Correlated annealing and formation of vacancy-hydrogen related complexes in silicon

We report on a deep level transient spectroscopy study of annealing kinetics of a deep, vacancy-hydrogen related level, labeled E5*, at 0.42 eV below the conduction band in hydrogen-implanted n-type silicon. The E5* annealing correlates with the formation of another commonly observed vacancy-hydrogen related level, labeled E5, at 0.45 eV below the conduction band. The annealing of E5* and the formation of E5 exhibit first-order kinetics with an activation energy of 1.61  ±  0.07 eV and a pre-factor of ~1013–1014 s−1. The pre-factor indicates a dissociation or structural transformation mechanism. The analysis of electron capture cross-sections for E5* and E5 reveals considerable transition entropies for both states and a temperature dependent capture cross-section for E5*. Two possible identifications of E5* and E5 are put forward. Firstly, E5* can be attributed to V2H2(−/0) or V2H3(−/0), which dissociate with the emission of VH (E5). Secondly, E5* and E5 can be assigned to two different configurations of V3H.


Introduction
Hydrogen (H) is probably the most common and, at the same time, controversial impurity in silicon. It can be found in a form of atomic interstitial species (H i ) [1], dimers (H * 2 ) [2] and molecules (H 2 ) [3]. Diffusion of different hydrogen species and their interactions with defects are still not fully understood (see [4] and references therein). Hydrogen is also a crucial impurity in silicon technology due to the ability to passivate dangling bonds. Moreover, recently there have been observed indications that hydrogen may have an effect on the so-called light-induced degradation of solar cells [5][6][7].
Recently, we have reported on a hydrogen-related level, labeled as E5 * [8]. The level forms during heat treatments in the temperature range 75 °C-95 °C and has a position at E c − 0.42 eV (E c being the conduction band minimum) and an apparent capture cross-section (CCS) of 4 × 10 −17 cm 2 . The CCS measured directly by filling pulse variation is found to be ~6 × 10 −18 cm 2 . This reveals a relatively high transition entropy for ionization of E5 * : ∆S/k ≈ 2, where k is Boltzmann constant. Such a high ∆S can indicate a complex process involving structural changes.
A detailed annealing study of the implantation-induced defects at 75 °C-95 °C observed two processes with different rates [8]. The process with a faster rate is related to the dissociation of phosphorus-hydrogen (P-H) pairs and formation of vacancy-oxygen-hydrogen (VOH) and divacancyhydrogen (V 2 H) complexes by reactions with H i released from P-H. The second process is slower by a factor of 4, and is associated with formation of E5 * . It has been observed that formation of E5 * correlates with annealing of the deep level transient spectr oscopy (DLTS) peak, labeled E4, that consists of overlapping contributions from single acceptor states of divacancy, V 2 (−/0), and divacancy-hydrogen, V 2 H(−/0). The correlation between E5 * and the double-acceptor state of divacancy, V 2 (=/−), has not been observed, and V 2 has been ruled out as a precursor for E5 * . Thus, it has been concluded that the precursor for E5 * is likely to be V 2 H.
The activation energies for the fast and slow rates have been found to be similar: 1.05 ± 0.04 eV and 1.10 ± 0.02 eV, respectively; with the pre-exponential factor for the fast rate in the confidence interval (2 ÷ 25) × 10 11 s −1 and for the slow one in the confidence interval (3 ÷ 14) × 10 11 s −1 . The difference in the deduced values lies close to the experimental uncertainty, and one can not conclusively claim the nature of the difference between the two rates. However, a small barrier (~0.05 eV) for interaction between V 2 H and H i or H * 2 has been suggested [8].
In the present work, we investigate possible identification of E5 * by studying the annealing kinetics. A series of isochronal and isothermal annealing is performed. We observe that the annealing kinetics of E5 * is consistent with a dissociation mechanism or structural transformation and deduce the activation energy. We also find an anti-correlation between E5 * and a commonly observed hydrogen-related state, labeled E5. The origins of both E5 * and E5 are discussed.

Experimental details
The samples in present work were phosphorus (P) doped n-type Czochralski-grown silicon described in our previous investigation [8]. The P concentration was derived from capacitance-voltage profiles to be 1.2 × 10 14 cm −3 . The oxygen and carbon concentrations were found by Fourier transform infrared spectroscopy to be 7 × 10 17 cm −3 and below 5 × 10 15 cm −3 , respectively. The wafers were chemically cleaned by standard RCA solutions, and then a dilute hydrofluoric acid was utilized to remove a native oxide layer. Schottky contacts were produced by 150 nm palladium deposition using a circle shadow mask. The samples were annealed at 300 °C during 2 h in nitrogen atmosphere to diffuse-out H that was introduced during chemical treatment. The backside contacts were formed by aluminum film or indium-gallium eutectic.
The H + -implantations were performed at room temperature through the Schottky contacts with six different energies in the range 300-600 keV and a total dose of 4 × 10 10 cm −2 to obtain uniform 'box-like' distributions of defects and H. One of the samples underwent 30 min isochronal annealings in the range of 75 °C-275 °C. Other samples underwent isothermal annealings at 75 °C-86 °C during 496-860 min, which resulted in formation of E5 * (as described in [8]). The samples were then heat treated at 170 °C for 30 min to anneal minor unstable defects prior to isothermal annealings at 190, 200, 210 or 220 °C.
The capacitance-voltage and DLTS measurements were performed using a refined version of the setup described in [9] with Boonton 7200 capacitance meter and a closed cycle helium cryostat. The DLTS signal was deduced by using GS4 weighting function to obtain higher energy resolution in the DLTS spectrum [10]. Figure 1 shows the DLTS spectra for the sample before and after 30 min isochronal annealings at different temperatures. The DLTS spectrum of the as-implanted sample shows signals of several defects: E1, E2, VO, V 2 (−/0), double acceptor state of trivacancy (V 3 (=/−)) [11], VOH, E4, E5 * and E5. The E1 and E2 peaks in figure 1 are observed together only in H-containing samples, and their energy positions are close to those attributed to carbon-oxygen-hydrogen complexes in [12,13] and labeled there E1 and E2 as well. The E4 peak consists of overlapping contributions from V 2 (−/0) and, presumably, V 2 H(−/0) [8,14,15]. In addition, E4 may have a contribution from the single acceptor state of trivacancy, V 3 (−/0) [11].

Experimental results
Before annealing the dominant part of H in the studied samples is stored in P-H [8]. P-H dissociates at ⩾75 °C with a release of H i that interacts with the defects. This is manifested in figure 1 in the growth of VOH and E5 * , and decrease of VO, V 2 (=/−) and E4 after annealing at 150 °C. After further heat treatment at 205 °C, VO decreases slightly, accompanied by a slight increase in VOH. The V 2 (=/−) and E4 peaks also decrease. E5 * , however, anneals out completely, and growth of E5 takes place. After annealing at 275 °C, VO and VOH remain to be the dominant peaks, with a slight decrease in VO and a slight increase in VOH. The V 2 (=/−) peak has disappeared, indicating that V 2 is annealed out. This suggest that the remaining amplitude of E4 is mainly due to V 2 H(−/0). E5 is completely annealed out.
Four samples have been annealed isothermally at temperatures of 190, 200, 210 and 220 °C to study the evolution kinetics of E5 * and E5. The DLTS spectra before and after different annealing steps at 190 and 220 °C are presented in figure 2. The as-implanted spectra are practically identical to that in figure 1. Prior to the isothermal annealings, the samples were annealed at 75 °C-86 °C to form the E5 * peak (see [8]) and then heat treated at 170 °C for 30 min to annealed out minor peaks. Subsequent isothermal annealings at 190 °C,    Moreover, the amplitudes of E5 * and E5 show a strong anti-correlation upon annealing. That is shown in figure 4(a) where the growth of E5 is plotted versus the loss of E5 * for all the samples isothermally annealed at 190 °C-220 °C during the investigation. The data points demonstrate a linear correlation and lie close to the line y = x/2, i.e. ∆E5 = − ∆E5 * /2. The Arrhenius plot for the E5 * annealing rate and the E5 formation rate is shown in figure 4(b). The rates follow Arrhenius behavior, and the deduced activation energy and the preexponential factor are 1.61 ± 0.07 eV and ~10 13 -10 14 s −1 , respectively.
CCS for both E5 * and E5 have been directly measured by varying the DLTS filling pulse duration at different measurement temperatures, and the experimental results are shown in figure 5. Similarly to E5 * , the transition entropy ∆S/k for E5 is close to 2. However, we observe a temperature dependence of CCS for E5 * , while CCS for E5 remains constant in the range 254-288 K. Assuming an activation mechanism for CCS, we deduce the energy barrier for electron capture to be 50 ± 10 meV. The relatively high entropies together with the temperature dependence of CCS for E5 * may indicate structural changes between occupied and empty levels, and could be attributed to a complex structure for both E5 * and E5.

E5 * as a divacancy-hydrogen complex
In the previous study [8], we have tentatively attributed E5 * to an acceptor state of a defect formed by reaction of V 2 H with H i or H * 2 , i.e. V 2 H 2 (−/0) or V 2 H 3 (−/0). The pre-exponential factor for E5 * annealing (~10 13 -10 14 s −1 ) found in the present study indicates a dissociation mechanism. The rates for the E5 * and E5 evolutions are similar at each studied temperature and follow Arrhenius behavior. The correlation between E5 * and E5 is valid for all the samples. All these facts lead to an assumption that E5 can be a product of the E5 * dissociation. Previously, E5 has been tentatively attributed to the acceptor state of VH [15]. Indeed, the electrical activity of VH also originates from the dangling bonds, and its electrical level VH(−/0) is predicted to have an energy position close to those of V 2 (−/0) and the vacancy-phosphorus (VP) state at ∼ E c − 0.42 eV [16]. E5 is known to form in H + -implanted samples [8,15], but not in irradiated material with subsequent hydrogenation [17,18]. This supports the identification of E5 as VH that requires simultaneous presence of both monovacancies and H atoms to form. One could expect that dissociation of hydrogen-vacancy complexes V 2 H 2 or V 2 H 3 (E5 * ) leads to emission of VH (E5). This mechanism, however, has difficulties in explaining the quantitative correlation between E5 * and E5, where ∆E5 = −∆E5 * /2. Indeed, annealing of, for instance, one V 2 H 2 (E5 * ) would result in formation of two VH (E5), while we observe the opposite: annealing of two E5 * centers is required to form one E5 center. Thus, one has to assume a more complex dissociation mechanism, where several reaction channels are possible, and the probability of VH formation is 50%.

E5 * as a trivacancy-hydrogen complex
On the other hand, the significant transition entropy ∆S/k ≈ 2, together with temperature dependent CCS for E5 * , can indicate another, more complex structure of E5 * . It is known that V 3 has a considerable concentration in irradiated and ion implanted silicon ( [11] and figure 1). V 3 can be presented in two configurations in silicon lattice: (i) so-called 'part of a hexagonal ring' (V . The transformation follows an activation mechanism with an activation energy of around 1.2 eV and a pre-factor of ~10 13 -10 14 s −1 [19]. It is tempting to suggest that, similarly to V 3 , trivacancyhydrogen complex (V 3 H) can also exist in PHR (V 3 H (PHR) ) and FFC (V 3 H (FFC) ) configurations. V 3 H (PHR) should exhibit a structure similar to V (PHR) 3 , with a H atom passivating one of the two dangling bonds [20]. The remaining dangling bond should give rise to one deep acceptor state, V 3 H (PHR) (−/0), close to that of V  from a Si dangling bond, while having different structure. Taking this into account one can suggest that the annealing of E5 * is not a dissociation, but a structural transformation from E5 * to E5, i.e. from one configuration of V 3 H to another. One can notice the similar pre-factors for the E5 * → E5 and V kinetics: ~10 13 -10 14 s −1 . We observe, however, that the amplitude of E5 * is almost double of that of V 3 (=/−) in as-implanted sample (figure 1), and the correlation of E5 * versus E5 is 2-to-1 ( figure 4(a)). One could speculate that, for instance, V 3 H (FFC) might be a negative-U defect with an acceptor and a donor levels [21], which emits two electrons upon the (−/+) charge transition, resulting in a doubled amplitude in the DLTS spectrum. Both V 3 H (PHR) and V 3 H (FFC) can have a (0/+) donor and a (0/−) acceptor transition similarly to V 2 , V 2 H and V 3 .
The negative-U behavior implies a considerable structural change of the center upon electron capture and emission. It is interesting to note that the temperature dependence of CCS for E5 * (figure 5) is consistent with the possible negative-U nature. Thus, an identification of E5 * as V 3 H (FFC) and E5 as V 3 H (PHR) can be tentatively put forward. This identification does not contradict our previous observation on the correlation of E5 * formation with the annealing of E4 [8], since E4 has a contribution from V 3 as well.
We can not conclude at the moment on the exact configurations for E5 * and E5, and a reverse identification of E5 * as V 3 H (PHR) and E5 as V 3 H (FFC) can not be ruled out. Theoretical studies on the atomic configurations and electronic properties of V 3 H are, thus, necessary to substantiate or rule out these assignments.

Conclusion
The annealing kinetics of the hydrogen-related states E5 * and E5 has been studied by DLTS. E5 * anneals out at around 200 °C with a correlated formation of E5. The kinetics exhibit a firstorder behavior with an activation energy of 1.61 ± 0.07 eV and a pre-factor of ~10 13 -10 14 s −1 . The pre-factor indicates a dissociation or structural transformation mechanism for the E5 * annealing and the E5 formation. The analysis of the electron capture cross-sections for E5 * and E5 reveals considerable entropy factors for both states and a temperature dependent capture cross-section for E5 * . Two possible identifications of E5 * and E5 are put forward. Firstly, E5 * can be attributed to V 2 H 2 (−/0) or V 2 H 3 (−/0), which dissociate with the emission of VH (E5). Secondly, E5 * and E5 can be assigned to two different configurations of V 3 H.