Modern silicon base component fabrication processes widely use the gettering of impurities and structural defects which deleteriously affect the electrical parameters of the devices [1]. These impurities are primarily Fe, Cu, Ni, Zn, Cr, Au, Hg, Ag etc. [2, 3]. The main sources of silicon wafer contamination with these impurities are bulk impurities forming during single crystal silicon growth and wafer fabrication as well as micro- and nanoparticles, chemical elements and their compounds remaining on the surface of the wafers after thorough cleaning. Furthermore, rapidly diffusing impurities are efficiently adsorbed by the wafer surface from the working media (vacuum, liquids and vapor or gas atmospheres). During further high-temperature process operations the impurities from the surface layer penetrate intensely to the surface region reaching significant depths. (The diffusion coefficients of metallic impurities are 3–4 orders of magnitude greater than the diffusion coefficients of doping impurities.) Metallic impurities penetrating into the working part of silicon wafers combine with structural defects to form generation/recombination centers which are a major cause of the increase in the leakage current [2, 4, 5]. This causes the degradation of a wide range of device parameters. In other words, the presence of metallic impurities in the crystal and on its surface is critical for a number of electronic components. Some applications require silicon wafers contain a bulk concentration of e.g. Fe of max. 1×10¹⁰ cm^-3, the surface impurity concentration also being controlled (less than 1×10¹¹ cm^-2) [1, 6]. These requirements are most critical for ICs with design rules of about decades of nanometers or less, low working currents and voltages and extremely low self-noise. Despite the large background of studies aimed at increasing the cleaning efficiency of as-grown wafers and process environments and providing cleanliness of process operations, the probability of silicon wafer contamination with detrimental rapidly diffusing impurities remains high. It should be noted that the development of absolutely clean conditions in the design of electronic structure processes, especially high-temperature ones, is a complex task for a number of technical and economic reasons. Therefore the probability of the formation of intrinsic point defects and their clusters including those with impurity atoms (e.g. generation/recombination centers) remains high. Thus gettering is primarily used nowadays for preventing the penetration of contaminating impurities into silicon wafer bulk and the formation of generation/recombination centers in the working regions of electronic structures fabricated. Below we will consider process techniques used for the formation of different origin gettering centers, their advantages and drawbacks and redistribution mechanisms of rapidly diffusing impurities and structural defects in silicon wafer bulk.

Gas phase gettering implies heat treatment of wafers and bulk specimens in vacuum and inert and chlorine-containing atmospheres. This method is based on point defect and impurity atom extraction from wafer bulk to gas phase or vacuum.

An illustrative example of gas phase impurity gettering is oxygen impurity drain from the surface regions of silicon wafers (see above). This process is accompanied by sublimation of surface silicon atoms. Possibly the silicon/impurity system is depleted of the more volatile component more intensely during evaporation. Silicon and impurity sublimation leads to vacancy formation on the surface and in the surface region thus destroying the equilibrium of the vacancies and the interstitial atoms between the surface and the bulk. This results in diffusion of interstitial atoms and vacancies toward the surface and further to the gas atmosphere. The density of the structural microdefects consisting of interstitial atoms, including stacking faults, decreases due to the supply of vacancies. This process is typical of wafer heat treatment at 800–1200 °C in an argon gas atmosphere. As a result the surface wafer layer becomes almost free of primary microdefects and thus the formation of oxidation induced stacking faults is suppressed during the further growth of the oxide film.

Gettering in a chlorine-containing atmosphere with hydrogen chloride addition is most often used for heat treatment of microelectronics components in order to control microdefect composition and impurity content in silicon wafer bulk. We will consider below the evolution of point defect clusters during active chemical hydrogen chloride etching of silicon wafers. According to earlier data [22, 23], the basic etching process includes several stages:

The chlorine/silicon interaction reactions are exothermal, favoring a significant weakening of the bonds between the reacting silicon atoms in the crystalline structure. The metal and carbon impurities on the surface and in the surface region are efficient catalysts of silicon chloride formation reactions. Metals also form volatile chlorides. Oxygen and water steam supplied to the specimen surface have a negative effect on chloride formation.

Primary silicon atom removal from the surface wafer layer during HCl etching results in the formation of vacancies (V) which have a low but still meaningful probability of diffusing into wafer bulk [24]. The generation of metastable vacancies from the surface leads to the formation of a supersaturated solid solution of V in the bulk, accompanied by the formation of a diluted solid solution of other defect type. As reported earlier [25] this nonequilibrium situation takes place if ∆С_v > 0, ∆C_i ≤0. Then V interaction with microdefects consisting of interstitial atoms reduces the microdefect size until complete dissolution. More complete data on the interaction between point defects and microdefects in silicon were reported elsewhere [26–28].

Experimental confirmation of A cluster type microdefect annihilation and oxidation induced stacking fault density reduction in silicon specimens after heat treatment in chlorine-containing atmosphere was reported earlier [29–30]. Figure 7 shows macro- and microdistribution of A clusters in the as-grown specimens and specimens heat treated in a chlorine-containing atmosphere. The A clusters had a swirl-type distribution in the as-grown specimen. The heat treatment parameters were 1200 °C, 8 min, in an H₂ + HCl atmosphere. After this heat treatment selective etching revealed no A clusters in the silicon specimens, and this was additionally confirmed by the absence of oxidation induced stacking faults after subsequent oxidation.

Heat treatment in a chlorine-containing atmosphere has also found practical application for oxide film growth on silicon wafers. Note that SiO₂ films grown by thermal oxidation, e.g. for MOS (metal/oxide/semiconductor) structures, does not always satisfy the strict requirements to surface charge, surface recombination rate and bulk trap concentration (more than 10¹⁶ cm^-3). It was reported that the degradation of these parameters is mainly caused by alkaline metal contamination of the material during high-temperature oxidation [31, 32]. A promising solution was to eliminate the effect of alkaline metals by adding HCl to the oxidizing atmosphere. Much chlorine is adsorbed by the growing oxide which is then strongly bound with mobile alkaline metal ions. It is probably due to their large size that chlorine ions are almost immobilized in the SiO₂ matrix and hence strongly bind mobile alkaline metal ions. Practical result are high-quality 5–100 nm thick SiO₂ films grown by two-stage dry oxidation with ~1–3% HCl addition.

Impurity and defect gettering in a chlorine-containing atmosphere finds general use in the fabrication of high-power high-voltage devices with the source material being zone melting grown dislocation-free single crystal silicon. The oxygen impurity concentration in this silicon is several orders of magnitude lower than in Czochralski grown silicon which is the main material of microelectronics. Moreover the working region of high-power devices covers the entire wafer bulk. Therefore getter formation on the basis of structural defects in zone melting grown silicon wafers for high-power devices is impossible. The problem of gettering microdefects and rapidly diffusing metallic impurities in zone melting grown silicon wafers proved to be solvable by using heat treatment in a CCl₄ base chlorine-containing atmosphere. By way of example Fig. 8 illustrates the effect of heat treatment atmosphere on microdefect evolution in zone melting grown bulk silicon crystals [33]. In the as-grown state the microdefects are distributed in the crystal bulk in the form of swirls (Fig. 8a), and after heat treatment in a chlorine-containing atmosphere at 1250 °C for 40 h no microdefects were not found in these crystals (Fig. 8b). Similar heat treatment of the reference specimens in air caused an increase in the microdefect size and density and a growth of swirl strips. This indicates drain of metastable intrinsic defects and impurity atoms to the as-grown microdefects. The chlorine-containing atmosphere used in these experiments was an argon and oxygen mixture with ~1 mol.% CCl₄ vapor addition. More detailed studies of chlorine-containing atmosphere heat treatment efficiency for microdefect-containing silicon crystals, including those with swirl-type distribution, were reported earlier [34, 35].

Another efficient application of heat treatment in a chlorine-containing atmosphere is radiation induced defect annealing in the neutron doped silicon technology. The source material for neutron doping is zone melt grown single crystal dislocation-free silicon. The specially prepared silicon ingots are neutron irradiated in a thermal neutron reactor. Detailed description of the neutron doping process of semiconductor materials was reported earlier [36]. Thermal neutron reactions

Si³⁰(n,γ)Si³¹(β^--decay) → P³¹

produce phosphorus atoms that are a donor impurity in silicon. Along with P³¹ atoms, a large number of accompanying radiation induced defects are generated in the doped silicon crystals as a result of bombardment with nuclear debris, fast (high-energy) neutrons and γ-particles and due to recoil atoms. As-irradiated ingots are heat treated at high temperatures for eliminating the negative effect of the radiation induced defects on the electrophysical properties. The completeness of radiation induced defect annealing is judged about by the recovery of the electrophysical parameters of the material, primarily the electrical resistivity ρ and minority carrier lifetime τ. It was shown that heat treatment at 800–900 °C for ~2 h recovers ρ to the target level, with τ always being several times lower than the initial level, sometimes even below the acceptable level. Despite the large number of studies [36] the τ recovery problem could not be solved for a long time until the effect of annealing atmosphere on microdefect transformation and change in bulk content of rapidly diffusing impurity was revealed. One of the earliest works on this theme showed that τ differs significantly depending on annealing atmosphere (Fig. 9) [37].

The highest τ are observed in specimens heat treated in a chlorine-containing atmosphere; air, vacuum and argon heat treated specimens follow in order of decreasing τ. It should be noted that τ degradation depends on microdefect density and presence of rapidly diffusing metallic impurity atoms in crystals. By way of example Fig. 10 shows τ as a function of microdefect density in neutron doped silicon heat treated in air [38] and Fig. 11 shows bulk τ distribution depending on gold atom profile in similar specimens heat treated in air and in a chlorine-containing atmosphere [39]. Gold was deposited on specimen butts after neutron irradiation before annealing. Noteworthy is the appearance of microdefects in some silicon specimens after irradiation and heat treatment in air in which τ were low [40, 41]. No microdefects were found in these as-grown ingots before irradiation and τ were sufficiently high. Note also that τ in specimens heat treated in a chlorine-containing atmosphere increase monotonically in depth (Fig. 11b) and drastically reach the as-grown ~1000 μs level beginning from a 3 mm depth unlike for air heat treated specimens in which τ reach 200–400 μs at best.

Thus gas phase gettering solves the following tasks:

Below we will consider possible methods of microdefect composition controlling and τ recovery.

It was shown above that generation/recombination centers are complex agglomerates of the microdefect + metal impurity atom type. The microdefects may have variable dimensions and mainly consist of interstitial atoms. It can therefore be concluded that introducing a nonequilibrium vacancy concentration into crystal bulk might help destroying the generation/recombination centers due to vacancy interactions with interstitial atoms in the microdefects. Note that heat treatment induces complex processes in silicon crystals including generation and recombination of intrinsic point defects and their interaction with different types of structural defects. They generally include the following:

The most complete analysis of the interactions occurring in this system was reported elsewhere [25]. Taking into account those data we will consider specific examples of microdefect transformation depending on heat treatment conditions.

Oxidation induced stacking fault size evolution during heat treatment in argon and in hydrogen as well as during thermal oxidation is shown as a function of exposure time in Fig. 12 [25]. It can be seen from Fig. 12 that heat treatment in argon and in hydrogen (Curves 1 and 2) gradually reduces the size of the oxidation induced stacking faults, the reduction being temperature-dependent. These data suggest that heat treatment in argon and in hydrogen at best retains the defect structure in silicon specimens close to the as-grown one although microdefect sizes change. Thus the system is slightly supersaturated with vacancies.

In turn the oxidation induced stacking fault sizes grow quite rapidly during heat treatment (Curves 3 and 4), the growth being slightly temperature-dependent. Presumably interstitial atoms dominate in the system. This nonequilibrium situation is described by the ratio ∆С_v ≤ 0, ∆C_i > 0. In other words the oxidation induced interstitial silicon atoms condensate intensely on microdefects, their size and density increasing as a result. An efficient source of nonequilibrium interstitial atoms is in this case the thermally oxidized silicon surface. A similar situation occurs in neutron doped silicon during air annealing: some crystals exhibit microdefect size and density growth. These crystals usually have a short carrier lifetime. Note that the origin of microdefect growth is associated with the precursor defects forming during single crystal growth and having small sizes in the as-grown state thus having a negligible effect on lattice deformation. Taking into account this observation one should preliminarily dissolve the precursor defects in order to avoid the formation of oxidation induced stacking faults in silicon specimens which will undergo oxidation. In other words additional heat treatment is intended to produce a supersaturated vacancy solid solution in the crystals: ∆С_v > 0, ∆C_i ≤ 0. As shown above, this method is implemented using heat treatment in a chlorine-containing atmosphere, e.g. HCl.

The interaction mechanism between the structural defects in neutron doped silicon is more complex since it develops in the presence of radiation induced defects the structure of which is incompletely clear. Assuming that neutron irradiation does not produce disordered regions with amorphization traces one can safely hypothesize that the radiation defects mainly contain point defects and their complexes including those with impurity atoms inherited from the as-grown ingots or formed as a result of nuclear reactions. This makes it possible that the condition ∆С_v > 0, ∆C_i > 0 is true. At an early stage of heat treatment in a chlorine-containing atmosphere part of the radiation induced interstitial atoms recombinate with vacancies while other part migrate to specimen surface due to a higher diffusion coefficient or settle on microdefect surfaces. With a decrease in the nonequilibrium concentration of interstitial atoms at subsequent annealing stage due to chlorine chemical reactions with silicon, the vacancy formation on crystal surface becomes predominant, with the vacancies intensely penetrating into the bulk and interacting with growth microdefects. In this case the relationship ∆С_I/∆С_V < D_V/D_I is valid and hence microdefects dissolve. Thus heat treatment of neutron irradiated silicon specimens in a chlorine-containing atmosphere destroys the generation/recombination centers, and the released metallic impurity atoms diffuse over interstitial sites and drain to the surface where they react intensely with chlorine atoms. This mechanism of defect annealing in a chlorine-containing atmosphere is confirmed by earlier experimental data [39] obtained in a radioactive marker study of gold and iron atom release efficiency from the bulk of silicon specimens heat treated in different atmospheres for radiation induced defect annealing. The test specimens with pre-deposited Au¹⁹⁵ and Fe⁵⁹ isotopes onto the butts were heat treated first in air and then in a chlorine-containing atmosphere and a separate series of specimens were only heat treated in the chlorine-containing atmosphere. Air annealing resulted in iron and gold atoms penetration to a 40 μm depth into the bulk. Subsequent heat treatment in a chlorine-containing atmosphere reduced the gold concentration by approx. one order of magnitude and the iron concentration almost by two orders of magnitude at a 15 μm depth. The specimens heat treated in the chlorine-containing atmosphere without preliminary air anneal had only traces of radioactive elements. Thus the possibility of back-diffusion of metallic impurities from the bulk of single crystal silicon specimens is confirmed if certain conditions are met.

We reviewed results showing that gettering of rapidly diffusing metallic impurities and structural defects is widely used in the fabrication of electronic components on the basis if single crystal silicon. Getters formed on structural defects are dislocation networks in case of extrinsic getters while intrinsic getters consist of oxygen/precipitate dislocation complexes. These getters also contain dislocation generation centers with a high density which during device operation may emit dislocations to the active device regions as a result of contingency external mechanical load application. Furthermore recrystallization may cause recontamination of the device working region with formerly gettered impurities. In either case device degradation is probable.

In turn, if silicon is supersaturated with vacancies, gas phase getters favor microdefect dissolution, generation/recombination center destruction and rapidly diffusing impurity drain to surface followed by the formation of volatile chlorine compounds. The fundamental difference in the nature of the two getter types shows good promise for improving the reliability of devices fabricated on silicon wafers using heat treatment in chlorine-containing atmospheres.