Monolithic Segmented 3D Ion Trap for Quantum Technology Applications

Abstract

Monolithic three-dimensional (3D) Paul traps combine the high-precision microfabrication of two-dimensional (2D) chip traps with the deep trapping potentials and low heating rates characteristic of macroscopic Paul traps, which are typically manually assembled. However, achieving low motional heating rates and optical access with a high numerical aperture (NA) while maintaining the high radio-frequency (RF) voltages required for heavy ionic species, such as Yb$^{+}$ and Ba$^{+}$, remains a significant technical challenge. In this work, we present a segmented, monolithic 3D fused silica blade trap, featuring an ion-electrode distance of 250 $μ$m with stable operation at high RF voltages. We benchmark the performance of the trap using Yb$^{+}$ ions, demonstrating axially homogeneous trapping potentials for 200 $μ$m around the axial center of the trap, high multi-directional optical access (up to 0.7 NA), and radial motional heating rate as low as 1.1 $\pm$ 0.1 quanta/s at radial trap frequencies about 3 MHz near room temperature. Furthermore, we observe a motional Ramsey coherence time, $T_{2}$, of around 95 ms for the radial center-of-mass mode. We demonstrate a two-qubit gate fidelity of ${99.3}^{+ 0.7}_{- 1.5}$$\%$ with state preparation and measurement correction. These results establish fused-silica monolithic blade traps as a scalable, modular platform for quantum simulation, computation, metrology, and networking with heavy ionic species.

Figures (20)

• Figure 1: (a) GEN3 segmented, monolithic 3D blade trap made of fused silica. (b) The monolithic trap assembly and the oven assembly, for Yb$^+$ and Ba$^+$ ions, before being inserted into the vacuum chamber. (c) A Doppler-cooled linear crystal of 19, $^{171}\rm{Yb}^{+}$ ions at $\sim3$ MHz radial confinement, with an inter-ion spacing of $4.2(3)\,\mu$m for the central 13 ions, for which the uniformity was optimized with 7% variability. (d) An exploded view of the trap stack detailing the metal and ceramic layers for heat dissipation, and the routing of 22 AWG copper wires from the trap's gold pads (through fuzz-buttons, not shown) to the in-vacuum filtering capacitors for the DC electrodes. The ground wire connects the trap's ground pad to the aluminum base plate. (e) Insets of critical features of the trap, which include the high multi-directional optical access from the blade geometry and axial trenches, segmented electrodes, and trench geometries for DC-DC and RF-GND isolation. (f) Thermal image of the trap stack under high-power RF voltage test in UHV conditions. The color-coded temperature map is measured at $V_{\rm{RF}}=680 \ V_{\rm{pk}}$, $\Omega_{\rm{RF}}/2\pi=23.26\ \rm{MHz}$. The thermal IR camera temperature reading is calibrated to an in-vacuum thermocouple, in contact with the top AlN ceramic of the trap stack.
• Figure 2: (a) Comparison of radial frequencies measured as a function of position along the $x$ axis for the GEN2 (orange) and GEN3 (blue) trap. Squares and triangles refer to HF and LF radial modes, respectively, along with their simulation-based fits (dashed lines). GEN2 trap measurements were at $V_{\rm{RF}}\approx380\,\rm{V_{pk}}$ and $V_{\rm{twist}}=-0.5\,\rm{V}$, while those of the GEN3 trap were at $V_{\rm{RF}}\approx\,480\rm{V_{pk}}$ and $V_{\rm{twist}}=-2\,\rm{V}$. The shallow concavity in the HF and LF modes is due to the $x$-axis 'push' that shifts the ion, and it is well captured by the simulation. (b) Axial frequency measurements of GEN2 (orange crosses) and GEN3 (blue circles) vs applied voltage on endcaps, and their common axial frequency simulation (black line).
• Figure 3: Residual axial micromotion profile of the GEN2 (orange) and GEN3 (blue) trap at $\omega/2\pi\sim3$ MHz radial confinement.
• Figure 4: (a) Heating rate of a single $^{171}\rm{Yb}^{+}$ ion extracted by measuring $\bar{n}$ (error bars representing standard deviation) using the sideband ratio method for different wait times after sideband cooling Monroe1995a, for the HF (blue) and LF (red) modes at $\sim480\ \rm{V}_{\rm{pk}}$ and -2 V of 'twist' voltage. (b) A collection of surface electric field noise measurements, $S_E(\omega)$, of the axial and radial modes in multiple Paul traps as a function of the ion-electrode distance (see references in table \ref{['tab_heatingrate_legend']}). The noise-spectral density is rescaled by the frequency, assuming $1/\omega$ scaling of $S_E(\omega)$. The red, open star represents $\dot{\bar{n}}= 1.1\pm0.1$ q/s at $\omega/2\pi= 2.9$ MHz for the GEN3 trap. The dashed gray line passing through our measurement is a guiding line for the typical scaling of $d^{-4}$ from fluctuating patch potentials or dipoles on the trap electrode surface safavi2011microscopicsafavi2013influenceBrownnutt2015.
• Figure 5: (a) Heating rate measurements at $V_{\rm{RF}}\sim480\ V_{\rm{pk}}$ for the 'cold' radial mode (squares) and 'hot' (triangles) mode as a function of radial frequency by changing the 'twist' voltage (magnitude indicated by the colorbar). For a given 'twist' voltage and heating rate per mode, the heating rates for each radial mode frequency are swapped by inverting the sign of the 'twist' voltage. (b) FEA simulation of the HF mode's projection angle ($\Theta$) with respect to the $z$ axis and radial mode frequency (HF-black; LF-gray) as a function of 'twist' voltage. A 3 V endcap voltage lifts the radial mode degeneracy for $V_{\rm{twist}}=0\,\rm{V}$, without rotating the principal axes. The blue ellipse, between the blades, denotes an equipotential contour of the radial confinement where the minor axis is the HF mode and the major axis is the LF mode. Due to the blade angle of $\sim30^\circ$ with respect to the $y$ axis, a strong positive (negative) twist rotates the LF (HF) mode to become the 'cold' mode, while the HF (LF) mode becomes the 'hot' mode. (c) Heating rate of both 'cold' (blue squares) and 'hot' (red triangles) radial modes measured as a function of trap frequency by changing both 'twist' and RF voltage. Solid blue (red) line is a fit to the 'cold' ('hot') mode. Error bars represent the standard deviation in the heating rate slope extracted from a linear fit.