Accumulation of mobile impurities at dislocations can be due to fundamentally different reasons:
1. segregation in the dislocation strain field;
2. chemical binding to the dislocation core;
3. precipitation at dislocations;
• due to a reduced nucleation barrier; or
• due to faster kinetic of impurity atom incorporation into growing precipitates;
4. accumulation close to dislocations as a result of fast equilibration of intrinsic point defects there.
The latter effect is restricted to those impurities that are mainly dissolved on substitutional sites but diffuse via the interstitial species like, e. g., gold and platinum in silicon or to those precipitates that must emit or absorb vacancies or self-interstitials for strain-free growth. Note that in these cases the accumulation would be a transient phenomenon.
Segregation in the dislocation strain field is due to an elastic interaction with impurities that mainly comes from the difference in size of a Si atom in the host lattice (r0 = 0.117nm) and the (covalent) radius rimp of the impurity atom . In many cases the dominant term of elastic interaction is proportional to p x (r0 – rimp)3, where p is the local hydrostatic pressure around dislocations. Since the edge component of dislocations provides both dilatation and compression field, negative and positive misfit will result in an attractive interaction. The elastic interaction energy depends on the impurity type and on type of dislocation. For dissociated 60° dislocations it usually does not exceed 0.4-0.5 eV at a distance of about 1 nm from the 90° partial dislocation. This means that elastic interaction can collect a significant number of impurity atoms only at quite low temperature, typically less than 400°C. At such temperature solid solutions of transition-metal impurities will be strongly supersaturated implying that precipitation could be an important competing process. For illustration, Figure 4.13 shows calculated
Figure 4.13 (a) Estimated concentration of impurity atoms for elastic interaction depending
on the distance from the dislocation core in the direction of maximal concentration in thermal equilibrium atT = 300C and 500°C after Bullough and Newman . The impurity concentration far away from the dislocation is 1015 cm-3, the elastic interaction energy is assumed to be 0.45 eV at a distance of 1 nm from the core. Please note, that the total number of impurity atoms collected by the dislocation is ofabout 106 cm-1 atT = 300°C but less than 5 x 104 cm-1 at T = 500°C. (b) Gold concentration measured in highly dislocated silicon masured by NAA  (square symbols and error bars) compared with calculated total concentrations assuming a binding energy AHb of 1.92 eV (curve 1), 1.72 eV (curve 2), and 1.52 eV (curve 3) according to . The dashed line represents the gold solubility in dislocation-free silicon.
concentration profiles of impurity atoms around dislocation caused by elastic interaction of about 0.45 eV at a distance of 1 nm. Even at 500°C the local impurity concentration close to dislocation is higher by almost 3 orders of magnitude (x850 at 500°C and x9000 at 300°C) than the average concentration in the bulk. However, the total number of atoms collected by dislocation is not very high. At 500°C it is less than 105 cm-1 and only at 300°C it reaches about 106 cm-1. The number of deep defects at dislocations typically observed by DLTS or EBIC for not very clean dislocations is more than 106 cm-1. Obviously, the elastic interaction can collect such a significant number of impurity atoms to dislocations only in the case of impurities that are sufficiently mobile at low temperature, but at the same time have a reasonably large interaction energy implying a large volume misfit. These conditions in many cases are inconsistent, implying that the elastic interaction cannot be considered as the main factor responsible for collection of impurities to dislocations. However, since the elastic interaction is a long – range (1/r) interaction that significantly increases the concentration of impurity atoms close to dislocation, it can strongly accelerate the kinetics of processes like chemical binding to the dislocation core and precipitation at dislocations. Although not being the main reason, elastic interaction can be involved in other processes of impurity collection to dislocations by enhancing the respective kinetics.
For some impurities, chemical bonding to the dislocation core can involve large binding energies of 2eV or more, which allow dislocations to collect a large number of atoms even at quite high temperature. Most important questions concern the electronic
properties (energy levels and capture cross section) of impurity species in the dislocation core and the actual values of their binding energies there. The former issue is related to the recombination activity of such decorated dislocations , while the latter is important when binding and dissociation kinetics are to be modeled in view of cleaning contaminated dislocations by external gettering techniques.
In order to get answers to these questions, sophisticated experiments and abinitio computer calculations are needed. At present, our knowledge about binding energies and electronic properties of transition-metal impurity atoms in dislocation cores is very limited, which consequently limits our ability for predictive computer simulations of get – tering in silicon with a high density of dislocations. Therefore, an intensive research in this field is fairly desirable. Below, we give a few examples of results obtained recently in this field. In a theoretical study, Fujita et al.  showed that the binding energy of substitutional copper atoms to 30°-partial and 90°-partial dislocations is about 1.5 eV and 2 eV, respectively. A comparison of the electrical levels of Cus in dislocation cores with electrical levels of Cus in the bulk shows that the donor level for Cus at dislocation is shifted upwards by more than 0.3 eV, while the single acceptor level is shifted downwards by approximately 0.15 eV. This means that binding of substitutional Cu atoms to the dislocation core fundamentally changes its electrical properties and even becomes a negative-U defect with its acceptor level below its donor level. In  copper at dislocations in p-type Si/SiGe/Si structures has been investigated experimentally and a new dislocation-related DLTS-level at EV + 0.32 eV was detected after intentional contamination with copper. This level is remarkably close to the theoretical acceptor level position at EC-0.77eV for Cus at 90°-partial dislocations with single-period reconstruction.
For gold in silicon, which is mainly dissolved on substitutional sites, such a large binding energy would lead to an almost complete accumulation of gold in the dislocation cores even at elevated temperature. Recent experiments for gold in silicon using neutron activation analysis (NAA) indeed show a significant increase of the total concentration of gold in silicon containing a high density of dislocations , as shown in Figure 4.13(b). The binding energy of Aus to dislocations, i. e., the enthalpy difference of Au atoms dissolved on substitutional lattice sites and Au atoms bound to dislocations, estimated from these experimental data by the authors is very large, i. e., about 2.7 eV. Re-analysis of the data presented by Vo6  results in a slightly smaller binding energy of Aus to dislocation core of about 1.7 eV. This means that dislocations in Si doped with gold must usually contain very high concentration of Aus atoms in the core. However, up to now no reliable DLTS data about electronic properties of gold at dislocations are available in the literature.
The precipitation of fast-diffusing transition-metal impurities is strongly affected by the presence of dislocations that in general serve as heterogeneous nucleation sites , which might be due to the fact that atomic configurations in the dislocation core and the long-range strain field can reduce the barrier for the formation of small precipitate nuclei. In addition, dislocations may also assist precipitate growth. One familiar way is the accommodation of volume misfit between precipitate phase and silicon matrix. Furthermore, dislocations can provide sites for incorporation of solute atoms into precipitates . In all cases reported in the literature, the precipitation at dislocations is in the form of those silicide phases described in Section 4.4.1 although dislocation-specific configurations have been reported .