Introduction to Silvaco ATHENA Tool and Basic Concets in Process Modeling Part - 1 Instructor: Dragica Vasileska Deartment of Electrical Engineering Arizona State University
1. Introduction to Process Simulation The farication rocess of an integrated circuit consists of the following main stes: ❶ Eitaxial growth ❷ oxidation, assivation of the silicon surface ❸ Photolithograhy ❹ diffusion ❺ metalization A schematic descrition of a lanar rocess for the farication of a n-junction, consists of the following stes: 1. Eitaxial growth: Eitaxial n-layer -sustrate High-temerature rocess (~1000 C) The amount of doant atoms determines the conductivity of the layer
. Oxidation and Photolithograhy Eitaxial n-layer -sustrate oxidation SiO Eitaxial n-layer -sustrate hotolithograhy Diffusion window Thermal oxidation leads to formation of oxide layer for surface assivation Photolithograhy allows roer formation of the diffusion window 3. Diffusion and Metalization stes n-layer -sustrate diffusion n-layer -sustrate metalization The diffusion rocess gives rise to the n-junction (takes lace at ~1000 C) Electrical contacts are formed via the metalization rocess
The sequence of events that lead to successful farication of the device structure are the following: otimization Faricate device structure Simulation relacing exerimental stes: ATHENA Process simulation tool Perform electrical characterization ATLAS Device simulation tool no yes Design condition met?
Physically-ased rocess simulation redicts the structure that results from secified rocess sequence Accomlished y solving systems of equations that descrie the hysics and chemistry of semiconductor rocesses Physically-ased rocess simulation rovides three major advantages: ❶ it is redictive ❷ it rovides insight ❸ catures theoretical knowledge in a way that makes the knowledge availale to non-exerts Factors that make hysically-ased rocess simulation imortant: ❶ quicker and cheaer than exeriments ❷ rovides information that is difficult to measure
The rocessing stes that one needs to follow, for examle, for faricating a 0.1 µm MOSFET device, include (in random order): ❶ Ion imlantation rocess ❷ Diffusion rocess ❸ Oxidation rocess ❹ Etching models ❺ Deosition models In the following set of slides, each of this rocess is descried in more details with the aroriate statements and arameter secification.
Some historical dates: - Biolar transistor: 1947 - DTL - technology 196 - Monocrystal germanium: 1950 - TTL - technology 196 - First good BJT: 1951 - ECL - technology 196 - Monocrystal silicon: 1951 - MOS integrated circuit 196 - Oxide mask, - CMOS 1963 Commercial silicon BJT: 1954 - Linear integrated circuit 1964 - Transistor with diffused - MSI circuits 1966 ase: 1955 - MOS memories 1968 - Integrated circuit: 1958 - LSI circuits 1969 - Planar transistor: 1959 - MOS rocessor 1970 - Planar integrated circuit: 1959 - Microrocessor 1971 - Eitaxial transistor: 1960 - I L 197 - MOS FET: 1960 - VLSI circuits 1975 - Schottky diode: 1960 - Comuters using - Commercial integrated VLSI technology 1977 circuit (RTL): 1961 -...
. Descrition of the Ion Imlantation Process Ion imlantation is the most-frequently alied doing technique in the farication of Si devices, articularly integrated circuits. Two models are frequently used to descrie the ion imlantation rocess: ❶ Analytical models: do not contriute to hysical understanding can e adequate for many engineering alications ecause of its simlicity ❷ Statistical (Monte Carlo technique): first rinciles calculation (time consuming) can descrie arasitic effects such as: - lattice disorder and defects - ack scattering and target suttering - channeling (imortant in crystalline mater.)
(A) Analytical Models For all of the analytical models, the real ion distriution in 1D is given the following functional form: C ( x) = Df ( x) D total imlanted dose er unit area f(x) roaility density function, frequency function - descried with the following four characteristic quantities: ❶ Projected range R : ❷ Standard deviation R P : = + R xf x) dx ( R = ( x R ) ❸ Skewness γ: ❹ Excess or kurtosis β: γ = + ( x R ) 3 ( R ) 3 f ( x) dx β = + + ( x R ) 4 ( R ) 4 f ( x) dx f ( x) dx 1/
Analytical distriutions most frequently used for descriing doing rofiles are: ❶ Simle Gaussian or normal distriution ❷ Joined half-gaussian distriution ❸ Pearson tye IV distriution Simle Gaussian or normal distriution 1D model Makes use of the rojected range R and the standard deviation R : ( ) D x R ( ) C( x) = ex π R R Has γ=0 and β=3. The aroximation of the true rofile is only correct u to first order, since it gives symmetric rofiles around the eak of the distriution. Range arameters R and R for all the imuritymaterial cominations are stored in the ATHENAIMP file.
The model is activated via the GAUSS arameter on the IMPLANT statement; R (RANGE) and R (STD.DEV) Other arameter that has to e secified is the dose D (via the arameter DOSE on the IMPLANT statement) Pearson distriution 1D model This is a standard model in SSUPREM4, and is used for generating asymmetrical doing rofiles. The family of Pearson distriution functions is otained as a solution of a differential equation: ( ) ( ) ( ) ( ) + + + + = + + = 1 0 1 0 1 1/ 0 1 1 0 4 arctan 4 / ex ) ( ) ( ) ( R x a R x R x K x f x x x f a x dx x df
The tye of the Pearson distriution deends uon the sign of the term: D = 4 0-1. Only the Pearson IV (D>0) distriution has the roer shae and a single maximum. The constants a, 0, 1 and are related to the moments of f(x) in the following manner: R a = γ β 3γ = A ( ) ( ) β + 3 R 4β 3γ A, 6, The vertical doant concentration is then roortional to the ion dose: C ( x) = Df ( x) This simle model can fail in the case when channeling effects are imortant (dual Pearson model has to e used) 0 = A A = 10β 1γ 8, 1 = a
The model is activated via the PEARSON arameter on the IMPLANT statement. Other arameters that can e secified in conjunction with the model choice include: ❶ Lattice structure tye: CRYSTAL or AMORPHOUS ❷ Imlant material tye: ARSENIC, BORON, etc. ❸ Imlant energy in kev via ENERGY arameter ❹ For dual-pearson model, another arameter is imortant and descries the screen oxide (S.OXIDE) through which ion imlantation rocess takes lace
Two-dimensional imlant rofiles D analytical imlant models are quite rudimentary and usually ased on a simle convolution of a quasi-one dimensional rofile C(x, t mask (y)) with a Gaussian distriution in the y-direction: MASK IONS C( x, y) = 1 πσ y + C( x, t mask ( y'))ex ( y y' ) σ y - indeendent of deth (rolem) dy' In the case of an infinitely high mask extending to the oint y = a, the convolution can e erformed analytically, to give: σ 0 y < a C( x, tmask( y)) = C( x) y a a y = C( x, y) 1 C( x) erfc ; erfc( x) = e σ y π x y y (lateral) x (deth) t dt
Additional Parameters that need to e secified for D ion-imlantation rofiles are: ❶ Tilt angle: TILT ❷ Angle of rotation of the imlant: ROTATION ❸ Imlant erformed atall rotation angles: FULLROTATIO ❹ Print moments used for all ion/material cominations: PRINT.MOM ❺ Secification of a factor y which all lateral standard deviations for the first and second Pearson distriution are multilied: LAT.RATIO1 and LAT.RATIO
(B) Monte Carlo Models Analytical models can give very good results when alied to ion-imlantation in simle lanar structures. For non-lanar structures, more sohisticated models are required. SSUPREM4 contains two models for Monte Carlo simulation: ❶ Amorhous material model ❷ crystaline material model The Monte Carlo model can also deal with the rolem of ion imlantation damage: Damage tyes: Frankel airs (Interstitial and Vacancy rofiles), <311> clusters, Dislocation loos Two models exist for ion imlantation damage modeling: ❶ Kinchin-Pease model (for amorhous material) ❷ Crystalline materials model
(C) Some examles for analytical models Imlant of hoshorus with a dose of 10 14 cm - and Gaussian model used for the distriution function. The range and standard deviation are secified in microns instead of using tale values. IMPLANT PHOS DOSE=1E14 RANGE=0.1 STD.DEV=0.0 GAUSS 100 kev imlant of hoshorus done with a dose of 10 14 cm - and a tilt angle of 15 to the surface normal. Pearson model is used for the distriution function. IMPLANT PHOSPH DOSE=1E14 ENERGY=100 TILT=15 60 kev imlant of oron is done with a dose of 4 10 1 cm -, tilt angle of 0 and rotation of 0. Pearson model for the distriution function is used that takes into account channeling effect via the secification of the CRYSTAL arameter. IMPLANT BORON DOSE=4.0E1 ENERGY=60 PEARSON \ TILT=0 ROTATION=0 CRYSTAL
(D) Characteristic values for the ion-imlantation rocess Dose: 10 1 to 10 16 atoms/cm Current: 1 µa/cm to 1 A/cm Voltage-energy: 10 to 300 kv After the fact annealing: 500 to 800 C Advantages of the ion imlantation rocess: Relatively low-temerature rocess that can e used at aritrary time instants during the farication sequence.