modeling of trench filling during ionized metal …

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

MODELING OF TRENCH FILLINGDURING IONIZED METAL PHYSICAL VAPOR DEPOSITION+

JLU_AVS 00

Junqing Lu* and Mark J. Kushner**

*Department of Mechanical and Industrial Engineering

**Department of Electrical and Computer Engineering

University of Illinois at Urbana-Champaign

October 2000

+Supported by SRC, NSF and DARPA/AFOSR

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

AGENDA

JLU_AVS 01

• Motivation and introduction to Cu IMPVD

• Description of the model and the sputter algorithm

• Metal densities in Cu IMPVD

• Ion and neutral distributions

• Surface diffusion model for profile simulation

• RF coil voltage

• Trench filling

• Radial locations

• Pressures

• Magnetron and ICP power

• Aspect ratio of the trench

• Conclusions

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

MOTIVATION FOR Cu IONIZED PVD

JLU_AVS 02

• The resistance of Cu is only half that of Al.

• To increase processor speed, Cu is replacing Al as the metal for interconnect wiring.

• The filling of large aspect ratio trenches benefits from ionized PVD.

• Ions are able to fill deep trenches because their angular distributions are narrowed by an rf bias.

METALATOMS

SiO2

METAL

VOID

METALIONS

SiO2

METAL

RFBIAS

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

COMPUTATIONAL PLATFORM

JLU_AVS 03

Hybrid Plasma Equipment Model(HPEM, including Plasma Chemistry Monte Carlo Simulation*)

Flux to wafer

Monte Carlo Feature Profile Model (MCFPM)

Angular and energy distributions

*PCMCS generates angular and energy distributions for the depositing fluxes, using species sources and time-dependent electric fields obtained by HPEM.

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

FEATURES OF SPUTTER MODEL

JLU_AVS 04

• Ion energy-dependent yield* for sputtered atoms.

• The effective yield of 1 for the reflected neutrals.

• Ion energy-dependent kinetic energy

• Sputtered atoms: Cascade distribution

• Reflected neutrals: TRIM** and MD***

• Cosine distribution in angle for sputtered and reflected atoms emitted from target.

• Momentum and energy transfer from sputtered and reflected atoms to background gas (sputter heating).

• Electron impact ionization for in-flight sputtered and reflected neutrals.

• Source terms for thermalized sputter species and gas heating.

*Masunami et al., At. Data Nucl. Data Tables 31, 1 (1984) . **D. Ruzic, UIUC.***Kress et al., J. Vac. Sci. Technol. A 17, 2819 (1999)

00 200 400 600

0.5

1.0

1.5

Incident Ar + Energy (eV)

2.0

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

Cu IMPVD: METAL DENSITIES

JLU_AVS 05

• Reactor is based on *Cheng et al.

• Operating conditions:

• 40 mTorr Ar

• 1.0 kW ICP

• 20 V rf voltage on coils

• 0.3 kW magnetron

• -25 V dc bias on substrate

• Cu peaks below the target since most of the sputtered Cu atoms are thermalized a few cm below the target.

• Cu+ peaks at the center due to peak in plasma potential.

*Cheng, Rossnagel, and Ruzic, JVST 13(2), 1995, p. 203.

Radius (cm)

Substrate

Target MagnetInlet

Density

(cm-3)

9.0 x 1010

9.0 x 109

1.0 x 109

Cu Density Cu+ Density

Cu Target

Target to Substrate Distance = 13 cm

Substrate

Coils

InletMagnet

Outlet

0 5 10 155101518 18

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

Cu IMPVD: ELECTRON TEMPERATURE AND DENSITY

JLU_AVS 06

• The electron temperature is > 3 eV throughout the reactor.

• The large Te near the coils is due to the large ICP power deposition in this region.

• The electron density peaks off center due to the magnetron effect and off-axis ionization by ICP power.

Te

Radius (cm)

Substrate

Target MagnetInlet

1.5 x 1012

1.6 x 1011

2.0 x 1010

0 5 10 155101518 18

3.7

3.4

3.0

e Density

eV cm-3

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

ION AND NEUTRAL SPECIES DISTRIBUTIONS

JLU_AVS 07

• Neither the ion energy nor the neutral energy are mono-energetic.

• The spread in ion energy is due to the rf voltage on the coil and collisional broadening.

• The neutral distributions in angle are broader than that for ions.

• Operating conditions: 40 mTorr Ar, 1.0 kW ICP, 20 V rf voltage on coils, 0.3 kW magnetron, -25 V dc bias on substrate

• Cu+

0 30-30Angle (deg.)

0

20

40

60

80

High

Low

• Cu*

0 50 90-50-900

0.5

1.0

1.5

High

Low

Angle (deg.)

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

MONTE CARLO FEATURE PROFILE MODEL (MCFPM)

JLU_AVS 08

• The MCFPM obtains the etch and deposition profile using ion and neutral distributions from the HPEM.

• Surface processes are implemented using a chemical reaction mechanism:

• Deposition: Cu(g) + Si(s) Cu(s) + Si(s)

• Resputtering: Ar+(g) + Cu(s) Ar(g) + Cu(g)

• The model takes account of angular and energy dependent etch and deposition rates.

• The model is able to simulate many different chemistries and materials.

SiO2

METAL

RFBIAS

Ar+ Cu+ Cu

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

TRENCH FILLING WITH AND WITHOUT DIFFUSION

JLU_AVS 09

• A diffusion algorithm was incorporated into MCFPM to reduce unphysical dendric growth.

• The diffusion probability of the depositing metal depends on the activation energy for each possible diffusion site.

• Without diffusion, the Cu films are unphysically porous and non-conformal.

• With diffusion, Cu species deposit compactly and conformally.

With Diffusion

Trench Aspect Ratio = 1.1, Trench Width = 600 nmCu+ : Cu Neutrals = 4:1

SiO2

SiO2

Width (µm)

0 0.4 0.8 1.2

0.8

0.4

0

1.2

SiO2

Width (µm)

0 0.4 0.8 1.2

0.8

0.4

0

1.2 Without Diffusion

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

TRENCH FILLING VS COIL VOLTAGE

JLU_AVS 10

• For ICP power of 500 W and 2 mTorr Ar, measured plasma-potential oscillation* ranges from 10 to 30 V, depending on the termination capacitance of the coil. • The oscillation in plasma potential extends the range of ion energies, thereby regulating the degree of sputtering.

• Operating conditions: 40 mTorr Ar, 1 kW ICP, 0.3 kW magnetron, -30 V dc bias on wafer.

Width (µm)0 0.4 0.8 1.2

0.8

0.4

0

1.2

rf coil voltage = 10 V 30 V20 V

Width (µm)0 0.4 0.8 1.2

Width (µm)0 0.4 0.8 1.2

ExcessiveSputtering

*Suzuki, Plasma Sources Sci. Technol. 9, 199 (2000).

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

TRENCH FILLING AT DIFFERENT RADIAL LOCATIONS

JLU_AVS 11

• Electric field perturbation at the edge of the wafer generates asymmetry in Cu+ distribution, which causes asymmetry in deposition profile.

• Operating conditions:

• 40 mTorr • 1 kW ICP • 20 V rf voltage on coils • 0.3 kW magnetron • -25 V dc bias on wafer

Width (µm)

0 0.4 0.8 1.2

0.8

0.4

0

1.2

Cu+ : Cu Neutrals = 4:1 2:12.3:1

Width (µm)

0 0.4 0.8 1.2

Width (µm)

0 0.4 0.8 1.2

0 40-40Angle (deg.)

0

40

80

R = 0.5 cm R=5.2 cm R = 9.9 cm Low

High

0 40-40Angle (deg.)

0 40-40Angle (deg.)

Radial Location = 0.5 cm 5.3 cm

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

TRENCH FILLING VS PRESSURE

JLU_AVS 12

• Voids form at low pressure and fill with increasing pressure.

• The ionization fraction increases with increasing pressure, due to slowing of Cu atoms which allows more ionization. • Reasons for pinch-off:

• Diffuse angular distribution of the neutrals

• Less sputtering of over-hanging deposits

• Operating conditions*:

• 1 kW ICP • 0.3 kW magnetron• -25 V dc bias on wafer

*Cheng, Rossnagel and Ruzic, JVST B 13, 203 (1995).

0.8

0.4

0

1.2

Cu+ : Cu Neutrals = 1:3 4:12:1

5 mTorr 20 mTorr 40 mTorr

600 nm

5 mTorr 20 mTorr 40 mTorr

1 µm

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

TRENCH FILLING WITH AND WITHOUT RESPUTTERING

JLU_AVS 13

• Without resputtering, the overhang deposits grow faster than the bottom deposits, leading to pinch-off at the top.

• Resputtering reduces the overhang deposits, opens up the top of the trench, and enables more fluxes to arrive at the bottom.

• Note that Ar+ contributes significantly to resputtering.

Cu+ : Cu Neutrals = 4 : 1

Width (µm)

0 0.4 0.8 1.2

0.8

0.4

0

1.2

Width (µm)

0 0.4 0.8 1.2

0.8

0.4

0

1.2 No Resputtering With Resputtering

Ar+ : Cu Total = 7 : 1

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

TRENCH FILLING VS MAGNETRON POWER

JLU_AVS 14

• As magnetron power increases, the incident ion flux and the target bias increase, and more Cu atoms are sputtered into the plasma.

• The ionization fraction of the Cu atoms decreases since more ICP power would be required to maintain the same ionization fraction.

• The small void at low magnetron power was caused by microtrenching.

• Operating conditions: 30 mTorr Ar, 1 kW ICP, -30 V dc bias on wafer.

Cu+ : Cu Neutrals = 3:1

Width (µm)

0 0.4 0.8 1.2

0.8

0.4

0

1.2

1.6:1

Magnetron Power = 0.3 kW 2 kW 0.3 kW3:1

Microtrenching

Width (µm)

0 0.4 0.8 1.2

Width (µm)

0 0.4 0.8 1.2

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

TRENCH FILLING VS ICP POWER

JLU_AVS 15

• As ICP power decreases, the power available for Cu ionization decreases, and the Cu ionization fraction decreases.

• The pinch-off at low ICP power is caused by low ionization fraction.

• Operating conditions: 30 mTorr Ar, 0.3 kW magnetron, -30 V dc bias on wafer.

Cu+ : Cu Neutrals = 3:1

Width (µm)

0 0.4 0.8 1.2

0.8

0.4

0

1.2

1.4:1

ICP Power = 1.0 kW 0.3 kW

Width (µm)

0 0.4 0.8 1.2

Width (µm)

0 0.4 0.8 1.2

2.5:10.6 kW

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

TRENCH FILLING AT DIFFERENT ASPECT RATIOS

JLU_AVS 16

• As the aspect ratio increases, trench filling becomes more difficult.

• The fluxes that are able to completely fill shallow trenches may left voids in deeper trenches.

• Operating conditions:

• 40 mTorr • 1 kW ICP • 0.3 kW magnetron • -25 V dc bias on wafer

Width (µm)

0 0.4 0.8 1.2

0.8

0.4

0

1.2

Width (µm)

0 0.4 0.8 1.2

Aspect Ratio 1.1

Width (µm)

0 0.4 0.8 1.2

Aspect Ratio 2

Aspect Ratio 3

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

COMPLETE TRENCH FILLING ATDIFFERENT ASPECT RATIOS

JLU_AVS 17

• The ionization fraction required for complete filling increases with the aspect ratio.

• The highest possible ionization fraction is about 90%, due to gas heating.

• The simulated results indicate the largest aspect ratio for a complete filling is 3, the consensus in literature for highest aspect ratio filling is 4.

• For aspect ratio > 4, experimental results suggest that tapered trench walls are needed for seed layer deposition at the bottom.

• Operating conditions:

• 1 kW ICP • 0.3 kW magnetron • -25 V dc bias on wafer • Radius = 0.5 cm

Tapered Trench

Aspect Ratio0 1 2 3

0

0.2

0.4

0.6

0.8

1.0

Void

Complete Filling

Increasing Pressure

UNIVERSITY OF ILLINOISOPTICAL AND DISCHARGE PHYSICS

CONCLUDING REMARKS

JLU_AVS 18

• An integrated plasma equipment-feature scale model has been developed and applied to IMPVD modeling.

• The depositing ions have a broadened energy distribution due to oscillation of the plasma potential.

• Surface diffusion is an important process in metal deposition.

• Electric field enhancement at the wafer edge may cause asymmetry in trench filling.

• Formation of voids in trench filling occurs when the ionization fraction of the depositing metal flux is low.

• As aspect ratio of the trench increases, the ionization fraction for complete filling also increases.

• The desirable conditions for complete trench filling are high pressure, low magnetron power, and high ICP power.