Introduction

The technically challenging integration of suspended single-wall cabon nanotubes into complex qantum devices has recently made significant advances1,2,3,4,5,6, as has also the integration of nanotube quantum dots into coplanar microwave cavities7,8,9. Both regarding their nanomechanical10,11 as well as their electronic properties12,13, carbon nanotubes are a prototypical experimental system. However, small vibrational deflection and length have made their optomechanical coupling to microwave fields14 so far impossible.

In this work, we demonstrate large optomechanical coupling of a suspended carbon nanotube quantum dot and a microwave cavity. The nanotube is deposited onto source and drain electrodes close to the coplanar waveguide cavity; a finger-like extension of the cavity center conductor, passing below the suspended nanotube, serves as capacitively coupling gate. We find that the optomechanical coupling of the transversal nanotube vibration and the cavity mode is amplified by several orders of magnitude via the inherent nonlinearity of Coulomb blockade. With this, full optomechanical control of the carbon nanotube vibration in the quantum limit15 is reachable in the near future. A unique experimental system becomes accessible, where the nanomechanically active part directly incorporates a quantum-confined electron system16.

Results

Device precharacterization

Our device, depicted in Fig. 1a, combines a half-wavelength coplanar microwave cavity with a suspended carbon nanotube quantum dot. Near the coupling capacitor, the center conductor of the niobium-based cavity is connected to a thin gate electrode, buried between source and drain contacts of the carbon nanotube, see the sketch of Fig. 1b. At the cavity center, i.e., the location of the voltage node of its fundamental mode, a bias connection allows additional application of a dc voltage Vg to the gate. The device is mounted at the base temperature stage (T ≃ 10 mK) of a dilution refrigerator; for details see Supplementary Note 4 and Supplementary Fig. 4.

Fig. 1: Integrating a suspended carbon nanotube into a microwave cavity.
figure 1

a Optical micrograph showing a niobium-based λ/2 coplanar waveguide cavity for transmission measurement, with carbon nanotube deposition areas and dc contact structures (see the red dashed squares) near the coupling capacitors. For fabrication redundancy, two deposition areas exist on the device, but only one is used here. Bond wires visible as dark lines connect different segments of the ground plane to avoid spurious resonances. b Simplified sketch of the nanotube deposition area, including source and drain electrodes, a carbon nanotube deposited on them, and the buried gate connected to the cavity center conductor. c dc transport characterization of the carbon nanotube at Tbase ≃ 10 mK. The plot of the absolute value of currrent \(\left|I\right|\) as function of gate voltage Vg and bias voltage Vsd displays the typical diamond-shaped Coulomb blockade regions of suppressed conductance12,40,41. d Using rf excitation with an antenna and dc measurement17,18, two transversal vibration modes can be traced across a large gate voltage range; the figure plots the detected resonance frequencies. The corresponding raw data as well as a fit can be found in Supplementary Figs. 6 and 7.

Full size image

At cryogenic temperatures, electronic transport through the carbon nanotube is dominated by Coulomb blockade, with the typical behavior of a small band gap nanotube12. Near the electronic band gap, sharp Coulomb oscillations of conductance can be resolved; measurements are shown in Fig. 1c and Supplementary Fig. 3. A well-known method to detect the transversal vibration resonance of a suspended nanotube quantum dot is to apply a rf signal and measure the time-averaged dc current17,18,19. On resonance, the oscillating geometric capacitance, effectively broadening the Coulomb oscillations, leads to an easily recognizable change in current. This was used to identify the transversal vibration resonances of the device; Fig. 1d plots the resonance frequencies over a wide gate voltage range. Two coupled vibration modes are observed (see also Supplementary Note 5), one of which clearly displays electrostatic softening20,21. At low gate voltages, ∣Vg∣ ≤ 1.2 V, where subsequent experiments are carried out, the resonance which we will utilize in the following is at ωm ≃ 2π ⋅ 502.5 MHz, with typical quality factors around or exceeding Qm ~ 104 observed in time-averaged dc current detection17.

The combined suspended nanotube—cavity device forms a dispersively coupled optomechanical system14. The cavity has a resonance frequency of ωc = 2π ⋅ 5.74 GHz with a decay rate of κc = 2π ⋅ 11.6 MHz, dominated by internal losses. Nevertheless, due to the large mechanical resonance frequency ωm of the carbon nanotube, the coupled system is far in the resolved sideband regime ωm ≫ κc, the most promising parameter region for a large number of optomechanical protocols including ground state cooling and quantum control.

Optomechanically induced transparency (OMIT)

To probe for optomechanical coupling, we perform an OMIT type experiment22, cf. Fig. 2a, b: a strong, red-detuned drive field (ωd ≃ ωc − ωm) pumps the microwave cavity; the transmission of a weak, superimposed probe signal ωp near ωc is detected. A distinct, sharp OMIT absorption feature within the transmission resonance of the cavity becomes visible in the measurements of Fig. 2c–e. It occurs due to destructive interference of the probe field with optomechanically upconverted photons of the drive field, when the two-photon resonance condition ωp − ωd = ωm is fulfilled22, and shifts in frequency as expected when ωd is detuned from the precise red sideband condition, see Fig. 2d, e. Fitting the OMIT feature allows to extract the optomechanical coupling parameter \(g=\sqrt{{n}_{{\rm{c}}}}(\partial {\omega }_{{\rm{c}}}/\partial x){x}_{{\rm{zpf}}}\), describing the cavity detuning per displacement of the mechanical harmonic oscillator14,22, see Supplementary Note 9 for details. Surprisingly, from Fig. 2c, one obtains a single-photon coupling on the order of \({g}_{0}=g/\sqrt{{n}_{{\rm{c}}}} \sim 2\pi \cdot 100\ {\rm{Hz}}\).

Fig. 2: Optomechanically induced transparency (OMIT) in the Coulomb blockade regime.
figure 2

a Frequency scheme and b detection setup of an OMIT measurement. A strong drive signal at ωd = ωc − ωm pumps the microwave cavity; the cavity transmission near the cavity resonance ωc is characterized using a superimposed weak probe signal ωp from a vector network analyzer (VNA). Device parameters are: ωc ≃ 2π ⋅ 5.74 GHz, κc = 2π ⋅ 11.6 MHz, ωm ≃ 2π ⋅ 502.5 MHz. c–e Probe signal power transmission \({|{S}_{21}({\omega }_{{\rm{p}}})|}^{2}\) for three different choices of cavity drive frequency ωd, at ωd = ωc − ωm (c) and slightly detuned (d, e). The gate voltage Vg = −1.1855 V is fixed on the flank of a sharp Coulomb oscillation of conductance; Vsd = 0. f Probe signal transmission as in c–e, now for a fixed cavity drive frequency ωd = 2π ⋅ 5.23989 GHz and varied gate voltage Vg across a Coulomb oscillation. The depth of the OMIT feature allows the evaluation of the optomechanical coupling g(Vg) at each gate voltage value. g Optomechanical coupling g(Vg) (left axis) and corresponding single photon coupling \({g}_{0}({V}_{{\rm{g}}})=g({V}_{{\rm{g}}})/\sqrt{{n}_{{\rm{c}}}}\) (right axis), extracted from the data of f; nc = 67,500. Error bars indicate the standard error of the fit result.

Full size image

Such a value of g0 strongly exceeds expectations from the device geometry23. For a mechanical oscillator dispersively coupled to a coplanar waveguide resonator, the coupling is given by

$${g}_{0}=\frac{{\omega }_{{\rm{c}}}}{2{C}_{{\rm{c}}}}\frac{\partial {C}_{{\rm{c}}}}{\partial x}{x}_{{\rm{zpf}}},$$
(1)

where Cc is the total capacitance of the cavity, x is the mechanical displacement, and xzpf the mechanical zero-point fluctuation length scale. Assuming a metallic wire over a metallic plane and inserting device parameters23, the coupling calculated from the change in geometric gate capacitance Cg(x) becomes ∂Cg/∂x ~10−12 F m−1. This leads to \({g}_{0}^{* }=2\pi \cdot 2.9\ {\rm{mHz}}\), more than four orders of magnitude smaller than the measured g0. To explain this discrepancy, we need to focus on the properties of the carbon nanotube as a quantum dot, with a strongly varying quantum capacitance CCNT(x) as the displacement-dependent component of Cc dominating g0.

  Figure 2f depicts OMIT measurements for similar parameters as in Fig. 2c–e, however, we now keep the drive frequency ωd constant and vary the gate voltage Vg across a Coulomb oscillation of conductance. The mechanical resonance frequency ωm shifts to lower frequencies in the vicinity of the charge degeneracy point. This electrostatic softening is a well-known characteristic of suspended carbon nanotube quantum dots18,24. More interestingly, the resulting gate-dependent coupling g(Vg) (along with g0(Vg)) is plotted in Fig. 2g. It is maximal at the edges of the finite conductance peak, whereas at its center and on the outer edges, the coupling vanishes; the enhancement of g0 is intrinsically related to Coulomb blockade.

Mechanism of enhanced coupling

  Figure 3 explores the nature of this enhanced coupling mechanism. We treat the nanotube as a single quantum dot; see Supplementary Note 3 for a discussion of the validity of this assumption. Further, we assume a full separation of time scales ωm ≪ ωc ≪ Γ, where Γ describes the tunnel rates of the quantum dot. We can then introduce the quantum capacitance25,26

$${C}_{{\rm{CNT}}}=e\frac{{C}_{{\rm{g}}}}{{C}_{{\rm{dot}}}}\frac{\partial \langle N\rangle }{\partial {V}_{{\rm{g}}}},$$
(2)

where \(\left\langle N\right\rangle ({V}_{{\rm{g}}})\) is the number of charge carriers (here holes) on the quantum dot averaged over the tunneling events, and Cdot is the total quantum dot capacitance; see Supplementary Note 12 for a derivation. In a quantum dot, each Coulomb oscillation corresponds to the addition of one electron or hole. The charge occupation \(\left\langle N\right\rangle ({V}_{{\rm{g}}})\) resembles a step function, with the sharpness of the step given for zero bias voltage by lifetime and temperature broadening. This is plotted in Fig. 3a, for the limit of kBT ≪ Γ. The quantum capacitance CCNT(Vg) becomes a Lorentzian, as plotted in Fig. 3b.

Fig. 3: Coulomb blockade enhanced optomechanical coupling mechanism.
figure 3

Solid lines correspond to the model of a Lorentz-broadened quantum dot level at kBT ≪ Γ. The Coulomb oscillation center \({V}_{{\rm{g}}}^{* }=-1.18841\ {\rm{V}}\), the line width ℏΓ = 0.673 meV, and a scaling prefactor a = 5.77 (see text) have been obtained by fitting to the OMIT data g(Vg). a Time-averaged charge occupation \(\left\langle N\right\rangle ({V}_{{\rm{g}}})\) of the quantum dot (note that we are in the hole conduction regime). b Conductance dI/dVsd(Vg) (left axis) and quantum capacitance CCNT (right axis), cf. Supplementary Fig. 12. c Coulomb-blockade enhanced optomechanical coupling g(Vg) (left axis) and single photon coupling g0(Vg) (right axis). The data points are identical to Fig. 2g; the calculation result has been scaled with 5.77 to fit the data. d–f Schemata for the situations corresponding to the dashed lines in a–c, see the text.

Full size image

Any motion δx modulates the geometric capacitance Cg(x). It thus shifts the position of the Coulomb oscillations in gate voltage, acting equivalent to an effective modulation of the gate voltage δVg. With this, the optomechanical coupling g, scaling with \(\left|\partial {C}_{{\rm{CNT}}}/\partial x\right|\), becomes proportional to the derivative ∂CCNT/∂Vg and thus the second derivative of \(\left\langle N\right\rangle ({V}_{{\rm{g}}})\), as is illustrated in Fig. 3c. The functional dependence has been fitted to the data points of Fig. 2g, here again shown in the background.

The three key situations depending on the gate voltage are sketched in Fig. 3d–f: away from the conductance peak, the charge on the nanotube is constant, and only geometric capacitances change, see Fig. 3d. On the flank of the conductance resonance, a small change δx (∝δCg) strongly modulates CCNT, see Fig. 3e. At the center of the conductance resonance, the charge adapts to x, but the derivative ∂CCNT/∂Vg and with it g ∝ |∂CCNT/∂x| is approximately zero.

The detailed derivation and the full expressions and values for Fig. 3 can be found in the Supplementary Information, Supplementary Note 12, and Supplementary Table 1. The parameter entering the optomechanical coupling in Eq. (1), the derivative of the quantum capacitance ∂CCNT/∂x, is found to be

$$\frac{\partial {C}_{{\rm{CNT}}}}{\partial x}=\eta \frac{\partial {C}_{{\rm{g}}}}{\partial x}=e\frac{{\partial }^{2}\left\langle N\right\rangle }{\partial {V}_{{\rm{g}}}^{2}}\frac{{V}_{{\rm{g}}}}{{C}_{{\rm{dot}}}}\frac{\partial {C}_{{\rm{g}}}}{\partial x},$$
(3)

indicating that for significant optomechanical coupling a sharp Coulomb oscillation (i.e., low temperature and low intrinsic line width Γ, leading to large values of \({\partial }^{2}\,\left\langle N\right\rangle /\partial {V}_{{\rm{g}}}^{2}\)) and a large Vg are required. From device data, we obtain an amplification factor η ~ 104. The experimental gate voltage dependence g0(Vg) is qualitatively reproduced very well. To obtain the quantitative agreement of Fig. 3c, we have introduced an additional scaling prefactor as free fit parameter, resulting in \({g}_{0}^{{{\exp }}}/{g}_{0}^{{\rm{th}}}=5.77\). Given the uncertainties of input parameters, this is a good agreement; see Supplementary Note 15 for a discussion of error sources.

Discussion

In literature, many approaches have been pursued to enhance optomechanical coupling26,27,28,29,30,31,32,33,34,35. Resonant coupling, with ωm = ωc, has been demonstrated successfully for a carbon nanotube quantum dot26, but does not provide access to the wide set of experimental protocols developed for the usual case of dispersive coupling and the “good cavity limit” ωm ≫ κc. The mechanism presented here is most closely related to those where a superconducting charge qubit was coherently introduced between mechanical resonator and cavity27. However, the impact of single electron tunneling and shot noise on the optomechanical system shall require careful analysis.

Given the sizeable coupling in the good cavity limit κc ≪ ωm, many experimental techniques for future experiments are at hand. First steps are demonstrated in Fig. 4 in a two-tone spectroscopy experiment: a mechanical drive signal ωa is applied simultaneously to a cavity pump signal at ωd = ωc − ωa; the plotted cavity output power at ωc clearly shows the optmechanical upconversion (anti-Stokes scattering) at mechanical resonance ωa = ωm. In Fig. 4a, the dc bias across the nanotube is set to zero, and the antenna drive kept at a minimum. In Fig. 4b, both antenna drive and bias voltage have been increased. A background signal independent of device parameters emerges; at the same time, the upconverted signal displays a phase shift and destructive interference with the background for parts of the gate voltage range, meriting further measurements and analysis.

Fig. 4: Two-tone spectroscopy.
figure 4

Via an antenna, the carbon nanotube is driven at ωa close its mechanical eigenfrequency; the microwave cavity is simultaneously pumped at ωd = ωc − ωa. The plots show the power output of the cavity at the upconverted frequency ωc, with the nanotube acting as nonlinear element. Drive power Pd = 20 dBm (nc ≃ 2.1 × 104), measurement bandwidth 5 Hz. a Antenna generator power Pa = −55 dBm, bias Vsd = 0; b antenna generator power Pa = −30 dBm, bias Vsd = 0.5 mV.

Full size image

Future improvements of the optomechanical coupling via drive power and device geometry and of the detection sensitivity via the output amplifier chain shall allow detection of the thermal motion of the carbon nanotube and subsequently motion amplitude calibration.

The observation of strong optomechanical coupling and the corresponding normal mode splitting requires a coupling g exceeding both mechanichal linewidth κm and cavity line width κc. Clean carbon nanotubes have reached mechanical quality factors up to36 Qm ~ 106, allowing for two orders of magnitude improvement and a line width of κm ~ 2π · 500 Hz. Regarding microwave resonators we have reached up to Qc = 105 in our setup so far, corresponding to κc = 2π · 57 kHz. This means that strong coupling should be reachable already at moderate increase of our so far rather low cavity photon number nc.

Regarding the cooperativity C = 0.0042 of our experiment (cf. Supplementary Table 1), already an improvement of the nanotube Qm by a factor 100 brings it into the same order of magnitude as the thermal mode occupation nm = 0.4, with significant further and independent room for improvement via the cavity photon number nc.

With this, a wide range of physical phenomena becomes experimentally accessible, ranging from side-band cooling of the vibration mode and potentially its quantum control37 all the way to real-time observation of its interaction with single electron tunneling phenomena38.