resonators. The resonator may also be presented equivalently with different terms such as an oscillator, a cavity or qubit-cavity. In contrast to the more common transversal qubit-cavity coupling, the magnitude of the resulting effective qubit-qubit interaction does not rely on a small perturbative parameter. As a result, this interaction strength can be made large, resulting in effective gate times and gate fidelities. The gate fidelity can be exponentially improved by squeezing the resonator field and the approach may be generalized to qubits coupled to separate resonators. The longitudinal coupling strength between a qubit and a resonator may also be referred to as longitudinal qubit-resonator interaction. Reference is made to the drawings throughout the following description.

In another set of embodiments, an effective quantum non-demolition (QND) qubit readout is disclosed by modulating longitudinal coupling strength between a qubit and a resonator. The longitudinal coupling strength between a qubit and a resonator may also be referred to as longitudinal qubit-resonator interaction. In one embodiment, the QND qubit readout is accomplished by modulating longitudinal coupling between a resonator and a qubit at the resonator frequency. The resonator may also be presented as an oscillator, a cavity or qubit cavity. The longitudinal coupling strength then provides a qubit-state dependent signal from the resonator. This situation is fundamentally different from the standard dispersive case. Single-mode squeezing can optionally be exploited to increase the signal-to-noise ratio of the qubit readout protocol. An exemplary implementation of the qubit readout is provided in circuit quantum electrodynamics (circuit QED) and a possible multi-qubit architecture is also exemplified. Reference is made to the drawings throughout the following description.

For quantum information processing, qubit readout is expected to be fast, of high-fidelity and ideally QND. In order to rapidly reuse the measured qubit, fast reset of the measurement pointer states is also needed. Combining these characteristics is essential to meet the stringent requirements of fault-tolerant quantum computation. Dispersive readout relies on coupling the qubit to an oscillator acting as pointer. With the qubit modifying the oscillator frequency in a state-dependent fashion, driving the oscillator displaces its initial vacuum state to qubit-state dependent coherent states. Resolving these pointer states by homodyne detection completes the qubit measurement. The dispersive readout approach is used with superconducting qubits and quantum dots, and is studied in a wide range of systems including donor-based spin qubits and Majorana fermions. The same qubit-oscillator interaction is used to measure the oscillator state in cavity QED with Rydberg atoms.

Embodiments of the present invention provide parametric modulation of longitudinal qubit-resonator interaction for a faster, high-fidelity and ideally QND qubit readout with a reset mechanism. Embodiments of the present invention show that the signal-to-noise ratio (SNR) of the qubit readout can be further improved with a single-mode squeezed input state on the resonator. Like dispersive readout, the approach presented herein is applicable to a wide variety of systems. Skilled people will readily recognize that the modulation principle presented herein could be applied outside of the superconductive context.

A quantum information unit (qubit) memory is provided with a qubit frequency ω_a characterized by {circumflex over (σ)}_z. A resonator is provided that is defined by a resonator damping rate κ, a resonator frequency ω_r, a resonator electromagnetic field characterized by â^† and â, a longitudinal coupling strength g_z and an output â_out.

Conventional dispersive readout of {circumflex over (σ)}_z relies on transversal qubit-resonator coupling defined with a Hamiltonian H_x=g_x(â^†+â){circumflex over (σ)}_z. Embodiments of the present invention rely on longitudinal coupling (or longitudinal interaction) between the resonator and the qubit memory defined with a Hamiltonian Ĥ_z=g_z{circumflex over (σ)}_z(â^†+â). Despite the apparently minimal change, the choice of focusing the qubit readout on longitudinal coupling improves qubit readout. First, longitudinal coupling leads to an efficient separation of the pointer states. Indeed, Ĥ_z is the generator of displacement of the oscillator field with a qubit-state dependent direction.

FIG. 1 of the drawings provides a schematic representation of pointer state evolution of the intra-resonator field â in a phase space from an initial resonator vacuum state. FIG. 2 provides a corresponding graph presenting pointer state separation in a phase space as a function of time (t). Full lines 1000 relate to longitudinal modulation readout while dispersive readout for a dispersive shift

$\chi =\frac{\kappa }{2}$

is illustrated by the dashed lines 1100. Evolution from the initial resonator vacuum state is illustrated in phase space by lines 1000 and 1100 after a certain time (t). Discrete times and steady-state (t→∞) are provided for illustrative purposes with similar times being depicted using similar hatching patterns on lines 1000 and 1100. Circle size around at each time are used to illustrate uncertainty on the corresponding measured value, but are not drawn to scale on FIG. 1. As can be appreciated, dispersive readout illustrated by the dashed lines 1100 on FIG. 1 provides for complex path in phase space and poor separation of the pointer states at short times.

For this reason, even for identical steady-state separation of the pointers, longitudinal readout is significantly faster than its dispersive counterpart. On FIG. 2, pointer state separation is depicted for the resonator output field â_out as a function of time t. Vertical dashed lines (101 to 104) correspond to the same lines in FIG. 1.

The results depicted on FIG. 1 may be obtained using a modulator periodically modulating the longitudinal coupling strength g_z with a signal of amplitude {tilde over (g)}_z greater than or equal to the resonator damping rate κ and of frequency ω_m with ω_m±κ resonant with ω_r±a correction factor. The correction factor is <ω_r/10, and may for instance be between 0 and ω_r/100. In some embodiments, {tilde over (g)}_z is at least 3 times κ or at least 10 times κ. The longitudinal coupling strength g_z varies over time (t) in accordance with g_z(t)=g_z+{tilde over (g)}_z cos (ω_mt) with g_z representing an average value of g_z. Using a homodyne detector, the value {circumflex over (σ)}_z of the qubit memory is measured from a phase reading of the output â_out.

FIG. 3 shows a logic depiction of an exemplary readout—reset cycle. After a measurement time τ, longitudinal modulation amplitude {tilde over (g)}_z is reversed (i.e., −{tilde over (g)}_z) during a time τ to move the pointer state to the origin irrespective of the qubit state.

As can be appreciated, larger pointer state separations is achieved using a longitudinal modulator because Ĥ_z commutes with the value of the qubit {circumflex over (σ)}_z (also referred to as the measured qubit observable), resulting in an ideally QND readout. The situation is different from the dispersive case because (Ĥ_x|{circumflex over (σ)}_z)≈0. In the dispersive regime, where the qubit-resonator detuning Δ is large with respect to g_x, non-QNDness manifests itself with Purcell decay and with the experimentally observed measurement-induced qubit transitions. For these reasons, the resonator damping rate κ cannot be made arbitrarily large using and the measurement photon number n is typically kept well below the critical photon number n_crit=(Δ/2g_x)². As a result, dispersive readout is typically slow (small κ) and limited to poor pointer state separation (small n).

Under longitudinal coupling, the qubit-resonator Hamiltonian reads (ℏ/2pi) Ĥ_z=ω_râ^†â+½ω_a{circumflex over (σ)}z+g_z{circumflex over (σ)}_z(â^†+â) (referred to as Equation 1 hereinafter).

In steady-state, Equation 1 leads to a qubit-state dependent displacement of the resonator field amplitude

$\pm \frac{g_z}{{\omega }_r+\frac{i\kappa }{2}}.$

A static longitudinal interaction is therefor of no consequence for the typical case where ω_r«g_z, κ.

It is proposed herein to render the longitudinal interaction resonant during qubit readout by modulating the longitudinal coupling at the resonator frequency: g_z(t)=g_z+{tilde over (g)}_z cos (ω_mt). From the perspective of the longitudinal coupling and neglecting fast-oscillating terms, the following Equation 2 is obtained: {tilde over (H)}_z=½{tilde over (g)}_z{circumflex over (σ)}_z(a^†+â)

From Equation 2, it can be appreciated that a large qubit-state dependent displacement

$\pm \frac{{\tilde{g}}_z}{\kappa }$

is realized. Even with a conservative modulation amplitude {tilde over (g)}_z˜10κ, the steady-state displacement corresponds to 100 photons and the two qubit states are easily distinguishable by homodyne detection. With this longitudinal coupling, there is no concept of critical photon number and a large photon population is therefore not expected to perturb the qubit. Moreover, as already illustrated in FIG. 1, the pointer states take an improved path in phase space towards their steady-state separation. As shown in FIG. 2, this leads to a large pointer state separation at short times.

The consequences of using longitudinal coupling for qubit measurement can be quantified with the signal-to-noise ratio (SNR). The SNR quantity is evaluated using {circumflex over (M)}(τ)=√κ∫₀^τδt[â^†_out(t)+â_out(t)] the measurement operator for homodyne detection of the output signal â_out with a measurement time τ. The signal is defined as |{circumflex over (M)}₁−{circumflex over (M)}₀| where {0,1} refers to qubit state, while the imprecision noise is [{circumflex over (M)}_N1²(τ)°{circumflex over (M)}_N0²(τ)]^1/2 with {circumflex over (M)}_N={circumflex over (M)}−{circumflex over (M)}.

Combining these expressions, the SNR for the longitudinal case reads in accordance with Equation 3:

${\mathrm{SNR}}_z=\sqrt{8}\frac{{\tilde{g}}_z}{\kappa }\sqrt{\kappa \tau }\left[1-\frac{2}{\kappa \tau }\left(1-e^{-\frac{1}{2}\kappa \tau }\right)\right]$

This is to be contrasted to SNR_χ for dispersive readout with drive amplitude ϵ and optimal dispersive coupling χ=g_x²/Δ=κ/2 in accordance with Equation 4:

${\mathrm{SNR}}_{\chi }=\sqrt{8}\frac{\in }{\kappa }\sqrt{\mathrm{\kappa \tau }}\left[1-\frac{2}{\kappa \tau }\left(1-e^{-\frac{1}{2}\kappa \tau }\cos \frac{1}{2}\kappa \tau \right)\right]$

Both expressions have a similar structure, making very clear the similar role of {tilde over (g)}_z and E, except for the cosine in Equation (4) that is a signature of the complex dispersive path in phase space. For short measurement times κτ«1, a favorable scaling is obtained for longitudinal modulation readout with SNR_z∝SNR_χ/κτ.

FIG. 4 shows a graphical representation of SNR versus integration time for longitudinal (107) and dispersive without Purcell decay (105) coupling. Even for equal steady-state separation ({tilde over (g)}_z=ϵ), shorter measurement time are obtained for longitudinal coupling. SNR is shown in units of {tilde over (g)}_z/κ as a function of integration time τ. Longitudinal coupling (107) is compared to dispersive coupling (105) with χ=κ/2 for the same steady-state separation, |{tilde over (g)}_z|=|ϵ|. Line (106) accounts for Purcell decay in dispersive readout. Line (108) shows exponential improvement obtained for when a single-mode squeezed input state with e²⁴=100 (20 dB).

FIG. 5 provides a graphical representation of the measurement time τ required to reach a fidelity of 99.99% as a function of {tilde over (g)}_z/κ (or ϵ/κ for the dispersive case). When taking into account the non-perturbative effects that affect the QNDness of dispersive readout, the potential of the present approach is made even clearer. Line (106) of FIG. 4 and FIG. 5 correspond to the dispersive case with Purcell decay. In this more realistic case, longitudinal readout outperforms its counterpart at all times. FIG. 6 provides a graphical representation of resonator damping rate κ/2π to reach a fidelity of 99.99% in τ=50 ns versus intra-resonator photon number n=({tilde over (g)}_z/κ)²=(ϵ/κ)². Squeezing (108) helps in further reducing the required photon number or resonator damping rate. The squeeze strength is optimized for each κ, with a maximum set to 20 dB reached close to κ/2π=1 MHz In FIGS. 5 and 6, results for the dispersive readout are stopped at the critical photon number obtained for a drive strength ϵ_crit=Δ/√8g_x for g_x/Δ= 1/10.

Up to this point, equal pointer state separation has been assumed for the longitudinal and the dispersive readouts. As already mentioned, dispersive readout is, however, limited to measurement photon numbers well below n_crit. This is taken into account in FIGS. 5 and 6 by stopping the dispersive curves at n_crit (black circle) assuming the typical value g_x/Δ= 1/10. FIG. 5 illustrates that only longitudinal readout allows for measurement times <1/κ. This is moreover achieved for reasonable modulation amplitudes with respect to the cavity linewidth As illustrated, the longitudinal coupling strength g_z with having a signal of amplitude {tilde over (g)}_z at least three (3) times greater than the resonator damping rate κ still allows for qubit readout. On FIG. 6, the resonator damping rate vs photon number required to reach a fidelity of 99.99% in υ=50 ns is illustrated. Note that line (106) corresponding to dispersive with Purcell is absent from this plot. With dispersive readout, it appears impossible to achieve the above target fidelity and measurement time in the very wide range of parameters of FIG. 6. On the other hand, longitudinal readout with quite moderate values of κ and n provide meaningful results. Further speedups are expected with pulse shaping and machine learning. Because the pointer state separation is significantly improved even at short time, the latter approach should be particularly efficient.

To allow for rapid reuse of the qubit, the resonator should be returned to its grounds state ideally in a time «1/κ after readout. A pulse sequence achieving this for dispersive readout has been proposed but is imperfect because of qubit-induced nonlinearity deriving from Ĥ_x.

As illustrated in FIG. 3, with an approach based on longitudinal modulation as proposed herein, resonator reset is realized by inverting the phase of the modulation. Since Ĥ_z does not lead to qubit-induced nonlinearity, the reset remains ideal. In practice, reset can also be shorter than the integration time. It is also interesting to point out that longitudinal modulation readout saturates the inequality Γ_φm≥Γ_meas linking the measurement-induced dephasing rate Γ_φm to the measurement rate Γ_meas and is therefore quantum limited.

Another optional feature of the longitudinal modulation readout to improve SNR (theoretically exponentially) by providing a single-mode squeezed input state on the resonator. The squeeze axis is chosen to be orthogonal to the qubit-state dependent displacement generated by g_z(t). referring back to the example of FIG. 1, squeezing would provide for a more defined value along the vertical axis. Since the squeeze angle is unchanged under evolution with Ĥ_z, the imprecision noise is exponentially reduced along the vertical axis while the imprecision noise is left to augment along the horizontal axis. The signal-to-noise ratio becomes e^rSNR_z, with r the squeeze parameter. This exponential enhancement is apparent from line (108) depicted in FIG. 4 and in the corresponding reduction of the measurement time in FIG. 5. Note that by taking g_z=0, the resonator field can be squeezed prior to measurement without negatively affecting the qubit.

The exponential improvement is in contrast to standard dispersive readout where single-mode squeezing can lead to an increase of the measurement time. Indeed, under dispersive coupling, the squeeze angle undergoes a qubit-state dependent rotation. As a result, both the squeezed and the anti-squeezed quadrature contributes to the imprecision noise. It is to be noted that the situation can be different in the presence of two-mode squeezing where an exponential increase in SNR can be recovered by engineering the dispersive coupling of the qubit to two cavities.

While the longitudinal modulation approach is very general, a circuit QED implementation 700 is discussed in greater details hereinafter with reference to FIG. 7. Longitudinal coupling of a flux or transmon qubit to a resonator of the LC oscillator type may result from the mutual inductance between a flux-tunable qubit and the resonator. Another example focuses on a transmon qubit phase-biased by the oscillator. FIG. 7 shows a schematically lumped version of an exemplary circuit QED implementation 700 considering the teachings of the present disclosure. In practice, the inductor L can be replaced by a Josephson junction array, both to increase the coupling and to reduce the qubit's flux-biased loop size. Alternatives (e.g., based on a transmission-line resonator) may also be realized as it is explored with respect to other embodiments described hereinbelow.

The Hamiltonian of the circuit of FIG. 7 is similar to that of a flux-tunable transmon, but where the external flux Φ_x is replaced by Φ_x+δ with δ the phase drop at the oscillator. Taking the junction capacitances Cs to be equal and assuming for simplicity that Z₀/R_K«1, with Z₀=√L/C and R_K the resistance quantum, the Hamiltonian of the circuit QED 700 may be expressed as Ĥ=Ĥ_r+Ĥ_q+Ĥ_qr, with H_r=ω_râ^†â resenting the oscillator Hamiltonian and Ĥ_q=½ω_a{circumflex over (σ)}_z presenting the Hamiltonian of a flux transmon written here in its two-level approximation.

The Hamiltonian of the qubit-oscillator interaction (or longitudinal coupling strength) takes the form Ĥ_qr=g_x{circumflex over (σ)}_z(â^†+â)+g_z{circumflex over (σ)}_z(â^†+â) when Equation 5 and Equation 6 are satisfied:

$g_z=-\frac{E_J}{2}{\left(\frac{2E_C}{E_J}\right)}^{1/2}\sqrt{\frac{\pi Z_0}{R_K}}\sin \left(\frac{{\mathrm{\pi \Phi }}_x}{{\Phi }_0}\right),g_x={{\mathrm{dE}}_J\left(\frac{2E_C}{E_J}\right)}^{1/4}\sqrt{\frac{\pi Z_0}{R_K}}\cos \left(\frac{{\mathrm{\pi \Phi }}_x}{{\Phi }_0}\right)$

where E_Jis the mean Josephson energy, d the Josephson energy asymmetry and E_C the qubit's charging energy. Skilled person will readily be able to locate expressions for these quantities in terms of the elementary circuit parameters. In the circuit QED implementation 700, E_J1=E_J(1+d)/2 and E_J2=E_J(1−d)/2 with dϵ[0,1]. As purposely pursued, the transverse coupling g_x vanishes exactly for d=0, leaving only longitudinal coupling g_z. Because longitudinal coupling is related to the phase bias rather than inductive coupling, g_z can be made large.

For example, with the realistic values E_J/h=20 GHz, E_J/E_C=67 and Z₀=50Ω, g_z/2π≈r=135 MHz×sin (πΦ_x/Φ₀) where Φ₀ represents the magnetic flux quantum. FIG. 8 shows a graphical representation of flux Φ_z/Φ₀ dependence of both g_z (full line) and g_x with d=0 (dash-dotted line) and d=0.02 (dashed line). Modulating the flux by 0.05Φ₀ around Φ_x=0, it follows that g_z=0 and {tilde over (g)}_z/2π˜21 MHz Conversely, only a small change of the qubit frequency of ˜40 MHz is affected, as can be appreciated from FIG. 9 showing transmon frequency versus flux in accordance with the preceding exemplary parameters. Importantly, this does not affect the

SNR.

As can be appreciated, a finite g_x for d≠0. On FIG. 8, a realistic value of d=0.02 and the above parameters, g_x/2π≈13 MHz×cos (πΦ_x/Φ₀). The effect of this unwanted coupling can be mitigated by working at large qubit-resonator detuning Δ where the resulting dispersive interaction χ=g_x²/2Δ can be made very small. For example, the above numbers correspond to a detuning of Δ/2π=3 GHz where χ/2π˜5.6 kHz. It is important to emphasize that, contrary to dispersive readout, the longitudinal modulation approach is not negatively affected by a large detuning.

When considering higher-order terms in Z₀/R_K, the Hamiltonian of the circuit QED 700 exemplified in FIG. 7 contains a dispersive-like interaction χ_zâ^†â{circumflex over (σ)}_z even at d=0. For the parameters already used above, χ_z/2π˜5.3 MHz, a value that is not made smaller by detuning the qubit from the resonator. However since it is not derived from a transverse coupling, χ_z is not linked to any Purcell decay. Moreover, it does not affect SNR_z at small measurement times.

In the absence of measurement, g_z=ĝ_z=0 and the qubit may advantageously be parked at its flux sweet spot (e.g., integer values of Φ₀). Dephasing due to photon shot noise or to low-frequency flux noise is therefore expected to be minimal. Because of the longitudinal coupling, another potential source of dephasing is flux noise at the resonator frequency which will mimic qubit measurement. However, given that the spectral density of flux noise is proportional to 1/f even at high frequency, this contribution is negligible.

In the circuit QED implementation, the longitudinal readout can also be realized with a coherent voltage drive of amplitude ϵ(t) applied directly on the resonator, in place of a flux modulation on the qubit. Taking into account higher-order terms in the qubit-resonator interaction, the full circuit Hamiltonian without flux modulation can be approximated to (g_z=0)

Ĥ=ω _r â ^†â+½ω_a{circumflex over (σ)}_z+χ_zâ^†â{circumflex over (σ)}_z+iγ(t)(â^†−â)

the well-known driven dispersive Hamiltonian where the AC-Stark shift interaction originating from the higher-order longitudinal interaction is given by

${\chi }_z=-\frac{\sqrt{E_JE_C}}{h}\frac{\pi Z_r}{R_K}$

Assuming a drive resonant with the resonator frequency ω_r with phase φ=0 for simplicity, in the rotating frame the system Hamiltonian becomes (neglecting fast-rotating terms)

$\hat{H}+{\chi }_z{\hat{a}}^{†}\hat{a}{\hat{\sigma }}_z+i\frac{ϵ}{2}\left({\hat{a}}^{†}-\hat{a}\right)$

Under a displacement transformation D(α)âD^†(α)→â−α and including the resonator dissipation, the following is obtained:

$\hat{H}+\chi \left({\hat{a}}^{†}-{\alpha }^{*}\right)\left(\hat{a}-\alpha \right){\hat{\sigma }}_z+i\frac{ϵ}{2}\left({\hat{a}}^{†}-\hat{a}\right)+i\frac{\kappa }{2}\left(\alpha {\hat{a}}^{†}-{\alpha }^{*}\hat{a}\right)$

Finally, choosing α=ϵ/κ, the system is now simplifed to

Ĥ=χ _z â ^†â{circumflex over (σ)}_z+g′_z(â^†+â){circumflex over (σ)}_z

with an effective (driven) longitudinal interaction with strength g′_z=χ_zϵ/κ. In a regime of large voltage drive amplitudes with ϵ/κ»1, the voltage drive performs the ideal longitudinal readout as the residual dispersive effects are mitigated g′_z»χ_z. As mentioned earlier, the absence of Purcell decay and of any critical number of photons in the system allows to push the standard dispersive readout mechanism towards the ideal limit of the longitudinal readout.

A possible multi-qubit architecture consists of qubits longitudinally coupled to a readout resonator (of annihilation operator â_z) and transversally coupled to a high-Q bus resonator (â_x). The Hamiltonian describing this system is provided by Equation 7:

$\hat{H}={\omega }_{\mathrm{rz}}{\hat{a}}_z^{†}{\hat{a}}_z+{\omega }_{\mathrm{rz}}{\hat{a}}_x^{†}{\hat{a}}_x+\sum _j\frac{1}{2}{\omega }_{\mathrm{aj}}{\hat{\sigma }}_{\mathrm{zj}}+\sum _jg_{\mathrm{zj}}{\hat{\sigma }}_{\mathrm{zj}}\left({\hat{a}}_z^{†}+{\hat{a}}_z\right)+\sum _jg_{\mathrm{xj}}{\hat{\sigma }}_{\mathrm{xj}}\left({\hat{a}}_x^{†}+{\hat{a}}_x\right)$

Readout can be realized using longitudinal coupling while logical operations via the bus resonator. Alternative architectures, e.g., taking advantage of longitudinal coupling may also be proposed. Here for instance, taking g_zj(t)=g_z+{tilde over (g)}_z cos (ω_rt+φ_j), the longitudinal coupling, from the perspective of the longitudinal interaction and neglecting fast-oscillating terms, is represented by Equation 8: Ĥ_z=(½g_zΣ_j{circumflex over (σ)}_zje^−iφj)â_z+H.c.

This effective resonator drive displaces the field to multi-qubit-state dependent coherent states. For two qubits and taking φ_j=jπ/2 leads to the four pointer states separated by 90° from each other or, in other words, to an optimal separation even at short times. Other choices of phase lead to overlapping pointer states corresponding to different multi-qubit states. Examples are φ_j=0 for which |01¢ and |10 are indistinguishable, and φ_j=jπ where |00 and |11 are indistinguishable. However, these properties may be exploited to create entanglement by measurement. As another example, with 3 qubits the GHZ state may be obtained with φ_j=j2π/3.

In the following pages, additional sets of embodiments are presented. In a first additional set of exemplary embodiments, longitudinal coupling is considered (A). In a second additional set of exemplary embodiments, longitudinal coupling with single-mode squeezed states is presented (B). Standard dispersive coupling (C) as well as innovative longitudinal coupling in presence of transverse coupling in the dispersive regime (D) are also considered.

A. Longitudinal coupling

1. Modulation at the Resonator Frequency

The first set of embodiments (A) considers a qubit longitudinally coupled to a resonator with the Hamiltonian:

Ĥ=ω _r â ^† â+½ω{circumflex over (σ)}_z+[g_z(t)â^†+g_z*(t)â]{circumflex over (σ)} _z.   (S1)

In this expression, ω_r is the resonator frequency, ω_a the qubit frequency and g_z is the longitudinal coupling that is modulated at the resonator frequency:

g _z(t)=g_z+|{tilde over (g)}_z| cos (ω_rt+φ).   (S2)

In the interaction picture and using the Rotating-Wave Approximation (RWA), the above Hamiltonian simplifies to

Ĥ=½[{tilde over (g)}_zâ^†+{tilde over (g)}_z*â]{circumflex over (σ)}_z,   (S3)

where {tilde over (g)}_z≡|{tilde over (g)}_z|e^iφ is the modulation amplitude. From Equation (S3), it is clear that the modulated longitudinal coupling plays the role of a qubit-state dependent drive. The Langevin equation of the cavity field simply reads

{dot over (â)}=−i½{tilde over (g)}_z{circumflex over (σ)}_z−½κâ−√κâ_in,   (S4)

where â_in is the input field. Taking this input to be the vacuum, the input correlations are then defined by:

Using the input-output boundary â_out=â_in+√κâ, it integration of the Langevin equation leads to {circumflex over (â)}_in(t)â_in^†(t′)=[â_in(t), â_in^†(t′)]=δ(t−t′)

$\begin{array}{cc}{\alpha }_{\mathrm{out}}\left(t\right)=-\frac{i{\tilde{g}}_z}{\sqrt{\kappa }}〈{\hat{\sigma }}_z〉\left[1-e^{-\frac{1}{2}\kappa t}\right],{\hat{d}}_{\mathrm{out}}\left(t\right)={\hat{d}}_{\mathrm{in}}\left(t\right)-\kappa {\int }_{-\infty }^t{\mathrm{dt}}^{\prime }e^{-\frac{1}{2}\kappa \left(t-t^{\prime }\right)}{\hat{d}}_{\mathrm{in}}\left(t^{\prime }\right),& \left(S5\right)\end{array}$

where α_out=â_out stands for the output field mean value and {circumflex over (d)}_out=â_out−α_out its fluctuations. Because here the qubit-dependent drive comes from modulations of the coupling, and not from an external coherent drive, there is no interference between the outgoing and the input fields. As a result, α=α_out/√κ and the intracavity photon number evolves as

$\begin{array}{cc}〈{\hat{a}}^{†}\hat{a}〉={\frac{{{\tilde{g}}_z}^2}{{\kappa }^2}\left[1-e^{-\frac{1}{2}\kappa t}\right]}^2.& \left(S6\right)\end{array}$

The measurement operator corresponding to homodyne detection of the output signal with an integration time τ and homodyne angle φ_h is

{circumflex over (M)}(τ)=√κ∫₀^τdt[â_out^†(t)e^iφh+â_out(t)e^−iφh].   (S7)

The signal for such a measurement is M while the noise operator is {circumflex over (M)}_N={circumflex over (M)}−{circumflex over (M)} In the presence of a qubit, the measurement signal is then

$\begin{array}{cc}{〈\hat{M}〉}_1-{〈\hat{M}〉}_0=4{\tilde{g}}_z\sin \left(\varphi -{\phi }_h\right)\tau \left[1-\frac{2}{\kappa \tau }\left(1-e^{-\frac{1}{2}\mathrm{\kappa \tau }}\right)\right].& \left(\mathrm{S8}\right)\end{array}$

On the other hand, the measurement noise is equal to {circumflex over (M)}_N²(τ)=κτ. Combining these two expressions, the signal-to-noise ratio (SNR) then reads

$\begin{array}{cc}\begin{array}{c}{\mathrm{SNR}}^2\equiv \frac{{{〈\hat{M}〉}_1-{〈\hat{M}〉}_0}^2}{〈{\hat{M}}_{N1}^2\left(\tau \right)〉+〈{\hat{M}}_{N0}^2\left(\tau \right)〉}\\ =\frac{8{{\tilde{g}}_z}^2}{{\kappa }^2}{\sin }^2\left(\varphi -{\phi }_h\right)\kappa {\tau \left[1-\frac{2}{\kappa \tau }\left(1-e^{-\frac{1}{2}\mathrm{\kappa \tau }}\right)\right]}^2.\end{array}& \begin{array}{c}\left(\mathrm{S9}\right)\\ \begin{array}{c}\\ \left(\mathrm{S10}\right)\end{array}\end{array}\end{array}$

The SNR is optimized by choosing the modulation phase φ and the homodyne angle such that φ−φh=modπ.With this choice, the optimized SNR finally reads:

$\begin{array}{cc}\mathrm{SNR}=\sqrt{8}\frac{{\tilde{g}}_z}{\kappa }\sqrt{\mathrm{\kappa \tau }}\left[1-\frac{2}{\kappa \tau }\left(1-e^{-\frac{1}{2}\kappa \tau }\right)\right].& \left(S11\right)\end{array}$

At long measurement times (τ»1/κ), the signal-to-noise ratio evolves as

$\mathrm{SNR}=\sqrt{8}\frac{{\tilde{g}}_z}{\kappa }\sqrt{\mathrm{\kappa \tau }}$

while in the more experimentally interesting case of short measurement times leads to

$\mathrm{SNR}{=}^{\frac{1}{\sqrt{2}}\frac{{\tilde{g}}_z}{\kappa }{\left(\kappa \tau \right)}^{3/2}}.$

In short, the SNR increases as τ^3/2, much faster than in the dispersive regime where the SNR rather increases as τ^5/2, as will become apparent in (C) below.

2. Measurement and Dephasing Rates

To evaluate the measurement-induced dephasing rate, a polaron-type transformation is applied on Hamiltonian from (S3) consisting of a displacement of a â by −i{tilde over (g)}_z{circumflex over (σ)}_z/κ. this transformation, the cavity decay Lindbladian κ[â]{circumflex over (ρ)}=[â]ρ=âρâ^†−½{â^†â, ρ}, leads to ½Γ_φm[{circumflex over (σ)}_z]{circumflex over (ρ)} where Γ_φm=2[ĝ_z]²/κ is the measurement-induced dephasing. On the other hand, the measurement rate is obtained from the SNR as Γ_meas=SNR²/(4τ)=2[{tilde over (g)}_z]²/κ.

The relation between the dephasing and the measurement rate is then Γ_meas=Γ_φm. This is the bound reached for a quantum limited measurement.

3. Modulation Bandwidth

A situation where the longitudinal coupling is modulated at a frequency ω_m≠ω_r is now considered. That is, g_z(t)=g_z+|{tilde over (g)}_z| cos (ω_mt+φ). Assuming that the detuning Δm=ω_m−ω_r is small with respect to the modulation amplitude {tilde over (g)}_z, the Hamiltonian in a frame rotating at the modulation frequency now reads under the RWA as:

Ĥ=−Δ _m â ^† â+½[{tilde over (g)}_zâ^†+{tilde over (g)}_zâ]{circumflex over (σ)}_a,   (S12)

The corresponding Langevin equation is then

{dot over (â)}=−i1/2{tilde over (g)}_z{circumflex over (σ)}_a−iΔ_mâ−1/2κâ−√κâ_in,   (S13)

yielding for the output field,

$\begin{array}{cc}{\alpha }_{\mathrm{out}}\left(t\right)=-\frac{i{\tilde{g}}_z\sqrt{\kappa }}{\kappa -2i{\Delta }_m}〈{\hat{\sigma }}_z〉\left[1-e^{\left(i{\Delta }_m-\frac{1}{2}\kappa \right)\tau }\right],{\hat{d}}_{\mathrm{out}}\left(t\right)={\hat{d}}_{\mathrm{in}}\left(t\right)-\kappa {\int }_{-\infty }^t{\mathrm{dt}}^{\prime }e^{\left(i{\Delta }_m-\frac{1}{2}\kappa \right)\left(t-t^{\prime }\right)}{\hat{d}}_{\mathrm{in}}\left(t^{\prime }\right).& \left(S14\right)\end{array}$

From these expressions, the measurement signal is then

$\begin{array}{cc}{〈\hat{M}〉}_1-{〈\hat{M}〉}_0=\frac{4{\tilde{g}}_z}{\sqrt{1+{\left(\frac{2{\Delta }_m}{\kappa }\right)}^2}}\sin \left[\varphi -{\phi }_h+\arctan \left(2{\Delta }_m/\kappa \right)\right]\tau -\frac{8{\tilde{g}}_z/\kappa }{1+{\left(\frac{2{\Delta }_m}{\kappa }\right)}^2}\sin \left[\varphi -{\phi }_h+2\arctan \left(2{\Delta }_m/\kappa \right)\right]+\frac{8{\tilde{g}}_z/\kappa }{1+{\left(\frac{2{\Delta }_m}{\kappa }\right)}^2}\sin \left[\varphi -{\phi }_h+2\arctan \left(2{\Delta }_m/\kappa \right)+{\Delta }_m\tau \right]e^{-\frac{1}{2}\kappa \tau }.& \left(\mathrm{S15}\right)\end{array}$

While the signal is changed, the noise is however not modified by the detuning. From the above expression, given a detuning Am and a measurement time T, there is an optimal angle φ that maximizes the SNR.

B. Longitudinal coupling with squeezing

In the second set of exemplary embodiments (B), a situation where the modulation detuning is zero and where the input field is in a single-mode squeezed vacuum is now considered. This leaves the signal unchanged, but as will be appreciated, leads to an exponential increase of the SNR with the squeeze parameter r. Indeed, in the frame of the resonator, the correlations of the bath fluctuations are now

$\begin{array}{cc}\left(\begin{array}{cc}〈{\hat{d}}_{\mathrm{in}}^{†}\left(t\right){\hat{d}}_{\mathrm{in}}\left(t^{\prime }\right)〉& 〈{\hat{d}}_{\mathrm{in}}\left(t\right){\hat{d}}_{\mathrm{in}}\left(t^{\prime }\right)〉\\ 〈{\hat{d}}_{\mathrm{in}}^{†}\left(t\right){\hat{d}}_{\mathrm{in}}^{†}\left(t^{\prime }\right)〉& 〈{\hat{d}}_{\mathrm{in}}\left(t\right){\hat{d}}_{\mathrm{in}}^{†}\left(t^{\prime }\right)〉\end{array}\right)=\left(\begin{array}{cc}{\sinh }^2r& \frac{1}{2}\sinh 2re^{2i\theta }\\ \frac{1}{2}\sinh 2re^{-2i\theta }& {\cosh }^2r\end{array}\right)\delta \left(t-t^{\prime }\right).& \left(\mathrm{S16}\right)\end{array}$

where is has been assumed broadband squeezing with a squeeze angle θ. The measurement noise is then

$\begin{array}{cc}〈{\hat{M}}_N^2\left(\tau \right)〉=\kappa {\int }_0^{\tau }\mathrm{dt}{\int }_0^{\tau }{\mathrm{dt}}^{\prime }\left[〈{\hat{d}}_{\mathrm{out}}^{†}\left(t\right){\hat{d}}_{\mathrm{out}}\left(t^{\prime }\right)〉+〈{\hat{d}}_{\mathrm{out}}\left(t\right){\hat{d}}_{\mathrm{out}}^{†}\left(t^{\prime }\right)〉+〈{\hat{d}}_{\mathrm{out}}\left(t\right){\hat{d}}_{\mathrm{out}}\left(t^{\prime }\right)〉e^{-2i{\phi }_h}+〈{\hat{d}}_{\mathrm{out}}^{†}\left(t\right){\hat{d}}_{\mathrm{out}}^{†}\left(t^{\prime }\right)〉e^{2{\mathrm{iO}}_h}\right].& \left(\mathrm{S17}\right)\end{array}$

The output-field correlations are easily obtained from

$\begin{array}{cc}\left(\begin{array}{cc}〈{\hat{d}}_{\mathrm{out}}^{†}\left(t\right){\hat{d}}_{\mathrm{out}}\left(t^{\prime }\right)〉& 〈{\hat{d}}_{\mathrm{out}}\left(t\right){\hat{d}}_{\mathrm{out}}\left(t^{\prime }\right)〉\\ 〈{\hat{d}}_{\mathrm{out}}^{†}\left(t\right){\hat{d}}_{\mathrm{out}}^{†}\left(t^{\prime }\right)〉& 〈{\hat{d}}_{\mathrm{out}}\left(t\right){\hat{d}}_{\mathrm{out}}^{†}\left(t^{\prime }\right)〉\end{array}\right)=\left(\begin{array}{cc}〈{\hat{d}}_{\mathrm{in}}^{†}\left(t\right){\hat{d}}_{\mathrm{in}}\left(t^{\prime }\right)〉& 〈{\hat{d}}_{\mathrm{in}}\left(t\right){\hat{d}}_{\mathrm{in}}\left(t^{\prime }\right)〉\\ 〈{\hat{d}}_{\mathrm{in}}^{†}\left(t\right){\hat{d}}_{\mathrm{in}}^{†}\left(t^{\prime }\right)〉& 〈{\hat{d}}_{\mathrm{in}}\left(t\right){\hat{d}}_{\mathrm{in}}^{†}\left(t^{\prime }\right)〉\end{array}\right)& \left(S18\right)\end{array}$

which holds here since the drive is ‘internal’ to the cavity. As a result))

{circumflex over (M)} _N ²(τ)={cos h(2r)−sin h(2r) cos [2(φ_h−θ)]}κτ.   (S19)

The noise is minimized by choosing θ according to

$\theta -\phi h=\frac{\pi }{2}\mathrm{mod}\pi .$

With this choice, the SNR reads

SNR(r)=e^rSNR(r=0).   (S20)

The SNR is thus exponentially enhanced, leading to Heisenberg-limited scaling.

A source of broadband Γ pure squeezing is assumed to be available. The effect of a field squeezing bandwidth r was already studied elsewhere, it only leads to a small reduction of the SNR for r»κ. On the other hand, deviation from unity of the squeezing purity P leads to a reduction of the SNR by 1/√P. The SNR being decoupled from the anti-squeezed quadrature, the purity simply renormalizes the squeeze parameter.

C. Dispersive coupling

For completeness, the SNR for dispersive readout is also provided, even though corresponding result may be found in the literature. In the dispersive regime, the qubit-cavity Hamiltonian reads

Ĥ=Φ _r â ^† â+½ω_a{circumflex over (σ)}_z+χâ^†â{circumflex over (σ)}_z,   (S21)

where χ=g_x²/Δ is the dispersive shift. The Langevin equation of the cavity field in the interaction picture then reads

{dot over (â)}=−iχ{circumflex over (σ)} _z â−½κâ−√κâ_in.   (S22)

With a drive of amplitude ϵ=|ϵ|e^−iφd on the cavity at resonance, the input field is defined by its mean a_in=â_in=−ϵ/√κ and fluctuations {circumflex over (d)}_in=â_in−α_in. Integrating the Langevin equation yields

$\begin{array}{cc}{\alpha }_{\mathrm{out}}\left(t\right)=\frac{ϵ}{\sqrt{\kappa }}e^{-i{\varphi }_{\mathrm{qb}}〈{\hat{\sigma }}_z〉}\left[1-2\cos \left(\frac{1}{2}{\varphi }_{\mathrm{qb}}\right)e^{-\left(i\chi 〈{\hat{\sigma }}_z〉+\frac{1}{2}\kappa \right)t+\frac{1}{2}i{\varphi }_{\mathrm{gb}}〈{\hat{\sigma }}_z〉}\right],& \left(S23\right)\\ {\hat{d}}_{\mathrm{out}}\left(t\right)={\hat{d}}_{\mathrm{in}}\left(t\right)-\kappa {\int }_{-\infty }^t{\mathrm{dt}}^{\prime }e^{-\left(i\chi 〈{\hat{\sigma }}_z〉-\frac{1}{2}\kappa \right)\left(t-t^{\prime }\right)}{\hat{d}}_{\mathrm{in}}\left(t^{\prime }\right).& \left(S24\right)\end{array}$

where φ_qb=2 arctan (2χ/κ) is the qubit-induced phase of the output field. Moreover, the intracavity photon number is as

$\begin{array}{cc}〈{\hat{a}}^{†}\hat{a}〉={\left(\frac{2ϵ}{\kappa }\right)}^2{\cos }^2\left(\frac{1}{2}{\varphi }_{\mathrm{qb}}\right)\left[1-2\cos \left(\chi t\right)e^{-\frac{1}{2}\kappa t}+e^{-\kappa t}\right].& \left(S25\right)\end{array}$

From the above expressions, the measurement signal is

$\begin{array}{cc}{M}_{S1〉}-M_{S0〉}=4ϵ\sin \left({\varphi }_{\mathrm{qb}}\right)\sin \left({\varphi }_d-{\phi }_h\right)\tau \left\{1-\frac{4}{\kappa \tau }{\cos }^2\left(\frac{1}{2}{\varphi }_{\mathrm{qb}}\right)\left[1-\frac{\sin \left(\chi \tau +{\varphi }_{\mathrm{qb}}\right)}{\sin \left({\varphi }_{\mathrm{qb}}\right)}e^{-\frac{1}{2}\kappa \tau }\right]\right\}& \left(\mathrm{S26}\right)\end{array}$

On the other hand, the measurement noise is simply equal to {circumflex over (M)}_N²(T)=κT. The measurement signal is optimized for

${\varphi }_d-{\phi }_h=\frac{\pi }{2}\mathrm{mod}\pi$

and at long integration times by

${\varphi }_{\mathrm{qb}}=\frac{\pi }{2},$

or equivalently χ=κ/2. For this optimal choice, the SNR then reads

At long measurement times, the SNR evolves as

$\mathrm{SNR}=\sqrt{8}\frac{ϵ}{\kappa }\sqrt{\kappa \tau }$

and at short measurement times it starts as:

$\begin{array}{cc}\mathrm{SNR}=\frac{1}{\sqrt{18}}\frac{ϵ}{\kappa }{\left(\kappa \tau \right)}^{5/2}\mathrm{SNR}=\sqrt{8}\frac{ϵ}{\kappa }\sqrt{\mathrm{\kappa \tau }}\left[1-\frac{2}{\mathrm{\kappa \tau }}\left(1-e^{-\frac{1}{2}\mathrm{\kappa \tau }}\cos \frac{1}{2}\mathrm{\kappa \tau }\right)\right].& \left(\mathrm{S27}\right)\end{array}$

The presence of Purcell decay γ_κ=(g/Δ)²κ is taken into account using the expression of {circumflex over (σ)}_z(t) for a Purcell-limited qubit, i.e., {circumflex over (σ)}_z(t)=(1+{circumflex over (σ)}_z(0))e^−γκt−1. The

$\begin{array}{cc}\mathrm{SNR}=\sqrt{2}\frac{ϵ}{\kappa }\sqrt{\mathrm{\kappa \tau }}\left\{1-\frac{2}{\mathrm{\kappa \tau }}\left(1-e^{-\frac{1}{2}\mathrm{\kappa \tau }}\cos \frac{1}{2}\mathrm{\kappa \tau }\right)-\frac{\kappa }{\tau }{\int }_0^{\tau }\mathrm{dt}{\int }_0^t{\mathrm{dt}}^{\prime }e^{-\frac{1}{2}\kappa \left(t-t^{\prime }\right)}\sin [\frac{1}{2}\kappa \left(t-t^{\prime }\right)-\frac{\kappa }{{\gamma }_{\kappa }}\left(e^{-{\gamma }_{\kappa }t^{\prime }}-e^{-{\gamma }_{\kappa }t}\right)]\right\}.& \left(\mathrm{S28}\right)\end{array}$

corresponding SNR is then, for χ=κ/2,

D. Effect of a residual transverse coupling

In the fourth set of exemplary embodiments (D), presence of a spurious transverse coupling g_x in addition to the longitudinal coupling g_z is considered whereby:

Ĥ=ω _r â ^† â+ω _a{circumflex over (σ)}_z+{[g_x+{tilde over (g)}_x cos (ω_rt+φ_x)]{circumflex over (σ)}_x+[g_z+{tilde over (g)}_z cos (ω_rt+φ)]{circumflex over (σ)}_z}(â^†+â).   (S29)

It is now assumed that g_x«Δ and we follow the standard approach to eliminate the transverse coupling. To leading order in g_x/Δ and under the RWA, the interaction picture is defined as:

{tilde over (H)}=½(χ_x−2χ_xz){circumflex over (σ)}_z+χâ^†â{circumflex over (σ)}_z+½[{tilde over (g)}_zâ^†+{tilde over (g)}_z*â]{circumflex over (σ)}_z,   (S30)

with the dispersive shifts χ=χ_x−4χ_xz·χ_x=g_x⁻²/Δ and χ_xz=g_xg_z/Δ.

Going to an interaction picture also with respect to the first term of Equation (S30), the starting point is

Ĥ=χâ ^† â{circumflex over (σ)} _z+½[{tilde over (g)}_zâ^†+{tilde over (g)}_z*â]{circumflex over (σ)}_z.    (S31)

This leads to the Langevin equation

{circumflex over ({dot over (a)})}=−i½g_z{circumflex over (σ)}_z−(iχ÷½κ)â−√κâ_in.   (S32)

In accordance with previously presented results in (A), the measurement signal is

$\begin{array}{cc}{〈\hat{M}〉}_1-{〈\hat{M}〉}_0=4{\tilde{g}}_z\sin \left(\varphi -{\phi }_h\right)\tau {\cos }^2\left(\frac{1}{2}{\varphi }_{\mathrm{gb}}\right)\left\{1-\frac{2}{\mathrm{\kappa \tau }}\left[\cos \left({\varphi }_{\mathrm{gb}}\right)-\cos \left({\varphi }_{\mathrm{qb}}+\mathrm{\chi \tau }\right)e^{-\frac{1}{2}\mathrm{\kappa \tau }}\right]\right\}.& \left(\mathrm{S33}\right)\end{array}$

where as before it is noted that the dispersive-coupling-induced rotation φ_qb=2 arctan (2χ/κ). Again as above, the measurement noise is not changed by the dispersive shift. Choosing

$\varphi -{\phi }_h=\frac{\pi }{2}\mathrm{mod}\pi ,$

the SNR finally reads

$\begin{array}{cc}\mathrm{SNR}\left(\chi \right)=\sqrt{8}\frac{{\tilde{g}}_z}{\kappa }\sqrt{\mathrm{\kappa \tau }}{\cos }^2\left(\frac{1}{2}{\varphi }_{\mathrm{qb}}\right)\left\{1-\frac{2}{\mathrm{\kappa \tau }}\left[\cos \left({\varphi }_{\mathrm{qb}}\right)-\cos \left({\varphi }_{\mathrm{qb}}+\mathrm{\chi \tau }\right)e^{-\frac{1}{2}\mathrm{\kappa \tau }}\right]\right\}.& \left(\mathrm{S34}\right)\end{array}$

The residual dispersive coupling reduces the value of the SNR, with the decrease behaving differently at long and short measurement times. At long measurement times, the dispersive coupling reduces the SNR by

$\begin{array}{cc}\mathrm{SNR}\left(\chi \right)\simeq {\cos }^2\left(\frac{1}{2}{\varphi }_{\mathrm{qb}}\right)\mathrm{SNR}\left(\chi =0\right)=\frac{{\kappa }^2}{{\kappa }^2+4{\chi }^2}\mathrm{SNR}\left(\chi =0\right),\mathrm{for}\tau 1/\kappa .& \left(\mathrm{S35}\right)\end{array}$

The SNR is not affected for χ«κ/2. Interestingly, at short measurement times the SNR is completely independent of the spurious dispersive shift to leading orders

$\begin{array}{cc}\mathrm{SNR}\left(\chi \right)\simeq \frac{1}{\sqrt{2}}\frac{{\tilde{g}}_z}{\kappa }{\left(\kappa \tau \right)}^{3/2}\left(1-\frac{1}{6}\kappa \tau \right),\mathrm{for}\tau 1/\kappa .& \left(S36\right)\end{array}$

In short, the SNR is not affected by a spurious transverse coupling for short measurement times τ«1/κ.

II. Circuit QED Realization

An exemplary realization of longitudinal coupling in circuit QED is now addressed. While a lumped circuit QED implementation 700 is presented in FIG. 7, focus is on put a transmon qubit that is phase-biased by a coplanar waveguide resonator. The lumped element results are recovered in the appropriate limit. Emphasis is put on numerical results obtained for this coplanar realization. For this reason these numerical values differ from, but are compatible with, what is found with reference to FIGS. 1 to 9.

As illustrated in FIG. 10, a transmon qubit (110) coupled at middle of the center conductor of a λ/2 resonator (109). To increase the longitudinal coupling strength, a Josephson junction can be inserted in the center conductor of the resonator at the location of the qubit (111). FIG. 10 shows a logical representation of physical characteristics of a transmon qubit coupling longitudinally to a resonator. The coplanar waveguide resonator is composed of a central electrode (109) surrounded by a ground plane (112). A coupling inductance (111) mediates the longitudinal coupling to the transmon qubit (110). A capacitively coupled transmission line (113) allows to send and retrieve input and output signals to and from the resonator. FIG. 11 shows a close-up of the transmon qubit with the definitions of the branch fluxes used herein. The qubit is composed of nominally identical Josephson junctions (114) and large shunting capacitances (115). Additional control lines are needed to modulate the flux through the loop φ_x and to perform single-qubit operations (not shown on FIG. 10 and FIG. 11). Ultrastrong transverse coupling of a flux qubit to a resonator has been presented by Bourassa, J. et. al. in “Ultrastrong coupling regime of cavity QED with phase-biased flux qubits”, which involve at least some inventors also involved in an invention claimed herein. The modelling of the present circuit closely follows Bourassa, J. et. al. and relevant details are provided herein for completeness.

The Lagrangian of this circuit, =r+q+qr, is composed of three parts consisting of the bare resonator r, qubit q and interaction qr Lagrangians. From standard quantum circuit theory, the resonator Lagrangian takes the form

$\begin{array}{cc}{\left(\frac{2\pi }{{\Phi }_0}\right)}^2{ℒ}_r={\int }_{-L/2}^{L/2}\left(\frac{C^0}{2}{\stackrel{.}{\psi }}^2\left(x,t\right)-\frac{1}{2L^0}{\left({\partial }_x\psi \left(x,t\right)\right)}^2\right)\mathrm{dx}+{\left(\frac{2\pi }{{\Phi }_0}\right)}^2E_{\mathrm{Jr}}\cos \left[\Delta \psi \right]+\frac{C_{\mathrm{Jr}}}{2}\stackrel{.}{\Delta }\psi .& \left(S37\right)\end{array}$

where ψ(x) is the position-dependant field amplitude inside the resonator and Δψ=ψ(x_a+Δx/2)−ψ(x_a+Δx/2) is the phase bias across the junction of width Δx at position x_a. In this expression, it is assumed that the resonator has total length L with capacitance C⁰ and inductance L⁰ per unit length. In the single mode limit, ψ(x, t)=ψ(t)u(x) where u(x) is the mode envelope. The Josephson junction in the resonator's center conductor has energy E_Jr and capacitance C_Jr. This junction creates a discontinuity Δψ≈0 in the resonator field that will provide the desired longitudinal interaction. The coupling inductance can be replaced by a SQUID, or SQUID array, without significant change to the treatment. This Lagrangian was already studied in the context of strong transverse coupling flux qubits to transmission-line resonators and for non-linear resonators.

The transmon qubit is composed of a capacitor to ground C_b and two capacitively shunted Josephson junctions of energies E_J1 and E_J2, and total capacitances C_q1=C_j1+C_s1 and C_q2=C_j2+C_s2 respectively. In terms of the branch fluxes defined on FIG. 11, the qubit Lagrangian taking into account the coupling to the resonator is

$\begin{array}{cc}{ℒ}_q+{ℒ}_{\mathrm{qr}}={\left(\frac{{\Phi }_0}{2\pi }\right)}^2\left[\frac{C_{q1}}{2}{\left({\stackrel{.}{\psi }}_1-\stackrel{.}{\phi }-{\stackrel{.}{\Phi }}_x\right)}^2+\frac{C_{q2}}{2}{\left({\stackrel{.}{\psi }}_2-\stackrel{.}{\phi }\right)}^2+\frac{C_b}{2}{\stackrel{.}{\phi }}^2\right]+E_{J1}\cos \left[{\psi }_1-{\phi }_1-{\Phi }_x\right]+E_{J2}\cos \left[{\phi }_2\right]& \left(S38\right)\end{array}$

Here ψ₁₍₂₎=ψ(x_a∓Δ_x/2) is defined for simplicity and the qubit capacitances C_q=C_j+C_s for each arm. Defining new variables θ=(ψ₁+ψ₂−2φ)/2 and δ=(ψ₁+ψ₂+2φ)/2, the above is obtained

$\begin{array}{cc}{ℒ}_q+{ℒ}_{\mathrm{qr}}={\left(\frac{{\Phi }_0}{2\pi }\right)}^2\hspace{1em}\left[\frac{C_{q1}+C_{q2}}{2}\left({\stackrel{.}{\theta }}^2+\stackrel{.}{\Delta }{\psi }^2/4\right)+\frac{C_b}{8}{\left(\stackrel{.}{\delta }-\stackrel{.}{\theta }\right)}^2+\frac{\left(C_{q1}-C_{q2}\right)}{2}\stackrel{.}{\theta }\stackrel{.}{\Delta }\psi \right]+E_{J1}\cos \left[\theta -\Delta \psi /2+{\Phi }_x\right]+E_{J2}\cos \left[\theta +\Delta \psi /2\right].& \left(S39\right)\end{array}$

Shifting the variable θ→θ+φ_x/2, a more symmetrical Lagrangian is obtained with respect to the external flux

$\begin{array}{cc}{ℒ}_q+{ℒ}_{\mathrm{qr}}={\left(\frac{{\Phi }_0}{2\pi }\right)}^2\left[\frac{C_{q1}+C_{q2}}{2}\left({\stackrel{.}{\theta }}^2+\frac{\Delta {\stackrel{.}{\psi }}^2+{\stackrel{.}{\Phi }}_x^2}{4}\right)+\frac{C_b}{8}{\left(\stackrel{.}{\delta }-\stackrel{.}{\theta }-{\stackrel{.}{\Phi }}_x/2\right)}^2+\frac{\left(C_{q1}-C_{q2}\right)}{2}\stackrel{.}{\theta }\left({\stackrel{.}{\Phi }}_x-\stackrel{.}{\Delta }\psi \right)+\frac{C_{q1}+C_{q2}}{4}\stackrel{.}{\Delta }\psi {\stackrel{.}{\Phi }}_x\right]+E_{J1}\cos \left[\theta -\Delta \psi /2-{\Phi }_x/2\right]+E_{J2}\cos \left[\theta +\Delta \psi /2+{\Phi }_x/2\right].& \left(S40\right)\end{array}$

The equation of motion for δ being {umlaut over (δ)}−{umlaut over (θ)}+{umlaut over (Φ)}_x/2=0, the term involving the capacitance to ground C_b is thus a constant and can be ignored. Assuming the resonator phase-bias to be small Δψ«1 and a DC-flux bias ({dot over (Φ)}_r=0), we find to zeroth-order in Δψ the usual Lagrangian of an asymmetric fluxbiased transmon qubit

$\begin{array}{cc}{ℒ}_q={\left(\frac{{\Phi }_0}{2\pi }\right)}^2\frac{C_{q1}+C_{q2}}{2}{\stackrel{.}{\theta }}^2+E_{J\Sigma }\left[\cos \left({\Phi }_x/2\right)\cos \theta -d\sin \left({\Phi }_x/2\right)\sin \theta \right],& \left(S41\right)\end{array}$

where E_JΣ=E_J1+E_J2 and d=(E_J2−E_J1)/E_JΣ is the junction asymmetry. Defining {circumflex over (n)} to be the conjugate charge to {circumflex over (θ)}, the corresponding Hamiltonian is

Ĥ _q =ΛE _C{circumflex over (n)}²−E_JΣ[cos (Φ_x/2) cos {circumflex over (θ)}−d sin (Φ_x/2) sin {circumflex over (θ)}].   (S12)

with the charging energy E_C=e²/2(C_q1+C_q2).

For clarity, projection is made on the qubit subspace {|0>, |1>} where the total Hamiltonian takes the form

$\begin{array}{cc}\hat{H}={\omega }_r{\hat{a}}^{†}\hat{a}+{K\left({\hat{a}}^{†}\hat{a}\right)}^2+\frac{{\omega }_a}{2}{\hat{\sigma }}_z+{\hat{H}}_{\mathrm{qr}}.& \left(S43\right)\end{array}$

In this expression, ω_a is the qubit transition frequency and K the Kerr non-linearity. The latter can be made small and in particular negligible with respect to the photon decay rate κ. To first order in Δ{circumflex over (ψ)} and in the same two-level approximation, the interaction Hamiltonian reads

$\begin{array}{cc}\begin{array}{c}{\hat{H}}_{\mathrm{qr}}=E_{J\Sigma }\frac{\Delta \hat{\psi }}{2}\left[\sin \left({\Phi }_x/2\right)\cos \hat{\theta }+d\cos \left({\Phi }_x/2\right)\sin \hat{\theta }\right]\\ =g_z\left({\hat{a}}^{†}+\hat{a}\right){\hat{\sigma }}_z+g_x\left({\hat{a}}^{†}+\hat{a}\right){\hat{\sigma }}_x,\end{array}& \left(S44\right)\end{array}$

where

$\begin{array}{cc}{g}_z\left({\Phi }_x\right)=-{\psi }_{\mathrm{rms}}\Delta u\frac{E_{J\Sigma }}{4}\left[m_{11}\left({\Phi }_x\right)-m_{00}\left({\Phi }_x\right)\right],& \left(S45\right)\\ g_x\left({\Phi }_x\right)={\psi }_{\mathrm{rms}}\Delta u\frac{E_{J\Sigma }}{2}m_{01}\left({\Phi }_x\right).& \left(S46\right)\end{array}$

To obtain these expressions, Δ{circumflex over (ψ)}=Δuψ_rms(â^†+â) is used, with ψ_rms=√4πZ_r/R_K. Here Z_r is the resonator mode impedance and R_K=h/e² the quantum of resistance. The mode gap Δu is found to be optimal for a junction placed in the center of the resonator center (see Bourassa, J. et. al.). At that location, Δu can be related to the participation ratio of the coupling inductance L_J⁻¹=E_Jr(2π/φ₀)² to the effective inductance of the resonator mode L_r as Δu≈2η=2L_J/L_r The factor of two originates from the fact that both resonator halves bias equally the qubit. The lumped element limit discussed in the previous discussion is obtained for η→1/2. Moreover, it has been defined that m_ij=i| sin (Φ_x/2) cos {circumflex over (θ)}+d cos (Φ_x/2) sin {circumflex over (θ)}|j. Equations (S46) and (S45) are calculated numerically by diagonalizing the transmon Hamiltonian.

B. Numerical evaluation of the coupling strength

A set of possible parameters is presented for this circuit. It is very important to emphasize that these numbers relate to the coplanar architecture discussed here and not to the lumped-element example described in relation to FIGS. 7 to 9. A λ/2 resonator is used of characteristic impedance Z⁰˜50Ω, total length 7.6 mm and capacitance per unit length C⁰=0.111 nF/m and inductance per unit length L⁰=0.278 μH/m. The coupling inductance consists of an array of 13 Josephson Junctions with a total Josephson energy E_Jr/h=420 GHz. Following the standard procedure, the first resonator mode is determined to be at frequency ω_r/2π=10.02 GHz, with a characteristic mode impedance Z_r=18.8Ω. The participation ratio of the array is found to be η˜0.3. With these parameters, the resonator bias on the qubit is ψ_rms≈0.096.

The transmon Josephson and charging energies are E_JΣ/h=20 GHz and E_c/h=0.3 GHz respectively, and the asymmetry d=0.02. These parameters yield ω_a/2π=6.6 GHz when evaluated at the flux sweet-spot Φ_x=0. The magnetic flux is modulated around this flux value with a maximum excursion of Φ_z/2π=0.1. At the sweet-spot, the qubit-resonator is Δ=ω_n(0)−ω_r≈2π×3.4 GHz The longitudinal coupling {tilde over (g)}_z the various spurious interactions found numerically using these numbers are summarized in Table I, which provides longitudinal coupling rate {tilde over (g)}_z and spurious leading-order and second-order couplings for a transmon coupled to a λ/2 resonator with a coupling inductance consisting of an array of 13 Josephson junctions as discussed herein above with reference to the coplanar architecture (i.e., not in relation to the lumped element version discussion in FIGS. 7 to 9).

		Rates
Couplings		(2π × MHz)
Longitudinal [Eq. (S48)]	{tilde over (g)}z	26.3
Transverse [Eq. (S49)]	gx	7.89
Dispersive	gx2/Δ	0.018
Non-linear dispersive [Eq. (S56)]	2Λz	3.97
Non-linear transverse [Eq. (S57)]	Λ x	0
Kerr nonlinearity	K	0.007
Qubit frequency shift (from flux-drive) [Eq. (S54)]	ϵq2/Δ	4.4 × 10−5

In particular, {tilde over (g)}_z/2π=26.3 MHz and a small transverse coupling resulting in a negligible dispersive shift χ_x=g_x²/Δ˜18 kHz. The maximum change in qubit frequency is

ω_a(0)−ω_a({tilde over (Φ)})˜2π×170 MHz.

The array of N junctions allows for strong longitudinal coupling while reducing the resonator nonlinearity. Compared to a single junction of equal energy, the non-linear Kerr effectively decreases as K=K₁/N² and from the parameters above K/2π=7 kHz using 13 junctions is obtained.

C. Asymptotic expression for the coupling strengths

To obtain asymptotic expressions for the qubit-resonator coupling, the transmon is considered as a weakly anharmonic oscillator. In this situation,

$\begin{array}{cc}\tilde{\theta }\approx {\left[\frac{2E_C}{E_J\left({\Phi }_s\right)}\right]}^{1/4}\left({\hat{b}}^{†}+\hat{b}\right)\mathrm{and}n_{i,j\ne i}\approx {d\left(\frac{2E_C}{E_J\left({\Phi }_x\right)}\right)}^{1/4}\cos \left({\Phi }_x/2\right)〈i\left({\hat{b}}^{†}+\hat{b}\right)j〉,m_{\mathrm{ii}}\approx -{\left(\frac{2E_C}{E_J\left({\Phi }_x\right)}\right)}^{1/2}\frac{\sin \left({\Phi }_x/2\right)}{2}〈i{\left({\hat{b}}^{†}+\hat{b}\right)}^2i〉& \left(S47\right)\end{array}$

Restricting to the {|0, |1} subspace and using Equations (S46) and (S45), the asymptotic expressions obtained are

$\begin{array}{cc}{g}_z\approx -{\frac{E_{J\Sigma }}{2}\left[\frac{2E_C}{E_J\left({\Phi }_x\right)}\right]}^{1/2}\sqrt{\frac{\pi Z_r}{R_K}}\sin \left({\Phi }_x/2\right)2\eta ,& \left(S48\right)\\ g_x\approx {{\mathrm{dE}}_{J\Sigma }\left[\frac{2E_C}{E_J\left({\Phi }_x\right)}\right]}^{1/4}\sqrt{\frac{\pi Z_r}{R_K}}\cos \left({\Phi }_x/2\right)2\eta .& \left(S49\right)\end{array}$

These correspond to Equations 5 and 6 previously discussed where η→1/2.

D. Upper bounds on the coupling strength

An upper bounds for the longitudinal coupling is obtained by expressing g_z in units of the qubit frequency. Using Equation S48:

$\begin{array}{cc}\frac{g_z}{{\omega }_a}\approx \frac{{\Phi }_{x}}{8}\sqrt{\frac{\pi Z_r}{R_K}}2\eta .& \left(S50\right)\end{array}$

For Z_r˜50Ω resonator, maximal coupling is reached in the lumped-element limit of the oscillator where η→1/2 and we get g_z/ω_a˜0.006, or about twice the coupling obtained in the previous example giving g_z/2π˜42.4 MHz.

Even larger values of g_z can be achieved by using lumped LC-circuit comprises of a superinductance of large impedance Z_r˜R_K/4. For participation ratios in the range 2η˜[10⁻², 1], the coupling is enhanced by a factor of ˜10 to g_z/ω_a=Φ×η√π/8˜[7×10⁻⁴, 7×10⁻²]. While the previous circuit model would have to be redefined to take the large bias Δψ into account, it is safe to say that the larger and more compact coupling inductance is, the stronger the longitudinal coupling will be.

III. Spurious Couplings and Imperfections

From the first term in Equation (S39), an oscillating external magnetic flux Φ_x(t)={tilde over (Φ)}_x cos (ω_rt+φ) at frequency ω_d leads to an effective voltage driver on the resonator

$\begin{array}{cc}{ℒ}_{r.\mathrm{drive}}={\left(\frac{{\Phi }_0}{2\pi }\right)}^2\frac{C_{q1}+C_{q2}}{2}{\stackrel{.}{\Phi }}_x\stackrel{.}{\psi }\left(0\right).& \left(S51\right)\\ {ϵ}_r=\frac{h{\omega }_r}{16E_{\mathrm{CJ}}}{\psi }_{\mathrm{rms}}{\Phi }_x{\omega }_d,& \left(S52\right)\end{array}$

This term leads to an effective driver on the resonator Ĥ_r,d=ϵ_re^{−i(ωrt+φ)}â^†+H.c. of drive amplitude ϵ_r

where E_CJ=e²/[2(C_q1+C_q2)]. With the above circuit, parameters, ϵ_r/2π˜5 MHz is obtained. If desired, the effect of this drive can be cancelled by an additional drive on the input port of the resonator. Otherwise, this simply leads to an additional qbit-state independent displacement of the cavity field that does not affect the SNR.

In the same way, flux modulation also directly drives the qubit. This is caused by the last charging energy term of Equation (S39) and yields

$\begin{array}{cc}{\hat{H}}_{q,d}={\left(\frac{{\Phi }_0}{2\pi }\right)}^2\frac{C_{q1}-C_{q2}}{2}{\stackrel{.}{\Phi }}_x\hat{q},& \left(S53\right)\end{array}$

From the asymptotic expression {circumflex over (q)}≈i(E_J/32E_C)^1/4(b^†−b), Ĥ_q,d=ϵ_qe^−k(ωrt+φ){circumflex over (b)}^†+H.C. is obtained where

$\begin{array}{cc}{ϵ}_q\approx \frac{d}{8}\frac{C_J}{C_{\Sigma }}{\left(\frac{E_J}{32E_C}\right)}^{1/4}{\Phi }_x{\omega }_r& \left(S54\right)\end{array}$

is identically zero for symmetric junctions (d=0). With the above parameters and C_J/C_Σ=0.01, then ϵ_q/2π˜0.18 MHz. Given the strong qubit-resonator detuning of several GHz, this is however of no consequences.

B. Higher-order interaction terms

To second order in Δψ in Equation (S39) yields the additional interactions

Ĥ _qr ⁽²⁾=(â^†+â)²(Λ_x{circumflex over (σ)}_x+Λ_z{circumflex over (σ)}_z).   (S55)

where

$\begin{array}{cc}{\Lambda }_z={\psi }_{\mathrm{rms}}^2\frac{E_{J\sum }}{16}\left(〈0\widehat{\mathrm{cc}}-d\widehat{\mathrm{ss}}[0〉-〈1\widehat{\mathrm{cc}}-d\widehat{\mathrm{ss}}1〉\right),& \left(\mathrm{S56}\right)\\ {\Lambda }_x={\psi }_{\mathrm{rms}}^2\frac{E_{J\sum }}{8}〈1\left[\widehat{\mathrm{cc}}-d\widehat{\mathrm{ss}}\right]0〉.& \left(\mathrm{S57}\right)\end{array}$

and where have been defined ĉc=cos (Φ_x/2) cos {circumflex over (θ)} and ŝs=sin (Φ_x/2) sin {circumflex over (θ)}.

Under flux modulation, Λ_k(t)=Λ_k+{tilde over (Λ)}_k cos (ω_rt+φ_r) with k=x,z. With the RWA this leads to

Ĥ _qr ⁽²⁾≈(2â^†â+1)(Λ_x{circumflex over (σ)}_x+Λ_x{circumflex over (σ)}_z).   (S58)

For d small and flux modulations around Φ_x=0 then Λ_x=0. On the other hand, the dispersive-like interaction of amplitude χ_z=2Λ_z has a magnitude of χ_z/2π≈3.97 MHz for the above circuit parameters. In the lumped-element limit with Z_r=50Ω, the result is χ_z/2π=5.3 MHz. Since this dispersive-like interaction originates from a second-order correction, it is always found that g_z/χ_z≈10. In practice, choosing χ_z<κ will allow this spurious coupling will not affect the SNR at short integration times.

IV. Multiple Qubit Read-Out in (X+Z) Circuit QED

FIG. 12 illustrates a possible multi-qubit architecture: qubits (110) are longitudinally coupled to a readout resonator (109) and transversally coupled to a bus resonator (120). The readout resonator is used only when the longitudinal couplings are modulated and individual flux control for each qubit is possible with separated flux lines. The system Hamiltonian reads

$\begin{array}{cc}\hat{H}={\omega }_{\mathrm{rz}}a_z^{†}{\hat{a}}_z+{\omega }_{\mathrm{rx}}{\hat{a}}_x^{†}{\hat{a}}_x+\sum _jg_{\mathrm{zj}}{\hat{\sigma }}_{\mathrm{zj}}\left({\hat{a}}_z^{†}+{\hat{a}}_z\right)+\sum _jg_{\mathrm{xj}}{\hat{\sigma }}_{\mathrm{xj}}\left({\hat{a}}_x^{†}+{\hat{a}}_x\right).& \left(\mathrm{S59}\right)\end{array}$

In the interaction picture and neglecting fast-oscillating terms, the longitudinal coupling becomes

$\begin{array}{cc}{\tilde{H}}_z=\left(\frac{1}{2}{\tilde{g}}_z\sum _j{\hat{\sigma }}_{\mathrm{zj}}e^{i_{\varphi ,i}}\right){\hat{a}}_z+H.c.& \left(\mathrm{S60}\right)\end{array}$

The position of the qubits and the coupling inductances can be adjusted to get equal longitudinal coupling strengths |g_zj|. By properly choosing the phases φ_j, it is possible to operate in the joint qubit readout mode or in the entanglement by measurement mode. Examples for 2 and 3 qubits are shown in FIGS. 13A, 13B, 13C and 13D (joint measurement of two qubits, synthesis of Bell states and 3-qubit GHZ state by measurement).

FIG. 12 shows a multi-qubit architecture consisting of qubits (110) longitudinally coupled to a readout resonator (109) and transversally coupled to a bus resonator (120) by way of bus-resonator fingers (119) capacitively coupled to the qubits. As in FIG. 10, the read-out resonator is coupled to an input transmission-line (113) but also to and output transmission-line (114). Similarly, the bus resonator is connected to both an input (116) and output (117) transmission-line for control. The ground plane (112) is placed on both sides of the resonators. Between the qubits (110), the pieces of ground planes are interconnected with air bridges (118). Flux control lines and qubit controls lines for each qubit are absent in this figure. FIGS. 13A, 13B, 13C and 13D show phase space representation of the pointer states for two and three qubits. On FIG. 13A, two-qubit joint readout (φ_j=jπ/2). On FIGS. 13B and 13C, two-qubit entanglement by measurement is depicted. The double-hatched circles centered at the origin circles correspond to the Bell states

${\Psi }^{-}〉=\frac{1}{\sqrt{2}}\left[0,1〉+1,0〉\right]$

and

${\Phi }^{-}〉=\frac{1}{\sqrt{2}}\left[0,0〉+1,1〉\right],$

obtained for φ_j=0 and φ_j=jπ, respectively. On FIG. 13D, the centered at the origin circle corresponds to the GHZ state

$\mathrm{GHZ}〉=\frac{1}{\sqrt{2}}\left[0,0,0〉+1,1,1〉\right].$

This configuration is obtained for φ_j=j2π/3

V. Qubit Dephasing

When not measuring the qubit, coupling to the resonator is turned off and the qubit is not dephased by photon shot noise or residual thermal population. In the superconducting qubit architecture studied in the previous section, another potential source of dephasing is flux noise. The influence of low-frequency flux noise on the transmon was already studied. Since the qubit is operated at its flux sweet spot, the effect of this noise can be negligible.

More interesting here is the contribution opened by the longitudinal coupling, that is flux noise at the resonator frequency which will effectively ‘measure’ the qubit. Following the standard procedure and using Equation (5) previously defined, in the presence of flux noise, longitudinal coupling leads to a decay of the qubit's off-diagonal density matrix element with

$\begin{array}{cc}{\rho }_{01}\left(t\right)={\rho }_{01}\left(0\right)e^{i{\omega }_nt}\exp \left\{\frac{-{\lambda }^2}{2}\frac{\omega }{2\pi }S_{\Phi }\left(\omega +{\omega }_r\right)\left[\left(2\stackrel{\_}{n}+1\right)\frac{{\sin }^2\left(\omega t/2\right)}{{\omega }^2}+2\stackrel{\_}{m}\cos \left({\omega }_rt-\theta \right)\frac{\sin \left(\omega t/2\right)\sin \left(\left(\omega +2{\omega }_r\right)t/2\right)}{\omega \left(\omega +2{\omega }_r\right)}\right]\right\}.& \left(S61\right)\end{array}$

In this expression, S₁₀₁ (ω) is the spectral density of flux noise, n=â^†â, me^iθ=â² and

$\begin{array}{cc}\lambda =\frac{\partial g_z}{\partial {\Phi }_x}{|}_{{\Phi }_x=0}=-\frac{\pi }{{\Phi }_0}\frac{E_J}{2}{\left(\frac{2E_C}{E_J}\right)}^{1/2}\sqrt{\frac{\pi Z_0}{R_K}}.& \left(\mathrm{S62}\right)\end{array}$

The term inρ₀₁(t) not proportional to n or m takes the standard form, with the crucial difference that here the noise spectral density is evaluated at ω+ω_r rather than at ω. With

S_Φ(ω)=2πA²/|ω|^a even at large frequencies, dephasing caused by longitudinal coupling is therefore expected to be negligible. In practice, the terms proportional to n or m are expected to contribute only in the situation where the intra-cavity field would be squeezed prior to the qubit measurement. In this situation, n=sin h²r and m=−sin h(2r)/2 with r the squeeze parameter.

Although decay is not exponential, the dephasing time can be estimated from the above expression by using the 1/e threshold as an estimate. For this an infrared cutoff must be introduced. Fortunately, the end result is only weakly dependent on this cutoff. For simplicity α=1 is taken in the noise spectral density with A=10⁻⁵Φ₀. Using the parameters given in the text, in the absence of squeezing but with a large spurious photon number n=0.1 a very large dephasing time of ˜10² seconds is found. For a squeezed state of 20 dB a dephasing time larger than a second is found. This mechanism is clearly not limiting the qubit dephasing time.

A method is generally conceived to be a self-consistent sequence of steps leading to a desired result. These steps require physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic/ electromagnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It is convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, parameters, items, elements, objects, symbols, characters, terms, numbers, or the like. It should be noted, however, that all of these terms and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities. The description of the present invention has been presented for purposes of illustration but is not intended to be exhaustive or limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments were chosen to explain the principles of the invention and its practical applications and to enable others of ordinary skill in the art to understand the invention in order to implement various embodiments with various modifications as might be suited to other contemplated uses.

• • We investigate an approach to universal quantum computation based on the modulation of longitudinal qubit-oscillator coupling. We show how to realize a controlled-phase gate by simultaneously modulating the longitudinal coupling of two qubits to a common oscillator mode. In contrast to the more familiar transversal qubit-oscillator coupling, the magnitude of the effective qubit-qubit interaction does not rely on a small perturbative parameter. As a result, this effective interaction strength can be made large, leading to short gate times and high gate fidelities. We moreover show how the gate fidelity can be exponentially improved with squeezing and how the entangling gate can be generalized to qubits coupled to separate oscillators. Our proposal can be realized in multiple physical platforms for quantum computing, including superconducting and spin qubits.

Introduction.—A widespread strategy for quantum information processing is to couple the dipole moment of multiple qubits to common oscillator modes, the latter being used to measure the qubits and to mediate long-range interactions. Realizations of this idea are found in Rydberg atoms, superconducting qubits and quantum dots amongst others. With the dipole moment operator being off-diagonal in the qubit's eigenbasis, this type of transversal qubit-oscillator coupling leads to hybridization of the qubit and oscillator degrees of freedom. In turn, this results in qubit Purcell decay and to qubit readout that is not truly quantum non-demolition (QND). To minimize these problems, the qubit can be operated at a frequency detuning from the oscillator that is large with respect to the transverse coupling strength g_x. This interaction then only acts perturbatively, taking a dispersive character. While it has advantages, this perturbative character also results in slow oscillator-mediated qubit entangling gates.

Rather than relying on the standard transversal coupling, H_x=g_z(â^†+â){circumflex over (σ)}_x, an alternative approach is to use a longitudinal interaction, H_z=g_z(â^†+â){circumflex over (σ)}_z. Since H_z commutes with the qubit's bare Hamiltonian the qubit is not dressed by the oscillator. Purcell decay is therefore absent and qubit readout is truly QND. The absence of qubit dressing also allows for scaling up to a lattice of arbitrary size with strictly local interactions.

By itself, longitudinal interaction however only leads to a vanishingly small qubit state-dependent displacement of the oscillator field of amplitude g_z/ω_r«1, with ω_r the oscillator frequency. In Paper A, it was shown that modulating g_z at the oscillator frequency ω_r activates this interaction leading to a large qubit state-dependent oscillator displacement and to fast QND qubit readout. In this Letter, we show how the same approach can be used, together with single qubit rotations, for universal quantum computing by introducing a fast and high-fidelity controlled-phase gate based on longitudinal coupling. The two-qubit logical operation relies on parametric modulation of a longitudinal qubit-oscillator coupling, inducing an effective {circumflex over (σ)}_z{circumflex over (σ)}_z interaction between qubits coupled to the same mode. We show that, with a purely longitudinal interaction, the gate fidelity can be improved exponentially using squeezing, and that the gate can be performed remotely on qubits coupled to separate but interacting oscillators. The latter allows for a modular architecture that relaxes design constraints and avoids spurious interactions while maintaining minimal circuit complexity.

In contrast to two-qubit gates based on a transversal interaction, this proposal does not rely on strong qubit-oscillator detuning and is not based on a perturbative argument. As a result, the longitudinally mediated {circumflex over (σ)}_z{circumflex over (σ)}_z interaction is valid for all qubit, oscillator and modulation parameters and does not result in unwanted residual terms in the Hamiltonian. For this reason, in the ideal case where the interaction is purely longitudinal (i.e. described by H_z), there are no fundamental bounds on gate infidelity or gate time and both can in principle be made arbitrarily small simultaneously.

Similarly to other oscillator-mediated gates, loss from the oscillator during the gate leads to gate infidelity. This can be minimized by working with high-Q oscillators something that is, however, in contradiction with the requirements for fast qubit readout. We solve this dilemma by exploiting quantum bath engineering, using squeezing at the oscillator input. By appropriately choosing the squeezed quadrature, we show how ‘which-qubit-state’ information carried by the photons leaving the oscillator can be erased. This leads to an exponential improvement in gate fidelity with squeezing strength.

Oscillator mediated qubit-qubit interaction.—Following Paper A, we consider two qubits coupled to a single harmonic mode via their {circumflex over (σ)}_z degree of freedom. Allowing for a time-dependent coupling, the Hamiltonian reads (h=1)

$\begin{array}{cc}\hat{H}\left(t\right)={\omega }_r{\hat{a}}^{†}\hat{a}+\frac{1}{2}{\omega }_{a1}{\hat{\sigma }}_{z1}+\frac{1}{2}{\omega }_{a2}{\hat{\sigma }}_{z2}+g_1\left(t\right){\hat{\sigma }}_{z1}\left({\hat{a}}^{†}+\hat{a}\right)+g_2\left(t\right){\hat{\sigma }}_{z2}\left({\hat{a}}^t+\hat{a}\right).& \left(1\right)\end{array}$

In this expression, ω_r and ω_ai are the frequencies of the oscillator and of the i^th qubit, respectively, while g₂(t) are the corresponding longitudinal coupling strengths.

In the absence of external drives or of modulation of the coupling, the longitudinal interaction only leads to a vanishingly small displacement of the oscillator field and can safely be dropped. This interaction can be rendered resonant by modulating g_i(t) at the oscillator frequency leading to a large qubit-state dependent displacement of the oscillator state. Measurement of the oscillator by homodyne detection can then be used for fast QND qubit readout. Consequently, modulating the coupling at the oscillator frequency rapidly dephases the qubits. To keep dephasing to a minimum, we instead use an off-resonant modulation of g₂(t) at a frequency ω_m detuned from ω_r by many oscillator linewidths g₂(t)=g_i cos(ω_mt), where g_1,2 are constant real amplitudes.

The oscillator-mediated qubit-qubit interaction can be made more apparent by applying a polaron transformation Û(t)=exp[Σ_i=1,2α_i(t){circumflex over (σ)}_ziâ^†—H.c.] with an appropriate choice of α_i(t). Doing this, we find in the polaron frame the simple Hamiltonian

Ĥ _pol(t)=ω_râ^†â+J_z(t){circumflex over (σ)}_z1{circumflex over (σ)}_z2.   (2)

The full expression for the {circumflex over (σ)}_z{circumflex over (σ)}_z-coupling strength J_z(t) is given “Supplemental Material for Fast and High-Fidelity Entangling Gate through Parametrically Modulated Longitudinal Coupling” , which we will henceforth refer to as SM B. In the following we will, however, assume two conditions on the total gate time, t_g, such that this expression simplifies greatly. For δt_g=n×2π and ω_mt_g=m×π, with n and in integers, we can replace J_z(t) by

$\begin{array}{cc}{\stackrel{\_}{J}}_z=-\frac{g_1g_2}{2}\left[\frac{1}{\delta }+\frac{1}{{\omega }_r+{\omega }_m}\right],& \left(3\right)\end{array}$

where δ≡ω_r−ω_m is the modulation drive detuning.

By modulating the coupling for a time t_g=θ/4/J_z|, evolution under Eq. (2) followed by single qubit Z-rotations leads to the entangling controlled-phase gate U_CP(θ) diag[1, 1, 1, e^iθ] Since U_CP (π) together with single qubit rotations forms a universal set, we only consider this gate from now on.

Note that the conditions on the gate time used in Eq. (3) are not necessary for the validity of Eq. (2), and the gate can be realized without these assumptions. However, as we will discuss below, these conditions are important for optimal gate performance: They ensure that the oscillator starts and ends in the vacuum state, which implies that the gate does not need to be performed adiabatically. Finally, not imposing the second constraint, ω_mt_g=m×π, only introduces fast rotating terms to Eq. (3) which we find to have negligible effect for the parameters used later in this Letter. In other words, this constraint can be ignored under a rotating-wave approximation.

The above situation superficially looks similar to controlled-phase gates based on transversal coupling and strong oscillator driving. There are, however, several key differences. With transversal coupling, the {circumflex over (σ)}_z{circumflex over (σ)}_z interaction is derived using perturbation theory and is thus only approximately valid for small g_x/{Δ, δ_d}, with Δ the qubit-oscillator detuning and δ_d the oscillator-drive detuning. For the same reason, it is also only valid for small photon numbers n«n_crit=Δ²/4g²_x. Moreover, this interaction is the result of a fourth order process in g_x/{Δ,δ_d}, leading to slow gates. Because of the breakdown of the dispersive approximation, any attempt to speed up the gate by decreasing the detunings or increasing the drive amplitude results in low gate fidelities. In contrast, with longitudinal coupling, the {circumflex over (σ)}_z{circumflex over (σ)}_z interaction is conveniently the result of a second-order process in g_1,2/δ. In addition, since Eq. (2) is obtained after exact transformations on the system Hamiltonian, it is valid for any value of g_1,2/δ and is independent of the oscillator photon number. As will become clear later, this implies that the gate time and the gate infidelity can be decreased simultaneously. Finally, with longitudinal coupling, there is no constraint on the qubit frequencies, in contrast with usual oscillator-induced phase gates where the detuning between qubits should preferably be small.

Oscillator-induced qubit dephasing.—FIG. 1 illustrates, for g₁=g₂, the mechanism responsible for the qubit-qubit interaction. Under longitudinal coupling, the oscillator field is displaced in a qubit-state dependent way, following the dashed lines in FIG. 1(a) (Panels (b) and (c) will be discussed later). This conditional displacement leads to a non-trivial qubit

FIG. 1. (color online) Schematic illustration, in a frame rotating at ω_r, of the qubit-state dependent oscillator field in phase space for g₁=g₂ starting and ending in the vacuum state. The oscillator's paths for 100 and 111 are shown in dashed lines. The qubit-state dependent oscillator state is shown in light (t=t_g/4) and dark greys (t=t_g/2). The oscillator's state associated to {101, 110} stays in the vacuum state for the duration of the gate. (a) No squeezing. (b,c) Squeezing can help in erasing the which-qubit-state information.

phase accumulation. This schematic illustration also emphasizes the main cause of gate infidelity for this type of controlled-phase gate, irrespective of its longitudinal or transversal nature: Photons leaking out from the oscillator during the gate carry information about the qubit state, leading to dephasing.

A quantitative understanding of the gate infidelity under photon loss can be obtained by deriving a master equation for the joint qubit-oscillator system. While general expressions are given in SM B. To simplify the discussion we assume here that g₁=g₂≡g. Following the standard approach, the Lindblad master equation in the polaron frame reads

$\begin{array}{cc}\stackrel{.}{\rho }\left(t\right)=-i\left[{\hat{H}}_{\mathrm{pol}},\rho \left(t\right)\right]+\mathrm{\kappa }\left[\hat{a}\right]\rho \left(t\right)+\Gamma \left[1-\cos \left(\delta t\right)\right]\left[{\hat{\sigma }}_{z1}+{\hat{\sigma }}_{z2}\right]\rho \left(t\right),& \left(4\right)\end{array}$

where κ is photon decay rate and [x] denotes the usual dissipation super-operator [x]•=x•x^†−1/2{x^†x, •}. The last term of Eq. (4) corresponds to a dephasing channel with rate Γ=2κ(g/2δ)². Since Ĥ_pol does not generate qubit-oscillator entanglement during the evolution, we can ensure that Γ is the only source of dephasing by imposing that the initial and final polaron transformations do not lead to qubit-oscillator entanglement. This translates to the condition α_i(0)=α_i(t_g)=0 and is realized for δt_g=n×2π, which is the constraint mentioned earlier (neglecting fast-rotating terms related to the second constraint ω_mt_g=m×π). More intuitively, it amounts to completing n full circles in FIG. 1, the oscillator ending back in its initial state. Note that these conclusions remain unchanged if

FIG. 2. (color online) Average gate infidelity 1- (full line) and gate time (dashed lines) of U_CP(π) as a function of (a) detuning, (b) coupling strength and (c) squeezing power. In panel (a) g/κ×10⁻³ is fixed at 2 (201), 3 (202), 4 (203). Note that the corresponding three infidelity curves are indistinguishable on this scale. In panel (b) δ/κ×10⁻⁵ is fixed at 0.75 (204), 1 (205), 1.25 (206). In panel (c) parameters are δ/2π=0.6 GHz, g/2π=60 MHz, t_g=42.7 ns, κ/2π=1 MHz. (207) rotating squeezing angle as illustrated in FIG. 1(b). (208) squeezing at ω_r as illustrated in FIG. 1(c), and κ(ω_m)=0 simulating a filter reducing the density of modes to zero at ω_m.

the oscillator is initially in a coherent state. As a result, there is no need for the oscillator to be empty at the start of the gate.

Based on the dephasing rate Γ and on the gate time t_g, a simple estimate for the scaling of the gate infidelity is 1-˜Γ×t_g˜κ/δ[1]. A key observation is that this gate error is independent of g, while the gate time scales as t_g˜δ/g². Both the gate time and the error can therefore, in principle, be made arbitrarily small simultaneously. This scaling of the gate error and gate time is confirmed by the numerical simulations of FIG. 2, which shows the dependence of the gate infidelity on detuning δ and coupling strength g, as obtained from numerical integration of Eq. (4). The expected increase in both fidelity (full lines) and gate time (dashed lines) with increasing detuning δ are apparent in panel (a). In addition, panel (b) confirms that, to a very good approximation, the fidelity is independent of g (full lines) while the gate time decreases as t_g˜1/g² (dashed lines).

This oscillator-induced phase gate can be realized in a wide range of physical platforms where longitudinal coupling is possible. Examples include spin qubits in inhomogeneous magnetic field, singlet-triplet spin qubits, flux qubits capacitively coupled to a resonator and transmon-based superconducting qubits. The parameters used in FIG. 2 have been chosen following the latter references. In particular, taking κ/2π=0.01 MHz, g/2π=60 MHz and δ/2π=537 MHz results in a very short gate time of t_g=37 ns with an average gate infidelity as small as 2×10⁻⁵. Taking into account finite qubit lifetimes T₁=30 μs and T₂=20 μs, we find that the infidelity is increased to ˜10⁻³. In other words, the gate fidelity is limited by the qubit's natural decoherence channels with these parameters.

A crucial feature of this gate is that the circular path followed by the oscillator field in phase space maximizes qubit-state dependent phase accumulation while minimizing de-phasing, allowing for high gate fidelities. Furthermore, we show below that this also allows for exponential improvement in gate fidelity with squeezing. It is therefore desirable to minimize, or avoid completely, dispersive coupling in experimental implementations [2].

Improved fidelity with squeezing—As discussed above, for fixed g and δ the fidelity increases with decreasing n. A small oscillator decay rate κ, however, comes at the price of longer measurement time if the same oscillator is to be used for readout. This problem can be solved by sending squeezed radiation to the oscillator's readout port. As schematically illustrated in FIG. 1, by orienting the squeezing axis with the direction of the qubit-dependent displacement of the oscillator state, the which-path information carried by the photons leaving the oscillator can be erased. By carefully choosing the squeezing angle and frequency, it is thus possible to improve the gate performance without reducing κ. We now show two different approaches to realize this, referring the reader to SM B for technical details.

A first approach is to send broadband two-mode squeezed vacuum at the input of the oscillator, where the squeezing source is defined by a pump frequency ω_p=(ω_r+ω_m)/2 and a squeezing spectrum with large degree of squeezing at ω_r and ω_m. A promising source of this type of squeezing is the recently developed Josephson travelling wave amplifiers. With such a squeezed input field, a coherent state of the oscillator becomes a squeezed state with a squeezing angle that rotates at a frequency δ/2. As illustrated in FIG. 1(b), this is precisely the situation where the anti-squeezed quadrature and the displacement of the oscillator's state are aligned at all times. This leads to an exponential decrease in dephasing rate

Γ(r)˜e^−2rΓ(0),   (5)

with r the squeezing parameter. This reduction in dephasing rate leads to the exponential improvement in gate fidelity with squeezing power shown by the brown line in FIG. 2(c). An interesting feature in this Figure is that increasing κ by 2 orders of magnitude to allow for fast measurement, leads to the same ˜10⁻⁵ gate infidelity obtained above without squeezing here using only ˜6 dB of squeezing. Since numerical simulations are intractable for large amount of squeezing, we depict the infidelity obtained from a master equation simulation by a solid line and the expected infidelity from analytical calculations by a dash-dotted line.

An alternative solution is to use broadband squeezing centered at the oscillator's frequency, i.e. a squeezing source defined by a pump frequency ω_p=ω_r. As illustrated in FIG. 1(c), using this type of input leads to a squeezing angle that is constant in time in a frame rotating at ω_r. With this choice, information about the qubits' state contained in the â^†+â quadrature of the field is erased while information in the i(â^†−â) quadrature is amplified (cf. FIG. 1). By itself, this does not lead to a substantial fidelity improvement. However, a careful treatment of the master equation shows that Eq. (5) can be recovered by adding a filter reducing the density of modes at ω_m to zero at the output port of the oscillator. Filters of this type are routinely used experimentally to reduce Purcell decay of superconducting qubits. As illustrated by the dark blue line in FIG. 2(c), using single-mode squeezing at ω_r and a filter at the modulation frequency, we recover the same exponential improvement found with two-mode squeezing, Eq. (5), in addition to a factor of two decrease in gate infidelity without squeezing.

Interestingly, rotating the squeezing axis by π/2 when squeezing at the oscillator frequency helps in distinguishing the different oscillator states and has been shown to lead to an exponential increase in the signal-to-noise ratio for qubit readout. In practice, the difference between performing a two-qubit gate and a measurement is thus the parametric modulation frequency (off-resonant for the gate and on resonance for measurement) and the choice of squeezing axis.

We note that Eq. (5) was derived from a master equation treatment under the standard secular approximation, which is not valid at high squeezing powers (here, ≳14 dB). At such high powers, the frequency dependence of n together with other imperfections are likely to be relevant.

Circuit QED implementation.—The proposed gate is applicable in a variety of physical systems where Eq. (1) is realizable. As examples, we mention flux qubits capacitively coupled to a resonator and spin qubits . We here propose an alternative realization based on transmon qubits, which are particularly attractive due to their excellent coherence times. The circuit we consider is illustrated in FIG. 3, and more details are given in SM B. This realization does not, in fact, give an exact realization of Eq. (1), as we show below that there are additional terms in the effective Hamiltonian, but these terms turn out to not be of any detrimental consequence for the gate. This circuit QED realization with transmon qubits is an extension to two qubits of the circuit introduced in Paper A. A detailed derivation of the Hamiltonian of this circuit yields in the two-level approximation

$\begin{array}{cc}{\hat{H}}^{\mathrm{circ}}\left(t\right)={\omega }_r{\hat{a}}^{†}\hat{a}-\varepsilon \left(t\right)\left(\hat{a}+{\hat{a}}^{†}\right)+\sum _i\left[\frac{{\omega }_{\mathrm{ai}}}{2}{\hat{\sigma }}_{\mathrm{zi}}+g_i\left(t\right)\left(\hat{a}+{\hat{a}}^{†}\right){\hat{\sigma }}_{\mathrm{zi}}+{\chi }_i{\hat{a}}^{†}\hat{a}{\hat{\sigma }}_{\mathrm{zi}}\right],& \left(6\right)\end{array}$

where modulation of the external fluxes indicated in FIG. 3 leads to the desired modulation of the longitudinal coupling, g_i(t)˜Φ_i(t), in addition to a small renormalization of the qubits frequencies. A large drive on the resonator leads to a similar effect, which means that in practice the longitudinal couplings can be modulated either through external fluxes or a strong resonator drive.

This Hamiltonian is different from Eq. (1) due to the presence of the λ_i ac-Stark shifts. We have also included a drive term ε(t), which includes a flux-induced contribution, but can be tuned via a gate voltage applied to the resonator (not shown in the figure). Note that the χ_i-terms are not the result of a dispersive approximation to a Jaynes-Cummings Hamiltonian but appear naturally in this circuit and does not introduce any Purcell decay or unwanted resonator-qubit hybridization. For reasonable circuit parameters, we find that χ_i˜g_i, which means that these ac-Stark shifts cannot be neglected. However, a {circumflex over (σ)}_z1{circumflex over (σ)}_z2 interaction can be made explicit through a polaron-type transformation which leads, in the symmetric case χ₁=χ2≡χ, g₁=g₂≡g (taken for simplicity), to the transformed

Hamiltonian

$\begin{array}{cc}{\hat{H}}_{\mathrm{pol}}^{\mathrm{circ}}\left(t\right)=\delta {\hat{a}}^{†}\hat{a}+\chi {\hat{a}}^{†}\hat{a}\left({\sigma }_{z1}+{\sigma }_{z2}\right)+J_z\left(t\right){\sigma }_{z1}{\sigma }_{z2}+\frac{\varpi \left(t\right)}{2}\left({\hat{\sigma }}_{z1}+{\hat{\sigma }}_{z2}\right),& \left(7\right)\end{array}$

where full expressions for J_z(t) and ω=(t) can be found in SM B. This equation is very similar to Eq. (2) in the sense that the ac-Stark shift does not play a role since the resonator stays in its ground state in this frame (assuming an initial vacuum state). The additional single qubit Z rotations can be corrected for through standard single-qubit gates.

There are however two main differences with Eq. (2). First, we have to impose an additional condition 2χt_g=2π×p with p an integer to avoid residual resonator-qubit entanglement at the end of the gate. Then, J_z(t) can be replaced by

$\begin{array}{cc}{J}_{z0}=\frac{-{\left(g\delta +\mathrm{\varepsilon \chi }\right)}^2}{2\delta \left({\delta }^2-4{\chi }^2\right)},& \left(8\right)\end{array}$

where we have assumed the ε(t)=ε cos(ω_mt) real, for simplicity. The second difference is that the paths in phase space differ from the paths shown in FIG. 1. Unfortunately we find that in the presence of ac-Stark shift, squeezing cannot be used to reduce the dephasing rate substantially since it is not possible to align the anti-squeezed quadrature and the displacement of the field at all times. These differences aside, the gate time and infidelity stay approximately the same in the absence of squeezing. For example, taking κ/2π=0.01 MHz, g/2π=40 MHz, χ/2π=75 MHz, ε=158 MHz and δ/2π=600 MHz results in a gate time of t_g=40 ns with an average gate infidelity of 5×10⁻⁵, which is comparable to the numbers based on Eq. (1).

Independent control of the couplings can be used for individual qubit readout while single-qubit X and Y rotations can be realized by driving weakly coupled voltage gates (not shown in the figure). It is interesting to note that longitudinal coupling can alternatively be obtained from the dispersive regime of transveral coupling under appropriate resonator drive. This approximate realization of longitudinal coupling however suffers from all of the above mentioned drawbacks of the transversal coupling from which it emerges.

Scalability.—So far we have focused on two qubits coupled to a single common oscillator. Longitudinal coupling of several qubits to separate oscillators that are themselves coupled transversely has favorable scaling properties. Interestingly, the gate introduced in this Letter

FIG. 3. (Color online.) Circuit QED implementation of longitudinal interaction with two qubits coupled to a common LC oscillator. Further details on the mapping of this circuit to Eq. (6) can be found in SM B.

can also be implemented in such an architecture. Consider two qubits interacting with distinct, but coupled, oscillators with the corresponding Hamiltonian

$\begin{array}{cc}{\hat{H}}_{\mathrm{ab}}={\omega }_a{\hat{a}}^{†}\hat{a}+{\omega }_b{\hat{b}}^{†}\hat{b}+\frac{1}{2}{\omega }_{a1}{\hat{\sigma }}_{z1}+\frac{1}{2}{\omega }_{a2}{\hat{\sigma }}_{z2}+g_1\left(t\right){\hat{\sigma }}_{z1}\left({\hat{a}}^{†}+\hat{a}\right)+g_2\left(t\right){\hat{\sigma }}_{z2}\left({\hat{b}}^{†}+\hat{b}\right)-g_{\mathrm{ab}}\left({\hat{a}}^{†}-\hat{a}\right)\left({\hat{b}}^{†}-\hat{b}\right).& \left(9\right)\end{array}$

In this expression, â, {circumflex over (b)} label the mode of each oscillator of respective frequencies ω_a,b, and g_ab is the oscillator-oscillator coupling. As above, g_1,2(t) are modulated at the same frequency ω_m, corresponding to the detunings δ_a≡ω_a−ω_m and δ_b≡ω_b−ω_m. Following the same procedure as above and performing a rotating-wave approximation for simplicity, we find a Hamiltonian in the polaron frame of the same form as Eq. (2), but now with a modified {circumflex over (σ)}_z{circumflex over (σ)}_z interaction strength

$\begin{array}{cc}{\stackrel{\_}{J}}_z=\frac{1}{2}\frac{g_1g_2g_{\mathrm{ab}}}{{\stackrel{\_}{\delta }}^2-g_{\mathrm{ab}}^2\left(1+{\zeta }^2\right)},& \left(10\right)\end{array}$

where δ=(δ_a+δ_b)/2 and ζ=(ω_b−ω_a)/(2g_ab) This implementation allows for a modular architecture, where each unit cell is composed of a qubit and coupling oscillators, used for both readout and entangling gates. Such a modular approach can relax design constraints and avoids spurious interactions with minimal circuit complexity.

Conclusion.—We have proposed a controlled-phase gate based on purely longitudinal coupling of two qubits to a common oscillator mode. The key to activating the qubit-qubit interaction is a parametric modulation of the qubit-oscillator coupling at a frequency far detuned from the oscillator. The gate infidelity and gate time can in principle be made arbitrarily small simultaneously, in stark contrast to the situation with transversal coupling. We have also shown how the gate fidelity can be exponentially increased using squeezing and that it is independent of qubit frequencies. The gate can moreover be performed remotely in a modular architecture based on qubits coupled to separate oscillators. Together with the fast, QND and high-fidelity measurement scheme presented in Paper A, this makes a platform based on parametric modulation of longitudinal coupling a promising path towards universal quantum computing in a wide variety of physical realizations.

[1] Note that 1- refers only to the error due to photon decay, excluding the qubits natural T₁ and T₂ times.

[2] This can be done by reducing the participation ratio, η, such that x≤κ.

Claims

1. A circuit quantum electrodynamics (circuit QED) implementation of a control-phase quantum logic gate UCP(θ)=diag[1,1,1,eiθ], the circuit QED implementation comprising: two qubits Qi, where i=1 corresponds to a first qubit Q1 and i=2 corresponds to a second qubit Q2, each having a frequency ωqi and being characterized by {circumflex over (σ)}zi;a first resonator Ra, associated with the qubit Q1, defined by: a resonator frequency ωra;a resonator electromagnetic field characterized by ↠and â;a longitudinal coupling strength g1z with the qubit Q1;a first longitudinal coupling g1z{circumflex over (σ)}1z(â†+â);wherein a second resonator Rb, such that, when the second resonator Rb is independent from Ra: Rb is associated with the qubit Q2;a longitudinal resonator-resonator coupling gab is defined and Rb is further defined by: a second resonator frequency ωrb;a second resonator electromagnetic field characterized by {circumflex over (b)}† and {circumflex over (b)};a second longitudinal coupling strength g2z with the qubit Q2;a second longitudinal coupling g2z{circumflex over (σ)}2z({circumflex over (b)}†+{circumflex over (b)}); andwhen the second optional resonator Rb is not independent from Ra and integrated into Ra: Ra is associated with the qubit Q2;the longitudinal resonator-resonator coupling gab=1;the second resonator electromagnetic field is characterized by ↠and â where {circumflex over (b)}†=↠and {circumflex over (b)}=â;the second resonator frequency ωrb=ωra;the second longitudinal coupling strength g2z is between the qubit Q2 and Ra;the second first longitudinal coupling g2z{circumflex over (σ)}2z({circumflex over (b)}†+{circumflex over (b)}) is defined by Ra as g2z{circumflex over (σ)}2z(â†+â); anda modulator periodically modulating, at a frequency ωm during a time t, the longitudinal coupling strengths g1z and g2z with respective signals of respective amplitudes {tilde over (g)}1 and {tilde over (g)}2, wherein selecting a defined value for each of t, g1z and g2z determines θ to specify the quantum logical operation performed by the control-phase quantum logic gate and wherein when the qubit Q1 and the qubit Q2 are decoupled when either one of the defined value of g1z and the defined value of g2z is to set to 0.

2. The circuit QED implementation of claim 1, further comprising a transmitter for selectively providing a modulator activation signal to the modulator for activating the modulator for the duration t.

3. The circuit QED implementation of claim 1, further comprising a signal injector providing a squeezed input to diminish a which-qubit-state information.

4. The circuit QED implementation of claim 3, wherein the squeezed input is a single-mode squeezed input.

5. The circuit QED implementation of claim 3, wherein the squeezed input is a two-mode squeezed input.

6. The circuit QED implementation of claim 3, wherein the signal injector relies on broadband squeezed centered at ωrb and/or ωra.

7. The circuit QED implementation of claim 1, wherein the qubit Q1 and the qubit Q2 are transmons each comprising two Josephson junctions with respectively substantially equivalent capacitive values and the modulator comprises an inductor-capacitor (LC) oscillator, the longitudinal coupling resulting from mutual inductance between the oscillator and the transmons, the oscillator varying a flux φ1 in the qubit Q1 and a flux φ2 in the qubit Q2.

8. The circuit QED implementation of claim 7, wherein a 3-Wave mixing Josephson dipole element is used to couple the qubit Q1 and the resonator Ra.

9. A method for specifying a quantum logical operation performed by a control-phase quantum logic gate UCP(θ) =diag[1,1,1,eiθ], wherein the circuit QED implementation comprises (I) two qubit Q1, where i=1 corresponds to a first qubit Q1 and i=2 corresponds to a second qubit Q2, each having a frequency ωqi and being characterized by {circumflex over (σ)}zi; (II) a first resonator Ra, associated with the qubit Q1, defined by a first resonator frequency ωra, a first resonator electromagnetic field characterized by ↠and â, a first longitudinal coupling strength g1z with the qubit Q1 and a first longitudinal coupling g1z{circumflex over (σ)}1z(â†+â); (III) a second resonator Rb, such that, when the second resonator Rb is independent from Ra, Rb is associated with the qubit Q2, a longitudinal resonator-resonator coupling gab is defined and Rb is further defined by: a second resonator frequency 107rb, a second resonator electromagnetic field characterized by {circumflex over (b)}† and {circumflex over (b)}, a second longitudinal coupling strength g2z with the qubit Q2, a second longitudinal coupling g2z{circumflex over (σ)}2z({circumflex over (b)}†+{circumflex over (b)}) and (IV), when the second optional resonator Rb is not independent from Ra and integrated into Ra, Ra is associated with the qubit Q2, the longitudinal resonator-resonator coupling gab=1, the second resonator electromagnetic field is characterized by ↠and â where {circumflex over (b)}†=↠and {circumflex over (b)}=â, the second resonator frequency ωrb=ωra, the second longitudinal coupling strength g2z is between the qubit Q2 and Ra, the second longitudinal coupling g2z{circumflex over (σ)}2z({circumflex over (b)}†+{circumflex over (b)}) is defined by Ra as g2z{circumflex over (σ)}2z(â†+â), the method comprising: periodically modulating, at a frequency ωm during a time t, the longitudinal coupling strengths g1z and g2z with respective signals of respective amplitudes {tilde over (g)}1 and {tilde over (g)}2;selecting a defined value for each of t, g1z and g2z thereby fixing θ to specify the quantum logical operation performed by the control-phase quantum logic gate; andsetting at least one of the defined value of g1z and the defined value of g2z is to 0 to decouple the qubit Q1 from the qubit Q2.

10. The method of claim 9, wherein selecting the defined value for each of t, g1z and g2z comprise a selectively providing a modulator activation signal to the modulator for activating the modulator for the duration t.

11. The method of claim 9 or claim 10, further comprising providing a squeezed input to diminish a which-qubit-state information.

12. The method of claim 11, wherein the squeezed input is a single-mode squeezed input.

13. The method of claim 11, wherein the squeezed input is a two-mode squeezed input.

14. The method of claim 11, wherein the squeezed input relies on broadband squeezed centered at ωrb and/or ωra.

15. The method of claim 9, wherein the qubit Q1 and the qubit Q2 are transmons each comprising two Josephson junctions with respectively substantially equivalent capacitive values and the modulator comprises an inductor-capacitor (LC) oscillator, the longitudinal coupling resulting from mutual inductance between the oscillator and the transmons, the oscillator varying a flux φ1 in the qubit Q1 and a flux φ2 in the qubit Q2.

16. The method of claim 15, wherein a 3-Wave mixing Josephson dipole element is used to couple the qubit Q1 and the resonator Ra.