A Study of the Activation of Ion Implanted Phosphorus Donors in Silicon Using ESR

N. Bulatovica, W.D. Hutchisona, P.G. Spizzirrib, J.C. McCallumb, N. Stavriasb and S. Prawerb

aCentre for Quantum Computer Technology, School of PEMS,

The University of New South Wales at ADFA, Canberra ACT 2600. bCentre for Quantum Computer Technology, School of Physics, University of Melbourne,

Parkville Vic 3010.

Electron spin resonance is applied to Si:P ensembles, ion-implanted at various energies and fluences, and measured at ~5 K. The experimental spectra are compared with theoretical estimates based on local donor concentrations predicted by SRIM simulations. The outcomes are relevant to proposed Si:P quantum computing.

1. Introduction

Phosphorus doped silicon (Si:P) was proposed as a potential basis for a quantum computer (QC) due to its long spin coherence and compatibility with existing silicon processing technologies [1]. The use of low energy ion implantation during the “Top-Down” fabrication of Si:P solid state qubits relies upon efficient dopant activation that is not dependent upon the incident ion energy [2]. Issues such as dopant gettering at the silicon interface and ionisation by interfacial traps could impact upon the successful fabrication and operation of any such device.

In theory, electron spin resonance (ESR) is ideal for probing Si:P spin qubits and metering the success of the implantation and activation processes. In addition, the use of molecular P2

+ implants plus ESR potentially should provide a way of observing exchange coupling of donor pairs. However previous ESR experiments show very low ESR signal levels, possibly reflecting reduced donor activation for low (<20 keV) energy implants. In this work, we return to a very fundamental scenario and report on the application of ESR to measurements on phosphorus P+ implants into silicon using various implantation energies. The characteristic ESR spectral lines (and intensities) are used to identify phosphorus dopants and radicals (e.g. Pb) which constitute electronic trap states and identify some issues that still need addressing.

2. Sample preparation

Twelve sample combinations were prepared by P+ ion implantation. Wafer pieces (~10mm x 10 mm) of high resistivity silicon (<100>), p-type (Boron) originally supplied by Topsil were used with either a 2 nm native oxide or a 5 nm (nominal without the DCE furnace bubbler) thermal oxide. Implantation was carried out with a set area of 4.84 cm2 which is greater than the sample size at 2 different doses (1 x1013 cm-2 and 1 x1012 cm-2), and 3 different implantation energies (15, 40, 70 keV). A set up error occurred during the 40keV implantation of the thermal oxide samples which saw them implanted at 35keV instead. After the implantation the samples were cleaned and given a rapid thermal anneal, 1050°C for 5 s. Then the samples were cleaved (to <5mm x 10 mm for ESR), labelled and cleaned again. Native oxide samples were chemically etched (alcoholic HF etch) and all samples were packed into a glass vial under DI water for shipment (from Melbourne to Canberra) during which time a “water oxide” was grown .

The ESR experiments were carried out at ~5 K using a Bruker ESP300 spectrometer with a standard x-band cavity coupled to a flow cryostat. Si:P wafers, were inserted to the cavity on the end of a silica glass rod. A field sweep of width 0.01 T (100 G) about a centre field of 0.3380 T (3380 G) was employed. Since for Si:P, g = 1.99875 ± 0.00010 [3], the phosphorus (P) donor hyperfine split doublet lines are close to centred with a microwave frequency of ~9.45 GHz. There are long spin-lattice relaxation rates (T1) at 5 K [4]. White light from a halogen lamp was focused into the cavity, this shortens T1 considerably by interaction of the donor spins with photo-electrons. It has been noted, previously, that this does not unduly alter the spectra produced but, makes for better signal-to-noise (s/n) than for the alternative, which would be measurements in the dark at ~15 K.

3. Results

The resulting ESR spectra for the 12 implanted samples are presented in Figures. 1 and 2. The resonance lines have been fitted with differential Lorentzians and thus identified according to their g-factor. The dominant line in all of the spectra is the Pb electron trap (dangling bond in SiO2) at ~3365 G (g ~ 2.006). A hyperfine split P doublet (42 G splitting centred at 3380 G or g =1.99875) is also present in all cases. A central line due to exchange

Fig. 1. ESR spectra and fits for the samples with ~2 nm native oxide: a) implanted with 1x1013 cm-2 fluence and

energies of 15, 40 and 70 keV, b) implanted with 1x1012 cm-2 fluence and the same energies.

Fig. 2. ESR spectra and fits for the samples with ~5 nm thermal oxide: a) implanted with 1x1013 cm-2 fluence

and energies of 15, 35 and 70 keV, b) implanted with 1x1012 cm-2 fluence and the same energies.

coupled P pairs or P clusters at g = 1.99875 is also observed in some cases. The integral signals (i.e. double- integrated signal of the differential absorption ESR spectra), based on the fits to these lines are summarised in Table 1. While in Table 2, experimental P signals (as a percentage of the total at 70 keV) are compared with a simple theoretical prediction of the spectra based on donor concentrations assigned by the SRIM simulations [6] of Figure 3. The curves of this figure are calculations of the local P donor concentrations for the different implant energies and have been used to estimate the fraction of donors that fall into each of

Table 1. Summary of experimentally determined integral ESR signals.

Integral SignalOxide D (cm-2) E (keV) P1 P2 Central Pb

15 2768.655 2055.965 350 97007.51Native 1x1013 40 3932 3994 24776 351402*

70 3892.452 3892.452 23773.56 130825.9

15 1600.74 1600.743 200 84058.89Native 1x1012 40 1200.901 3501.102 154181.8

70 5504 3559 154004

15 1300.403 1299.911 26800.6 113487.3Thermal 1x1013 35 1403.357 3195.852 34963.19 145901.5

70 3603.461 3402.655 30556.15 53346.01

15 1100.509 1600.29 103258.2Thermal 1x1012 35 1601.586 1951.465 169992

70 3604.444 3604.444 801.189 86304.71Native Blank 96500

* Possible extra damage from handlingNotes: “D”: implantation fluence; “E”: Implantation energy; “P1”: Integral signal of the “first” P doublet line; “P2”: Integral signal of the “second” P doublet line; “Central”: Integral signal of the exchange coupled pairs and clusters; “Pb”: Integral signal of the oxygen radical (trap) in SiO2.

Table 2. Comparison of experimental data and theoretical allocations for the P donor signals.

Exp. Results Theor. ResultsOxide D (cm-2) E (keV) Doublet S% Central S% Doublet S% Central S% %Lost in O

15^ 15.29 1.11 7 93Native 1x1013 40 # 24.24 75.76 14 86

70 24.66 75.33 33 67

15^ 35.32 2.21 57 43Native 1x1012 40^ 51.88 0 69 31

70 # 100 0 100 0

15^ 6.92 71.35 5 93 2Thermal 1x1013 35 11.62 88.37 14 86

70 9.59 90.41 33 67

15^ 33.72 0 54 43 2Thermal 1x1012 35^ 44.36 0 69 31

70 90 10 100 0

^ Normalised against 70 keV ESR signal

Fig. 3. SRIM simulations [6] of P implants at various energies into silicon with a 5 nm thermal oxide.

three categories (described below), for each implant energy and fluence. Theoretical spectra are then constructed as follows: Donors > 20 nm apart (i.e. concentration < 1.25 x 1017 cm-3) are considered isolated and assigned to an ESR doublet signal. Those < 11 nm (i.e. concentration > 7.0 x 1017 cm-3) are considered clustered (i.e. hyperfine interaction collapsed) yielding a single line [5]. While donors at intermediate separations are assumed to be exchange coupled pairs and are assumed to give a triple line spectrum. 4. Discussion

The overwhelming feature of the ESR spectra presented here are the Pb signals. This is indicative of high density of charge traps and although it might be expected for silicon covered with native oxide, here the Pb signal is also large in this batch of thermal oxides. This is due, in part, to implantation damage, but also indicates that the oxide is not of sufficient quality for QC devices. An over abundance of charge traps are not desirable if donors are to be placed near the surface.

Indeed we find that the ESR signal from the lower energy implants, as the donors are placed nearer the interface (mean depth 25 nm at 15 keV), is severely diminished. This is especially true for the central line (exchange coupled or clustered donors). The loss of P donor signals is more pronounced in the native oxide capped samples where there are a higher density of charge traps and surface defects. Overall there is a strong indication that donor electrons are being poached by traps where the donors are placed near to the surface (i.e. donors effectively compensated by traps). This represents a significant challenge for quantum computer engineering, using the ‘Top Down’ approach, (i.e. ion implantation) and best quality oxides will be required as a minimum starting point.

Acknowledgments

This work was supported by the Australian Research Council, the U.S. National Security Agency and the Army Research Office under contract DAAD19-01-1-0653. References [1] B.E. Kane, Nature, 393 (1998) 133. [2] R.G. Clark et. al., Phil. Trans. R. Soc. Lond. A 361, (2003) 1451-1471. [3] G. Feher, Phys. Rev., 114 (1959) 1219 [4] G. Feher and E.A. Gere, Phys. Rev., 114 (1959) 1245. [5] S. Maekawa and N. Kinoshita, J. Phys. Soc. Japan, 20 (1965) 1447. [6] J.F. Ziegler, J.P. Biersack and U. Littmark “SRIM - The Stopping and Range of Ions in

Solids,” Pergamon Press, New York, (1985).