A. Boogaard

Low-Temperature Fabricated TFTs on Polysilicon Stripes

This paper presents a novel approach to make high-performance CMOS at low temperatures. Fully functional devices are manufactured using back-end compatible substrate temperatures after the deposition of the amorphous-silicon starting material. The amorphous silicon is pretextured to control the location of grain boundaries. Green-laser annealing is employed for crystallization and dopant activation. A high activation level of As and B impurities is obtained. The main grain boundaries are found at predictable positions, allowing transistor definition away from these boundaries. The realized thin-film transistors (TFTs) exhibit high field-effect carrier mobilities of 405 cm2/Vmiddots (NMOS) and 128 cm2/Vmiddots (PMOS). CMOS inverters and fully functional 51-stage ring oscillators were fabricated in this process and characterized. The process can be employed for large-area TFT electronics as well as a functional stack layer in 3-D integration.

Langmuir-probe Characterization of an Inductively-Coupled Remote Plasma System intended for CVD and ALD

We measured electron density and electron energy distribution function (EEDF) vertically through our reactor for a range of process conditions and for various gases. The EEDF of Ar plasma could largely be described by the Maxwell-Boltzmann distribution function, but it also contained a fraction (~ 1E-03) of electrons which were faster (20-40 eV). At low pressures (6.8-11 µbar), the fast-electron tail shifted to higher energies (Emax ~ 50 eV) as we measured more towards the chuck. The fast-electron tail shifted to lower energies (Emax ~ 30 eV) when we increased pressure to 120 µbar or applied an axial magnetic field of 9.5 µT. Addition of small amounts of N2 (1-10%) or N2O (5%) to Ar plasma lowered the total density of slow electrons (approx. by a factor of two) but did not change the shape of the fast-electron tail of the EEDF. The ionization degree of Ar-plasma increased from 2.5E-04 to 5E-04 when a magnetic field of 9.5 µT was applied.

Green laser crystallization of α-Si films using preformed α-Si lines

In this work, amorphous silicon films with preformed a-Si lines were crystallized using a diode pumped solid state green laser irradiating at 532 nm. The possibility of controllable formation of grain boundaries was investigated. The crystallization processes in the rapidly melted silicon films were discussed. The influence of the crystallization parameters (i.e., energy density, scan velocity, etc.) and structure type (i.e., with and without preformed lines) on properties of the crystallized films was studied. The laser treatment with an energy density of 1.00 J/cm2 at a laser pulse overlapping of 90% provided the optimal crystallization process with predefined grain boundary location. X-ray diffraction (XRD), SEM and AFM microscopy have been used to characterize the crystallized silicon films.

On the verification of EEDFs in plasmas with silane using optical emission spectroscopy

We measured the electron density and electron energy distribution function (EEDF) of plasmas in our reactor which is intended for silicon oxide and nitride deposition. Langmuir-probe measurements showed that the EEDF of Ar plasma could largely be described by the Maxwell-Boltzmann (MB) distribution function, but it also contained a fraction (~0.5 %) of fast electrons in the energy range between 20 and 40 eV, strongly deviating from the MB distribution. We also measured relative mean electron temperatures (kTe) by optical emission spectroscopy (OES) which were calibrated by the absolute Langmuir-probe measurements. The kTe as measured by OES in Ar plasma decreased from 1.7 eV at 1.1 Pa to 1.4 eV at 12 Pa, while Langmuir-probe measurements showed a decrease from 1.7 eV to 0.8 eV. This difference is caused by the OES method, which is especially sensitive to the fraction of fast electrons in the plasma. OES can be used instead of Langmuir-probe measurements when depositing plasmas are used. Combining both methods, we demonstrated that EEDFs as measured by the Langmuir probe in Ar-N2, and Ar-N2O plasmas, resemble EEDFs in plasmas with small additions of silane, provided that (a) precursor fractions in plasma are small (SiH4 {less than or equal to} 0.8 % and N2O {less than or equal to} 15 %), and (b) total pressure does not exceed 3.6 Pa (27 mTorr). As such, the measured EEDF without silane can be used as input for chemical modeling and optimization of deposition processes in plasmas containing silane.

Electrical properties of plasma-deposited silicon oxide clarified by chemical modeling

Our study is focused on Plasma Enhanced Chemical Vapor Deposition (PECVD) of silicon dioxide films at low temperatures (< 150 oC) using Inductively Coupled (IC) High-Density (HD) plasma source. We recently fabricated Thin Film Transistors (TFTs) with high-quality ICPECVD gate oxides, which exhibited a competitive performance. For better understanding of the influence of deposition parameters on both the deposition kinetics and oxide quality, we have modeled the Ar-SiH4-N2O plasma system with 173 chemical reactions. We simulated concentrations of 43 reactive species (such as e.g. SiHx radicals and SiHx + (x=0-3) ions, polysilanes, SiO, SiN, SiH3O, SiH2O, HSiO, etc., as well as atomic hydrogen, nitrogen and oxygen) in plasma. We further used our simulations to qualitatively explain (in terms of concentrations of the reactive species) the influence of SiH4/N2O gas-flow ratio and total gas pressure on film electrical properties and deposition rate.

Impact of Small Deviations in EEDF on Silane-based Plasma Chemistry

In this work, we emphasize the importance of using a correct Electron Energy Distribution Function (EEDF) to model chemical reactions in High-Density (HD) low-pressure silane-containing plasmas. We have modeled chemical reactions in Ar-SiH4-N2O- (N2-H2-O2) Inductively Coupled Plasma Enhanced Chemical Vapor Deposition (ICPECVD) system, intended for deposition of silicon oxide and silicon nitride layers. For the modeling, we used the experimentally measured EEDF, deviating from the Maxwell-Boltzmann (MB) EEDF. We demonstrate that the use of an inappropriate (i.e. MB in our example) EEDF, only slightly deviating from the experimental (i.e. actual) distribution, could lead to significant discrepancies (1-2 orders of magnitude) between the calculated and actual radical densities.

Negative Charge in Plasma Oxidized SiO2 Layers

Silicon dioxide (SiO2) gate dielectric layers (4-60 nm thick) were deposited (0.6 nm/min) on n-type Si by inductively-coupled plasmaenhanced chemical vapor deposition (ICPECVD) in strongly diluted silane plasmas at 150°C . In contrast to the well-accepted positive charge for thermally grown SiO2, the net oxide charge was negative and a function of the layer thickness. Our experiments suggested that the negative charge was created due to unavoidable oxidation of the silicon surface by plasma species, and the CVD component adding a positive space charge to the deposited oxide. The net charge was negative under process conditions where plasma oxidation played a major role. Such conditions included low deposition rates and relatively thin grown layers. Additional measurements showed that the negative charge in SiO2 also persisted on p-type substrates. We suggest that plasma oxidation of the silicon surface results in SiO2 layers with a surplus of oxygen. This surplus of oxygen is able to accumulate a negative charge. This assumption is addressed in this paper by a review of earlier work on silicon oxidation, and by a first series of experiments wherein oxygen is implanted into thermal SiO2. It is shown that the implantation can result in a negative charge to the bulk oxide layer. The effect of the negative charge on the flatband voltage can be described by the implantation profile.