Taras A. Kirichenko

Boron diffusion and activation in the presence of other species

Modeling and experimental investigation of B equilibrium diffusivity and its activation in Si in the presence of other species, including ab initio calculations, are presented here. The results suggest that incorporating other species along with B into the Si substrate can achieve shallower junctions and higher B activation in semiconductor device applications.

Boron diffusion in silicon in the presence of other species

We model B diffusion in the category of equilibrium diffusivity in the presence of other species using ab initio calculations. The result shows that in the presence of other atoms X (X=F, N, C, Al, Ga, In, and Ge), the migration energy for B along the migration pathway is larger than the case if X=Si (0.33 eV). This suggests the reduction of B diffusion. If another B occupies that position (X=B), then a smaller migration energy is observed and enhanced diffusion is expected. The simulation results are consistent with experiments. The reason for the migration energy difference can be understood in terms of electronic and strain effects. Estimates of the enthalpy of bond formation agree with the simulation except for the low-enthalpy cases where the strain effect caused by atom-size misfit dominates.

Enhanced and retarded diffusion of arsenic in silicon by point defect engineering

Arsenic enhanced or retarded diffusion is observed by overlapping the dopant region with, respectively, interstitial-rich and vacancy-rich regions produced by Si implants. Enhanced diffusion can be attributed to interstitial-mediated diffusion during postimplant annealing. Two possible mechanisms for diffusion retardation, interstitial-vacancy recombination and dopant clustering, are analyzed in additional experiments. The point defect engineering approach demonstrated in this letter could be applied to fabrication of n-type ultrashallow junctions.

Origin of vacancy and interstitial stabilization at the amorphous-crystalline Si interface

Using plane-wave pseudopotential density functional theory calculations, we have investigated the behaviors of neutral interstitials and vacancies at the amorphous-crystalline (a–c)Si interface. A continuous random network model is employed in the construction of defect-free a-c interface structure. We find that both vacancies and interstitials prefer to reside on the amorphous side of the interface. In both cases, the most stable defects occur 3–4A from the a-c interface. Vacancy stabilization is found to be due to strain relief provided to the substrate lattice while interstitial stabilization is due largely to bond rearrangement arising from interstitial integration into the substrate lattice. We also discuss the effect of the “spongelike” behavior of the amorphous phase toward native defects on ultrashallow junction formation in the fabrication of ever-shrinking electronic devices.

Boron diffusion in strained Si: A first-principles study

We investigate B diffusion in strained Si by using first-principles density functional theory calculations. An enhancement and an anisotropy of B diffusion in biaxial tensile strained Si are found. The diffusion barrier along the strain plane (channel) is decreased while the barrier in the vertical direction (depth) remains unchanged. This anisotropy comes from the orientation dependence of the saddle point in the diffusion pathway. The formation enthalpy of B-I pair also decreases in strained Si. According to our calculations, for strained Si on a Si0.8Ge0.2 buffer layer, which is widely used in strained metal oxide semiconductor field-effect-transistor, an enhancement of B diffusivity along the channel by a factor ∼4 and a factor ∼2 in the vertical direction are expected for typical rapid thermal anneals.

Indium out-diffusion from silicon during rapid thermal annealing

The out-diffusion of indium (In) from In-implanted silicon (Si) samples that includes bare Si, samples with an oxide-cap layer, nitride-cap layer, and nitride/oxide/Si sandwiched samples, is investigated. The dose loss of In with respect to different implant energies, doses, and soak times during rapid thermal annealing (RTA) is quantified. Experimental results of bare Si samples show that over 90% of In out-diffusion happens within 1 sec of soak time in the RTA process. In the capped samples, In rapidly diffuses through the oxide layer and stops at the nitride/oxide interface. In gets piled up at the interface of Si/oxide and oxide/nitride, and nitride very efficiently prevents In out-diffusion from the oxide layer out to the nitride layer. In addition, In gets more segregated in the Si surface in the presence of boron.

Comparison of low energy BF2+, BCl2+, and BBr2+ implants for the fabrication of ultrashallow P+-N junctions

We present a comparative study of BCl2+ and BBr2+, and the traditionally used BF2+, as implant species for the formation of ultrashallow P+-N junctions. From “as-implanted” profiles, a large reduction in channeling tail has been observed for the BCl2+ and BBr2+ implants, relative to BF2+ implantation; the depths are reduced by over 200 A compared to the “equivalent” energy BF2+ implants. After annealing, the 5 keV BCl2+ implanted junctions are shallower than the 5 keV BF2+ junctions by over 150 A. The 18 keV BBr2+ implants yield junctions as shallow as 100 A, but suffer from severe doss loss problems—probably caused by enhanced surface scattering and etching for the deceleration mode implant. The 20 keV BBr2+ implants done without deceleration show no dose loss, and result in junction depths identical to those in the 5 keV BF2+ implants. Leakage current measurements indicate an increase of three orders of magnitude for the BCl2+ implanted junctions over BF2+ ones; BBr2+ implanted junctions exhibit lower l...

Physically based kinetic Monte Carlo modeling of arsenic-interstitial interaction and arsenic uphill diffusion during ultrashallow junction formation

A kinetic arsenic-interstitial interaction model has been developed to study and predict arsenic transient enhanced diffusion (TED) and deactivation behavior during ultrashallow junction (USJ) formation. This model is based on density functional theory and has been verified by previous experiments in which the significant role of interstitial mechanism in arsenic TED was revealed. The mechanism of enhanced and retarded arsenic diffusion in different point defect environments is investigated by utilizing this model in kinetic Monte Carlo simulation. The arsenic-interstitial pair, with low binding energy and low migration energy, is shown to be the major contributor to arsenic TED in silicon interstitial-rich situations. In addition, by using this model, we demonstrate the transient existence of arsenic-interstitial clusters (AsnIm) during postimplant annealing and propose their possible role in deactivation for short time annealings such as laser annealing and spike annealing. Moreover, we have developed a...

Interstitial-based boron diffusion dynamics in amorphous silicon

Using density-functional theory calculations we identified an interstitial-based fast boron diffusion mechanism in amorphous silicon. We found that interstitial-like point defects, omnipresent in as-implanted silicon, to be very stable in an amorphous network and can form highly mobile pair with Boron atoms. The transient existence of such point defects in amorphous silicon is suggested to play an important role in boron diffusion. We found the activation energy for this pathway to be 2.73 eV, in good agreement with experimental results. In addition, this mechanism is consistent with the experimentally reported transient and concentration-dependent features of boron diffusion in amorphous silicon.

Gate Disturb Reduction in a Silicon Nanocrystal Flash EEPROM by Means of Natural Threshold Voltage Reduction

Introduction As CMOS technology is scaled to the 90nm node and beyond, silicon nanocrystal nonvolatile memories are receiving increased attention as a replacement for floating gate nonvolatile memories [1, 2]. The thin dielectrics in these memories can lead to excessive gate disturb during the read operation. Of primary concern is the loss of electrons of the program state to the gate through the top oxide overlying the nanocrystals. This loss is the result of tunneling due to the high electric field between the gate and the nanocrystals. It has been shown that reducing the natural threshold voltage (Vt,nat) of the memory cell leads to a reduction in gate disturb [3]. Simple reduction of the Vt,nat by decreasing the substrate doping concentration can result in severely degraded short channel performance, as well as degraded hot carrier injection (HCI) performance during the program operation. Thus, it is desired to construct a substrate doping profile with a light surface concentration to obtain a low Vt,nat, and a heavy doping concentration just below the surface to provide robust short channel performance and good HCI programmability.