Improved placement precision of implanted donor spin qubits in silicon using molecule ions

TL;DR

Donor spins in silicon offer long coherence times but achieving precise, deterministic placement remains challenging due to ion straggling. The authors demonstrate PF$_2^+$ molecule ions, which co-implant phosphorus with fluorine bystander ions, boosting the ion beam induced charge signal and allowing high-detection-confidence placement at shallow depths. SIMS shows fluorine diffuses away during the donor-activation anneal while phosphorus remains near its target location, and ESR measurements yield $T_2^* = 20.5 ± 0.5 μs$ and $T_2^{Hahn} = 424 ± 5 μs$ with no detectable coupling to ^{19}F; the qubit is therefore not significantly magnetically noisy from fluorine. This work provides a scalable route to high-precision donor qubit arrays compatible with standard MOS fabrication.

Abstract

Donor spins in silicon-28 ($^{28}$Si) are among the most performant qubits in the solid state, offering record coherence times and gate fidelities above 99%. Donor spin qubits can be fabricated using the semiconductor-industry compatible method of deterministic ion implantation. Here we show that the precision of this fabrication method can be boosted by implanting molecule ions instead of single atoms. The bystander ions, co-implanted with the dopant of interest, carry additional kinetic energy and thus increase the detection confidence of deterministic donor implantation employing single ion detectors to signal the induced electron-hole pairs. This allows the placement uncertainty of donor qubits to be minimised without compromising on detection confidence. We investigate the suitability of phosphorus difluoride (PF$_2^+$) molecule ions to produce high quality P donor qubits. Since $^{19}$F nuclei have a spin of $I = 1/2$, it is imperative to ensure that they do not hyperfine couple to P donor electrons as they would cause decoherence by adding magnetic noise. Using secondary ion mass spectrometry, we confirm that F diffuses away from the active region of qubit devices while the P donors remain close to their original location during a donor activation anneal. PF$_2$-implanted qubit devices were then fabricated and electron spin resonance (ESR) measurements were performed on the P donor electron. A pure dephasing time of $T_2^* = 20.5 \pm 0.5$ $μ$s and a coherence time of $T_2^{Hahn} = 424 \pm 5$ $μ$s were extracted for the P donor electron-values comparable to those found in previous P-implanted qubit devices. Closer investigation of the P donor ESR spectrum revealed that no $^{19}$F nuclear spins were found in the vicinity of the P donor. Molecule ions therefore show great promise for producing high-precision deterministically-implanted arrays of long-lived donor spin qubits.

Figures (5)

• Figure 1: A donor spin qubit device fabricated using the implantation of PF$_2^+$ molecule ions. The bystander $^{19}$F ions diffuse away from the active region of the device upon a donor activation anneal, while the $^{31}$P donor qubit remains close to the original stopping location.
• Figure 2: SRIM simulation of the P ion distribution resulting from the implantation of a) 15 keV P$^+$ ions and b) 19.5 keV PF$_2^+$ ions into Si. The colour scale represents the probability distribution of the P donors. Effects from ion channelling were neglected.
• Figure 3: SIMS measurements showing the concentration of P and F as a function of depth below the Si surface for a) sample A and b) sample B. The coloured dashed lines show the depth profiles of P and F simulated using SRIM for an implantation of 20 keV PF$_2$ at a fluence of $3\times10^{14}$ cm$^{-2}$. The black dashed line in b) shows the background F concentration present in sample B outside of the implanted region.
• Figure 4: a) Adiabatic ESR spectrum of a $^{31}$P donor electron in Si. Two ESR peaks are visible with a splitting of $\sim$115 MHz due to the hyperfine coupling to the $I=1/2$ nuclear spin of the $^{31}$P donor. b) Ramsey measurement on the $^{31}$P donor electron with the corresponding pulse sequence shown in the top right. The fit yields a pure dephasing time $T_2^*=20.5\pm0.5$$\mu$s. c) Hahn echo measurement on the $^{31}$P donor electron with the corresponding pulse sequence shown in the top right. The data is fit with a sinusoidal Gaussian decay, yielding a coherence time $T_2^{\text{Hahn}}= 424 \pm 5$$\mu$s.
• Figure 5: a) Pulse sequence used to flip the spins of $^{29}$Si or $^{19}$F nuclei while monitoring the P donor ESR spectrum. b) The Bohr radius of the P donor electron encompasses multiple $^{29}$Si spins but no $^{19}$F spins. c) Coherent ESR spectrum of the lower hyperfine split P donor ESR peak. The P donor electron spin up probability is measured as a function of ESR frequency for multiple repetitions, interleaved with $^{29}$Si NMR $\pi/2$ pulses. The instantaneous ESR frequency is seen to jump frequently between discrete values, determined by the orientation of a few $^{29}$Si nuclear spins, hyperfine coupled to the P electron with strengths of order 100 kHz. d) The same experiment repeated with NMR $\pi/2$ pulses applied at the expected resonance of $^{19}$F nuclei, interleaved with the collection of coherent ESR spectra. The ESR frequency remains completely stable, indicating that no $^{19}$F are coupled to the P donor electron.