Regulating Solid State Diffusion in Semiconductor …eseebauer/talks/Ssdiffusion.pdf1 Regulating Solid State Diffusion in Semiconductor Processing Edmund G. Seebauer Department of

1

Regulating Solid State Diffusion in Semiconductor Processing

Edmund G. SeebauerDepartment of Chemical Engineering

University of Illinois at Urbana-Champaign

Support: NSF, DOE, Chartered Semiconductor

2

MOSFET Schematic

Source Gate Drain

Silicon substrate

Interconnect

Glass

SiO2Channel

Silicide

Poly-Si

Heavily-doped Si

3

International Technology Roadmap for Semiconductors

! Feature size decreases 30% per 3 yrs! Cost/function decreases 50% per 3 yrs

4

Transistor Scaling and Diffusion

! Surface/volume ratios increase! Surface/interface diffusion can dominate transport! Nonthermal means needed to enhance surface diffusion

5

Application: Hemispherical Grained Silicon for DRAMs

! Formed by heating amorphous Si

! Crystallization begins at free surface

! Grains mushroom from surface by surface diffusion

! Rough surface increases capacitance/area

6

UIUC-grown HSG

Top view Cross-section

100µµµµm100µµµµm

7

Application: Reflow

! Smooths surfaces of! Al, Cu for metallization

! Glasses for passivation

! Submicron length scales → surface diffusion dominates

! Simulators model effects accurately

Cu on 0.35µm trench

450°C375°C

8

Application: Sputter DepositionBallistic flux shadowing Diffusional smoothing

Simulation byGROFILMS

Actual Ti film

Process simulators can predict growth behavior

9

Surface Second Harmonic Generation

! Interface polarization P ~ χ1•E + χ2:EE + …

! χ1 yields everyday reflection: adsorption insensitive

! χ2 yields SHG: adsorption sensitive

! For centrosymmetric bulk, χ2 ≠ 0 only at interface

ν2ν

Laser PMT

Filter

10

Second Harmonic Microscopy

Laser

Array

! Spatially resolved SHG, D from profile spreading

! Image before and after heating

! Magnification ~ 20x, diffraction limited

11

Experimental Set-up

! Signal averaging at 10 Hz takes ~ 5 min

! Imaging does not induce profile spreading

J. Opt. Soc. Am. B 10 (1993) 546

12

0.0 0.5 1.0 1.50

2

4

6

θθθθ

SH S

igna

l (ar

b un

its)

Calibration Curve: Ge on Si(111)

! Developed by Auger spectroscopy

! Linear up to ~0.7 ML

13

Diffused Step

SH S

igna

l (ar

b un

its)

Pixel Number

350 400 450 500 5500.2

0.4

0.6

0.8

1.0

1.2

Diffused Step

θ θθθ Ge

Position (µm)300 450 600 750

0.0

0.2

0.4

0.6

Second Harmonic Images: Ge on Si(111)

Unprocessed profiles After image processing

14

T (oC)

D (c

m2 /s

)

1000/T (K-1)0.85 0.90 0.95 1.00 1.05 1.10

10-11

10-10

10-9

10-8

Do = 6 x 102 cm2/s

840 780 730 680

Ea= 2.48 eV

Arrhenius Plot: Ge on Si(111)

Phys. Rev. B 55 (1997) 13304

! Activation energy, prefactor both enormous

! Independent of surface concentration

15

Surface Diffusion Physics on Si at High T

" Composite surface diffusivity DM depends on:

" Mobility of mobile species

" Number of mobile species

" Etotal = Ehop + Epair formation ≈ 0.5 + 2.0 eV

" Do,total = (10-3 cm2/s) exp(∆Spair formation/kT)

≈ 106

16

Illumination-Influenced Surface Diffusion: Experimental Results

390440 350 315635 560 495

T (oC)

D (c

m2 /s

)

1000/T (K-1)1.1 1.2 1.3 1.4 1.5 1.6 1.7

10-11

10-10

10-9

10-8

10-7

,, n

p

Illuminated

Thermal

! In, Ge, Sb on Si(111)Phys. Rev. Lett. 81 (1998) 1259Phys. Rev. B 61 (2000) 13710

! Measured by second harmonic microscopy

! D increases for n-type, decreases for p-type material

17

Surface Diffusion Physics on Si at High T

" Composite surface diffusivity DM depends on:

" Number of vacancies,mobile atoms

" Numbers depend on charge state

" For all adsorbates (at 2 W/cm2):" Ediff changes by 0.3 eV" Prefactor changes by ~ 100×

" Invariance with adsorbate suggests underlying commonality: Ionization of surface vacancies

18

Quantum Calculations

MotivationCharge states of Si surface vacancies unknown, difficult to measure experimentally

Quantum calculations can predict vacancy energy levels and formation energies

ImplementationAb-initio pseudopotential calculation using DFT

Supercell approach, periodic boundary conditions

CASTEP software (Accelrys Inc.)

Si(100)-(2×1) surface: mono- and divacancy

19

Formation Energies for Si Monovacancy

" Four stable charge states (+2,0,-1,-2)

" Comparable results for other vacancy types

Si surface supports a variety of charged vacancies, consistent with picture for optically driven diffusion

Fermi Energy (eV)0.0 0.2 0.4 0.6 0.8 1.0 1.2

Form

atio

n En

ergy

(eV)

-1

0

1

2

-2

-1

+2

+1

0

Upper Monovacancy0 K

20

Analogies between Surface, Bulk Diffusion

Bulk Surface

Hopping Interstitial motion Adatom motion

diffusion Vacancy motion Vacancy motion

Kick-in/kick-out Exchange

Overall mass Clustering Islanding transport Vacancy-interstitial

formation Vacancy-adatom

formation

Point defect ionization

Point defect ionization

21

Bulk Diffusion Physics in Si at High T

" Rate of total mass transport depends on:

" Mobility to mobile species

" Number of mobile species

" Numbers and mobilities may depend on charge state → photoexcitation

22

Ultrashallow Junction Formation

" Implantation leaves lattice damage

" Dopant must be activated electrically

" High-powered lamps

" Rapid heating/cooling

" Defects promote unwanted diffusion

Ion Implantation

Rapid Thermal Annealing

23

Measurement of Bulk Diffusion

" Isotopically-depleted layer grown epitaxially via LPCVD

30Si28Si (natural)

30Si (depleted)

30Si (natural)

28Si (enriched)

" 30Si step profile created

" Doping level uniform (eliminates drift effects)

24

1e+17

1e+18

1e+19

1e+20

1e+21

1e+22

100 200 300 400 500 600

x (nm)

30Si

(cm

-3)

hvdarkas-grown

Non-thermal Illumination Effects: p-type

! 850ºC, 4 hr, 1×1019 cm-3 p-type, 0.05 W/cm2

D(hν) = 7×10-16 cm2/s

D(dark) = 1×10-15

cm2/s

Diffusion inhibition by a factor of 1.5

25

Non-thermal Illumination Effects in Bulk Si

! 900ºC, 30 min, 1×1019 cm-3 n-type, 2 W/cm2

1e+17

1e+18

1e+19

1e+20

1e+21

1e+22

100 200 300 400 500 600

x (nm)

30Si

(cm

-3)

hvdarkas-grown

D(hν) = 9×10-14 cm2/s

D(dark) = 1×10-15

cm2/s

Diffusion enhancement by

a factor of 90

Results like surface diffusion, impt in rapid thermal processing

26

Application for Nonthermally-Driven Diffusion: Nanoparticle Growth on a-Si

• Anneal (645°C)

75 sec 90 sec 105 sec

Motivated by Hemispherical Grained Si growth for DRAMs

• Grow amorphous Si by chemical vapor deposition

• Grain growth rate may be nonthermally influenced

27

Directed Self-Assembly Using Amorphous Semiconductors

Ostwald ripening

Expose a-Si with subcritical nuclei to hν or e-beam just below crystallization temp

100µµµµm

Dots, walls 2 – 200+ nm

28

Applications of Self-Assembly Method

! Nonvolatile memory at high density, low power! Sensors! Photonic band gap materials! Flat panel displays! Solar cells! On-chip nanopore devices for probing electrical

activity of biological molecules

Particle arrays, walls, narrow pores for:

29

Diffusion at Solid Interfaces: Copper in Interconnect Structures

! Diffusion of Cu at interfaces remains uninvestigated! Barrier layers

! Etch stops

! Interlayer dielectrics

! Could be major means of transport during processing, use

! Problems with line-to-line shorting, Si deep levels…

We seek to make first measurements of Cu diffusion at these interfaces

30

Problems

! Inaccessibility of solid-solid interfaces! Electron, ion spectroscopies no good

! Small number of diffusing atoms (< 1 monolayer)

! Need for probe with interface specificity! Raman, IR, ellipsometry no good

We employ optical second harmonic microscopy

31

Second Harmonic Microscopy at Solid Interfaces

Laser

Array

Solid 1

Solid 2

! Implementation similar to free surfaces

! Create profile ex situ

! Image through transparent overlayer

32

Measurement Concept

! Position Cu source near one end of interface

! Heat structure

! Freeze concentration profile and image

! Structure below made at Chartered Semiconductor

SiO2

Si3N4

Si3N4

Cu! 5 nm Ta

separates Cu from ILD

33

Preliminary Imaging Results

700°C

! Profiles are very flat! Diffusion through Ta barrier rate limiting

! Diffusion at upper or lower interface?

34

Low-Energy Ion-Surface Interactions

! In beam-assisted deposition, plasma etching, effects of substrate T usually ignored

! In much early work, Eion >> kT

! Repulsive potentials govern

! Now Eion closer to kT, leading to possible E-T interactions

! Attractive potentials more important

! Question: What are mechanisms for these interactions?

! Now nearly virgin territory

35

Ge/Si(111) Surface Diffusion: Physical-Chemical Interactions

Dion, 54 eV

860 770 690 620T (oC)

D (c

m2 /s

)

1000/T (K-1)

0.88 0.96 1.04 1.1210-11

10-10

10-9

Dion, 35 eV

Dthermal

Dion, 65 eV

Two temperature regimes:

Low T → only prefactor increasesHigh T → prefactor and Ediff decrease

Phys. Rev. Lett., 82 (1999) 1185

36

Calculation Method: MD Simulation

! Si or Ge (111)-(1×1)! Ne+, Ar+, Xe+ < 65 eV! Stillinger-Weber/

Universal pot.

! Monitor frequency (per ion) of 4 possible events:

37

Energy Effects, Constant T

E (eV)10 20 30 40 50 60 70

Per

Impa

ct P

roba

bilit

ies

0.0

0.1

0.2

0.3

0.4 SputteringAdatom formationKnockinBulk vacancy

Ar-Si, 1100K

Thresholds

Determine EThres where f ≤ 0.01

Fits use f = A (E1/2 – EThres1/2)

38

Temperature Effects, Constant E

T (K)0 200 400 600 800 1000 1200

Per i

mpa

ct p

roba

bilit

y

0.0

0.1

0.2 Ar-Si sputtering65eV

Threshold temperature exists!

Occurs for all events, such as sputtering, knock-in

Fits use f = A(T – TThres)

39

At Threshold, T and Ion Energy are Coupled

T (K)0 200 400 600 800 1000 1200

E Thre

s(eV)

0

20

40

60

80

Si

Ge

Direct trade-off between E and TETotal = σkT + EThres (new result!)

~100eV ~10-65eV ~10-65eV

Large amplification factor σ ~ 103

Ar-Si sputtering

40

Physics: Ion Mass Variation

Atomic Number

0 10 20 30 40 50 60

E Tota

l(eV)

80

120

160

200

240

σ σσσ

-2400

-2000

-1600

-1200

-800

-400

(Ne) (Ar) (Xe)

Si Sputtering

Look for “standard” momentum matching effectsExpect big variation with mass, but observe little

41

Chemistry: Substrate Variation

T (K)0 200 400 600 800 1000 1200

E Thre

s(eV)

0

20

40

60

80

Si

Ge

σ (slope) invariant

ETotal (intercept) scales with cohesive energy, melting temp

! Look for bond strength effects

42

Experimental Energy Effects

E (eV)0 20 40 60

D (1

0-10 c

m2 /s

)

0.5

1.0

1.5

2.0

2.5

3.0Ar-Si(111) 900K

Threshold

Determine EThres where D increases, using SHM

Fits use: Dion = A (E1/2 – EThres1/2) + Dthermal

43

E-T Coupling Strength Matches Simulation

T (K)880 920 960 1000 1040

E Thre

s (eV

)

12

14

16

18

20

22

24

Ar-Si

ETotal σσσσ

Sim. 70 to 100 700 Expt. 110 1200

44

Unimportance of Ion Mass Matches Simulation

Atomic Number

0 10 20 30 40 50 60

E Tota

l(eV

)

80

120

160

200

240

σ σσσ

-2400

-2000

-1600

-1200

-800

(Ne) (Ar) (Xe)

Low T (900-1000K)

Observe almost no variations!

45

Apparent Mechanism

! Low E! Long-range interaction! Corrugation flat! Nonlocal momentum transfer

! High E! Short-range interaction! Corrugation rough! Local momentum transfer

! Intermediate E! Medium-range interaction! Corrugation asperities only if

target atom ~0.3 A off site

46

Summary

! New, nonthermal ways (hν, ions) to transport semiconductor atoms have come to light

! Numerous applications in microelectronic processing (RTP, dry etching, ion-assisted deposition) and nanotechnology

! Solid interfaces SHM imaging demonstrated

47

Acknowledgements to:

! Postdocs! Harry Ho Yeung Chan

! Graduate Students! Kapil Dev! Rod Ditchfield! Mike Jung! Diana Llera-Hurlburt! Shih Hwee Tey! Mike Wang