`
`
`
`65.207.2.2Downloaded on 2017-06-19 to IP
`
` address. Redistribution subject to ECS terms of use (see
`
`ecsdl.org/site/terms use
`
`) unless CC License in place (see abstract).
`
`Page 1 of 7
`
`TSMC Exhibit 1047
`TSMC v. IP Bridge
`IPR2016-01377
`
`
`
`Vol. 130, No. 2
`
`THREE-LAYER RESIST SYSTEM
`
`479
`
`layer perpendicular to the underlying grating struc-
`tures. Exposures were performed either with a direct
`write Vector Scan electron beam system (39) or with
`a 5x step-and-repeat optical projection printer (40).
`The exposed and developed image pattern in the AZ
`resist layer was replicated in the sandwiched Si film
`by RIE in a CF4 plasma (41). Transfer of the Si pattern
`to the underlying PMMA layer was performed either
`by flood exposure to deep u.v. radiation (ZOO-.260 nm)
`followed by development of the image pattern in or-
`ganic solvents, or by in situ RIE in an oxygen plasma
`(11-14, 19-29). A typical process sequence is outlined
`below and is illustrated schematically in Fig. 1. Typi-
`cal patterns of PMMA resist lines crossing over SiOz
`steps are shown in Fig. 2.
`Process sequence.—Spin-coat HMDS. Spin-coat 1 pm
`PMMA.
`(Dupont Elvacite 2041, 10% by weight
`in
`2-methoxyethyl other). Bake 15 min at 85°C. Repeat,
`spin-coat 1 pm PMMA. Bake 1 hr at 160°C. Deposit
`amorphous Si 30-200 nm by plasma CVD of SiH4 or by
`electron gun evaporation of Si. Bake 1 hr at 100°C in
`air. Spin-coat HMDS. Spin-coat 0.45 am A21350J using
`2 AZ1350J21 AZ thinner. Bake 15 min at 85°C. Expose
`test pattern either by (i) electron beam at 30 HC/sz,
`20 keV, or (it) on a 5x optical projection printer at
`15 mJ/cm“ using 405 nm monochromatic radiation. De-
`velop image in AZ developer,
`then bake 30 min at
`100°C. RIE Si in CF4. RIE patterning:
`(i) In situ RIE
`PMMA in 02 plasma, and (ii) dip etch 5-10 sec in buf-
`fered HF to clean off residues. Deep u.v. patterning: (i)
`flood expose to deep u.v. at 0.5-1 J/cm’, and (ti) de-
`velop image in either chlorobenzene or methyl isobutyl
`ketone (MIBK).
`
`Resist films—Although resist films other than PMMA
`were used for planarization, PMMA was selected for
`
`
`
`Fig. 2. Typical high resolution line/space test patterns in the
`three-layer system AZ/Si/PMMA defined over i am high grating
`structures etched in an Sl°2 film. Top: Optical micrograph of part
`of the test pattern showing line/space features ranging between
`0.5-5 urn. Bottom: SEM of equal line/space features 3 am wide in
`3 pm thick PMMA generated by RIE in oxygen.
`
`EXPOSE
`I
`_
`a beam
`- PROJECTION
`
`Dix/£4)?
`
`EYCH
`:75.
`
`,
`
`
`
`Lyflf
`
`3
`
`[IMAGING
`ncss‘r
`’9
`,
`r
`PLANARIz Nc
`fissusr
`suaswni
`
`IQESIST
`a- S
`_
`::.--
`
`>
`[—w—-
`V: eaESIsr
`‘
`A
`suestnrre
`
`,—
`.—_
`.rRtSISl
`Va,
`mieem,k,—:tiy a_s
`_.i::'i_,, ,,
`i
`RtSIsI
`,
`,3
`SLESYFATE
`2,,
`7 H
`EXPCSE r/
`
`\,
`
`detailed study because of its deep u.v. sensitivity (42)
`and its high etching rate in an 02 plasma. PMMA films
`>1 um thick were obtained by applying multiple coat-
`ings in order to avoid cracking, and to promote better
`planarization of the substrate surface as discussed be-
`low. Adhesion of PMMA to Si and SiOg surfaces was
`enhanced by prior coating with the adhesion promoter
`hexamethyl disilazane (HMDS). However, the use of
`HMDS was essential to ensure adequate adhesion of
`AZ1350J photoresist to the Si surface.
`Despite its low sensitivity, AZ1350J resist was se-
`lected as an imaging layer because it offered the con-
`venience of both optical and electron beam sensitivity.
`The initial thickness of AZl350J resist was reduced by
`~0.1 um after development to a final nominal thick-
`ness of 0.4 am. The latter was quite adequate to mask
`the underlying Si layer in a CF4 plasma during the pat-
`tern transfer process.
`Silicon films—Plasma CVD silicon films were depos-
`ited in the plasma reactor illustrated in Fig. 3 using a
`mixture of Sin and He at 4 and 65 mTorr partial pres-
`sures, respectively. The wafers were placed on the 125
`mm diameter lower electrode, and after pumping to
`a pressure of 5 x 10-7 Torr, the deposition was per-
`formed at 50°C substrate temperature, at a rate of 15
`nm/min using a power density of 0.6 W/cm”. Films de-
`posited under
`these conditions were hydrogenated
`amorphous silicon containing as much as 30% H2 (43-
`46). Films obtained by electron gun evaporation of
`pure silicon on substrates maintained at or slightly
`above room temperature were functionally indistin-
`guishable from CVD Si films.
`Thin films obtained by plasma CVD of silane and
`electron gun evaporation of silicon were particularly
`Fig. I. Deep in. and RIE processes for patterning over topography
`weu suited for lithographic processing because the
`using the three-layer resist system AZl350J/Si/PMMA. The elec.
`films were clear, smooth, and amorphous. The optimum
`tron beam or optically exposed pattern is generated in the thin top
`thickness for most patterning applications was 90 1- 20
`AZ layer, and then replicated in the sandwiched amorphous silicon
`nm although patterns were defined in films 30-200 nm
`film by II! in Ch. The Si pattern is transferred to the thick bot-
`thick. Silicon films were transparent
`in the visible
`tom PMMA planarizing layer by deep I.v. exposure and solvent
`development or by II! in oxygen.
`spectrum as shown in Fig. 4, but more importantly.
`Dmbaded on 2017-06-19 to IP 65207.22 addrm. Redstribution subject I) ECS terms of me (see ecsdl.orysitelterms use) urless CC License in phoe (see distract).
`
`x-aooeaoorn”‘
`
`.
`.
`;
`L
`,;
`
`\
`
`PESISY
`rs.
`vuun
`SLBSTFATE
`
`DEVELOP.
`
`I
`‘....._
`
`\\
`‘\
`. . . . ._
`l
`~L-uu-i
`.
`,
`
`‘
`an:
`
`er
`
`..
`
`moi
`HANARIZING
`PESis‘
`»_J
`j sunSYPATF
`
`Page 2 of 7
`
`
`
`480
`
`J. Electrochem. Soc.: SOLID—STATE SCIENCE AND TECHNOLOGY
`
`February 1983
`
`PLASMA CVD DEPOSITION SYSTEM
`
`R F
`
`GENERATOR
`
`MAl'CHING
`
`
`
`COOLING
`
`
`
`
`PLASM A
`
`SUBSTRATES
`
`___TO meme
`STACK
`
`SPUTTER
`
`THERMO
`COUPLE
`BIAS
`VOLTAGE
`
`
`
`GENERATOR
`
`
`
`
`Fig. 3. Sketch of plasma CVD system for the deposition of‘hydro-
`genated amorphous silicon films from a silone/helium mixture. The
`14 cm diameter electrodes 6 cm apart operated at 13.56 MHz.
`Typical conditions: Sit-l4 4 mTorr, He 65 InTorr, upper electrode
`850V,
`lower (substrate) electrode 0V, power 0.6 W/cmz, deposition
`rate is nm/min.
`
`they were totally opaque in the deep u.v. region below
`300 nm. The films were excellent blocking layers for
`deep u.v. pattern transfer applications as suggested by
`the PCM techniques discussed by Lin (1, 4, 16-18).
`Pattern replication—Transfer of the image pattern
`from the top AZ layer to the Si film was performed by
`RIE in a diode reactor with a perforated grounded alu-
`minum catcher plate positioned between the electrodes
`to minimize backseattering of sputtered aluminum
`
`AMORPHOUS Si (0mm)
`
`0.62pm
`
`'00
`90
`
`80
`
`NI0
`
`(47). The system was first pumped dam to 10" Torr,
`and Si was then etched in CF4 at 2.5 m'I'orr pmure,
`0.25 W/cm2 rf power density at 13.56 MHz. Using end-
`point laser detection (48) and 50% overetching beyond
`and point ensured complete and uniform removal of
`the Si layer.
`Etching of the underlying PMMA or polymer layer
`was performed in an oxygen plasma at 20 mTorr pres-
`sure and 0.06 W/cm2 rf power density. Overetching
`Varied between 50 and. 100% depending on the topog-
`raphy present and the maximum thickness of resist.
`The etch rates of the 3 layers viz., Si, A21350J, and
`PMMA in CF; were 40, 50, and 70 nm/min, respec-
`tively, while in 02 the etch rates of the same films
`were 0, 35, and 70 nm/min, respectively. Replication
`of the AZ image pattern in an Si film 100 nm thick re-
`sulted in an AZ film loss of 200 nm, while replication
`into a PMMA layer £1 urn resulted in the total loss or
`the AZ film.
`'
`Transfer of the Si image pattern to the PMMA layer
`was also performed by flood exposure to deep u.v. ra-
`diation, followed by development in either chloroben-
`zene or MIBK at room temperature. A 1 kW Hg-Xe
`deep u.v. source manufactured by Hanovia was used
`to expose individual wafers in air. The exposure time
`varied between 2-10 min depending on the sensitivity
`and thickness of the planarizing resist as well as on
`the composition of the developing solvent. To obtain_
`the same dissolution rate of PMMA, the deep u.v. ex-
`posure necessary for MIBK development was approxi-
`mately twice that required for chlorobenzene.
`
`Results and Discussion
`
`High-resolution line/space features over 1 pm topog-
`raphy were defined in the three-layer system AZ/Si/
`PMMA using RIE and deep u.v. patterning techniques.
`The minimum dimension defined in PMMA layers 1-3
`pm thick was 0.5 pm using electron beam exposure of
`the top AZ layer. Optical exposure on a 5X step-and-
`repeat projection system using 405 nm monochromatic
`radiation produced minimum line/space features of 1
`am. Figure 5 shows 1 pm and 0.75 pm line/space fea-
`tures defined in the Si layer, and replicated in 2 nm
`thick PMMA by deep u.v. exposure, and development
`in chlorobenzene. Similar test patterns 1
`rim/0.5 pm
`
`
`
`m9
`
`'36TRANSMISSION4)0|0O
`
`T_
`
`0‘O
`
`N0
`
`l0
`
`0
`020
`
`025
`
`Page 3 of 7
`
`j_L_LJ__l_.I_L_g
`05
`L0
`030
`WAVELENGTH (pm)
`
`Fig. 5. SEM micrographs of high resolution line/space test pat-
`terns defined in the AZ/Si/PMMA system over i am high topog-
`Fig. 4. Transmission spectrum of a 1” nm thick film at hydro-
`raphy. A: Equal
`'l
`,urn line/space features in AZI3501 and Si ob-
`genated amorphous silicon deposited by plasma CVD of a silane/
`tained by electron beam exposure followed by RIE etching of Si. B:
`helium mixture on a quartz (suprasil) substrate. The film is trans-
`Same pattern alter deep u.v. exposure and development of PMMA
`parent in the visible region with a peak at 620 nm and is totally
`2 pro thick. C: Same as 3 showing vertical sidewalls of PMMA
`opaque below 400 nm. Similar spectra are exhibited by films as thin
`crossing over an oxide step i ,um high. D: Equal 0.75 pm line/space
`features similar to C.
`as 20 not including those obtained by electron-gun evaporation.
`Downloaded on 2017-06-19 to IP 65207.22 addrm. Reddribution subject to £08 terms of use (see ecsdl.orysitdterms use) unles CC License in phoe (see distract).
`
`
`
`Vol. 130, No. 2
`
`THREE-LAYER RESIST SYSTEM
`
`481
`
`and 1 uni/1 um were obtained by RE of PMMA in an
`oxygen plasma as shown in Fig. 6.
`The RIE and deep u.v. patterning techniques were
`equally capable of generating uniform resist features
`with good linewidth control, quasi-vertical sidewalls,
`and high aspect ratio over 1 pm topography as illus-
`trated inFig. 2, 5, and 6. Such patterns are highly de-
`sirable for subtractive and additive processes in VLSI
`fabrication The differences between the RIE and deep
`u.v. patterning techniques were mainly determined by
`the characteristics of the planarizing layer as dis-
`cussed below.
`
`RIE' Dafleffliflgv—RIE processing was more toler-
`ant of variations in thicknms and composition of the
`planarizing layer by virtue of the anisotropy of RISE,
`and the high etch rate of organic resist materials rela-
`tive to Si in an oxygen plasma. Radiation sensitivity of
`the planarizing layer was also not a requirement in
`BE processing. The selection of useful organic coat-
`ings was,
`therefore, wider, and the processing flex-
`ibility, for example with respect to high-temperature
`baking and thermal cycling, was greater than deep u.v.
`processing.
`Undesirable resist residues, which appeared on sur-
`faces after etching by RE, were removed by a clean-
`ing procedure such as dip-etching in dilute HF solu-
`tion. Figure 7 illustrates the effectiveness of the HF
`dip-etching technique in cleaning 01! all traces of resi-
`dues from RIB-patterned surfaces. Where exposure to
`HF was incompatible with the materials on the water
`surface, alternate cleaning procedures were used in
`conjunction with modified RLE conditions. Potential
`contamination from metals sputtered from the sub-
`strate holder and radiation damage from energetic pho-
`tons could be drawbacks in some applications, but their
`
`
`
`Fig. 6. SEM micrographs of line/space patterns defined in the
`AZ/Si/PMMA three-layer system over 0.8 um high topography. Top:
`Line/space pattern 1 urn/0.5 1.1m obtained by RIE of PMMA 2.5'um
`thick. Bottom: Similar pattern with line/space features ‘I inn/l pm.
`Decrease in linewidth is less than 0.1 pm.
`
`
`
`I m
`’1'
`
`t——1
`Info
`
`Fig. 7. SEM micrographs of patterns defined in c PMMA film 2
`.um thick by ME in oxygen. Resist residues on etched surfaces were
`cleaned off by clip etching in buffered HF for 10-15 sec. A,C: After
`RIE. B,D: After cleaning.
`
`efiects could be eliminated or minimized by adjusting
`RISE conditions (47).
`
`Deep u.v. patterning.—In deep u.v. patterning, resist
`residues seldom appeared after solvent development
`of adequately exposed PMMA. Residues tended to ap-
`pear when the overall PMMA thickness was excessive,
`or the thickness variation over steps was large. This
`could be attributed to the strong attenuation of the
`deep u.v. intensity in thick layers, resulting from the
`strong absorption of PMMA in the spectral region be-
`tween 200-260 nm (42). As the solubility of PMMA was
`a strong function of the energy absorbed, the attenu-
`ation from top to bottom of thick PMMA layers re-
`sulted in a corresponding decrease in development rate.
`This phenomenon influenced residue-formation, as well
`as the cross-sectional profile and aspect ratio of resist
`features. Where PMMA thickness variations due to to-
`pography were large, development in chlorobenzene
`resulted in better linewidth control
`than undiluted
`MIBK Development in chlorobenzene could also be
`used to retain the top imaging resist film, and generate
`a three-layer structure which is potentially uSeful in
`lift—QR applications (49-51). Similar structures could
`also be generated by RIE using a top imaging resist
`film which is stable in an oxygen plasma.
`An advantage of the deep u.v. patterning process
`was its relative insensitivity to pinholes or cracks in
`the barrier layer when the top imaging film was intact
`and opaque in the deep-u.v. In the absence or the Si
`barrier layer, a film of AZISSOJ photoresist 0.3 um
`thick is sufficient to block deep u.v. radiation from ex-
`posing the PMMA layer
`(4). Despite the infinite
`solubility of exposed PMMA relative to silicon,
`the
`maximum useful thickness of PMMA was limited to
`8 pm due to the severe attenuation of the deep u.v. in-
`tensity,’and the tendency of thicker layers to crack
`after baking or pattern development.
`
`Film integrity.—4uccessful patterning of the AZ/Si/
`PMMA system required careful optimization of the film
`thicknesses and the experimental conditions which af-
`fected the integrity of the 3 layers. In particular, the
`stress, the adhesion, and defect levels in one layer af-
`fected directly the integrity of the two other layers
`during processing. The film characteristics and the pro-
`cessing toleranCes varied with the horizontal and verti-
`cal geometry ot the underlying topography.
`
`DMIIoaded on 2017-06-19 to IP 65207.22 addrm. Minion subject to £06 terms of use (see ecsdl.orysiteltenns use) unless 00 License in place (see wad).
`
`Page 4 of 7
`
`
`
`482
`
`J. Electrochem. 5°C.: SOLID-STATE SCIENCE AND TECHNOLOGY
`
`February 1983
`
`films deposited by
`Hydrogenated amorphous Si
`plasma CVD often formed blisters (Fig. 8) on poorly
`planariaed PMMA surfaces. Such blister formation was
`presumably due to excessive stress induced in theSi
`film by the underlying topography. Silicon films de-
`posited at room temperature by electron gun evapora-
`tion were more susceptible to loss of adhesion than
`plasma CVD Si films. The evaporated films were more
`specular and uniform, resulting in better defined pat-
`terns. The etching and deep u.v. blocking character-
`istics of evaporated and CVD films were identical.
`Poorly adhering and highly stressed silicon films often
`cracked and peeled after the imaging layer was pat-
`terned. Deposition above room temperature and an-
`nealing of the films prior to the application or the top
`imaging layer alleviated this problem. Similar be-
`havior was observed with evaporated films of A1 and
`Ti used as barrier layers between PMMA and A21350J.
`The thickness of the top imaging resist layer and the
`baking conditions also influenced the integrity of the
`silicon barrier layer. Resist films 50.5 m thick baked
`
`
`
`
`i——l
`
`5pm
`
`at about 100°C were readily processed over a wide
`range of PMMA and Si film thicknesses. Good step cov-
`erage and adequate planarization were essential to ob-
`tain a uniformly thick imaging layer. Inadequate pla-
`narization resulted in variations in thickness of the
`imaging layer with conaponding variations in line-
`width and profile geometry. Multiple reflections at the
`resist-silicon interface also contribute to linewidth va-
`riations in the case of optically exposed patterns. The
`efiect is clearly illustrated in Fig. 9 which shows the
`systematic variation in linewidth of 5 parallel resist
`lines nominally 2 m wide crossing over 2 raised oxide
`steps 1 m thick and 2 m wide.
`
`Plana'rization.—Due to the importance of the pla-
`narizing function of PMMA. an empirical evaluation of
`the coating characteristics of a variety of films was
`made using single and multiple coatings on stepped
`surfaces. Oxidized wafers, with sets of periodic line/
`space structures 2.5-25 am wide, were coated by con-
`ventional resist spinning and baking methods. The un-
`dulations of the resist surface profile conforming to the
`substrate topography were scanned with a mechanical
`stylus to determine the planarization or leveling ac-
`complished by a coating layer.
`The curves in Fig. 10 show the peak-to-valley am-
`plitude of the resist surface profile covering surface
`steps 0.8 m high. The degree of planarization, repre-
`sented by the amplitude, varied with the molecular
`weight of PMMA, the thickness and number of PMMA
`layers, the line/space dimensions of the substrate top-
`ography, and their periodicity or pitch. As the average
`molecular weight of PMMA decreased, the amplitude
`of the surface undulations decreased indicating en-
`hanced planarization. This effect is illustrated in Fig.
`10(a) with three PMMA compositions of average mo-
`lecular weight 500,000, 150,000, and 33,000. The films
`were nominally 2 pm thick coated on sets of parallel
`steps 5 m wide and varying pitch.
`The expected improvement in planarization with in-
`creasing PMMA thickness is illustrated in Fig. 10(b)
`for the same sets of steps 5 m wide and 0.8 um high.
`The amplitude of the surface undulations also pro-
`gressively decreased as the number of PMMA layers
`increased. It is interesting to note that better planar-
`
`
`
`Fig. 8. Hydrogenated amorphous silicon films 100 "In thick de-
`posited on PMMA at room temperature. Blisters indicative of corn-
`prestive stress in the film are randomly distributed on the surface.
`Top: Single blister after film deposition. Bottom: Appearance of
`blister after patterning.
`Downloaded on 2017-06-19 to IP
`
`lines that run from
`Fig. 9. Linewidth variation of PMMA resist
`top to bottom of the micrograph traversing two i
`inn high Si°2
`steps. Narrowing at the resist line: occurs at the top of the steps
`where the resist is thinnest.
`address. Redistribution stbjecl to ECS terms of use (see
`
`) unless OC license in place (see abstrad).
`
`Page 5 of 7
`
`
`
` (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12)
`
`
`
`65.207.2.2Downloaded on 2017-06-19 to IP
`
` address. Redistribution subject to ECS terms of use (see
`
`ecsdl.org/site/terms use
`
`) unless CC License in place (see abstract).
`
`Page 6 of 7
`
`
`
`
`
`
`
`65.207.2.2Downloaded on 2017-06-19 to IP
`
` address. Redistribution subject to ECS terms of use (see
`
`ecsdl.org/site/terms use
`
`) unless CC License in place (see abstract).
`
`Page 7 of 7
`
`