Physical Vapor Deposition (PVD): SPUTTER DEPOSITION

‹ We saw CVD Gas phase reactants: Pg ≈ 1 mTorr to 1 atm. Good step coverage, T > > RT

‹ …PECVD Plasma enhanced surface diffusion without need for elevated T

−6 ‹ We will see evaporation: Evaporate source material, Peq.vap. Pg ≤10 Torr (another PVD) Poor step coverage, alloy fractionation: ∆ Pvapor

‹ We will see Dry etching Momentum transfer and chemical reaction from plasma to remove surface species

‹ Now sputter deposition. Noble or reactive gas P ≈ 10 -100 mTorr

What is a plasma?

Oct. 17, 2005 6.152J/3.155J 1 What is a plasma? A gas of ionized particles, typically noble gas (e.g. Ar+ + e-)

P ≈ 10-100 m Torr cathode Ionization Ionization anode Ionization E potential of Ar event x potential of Ar v − - v + e ≈ 16 eV Ar ⊕ ≈ 16 eV V ≈ 1kV

Plasma is only self-sustaining over a range of pressures: typically 1 or 10 mT > P > 100 mT.

To understand “Why this pressure range?”, we need to understand what goes on inside a plasma?

Oct. 17, 2005 6.152J/3.155J 2 First, what is molecular density at 10 mT?

k BT Ideal gas: n = P/(kBT) λ = 2π d 2 P 2.5 x 1025 m-3 25 λ (cm)

24 10-3 Log[n (#/m3)] λ ≈ 3 cm 23 Ar 10-2 ( λ[Ar+ ] much less) 1 Atm= -1 22 10 0.1 MPa ≈14 lb/in2 21 100 1 Torr n = 3.2 x 1020 m-3 10 mT 20 101

012345 Log[P (N/m2)]

At 10 mT, molecular spacing ≈ n-1/3 = 0.15 microns. Is this = λ ?

Oct. 17, 2005 6.152J/3.155J 3 z -1/3 z z z z Spacing between molecules ≈ n = 0.15 microns. z z z z z z

kB T λAr ≈ 3 cm λ = 2 2πd P ( λ[Ar+ ] much less) 3k T Thermal velocity of Ar, v ≈ B ≈ 103 m/s m Ar

What’s final velocity of ions at x = λ? E field accelerates Ar+, e- between collisions.

Eq 2 2 5 7 v f = v 0 + 2ax ≈ 2 λ v ≈ 4 ×10 m/s v − ≈ 2 ×10 m/s, m Ar e

(only 0.1% to 1% of nAr are ions):

Oct. 17, 2005 6.152J/3.155J 4 3 3k B T Be clear about different velocities: vkT ≈ 10 m/s v ≈ 5 m Ar vAr+ ≈ 4 × 10 m/s 7 ve- ≈ 2 × 10 m/s

So we have: low-vel. Ar+ high vel. e-

Ionization cathode anode event Ex ve− - vAr+ ⊕ V ≈ 1kV

The plasma is highly conducting due to electrons:

J = nqv x Thus Je- >> JAr+

at Eq λ nq2 λ ne2τ J = σE = nqv ≈ nq ≈ nq σ e− ≈ = e− x 2 2m v 2m v 2m

Oct. 17, 2005 6.152J/3.155J 5 Plasma is self-sustaining only for 1 or 10 mT < P < 100 mT.

“Why this pressure range?”

Necessary conditions: If pressure is too low, λ is large, too few collisions in plasma to sustain energy 1) λ < L k T λ = B so collisions exchange energy 2π d 2 P within plasma

If pressure is too high, λ is small, very little acceleration between collisions

1 2 + mv f = 2ax ≈ Eqλ 2) EK > ionization potential of Ar 2

Oct. 17, 2005 6.152J/3.155J 6 Plasma is self-sustaining only for 1 or 10 mT < P < 100 mT. 1 V k BT 2 λ = mv f ≈ Eqλ ≈ λ(eV) 2 Self-sustained 2 L 2π d P plasma Log[λ (m)] 100 V 1000 V EK (eV) 0 λ ≈ 3 cm: E ≈ 0.3 x V 2 K Ne+ 22 + + Ar 16 -1 2) E > Ar 1 + 1) λ < L ≈ 10 cm K Kr 13

-2 0

-3 -1

-4 -2 10 mT: λ ≈ 3 cm

-5 -3 -1012345 Log[P (N/m2)]

Oct. 17, 2005 6.152J/3.155J 7 What about Glow? Cathode Anode Ex Ar+ impact v − on cathode e => Lots more electrons, plus v Ar + sputtered atoms x 2 EK ~ ve − 3 eV ¨ v 2 = 2ax cathode anode § f 1.5eV ¦ Ionization - ⊕ glow Faraday dark space x e-s swept out

e − e− ¦ EK <~ 1.5eV § 1.5 ~~< E K < 3eV ¨ EK > 3eV

Cathode dark space, e- - induced => ionization, no action optical excitation High conductivity plasma of Ar ⇒ visible glow

Oct. 17, 2005 6.152J/3.155J 8 3 eV ¨ Inside a plasma 2 E ~ v − K e § (D.C. or “cathode sputtering”) 1.5eV Ionization ¦ glow x

cathode anode

- ⊕

J e− , v e− >> J Ar + ,v Ar + ⇒

Surfaces in plasma charge negative, Cathode Anode attract Ar+ repel e- ∴ plasma ≈ 10 V positive Target + VD.C. Ar+ relative to anode Target species e- Cathode sheath: Substrate Plasma low ion density High conductivity Oct. 17, 2005 6.152J/3.155J 9 Which species, e- or Ar+ , is more likely to dislodge an atom at electrode ?

P ≈ 10-100 m Torr cathode anode Cathode Ex v − is “target”, e - vAr+ ⊕ source material V ≈ 1kV

Sputter removal (etching) At 1 keV, vAr+ = ve/43 occurs here

pe = mv Momentum transfer: PAr = MV = 1832mv/43 ≈ 43pe-

No surprise. from ion implantation, 4M1M2 ∆E = E1 2 most energy transfer when: ()M1 + M2 i.e. incoming particle has mass close to that of target. Oct. 17, 2005 6.152J/3.155J 10 Sputtering process Ar+ impact, momentum transfer at cathode ⇒ 1) e- avalanche and 2) released target atoms, ions.

Atomic billiards

Elastic energy transfer

E2 E 4M M 2 ∝ 1 2 cos2 θ θ 2 E2 greatest for M1 ≅ M2 E1 ()M1 + M2

E1 E 4M - 2 1 For e hitting anode, substrate, M1 < < M2 ≈ (small) E1 M 2 - But e can give up its EK in inelastic collision: 1 m v2 ⇒∆U Excitation of atom or ion 2 e e Oct. 17, 2005 6.152J/3.155J 11 Sputtering process: ablation of target

P ≈ 10-100 m Torr cathode anode Cathode Ex v − is “target”, e - vAr+ ⊕ source material V ≈ 1kV

cathode anode - + ⊕ Ar V ≈ 1kV

Momentum transfer of Ar+ on cathode erodes cathode atoms Mostly-neutral source atoms Target material (cathode) ⇒ flux to anode, substrate. (lots of e’s around) must be conductive …or must use RF sputtering (later) Oct. 17, 2005 6.152J/3.155J 12 How plasma results in deposition 5) Flux => deposition at anode: 1) Ar+ accelerated to cathode Al, some Ar, some impurities

cathode anode 2) Neutral + Ar+ target species (Al) Al kicked off; Ar =Ar+ + e- Ar 3) Some Ar, - ⊕ V ≈ 1kV Ar+ and e- also. Al e- - O Al 4) e- may ionize + Ar+ impurities Al (e.g. O => O-) 6) Some physical resputtering of film (Al) by Ar

Oct. 17, 2005 6.152J/3.155J 13 Sputtering rate of source material in target is key parameter. Typically 0.1 - 3 target atoms released/Ar incident Sputtering rates vary little from material to material.

Vapor pressure of source NOT important (this differs greatly for different materials).

cathode anode 2) Neutral + Ar+ target species (Al) Al kicked off. Ar =Ar+ + e- Ar - ⊕ V ≈ 1kV Al

Al + Ar+ 6) Some physical Al resputtering of Al by Ar cos θ Oct. 17, 2005 6.152J/3.155J 14 Sputtering yield # atoms, molecules from target S = Sputtering yield = # incident ions Surface E1 binding energy ⎡ ⎤ E ⎛ E ⎞ Eb 2 ⎢ 2 ⎥ S = σ 0nA × 1+ ln⎜ ⎟ 4E thresh ⎣⎢ ⎝ E b ⎠ ⎦⎥ E2 >Et

Random walk 2 # excited in E 2 4M1M2 2 π d # / area ∝ 2 cos θ each layer; to surface; E1 ()M1 + M2 E = surface E = energy b thresh binding energy to displace interior atom φ

Oring, Fig. 3.18 S

Sputter rate depends on angle of incidence, relative masses, kinetic energy. 90° φ Oct. 17, 2005 6.152J/3.155J 15 Figure removed for copyright reasons.

Table 3-4 in Ohring, M. The Materials Science of Thin Films. 2nd ed. Burlington, MA: Academic Press, 2001. ISBN: 0125249756.

Oct. 17, 2005 6.152J/3.155J 16 Figure removed for copyright reasons.

Figure 12.13 in Campbell, S. The Science and Engineering of Microelectronic Fabrication. 2nd ed. New York, NY: Oxford University Press, 2001. ISBN: 0195136055.

Sputter yield vs. ion energy, normal incidence

Oct. 17, 2005 6.152J/3.155J 17 Sputtering miscellany

Isotropic flux: Anisotropic flux: cos θ = normal component cosn θ: more-narrowly of flux directed at surface (surface roughness) Higher P Poor step coverage. Lower P

Large target, small substrate => Good step coverage, more uniform thickness Moving substrates

k T λ = B p λ (cm) Higher pressure => shorter λ, 2πd 2 P better step coverage …but more trapped gas 10 mT 2 1 mT 20

Oct. 17, 2005 6.152J/3.155J 18 Target composition vs. film composition Sputtering removes outer layer of target; can lead to problem with multi-component system, but only initially

Initial target composition A B 3 If sputter yield A > B

A3B A A B Surface 2.5 A B composition 2 of target B

time Deposited film initially richer in A than target; film composition eventually correct. A B Another approach: Co-sputtering => composition control; multiple targets & multiple guns. Also can make sample “library”.

A2B AB AB2 Oct. 17, 2005 6.152J/3.155J 19 Varieties of sputtering experience

D.C. sputter deposition: Only for conducting materials. if DC sputtering were used for insulator. e.g. carbon, charge would accumulate at each electrode and quench plasma within 1 - 10 mico-sec.

cathode anode

t < 1 micro sec C Ar =Ar+ + e-

C Ar - ⊕ V ≈ 1kV e- O- What does 1 microsec + Ar+ tell you about ion speed? C (>106 cm /sec)

‹ Therefore, use RF plasma…

RF -sputter system is basically a capacitor with gas dielectric. Energy density ↑ as f ↑

Oct. 17, 2005 6.152J/3.155J 20 ‹ RF plasma sputtering Vplasma 7 −1 Target Plasma conducts at low f ω p ~10 s Thus, V = 0 in plasma

Substrate Plasma potential still V > 0 due to high e- velocity, (-) charging of surfaces. Potential now symmetric.

Mobility of target species (concentration gradient )still => sputtering in 1/2 cycle: 2eV v ≤ ≈ 7 ×104 m /s M

Vplasma Smaller target => higher field Target

m ⎛ ⎞ + V1 A2 + Ar + = ⎜ ⎟ m ≈ 1,2 + Ar V2 ⎝ A1 ⎠ Substrate Some re-sputtering of wafer F.C.C. reserves: 13.56 MHz for sputtering

Oct. 17, 2005 6.152J/3.155J 21 Varieties of sputtering experience V plasma V ≈ ‹ RF Bias sputtering Target bias -100 V Negative wafer bias + enhances re-sputtering of film V ≈ + Ar+ Ar + -1 kV Substrate Some re-sputtering of wafer

Bias is one more handle for process control.

Used in SiO2: denser, fewer asperities. Figure removed for copyright reasons. Bias affects stress, resistivity, Figure 3-24 in Ohring, 2001. density, dielectric constant…

Oct. 17, 2005 6.152J/3.155J 22 Film stress in RF sputter deposition

Cathode

Film Species Ar+

-Vbias

Better step coverage

Oct. 17, 2005 6.152J/3.155J 23 Varieties of sputtering experience

‹Magnetron sputtering use magnets B Cathode ur v ur 2 mv N S N F = qv× B + E q mv r = target ()x F = , r qB N S ¦ vz Bx ⇒ Fy ⇒ vy S Ions spiral around B field lines § vy Bx ⇒ Fz ⇒ vz

v ≈ 6 × 106 m/s for electron, r < 1 mm for 0.1 T

B field enhances time of e- in plasma V+

+ ⇒ more ionization, greater Ar density. Ex

v0

Oct. 17, 2005 6.152J/3.155J 24 ‹ Reactive sputter deposition: Mix reactive gas with noble gas (Ar or Ne). analogous to PECVD

Ti or TiN target

Ar +N2 TiN

Also useful for oxides: SiOx, TiO, CrO… other nitrides: SiN, FeN,…

Oct. 17, 2005 6.152J/3.155J 25 Figure removed for copyright reasons. Figure 12.27 in Campbell, 2001.

Resistivity and composition of reactivity sputtered TiN vs. N2 flow.

Oct. 17, 2005 6.152J/3.155J 26 Thin film growth details (R < 1) Rate of arrival R ≡ Diffusion rate

1) Arrival rate, 3) Chemical physical reaction 5) Growth adsorption

Anode

4) Nucleation 6) Bulk diffusion

2) Surface diffusion

If R > 1, these processes have reduced probability

Oct. 17, 2005 6.152J/3.155J 27 ‹ Sputtering metals Can use DC or RF

‹ Sputter alloys, compounds Concern about different sputter yield S

e.g. Al Si But S does not vary as much as Peq.vap S = 1.05 0.5 which controls evaporation

Sputter target T < < Tevap No diffusion, High diffusion, surface enriched in composition change, species low S components distributed over entire source

Oct. 17, 2005 6.152J/3.155J 28 Figure removed for copyright reasons. Table 3-7 in Ohring, 2001.

Oct. 17, 2005 6.152J/3.155J 29