Isotopic enrichment of silicon by high fluence ²⁸Si^- ion implantation

Donor and quantum dot spin qubits in silicon (Si) are attractive candidates for high-fidelity scalable quantum computing architectures Kane (1998); Loss and DiVincenzo (1998); Zwanenburg et al. (2013). Si provides a desirable matrix for hosting spin qubits due to its important role in the microelectronics industry, weak spin-orbit coupling and the existence of isotopes with zero nuclear spin. Natural Si consists of three isotopes: ²⁸Si (92.23%), ²⁹Si (4.67%) and ³⁰Si (3.1%) De Groot (2004). The dominant source of qubit decoherence in ${}^{\text{nat}}$Si is due to coupling with the surrounding ²⁹Si nuclei, which possess a nuclear spin of $I=1/2$. Dipolar fluctuations of ²⁹Si spins cause perturbations in the local magnetic field, resulting in a time-varying qubit resonance frequency de Sousa and Sarma (2003); Witzel et al. (2010). This spectral diffusion limits the spin coherence time to around 200 $\mu$s for electrons Pla et al. (2012) and 60 ms for ionised donor nuclei Pla et al. (2013), as measured for a single phosphorus (P) donor at low temperature using the Hahn-echo pulse sequence.

Fortunately, ²⁸Si has no nuclear spin and can therefore provide an ideal low-noise environment for spin qubits. Long coherence times for donor spin qubits in a ²⁸Si epilayer with 800 ppm residual ²⁹Si Itoh and Watanabe (2014) have been demonstrated, with Hahn-echo decay times of around 1 ms for electrons and 1.75 s for ionised donor nuclei measured for a single P donor at low temperature, which can be further extended with dynamical decoupling Muhonen et al. (2014). Isotope engineering of semiconductor materials also has applications for increased thermal conductivity Capinski et al. (1997); Ruf et al. (2000); Inyushkin et al. (2018), capable of improved heat dissipation in Si integrated circuits Kizilyalli et al. (2005).

Isotopically enriched ²⁸Si can be produced by various methods, many of which involve the centrifugation of silicon tetrafluoride gas to produce high purity ²⁸SiF₄ Takyu et al. (1999); Bulanov et al. (2000); Itoh et al. (2003); Becker et al. (2009, 2010); Sennikov et al. (2012); Abrosimov et al. (2017). In the Avogadro Project Becker et al. (2010), ²⁸SiF₄ is converted into isotopically pure silane gas, ²⁸SiH₄, which is used to grow polycrystalline ²⁸Si by chemical vapour deposition (CVD). Float-zone growth is then used to produce ²⁸Si single crystal rods. An isotopic purity of $<10$ ppm ²⁹Si and the highest chemical purity to date ($<4\times 10^{15}$ cm^-3 for C and $<4\times 10^{14}$ cm^-3 for O) was achieved Abrosimov et al. (2017). Epilayers of ²⁸Si can be grown on ${}^{\text{nat}}$Si substrates by CVD Mazzocchi et al. (2019); Sabbagh et al. (2019) (a method employed by the Isonics Corporation Itoh and Watanabe (2014)) or molecular beam epitaxy Kojima et al. (2003); Bracht et al. (2012); Nakabayashi et al. (2000) (suitable for encapsulation of scanning tunneling microscopy-placed donors in ²⁸Si O’Brien et al. (2001); Oberbeck et al. (2002)). These methods rely on isotopically enriched sources of silane gas or solid-state Si. Mass spectrometry, on the other hand, can be used to separate ²⁸Si ions from a natural silane gas source. This has been used in conjunction with hyperthermal energy ion beam deposition to achieve a residual ²⁹Si fraction of $<1$ ppm in a layer of ²⁸Si Dwyer et al. (2014); Tang et al. (2020). However, the concentration of C and O was greater than $1\times 10^{19}$ cm^-3 Dwyer et al. (2014). The epitaxial growth of ²⁸Si thin films by ion beam deposition with a solid ${}^{\text{nat}}$Si source has also been achieved by depositing low energy ($\sim$40 eV) ²⁸Si^- ions Tsubouchi et al. (2001).

In this work, we achieve isotopic enrichment by implanting a high fluence of ²⁸Si^- ions, mass-separated from an accelerated ion beam produced from a solid-state ${}^{\text{nat}}$Si source, into a ${}^{\text{nat}}$Si substrate, shown schematically in Fig. 1. Since any isotope in the substrate can be sputtered from the surface, but only ²⁸Si ions are implanted, the levels of ²⁹Si and ³⁰Si are depleted with increasing fluence. An enriched surface layer of ²⁸Si can be produced with sufficient volume for donor qubits; typically implanted to a depth of around 20 nm below the Si surface for effective control and readout by surface nanocircuitry Tosi et al. (2017). This method has the advantage of using standard ion beam laboratory equipment, enabling the integration of in-situ enrichment with existing ion-implanted donor qubit fabrication Van Donkelaar et al. (2015). Additionally, the creation of an amorphous ²⁸Si layer increases the placement precision of implanted donors by suppressing ion channeling Schreutelkamp et al. (1991) and can increase the donor electrical activation yield Holmes et al. (2019). Solid phase epitaxy (SPE) of an amorphous Si ($a$-Si) surface layer formed by Si implantation can produce near-perfect single-crystal Si ($c$-Si) Zhu et al. (1999a) with a smooth surface Zhu et al. (1998), in which no long-range atomic displacement occurs Radek et al. (2015). A one-for-one replacement regime resulting in a planar Si surface (suitable for post-fabrication of nanocircuitry) could allow for economical local enrichment of regions using a focused ²⁸Si ion beam into which donors are implanted. Minimising the number of ²⁹Si nuclei within the Bohr radius of the donor electron ($\sim$2 nm for ³¹P in Si Kohn and Luttinger (1955)) reduces the coupling of the donor electron to the dynamics of the ²⁹Si spin-bath through the contact hyperfine interaction Witzel et al. (2010).

The concentration of impurities introduced to the enriched layer during high fluence ²⁸Si implantation must be minimised, with particular care taken to avoid co-implantation of the molecular isobars CO and N₂. The presence of C, N and O in the $a$-Si layer at levels of around 0.5 at. % cause retardation of the SPE regrowth rate Kennedy et al. (1977) and can lead to polycrystalline nucleation during annealing at increased implant concentrations Williams (1983). The use of a negative ²⁸Si ion beam ensures a negligible component of the isobars CO and N₂, since these negatively charged molecular ions are electronically unstable.

In the present work, 30-45 keV ²⁸Si^- ions were implanted at a high fluence into ${}^{\text{nat}}$Si. The crystallisation kinetics were determined using time resolved reflectivity (TRR) and the coherence time of P donors implanted into the enriched ²⁸Si layer was measured using pulsed electron spin resonance (ESR). The extent of ²⁹Si depletion and impurity levels introduced were measured using secondary ion mass spectrometry (SIMS). The concentration of impurities was sufficiently low to allow for crystallisation to take place via SPE, resulting in a single crystal layer of ²⁸Si, as shown by transmission electron microscopy (TEM).

A 150 keV ion implanter, equipped with a SNICS II ion source and a 90^∘ double focusing magnet, was used for all implants. Fig. 1 schematically shows the mass spectrum of this implanter with a ${}^{\text{nat}}$Si source, demonstrating the mass resolution of the Si isotopes. Near the start of an implant run, the ion beam currents for ²⁹Si^- and ³⁰Si^- are higher than their natural abundance as they contain a significant fraction of ²⁸SiH^- and ²⁹SiH^-, respectively. ²⁸Si^- implants were performed at room temperature and with a 7^∘ tilt off the incident beam axis to suppress ion channeling. A Si aperture, prepared from a wafer of ${}^{\text{nat}}$Si, was used to collimate the beam and prevent contamination from forward recoils of foreign atoms. Substrates were given a degreasing clean and a HF etch to remove surface hydrocarbons and native SiO₂ before entering the implant chamber to further reduce contamination. A vacuum of less than $1\times 10^{-7}$ Torr was maintained in the target chamber with a cryopump to reduce impurity incorporation from residual gas by ion bombardment.

A highly intrinsic (4-10 k$\Omega$.cm) float-zone uniform high purity ${}^{\text{nat}}$Si (UHPS Topsil) substrate was implanted with 30 keV ²⁸Si^- at a fluence of $4\times 10^{18}$ cm^-2 followed by ³¹P^- implantation (30 keV, $6.5\times 10^{11}$ cm^-2 then 10 keV, $1.5\times 10^{11}$ cm^-2), resulting in a P concentration of $<1.4\times 10^{17}$ cm^-3 throughout the enriched ²⁸Si layer. To investigate the crystallisation kinetics of the $a$-Si layer, TRR Olson and Roth (1988) was used with a laser wavelength of $\lambda=632.8$ nm during an anneal in air at 609 ^∘C. The rate of crystallisation was compared to an $a$-Si standard: n-type ${}^{\text{nat}}$Si amorphised with a much lower fluence of ²⁸Si^- ions with the following implantation scheme: (0.5 MeV, $3\times 10^{15}$ cm^-2), then (1 MeV, $1\times 10^{15}$ cm^-2) and finally (2 MeV, $1\times 10^{15}$ cm^-2). The SPE growth rate of this $a$-Si standard during the initial stages of the anneal was used to calibrate the temperature of the TRR hot plate. Refractive indices of $n_{c}=4.086$ for $c$-Si Jellison Jr and Modine (1994) and $n_{a}=4.831$ for $a$-Si Olson and Roth (1988) were used for the SPE rate calculation.

Pulsed ESR was performed on P donors implanted in the ²⁸Si layer of the ²⁸Si^- (30 keV, $4\times 10^{18}$ cm^-2) implanted sample after SPE. The sample was mounted onto the surface of a superconducting cavity made by dry etching a 100 nm-thick NbTiN film Kobayashi et al. (2020); Weichselbaumer et al. (2019). The sample, along with the cavity, was mounted on the mixing chamber of a dilution refrigerator, with a base temperature of $\sim$16 mK. Pulses were sent to the cavity using a vector source at the resonant frequency of the cavity ($f=6.028$ GHz), and the detected echo signal was then pre-amplified and measured using a digitizer Kobayashi et al. (2020). The ESR spectrum was obtained by varying the external magnetic field, $B_{0}$, which confirmed the presence of P by the observation of the two hyperfine-split peaks due to the nuclear spin of ³¹P Feher (1959). The $T_{2}$ was measured by setting the magnetic field to the value corresponding to the upper hyperfine-split P peak and using a standard Hahn echo pulse sequence Kobayashi et al. (2020), where the pulse length (400 ns) and power were chosen such that the spins undergo a $\pi/2$ rotation for the first pulse. Due to the long $T_{1}$ of P donor electrons, the sample was illuminated with light of wavelength 1025 nm for 100 ms between each repetition of the pulse sequence in order to thermalize the donor spins faster Tyryshkin et al. (2012).

To improve the enrichment process, the implantation of ²⁸Si at various energies and fluences into ${}^{\text{nat}}$Si was simulated using TRIDYN, a binary collision Monte Carlo simulation package Möller et al. (1988). An initial interval spacing of $5$ Å was chosen to be longer than the mean free path but small enough to avoid artifacts from a coarse grid Möller and Eckstein (1985). A high statistical quality was achieved using a precision of 0.02 to keep the maximum relative change of layer areal density per projectile to $<0.2$%.

An additional ${}^{\text{nat}}$Si (UHPS Topsil) substrate was implanted with 45 keV ²⁸Si^- ions to a fluence of $2.63\times 10^{18}$ cm^-2. A piranha (4:1 98% H₂SO₄ : 30% H₂O₂, 90 ^∘C) and RCA-2 (5:1:1 H₂O : 30% H₂O₂ : 36% HCl, 70 ^∘C) clean was then performed followed by a thermal anneal at 620 ^∘C for 10 min to facilitate SPE growth and a rapid thermal anneal at 1000 ^∘C for 5 s, suitable for donor activation, both performed in an Ar atmosphere.

The composition with depth of the 45 keV, $2.63\times 10^{18}$ cm^-2 implanted sample after annealing was obtained with SIMS (IONTOF GmbH, TOF.SIMS 5). The Si isotopes were measured in positive polarity with a 1 keV O${}_{2}^{+}$ beam used for sputtering and a 30 keV Bi⁺ beam used for analysis. C and O impurities were measured in negative polarity with a 1 keV Cs⁺ beam used for sputtering and a 30 keV Bi⁺ beam used for analysis.

TEM was used to determine the crystal quality of the enriched ²⁸Si layer in the 45 keV, $2.63\times 10^{18}$ cm^-2 implanted sample after annealing. Before TEM lamella preparation, the sample was coated with a thin carbon layer ($\sim$20 nm). To prepare the sample, a focused ion beam (FEI, Nova Nanolab 200) was used to grow a 300 nm thick layer of Pt via electron-beam assisted deposition. This was followed by a further 2.5 $\mu$m Pt layer deposited via a 30 keV Ga ion beam. A lamella was then extracted and thinned to a thickness of $\sim$100 nm, with a final polishing step performed with a 5 keV Ga ion beam. A TEM (FEI, Tecnai TF20) was used to take high-resolution cross-sectional images in which a 200 keV electron beam was transmitted down the [110] direction through the lamella to view the atomic arrangement.

TRR was used to show the complete crystallisation of the $a$-Si layer in the 30 keV, $4\times 10^{18}$ cm^-2 implanted sample via SPE; a thermally activated process with an intrinsic rate described by the Arrhenius relationship:

with $E_{A}=2.70$ eV and $v_{0}=4.64\times 10^{16}$ Å/s Roth et al. (1990). The SPE rate of the $a$-Si standard was calculated from the TRR curve to be 17.3 Å/s, whereas that of the 30 keV, $4\times 10^{18}$ cm^-2 implanted sample was 8.9 Å/s. The increased level of impurities Kennedy et al. (1977) and open-volume defects Zhu et al. (1999b) introduced by high fluence implantation slows the progression of the $a/c$ interface.

The isotopic enrichment level of the 30 keV, $4\times 10^{18}$ cm^-2 implanted sample was measured using SIMS after annealing. The left axis of Fig. 2a shows the high fluence 30 keV ²⁸Si implantation depleted ²⁹Si and ³⁰Si to concentrations of around 3000 ppm and 2000 ppm, respectively, in a surface layer $\sim$50 nm thick. The level of isotopic enrichment achieved with 30 keV is worse than that achieved with 45 keV, despite the higher implant fluence, as it is limited by self-sputtering of Si due to the sputter yield being greater than one. ³¹P^- was implanted into the ²⁸Si layer with a depth profile, simulated using SRIM, shown on the right axis of Fig. 2a. The maximum P concentration was confirmed to lie below the SIMS detection limit ($\sim 2\times 10^{17}$ cm^-2).

Pulsed ESR measurements were performed on the implanted P donors, as shown in Fig. 2b. The transverse relaxation time $T_{2}$ obtained using the Hahn echo pulse sequence gives $T_{2}=285$ $\mu$s $\pm 14$ $\mu$s, which is comparable to previous reports Chiba and Hirai (1972); Tyryshkin et al. (2003) and longer than the coherence time obtained in ${}^{\text{nat}}$Si for similar doping concentrations ($\sim$100 $\mu$s Chiba and Hirai (1972)). Moreover, the echo decay is mono-exponential which confirms that the instantaneous diffusion between the donor electrons is the dominant decoherence mechanism here, instead of the spectral diffusion caused by ²⁹Si nuclear spins which dominates in ${}^{\text{nat}}$Si and adds an additional cubic term to the exponent Chiba and Hirai (1972); Tyryshkin et al. (2003); Abe et al. (2010); Tyryshkin et al. (2012). These ESR measurements demonstrate that the P donors have been implanted into an isotopically enriched ²⁸Si environment and successfully activated. In future, the concentration of P and residual ²⁹Si can be reduced to improve coherence time.

To improve the enrichment process, TRIDYN simulations were performed. The sputter yield as a function of implantation energy was determined for an implantation of $1\times 10^{17}$ cm^{-2 28}Si ions at normal incidence, as shown in Fig. 3. The sputter yield dependence on implantation energy shown here is in agreement with previous experimental Si sputter yields Andersen and Bay (1975); Fröhlich et al. (1990) and theoretical fits Eckstein (2007). An energy of $<3$ keV results in the deposition of ²⁸Si onto the Si surface. If the sputter yield is greater than 1, the surface layer will be eroded faster than it can be isotopically enriched, resulting in a thin ²⁸Si surface layer with reduced enrichment. ²⁸Si ions with energies $>45$ keV are implanted deeper below the surface and sputtering is suppressed, resulting in accumulation. This is desirable for producing a thick layer of ²⁸Si with a high level of enrichment, however, the surface will not be planar. A sputter yield of 1 is achieved at energies around 3 keV and 45 keV, both of which result in a planar surface; desirable for surface nanocircuitry fabrication. 45 keV was selected in order to produce a ²⁸Si surface layer thicker than the qubit target depth of $\sim$20 nm in the one-for-one replacement regime and to optimize the transmission of the ion beam through the implanter. The sputter yield is independent of angle of incidence for angles below 10^∘ for self-implanted Si Fröhlich et al. (1990) and so the TRIDYN simulations performed here at normal incidence are applicable for our experimental implants performed with a 7^∘ tilt.

The depth profiles of Si isotopes in ${}^{\text{nat}}$Si after the simulated implantation of 45 keV ²⁸Si at a fluence of $5\times 10^{18}$ cm^-2 are shown in Fig. 4a. This shows that an isotopically enriched surface layer $\sim$100 nm thick is created. The resultant concentrations of ²⁹Si and ³⁰Si at a depth of 20 nm below the surface as a function of fluence of 45 keV ²⁸Si are shown in Fig. 4b. This shows the trend of an increased isotopic purity resulting from an increased implant fluence. The isotope concentrations at a depth of 20 nm realised in this work with an implantation of ²⁸Si^- ions (45 keV, $2.63\times 10^{18}$ cm^-2) as discussed below are indicated with star symbols in Fig. 4b.

The concentrations of Si isotopes and ¹²C and ¹⁶O impurities in the ²⁸Si^- (45 keV, $2.63\times 10^{18}$ cm^-2) implanted sample after annealing were measured with SIMS and are displayed as a function of depth below the surface in Fig. 5. Fig. 5a shows the high fluence ²⁸Si^- implantation depleted ²⁹Si and ³⁰Si to concentrations of around 250 ppm and 160 ppm, respectively, in a surface layer of thickness $\sim$100 nm. The shape of the isotope concentration profiles agree well with the TRIDYN simulation shown in Fig. 4a. A higher level of enrichment was achieved experimentally than predicted by TRIDYN, as shown by the star symbols in Fig. 4b. This could be due to the experimental sputter yield being slightly less than 1, leading to the accumulation of a thicker isotopically enriched layer, as evidenced by the depth where the isotope concentrations reach natural abundance: $\sim$180 nm for the TRIDYN simulation (Fig. 4a) and $\sim$220 nm for the experimental measurement (Fig. 5a). This accumulation was shown to result in lower ²⁹Si and ³⁰Si concentrations, demonstrated by TRIDYN simulations implanting $>$45 keV Si (not shown). The discrepancy in sputter yield, sensitive to the target surface binding energy, could be due to the impurity content of the substrate Eckstein (2007). A smaller contribution could come from the uncertainty in the experimental implantation fluence. The residual ²⁹Si concentration achieved here is around 3 times lower than that found in a commercially-produced ²⁸Si wafer (Isonics) which, with 800 ppm ²⁹Si Itoh and Watanabe (2014), has previously demonstrated increased coherence times of implanted donors Muhonen et al. (2014).

Fig. 5b shows the concentrations of ¹²C and ¹⁶O in an implanted region (solid lines) and in a non-implanted region of the substrate (dashed lines). The concentrations are increased above the background levels to around $1\times 10^{17}$ cm^-3 for C and $3\times 10^{17}$ cm^-3 for O by the process of high fluence implantation of ²⁸Si^- ions and subsequent annealing. The concentrations of these impurities were calibrated by assuming that the background levels at a depth of $\sim$300 nm, which match for the implanted and non-implanted regions, were $5\times 10^{15}$ cm^-3; the maximum expected background contamination for UHPS Topsil quoted by the supplier. An increase in impurity levels, significantly above the background level in the surface $\sim$30 nm for C and surface $\sim$20 nm for O, is present in the non-implanted substrate. This accounts for some of the near-surface impurity content in the implanted region. If native SiO₂, typically $\sim$2 nm thick, was present, TRIDYN simulations (not shown) confirm the majority would be sputtered away during high fluence implantation, resulting in negligible O contamination from this source. The Si aperture reduced forward recoils of impurities, with no trace of heavy metals detected with high-resolution Rutherford backscattering spectrometry (not shown). The majority of the C and O contamination is proposed to be incorporated into the implanted layer from the imperfect vacuum, as seen before with high fluence implantation in a cryopumped target chamber Singer and Barlak (1983). These levels of contamination are comparable to those present in Czochralski-grown Si ($4\times 10^{17}-2\times 10^{18}$ cm^-3 for O and $2\times 10^{16}-4\times 10^{17}$ cm^-3 for C Liaw (1981)) and indeed are shown to be low enough to allow for the successful crystallisation of the enriched layer by SPE. A peak in the concentration of C and O impurities occurs at around 190 nm below the surface of Si. This depth is significantly shallower than the depth of the end of range defects ($\sim$290 nm) visible with TEM as a dark band of dislocation loops, as shown in Fig. 6a. The peak in impurity concentration at an intermediate depth between the surface and the end of range could be associated with the presence of open volume defects arising from vacancy clustering in this region and invisible to TEM. These open volume defects have been observed to act as gettering sites for impurities during annealing Tamura et al. (1991); Brown et al. (1998). Additionally, Fig. 5b shows preferential diffusion of C and O towards the surface, known to be a vacancy-rich region after ion implantation Servidori et al. (1987).

The crystal quality of the ²⁸Si^- (45 keV, $2.63\times 10^{18}$ cm^-2) implanted sample after annealing was determined using cross-sectional TEM, shown in Fig. 6. End of range defects, visible as a dark band $\sim$290 nm below the surface in Fig. 6a, indicate the location of the $a/c$ interface before annealing. The $a$-Si layer is extended to greater depths during continued ion bombardment above the Si amorphisation threshold Claverie et al. (1988) (typically around $1\times 10^{15}$ cm^-2 for keV Si ions Mori et al. (2001)). The excess of interstitials at the end of range produced during ion implantation can evolve into dislocation loops during SPE regrowth of the $a$-Si layer Jones et al. (1988). These dislocation loops are stable up to temperatures of 1100 ^∘C Pan et al. (1997), whereby they release self-interstitials into the surrounding substrate. This could cause undesired transient-enhanced diffusion Stolk et al. (1997) of implanted P donor qubits in this enriched layer and so lower thermal budgets, supplied by low temperature SPE and rapid donor activation anneals, are preferred. Fast Fourier transforms (FFTs) of regions of the TEM image in Fig. 6b were taken to give diffraction patterns indicating the crystal structure of the lamella. The diffraction pattern for the implanted region, shown in Fig. 6c, indicates good crystal quality and matches that of the non-implanted $c$-Si substrate beneath, shown in Fig. 6d. This shows the success of the crystallisation during post-implantation annealing. The contamination level introduced during the high fluence implantation is therefore low enough to avoid the formation of a polycrystalline ²⁸Si layer, which would contain undesirable charge traps and dangling bonds at grain boundaries Seager (1985). We expect that this single crystal layer of isotopically enriched ²⁸Si will provide an ideal environment for implanted donor qubits, with high activation and long coherence times.

In conclusion, a Hahn echo measurement of P donors implanted into a ²⁸Si layer with $\sim$3000 ppm ²⁹Si, produced by high fluence implantation of 30 keV ²⁸Si^- ions, was fitted with a mono-exponential decay, suggesting an isotopically enriched ²⁸Si donor environment. The extracted coherence time of $T_{2}=285\pm 14$ $\mu$s is longer than that found with ${}^{\text{nat}}$Si for similar P concentrations. The residual level of ²⁹Si was further decreased by implanting 45 keV ²⁸Si^- ions in the one-for-one sputtering regime. A high fluence ($2.63\times 10^{18}$ cm^-2) implant of ²⁸Si^- ions at this energy into ${}^{\text{nat}}$Si results in a depletion of ²⁹Si down to 250 ppm in a surface layer of thickness $\sim$100 nm, as measured with SIMS. The drastically reduced concentration of ²⁹Si spin-1/2 nuclei in this isotopically enriched layer should further extend the coherence time of implanted donors beyond that achieved with commercial Isonics epilayers. Care was taken to limit the level of contamination introduced during high fluence implantation and concentrations were found to be below $1\times 10^{17}$ cm^-3 for C and $3\times 10^{17}$ cm^-3 for O, comparable to those in Czochralski-grown Si. The levels of contamination in this isotopically enriched $a$-Si layer are low enough to allow for successful crystallisation by SPE. The quality of the single crystal surface layer of ²⁸Si was shown to be equivalent to the non-implanted region of the $c$-Si substrate using high-resolution TEM, in which the end of range defects were still visible after annealing. This work shows the high fluence implantation of ²⁸Si^- ions at energies around 45 keV is an effective method for isotopic enrichment which could be incorporated in-situ into the fabrication of ion implanted donor spins in ²⁸Si for quantum devices with increased coherence times.