Tool- and pattern-dependent spatial variations in silicon deep reactive ion etching

Tool- and pattern-dependent spatial variations in silicon deep reactive ion etching Hayden Taylor Microsystems Technology Laboratories Massachusetts Institute of Technology 12 May 2006

Coping with spatial variation in DRIE Why non-uniformity is a problem Controlling uniformity with the mask design Characterizing tools performance Wafer-to-wafer effects Improving the mask design process 2

Non-uniformity problems in MEMS Aim: remove imbalance in MIT Microengine rotor ~10 mm SEM picture of turbine blades Mask layout e.g.: A.H. Epstein et al., Proc. Transducers 97 3

Non-uniformity problems in MEMS also: embossing stamp fabrication for microfluidics manufacture Si stamp Thermoplastic polymer Cover perhaps too: avoiding footing during SOI etch-through 4

Inductively-coupled plasma in DRIE chamber cross-section vacuum chamber ~10-100 mtorr gas inlet wafer plasma X ~ r.f. supply to excite plasma to wafer load lock chuck exhaust ~ independent control of ions acceleration towards wafer 5

Time-multiplexed Bosch processing C 4 F 8 passivation SF 6 etch mask: e.g. SiO 2, photoresist X C 4 F 8 ~ every ~10 s S a F b+, F SF 6 ~ X ~ ~ Journal of The Electrochemical Society, 146 (1) 339-349 (1999); Robert Bosch GmbH, Pat. 4,855,017 and 4,784,720 (USA) and 4241045C1 (Germany) (1994) 6

Non-uniformity at three length scales device/ die spatial variation wafer in cross-section feature-scale inter- and intradevice wafer/chamber-scale aspect ratiodependent etching (ARDE) competition for reactants; diffusion ion and radical flux distribution waferlevel loading F 7

Strategies for improving uniformity Force etching to be reaction-limited Lower chamber pressure Wafer cooling Conservative mask design Improved tool design Relate mask design directly to non-uniformity Design mask according to desired uniformity Perturb etch rates constructively 8

Previously observed chamber-scale variation 1% 5% 1 20% 81 position index 70% 95% pattern density test patterns H.K. Taylor et al., submitted for publication 10

Previously observed pattern-dependent variation Average pattern density 5% throughout Localized to differing extents H.K. Taylor et al., submitted for publication 12

A two-level model, tuned for each tool + recipe characterization wafers characterization wafers A B filter magnitude radial distance + 2 scalar variables two-level model H. Sun et al., Proc. MEMS 05 + work submitted for publication 13

Basis for chamber-scale model? ion flux, ion J i J i Silicon siliconsilicon F Neutral neutral flux, J n J n adsorbed Adsorbed neutrals neutrals lateral transport J i (x, y) C(x, y) generation, recombination R: etch rate Θ: surface coverage ke i : activity constant for ions vs 0 : activity constant for radicals R = kθe J i i R = vs0( 1 Θ) J n 1 = 1 + 1 J R keij i vs0 n Synergism model: Gottscho et al., J Vac Sci Tech B 10 2133 (1992). Right: H.K. Taylor et al., submitted for publication Consumption: J n (x, y) C: fluorine concentration G: fluorine generation rate ρ: wafer-average pattern density C = α τ Gτ [ ρ + α ( 1 ρ) ] 1 1 2 + rate ct. selectivity loading mask silicon 14

Basis for chamber-scale model? X Assuming: A, J i B, G n + 2 scalar variables ion flux independent of etched pattern F generation (G) independent of etched pattern neutral flux (J n ) F concentration F concentration depleted as loading increases F concentration depends on G H.K. Taylor et al., submitted for publication 15

Accounting for spontaneity in Si etching 5 SF 6 etch with no passivation steps usual condition for recipe used SF 6 ~ x S a F b+, F platen power ~ Etch rate cut by ~15% when ion flux minimized But spontaneous etching is readily incorporated into map A: R R ( ke J c) i i + ( ) 0 Θ J n = Θ = vs 1 17

Accounting for time-multiplexing T tot flow rate T etch SF 6 T pass T Si-etch C 4 F 8 time Non-uniformity of CF x flux may matter Ion flux definitely important (T strip ) Ion flux important to an extent (T Si-etch ) 18

tot Time-multiplexing model R Si is the etch rate our previous model would have predicted. Overall, modified rate prediction: Tpassk passj CF R x ( ) Si Tetch T R J Si etchrsi Si Tetch T strip ik strip R' = = = T T T tot If we define a = T etch /T tot, p = J i k strip, q = k pass J CFx, we have: tot R' = [ a( p + q) q] p R Si R = ma + c with m = (p + q)r Si /p and c = qr Si /p. 19

Rate of passivation removal (R ) Here, etch step only just removes passivation ρ = 1% ρ = 40% ρ = 95% ρ = 5% qr Si /p (p + q)r Si /p Basic recipe (a) 20

(Short-term) memory effect in chamber Thermal diffusivity of aluminum ~ 10-4 m 2 /s Over 5, characteristic length ~0.17 m Over 30, characteristic length ~0.42 m 5 5 0.1% 5% 99.9% 5% several cycles 21

(Short-term) memory effect in chamber 0.1% 5 5% 99.9% several cycles 5 5% 0.1% 30 5% 99.9% 30 5% several cycles 22

Pattern component of memory effect? 5% 5 X Reference 5% (B) after Y Reference 5% (A) after X 5% A 5% 5% 5 Y B 5% (even) X 5% (concentrated into eighth) Y 23

Pattern component of memory effect? 5% (even) X Reference 5% (B) after Y Reference 5% (A) after X 5% (concentrated into eighth) Y Etch rate difference A B (μm/min) 0.07 0.065 0.06 0.055 0.05 0.045 0.04 0.035 0.03 5 10 15 20 25 30 Diametrical position /4 mm 24

Putting two-level model into action discretized mask design + scalar constants two-level model takes a few seconds to run drafting software refine mask design highlight problems on-screen 25

CAD tool for nonuniformity prediction Die-scale variation Chamber-scale variation Combined prediction Discretized mask design Ali Farahanchi 26

Acknowledgements Ali Farahanchi, Tyrone Hill, Hongwei Sun and Duane Boning Leslie Lea, Janet Hopkins and Nick Appleyard (Surface Technology Systems Ltd.) Ciprian Iliescu and Franck Chollet Singapore-MIT Alliance Cambridge-MIT Institute 27