Chapter 1

0 downloads 0 Views 2MB Size Report
selective epitaxial growth, SiH4 decomposes dominantly to silylene. (SiH2) and two hydrogen atoms which these Hydrogen atoms occupy the dangling bonds ...
PREFACE B-doping of group IV materials using B2H6 is widely performed in chemical vapor deposition (CVD) technique. The B-doped SiGe layers are grown epitaxially as the base layer in HBTs for increased frequency performance in mixed signal radio frequency (RF) applications. These layers may also apply as a stressor material in source/drain of pMOSFETs to enhance carrier mobility in a uniaxially strained channel. Furthermore, contact layers performance in terms of thermal stability and resistivity are improved by implementing highly boron doped (B-doped) layers in various electronic components. However, high diffusion of boron can limit the thermal budget for fabrication of the devices. One way to suppress this problem is integration of carbon in B-doped layers where the carbon diffuses out but boron stays in the SiGe layers. Therefore, growing highly B-doped group IV materials with high thermal stability and layer quality is a challenging issue. This chapter deals with growth kinetics, dopant incorporation, thermal stability, strain compensation, strain relaxation and defect formation of B-doped SiGe layers grown by reduced pressure CVD. The ion implantation and some of its processing issues regarding B-doping will be discussed.

Chapter 1

INTRODUCTION SiGe(C) materials have acquired great importance for many applications; e.g. high frequency heterojunction bipolar transistors (HBTs) in telecommunication [1-6], as the stressor material applied in source/drain regions to enhance the carrier mobility in the channel [7-11] and as virtual substrates for strained-Si CMOS logic applications [11]. Nowadays, every GPS unit or cell phone has a SiGe(C) front end in it. In most cases, SiGe layers are boron-doped and have grown either selectively or non-selectively on the active areas of the chip. Several high-temperature treatments are involved during the fabrication of HBTs or MOSFETs. The out-diffusion of B from the SiGe is the most important issue since it results in degradation of high frequency behavior or short channel effect of the source and drain junctions. Design of the base layer is one of the key issues for the high frequency behavior of HBTs. The layer profile of transistor’s base is recognized by the Ge and B profiles. The incorporation of boron depends on the Ge content and the growth rate [12]. As illustrated in Figure 1, there are three types of Ge profiles for base layers; box, trapezoid and triangle (graded Ge profiles) [13]. To achieve a box B profile inside a SiGe layer is not an easy task and requires a good understanding of the process.

4

M. Kolahdouz, H. H. Radamson and R. Ghandi

Figure 1. Germanium profiles applied for HBT structures [13].

The three important DC and AC characteristics of HBTs are current gain β, transit frequency fT and maximum oscillation frequency fmax, which are expressed by the following equations:

where ΔEg,t,app is band off-set term and it relates to the strain in SiGe base layer and the heavy boron doping. In the above equation, D is the electron diffusivity constant, W is the layer width and N stands for doping concentration (where the subscripts E and B stand for emitter and base, respectively). In this case, β increases for high Ge content. The transit frequency fT is given by:

where τ is the transit time and the subscripts e, b and c stand for emitter, base and collector, respectively. The fT in HBTs increases by thinning the SiGe base layer. In case of a triangle base design (graded Ge profile), a drift field is created which is necessary for decreasing the base transit time. The maximum oscillation frequency fmax can be written as:

Introduction

5

The fmax is determined by fT, the base resistance (Rb) and the collectorbase capacitance (Ccb). An implantation of extrinsic base layer decreases the base resistance remarkably. Further improvement may be obtained when Ni silicide is formed to decrease the contact resistance. The capacitance term, Ccb in fmax equation is affected by different factors but the important one is a self-aligned collector base junction. This can be achieved through the selective epitaxy growth of the collector layer [14] (see Figure 2). A chemical mechanical polish step is required to smoothen the selectively grown collector layer prior to the growth of non-selective SiGeC base layer.

Figure 2. TEM cross-section of a SiGeC HBT with a fully silicided extrinsic base region and NiSi emitter and collector (not shown) contacts, drawn emitter width of 0.5μm, and an emitter-poly overlap of 0.6 μm [14].

Many reports demonstrated that the collector layers can be created by P-implantation through the base layer [15-17]. Thermal treatment is involved to activate the dopants after ion implantation of the extrinsic and intrinsic part of the transistors. As mentioned earlier, one of the critical issues during these steps is boron out-diffusion from the doped SiGe base layers. Through ion implantation, the crystal structure of the target material is damaged by the high energetic ions. The damage concerns a distribution of interstitials and vacancies [18-24]. These extremely

6

M. Kolahdouz, H. H. Radamson and R. Ghandi

mobile point defects may form stable extended defects such as clusters or dislocations. Thermal annealing is necessary to rearrange the original crystal structure. However, supersaturation of these point defects caused by implantation results in an enhanced diffusion of the dopants. In case of boron-doped Si-based materials, the boron atoms located at the tail of the implanted profile diffuse by several orders of magnitude faster than the normal thermal diffusion. It is thus referred to as “transient enhanced diffusion” (TED). This phenomenon is temperature-, depth- and concentration-dependent in the early stages of the annealing process. An example of TED due to interstitials injected from the surface is shown in Figure 3 [25]. It is demonstrated in this figure that the displacement due to thermal diffusion is about 5–10 nm but TED displacement is about 200 nm.

Figure 3. (a) SIMS measurements of a B doping spike in MBE grown Si layers before and after diffusion at 810°C or 15 min. (b) same as (a) but the layer implanted with 1×1015 /cm2, 40 keV Si before diffusion [25].

TED has created an obstacle for fabrication of Si-based devices. The enhanced B diffusivity can be elucidated through the diffusion of highly mobile pairs of B and interstitial Si atoms. A Si interstitial can also ‘‘kick out’’ the substitutional boron atom to an interstitial position where it can diffuse easily (thermal diffusion). This means that the Si interstitial concentration is the first requirement for TED of boron atoms. Regardless of ion implantation, Si interstitials are also injected by the oxidation process [26,27] which in term increases the diffusivity of the dopant atoms. TED even at low temperatures (e.g. 800˚C) may last for

Introduction

7

more than one hour which is due to the existence of interstitial boron and Si clusters ({311} extended defects and dislocation loops). These clusters act as the interstitial reservoirs. The clusters dissolve and emit interstitials when the supersaturation becomes week. This process supports the transient enhanced diffusion. The rate of the interstitial emission by the clusters is determined by Ostwald [28] ripening of both the dislocation loops [29] and {311} extended defects [30] which was confirmed by the measurements of the size distributions of the clusters at different temperatures and times of annealing. It was shown that at high temperatures, TED lasts just for a short time due to instability of the clusters and coupling between interstitials and boron atoms [31]. It was also observed that at these temperatures, the tail displacement is small. Therefore, “spike anneal” using rapid thermal annealing (RTA) with very high ramp rates can be a solution to avoid the TED of boron [32,33]. Another solution to TED was reported [34] to be pre-amorphization of the Si surface prior to boron implantation. This process reduces the channeling tail, thereby allowing for shallower, more abrupt junctions. The effect of boron out-diffusion from the base layer into the emitter and collector adjacent regions is degradation of the transistor characteristics. This out-diffusion causes that the junctions shift outwards and at the same time the base widens. Since the additional base regions are silicon (not SiGe), these parts have wider bandgap than SiGe ones. This affects the emitter-base junction and decreases the DC current gain β. The lack of Ge at the CB junction establishes parasitic barriers which decrease the electron flux into the collector. On the other hand, these barriers control (degrade) the collector current, Early voltage and also AC performance of the transistor. Two earliest studies about the effect of boron out-diffusion on transistor characteristics were presented by Prinz et al [35] and Slotboom et al. [36]. Figure 4 shows the parasitic barriers due to out-diffusion of boron in HBTs. In this figure, the width of B profile in the HBT has become larger than the Ge profile width due to the boron out-diffusion.

8

M. Kolahdouz, H. H. Radamson and R. Ghandi

Figure 4. The SIMS profile and conduction-band energy diagram of a Si0.80Ge0.20 HBT and the corresponding Si transistor with the equivalent doping profile [36].

The calculated conduction band edges for a typical boron box-profile show that at a small CB bias, a large parasitic barrier establishes which makes the flux of minority electrons into the collector region difficult. Meanwhile, these parasitic barriers are reduced when large CB bias is applied and the depletion layer widths widen [36]. One way to retard the boron out-diffusion is integration of C in SiGe layer. Many reports [37,38] demonstrated carbon concentration about 0.04 - 0.5% over the whole base layer is necessary to suppress the boron out-diffusion (see Figure 5). The base current for SiGeC layers was not ideal; it was considerably larger compared to SiGe transistors. The behavior of base current in SiGeC transistors was recognized to deep-level defects related to the presence of C [39-44]. During the thermal treatment, the carbon atoms diffuse to the adjacent layers and may precipitates in emitter extrinsic

Introduction

9

contact. The carbon precipitates are not electrically charged and will not increase the noise level of the transistor.

Figure 5. SIMS plots of emitter and base profile with and without C-doping. The sample has been exposed to RTA treatment at 950°C for 10 sec [38].

Chapter 2

BORN INCORPORATION IN SI-BASED MATERIALS SiGe layers have been recently integrated in source/drain (S/D) regions of pMOSFETs to induce uniaxial strain into the Si channel and increase the hole mobility in the transistors. In this design, incorporation of high B concentration is a crucial point for reducing S/D parasitic resistances. Selective epitaxy growth (SEG) is applied to fill out the recess S/D regions. The SiGe layer profile (composition and the growth rate) varies over a chip due to the pattern dependency behavior of the growth. In general, the higher incorporation of boron in SiGe compared to Si layer is believed to originate from the presence of strain and higher growth rate of SiGe epitaxy compared to that of Si. The former should be dominant for highly B-doped SiGe layers with low Ge contents since the strain in the SiGe may be totally-compensated; whereas, the growth rate effect must be more obvious for high Ge contents. However, any change in Ge or B content will affect both these parameters simultaneously. Thus, separation of the growth rate and strain effects to estimate the B incorporation in SiGe layers is not a suitable solution. Jang et al. [45] summarized the observed total dopant incorporation results as a function of gas-phase diborane, for different germanium compositions (Figure 6 and 7) and showed that the values of boron incorporation in Si1-xGex with germanium content of 0, 0.13, and 0.20 fall on the same straight line (Figure 6). They concluded that the

12

M. Kolahdouz, H. H. Radamson and R. Ghandi

incorporation of boron atoms into the Si1-xGex layer is independent of germanium content and it has no effect on Si1-xGex growth rate.

Figure 6. Boron and arsenic incorporation as a function of gas-phase dopant fraction for different germanium contents [45].

Figure 7. Growth rate as a function of gas-phase diborane and arsine fraction for different germanium contents [45].

Born Incorporation in Si-Based Materials

13

Figure 8. Dependency of the (a) deposition rate, (b) Ge fraction and (c) boron concentration on the B2H6 partial pressures. The total pressure and the SiH4 partial pressure are 30 Pa and 6.0 Pa, respectively. The deposition temperature is 550 °C [46].

14

M. Kolahdouz, H. H. Radamson and R. Ghandi

Murota et al. [46] used the modified Langmuir-type surface adsorption and reaction scheme for explanation of the growth behavior (growth rate, Ge content and dopant concentration) for SiH4 – GeH4 – B2H6 –H2 gas mixture. It was reported that the deposition rate decreases with increasing B2H6 partial pressure at the high PGeH4 (Figure 8a), and the Ge fraction is independent of the PB2H6 (Figure 8b). However, the B2H6 partial pressure is in linear relation with boron concentration up to the 1022 cm-3 (Figure 8c). Kuhne et. al. [47] showed that high boron concentration (>1019 cm-3) during SiGe epitaxial growth can degrade the Ge profile. For singlewafer atmospheric-pressure chemical vapor deposition (APCVD), they reported that no Ge concentration depression occurs at deposition temperatures above 675°C. However, the Ge depression occurs increasingly along the wafer stack in a multi-wafer APCVD even at temperatures above 675°C. They also reported the possibility of highlevel in-situ boron doping of SiGe layers in low-pressure processes at 550°C without any degradation of the Ge profile (Figure 9) and concluded that LPCVD is superior to APCVD with respect to high-level in-situ B-doping of SiGe.

Figure 9. The calculated fractional atomic Ge content, xGe, of SiGeB thin films as a function of temperature for three different PB2H6 values: PB2H6 ≤ 10-4 Pa represents the behavior of B-free and weakly B-doped SiGe films; PB2H6 = 10-2 Pa and PB2H6 = 1.7×10-1 Pa show the increasing depression of xGe due to B2H6induced Si deposition enhancement. B effect on Ge content arises at growth temperatures below 675 °C [47].

Born Incorporation in Si-Based Materials

15

Other groups [48,49] announced the increase of the growth rate of SiGe layers with increase of diborane flow independent of Ge content of the layers. Meanwhile, increase of diborane flow results in decrease of the apparent Ge content of the layers. Since the Ge content in epi-layers is extracted from the measured strain by HRXRD and these values are compensated in presence of B, apparent Ge content is mentioned here in the text. This strain compensation is just due to substitutional dopant atoms (B) and the total dopant concentration (including the interstitial concentration) is unknown [50]. The atomic dopant concentration obtained by Secondary Ion Mass Spectroscopy (SIMS) may differ from substitutional (active or incorporated) dopant concentration in the lattice. In case of SiGe, performing the electrical measurements to obtain active dopant concentration of p-type dopant is not quite straight forward since the induced strain by Ge in Si crystal causes variation in both the hole mass and the warping of the valence band. In this case, the film composition is calculated from the mismatch values obtained from ω-2θ rocking curves (ω and 2θ are incident and diffracted angles, respectively) by scanning a focused x-ray beam on a specific chip in high resolution x-ray diffraction (HRXRD) mode. In these rocking curves, the position of the layer peak relative to the substrate peak provides the lattice mismatch perpendicular to the surface. The Ge content was obtained from simulation of the rocking curves by using the Takagi-Taupin equations. This type of measurement is one dimensional analysis, but, all lattice mismatch parameters can also be measured by high-resolution reciprocal lattice mapping (HRRLM) around (113) reflection. The low angle of the incident beam in this reflection (about 2.8˚) makes it extra sensitive for revealing the defects in the SiGe layers. The mismatch parameters perpendicular and parallel to the surface are obtained from the following equations: θ θ

ω ω

θ θ

θ θ

ω ω

θ θ

where the indices s and l stand for the substrate and the layer, respectively. The lattice mismatch can be written as:

16

M. Kolahdouz, H. H. Radamson and R. Ghandi

where υ is the Poisson ratio (υ = 0.278) for Si1−xGex/Si and the incorporation of B in these heterostructures is believed to have only a minor effect on the Poisson value [51]. For strained epitaxial layers, f║ is negligible. When the SiGe crystal is doped with B (substitutional dopant concentration), strain compensation occurs. This is due to the fact that the size of the dopant atom is smaller than Si and Ge atoms. As a result, the layer peak in ω-2θ rocking curve shifts towards the substrate peak. Thus, the substitutional (electrically active) dopant concentration can be calculated as follows:

where, f is the compensated mismatch (between the intrinsic and doped SiGe) and β is the lattice contraction coefficient which can be obtained from:

where NSi is the density of Si atoms in a unit volume and r is the atomic radius. In these calculations, the Ge content is assumed to be constant [52]. This method (HRXRD) has been recognized as a fast, easy and indestructible technique which measures the induced strain in Si or SiGe layers. The maximum B concentration incorporated substitutionally without any apparent crystal quality degradation are reported as 2.3×1020, 2.4×1020 and 2.9×1020 cm-3 for Si0.7Ge0.3, Si0.6Ge0.4 and Si0.5Ge0.5, respectively [48]. The measured SIMS profile of the total amount of B atoms in the reported layers has yielded slightly superior values concluding that some of the B atoms are incorporated in interstitial sites (Figure 10-12).

Born Incorporation in Si-Based Materials

Figure 10. SiGe:B growth rate as a function of the F(B2H6)/[F(SiH2Cl2)+F(GeH4)] mass-flow ratio (MFR) [48].

Figure 11. ‘‘Apparent’’ Ge concentrations as a function of the F(B2H6)/[F(SiH2Cl2)+F(GeH4)] mass-flow ratio (MFR) [48].

17

18

M. Kolahdouz, H. H. Radamson and R. Ghandi

Figure 12. B concentrations obtained from SIMS measurements (full symbols linked by full curves) in boron-doped layers as a function of the F(B2H6)/[F(SiH2Cl2)+F(GeH4)] mass-flow ratio. For comparison purpose, the substitutional B concentrations are also plotted [48].

Ghandi et.al.[12] reported that both the growth rate and Ge content (or strain) affect the B incorporation in SiGe layers. By applying a high Ge content (25–30%) and low growth rate, a high B concentration in the range 4-6×1020 cm−3 can be incorporated in SiGe layers (Figure 13). Also the selective epitaxy of elevated SiGe layers suffers from the facet formation ((311) and (111)) which results in pile-up features at the edges of the openings. These features are connected to the high growth rate in the presence of defects which occurs through the strain relaxation of SiGe layers grown on the facets. Therefore, the detected B concentration at the edges of the opening is believed to be lower than the concentration at the center. The pile-up issue is less dominant for SiGe layers grown inside the recessed openings. A proposed explanation was that the deposition occurs inside openings with Si surrounding for recessed case compared to elevated SiGe layers with SiO2 oxide openings.

Born Incorporation in Si-Based Materials

19

Figure 13. Comparison of Ge content and substitutional B concentration in SiGe layers grown in recessed and elevated S/D using a reduced-pressure CVD reactor [12].

Chapter 3

MODELING OF BORON INCORPORATION IN SI/SIGE LAYERS Chemical vapor deposition (CVD) and its related techniques e.g. reduced pressure CVD, metallorganic CVD, low pressure CVD and ultrahigh vacuum CVD have been widely utilized to grow epitaxial doped semiconductor layers. In order to achieve this goal, an absolute understanding of its reaction mechanism, kinetics and thermodynamics is required. This understanding can be reported as chemistry, hydrodynamic or microscopic models. Modeling of the CVD process has been performed to connect macroscopic process conditions (such as precursors’ partial pressure, the growth pressure and temperature) to macroscopic and microscopic film properties (such as thickness, uniformity of the profile, crystallinity, morphology, chemical composition and purity). Since 1970s, many models have been developed to comprehensively describe the physical and chemical phenomena occurring during this process. From 1990, by developing the computational technology, it has become achievable to integrate detailed descriptions of multi-species and multi-reaction chemistry into a single computational model. Inherently CVD process is the decomposition of the precursor gases inside the chamber over the substrate and deposition of a film from contained material in these gases. Several presursors are available in the market for Si epitaxy such as silane (SiH4), dichlorosilane (SiH2Cl2), trichlorosilane (SiHCl3), tetrachlorosilane (SiCl4), disilane (Si2H6) and trisilane (Si3H8). The most commonly used precursor for p-type doping is

22

M. Kolahdouz, H. H. Radamson and R. Ghandi

diborane (B2H6). Several groups have investigated the incorporation of B in Si epitaxy from various precursors [53-57]. There are a few models providing reasonable explanation for some aspects of Si doping by CVD process using a special precursor. In this chapter, we go through the models which employ the most common Si source (silane), and diborane to dope the epitaxial films. During nonselective epitaxial growth, SiH4 decomposes dominantly to silylene (SiH2) and two hydrogen atoms which these Hydrogen atoms occupy the dangling bonds on the surface. However, in order to perform selective epitaxy, SiH2Cl2 (DCS) is the preferred source. The selective epitaxy growth of Si is a deposition process where the layer is only deposited on exposed crystalline areas and not on the oxide (or nitride). This type of growth can be obtained by introducing SiH2Cl2 and HCl to the deposition chamber in order to suppress the formation of nucleation sites on the oxide. In the case of DCS, similar decomposition phenomenon happens but there are three possibilities for a silicon atom on the surface; Hterminated silicon site, Cl-terminated silicon site or it is a free silicon site and ready to adsorb another atom (dangling bond). Hydrogen desorbs at much lower temperatures than chlorine which explains the smaller deposition rates obtained with chlorosilanes compared to that of silane [58]. In Si CVD, B2H6 is the typical doping precursor for manufacturing p-type material. The B incorporation of in Si epitaxy has been investigated by several groups using various precursors [53-57]. An increase in the Si growth rate has been reported at low temperatures [5357]. The reports explained this behavior by increased H2 desorption due to presence of B on the surface [59]. Grutzmacher [60] has made an analytical approach to study the effect of B-doping on the growth rate of atmospheric pressure CVD of Si. A simple model based on the assumption that each B atom on the surface generates a two-dimensional nucleation sites for Si was proposed to interpret the growth rate enhancement. The article formulated the B incorporation into the B growth rate ( B) which is hypothetically the B-crystal deposition rate on Si substrate. The equation for B is given by: (1)

Modeling of Boron Incorporation in Si/Sige Layers

23

where N, NB and are the number of atoms in a unit volume of the Si crystal, B concentration and the growth rate of B-doped Si film, respectively. They observed that at 750˚C, B is linearly proportional to P(B2H6) with slope 1. As illustrated in Figure 14, the growth rate enhancement due to B becomes more sensible by decreasing the temperatures (slope of about 1.3 in the log-log plot).

Figure 14. The B growth rate vs. the B2H6 partial pressure at 550, 625 and 750°C [60].

They clarified this situation by considering pseudo-equilibrium between adsorbing B at the surface and desorbing B or BnHm compounds from the surface in this temperature range which is given by: (2) where the activation energy, Ea, is 1.09 ± 0.13 eV between 550 and 750 ˚C. By a Phenomenological Consideration, the overall growth rate can be written as a summation of “undoped Si growth rate”, “the effective increase in the Si growth rate” and “B growth rate” as follows:

24

M. Kolahdouz, H. H. Radamson and R. Ghandi (3)

By introducing the doping factor m = be rewritten as:

Si/B/

B,

the total growth rate can

(4) The doping factor is an important parameter which exposes the growth rate mechanism (see Figure 15). It can be defined as the number of Si atoms incorporated additionally for each B atom into the film. This factor decreases with increasing boron concentration. This means more Si atoms can incorporate per B atom when small number of B atoms exists on the surface. m can be calculated as follows: (5) where NML is the number of atomic sites per monolayer and NB is extracted from the SIMS data. NSi is deduced from the lowest doping level showing a growth enhancement by fitting the calculated m at this data point. NSi =l×1012 cm-2and NSi =2×1012 was determined at 550 and 625°C, respectively.

Figure 15. Comparison of experimental data (data points) and model calculations (solid and dashed lines) of the growth rate doping factor vs. the boron concentration [60].

Modeling of Boron Incorporation in Si/Sige Layers

25

Remarkably good agreement between experimental data and the calculated factor m is demonstrated, indicating that Eq. (5) gives the correct analytical relation between the growth rate enhancement and the B concentration in the film. Therefore, they concluded that the growth rate enhancement of B doped Si films observed using APCVD can be described thoroughly by the formation of additional growth nuclei from each B atom exist on the surface. They declared that hydrogen desorption increase due to the presence of B is not the main reason behind the enhanced growth rate phenomena. According to this report, the limiting factor in Si APCVD is the formation of growth nuclei at the surface. Mehta at al. [61] have presented a kinetic model for B doping in Si thermal CVD using SiH4 as the precursor. The model is based on (1) the collision theory of heterogeneous unimolecular elementary reactions, (2) statistical physics, and (3) the concept of competitive adsorption. The activated flux of a precursor on the surface [62,63] can be approximated by: (6) where p, m, and Ea are precursor’s partial pressure, molecular mass, and activation energy for its heterogeneous decomposition reaction on the surface, respectively. In the gas-phase, diborane homogeneously decomposes to 2BH3 (B2H6 →2BH3). After the heterogeneous reactions involved by employing SiH4 and B2H6, the number of H-terminated Si and B sites increases. Since the activation energy for reactions on Hterminated Si and B sites are larger compared to that of H-free ones, these fluxes can be neglected. Thus, the total Si and B fluxes are given by: (7) (8) By substituting Eq. 6 in Eq.7 and Eq. 8 the primary Si and B fluxes can be written as:

26

M. Kolahdouz, H. H. Radamson and R. Ghandi

θH(Si) and θH(B) are introduced as the ratio of H-terminated Si sites to all Si sites and the ratio of H-terminated B sites to all B sites, respectively. p and m are signs for gas partial pressure and molecular mass. Each equation has its individual activation energy unique to the particular reaction. γ represents the ratio of B sites to all surface sites and is calculated as [B]/N0, where [B] is the B concentration and N0 is the Si atomic density (5×1022 atoms/cm3). The hydrogen surface coverage plays an important role in the kinetics behavior. During the epitaxy, there is a balance between H2 desorption and SiH4 adsorption which bring H to the surface. The mentioned reactions are listed below:

The equilibrium constant for reaction [R3] is given by: (9) where A is the pre-exponential factor, H is the reaction enthalpy, and pH2 is the H2 partial pressure. The best fit to experimental data in the temperature range 425-700°C and the pressure range 0.3-3 mtorr was obtained with A = 3.036×109 and H = 50 kcal/mol [61]. The total growth rate (G) is proportional to the total flux which is the combination of Eq.7 and Eq.8. However, they neglected B fluxes in Eq.8

Modeling of Boron Incorporation in Si/Sige Layers

27

due to low B concentration [60].The total growth rate can thus be expressed by: (10) By substituting the fluxes, Eq.10 can be rewritten as:

(11)

At very low B2H6 partial pressure, growth rate enhancement is not observed and thus, the first part of Eq. 10 is dominant. In this case, ESiH4 on Si=28.9 kcal/mol was calculated from Arrhenius plot of the growth rate. As illustrated in Figure 16, Grutzmacher [60] reported that the nonlinearity of the growth rate vs. B2H6 partial pressure originates from JSiH4 on B.

Figure 16. Dependence of overall growth rate on B2H6 partial pressure in the temperature range 550-850°C: (—) from Mehta at al.[61] model and data points from [60].

28

M. Kolahdouz, H. H. Radamson and R. Ghandi

γ (m in the previous model [60]) is experimentally found to be linearly proportional to B2H6 partial pressure, therefore, 1-θH(B) is nonlinear with B2H6 partial pressure and suggested to be estimated from: (12) Eq. 12 implies that θH(B) increases with B2H6 partial pressure at a given temperature. Although, This parameter (θH(B)) plays a critical role in growth rate enhancement, they could not clarify the behavior given by Eq.12. According to this article, the increased growth rates at low temperatures are attributed to enhanced H2 desorption from B sites on the surface. By substituting Eq.12 in eq.11, the total growth rate can then be given as: (13) The hole (positive carrier) concentration is proportional to the B concentration in the film which can be extracted from the fraction of the B and Si fluxes as follows: (14) By considering complete decomposition reaction of B2H6 (PBH3→2PB2H6) and substituting the fluxes and Eq.12 in Eq.14, the following equation is obtained: (15) Mehta at al. [61] have also presented the following table composed of the constants used in modeling the B concentration and the Si growth rate (Eq. 13 and 15)

Modeling of Boron Incorporation in Si/Sige Layers

29

In the last two decades, silicon-based group IV hetero-devices have attracted attention for manufacturing of high-speed devices in telecommunication. Fabrication of composed Si-based films, e.g. Si1xGex requires advanced techniques (CVD) with atomic-order surface reaction control. H. Kuhne et al. [64] established a model in which epitaxial Si1-xGex:B thin films with very low boron contents were formulated. It is suggested in the article that in the kinetically controlled process, thin film composition is determined by the competition of the different species involved in the film growth. This competition is governed by the partial pressures and the apparent reaction rate constants corresponding to the decomposition and the incorporation reactions. These effects can quantify by: (16) where VL is the total thin film growth rate, Ii is the incorporation flux of the ith thin film constituent, NL is the total number of atoms per cm3 of the thin film, constj(T) is the apparent reaction rate constant related to j , Pj is the partial pressure related to j, j is the source gas related to i, l/n is the apparent order of the chemical decomposition reaction of j and Vi is the partial growth rate of the ith constituent. n for silane , germane and diborane are respectively 2,1 and 1. The enhancement in the growth rate, as previously mentioned in Eq.4, is acknowledged by representing additional partial rates as multiples of the Ge and B growth rates as follows: (17) (18)

30

M. Kolahdouz, H. H. Radamson and R. Ghandi

In order to perform a high quality epitaxy growth, due to low B concentration, VB and its additional partial rate (nVB) may be neglected in the total growth rate calculation. (19) Since there is no significant interaction between the B2H6 and GeH4, a change in the B partial pressure does not perturb the Ge contribution in the growth rate ((1+m)VGe). However, by variation in the layer composition, NL in Eq.4 will change. The modified NL can be estimated by: (20) where NA is Avogadro's number, ρ and M are the specific density and molar mass. This equation can be applied when NB 0.1 mtorr became amorphous in the small openings and the HRXRD layer peak disappeared. It was announced in this article that the maximum B active concentration is size dependent so the process has to be calibrated for the opening sizes of interest. Radamson et al. [71] that the feature of B concentration in SiGe layers (shown in Figure 22) is similar to Figure 20, since the incorporation of B in SiGe depends strongly on the growth rate and Ge content (or strain). Thus, any increase (or saturation) of these parameters may influence the B concentration in these layers.

38

M. Kolahdouz, H. H. Radamson and R. Ghandi

Figure 22. Growth rate and active boron concentration for B-doped SiGe layers grown at 650 °C depending on the chip exposed Si coverage. (The applied partial pressure of DCS, HCl and GeH4 are 60, 20 and 0.5 mtorr, respectively) [71].

All the curves in Figures 20 and 22 illustrate a linear increase of growth rate with decreasing the exposed Si coverage until a saturation region is reached. In epitaxy, the growth rate amount relates to the availability of the molecules and gas consumption rate over a chip [84]. This becomes in balance for low exposed Si coverage which may lie below 1%. This is another good reason for the fact that the exposed Si coverage of the chip is the dominant factor in the pattern dependency. According to the previous report [50], the B-doped SiGe layers in both recessed and unprocessed openings have higher growth rate compared to intrinsic layers. The B atoms on the surface enhance the growth rate by acting as desorption sites for Cl and H. This increase in the growth rate of B-doped layers is diminished when carbon is also introduced in SiGe layers.

Modeling of Boron Incorporation in Si/Sige Layers

39

Figure 23. Dopant concentration calculated from the shift of the layer peaks in HRXRD rocking curves. Identical dots were measured on different chips (different exposed Si coverage) of one wafer. Dichlorosilane, germane, methylsilane, HCl and diborane partial pressures were 60, 1.2, 0.3, 20 and 3.6 mtorr, respectively [50].

It is illustrated in Figure 23 that carbon doping level follows inversely the Ge content (see Figure 20). C concentration monotonically decreases in SiGe layers grown in smaller exposed Si coverage. Meanwhile, the boron doping level follows the Ge content and growth rate as expected. The results in this article indicate that the measured strain compensation of each dopant (B or C) in SiGe layers is additive. This has been concluded under the assumptions of high epitaxial quality and epi-layers with no strain relaxation.

Chapter 5

CONCLUSION B-doped SiGe(C) layers have been widely used as the base region in the high frequency HBTs and also as the stressor material in source/drain in pMOSFETs. Several high-temperature treatments are involved during the fabrication of HBTs and MOSFETs. The out-diffusion of B from the SiGe is the most important issue since it results in degradation of high frequency behavior or short channel effect of the source and drain junctions. Spike anneal using rapid thermal annealing (RTA) with very high ramp rates can be one of the solutions to avoid the TED of boron. Another way to retard the boron out-diffusion is integration of C in SiGe layer. B incorporation in SiGe layers is affected by the growth rate and Ge content (or strain). In selective epitaxy growth, the B incorporation also suffers from pattern dependency in recessed and unprocessed openings. The exposed Si coverage of a chip is one of the most important factors to control layer profiles of epitaxial grown SiGe. A mask with non-uniform pattern experiences inhomogeneous gas distribution and the part of chip with less exposed Si coverage has to be compensated in order to obtain a uniform layer profile.

REFERENCES [1]

[2]

[3]

[4]

P. Chevalier, B. Barbalat, L. Rubaldo, B. Vandelle, D. Dutartre, P. Bouillon, T. Jagueneau, C. Richard, F. Saguin, A. Margain, and A. Chantre, “300 GHzfma self-aligned SiGeC HBT optimized towards CMOS compabitility,” IEEE Bipolar/BiCMOS Circuits and Technology Meeting, 2005, pp. 120-123. S. Van Huylenbroeck, A. Sibaja-Hernandez, R. Venegas, S. You, G. Winderickx, D. Radisic, W. Lee, P. Ong, T. Vandeweyer, N.D. Nguyen, K. De Meyer, and S. Decoutere, “A 400GHz f MAX Fully Self-Aligned SiGe : C HBT Architecture,” IIEEE Bipolar/BiCMOS Circuits and Technology Meeting, 2009, pp. 811. M. Khater, J.-s Rieh, T. Adam, A. Chinthakindi, J. Johnson, R. Krishnasamf, M. Meghelli, F. Pagette, D. Sanderson, C. Schnabel, K.T. Schonenberg, P. Smith, K. Stein, A. Stricker, S.-j Jeng, D. Ahlgren, and G. Freeman, “SiGe HBT Technology with fmU / fT = 350 / 300 GHz and Gate Delay Below 3 . 3 ps,” Electron Devices Meeting. IEDM Technical Digest., 2004, pp. 247-250. J. Grahn, H. Fosshaung, M. Jargelius, P. Jonsson, M. Linder, B.G. Malm, B. Mohadjeri, J. Pejnefors, H.H. Radamson, M. Sanden, Y.-B. Wang, G. Landgren, and M. Ostling, “A lowcomplexity 62-GHz fT SiGe heterojunction bipolar transistor process using differential epitaxy and in situ phosphorus-doped poly-Si emitter at very low thermal budget,” Solid-State Electronics, vol. 44, Mar. 2000, pp. 549-554.

44 [5]

[6]

[7]

[8] [9]

[10]

[11]

[12]

[13]

M. Kolahdouz, H. H. Radamson and R. Ghandi J. Eberhardt and E. Kasper, “Modelling of SiGe heterobipolar transistors: 200 GHz frequencies with symmetrical delay times,” Solid-State Electronics, vol. 45, Dec. 2001, pp. 2097-2100. K. Oda, E. Ohue, I. Suzumura, R. Hayami, A. Kodama, H. Shimamoto, and K. Washio, “Self- Aligned Selective-EpitaxialGrowth Sil-x-yGexCy HBT Technology Featuring 170-GHz fmax,” Electron Devices Meeting. IEDM Technical Digest., 2001, pp. 332-335. T. Ghani, M. Armstrong, C. Auth, M. Bost, P. Charvat, G. Glass, T. Hoffiann, K. Johnson, C. Kenyon, J. Klaus, B. Mclntyre, K. Mistry, A. Murthy, M. Silberstein, S. Sivakumar, P. Smith, K. Zawadzki, S. Thompson, and M. Bohr, “A 90nm High Volume Manufacturing Logic Technology Featuring Novel 45nm Gate Length Strained Silicon CMOS Transistors,” Electron Devices Meeting, 2003. IEDM Technical Digest. IEEE International, 2003, pp. 978-980. N. Mohta and S.E. Thompson, “Mobility Enhancement,” IEEE CIRCUITS & DEVICES MAGAZINE, 2005, pp. 18-23. S.E. Thompson, M. Armstrong, C. Auth, S. Cea, R. Chau, G. Glass, T. Hoffman, J. Klaus, Z. Ma, B. Mcintyre, a Murthy, B. Obradovic, L. Shifren, S. Sivakumar, S. Tyagi, T. Ghani, K. Mistry, M. Bohr, and Y. El-Mansy, “A Logic Nanotechnology Featuring Strained-Silicon,” IEEE Electron Device Letters, vol. 25, Apr. 2004, pp. 191-193. S. Gannavaram, N. Pesovic, and M.C. Ozturk, “Low Temperature (