reactor- and feature-scale simulations for semiconductor...

G. Schulze-IckingMP PDT CS SIM2003-05-15

Reactor- and Feature-Scale Simulationsfor Semiconductor Manufacturing

- an excursion into different scales & complexities -

Dr. Georg Schulze-Icking

Infineon Technologies AG


Content

! equipment + feature ⇔ process + device simulation

! motivation

! equipment & feature scale simulation at IFX

� overview

� computational fluid dynamics

� coupling of scales

� feature scale simulation

! summary & outlook

! acknowledgement


Difference to Prozess & Device Simulation

! process & device (bulk):� dopant concentrations, (crude) topography,...� electrical behaviour (e.g. I-V curves),...

! equipment & feature scale (gas phase + surface):� dep/etch uniformity, precursor depletion, T distribution,... (wafer scale)� step coverage, selectivity, microtrenching,... (sub-µ scale)


Motivation

! production issues:� shrinking structures ⇒ more stringent specifications

! dep/etch uniformity across wafer, high aspect ratios,...� novel production processes

! high-k materials, atomic layer deposition,...� tough competition and timelines...

! simulation support:� allows separation of competing effects� helps understanding of complex processes → improvement of recipes� fast optimization cycles (after calibration)� cheap...


Equipment & Feature Scale Simulation at IFX

! equipment simulation:� reaction kinetics (Chemkin)� fluid dynamics & radiation (CFD-ACE)� plasma & sheath (several)� molecular dynamics (custom)� quantum chemistry (Material Studio)� ...

! feature scale simulation:� AP-CVD, electroplating (Evolve)� LP-CVD, plasma etch,... (Topsi, custom)� �


Computational Fluid Dynamics

! simulation of thermal processes (e.g. CVD) needs to account for several effects:

� gas flux (Navier-Stokes)� gas phase & surface chemistry� heat (convection, conduction, radiation, and reaction enthalpy)

! simulation yields e.g. dep/etch rate on wafer, bulk stoichiometry! issues: availability of models, breakdown at low-p (Knudsen limit),...

0.05 0.10 0.15

1.0

1.1

1.2

D [a

.u.]

r [m]

⇒


CFD: Chamber Clean

! CVD chambers need to be cleaned frequently to avoid particles� for SiO2, NF3 clean (with in- or ex-situ plasma) common� NF3 most expensive process gas → large gain from optimizing clean

! example: impact of gas inlet on clean time (transient):

original inlet modified inlet


CFD: HDP-CVD

! deposition across wafer needs to be very uniform (typically <2%)! issues becomes worse for larger wafers (12'' vs. 8'')! example: impact of reactor conditions on deposition uniformity

experiment simulationpres [SiH2]


Coupling of Scales: IPVD

! for plasma processes, effects on very different scales need to be considered:

Plasma Model

Sheath Model

Feature Model

Molecular Dynamics Model

M+

M

Plasma

Angular andEnergy Distribution

Transportof M, M+,Ar+

and Deposition

Reaction RatesM, Ar =>Surface

Wafer

1 mm

<1 µ

1 nm

Plasma

Pkap

Pind


Plasma Simulation

! plasma simulation (cont. or MC) yields e.g. plasma density, electron temperature, and bias voltage. Has to account for:

� coupling of different phases (neutrals, ions, electrons), each with separate T� complex chemistry (Tgas ≤ 1000 K; Tel ≈ 5⋅104 K)� self-consistent solving of charge distribution and E-fields (incl. external)� ...


Plasma Sheath

! existence of a plasma sheath (e- depletion zone):� electron are orders of magnitude faster than ions

→ e- depletion at surfaces (similar to diode...)→ strong electric field, acceleration of ions towards surface (increased by

external RF bias)→ deceleration of e-

� in steady state ion and electron flux is equal!

∆t ≈ 10-8 s


Plasma Sheath Simulation

! sheath important, determines ion energy & angle distribution on substrate. Hybrid FD-MC sheath simulator

� solves E-field and charge distribution self-consistently� accounts for gas phase collisions incl. charge exchange� applied RF bias,...

0.002

0.006

0.01

0.014

200 600 1000

Ion Energy [eV]

IED

[a.u

.]

5 mTorr; nPlasma=1010 cm-3

eion+

neutral

plasma

substrate


Molecular Dynamics

! ions sputter surface, depending on energy and angle of incidence.! experimentally hard to access (low E) → obtain probability & yield from MD,

"simple" Newtonian dynamics but with� complex multi-body potentials� vastly different timescales btw. events: O(ns) for impact, O(s) for diffusion

many!

⇒


Feature Scale Simulation

! using the above results, e.g gap fill or sputter damage can be studied (details of simulator below):

Ti+Ti+ / Ar+

Ti

Ti / Ti+

deposition

sputteringreflection


Feature Scale Simulator: Motivation & Requirements

! motivation� extreme geometries → transport important� processes increasingly complex (plasma, superfill,�)� more stringent specifications with decreasing groundrule

! requirements� high accuracy (processes already good�)� higher order reactions� transient surface chemistry� 2D, 2.5D, and 3D� speed�

! currently no simulator available that fulfills above requirements...


Topsi Feature Scale Simulator: Structure

! goal: temporal evolution of sub-µ structure subject to process:

! program structure:� flux calculation� surface chemistry� front propagation

time

front propagation

flux calculation

surface chemistryfront

local dep/etch rate


Level Set Front Propagation: Basic Idea

� represent front as zero level of the distance function φ(x,y)� calculate normal velocity FN at front and extend to grid� evolve φ(x,y) (instead of front itself�)� extract new front from φ�(x,y)

φ(x,y)

front (φ=0)front (φ=0)

φ<0: insideφ<0: inside

φ>0: outsideφ>0: outside distancesdistances

FNFN


Level Set Front Propagation: Equation

for every point at the front holds:

0ty

ytx

xt=

∂∂

∂φ∂+

∂∂

∂φ∂+

∂φ∂

0vt

=φ∇+∂φ∂ r

0NF||t

=φ∇+∂φ∂

0)t),t(y),t(x( =φ

total derivative wrt. t:

with normal velocity: vNFr

φ∇φ∇=

NFtttt ⋅φ∇⋅∆−φ=∆+φ→

φ<0: inside

φ>0: outside

tt ∆+φ

tφ

� LS equation is solved on grid points for time-step ∆t.

� after time-step a new front is extracted as zero level of φt+∆t(x,y).


Level Set Front Propagation: Advantages

shock front topology Change accuracy

readily extended to 3D!


Monte Carlo Transport: Motivation

transport effects important: - precursor depletion in structure- energetic sputtering of material- Reactive Ion Etch (RIE) lag...

tracing of particles (MC) accounts for these and...- is applicable to high aspect ratios (DT)- is easily extented to 3D- has no principal errors, only statistical...


Monte Carlo Transport: Efficiency

desired accuracy requires many MC particles (≥105 per step)

MC flux calculation in Topsi is accelerated using:- adaptive quad-/octtrees- parallelization using MPI- several statistical enhancements

growingfront

initialfront

quadtree

particle


Chemistry Solver: Requirements

! common processes exhibit:� higher order reactions� energetic reactions (plasma processes)� chemi- & physisorbed layers (e.g. ALD and DT etch)� complex surface chemistry (stiff ODE) � transient surface quantities (e.g. ALD)

! chemistry solver needs to account for these...


Simple CVD

! precursor depletion results in non-conformal deposition. Gap fill improves with decreasing taper & sticking coefficient:

taper

a=85°s=10-2

a=87.5°s=10-2

a=90°s=10-2

a=87.5°s=10-3

stick


HDP-CVD: Impact of Bias

! energetic ions from plasma:� sputter surface (high bias only)� activate surface� both yields depends on ion angle & energy

! neutrals deposit on active sites! re-deposition of sputtered material

high bias: low bias:


HDP-CVD : Impact of Environment

! ideal CVD process:� fills structures of different AR void-free� has deposition height which is independent of design (for CMP)

! however, reality is not ideal...


ALD: Transient Coverage

! Transport effect in (ideal) ALD� steady state obvious� study temporal evolution

DR=0

DR=0

DR=x

S=1

S=0

S=x


Simple CVD 3D

! simple CVD (s=0.02) into comb-like structure. 3D geometry introduces complex shadowing effects (visualization with vtk).


Summary & Outlook

! summary� equipment & feature scale simulation support unit process development � overview of IFX simulation activities� coupling of scales important (e.g. plasma)� computational fluid dynamics (briefly)� custom feature scale simulator Topsi� versatile problems...

! outlook� develop new chemistry models (bottleneck!)� improve support for current processes� develop support for future processes� ...


Acknowledgement

! thanks to my colleagues who performed most of the presented simulations:

� Gerd Enders� Dr. Werner Jacobs� Dr. Alfred Kersch� Dr. Gerhard Prechtl� Dr. Winfried Sabisch

reactor- and feature-scale simulations for semiconductor...

Documents