M] CROLITHOGRAPHY
`
`Science and Technology
`
`edited by
`
`James R. Sheats
`Hewlett-Packard Laboratories
`
`Palo Alto, California
`
`Bruce W. Smith
`Rochester Institute of Technology
`Rochester, New York
`
`MARCEL
`
`DEKKER
`
`' MARCEL DEKKER, INC.
`
`NEW YORK - BASEL - HONG KONG
`
`IP Bridge Exhibit 2017
`
`TSMC v. IP Bridge
`IPR2016-01376
`
`Page 0001
`
`IP Bridge Exhibit 2017
`TSMC v. IP Bridge
`IPR2016-01376
`Page 0001
`
`

`

`Library of Congress Cataloging-in-Publication Data
`
`Sheats, James R.
`
`Microlithography: science and technology / James R. Sheats, Bruce W. Smith.
`p.
`cm.
`
`Includes bibliographical references and index.
`ISBN 0—8247—9953-4 (alk. paper)
`1. Microlithography. 2. Integrated circuits—Masks. 3. Metal oxide semiconductors,
`Complementary—Design and construction. 4. Manufacturing processes.
`1. Smith,
`Bruce W.
`11. Title.
`
`TK7836.S46 1998
`
`621.3815'31--dc21
`
`98-16713
`CIP
`
`The publisher offers discounts on this book when ordered in bulk quantities. For
`more information, write to Special Sales/Professional Marketing at the address
`below.
`
`This book is printed on acid—free paper.
`
`Copyright © 1998 by MARCEL DEKICER, INC. All Rights Reserved
`
`Neither this book nor any part may be reproduced or transmitted in any form or
`by any means, electronic or mechanical, including photocopying, microfilming,
`and recording, or by any information storage and retrieval system, without
`permission in writing from the publisher.
`
`MARCEL DEKKER, INC.
`
`270 Madison Avenue, New York, New York 10016
`
`http://www. dekker. com
`
`Current printing (last digit):
`10 9 8 7 6 5 4 3 2 1
`
`PRINTED IN THE UNITED STATES OF AMERICA
`
`IPR2016-01376 Page 0002
`
`IPR2016-01376 Page 0002
`
`

`

`9
`
`Resist Processing
`
`Bruce W. Smith
`
`Rochester Institute of Technology
`Rochester, New York
`
`1
`
`INTRODUCTION
`
`F
`
`For the most part, conventional single-layer photoresists have been based on com—
`ponents with two primary functions. Whether considering older bis-arylazide
`cis—polyisoprene resists, diazonapthoquinone (DNQ)/novolac g/i—line resists, or
`chemically amplified polyhydroxystyrene (PHS) deep—UV (DUV) resists, an ap-
`proach has been utilized wherein a base resin material is modified for sensitivity to
`exposure by a photoactive compound or through photoinduced chemical amplifica-
`tion. The resist base resin is photopolymeric in nature and is responsible for etch
`resistance, adhesion, coat—ability, and bulk resolution performance. These resins
`generally do not exhibit photosensitivity on the order required for integrated circuit
`(IC) manufacturing. Single—component polymeric resists have been utilized for mi-
`crolithography, including methacrylates, styrenes, and other polymers or copoly-‘
`mers, but sensitization is generally low and limited to exposures at very short
`ultraviolet (UV) wavelengths or with ionizing radiation. Inherent problems associ—
`ated with low absorbance and poor radiation resistance (required, for example, dur—
`ing ion implantation or plasma etching steps) generally limit the application of these
`types of resists to low volumes or processes with unique requirements.
`Sensitization of photoresist materials has been accomplished by several
`methods. In the case of conventional g/i—line resists, a chemical modification of
`a base—insoluble photoactive compound (PAC), the diazonaphthoquinone, to a
`base-soluble photoproduct, indene carboxylic acid (ICA), allows an increase in
`
`515
`
`IPR2016-01376 Page 0003
`
`IPR2016-01376 Page 0003
`
`

`

`516
`
`Smith
`
`aqueous base solubility. For chemically amplified PHS—based resists, exposure
`of a photoacid generator (PAG) leads to the production of an acid, which sub-
`sequently allows polymer deprotection (positive behavior) or crosslinking (neg-
`ative behavior). Other similar processes have been developed (as discussed in
`Chapter 8) and may involve additional components or mechanisms.
`For any resist system, the thermodynamic properties of polymeric resins play
`an important role in processibility. During the coating, exposure, and development
`processes of a resist, an understanding of the thermodynamic properties is desir—
`able, as the glass transition temperature (Tg) of a polymer influences planariz—
`ability, flow, and diffusion. Although reasonably high 7; values may be desirable,
`glassy materials with values above 200°C are not suitable because of poor me-
`chanical performance. Once three-dimensional resist features are formed, how—
`ever, a thermoset material may be desired in which the polymer does not flow
`with temperature and a Tg essentially does not exist. This ensures the retention of
`high—aspect—ratio features through subsequent high—temperature and high-energy
`processes. By appropriate engineering of bake steps during single-layer resist pro-
`cessing, the control of polymer thermoplastic and thermoset properties can be
`made possible. For negative resists, the situation is inherently simplified. Coated
`negative resists are thermoplastic in nature, with a well-defmed Tg range. Upon
`exposure and subsequent secondary reactions, crosslinking leads to a networked
`polymer that will not flow with temperature. At some high temperature of de-
`composition (Td) the polymer will break down and begin to lose significant vol—
`ume. Imaging steps are therefore responsible for the production of thermally stable
`resist features. Operations are often included in the processing of positive resists
`that can accomplish similar thermal stability enhancements.
`This chapter addresses the critical issues involved in the processing of single—
`layer resists materials. Process steps to be discussed include:
`
`Resist stability, contamination, and filtration
`
`Substrate priming
`Resist coat
`
`Soft bake
`
`Exposure
`
`Postexposure bake
`
`Development
`
`Swing effects
`
`Hard bake and postdevelopment treatment
`
`The step—by—step process flow for DNQ/novolac resists has been covered else-
`where, and the reader is directed to these references for additional description
`[1—3]. Specific details are given here for positive DNQ/novolac resists and both
`positive and negative DUV chemically amplified resists based on PHS.
`
`IPR2016-01376 Page 0004
`
`IPR2016-01376 Page 0004
`
`

`

`Resist Processing
`
`51 7
`
`z RESIST STABILITY, CONTAMINATION, AND FILTRATION
`
`2.1 DNQ/Novolac Resist Stability and Filtration
`
`DNQ/novolac resists have proved to be robust materials with respect to sensi—
`tivity to thermodynamic and aging effects while stored in uncast form. A re—
`sist shelf life of several months can be expected with no significant change in
`lithographic performance. As resists are considered for application in produc—
`tion, the stability of materials at various points of the process also needs to be
`considered.
`For DNQ/novolac resists, aging can lead to an increase in absorption at
`longer wavelengths. Resist materials are susceptible to several
`thermal and
`acid/base (hydrolytic) reactions when stored [4]. These include thermal degrada—
`tion of the DNQ to ICA followed by acid—induced azo dye formation and azo
`coupling of the DNQ and novolac. A characteristic “red darkening” results from
`this coupling, induced by the presence of acids and bases in the resist. Although
`long—wavelength absorbance is altered by this red azo dye, the impact on UV
`absorbance and process performance is most often negligible. Degradation
`mechanisms can also result in crosslinking,
`leading to an increase in high—
`molecular-weight components. Hydrolysis of DNQ may occur to form more
`soluble products and hydrolysis of solvents is possible, which can lead to the
`formation of acids [5]. The practical limitation of shelf life for DNQ/novolac
`resists is generally on the order of 6 months to 1 year. Once coated, resist films
`can absorb water and exhibit a decrease in sensitivity, which can often be re-
`gained through use of a second soft bake step. As will be described, process de-
`lays for chemically amplified PHS resists are much more critical
`than for
`DNQ/novolac materials.
`.
`A larger problem encountered when storing DNQ/novolac resists is sensitizer
`precipitation. With time, DNQ PAC can fall out of solution, especially at high
`temperatures. These crystallized precipitates can form most readily with high
`loading levels of DNQ. In addition, resist particulate levels can be increased by
`the formation of gel particles, a result of acid—induced novolac crosslinking via
`thermal decomposition of DNQ. Any of these routes to particulate formation
`can lead to levels exceeding that measured by the resist manufacturer. Because
`of this, point—of—use filtration has become common practice for most production
`applications to ensure photoresist consistency [6]. Resist materials are com—
`monly filtered at a level of approximately 25% of the minimum geometry size.
`As the geometry size approaches sub—0.35 um, filtration requirements may ap-
`proach 0.05 pm. Such ultrafiltering will have an impact on how resists can be
`manufactured and used. Filtration speed is dramatically reduced and material
`preparation becomes more costly. Similar concerns can exist for pump through—
`put during resist dispensing. Fractionation of a resist material can also occur,
`
`IPR2016-01376 Page 0005
`
`IPR2016-01376 Page 0005
`
`

`

`518
`
`Smith
`
`resulting in the removal of long polymer chains and a change in process per—
`formance. To illustrate this, consider an i-line resist (DNQ/novolac) with a
`molecular weight on the order of 10—20 X 103 g/mol (number average). The
`resulting average polymer chain size is nearly 5 to 6 nm with a maximum as
`large as 40 nm. In highly concentrated resist formulations (>30 Wt. %), inter-
`twisting of polymers can result in chain sizes greater than 80 nm. If such a re-
`sist is filtered to 0.05 gm,
`the largest polymer chains can be removed. As
`technology progresses toward smaller feature resolution, it is clear that partic-
`ulate and filtration issues need to be carefully considered.
`
`2.2 Stability Issues for Chemically Amplified PHS Resists
`
`Filtration concerns for i-line resists are extended to deep UV lithography as
`
`PHS resists are considered. As sub—0.25. um geometry is pursued, the issue of
`ultrafiltering becomes an increasingly important problem. In addition, environ-
`mental stability issues are present for many chemically amplified resists
`(CARS) that are not issues for DNQ/novolac resists, especially for resists based
`on acid—catalyzed reactions and PHS resins. Ion exchange methods are conven—
`tionally used to reduce ion contamination levels in resists below 50 ppb. Ionic
`contamination reduction in both positive and negative CAR systems needs to
`be carefully considered. Deprotection of acid—labile components can result from
`reaction with cationic exchange resins. The catalytic acid produced upon expo—
`sure of these resists is also easily neutralized with base contamination at ppb
`
`levels. These contaminants can include such things as ammonia, amines, and
`NMP, which are often present in IC processing environments [7]. Any delay be—
`tween exposure and postexposure bake (PEB) can result in a decrease in sensi—
`tivity and the formation of a less soluble resist top layer or “T—top.”
`To reduce the likelihood of base contamination of these resists, several im—
`
`provements have been made in resist formulations. One method explored to re—
`duce acid loss is the use of low—activation—energy (Ea) polymers with highly
`reactive protection groups. These resists are sufficiently active that deprotection
`can occur immediately upon exposure, significantly reducing the sensitivity to
`PEB delay effects [8,9]. Additives have also been incorporated into DUV PHS
`resists to improve their robustness to contamination effects [10,11] and resist top-
`coating approaches have been introduced [12]. By coating a thin water—soluble
`transparent polymeric film over a resist layer, protection from airborne contam—
`ination can be made possible. This “sealing” layer is removed prior to devel—
`opment with a water rinse. Although such a solution leads to minimal additional
`process complexity, it is still desirable to use resist techniques that do not re—
`quire additional material layers. An alternative route for the reduction of con—
`tamination effects is the use of high—activation—energy resist materials. By
`
`reducing the reactivity of a resist, much higher bake processes are allowed. In—
`
`IPR2016-01376 Page 0006
`
`IPR2016-01376 Page 0006
`
`

`

`Resist Processing
`
`519
`
`creasing the bake temperatures above a polymer’s Tg results in a densification
`of the photoresist. This leads to a significant decrease in the diffusion rate of
`airborne base contamination prior to or after exposure [13,14]. These high-Ea
`resists also require that a photoacid generator be chosen that can withstand
`high temperatures. Other methods used to reduce base contamination of acid—
`acatalyzed resists include the use of activated charcoal air filtration during re-
`sist coating exposure, and development operations [15]. This is now considered
`a requirement for processing of PHS CAR resists. Environmental base contam-
`ination can be neutralized and further reduced by adding weak acids to these
`filters.
`
`The stability or shelf life of PHS-based resists is also influenced by the struc—
`ture of polymer protective groups. This is especially true for low—activation—
`energy (high—reactivity) resists, for which the liability of protective groups may
`decrease usable resist shelf life. Conversely, the more stable protective groups
`utilized with high—activation—energy (low—reactivity) resists lead to a higher de-
`gree of stability.
`
`3 RESIST ADHESION AND SUBSTRATE PHIMING
`
`Adequate adhesion of photoresist to a wafer surface is critical for proper process
`performance. Resist adhesion failure can occur not only during photolithography
`operations but also in subsequent etch, implant, or other masking steps. Nega—
`tive resists are less prone to adhesion failure, as crosslinking results in a net—
`worked polymer that is bound to the wafer surface. Positive resists (especially
`phenolic—based materials such as novolac or PHS resists) are more likely to be
`single-polymer chains and rely on weaker physical and chemical forces for ad—
`hesion. Etch process undercutting can often result from inadequacies at the re—
`sist interface, resulting in loss of etch line width control. The causes of resist
`adhesion failure are generally related to dewetting of a photoresist film. This can
`result from a large discordance between the surface tension of the wafer and that
`of the resist material, especially when coating over silicon oxide. Silicon diox—
`ide is an especially difficult layer to coat over because it provides a hydrophilic
`surface (water attracting) to a hydrophobic resist (water repelling). The surface
`tension of thermal silicon dioxide may be on the order of 15 dynes/cmz, whereas
`the surface tension of phenolic resists in casting solvent may be near 30
`dynes/cmz. Surface defects can also cause adhesion failure as surface free en—
`ergy can result in dewetting.
`Methods of adhesion promotion can be used for most silicon oxide lay—
`ers, whether thermally grown, deposited, native, or glasslike. Chemical pas—
`sivation of these surfaces is generally carried out using silylating priming
`agents, which act to modify the wafer surface. Some benefit can be realized
`With priming of layers other than oxides if techniques promote a closer match—
`
`IPR2016-01376 Page 0007
`
`IPR2016-01376 Page 0007
`
`

`

`520
`
`Smith
`
`ing of material surface tension. Alkylsilane compounds are generally used to
`prime oxide surfaces, leading to a lowering of surface hydrophilicity. The most
`commonly used silane-type adhesion promoter is hexamethyldisilazane (HMDS).
`Other
`similar promoters are available,
`including trimethylsilyldiethylamine
`(TMSDEA), which can be more effective but also less stable, resulting in lower
`shelf and coated lifetimes. Reduction of substrate surface tension is carried out
`in two stages, as shown in Fig. 1. Shown here is a silicon oxide surface with ad-
`sorbed water and OH groups. An initial reaction of water with an alkylsilane
`(HMDS) produces an inert hexamethyldisiloxane and ammonia, resulting in
`a dehydrated surface. Further reaction with HMDS produces a trimethylsilyl—
`substituted hydroxyl or oxide species and unstable trimethylsilylamine. With
`heat, this unstable compound reacts with other surface hydroxyl groups to pro-
`duce further ammonia and a trimethylsiloxy species. The process continues until
`steric hindrance (via the large trimethylsilyl groups) inhibits further reaction.
`Surface priming using HMDS, TMSDEA, or similar agents can be carried
`out in either liquid— or vapor—phase modes. In either case, elevated process tem-
`peratures (~100°C) must be reached to complete the priming reaction. Sub—
`strates should be cleaned prior to application using UV ozone, HF dip, plasma,
`
`
`
`(a)
`
`Si
`
`Si
`
`Si
`
`Si
`
`(b)
`
`
`
`sum); 83:93 S{:H3)3 :(CH3)3ofi(CHf;H.
`
`
`(C) WM
`
`Figure 1 Adhesion promotion of a silicon oxide surface with HMDS surface prim’
`ing. The substrate is first dehydrated upon reaction with silane promoter. Further reac-
`tion with heat leads to a hydrophobic surface.
`
`IPR2016-01376 Page 0008
`
`g
`
`IPR2016-01376 Page 0008
`
`

`

`Resist Processing
`
`52 1
`
`or other “oxidative” cleaning methods. Adhesion of photoresist to silicon nitride
`or deposited oxide layers can be enhanced by using an oxygen/ozone plasma
`treatment. Priming agents are generally best applied using vapor prime meth-
`ods, either in line or in batch vacuum ovens. Uniformity and reduced chemical
`usage make this more attractive than liquid methods.
`Overpriming of a wafer surface can result in dewetting and lead to further
`adhesion problems. This can occur with repeated treatment or by using exces—
`sive vapor times. Problems are often noticed in isolated substrate areas, de—
`pending on device topography or condition. A phenomenon known as resist
`“popping” can also occur as a result of overpriming; in this case high—fluence
`exposure (such as that encountered with heavy UV overexposure in ion im—
`plantation steps) can cause failure of weakened resist adhesion. Deposition of
`resist debris onto adjacent substrate areas can result. Measurements of resist
`surface tension using water contact angle techniques can identify such over—
`priming problems. Remedies include the use of shorter priming times, resist
`solvents with lower surface tension, double resist coating steps, or a pretreat—
`ment of the wafer surface with the resist casting solvent. Oxygen or ozone
`plasma treatments can also correct an overprimed wafer surface and allow
`repriming under more appropriate conditions.
`The strength of adhesion bonds between a photoresist and a substrate has
`also been shown to influence the Tg and thermal expansion coefficient of a thin
`film. The impact is greatest as resist films approach 1000 A thicknesses [16].
`
`4 RESIST COATING
`
`4.1 Resist Spin Coating Techniques and Control
`
`Photoresist can be dispensed by several methods, including spin coating, spray
`coating, and dip coating. The most widely used methods for coating resist onto
`wafer substrates are spin coating methods. During spin coating, resist is dis—
`pensed onto a wafer substrate (either statically or dynamically), accelerated to
`a final spin speed, and cast to a desired film thickness. Variations on this
`process have been suggested, including the use of a short—term high—speed ini—
`tial coating step followed by a slow drying stage [17]. Spin coating processes
`use the dynamics of centrifugal force to disperse a polymeric resist material
`over the entire wafer surface. The flow properties (rheology) of the resist in—
`fluence the coating process and need to be considered to achieve adequate re-
`sults [18]. In addition, solvent transport through evaporation occurs, which can
`result in an increase in resist viscosity and shear thinning, affecting the final
`film properties. As a resist—solvent material is spin cast, the film thickness de—
`creases uniformly, at a rate dependant on the spin speed (0)), kinematic viscos-
`
`IPR2016-01376 Page 0009
`
`IPR2016-01376 Page 0009
`
`

`

`522
`
`Smith
`
`ity (1)), solids concentration (c), solvent evaporation rate (e), and initial film
`thickness, expressed by the following rate equations:
`
`é _ —620)2h3
`dt
`31)
`
`dL
`dt
`
`(
`
`
`2032/13
`31)
`
`)
`
`e
`
`(1)
`
`where dS/dt and dL/dt are rate of change of solids (S) and solvents (L), re—
`spectively [19]. The results are shown in Fig. 2 for a 1—}.Lm film, where both
`solids and solvent volumes are plotted against spin time. Initially, concen—
`tration changes little as resist spread dominates. When the resist thickness
`drops to one third of its original value, evaporation dominates and solvent
`content reaches its final value. The high viscosity of the resist eliminates fur—
`ther flow.
`
`The primary material factors that influence spin-coated film properties in—
`clude the resist polymer molecular weight, solution viscosity, and solvent boil—
`ing point (or vapor pressure). Primary process factors include wafer spin speed,
`acceleration, temperature, and ambient atmosphere. The thickness of a resist
`film can be modified to some extent through control of the rotation speed of
`
`THICKNESS
`
`RELATIVE
`
`
`I
`I
`L_
`L__
`IO
`20
`3O
`4O
`TIME (s)
`
`O
`
`0
`
`50
`
`Calculated time dependance during spin coating on the volume of solids
`Figure 2
`(S) and solvent (L) per unit area normalized to initial values. When the resist thickness
`drops to one third of its original value, evaporation dominates and the solvent content
`reaches its final value. (From Ref. 19.)
`
`IPR2016-01376 Page 0010
`
`IPR2016-01376 Page 0010
`
`

`

`Resist Processing
`
`523
`
`the substrate. Resist thickness is inversely proportional to the square root of
`spin speed (co):
`
`Thickness cc —1-
`5
`
`(3)
`
`To achieve large thickness changes, modification of the resist solution viscos-
`ity is generally required, as coating at excessively low or high speeds results in
`poor coating uniformity. At excessively high speeds, mechanical vibration and
`air turbulence result in high levels of across—wafer nonuniformity. At low spin
`speeds, solvent loss of the resist front as it is cast over the substrate results in
`a situation of dynamic resist viscosity, also resulting in high levels of nonuni—
`formity. The optimal spin speed range is dependent on wafer size. Wafers up to
`150 mm can be coated at rotation speeds on the order of 4000 to 5000 RPM.
`Larger substrates require lower speeds.
`The optimum coating thickness for a resist layer is determined by the posi-
`tion of coherent interference nodes within the resist layer. Standing waves (see
`Section 9) resulting from reflections at the resist/substrate interface result in a reg—
`ular distribution of intensity from the top of the resist to the bottom. This distrib-
`ution results in a “swing” in the required clearing dose (E0) for a resist, as shown
`in Fig. 3. Three curves are shown, for polysilicon, silicon nitride (1260 A), and
`silicon dioxide (3700 A) coated substrates. A general upward trend in E0 is seen
`as resist thickness increases. This is due to the residual nonbleachable absorp-
`
`
`
`
`
`[mJ/cm2]
` Eo(clearingdose)
`
`
`
`
`
`
`i‘i'I—Llh‘i'l'l'l'i
`
`J
`
`l...¢_|
`
`3. 8 3 E>1 Q
`u—4
`I-l
`—:
`
`1.14
`
`I
`l
`J
`I
`I
`I
`I
`\D
`00 O
`—: H N
`
`Resist thickness [um]
`
`Clearing dose (E0) swing curves for an i—line resist over polysilicon, silicon
`Figure 3
`dioxide (3700A), and silicon nitride (1260A). The increasing trend in required dose is a
`function of residual absorption. Conditions of minimum interference leads to maximum
`E0 values but minimal scumming.
`
`IPR2016-01376 Page 0011
`
`IPR2016-01376 Page 0011
`
`

`

`524
`
`Smith
`
`tion of the resist, which can be significant. (For a resist with a residual ab—
`sorption of 0.10 um—l, the intensity at the bottom of a l—um resist layer is 90%
`of that experienced at the top.) In addition to this E0 trend, sensitivity oscillates
`from minimum to maximum values with thickness. Within one swing cycle, an
`exposure dose variation of 32% exists for the polysilicon substrate, 27% for sil—
`icon nitride, and 36% for silicon dioxide. When resist is coated over a dielec—
`tric layer, such as silicon dioxide, silicon nitride, or an antireflective coating
`(ARC), there will be a shift in the phase of E0 oscillations. Analysis of swing
`behavior may therefore be unique for various lithographic levels. Coated thick—
`ness optimization can be performed using these swing curves, determined ex—
`perimentally through open frame exposure of resist coated within a small range
`of thicknesses. Lithographic modeling can aid in generation of such relation-
`ships using knowledge of the resist refractive index, absorption properties (pos—
`sibly dynamic), exposure wavelength, and resist/substrate reflectivity.
`Inspection of the E0 swing curve in Fig. 3 suggests several possibilities for
`resist thickness, of which only a few are desirable. For polysilicon, there is a
`minimum dose requirement at a thickness of ~1.01 um, where constructive in—
`terference occurs, and there is maximum intensity at the resist base. At a resist
`thickness over polysilicon of ~1.06 um, destructive interference leads to a max—
`imum E0 requirement. Other alternatives might include positions on either side
`of these values (between nodes). Thicknesses corresponding to these midnodal
`positions allow the least amount of coating process latitude, as small devia—
`tions from the targeted film thickness lead to significant changes in dose re—
`quirements. Greater latitude exists at maximum interference positions, where
`there is a minimum requirement for exposure dose, which may be an attrac—
`tive choice. Small changes in film thickness result only in small E0 variations
`but the direction of these changes is toward higher clearing dose values. The
`result may be scumming of resist features resulting from underexposure, a sit—
`uation that is unacceptable. The best choice for targeted film thickness may be
`at a corresponding interference minimum, where small thickness changes re—
`sult in a small decrease in the dose requirement. Slightly lower throughput
`may result (generally not a gating factor in today’s exposure operations) but
`this will ensure no resist scumming related to underexposure. Image fidelity at
`the top surface of a resist film is also influenced by film thickness and posi—
`tions on the interference curve. By coating at a midnodal thickness, top sur-
`face rounding or T—topping can result (see the Section 9 for further discussion).
`During spin coating, a large amount of resist “free volume” can be trapped
`within a resist layer. A simplified free-volume model of molecular transport can
`be quite useful for correlation and prediction of diffusion properties of resist
`materials [20,21]. (The reader is directed to Refs. 20 and 21 for a detailed dis—
`cussion of diffusion in polymer-solvent systems.) Volumetric expansion en-
`
`IPR2016-01376 Page 0012
`
`IPR2016-01376 Page 0012
`
`

`

`Resist Processing
`
`525
`
`hances polymer chain mobility and acts similarly to the addition of plasticizers.
`The resist’s glass transition temperature (Tg)
`is lowered and the dissolution
`properties of novolac— and PHS—based resist can be increased [22]. Coating—
`induced free volume has been shown to affect acid diffusion as well and be—
`comes a concern when considering reduction of airborne base contamination
`and postexposure delay.
`
`4.2 Solvent Contribution to Film Properties
`
`Residual casting solvent can act as a plasticizer and can reduce the Tg of a re—
`sist. Resist solvent content has been shoWn to be dependent on film thickness.
`A 1000 A resist film may, for example, exhibit 50% more solvent retention
`than a 10,000 A film. Figure 4 shows residual solvent in PHS polymer films
`coated at thicknesses of 12,000 A and 1100 A. Only near the resist Tg (135°C)
`does the solvent content for the 1100 A film approach that of the thicker film.
`Table 1 shows diffusion coefficients for PGMEA solvent in the same PHS film
`thicknesses, determined by diffusion analysis during 2 hours of baking. These
`results may be due to a smaller degree of inter— or intramolecular hydrogen
`bonding in thinner films [23—25], which can allow a stronger polymer inter—
`action with the casting solvent and lower solvent evaporation rates. A higher
`solvent content leads to an increased dissolution rate and increased diffusivity
`levels. When considering various resist solvent systems, it might also be ex-
`pected that lower boiling point (Tb) solvents would lead to lower solvent
`
`(weight%)
`Residualsolvent
`
`
`20
`
`40
`
`60
`
`80
`
`1 00
`
`1 20
`
`140
`
`Annealing temperature (°C)
`
`Figure 4 Bake temperature dependence of residual PGMEA solvent in 1,100A and
`12,000A spin cast films annealed for 1300 minutes. (From Ref. 24.)
`
`IPR2016-01376 Page 0013
`
`IPR2016-01376 Page 0013
`
`

`

`526
`
`Smith
`
`Table 1
`
`Diffusion Coefficients of PGMEA Solvent
`
`in PHS Films for 2 Hours of Baking
`
`Dijfusion coefi‘icient (cmZ/s)
`
`
`
` Temperature 0.11 -}.Lm film 1.2-um film
`
`
`
`70°C
`
`4.2 x 10—14
`
`1.2 X 10—12
`
`90°C
`9.4 x 10—14
`4.4 X 10—12
`1.1 x 10—13110°C 1.4 x 10—11
`
`
`
`
`
`retention than higher Tb solvents. The opposite, however, has been demon—
`strated [26]. PGMEA, for instance, has a boiling point of 146°C and an evap—
`oration rate of 0.34 (relative to n—butyl acetate). Ethyl lactate has a higher Tb of
`154°C and an evaporation rate of 0.29. Despite its lower boiling point, PGMEA
`is more likely to be retained in a resist film. The reason for this is a skin for—
`mation that results from rapid solvent loss during the coating process [27]. The
`resist viscosity at the surface increases more rapidly for PGlVIEA as solvent is
`exhausted,
`leading to more residual solvent remaining throughout the resist
`film. If a resist film is then baked at temperatures below the bulk Tg, densifi—
`cation of surface free volume is allowed only at the top surface of the film and
`is prevented throughout the bulk. Entrapped solvent therefore leads to an ap—
`parent surface induction effect, which can be reduced only if the resist is baked
`above its Tg. Because solvent content plays an important role in determining the
`ultimate glass transition temperature of the resist (and therefore its dissolution
`properties), any postcoating incorporation of solvents can also have an adverse
`impact on performance. Such additional solvent may be encountered, for in—
`stance, when using an edge bead removal process based on acetone, ethyl lac—
`tate, pentanone, or other organic solvents.
`
`4.3 Substrate Contribution to Resist Contamination
`
`Continuous improvements have been made in the materials and processes used
`for DUV PHS chemically amplified resists to reduce top—surface base contam—
`ination effects. An additional contamination problem occurs when processing
`PHS resists over some substrates. Resists coated over Si3N4, BPSG, SOG, A1,
`
`and TiN have seen shown to initiate a substrate contamination effect that can
`result in resist scumming or “footing.” With TiN substrates, the problem has
`been attributed to surface N—3 and TiOz, which can act to neutralize photogen—
`erated acid, resulting in a lowering of the dissolution rate at the resist/substrate
`interface [28]. Sulfuric acid/hydrogen peroxide and oxygen plasma pretreat—
`ments have been shown to reduce contamination effects when coating over
`
`Si3N4 and other problematic substrates [29].
`
`IPR2016-01376 Page 0014
`
`IPR2016-01376 Page 0014
`
`

`

`Resist Processing
`
`52 7
`
`4.4 Edge Bead Removal
`
`After a spin—coating process, a bead of hardened resist exists at the edge of a
`wafer substrate. Formation of this edge bead is caused in part by excessive re—
`sist drying and can result in resist accumulation up to 10 times the thickness of
`the coated film. Elimination of this edge bead is required to reduce contamina—
`tion of process and exposure tools. Solvent edge bead removal (EBR) tech—
`niques can be utilized to remove this unwanted resist by spraying a resist
`solvent on the back side of the wafer substrate. Surface tension allows removal
`
`of a 2—3—mm resist edge from the front resist surface while removing any back
`side resist coating. Acetone, ethyl lactate, and pentanone are possible solvent
`choices for edge bead removal processes.
`
`5 RESIST BAKING—SOFTBAKE
`
`5.1 Goals of Resist Baking
`
`Baking processes are used to accomplish several functions and generally alter
`the chemical and/or physical nature of a resist material. Goals of resist baking
`operations may include the following, which are accomplished at various stages
`during resist processing:
`
`Solvent removal
`
`Stress reduction
`
`Planarization
`
`Reduction of resist voids
`Reduction of standing waves
`Polymer crosslinking and oxidation
`Polymer densification
`Volatilization of sensitizer, developer, and water
`Induction of (acid) catalytic reactions
`Sensitizer/polymer interactions
`
`Polymeric resist resins are thermoplastic in nature, m