Table of contents

This issue contains a selection of papers presented at the First International Workshop of the European project entitled Coordination Action on Defects Relevant to Engineering Silicon-Based Devices (CADRES) held in Catania, Sicily, 26--28 September 2004. The CADRES project is sponsored by the European Commission in the Framework 6 IST programme.

The Workshop was attended by about 107 delegates, from many European countries, who heard presentations from speakers prominent in their fields from all over the world, plus several excellent student presentations. Over the three days there were opportunities for very focussed discussion, and all who attended could benefit from new collaboration and training opportunities available as a result of this meeting.

I would like to thank the local organizers, Professor Francesco Priolo and his students for the smooth running of the workshop, and Professor Bengt Svensson for acting as the Programme Chairman. I would also like to thank Professors Svensson and Priolo for their help with the selection of papers for the workshop and with the Proceedings.

Hybrid functional calculations within density functional theory are carried out to investigate the electronic structure of boron-interstitial clusters (BICs). A one-parameter hybrid functional is chosen is to give accurate results for the whole electronic structure (including the gap) and the elastic properties of crystalline silicon. It is shown that this approach provides dependable defect level positions in the gap. Investigation of the boron+vacancy and boron+self-interstitial centres gives a consistent description of the experimentally observed G10 and G28 centres. The electronic structure of BICs, which may affect the activation rate of boron implantation, are reported. The one-electron level positions of isolated B_nI_m defects are given.

The local vibrational modes arising from single interstitial hydrogen centres in Si, Si-rich SiGe, Ge-rich SiGe, and Ge crystals are modelled by an ab initio supercell method. The stress response of the 1998 and 1794 cm⁻¹ bands that appear in proton-implanted Si and Ge samples is well reproduced, further confirming their assignment to bond-centred H⁺ defects. It is shown that H⁻ in Ge is anti-bonded to a Ge atom, and is likely to be considerably less mobile than in Si. Although H⁺ is not trapped by the minority species in both Si-rich and Ge-rich alloys, we find that H⁻ can be stabilized by forming anti-bonded H–Si structures.

Using ab initio density functional theory the configuration space of indium clusters in silicon up to size 4 is explored. The strongest binding energy corresponds to a cluster containing one indium and three self-interstitials. Two plausible configurations almost degenerate in energy are found: the model and an alternative which combines two split di-interstitials. The segregation of indium to a InI₄ cluster is found to capture the dual peak and the low temperature annealing behaviour of indium.

C_i and C_iC_s are two main defects produced via the Watkins replacement mechanism when Si-based materials containing carbon are subjected to particle irradiation. In the present article we re-examine our experimental observations reported on these defects in relaxed Si_1−xGe_x alloy layers in the light of very recent and thorough theoretical investigations (Venezuela et al 2004 Phys. Rev. B 69 115209 and Balsas et al 2004 Phys. Rev. B 70 085201). In addressing these defects two main issues are taken up. Firstly, the role of alloying upon the position of the electrical levels in the bandgap is considered: a linear shift towards the valence band with the same rate of all related levels is observed experimentally and predicted theoretically. Secondly, the dynamics of migration of interstitial carbon (C_i) in the process of forming C_iC_s is analysed. Our suggestion that C_i migrates via Si-based paths has now found theoretical justification.

Impurity-related electronic states in high-purity high-resistivity n-type float-zone (FZ) Si have been studied by deep level transient spectroscopy (DLTS). FZ-Si with a doping concentration of ∼1 × 10¹² cm⁻³ was chemically etched in 10% diluted hydrofluoric acid (HF) prior to thermal deposition of a Schottky gold contact. DLTS measurements of the as-prepared diode reveal the presence of a deep electronic trap at 0.52 eV below the conduction band. Depth profiling measurements show that the trap concentration increases towards the surface, indicating that it is due to an impurity diffusing from the surface. Heat treatments for 15 min at 150–300 °C lead to annealing of the 0.52 eV trap and formation of a 0.17 eV trap that is uniformly distributed versus depth. A reverse formation of the 0.52 eV trap with a uniform depth distribution, accompanied by annealing of the 0.17 eV trap, is observed after storage of the sample at room temperature for 20–50 days. A subsequent heat treatment at 300 °C results in the repeated annealing of the 0.52 eV trap and formation of the 0.17 eV trap, which is followed by the reversed formation of the 0.52 eV trap and annealing of the 0.17 eV trap at room temperature (RT). Several such annealing/formation cycles can be performed with these centres with a good reproducibility of the data. It might be suggested that both centres are related to hydrogen, although the exact origin is not identified.

Energy transfer to free carriers in an Auger process is well known to hamper emission of rare-earth dopants in semiconductors. In particular, this process limits the excitation mechanism and is partly responsible for the thermal quenching of the ∼1.5 µm photoluminescence from Er³⁺ ions embedded in the crystalline silicon matrix. In this contribution, we investigate the excitation cross section and the free-carrier Auger process in Er-doped silicon multinanolayer structures. This novel Si-based material has recently been shown to exhibit very interesting properties as regards photonic applications.

The electrical activity of defects present in strained silicon (SSi) on thin strain-relaxed Si_1−xGe_x buffer layers (SRBs) is evaluated using deep submicron CMOS compatible n⁺/p and p⁺/n shallow junctions. Strain relaxation has been achieved here by introducing a thin carbon-rich layer in an otherwise uniform Si_0.78Ge_0.22 epitaxial layer, resulting in an SRB thickness of 248 or 348 nm and processing compatible threading dislocation densities in the range of a few 10⁶ cm⁻². From a combination of electrical measurements (current– and capacitance–voltage; microwave absorption (MWA) recombination lifetime) and microscopic techniques (electron-beam-induced current; emission microscopy), it is concluded that generation centres associated with the C layer can play an important role. Their electrical activity is shown to depend strongly on the relative position of the C-doped layer with respect to the junction depth. The type of well dopant (implantation) also has a strong impact on the electrical activity of the different defect types present in the epitaxial layers. It is generally found that the 348 nm junctions show a lower reverse current at practical operation temperatures and voltages, while the p⁺/n diodes exhibit a better performance compared with their n⁺/p counterparts.

Using measured isotope shifts for the 607 cm⁻¹ local vibrational mode, LVM, of substitutional carbon, C_s, we demonstrate that isotope disorder contributes ∼0.5 cm⁻¹ to the width of that LVM in natural silicon. The width of the LVM of C_s also depends on its energy relative to the density of two-phonon states, and increases along the sequence ¹³C in natural silicon, ¹²C in natural silicon and ¹²C in ³⁰Si. Other LVMs show larger isotope effects, and so discrete structure rather than line broadening. In the case of zero-phonon lines, we take the 3942 cm⁻¹ line as a potentially favourable example. We estimate that the isotope disorder contributes only 0.09 cm⁻¹ to the linewidth in a natural silicon sample, a contribution that is negligible compared to typical strain-broadening effects, but would be a lower limit to the linewidth in high-purity natural silicon.

Deep level transient spectroscopy (DLTS) and high resolution Laplace DLTS (LDLTS) have been applied to p-type Czochralski silicon that contains dislocations that have and that have not been locked by oxygen. The stress-induced dislocations have been immobilized by oxygen during heat treatment, which prohibits glide under certain applied shear stresses. The DLTS spectra show typical broad features between 100 and 320 K, characteristic of those seen in other dislocated silicon reported in the literature, and several components are present in the LDLTS spectra. In addition, DLTS spectra show a sharp narrow peak at 40 K at a rate window of 200 s⁻¹ in the case of the locked dislocations, but not in the case of the sample where there is no oxygen locking. LDLTS shows that this deep level consists of more than one component and it is proposed that this peak is likely to be due to electrical activity associated with oxygen at the dislocation core. For hole emission at temperatures above 100 K, in the sample with unlocked dislocations, LDLTS detects a change of the emission rate of the carriers from some, but not all, of the components of the broad peak when the LDLTS fill pulse length is changed. This change is ascribed to band edge modification as the electronic states associated with the dislocation charge up during the fill pulse, and causes local electric field-driven emission of trapped charge during the reverse bias phase of the measurement. The LDLTS features which remain constant with fill pulse are proposed to be due to point defects in the material, which are not physically near dislocations.

Electrically active defects induced by neutron irradiation in n-type Czochralski-grown (Cz) Si crystals have been studied by means of capacitance transient techniques. These neutron-induced defects are compared with those created by electron irradiation and self-ion implantation. Four electron traps with the activation energies for electron emission of 0.12, 0.16, 0.24 and 0.42 eV were observed after neutron irradiation in phosphorous-doped Cz Si crystals. It is inferred that the E(0.12) and E(0.16) traps are related to the single-acceptor states of the silicon self-interstitial–oxygen dimer complex (IO_2i) and the vacancy–oxygen pair (VO), respectively. The E(0.24) trap is associated with the electron emission from the double-acceptor state of the divacancy (V₂). However, an asymmetric peak with its maximum at around 220 K and an activation energy for electron emission of 0.42 eV dominated the spectra. We used high resolution Laplace DLTS to investigate the structure of E(0.42) and found that this signal is complex, consisting of contributions from several defects. From the annealing behaviour, it was revealed that as some of these defects anneal out they are sources of vacancies evidenced by an increase in the concentration of VO and V₂. It is suggested that some of the defects contributing to the E(0.42) peak are related to small vacancy clusters.

We have used local vibrational mode (LVM) spectroscopy to monitor the formation of oxygen-related thermal double donors (TDDs) at 450 °C and their annihilation at 650 °C in carbon-lean Czochralski-grown (Cz-) Si crystals. A few samples were treated at 650 °C under high hydrostatic pressure. It is found that the annihilation of TDDs at 650 °C results not only in a partial recovery of the interstitial oxygen, but also in the appearance of a number of new O-related LVM bands in the range 990–1110 cm⁻¹. The positions of these lines and their shapes are identical to those observed for Cz-Si irradiated with electrons or neutrons and annealed at 600–700 °C. Since the lines appear upon annealing out of V O₃ and V O₄ defects in irradiated samples, they are suggested to arise from V O_m (m>4) complexes. In both kinds of samples, pre-annealed and pre-irradiated, the new LVM bands disappear upon prolonged annealing at 650 °C while enhanced oxygen precipitation occurs. The V O_m defects are suggested to serve as nuclei for oxygen precipitates developing at around 650 °C. High hydrostatic pressure is found to enhance further (up to 4–5 times) the oxygen precipitation process at 650 °C in the samples pre-annealed at 450 °C.

A deep level transient spectroscopy (DLTS) study of electrically active defects in electron irradiated silicon detectors has been performed. Two types of materials have been studied and compared: carbon-lean magnetic Czochralski (MCZ-) Si, and high purity, diffusion oxygenated float-zone (DOFZ-) Si. In both materials we observed an earlier reported shift in position of peaks associated with the divacancy (V₂) at 250–325 °C, indicating a gradual transition from V₂ to the divacancy–oxygen complex (V₂O). Heat treatments at higher temperatures reveal a difference in annealing behaviour of defects in DOFZ- and MCZ-Si. It is observed that VO and V₂O anneal with a higher rate in DOFZ-Si. The appearance of a hydrogen related level only in the DOFZ-Si reveals a small presence of H and it is suggested that the difference in annealing behaviour is due to defect interaction with H in the DOFZ-Si. Our findings also suggest that dissociation may be a main mechanism for the annealing of V₂O in MCZ-Si.

n-type Czochralski silicon was doped or co-doped in the melt with various group IV elements (Sn, C, Pb) and has been irradiated with 1 MeV electrons to a fluence of 1 × 10¹⁶ cm⁻². The irradiation-induced electrically active defects have been studied by deep level transient spectroscopy (DLTS). It is shown that while Sn is an efficient vacancy trap, leading to the formation of SnV centres, no specific Pb-related deep levels have been found in the upper half of the bandgap. The dominant electron trap is the A centre, while similar concentrations of SnVs are formed in Sn- and Pb+Sn-doped n-Cz material. A number of as yet unidentified deep levels with smaller concentrations has also been observed, together with some grown-in peaks, whereof some could be hydrogen or carbon and lead related.

In this study the alloy effect for iron-related defects in p-type unstrained B-doped Si_1−xGe_x (0<x<0.071) bulk crystals grown by the Czochralski technique have been analysed. This effect has been studied as the defect level splitting observed by the use of high-resolution Laplace DLTS. Two configurations of iron have been compared: isolated interstitial and paired with boron. For the interstitial configuration the alloy pattern suggests that the first- and second-nearest neighbours equally influence the defect energy level. For the iron–boron pair the observed DLTS signal is due to the ionization process of iron, and thus the alloy pattern represents the siting of iron in the more dilute samples (Si_0.98Ge_0.02), while for alloy compositions containing more Ge the signal comes from the ionization of boron, and so the observed pattern represents the siting of boron.

The displacement of B from substitutional lattice sites during irradiation with a high energy (650 keV) proton beam is measured by channelling analyses along the and using the ¹¹B (p,α)⁸Be reaction. The normalized B yield, χ, increases with the ion fluence and saturates at a value (χ_F<1) that depends on the channelling axis, being minimum for channelling along . Therefore, displaced B is not randomly located in the lattice. The displacement rate is shown to be consistent with a model involving Si interstitial–B interaction and to depend on the local Si self-interstitial production rate, rather than long range interstitial migration. This was demonstrated by comparing results from samples with and without a Si layer containing 1 at.% C (known to be a trap for interstitial Si) interposed between the B doped layer and the substrate. The B displacement rate does not change in this sample indicating that self-interstitials produced at the end of the range do not play a role in this process. It is therefore concluded that only the Si interstitials produced in the B doped layer contribute to the B displacement and we can fit the damage rate with the following formula: χ = χ_F− [χ_F− χ₀]^*exp(−σ^*N_I), where χ₀ is the χ of the non-irradiated sample and N_I is the number of (Si interstitials) cm⁻² calculated by TRIM, with σ∼10⁻¹⁶ cm².

Substitutional impurities (B, Ga) in Si experienced an off-lattice displacement during ion-irradiation using a H⁺ or He⁺ beam at room temperature in random incidence. Samples were prepared by solid phase epitaxy (SPE) of pre-amorphized Si subsequently implanted with B and Ga at a concentration of about 1 × 10²⁰ at. cm⁻³ confined in a 300 nm thick surface region. The lattice location of impurities was performed by a channelling technique along different axes (, ) using the ¹¹B(p,α)⁸Be reaction and standard RBS for B and Ga, respectively. The normalized channelling yield χ of the impurity signal increases with the ion fluence, indicating a progressive off-lattice displacement of the dopant during irradiation in random incidence, until it saturates at χ_F<1, suggesting a non-random displacement of the dopant. In particular, at saturation the off-lattice displacement of B and Ga was investigated by angular scanning, revealing different positions for each dopant. This effect has been related to the interaction of impurities with the Si self-interstitials (Si_I) generated by the impinging beam in the doped region.

A theoretical modelling of the oxygen diffusivity in silicon and germanium crystals both at normal and high hydrostatic pressure has been carried out using molecular mechanics, semiempirical and ab initio methods. It was established that the diffusion process of an interstitial oxygen atom (O_i) is controlled by the optimum configuration of three silicon (germanium) atoms nearest to O_i. The calculated values of the activation energy ΔE_a(Si) = 2.59 eV, ΔE_a(Ge) = 2.05 eV and pre-exponential factor D₀(Si) = 0.28 cm² s⁻¹, D₀(Ge) = 0.39 cm² s⁻¹ are in good agreement with experimental ones and for the first time describe perfectly the experimental temperature dependence of the O_i diffusion constant in Si crystals (T = 350–1200 °C). Hydrostatic pressure (P≤80 kbar) results in a linear decrease of the diffusion barrier ( for Si crystals). The calculated pressure dependence of O_i diffusivity in silicon crystals agrees well with the pressure-enhanced initial growth of oxygen-related thermal donors.

The thermal stability and electronic properties of the vacancy–donor complexes, often referred to as the E centres, have been studied in silicon, unstrained silicon–germanium and pure germanium. The E centres have been introduced by electron irradiation or gamma rays. In silicon, Laplace deep level transient spectroscopy has been used to separate the E centre emission from the di-vacancy, thus enabling very reliable data to be obtained for the vacancy complexes with P, As and Sb. In pure Ge only the E centres associated with P and Sb are reported and in Ge rich SiGe only V–P. In all the samples measured the thermal stability of V–Sb has been found to be significantly higher than V–P. With regard to the energy levels, the activation energy of electron emission from the single-acceptor level of the E centre in silicon are for V–Sb 0.40 eV and for V–P 0.46 eV. For the pure Ge case, the single acceptor is a hole trap with emission to the valence band having energies for V–P of 0.35 eV and V–Sb of 0.31 eV. Similar values are found for Ge rich SiGe. The double-acceptor state is not seen in silicon but in germanium produces a state with an activation energy for electron emission of 0.30 eV for V–P and 0.38 eV for V–Sb. This is also reflected in the Ge rich alloys of SiGe:P that have been measured in this work.

After heat treatment, silicon samples implanted with high doses of hydrogen exhibit blistering and defoliation of thin silicon layers. The process is used commercially in the fabrication of thin silicon-on-insulator layers (Smart Cut^®). In the present study we investigate the behaviour of hydrogen after different processing steps, which lead to thin Si layers bonded to glass substrates. A set of hydrogen implanted samples is studied by means of low temperature photoluminescence, Raman spectroscopy, x-ray diffraction and optical microscopy (visible and infrared). The formation of Si–H bonds is detected after implantation together with a build-up of internal strain. After annealing, the relaxation of the implanted layers is found to be connected with the formation of hydrogen saturated vacancies and the formation of H₂ molecules filling up larger voids. A comparison is made with hydrogen plasma treated samples, where well defined platelets on {111} planes are found to trap hydrogen molecules. No direct evidence of the role of {111} and {100} platelets in the blistering process is found in the implanted layers from our study. We determine considerable compressive stresses in the bonded Si layers on glass substrates. The photoluminescence is strongly enhanced in these bonded layers but red-shifted due to a strain reduced band gap.

The application of high-resolution x-ray diffraction for detecting and distinguishing defects in SiGe(C) layers is presented. A depth profile of the defects in SiGe/Si multilayers has been performed by using high-resolution reciprocal lattice mapping at different asymmetric reflections. Transmission electron microscopy was also applied in order to observe defects in the layers and these results were linked with the x-ray analysis. The substitutional C or B concentration in SiGe was measured by the shift of layer peak compared to the intrinsic layers. The thermal stability of the SiGe layers was investigated in order to rank the epitaxial quality of the SiGe below the detection limit of x-ray technique. It has also been demonstrated that x-ray analysis can be used for in-line process monitoring of layers grown in small device openings on patterned substrates. These types of analysis have also been used routinely for the evaluation of processed samples.

The depth distribution of open-volume point defects created by room temperature implantation of Cz silicon by 100 keV B⁺ ions at a dose of 5 × 10¹⁴ cm⁻² has been determined by enhanced-resolution beam-based positron annihilation spectroscopy (PAS). By incremental controlled etching (via anodic oxidation of 50–100 nm layers) the depth resolution of the PAS is maintained at ∼50 nm by using positrons implanted at energies below 2 keV to probe each layer as it brought close to the surface by the etching process. The etch depths have been verified by using secondary-ion mass spectrometry to profile the boron depth distribution. The results are in good agreement with Monte Carlo simulations, particularly in the traditionally difficult-to-measure deep tail region.

Electrically active defects induced by electron irradiation in Czochralski (Cz)-grown Si crystals with low carbon content (N_C≤2 × 10¹⁵ cm⁻³) have been studied by means of Hall effect measurements, deep level transient spectroscopy (DLTS) and high-resolution Laplace DLTS (LDLTS). It has been found that in n-type carbon-lean Cz-Si irradiated at room temperature a centre with an acceptor level at E_c−0.11 eV (E_0.11) is one of the dominant radiation-induced defects. This centre is not observed after irradiation in Cz-Si crystals with N_C> 10¹⁶ cm⁻³. The E_0.11 trap anneals out in the temperature range 100–130 °C with the activation energy 1.35 eV.

In p-type Cz-Si crystals with low carbon content and boron (N_B≤2 × 10¹⁴ cm⁻³) one of the dominant radiation-induced defects has been found to be a bistable centre with an energy level at E_v+0.255 eV (H_0.255). It has been inferred from the analysis of temperature dependences of electron occupancy of this level that it is the E(0/++) level of a defect with negative Hubbard correlation energy (negative U). The activation energy for hole emission from the doubly positively charged state of the H_0.255 centre has been determined as 0.358 eV from LDLTS measurements.

It is argued that the E_0.11 and H_0.255 energy levels are related to a complex incorporating an oxygen dimer and Si self-interstitial.

Thermal treatments of Czochralski-grown Si at T = 450, 600 and 650 °C, under high hydrostatic pressure of P≈11 kbar, introduce thermal donors and various structural defects, as for example oxygen precipitates. Neutron irradiation of such samples results first in the formation of oxygen–vacancy complexes, mostly VO defects. Upon annealing, the VO defects evolve in larger clusters via the accumulation of oxygen atoms and vacancies in the initial VO core, leading to the formation of V_mO_n defects. We focus on the study of the effect of pre-treatments on the production and evolution of the various V_mO_n defects upon isochronal annealing. The observed changes and variations in the IR spectra and the evolution curves in comparison with the corresponding ones of an initially untreated sample are discussed and some explanations are offered. The most important finding of this work is that the concentrations of the VO₂ and the VO₃ defects are reduced in the sample pre-treated at 450 °C, an indication of interaction between thermal donors and radiation-induced defects.

Many of the defects created by ion implantation into silicon can be detected by electron paramagnetic resonance (EPR). The main defects observed are isolated point defects such as Si-P3 centres (neutral 4-vacancies), silicon dangling bonds in amorphous silicon and defects associated with the so-called Σ resonance—a single broad anisotropic line; particular attention is paid to the latter defects as they have the highest population over a wide fluence range. This paper reports the effects on the nature and population of these defects of changing the fluence, mass and energy of the implanted ions as well as the effects of changing the implantation temperature.