Etching terminology h ilm d mask etch mask ilm substrate d mask bias B B d - d m (i.e., twice the undercut) anisotropy A d ilm d ilm A ilm 1 - v l / v v v l lateral etch rate v v vertical etch rate or ilms etched just to completion A = 1 - B / 2h h ilm thickness A = 0 isotropic A = 1 perectly anisotropic Dean P. Neikirk 2001, last update February 12, 2001 1 Dept. o ECE, Univ. o Texas at Austin
Impact undercut and eature aspect ratio d mask h ilm d ilm d ilm d ilm d ilm example: want ilm to have equal lines and spaces ater etch can we compensate or bias (undercut) by adjusting the mask? recall B = d d m «d m = d - B but or ilms etched just to completion B = 2 h ( 1 - A ) so h d mask d 1 2 1 d h / d is the eature aspect ratio since d m > 0 h ( A ) = Dean P. Neikirk 2001, last update February 12, 2001 2 Dept. o ECE, Univ. o Texas at Austin ( 1 A ) 0 1 2 > d A 1 2 1 ( h d ) i aspect ratio is << 1 anisotropy is not needed i aspect ratio is >> 1 need high anisotropy (i.e., A ~ 1)!
Example: undercut and eature aspect ratio d mask h ilm d ilm d ilm d ilm d ilm example: want ilm to have equal lines and spaces ater ISOTROPIC etch or isotropic etch, etched just to completion, undercut = ilm thickness compensate or bias (undercut) by adjusting the mask eatures here d = d m + 2h «d m = d -2h or d mask = h d 1 2 d h / d is the eature aspect ratio h / d = 0.1 Í d mask = 0.8 * d ilm h / d = 0.5 Í d mask = 0!!!!!! i aspect ratio is 0.5 you cannot make equal lines / spaces using an isotropic etch!!!! Dean P. Neikirk 2001, last update February 12, 2001 3 Dept. o ECE, Univ. o Texas at Austin
Selectivity o etches need etch that removes ilm much aster than either the etch mask or substrate required selectivities depend on uniormity o ilm thickness uniormity o etch rates anisotropy o etch rates overetch required acceptable loss o linewidth acceptable substrate loss Dean P. Neikirk 2001, last update February 12, 2001 4 Dept. o ECE, Univ. o Texas at Austin
Selectivity and Over-etch time how long must you etch? etch time = ilm thickness / ilm etch rate BUT what i ilm thickness is not uniorm? etch time = largest ilm thickness / ilm etch rate example: conormal ilm over a step, perectly anisotropic etch h nominal etch time t = h / v h step h step must continue etch or time = h step / v ilm to clear residue total etch time is then T = h /v + h step /v = h /v ( 1 + ) is the ractional over-etch time, here = h step / h ilm Dean P. Neikirk 2001, last update February 12, 2001 5 Dept. o ECE, Univ. o Texas at Austin
Film -to- substrate selectivity does over-etch matter? what about substrate exposed during over-etch period? over-etch etch time t = h step / v h sub = t v sub substrate is lost during overetch h sub = t over v sub = ( h step / v ) v sub = h step (v sub / v ) but v / v sub is the ilm-to-substrate selectivity S s, so S s = h step / h sub recalling = h step / h ilm we get S s = ( h ilm / h sub ) Dean P. Neikirk 2001, last update February 12, 2001 6 Dept. o ECE, Univ. o Texas at Austin
Eects o non-uniormities what i things aren t all uniorm: etch rate: v (1 ± φ ) thickness: h (1 ± δ ) h ( 1 ± δ ) so now etch time is t = 1+ v 1 ± φ ( 1+δ ) ( 1+ ) ( 1 φ ) ( ) Dean P. Neikirk 2001, last update February 12, 2001 7 Dept. o ECE, Univ. o Texas at Austin ( 1 δ ) ( 1+φ ) ( ) ( 1+δ ) ( ) ( + ) 1 φ h = 1 v to ensure complete etch you must use the longest time: t how long is the substrate exposed? shortest time substrate could be coverd by ilm: consider thinnest ilm removed at astest ilm etch rate! h ( 1 δ ) t covered = v 1+ φ so the substrate is exposed to the etch or a time so the amount o substrate lost is t exposed ( ) ( ) h h +δ + δ v 1 1 1 s s = vs t exposed = vs = h v v ( ) { 1 φ 1 +φ = h S s 1 ( ) ( 1+δ ) ( ) ( ) h ( 1 δ ) 1+ 1 φ v ( 1+φ ) v 144244 3 total etch time 14243 time sub. covered ( ) ( )
Example results or ilm - substrate selectivity ( 2 + + δ ) + δ ( 2 + ) 1 ( φ ) v h φ S s = = 2 vs hs + h h s selectivity 80 60 40 20 = 2 = 1 = 0.1 etch rate uniormity φ = 0.1 ilm thickness uniormity δ = 0.05 note h / h s tends to be BIG you may not be willing to lose much substrate! 0 10 20 30 40 h / h s Dean P. Neikirk 2001, last update February 12, 2001 8 Dept. o ECE, Univ. o Texas at Austin
Film -to- mask selectivity mask ilm perectly anisotropic vertical etch, both mask and ilm v vert, mask v vert, ilm A ilm = 1, but A mask < 1 no linewidth loss A ilm < 1, A mask < 1 v horz, mask v vert, mask vhorz, mask v vert, mask v vert, ilm v vert, ilm O ilm O mask loss o linewidth W = O mask - O ilm v horz, ilm need to consider loss o linewidth due to mask erosion unction o mask edge proile and anisotropy o etch (both mask and ilm!) Dean P. Neikirk 2001, last update February 12, 2001 9 Dept. o ECE, Univ. o Texas at Austin
Film -to- mask selectivity need to consider loss o linewidth due to mask erosion unction o: anisotropy o etch (both mask and ilm!) also a unction o the mask edge proile v horz, mask v vert, mask v vert, ilm W/2 should also include impact o various non-uniormites ilm thickness: h (1 ± δ ) etch rates: mask: v m (1 ± φ m ) ilm: v (1 ± φ ) v horz, ilm Dean P. Neikirk 2001, last update February 12, 2001 10 Dept. o ECE, Univ. o Texas at Austin
Film -to- mask selectivity v horz, mask v vert, mask v vert, ilm assume non- uniormities: perectly etch rate: v (1 ± φ ) anisotropic etch o ilm thickness: h (1 ± δ ) to ensure complete etch you must use the longest time: t etch thickest 64748 h ( 1 + δ ) = 1 v ( ) 123 1 φ 14243 ractional slowest etch ilm rate ( + ) over etch during this time two terms contribute to lateral mask erosion: vertical etch o mask combined with slope W/2 θ δx vert δz δ x vert δ z = vvert, = cos sin θ θ mask δz δt plus simple horizontal etch total mask erosion is just sum o terms: cos θ δ xtot = vhorz, mask δt + vvert, sin θ mask δt Dean P. Neikirk 2001, last update February 12, 2001 11 Dept. o ECE, Univ. o Texas at Austin
Film -to- mask selectivity so total mask edge movement is or W = 2 v vert, mask ( 1 + φ ) vhorz, mask vvert, mask 14243 1 A mask let s use worst case mask etch so W W h + cot θ v ( v horz, mask + v vert mask θ) t etch = cot 2, ( 1 + δ ) ( ) ( 1 + ) 1 φ ( 1 + δ ) ( 1 φ ) v h mask m = 2 ( [ 1 Amask ] + cot θ) 1 v 144 243 4 1 S m ( + ) or S m = 2 h ( W 1+φ m ) ( 1+δ ) 1+ 1 φ ( ) ( ) ( [ 1 A mask ]+ cotθ ) NOTE THIS ASSUMED PERFECTLY ANISOTROPIC ETCH OF FILM!!! Dean P. Neikirk 2001, last update February 12, 2001 12 Dept. o ECE, Univ. o Texas at Austin
Sample results or ilm - mask selectivity etch rate uniormity φ & φ = 0.1 m ilm thickness uniormity δ = 0.05 selectivity 80 60 40 20 = 1, A m = 0, θ = 60Û = 1, A m = 0, θ = 90Û = 0.2, A m = 0, θ = 60Û = 1, A m = 1, θ = 60Û = 0.2, A m = 1, θ = 60Û 0 = 1, A 2 4 6 8 m = 1, θ = 90Û h / W Dean P. Neikirk 2001, last update February 12, 2001 13 Dept. o ECE, Univ. o Texas at Austin
Lit-o patterning no-etch thin ilm patterning process do the lithography BEFORE ilm deposition extensively developed by IBM in the late 70 s or patterning o metals Au and Pb or Josephson junction computers! [1] C. Li and J. Richards, A High Resolution Double Layer Photoresist Structure or Lit-O Technology, presented at IEDM, Washington, DC, 1980. [2] M. Hatzakis, B. J. Canavello, and J. M. Shaw, Single-Step Optical Lit-O Process, IBM Journal o Research and Development, vol. 24, pp. 452-460, 1980. [3] R. M. Halverson, M. W. MacIntyre, and W. T. Motsi, The Mechanism o Single-Step Lito with Chlorobenzene in a Diazo-Type Resist, IBM Journal o Research and Development, vol. 26, pp. 590-595, 1982. [4] G. G. Collins and C. W. Halsted, Process Control o the Chlorobenzene Single-Step Lito Process with a Diazo-Type Resist, IBM Journal o Research and Development, vol. 26, pp. 596-604, 1982. [5] P. L. Pai and W. G. Oldham, A Lito Process Using Edge Detection (LOPED), IEEE Transactions on Semiconductor Manuacturing, vol. 1, pp. 3-9, 1988. [6] Y. Homma, A. Yajima, and S. Harada, Feature Size Limit o Lito Metallization Technology, IEEE Transactions on Electron Devices, vol. ED-29, pp. 512-517, 1982. [7] S. Lyman, J. Jackel, and P. Liu, Lit-o o thick metal layers using multilayer resist, Journal o Vacuum Science Technology B, vol. 19, pp. 1325-1328, 1981. [8] K. Grebe, I. Ames, and A. Ginzberg, Masking o Deposited Thin Films by Means o an Aluminum-Photoresist Composite, Journal o Vacuum Science Technology B, vol. 11, pp. 458, 1974. Dean P. Neikirk 2001, last update February 12, 2001 14 Dept. o ECE, Univ. o Texas at Austin
Basic lit-o concept use resist with a re-entrant cross-sectional proile use deposition process with poor step coverage now use solvent to dissolve resist, liting o the material to be removed problem: resist edge proile requently is not re-entrant Dean P. Neikirk 2001, last update February 12, 2001 15 Dept. o ECE, Univ. o Texas at Austin
Chlorobenzene process objective: produce a lip on the surace o the photoresist chlorobenzene soak process soaking photoresist in solvent alters exposure or/and development characteristics process: apply resist, expose soak in chlorobenzene develop exposure Dean P. Neikirk 2001, last update February 12, 2001 16 Dept. o ECE, Univ. o Texas at Austin
Bilayer lit-o use two separate layers bottom layer: sacriicial undercut top layer to orm lip 1st layer resist buer layer substrate (a) buer layer 2nd layer resist (b) two layers o resist (c) bottom layer o resist photoresist bridge (d) undercut top layer o resist (e) Dean P. Neikirk 2001, last update February 12, 2001 17 Dept. o ECE, Univ. o Texas at Austin
Wet chemical etching dominant etch process through late 1970 s tends to be isotropic A ~ 0 exception: anisotropic crystal etches or silicon etch rates along various crystal directions can vary widely (111) tends to be slowest tends to produce high selectivities Dean P. Neikirk 2001, last update February 12, 2001 18 Dept. o ECE, Univ. o Texas at Austin
Oxide wet etch basic oxide etch: hydroluoric acid (HF) SiO 2 + 6HF Í H 2 SiF 6 + 2H 2 O BHF: buering agent (NH 4 F) requently added to keep [HF 2 - ] constant etch rate thermally grown oxide: ~ ew tenth s micron per minute PECVD deposited oxide: ew times aster than thermal doped oxides (BSG, PSG): ~2-5X aster than thermal selectivities very high to photoresist undercutting or resist adhesion ailure biggest problem or long etches very high to silicon Dean P. Neikirk 2001, last update February 12, 2001 19 Dept. o ECE, Univ. o Texas at Austin
Silicon nitride wet etch HF, BHF etch typically much slower than oxide etch rates slowest or high quality, oxygen ree nitride rate in concentrated HF ranges rom 0.01 0.1 micron / minute about 10x slower in BHF obviously no selectivity to oxide boiling H 3 PO 4 (phosphoric acid) nitride etch rate ~ 0.01 micron / minute oxide etch rate ~ 0.001 micron / minute silicon etch rate ~ 0.0003 micron / minute Dean P. Neikirk 2001, last update February 12, 2001 20 Dept. o ECE, Univ. o Texas at Austin
Isotropic silicon wet etch: HNA mixture o HF (hydroluoric acid), HNO 3 (nitric acid), and CH 3 C00H (acetic acid) basic mechanism injection o holes into Si to orm Si 2+ ( anodic reaction) attachment o OH - to the Si 2+ to orm Si(OH) 2 2+ (oxidation step) reaction o the hydrated silica to orm soluble product luoride ions rom HF orm H 2 SiF 6 dissolution o products into solution overall reaction or HNA: 18HF + 4HNO 3 + 3Si Í 2H 2 SiF 6 + 4NO (gas) +8H 2 O recall mechanism required holes: HNO 2 + HNO 3 + H 2 O Í 2HNO 2 + 2OH - + 2h + note this is an electrochemical reaction is inluenced by dopant type and concentration, as well as any external applied electrical bias Dean P. Neikirk 2001, last update February 12, 2001 21 Dept. o ECE, Univ. o Texas at Austin
HNA ormulations HF : HNO 3 : water/ch 3 COOH 1 : 3 : 8 etch rate: 0.7 3 microns/minute doping dependence: light doping < 10 17 REDUCES rate ~ 150X SiO 2 etch rate: 0.03 microns/minute 1 : 2 : 1 etch rate: 4 microns/minute doping dependence: none SiO 2 etch rate:? 1 : 8.3 : 3.3 etch rate: 7 microns/minute doping dependence: none SiO 2 etch rate: 0.07 microns/minute Dean P. Neikirk 2001, last update February 12, 2001 22 Dept. o ECE, Univ. o Texas at Austin
Crystallographic etching recall crystal lattice is ace centered cubic (FCC), with two atom basis [at (0,0,0) and (1/4, 1/4, 1/4) ] two interpenetrating FCC lattices a (100) (110) (111) Dean P. Neikirk 2001, last update February 12, 2001 23 Dept. o ECE, Univ. o Texas at Austin
(111) planes (111) planes etch the slowest, tend to be cleavage planes 54.74 (111) wrt (100) (110) edge (111) atomic plane (111) atomic plane (100) plane edge o pit lines up with (110) (110) edge Dean P. Neikirk 2001, last update February 12, 2001 24 Dept. o ECE, Univ. o Texas at Austin
Masking assume bulk crystalline (100) silicon substrate combined with anisotropic etch results in pyramidal shape bounding (111) planes can be reached using a variety o mask shapes square mask opening, (100) waer orientation, side o square is aligned to the (110) lat what happens i you use a circular mask opening? undercutting o the mask occurs until the (111) planes are reached still orms a pyramidal pit! Dean P. Neikirk 2001, last update February 12, 2001 25 Dept. o ECE, Univ. o Texas at Austin
Other mask openings in general mask is undercut until (111) planes are reached bars undercut until bounding planes are reached cross -shaped mask opening will also undercut to orm pyramidal pit Dean P. Neikirk 2001, last update February 12, 2001 26 Dept. o ECE, Univ. o Texas at Austin
Other shapes? have to have a texana shape Texas star Longhorn obtuse corner cross dierent bending stress / properties Dean P. Neikirk 2001, last update February 12, 2001 27 Dept. o ECE, Univ. o Texas at Austin
Other orientations note (110) planes are perpendicular to one another (110) anisotropic etch rates astest to slowest (110) : (100) : (111) can etch rectangular trenches in (110) orientated substrates sides o trenches are (110) Dean P. Neikirk 2001, last update February 12, 2001 28 Dept. o ECE, Univ. o Texas at Austin
Anisotropic wet etch ormulations alkali metal hydroxide etchants examples: KOH, NaOH typically 15-40w% concentration, dilutant water or isopropanol, ~70 C proposed reactions: Si + 2OH - Í Si(OH) 2 2+ + 4e - 4H 2 O + 4e - Í 4OH - + 2H 2 (gas) Si(OH) 2+ 2 + 4OH - Í SiO 2 (OH) 2-2 + 2H 2 O overall: Si + 2OH - + 2H 2 O Í SiO 2 (OH) 2-2 + 2H 2 (gas) selectivities: (100) : (111) about 400 : 1, (111) rate ~ 1 µm/min p type doping > ~2x10 19 : reduces etch rate (oxide) : (111) about 1 : 1000 choice o concentration, etch temperature aects etch rate, surace smoothness generally, KOH tends to give very smooth (111) planes Dean P. Neikirk 2001, last update February 12, 2001 29 Dept. o ECE, Univ. o Texas at Austin
Other ormulations similar: ammonium hydroxide NH 4 OH 3.7wt% @ 75 C, (100) : (111) o 8,000 : 1, but < 0.1 µm/min etch rate hillock ormation also a problem TMAH (tetramethyl ammonium hydroxide (CH 3 ) 4 NOH ) 90 C, 10-40%, ~1 µm/min; surace roughness can be problem (100) : (111) selectivity 10-35 : 1 boron doping stop selectivity against oxide >1000 low aluminum etch rate EDP (ethylene diamine / pyrochatechol / water) 115 C, ~1 µm/min, (100) : (111) selectivity 35 : 1 oxide selectivity >1000:1, etches aluminum hydrazine Dean P. Neikirk 2001, last update February 12, 2001 30 Dept. o ECE, Univ. o Texas at Austin
Various issues saety hazards EDP, hydrazine potentially quite hazardous ammonia released rom TMAH at elevated temperatures hydrogen bubbles all the reactions tend to produce H 2 bubble ormation can locally mask etch leading to rough suraces bubbles trapped inside sacriicial regions can stop etch or cause breakage Dean P. Neikirk 2001, last update February 12, 2001 31 Dept. o ECE, Univ. o Texas at Austin
Plasma assistant pattern transer includes ion milling, sputtering plasma etching reactive ion etching (RIE) all use plasmas typical pressures 10-3 - 10 Torr mean ree paths 10 mm - 5 µm number density 10 13-10 17 cm -3 typical ion densities ~10 9-10 12 cm -3 most gas molecules are NOT ionized temperature electron: ~10 4 K gas: ~ambient Dean P. Neikirk 2001, last update February 12, 2001 32 Dept. o ECE, Univ. o Texas at Austin
RF discharges and potentials electron and ion mobilities are very dierent leads to separation o charge in r discharge electrons can ollow ield reversals, ions cannot plasma can act like a diode dc potentials can be developed even or pure ac drive ion sheaths driven electrode loating surace plasma dc voltage 0 V p V powered ground electrode grounded electrode electrode typical parameters ac requency: 13.56 MHz, other industrially assigned dc voltages depend on ratio o powered -to- grounded electrode areas tens (~equal areas) to hundreds o volts possible (small powered area wrt grounded) Dean P. Neikirk 2001, last update February 12, 2001 33 Dept. o ECE, Univ. o Texas at Austin V t
Ion bombardment in plasma discharge dc bias voltage / ield between plasma and electrode accelerates ions towards surace positive ions strike surace anisotropically V p + V t +V p + - + + + - + dark space + - + + - + + - n - electric ield recall most gas molecules are neutral still strike surace isotropically - V t sample on powered electrode Dean P. Neikirk 2001, last update February 12, 2001 34 Dept. o ECE, Univ. o Texas at Austin
Sputtering and ion milling i ion energy is ~500 ev substantial sputtering o target occurs inert gas (Ar) typically used sputtering systems accelerating potential rom sel bias ion mill powered electrode area << ground area sample placed on powered electrode separate ion generation, acceleration, and sample chamber process is purely physical everything sputters very low selectivity Dean P. Neikirk 2001, last update February 12, 2001 35 Dept. o ECE, Univ. o Texas at Austin
Plasma etching use reactive species produced in a plasma discharge to drive chemistry need to produce volatile reaction products usually try to avoid ion bombardment keep accelerating voltages small process is mainly chemical high selectivity low anisotropy driven electrode waers plasma waers planar system ground electrode plasma barrel system Dean P. Neikirk 2001, last update February 12, 2001 36 Dept. o ECE, Univ. o Texas at Austin
Reactive Ion Etching (RIE) similar to sputter etcher but replace noble gas with reactive gases like to those used in plasma etching want high energy ions at surace high accelerating voltages substrates on powered electrode asymmetric electrodes area o powered electrode < grounded electrode low pressures 10-3 -10-1 Torr in both plasma etching and RIE eed- gas composition produces the reactive species necessary or etching chemistry tends to be isotropic ion bombardment o suraces generates the anisotropy in plasma assisted pattern transer Dean P. Neikirk 2001, last update February 12, 2001 37 Dept. o ECE, Univ. o Texas at Austin
Ion- induced and ion- enhanced gas phase etching ion-induced reactions expose Si to Cl 2 : no etching o the Si occurs expose Si to Ar + ion mill beam: etch rate < 0.5 nm/min expose to both: etch rate ~ 10 nm/min note NO Cl 2 plasma was present in this example ion-enhanced reactions XeF 2 will spontaneously etch Si: ~0.5 nm/min but i expose to both XeF 2 and Ar + ion beam etch rate increases dramatically (~6 nm/min) Dean P. Neikirk 2001, last update February 12, 2001 38 Dept. o ECE, Univ. o Texas at Austin
CF 4 plasma etching electron impact in plasma produces reactive radicals CF 4 + e D CF + 3 + F * + 2e competing reactions between ree F and CF 3 tends to keep F concentration low ree luorine etches both Si and SiO 2, but etches Si aster Si + 4F " SiF 4 (g) SiO 2 + 4F " SiF 4 (g) + O 2 these are isotropic! add O 2 to gas mix CF 3 + + O + e " COF 2 + F * competes with F or CF 3, drives [F] up etch rates increase peak selectivity ~15 (Si:SiO 2 ) Si etch rate (Å/min) 5000 4000 3000 2000 1000 rom: Sze, 1st ed., p. 325. 350 280 210 140 70 SiO 2 etch rate (Å/min) 10 20 30 40 50 percent O 2 in CF 4 Dean P. Neikirk 2001, last update February 12, 2001 39 Dept. o ECE, Univ. o Texas at Austin
Ion assisted CF x etching add H 2 to CF 4 gas mix CF 4 + e D CF + 3 + F * + 2e H 2 + F* " HF so addition o H 2 drives [F] down reduces Si etch rate BUT CF 3 will NOT etch SiO 2 UNLESS there is ion bombardment CF x + SiO 2 + damage " SiF 4 (g) + (CO, CO 2, COF 2, etc.) CF x + Si " SiF 4 (g) + C-F (polymer) " stops etch i no ion bombardment! summary: CF 4 + H 2 (40%) : SiO 2 / Si selectivity ~ 10:1 can improve oxide-to-silicon selectivity by decreasing F:C ratio: use CHF 3 CF 4 + O 2 (10%) : Si / SiO 2 selectivity ~ 15:1 etch rate (Å/min) 600 500 400 300 200 100 rom: Sze, 1st ed., p. 326. SiO 2 with ion bomb. pmma pr Si 10 20 30 40 percent H 2 in CF 4 Dean P. Neikirk 2001, last update February 12, 2001 40 Dept. o ECE, Univ. o Texas at Austin
Mechanisms or ion-enhancement and induced anisotropy ion bombardment ion bombardment polymer damage surace damage induced anisotropy surace inhibitor mechanism anisotropy example o polymer ormation: Si etched in Cl 2 plasma (~isotropic) e + Cl 2 e + 2Cl Si + xcl SiCl x recombinant species C 2 F 6 e + C 2 F 6 2CF 3 + e [CF 3 + Cl] x [CF 3 Cl] 3 (polymer) at ~85% C 2 F 6 no undercutting occurs Dean P. Neikirk 2001, last update February 12, 2001 41 Dept. o ECE, Univ. o Texas at Austin
Aluminum plasma etch use volatile aluminum chloride Al + CCl 3+ + (bmbrdmnt) " AlCl 3 (g) + C Al + 3Cl * " AlCl 3 (g) problems initial Al 2 O 3 on surace harder to etch (mostly by CCl + 3 ) selectivity wrt SiO 2 < ~20 selectivity wrt photoresist < ~15 can also use BCl 3, may include some O 2 Dean P. Neikirk 2001, last update February 12, 2001 42 Dept. o ECE, Univ. o Texas at Austin
Etch summary material etched etch gas(es) volatile product selectivities Si CF 4 + O 2, etc. SiF 4 15:1 (Si:SiO 2 ) SiO 2 CF 4 + H 2, etc. SiF 4 20:1 (SiO 2 :Si) organics O 2 CO 2, H 2 O high Al CCl 4, BCl 3, etc. AlCl 3 15:1 (Al:SiO 2 ); ew:1 (Al:Si) Mo CF 4 MoF 6 W CF 4 WF 6 things that are hard to dry etch copper: no volatile reaction products use CMP instead Dean P. Neikirk 2001, last update February 12, 2001 43 Dept. o ECE, Univ. o Texas at Austin
Si Deep RIE desire: high etch rates, high aspect ratios requires very high degrees o anisotropy Bosch process (German patent: Larmer and Schilp, 1994) uses recombinant species and side-wall polymer ormation sequential etch / polymer deposition high bias reactive ion etch: SF 6 / Ar typical low bias polymerization: C 4 F 8, CHF 3 repeat usually need high density plasma source inductively coupled aspects ratios up to about 30:1 etch rate: ew microns per minute selectivities to PR : 50-100 to oxide 100-200 Dean P. Neikirk 2001, last update February 12, 2001 44 Dept. o ECE, Univ. o Texas at Austin
Deep RIE pictures at http://www.ee.washington.edu/class/539/lectures/lecture3/s ld011.htm STS (http://www.stsystems.com/) BPS/Plasma Therm (model 790, ICP type) Dean P. Neikirk 2001, last update February 12, 2001 45 Dept. o ECE, Univ. o Texas at Austin
in multi- level interconnect planarization is limiting problem CMP Chemical-mechanical polish (CMP) lithography/etch to pattern metals deposit dielectric polish to planarize dielectric surace deposit dielectric etch contact openings (vias), deposit via plug material, etch/polish as needed repeat starting with deposition/patterning o next metal lines Dean P. Neikirk 2001, last update February 12, 2001 46 Dept. o ECE, Univ. o Texas at Austin
Damascene process deposit dielectric etch trenches that are where you want metal lines line trenches i necessary blanket deposit metal polish metal back to dielectric surace deposit dielectric open vias deposit via plug material repeat by depositing dielectric, etching trenches to be illed with metal Dean P. Neikirk 2001, last update February 12, 2001 47 Dept. o ECE, Univ. o Texas at Austin
Damascene/CMP example re: R. W. Mann, L. A. Clevenger, P. D. Agnello, and F. R. White, Salicides and local interconnections or highperormance VLSI applications, IBM J. Res. Develop., vol. 39, pp. 403-417, 1995. Photo p. 414. dielectric layers have been etched away to reveal metal lines SRAM cell green: word lines, salicided poly yellow: 1st global, Ti/Al(Cu)/Ti/TiN pink: local (intra-cell), tungsten grey: contact studs, tungsten Dean P. Neikirk 2001, last update February 12, 2001 48 Dept. o ECE, Univ. o Texas at Austin
Copper Interconnects: CMP example multi-level copper interconnects reduce electrical resistance, improve electromigration resistance problems C. Wu, Computer Chips Take a Leap Forward, in Science News, vol. 152, 1997, pp. 196. dry etch o copper is challenging L. Geppert, Technology 1998, Analysis and Forcast: Solid State, in IEEE Spectrum, vol. 35, 1998, pp. 23-28. corrosion cover sidewalls with TiN, other reractory, beore copper deposition Dean P. Neikirk 2001, last update February 12, 2001 49 Dept. o ECE, Univ. o Texas at Austin
Problems in CMP dishing when making large area eatures the removal rate may not be uniorm across the eature copper: tends to get thinner near middle o large eatures dielectrics: similar behavior constrains ability to make large (i.e., width, length >> thinckness) structures Dean P. Neikirk 2001, last update February 12, 2001 50 Dept. o ECE, Univ. o Texas at Austin