Transformation of the nickel donor center by annealing in silicon measured by deep-level transient spectroscopy

Before measuring the DLTS spectrum, the electric characteristics of the Schottky barrier diodes were examined. Since sound diodes were always formed without etching the surfaces of the samples, no nickel precipitate was considered to form on the surfaces of the samples; all the contaminating nickel was dissolved in the bulk and no outdiffusion of the element occurred during the cooling of the samples. As reported in a previous paper,²⁴⁾ the depth distribution of the nickel center was almost homogeneous throughout the depth (from 2.0 × 10¹³ near the surface to 1.5 × 10¹³/cm³ deep in the bulk). Since the bulk concentration of nickel is 3.8 × 10¹⁴/cm³ when all the contaminating nickel homogeneously dissolves, the concentration fractions of the nickel center to the dissolved nickel are estimated to be from 4 to 5%.

Both EL-formed and EL-free samples were stored for a long time in a clean room at room temperature. For the EL-free samples, no appreciable change in the spectrum was observed after storage for 3.5 years. Figure 1 shows the change in the DLTS spectrum of the same EL-formed sample with cumulative storage time. For the as-diffused sample, four peaks were clearly observed at approximately 85, 165, 260, and 320 K, which are denoted as H(85), H(165), H(260), and H(320), respectively, after their peak temperatures. The largest peak H(85), whose activation energy was 0.17 ± 0.01 eV as determined by thermal emission measurement,²⁸⁾ corresponds to the nickel donor center.^13–23) The species H(165) and H(260), whose activation energies were 0.27 ± 0.02 and 0.45 ± 0.02 eV, respectively, correspond to the hydrogenated nickel centers^20,23) that appeared after wet chemical etching (hydrogenation). The species H(320), whose activation energy was 0.61 ± 0.05 eV, has been assigned as the nitrogen-related defect originally included in the crystal,²⁹⁾ although its structure has not yet been identified. The spectral feature shown in Fig. 1(a) resembles the features in the samples diffused with deposited nickel film at high temperatures (≥950 °C) reported earlier.^15–18) The highest temperature peak observed by Lemke¹⁵⁾ at 300K (E₁ denoted by him), however, does not appear in this study and other reports.^16–18) After 1.0 year of storage, while the intensity of H(260) slightly increased, the intensities of H(85) and H(165) negligibly changed. After 3.5 years of storage, a further increase in the intensity of H(260), a slight decrease (<10%) in that of H(85), and a slight increase in that of H(165) were observed. Note that the increase in the concentration of H(260) is larger than the decrease in the concentration of H(85) after 3.5 years of storage.

Zoom In Zoom Out Reset image size

Figure 2 shows the change in the typical DLTS spectrum by annealing for the EL-free samples after 3.5 years of storage at room temperature. After annealing at 150 °C for 10 min, the decrease in the concentration of H(85) and the increases in the concentrations of both H(165) and H(260) were observed. After annealing at 200 °C, while the H(85) concentration further decreased, both H(165) and H(260) concentrations began to decrease. Two new peaks, H(125) and H(200), whose activation energies are 0.20 ± 0.02 and 0.39 ± 0.02 eV, respectively, appeared at this temperature. The appearance of these peaks was not reported in previous papers.^15–23) After annealing at 350 °C for 60 min, all the peaks disappeared.

Zoom In Zoom Out Reset image size

The detailed concentration changes of the peaks by isochronal (10 min) annealing are shown in Fig. 3. It is clearly seen that while the H(85) concentration continuously decreased with increasing annealing temperature, the concentrations of H(165) and H(260) independently changed. No recovery of the H(85) concentration after annealing was observed. Since the concentrations of H(125) and H(200) identically changed with annealing temperature, suggesting that they originate from the same defect, they are denoted as H(125/200). Since H(125/200) began to appear after the decays of H(165) and H(260), the former species is suggested to be the extended product of either of the latter species. Since no DLTS peak appeared after annealing at higher temperatures than 350 °C, insoluble species such as nickel precipitate is inferred to evolve above this temperature.

Zoom In Zoom Out Reset image size

Figure 4 shows the change in the DLTS spectrum by annealing of the EL-formed samples. After annealing at 115 °C for 20 min, while the H(85) concentration decreased, the H(165) concentration considerably increased. The small pair peaks H(125/200) appeared after this annealing. The species H(150) (activation energy: 0.20 ± 0.02 eV), which did not appear in the EL-free samples, appeared after annealing at 140 °C. This peak formed a maximum after annealing at 160 °C for 5 min and gradually decreased to disappear by annealing for a long time at this temperature. The concentration changes of these peaks after isochronal (5 min) annealing are shown in Fig. 5. The concentration change of each peak progressed within lower annealing temperatures and shorter times in the EL-formed samples than in the EL-free samples. All the peaks disappeared after annealing at 170 °C for 5 min. Note that the maximum intensities of H(125/200) were much smaller in the EL-formed samples than in the EL-free samples.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Since the concentration (N) of the nickel center showed an exponential decay with annealing time (t) by isothermal annealing (not shown) as in the case of the copper center,²⁷⁾ the decay time constant τ of the center is given by the equation

$$\begin{equation} N = N_{0} \exp \left(-\frac{t}{\tau}\right), \end{equation} \tag{ 1 }$$

in which N₀ is the initial concentration of the center before annealing. Figure 6 shows the Arrhenius plots of τ's for the EL-free and EL-formed samples, in which T_A is the absolute annealing temperature. For the EL-free samples, the τ values in log-scale vary linearly with reciprocal temperature between 150 and 350 °C. The activation energy obtained from the slope between these temperatures as shown by the broken line is 0.26 ± 0.01 eV. For the EL-formed samples, τ has a threshold at approximately 100 °C, sharply decreases with increasing temperature, and linearly decreases with reciprocal temperature above 125 °C. The activation energy obtained from the linear slope shown by the dotted line is 1.06 ± 0.02 eV, indicating that the decay reaction of the nickel center in the SCR is different from that in the bulk.

Zoom In Zoom Out Reset image size

A possible explanation of the annealing behavior of the nickel-related species observed in the different types of sample is discussed. Since H(165) and H(260) are hydrogenated substitutional nickel complexes (Ni_sH_x, x: integers) according to the analyses by Shiraishi et al.²⁰⁾ and Scheffler et al.,²³⁾ there is no doubt that hydrogen is included in the as-diffused samples. Because the wafer surfaces were covered or bonded with hydrogen after their immersion in the HF aqueous solution as stated in Experimental Methods, it is inferred that a considerable amount of hydrogen has been introduced into the bulk during the cleaning and diffusion processes of the wafers. A part of the neutral hydrogen is known to reside at the tetrahedral (T) interstitial site and diffuses between adjacent T sites with a small diffusion barrier (<0.2 eV), while most of the hydrogen resides at the bond center (BC) site in p-type Si.^30–32) Since all the contaminating nickel dissolves in the bulk and the total fraction of the measured nickel-related species of the dissolved nickel is small (<10%) as described earlier, a considerable fraction of the diffused nickel is assumed to be accumulated as Ni_i in the bulk at room temperature. Ni_i has no level in the bandgap and diffuses as a neutral species from a T site to adjacent T sites with the diffusion barrier of 0.21 eV estimated by first-principles calculation^33,34) (0.15 eV measured in a recent experiment³⁵⁾). Accordingly, hydrogen and Ni_i are the main impurities that react with the nickel center and other nickel-related species during annealing. Since the Fermi level of the sample in this study is low (E_v + 0.24 eV) at room temperature estimated from the ordinary semiconductor theory,³⁶⁾ a considerable amount of the nickel center is assumed to be positively charged (Ni_s⁺) in the EL-free samples.

First, the decay of the nickel center in the EL-free samples is considered. By comparing the low activation energy of the decay of the center obtained in this study with the binding energy of the center (2.68–2.60 eV) estimated by first-principles calculation,³³⁾ the direct dissociation of the center (Ni_s → Ni_i + V, V: vacancy) is difficult. The center is assumed to decay owing to the actions of hydrogen and Ni_i by annealing as

$$\begin{align} &\text{Ni$_{\text{s}}$} + x\text{H} = \text{Ni$_{\text{s}}$H$_{x}$}, \end{align} \tag{ 2 }$$

$$\begin{align} &\text{Ni$_{\text{s}}$} + \text{$y$Ni$_{\text{i}}$} = \text{Ni$_{\text{s}}$(Ni$_{\text{i}}$$^{*}$)$_{y}$}, \end{align} \tag{ 3 }$$

in which y takes integers and Ni_i^* denotes the bonding state of Ni_i, although its actual structure is not specified in this study. Since the increase in the total concentration of H(165) and H(260) from room temperature to 150 °C is comparable to the decrease in the concentration of the nickel center between these temperatures (Fig. 3), the decay of the center is assumed to be caused by its hydrogenation [Eq. (2)]. The decay activation energy of the nickel center obtained in this study (0.26 eV) is accordingly attributed to the required energy to bond free hydrogen to the nickel center. Since H(165) and H(260) began to decay at 200 °C with the continuous decay of the nickel center, its hydrogenation and the successive production of new species to annihilate Ni_sH_x are considered to occur. H(125/200) can be assumed to be one of these new products. Since the energy levels of this species do not correspond to any known levels of the hydrogenated species,^20,23) the former species should form by the bonding of Ni_i to Ni_sH_x. The newly produced species are then formally expressed as Ni_sH_x(Ni_i^*)_z, in which z takes several integers. When H(125/200) grows to electrically inactive products, such as precipitates, by the further bonding of Ni_i at higher temperatures, the disappearance of all the nickel-related species after annealing at or above 350 °C is explained. Since the nickel center is considered to be immobile at the examined annealing temperatures, the formation of complexes containing several Ni_s seems difficult.

Next, the concentration changes in the nickel-related species in the EL-formed samples are considered. Note that all the nickel-related species in the SCR tend to become neutral owing to the bending of the electronic bands of the Schottky diode³⁷⁾ and then the nickel center easily forms complexes with Ni_i at low temperatures. Since the increase in the concentration of H(165) from room temperature to 135 °C is as small as approximately 1/5 of the decrease in the concentration of the nickel center between these temperatures (Fig. 5), the major part of the center decays owing to the reaction (or reactions) other than that given by Eq. (2). When Ni_s(Ni_i^*)_y is assumed to be electrically inactive for small values of y, the decay of the center is mostly attributed to the reaction given by Eq. (3). The activation energy obtained in this study is therefore attributed to the required energy to bond free Ni_i to the center. The formation of a small amount of H(125/200) between 115 and 140 °C is suggested to be caused by the small decrease in the H(260) concentration, i.e., the former species is the product of the latter species and Ni_i. H(165) is considered to become electrically inactive after bonding Ni_i. Since H(150) appears only in the EL-formed samples, the species is produced independently of the hydrogen-containing species. When the electrically inactive Ni_s(Ni_i^*)_y becomes electrically active by further bonding several Ni_i to the species, H(150) can be regarded as this product. All the nickel-related species are considered to precipitate by the further bonding of Ni_i at or above 170 °C.

The increase in the concentration of H(260) without a corresponding decrease in the concentration of the nickel center in the EL-formed samples (Fig. 1) cannot be explained by the reaction given by Eq. (2). A probable explanation is that the formation of Ni_s and its subsequent hydrogenation easily take place at room temperature in the SCR through the action of a special impurity complex (or complexes), which contains V (or V's). Approximately 10¹⁶/cm³ carbon and 10¹⁵/cm³ oxygen are generally included in as-grown FZ silicon crystals.³⁸⁾ Since carbon is tightly bonded at the lattice site with a smaller volume than a silicon atom,³⁹⁾ it is generally accepted that the formation of a carbon–V complex is difficult (a carbon-self-interstitial complex is easy to form⁴⁰⁾). Oxygen (O)–V complexes, such as OV and O₂V, are estimated to accumulate at a maximum concentration of approximately 10¹²/cm³ in commercial Czochralski (CZ) crystals (oxygen concentration ∼10¹⁸/cm³) at room temperature.⁴⁰⁾ Since the concentrations of O–V complexes are estimated to be several orders lower in FZ crystals than in CZ crystals owing to the lower concentrations of oxygen in the former crystals, the concentrations of the complexes are insufficient to explain the measured concentration of the nickel center (∼10¹³/cm³). Accordingly, it is also difficult to assume that O–V complexes take part in the formation of Ni_s in the studied samples.

As reported in a preceding paper,²⁴⁾ the inclusion of approximately 10¹³/cm³ nitrogen complex was always observed in the sample used in this study. Several nitrogen defects are considered to contain a vacancy,^40–45) which assists in the formation of Ni_s from Ni_i, as suggested in previous papers.^23,24) Since the diffusion barrier of Ni_i is very low,^33–35) the diffusivity of this species is estimated to have a sufficient value to have the species reach a nitrogen complex at room temperature. The neutralization of all the species including the nitrogen defects in the SCR can facilitate the transformation reaction of Ni_i and hydrogenation, although the structure of the relevant defect and the path to transform Ni_i to Ni_s are not determined yet.