aqueous-develop, photosensitive polynorbornene dielectric...

Aqueous-Develop, Photosensitive Polynorbornene Dielectric:Optimization of Mechanical and Electrical Properties

MEHRSA RAEIS-ZADEH,1 NORAICA DAVILA MELENDEZ,2

YU-CHUN CHEN,1 and PAUL A. KOHL1,3,4

1.—School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, 311 FerstDr., Atlanta, GA 30332-0100, USA. 2.—Department of Electrical and Computer Engineering,University of Puerto Rico, Call Box 9000, Mayaguez, PR 00681-9042, USA. 3.—e-mail: [email protected]. 4.—e-mail: [email protected]

The impact of thermal cure conditions on the mechanical and electricalproperties of an epoxy cross-linked network incorporating a polynorbornene(PNB) dielectric polymer was studied. The cross-linking of the dielectriccomposition was achieved by an acid-catalyzed cationic cure reaction initiatedby either thermal or photolytic activation of a photoacid generator. It is pro-posed that the observed mechanical and electrical properties of the fully curedpolymer composition are the result of the development of a three-dimensionalcross-linked network tying together the PNB polymer and multifunctionalepoxy additives. The epoxy ring-opening reaction was measured using Fou-rier-transform infrared spectroscopy. The reduced modulus, internal filmstress, dielectric constant, and swelling behavior of cross-linked films werestudied as a function of curing temperature. Trends in the observed propertiesare explained by formation of a three-dimensional cross-linked network anddegradation of the cross-links between the multifunctional epoxy additives athigh temperature. It was also found that exposure of the film to aqueous baseplays a role in the cure process and has a positive effect on the final properties.The optimum values of modulus, dielectric constant, residual stress, andmoisture content were found for films cured at 160�C for 1 h. This relativelylow cure temperature is potentially advantageous in device assembly andprocessing.

Key words: Polynorbornenes, cross-linking, curing, FT-IR, electrical andmechanical properties

INTRODUCTION

Advances in microelectronic devices and micro-electromechanical systems (MEMS) can benefit fromincreases in device speed, level of functionality, andshrinkage in component size. Advanced materialswith improved electrical and mechanical propertiescompatible with existing MEMS and microelec-tronic processing techniques play an important rolein system-level improvements. There has beenextensive research and development focused on

low-dielectric-constant polymers for various MEMSand microelectronics applications.1–6 Dielectricsbased on a polynorbornene (PNB) backbone are ofspecial interest because PNB has low dielectric con-stant and moisture uptake.1,7,8 However, MEMS andmicroelectronics industries also require dielectricswith superior mechanical, electrical, and chemicalproperties. Photosensitive dielectric materials arevaluable because of the ease of processing. Aqueousbase developing is environmentally friendly.

Previously, an aqueous-base-developable PNB-based dielectric polymer, Avatrel 8000P shown inFig. 1, was disclosed.9,10 In this polymer, carboxylic-acid-functionalized side-chains on the PNB back-bone provide reactive sites for cross-linking with

(Received February 25, 2011; accepted June 13, 2011;published online July 20, 2011)

Journal of ELECTRONIC MATERIALS, Vol. 40, No. 10, 2011

DOI: 10.1007/s11664-011-1704-z� 2011 TMS

2126

multifunctional epoxy additives. Fluorinated alco-hol and carboxylic acid groups provide solubility inaqueous base for environmentally friendly develop-ing in 0.26 N tetramethyl ammonium hydroxide.

The epoxy-based cross-linking of the functional-ized PNB can be initiated by acid catalysis leadingto a negative-tone, photodefinable dielectric mate-rial. This reaction has been studied for numerousepoxy systems.11–16 The photoinitiated reactionleads to the formation of a three-dimensional cross-linked network.17–19 During exposure to ultraviolet(UV) radiation or elevated temperature, the photo-acid generator (PAG) decomposes to form a proticacid (HX).20,21 The acidic proton (H+) reacts with alone pair of electrons on the epoxide oxygen, leadingto epoxy ring opening and the formation of ahydroxyl group and a carbocation (C+). Cross-linkingoccurs when the carbocation reacts with either apendant carboxylic acid group of a neighboringpolymer chain to form a carboxylic ester linkageor with a second epoxy ring to form a polyetherlinkage.

In epoxy resins, degradation of the polyether-typelinkages is known to occur at temperatures inexcess of 250�C.19 Previously, Chiniwalla et al.proposed a similar degradation scheme for poly-ether cross-linking degradation in cationically curedepoxy cross-linked PNB polymers at temperaturesapproaching 160�C based on the observation ofchanges in mechanical properties of the cross-linkednetwork with increasing cure temperature.17,18 Inthe study, the PNB is linked to a three-dimensionalcross-linked epoxy network through carboxylic esterbonds which prevent severing of the PNB polymerbackbone from the cross-linked epoxy networkduring degradation. Such degradation must, there-fore, occur in the multifunctional epoxy additivesincluded in the formulation. The proposed degra-dation mechanism results in cleavage of the poly-ether linkages, reducing the density of cross-links inthe developing three-dimensional cross-linked net-work. Similar decomposition mechanisms have beenproposed for diglycidyl ethers of bisphenol A(DGEBA) type epoxies at temperatures in excess of250�C.17,18,22

In this report, the reactions that occur duringepoxide ring opening, cross-linking, and degrada-tion of Avatrel 8000P were tracked by Fourier-transform infrared (FT-IR) spectroscopy. Filmproperties were a function of the degree of cross-linking in the polymer.23,24 The impact of cureconditions and degree of cross-linking on the elec-trical and mechanical properties of Avatrel 8000Pwas evaluated. The dielectric constant, reducedmodulus, film residual stress, and solvent absorp-tion in polymer films cured at different tempera-tures were studied. The optimum properties wereobtained at a relatively low cure temperature,160�C. Also, improved mechanical properties wereobtained compared with previous studies of similarmaterials.9 The relationship between the cure con-ditions and material properties was optimized toobtain high-quality Avatrel 8000P films. It is alsoshown that the properties of the cured film wereimproved by exposure of the polymer to aqueousbase.

EXPERIMENTAL PROCEDURES

PNB polymer, Avatrel 8000P, was obtained fromPromerus LLC (Brecksville, OH). Photosensitivepolymer mixtures were formulated by mixing thePNB polymer with a PAG, a UV-absorbing sensi-tizer, multifunctional epoxy cross-linkers, and anadhesion promoter in propylene glycol monomethylether acetate (PGMEA). Samples were spin-coatedon 100-mm-diameter h100i silicon wafers using aCEE 100CB spinner at 1500 rpm for 30 s, produc-ing �25-lm-thick films. To enhance film–substrateadhesion, 1 wt.% solution of 3-aminopropyltrieth-oxy silane in ethanol was applied to the substratesurface in all experiments according to the proce-dure of Raeis-Zadeh et al.25 UV exposure wasperformed using a Karl Suss MA-6 mask alignerwith a 365-nm filter. Polymers were soft-baked andpostexposure baked at 100�C for 5 min on a hot-plate in air. The latent image in the polymer filmswas developed for 3 min using Shipley MF-3190.26 N tetramethyl ammonium hydroxide (TMAH)developer. Curing was performed in a nitrogen-purged furnace, and the samples were ramped totheir cure temperature at 5�C/min and held attemperature for 1 h. The furnace was allowedto cool slowly to ambient temperature by naturalconvection.

Film thickness was measured after cure by usinga Veeco Detak profilometer. Swelling of cross-linkedpolymer films on silicon wafer substrates was eval-uated by measuring the increased sample weightversus time of exposure to the swelling solvent. Themeasurements were taken using an Ohaus VoyagerPro balance with readability of 0.0001 g and line-arity of ±0.0002. FT-IR spectroscopy was used tofollow the epoxy ring-opening reactions using aMagna 560 spectrometer (Nicolet Instruments).Scans were collected in transmission mode on

RFARCOOH

Fig. 1. Chemical structure of PNB polymer (Avatrel 8000P).

Aqueous-Develop, Photosensitive Polynorbornene Dielectric: Optimization of Mechanicaland Electrical Properties

2127

potassium bromide (KBr) substrates with 512 scansbeing averaged for each measurement at a resolutionof 2.00 cm�1. Samples were prepared by coatingeach KBr plate at 1500 rpm for 30 s, producing �25-lm-thick films. Samples were soft-baked and post-exposure baked at 100�C for 5 min. After exposureto UV light, the edge and back of the KBr disk werecovered by parafilm to prevent film delaminationand dissolution of the KBr disc in the TMAHdeveloper. Samples were developed in TMAHdeveloper for 3 min. To monitor conversion, theepoxy ring-opening reaction was monitored by fol-lowing the disappearance of the 844 cm�1 peak,associated with the oxirane ring C–O–C stretch.

Dielectric measurements were conducted by fab-rication of parallel-plate capacitors. The bottomplate of the capacitor was a full surface film ofsputtered aluminum (400 nm thick) deposited on asilicon surface by direct-current (DC) sputtering.After the adhesion promoter was applied onto thealuminum layer, the polymer was spin-coated onto a100-mm-diameter h100i silicon wafer at 1000 rpmfor 30 s, producing a �50-lm-thick film. The samplewas soft-baked and postexposure baked at 100�C for5 min. The edge of the sample was covered withparafilm to prevent delamination during developing.The sample was developed in TMAH developer for3 min, and cured at the desired temperature. Thetop electrode of 400 nm aluminum was deposited ontop of the polymer film and patterned with photoli-thography and wet etching to form the top plate ofthe parallel-plate capacitor with capacitance givenby Eq. (1).

C ¼ ere0A

t

� �; (1)

where C is the capacitance, er is the relativedielectric constant, e0 is the permittivity of freespace, A is the area, and t is the thickness of thedielectric. Conductance and capacitance were mea-sured at 10 kHz using a Hewlett Packard 4236 LCRmeter on a Karl Suss probe station. The ASTMcorrection, D 150, for fringing fields was performedaccording to Eq. (2).26

Ce ¼ 0:0041jv0 � 0:00334 ln tþ 0:0122� �

P; (2)

where Ce is the corrected capacitance, jv0 is theapproximate value of the dielectric constant and P isthe perimeter of the top measuring electrode.

Residual film stress was measured at room tem-perature using a He-Ne laser-based Flexus Tencorstress analyzer (model F2320).27 Stress measure-ments were performed on the samples prepared fordielectric measurements. The residual stress inpolymer films was calculated by monitoring thedifference in the wafer curvature between anAl-coated silicon wafer and one coated with thepolymer. The residual stress was calculated usingStoney’s equation, which relates the residual stress

of a film to the change in the radius of curvature ofthe supporting substrate (Eq. 3).

r ¼ E

1� m

� �h2

6Rt0; (3)

where r is the residual stress, E=1� m is the biaxialelastic modulus of the substrate (1.805 9 1011 Pafor h100i-oriented silicon), h is the substrate thick-ness, and t is the thickness of the film. R is thereduced change in radius, given by Eq. (4).

1

R¼ 1

R2� 1

R1; (4)

where R1 is the radius of curvature of the uncoatedsubstrate and R2 is the radius of curvature of thesubstrate after film coating and processing. Stoney’sequation correctly represents the average biaxialstress acting within a film deposited on a substratefor film thickness less than 10% of the substratethickness.28

To characterize mechanical properties of thesamples such as reduced modulus and hardness,nanoindentation was used as a depth-sensing tech-nique. Researchers have used this technique tocharacterize materials such as metallic alloys29 andceramics.30 However, nanoindentation of polymericmaterials is challenging due to their viscoelasticand viscoplastic response,31 low hardness,32 andresulting strain-rate dependence of deformation.33

In this work, the hardness (H) was defined as theapplied load per unit area of indentation, as givenby Eq. (5).34

H ¼ Pmax

A hcð Þ; (5)

where Pmax is the maximum load, and the projectedcontact area, A(hc), for an indenter with a tipimperfection as defined by Eq. (6).

A hcð Þ ¼ 24:5h2c þ

X7

i¼0

aih1=2i

c ; (6)

where hc was estimated for a geometrical constant(e) by using the Oliver and Pharr model (Eq. 7).

hc ¼ hmax � ePmax

S: (7)

Fused silica (quartz) is a commonly used standardbecause its elastic modulus does not change signif-icantly with indentation depth and, unlike metals, itdoes not have a surface oxide.35 However, it hasbeen shown that it is necessary to calibrate the tipwith a standard having a modulus similar to thesample to obtain an area function valid over theindentation depths of the sample.9 In this work, aHysitron polycarbonate calibration standard with a

Raeis-Zadeh, Melendez, Chen, and Kohl2128

modulus of 3.1 GPa was used to calculate the areafunction.

Quasistatic nanoindentation was performed onpolymer samples using a Triboindenter nanoinden-ter (Hysitron Inc., Minneapolis, MN) with a Berko-vich tip. The indenter was located on anantivibration table and enclosed in an acoustichousing. The peak load was varied between 250 lNand 6250 lN. The tip was loaded to maximum loadwithin 10 s, held for 10 s, and unloaded in 2 s. Tominimize the impact of the substrate on the inden-tation results, the maximum force was chosen so asto indent less than 5% of the total polymer filmthickness, which was 25 lm. Additionally, a 5 9 5array of points was indented in the center of thesamples to exclude edge effects. The maximum driftrate of the experiments was set at 0.1 nm/s and wasdetermined over a period of 40 s. The curvature ofthe Berkovich tip was between 250 nm and 970 nm.The Oliver–Pharr model was used to analyze theload–depth curves.9 The reduced modulus wasextracted from the 20–95% portion of the unloadingcurve.9,36 To eliminate the impact of thermal drift,the first data points were discarded so that theaverage hardness and modulus only includedindents above 500 nm.

RESULTS

The multifunctional epoxide cross-linking andpostcure degradation reactions determine the over-all degree of cross-linking in Avatrel 8000P andaffect the electrical and mechanical properties of thepolymer film.9 In this work, the cure temperature ofAvatrel 8000P was varied so as to optimize thedielectric and mechanical properties in the curedpolymer film. The extent of epoxide ring openingwas studied using FT-IR spectroscopy to helpdetermine the degree of cure. A 25-lm-thick film ofAvatrel 8000P was spin-coated onto a KBr disk and

taken through a process sequence consisting of soft-bake, exposure, postexposure bake, and three suc-cessive cure cycles, each for 1 h at 160�C. Infraredspectra from 700 cm�1 to 950 cm�1 were obtainedafter each process step in roughly the same locationon the KBr disk (Fig. 2). Three peaks correspondingto asymmetric and symmetric epoxide ring stretcheswere observed at 913 cm�1, 844 cm�1, and760 cm�1. A significant degree of epoxy ring openingoccurs as a result of the UV exposure and postex-posure bake processes, resulting in some degree ofcross-linking before the final cure step.

In this experiment, the postexposure bake stepswere performed using a KBr disk inverted onto asilicon wafer when placed on the hotplate. Theinverted structure was needed because the 4-mm-thick KBr disk has lower thermal conductivity thanthe 0.525-mm-thick silicon wafer, which was used asthe substrate in all other experiments. When the KBrdisk was postexposure baked in the right-side-upconfiguration, the film on the top surface of the KBrdisk did not reach the desired temperature duringthe soft-bake and postexposure bake. Inversion of theKBr disk onto a silicon wafer during hotplate bakebrings the polymer film into thermal contact withthe hotplate in a way that more closely replicatesthat of an Avatrel 8000P film directly on a siliconwafer baked in the right-side-up configuration.

After curing at 160�C for 1 h, the magnitude ofthe FT-IR epoxy peaks decreased. When the curedsample was heated again to 160�C for 1 h, there wasa small decrease in the peaks, suggesting that theepoxy ring-opening reaction had not reached com-pletion after the first 1 h cure. The third cure cycleat 160�C (Fig. 2) showed no significant change inthe FT-IR spectra, indicating that complete epoxidering opening of the Avatrel 8000P had occurredafter 2 h at 160�C.

A significant reduction in the magnitude of theepoxide ring peak height was observed in Fig. 2

0.2

0.4

0.6

0.8

1

1.2

1.4

700 750 800 850 900 950

Ab

sorb

ance

(A

.U.)

Wavenumbers (cm-1)

Soft Bake

Exposure

Post Exposure Bake

Cure

2nd Cure

3rd Cure

Fig. 2. Changes in infrared spectrum of thin-film Avatrel 8000P on KBr disc from 700 cm�1 to 950 cm�1 as a function of processing steps (softbake: 100�C for 5 min; exposure dose: 170 mJ/cm2; postexposure bake: KBr plate inverted onto the hotplate, 100�C for 5 min; cure at 160�C for1 h).


2129

after the UV exposure and postexposure bake. Theabsorbance after postexposure bake was furtherinvestigated by repeating the experiment with theKBr disk placed on the hotplate in the face-up con-dition, rather than in the face-down configuration,as shown in Fig. 2. Figure 3 shows the epoxideabsorbance after postexposure bake in the face-upcondition, where the lower conductivity of the KBrdisk affected the polymer heating. As shown inFig. 3, there was no significant change in the IRspectra between the soft-bake and postexposurebake steps. This confirms that the thermal insula-tion of the KBr disk indeed mitigates the effect ofthe postexposure bake heat treatment. The peakheight and area corresponding to the epoxide ringafter the first cure at 160�C for 1 h were similar(relative to the background) to the correspondingones in Fig. 2. To quantify the chemical changes inthe cured film, the relative absorbance of the twosamples was compared after the first cure step byusing the height of the soft-bake peak (hsoft bake) at844 cm�1 as a spectral reference. The infraredabsorbance corresponding to the epoxy group wascompared using Eq. (8).

Normalized absorbance ¼ hcure@844cm�1

hsoft bake@ 844 cm�1

; (8)

where h is the peak height using a backgrounddrawn by connecting the points at 800 cm�1 and870 cm�1 of the spectrum. The peak height of thecured spectrum (hcure) was used to quantify changesin absorbance. The normalized absorbance aftercure for the Avatrel 8000P film inverted on thehotplate (Fig. 2) and the sample baked in the face-up position (Fig. 3) were 0.18 and 0.27, respectively.The results show that the sample baked in theinverted postexposure bake condition had a higherdegree of epoxy ring opening after cure at 160�C for1 h than the one which was postexposure baked in

the film-up configuration. Thus, not only does thepostexposure bake assist in creating a latent imagefor later development (the usual benefit of postex-posure bake), but it also has an effect on how rapidlythe film is cured at the final cure temperature.Hereafter, all FT-IR samples on KBr in this studywere postexposure baked in the inverted configu-ration. The silicon substrates were all postexposurebaked in the upright configuration.

To examine the effect of exposure of the Avatrel8000P polymer film to developer (e.g., TMAH) onthe extent of the epoxy ring-opening reaction, newsamples were prepared and were exposed to TMAHdeveloper for 3 min after exposure to 170 mJ/cm2 of365-nm UV radiation. As described above, all sam-ples were postexposure baked. This exposure dosewas chosen because it represents full exposure ofthe Avatrel 8000P, as shown later. Figure 4 showsthat there is very little change in the magnitude ofthe FT-IR epoxy peak after the postexposure bakeand develop steps. However, a notable differencewas observed between the sample cured after it wasdeveloped (Fig. 4) and the sample cured withoutbeing developed (Fig. 2). The normalized absor-bance after cure was 0.18 for the undevelopedsample (Fig. 2) and 0.12 for the developed sample(Fig. 4). This shows that a higher degree of epoxyring opening occurred when the sample was exposedto the aqueous base developer prior to the first curefor 1 h. Further, there was no significant change inthe FT-IR after multiple cure cycles (each for 1 h)when the sample had been developed (Fig. 4),showing that complete epoxy ring opening occurredafter the first cure at 160�C for 1 h. Previous studiesignored the develop step during their cure studies,assuming that it had no effect on curing.9 However,the corresponding sample cured without havingbeen developed (Fig. 2) did eventually come to thefully cured state after multiple cure cycles at 160�Cfor 1 h in each cure cycle. This result shows that

0.2

0.4

0.6

0.8

1

1.2

1.4

700 750 800 850 900 950

Ab

sorb

ance

(A

.U.)

Wavenumber (cm-1)

Soft Bake

Exposure

Post Exposure Bake

1st Cure

2nd Cure

3rd Cure

Fig. 3. Effect of postexposure method on Avatrel 8000P cross-linking (soft bake: 100�C for 5 min; exposure dose: 170 mJ/cm2; postexposurebake: KBr disc in upright configuration on the hotplate, 100�C for 5 min; cure at 160�C for 1 h).


developing in TMAH assists in the epoxy ringopening and final cure.

There are several possible reasons why exposureto TMAH would assist in the final cure. Exposure ofthe Avatrel 8000P film to aqueous base could causesome degree of neutralization and dissolution of thephotogenerated acid created during exposure. ThePAG itself and byproducts (e.g., conjugate base)may also dissolve during developing. While leachingout the acid from the polymer film decreases theamount of acid catalyst, it may provide more freevolume and thus higher mobility for the activatedepoxy moieties within the film to move. Alterna-tively, incorporation of the TMAH base in the PNBcan plasticize the film. This observation is congru-ent with the fact that the residual stress in the filmdecreases after developing in TMAH, as shown laterin this paper. Additionally, the TMAH developercan act as a catalyst for the reaction of the carbox-ylic acid pendant on the polymer with the epoxygroups contained in the multifunctional cross-link-ers. This would improve the cross-linking within thepolymer. In any event, the beneficial effects of thedeveloping step are clear. The exact nature ofthe improved cross-linking with development maybe the subject of a future publication.

Cross-linking of Avatrel 8000P films can be acti-vated either photolytically or thermally. In photo-cross-linking of the polymer, the PAG is activatedby UV radiation, creating a proton which can cata-lytically activate epoxy ring opening on the multi-functional additive cross-linkers. The activatedepoxy can react with the carboxylic acid groupspendant from the PNB backbone. In thermal cross-linking of Avatrel 8000P, the PAG can be thermallyactivated, followed by acid-catalyzed epoxy activa-tion. Figure 5 shows the contribution of the thermalactivation of the PAG on the epoxy ring-openingreaction. The sample was not exposed to UV radia-tion and was cured at the same condition as in

Fig. 2 (160�C for 1 h). The 100�C, 5 min bake, whichwas identified as postexposure bake in Fig. 2,resulted in no ring opening in Fig. 5, where noexposure occurred; that is, the postexposure bakeresults in ring opening only when the sample isphotolytically activated (exposed to UV radiation).However, the 1 h, 160�C cure does result in epoxyring opening regardless of UV exposure, althoughthe UV exposure increases the percentage of epoxyring opening. The normalized absorbance of theepoxy groups after cure was 0.29 in Fig. 5, com-pared with 0.18 for the exposed sample in Fig. 2.These results show that, to achieve complete cross-linking of Avatrel 8000P at low cure temperature of160�C, both photo and thermal activation of epoxiderings are required.

Since UV exposure is critical to developing andcuring of the polymer, the effect of UV exposuredose on the degree of conversion of the epoxy ring-opening reaction in the polymer film was studied. AUV exposure dose sufficient to fully activate thePAG is desired, after which higher doses have littleor no effect on cure. An Avatrel 8000P film wasexposed to a 450 mJ/cm2 UV dose (compared withthe 170 mJ/cm2 used in Fig. 2), and the FT-IRspectrum was obtained (Fig. 6). After the PEB step,it can be seen that the 450 mJ/cm2 exposure dosecaused more epoxy conversion compared with the170 mJ/cm2 dose. After PEB, the peak at 844 cm�1

decreased by 55% for the 450 mJ/cm2 exposure dose,compared with a 38% decrease for the 170 mJ/cm2

exposure dose (Fig. 2). Additionally, there is essen-tially no change in the epoxy absorbance betweenthe first cure at 160�C for 1 h and the second andthird cure at 160�C (1 h each) because the film wasfully cured after the first cure cycle, whereas therewas a slight decrease in the epoxy peak after thefirst cure in Fig. 2. This shows that there is a slightdifference in the cure time when a higher UV dose isused. This change in epoxy absorbance between the

0.2

0.4

0.6

0.8

1

1.2

1.4

Ab

sorb

ance

(A

.U.)

Wavenumbers (cm-1)

Soft Bake

Exposure

Post Exposure Bake

Develop

Cure

2nd Cure

3rd Cure

700 750 800 850 900 950

Fig. 4. Effect of developing on epoxide ring opening of Avatrel 8000P (soft bake: 100�C for 5 min; exposure dose: 170 mJ/cm2; postexposurebake: KBr disc inverted onto the hotplate, 100�C for 5 min; develop for 3 min; cure at 160�C for 1 h).


2131

first and second cure (Fig. 2) was not viewed as asignificant change, since 170 mJ/cm2 was a suffi-cient dose for full development of the latent image.

The effect of exposure dose on the mechanicalproperties of cured Avatrel 8000P was investigatedfor UV doses ranging from 100 mJ/cm2 to 1728 mJ/cm2. The mechanical properties were measured bynanoindentation as a function of exposure dose for aset of samples cured at 160�C for 1 h, and for a set ofsamples cured at 200�C for 1 h. Nanoindentationwas performed on the samples, and the reducedmodulus was calculated. Higher values of reducedmodulus indicate a higher cross-link density for thecured films. The reduced modulus values increasednearly 45% for the samples exposed at doses from100 mJ/cm2 to 200 mJ/cm2, as shown in Fig. 7a. Theincrease in modulus is attributed to a higher cross-link density due to UV exposure, all other variablesheld constant. At exposure doses above 400 mJ/cm2,the reduced modulus was essentially constant at2.8 GPa (standard deviation 0.02). This suggests

that, at doses above 400 mJ/cm2, the maximumPAG activation was achieved and doses higher than400 mJ/cm2 resulted in no further increase in acidcatalyst concentration. These results are consistentwith the results obtained from the samples cured at200�C for 1 h, as shown in Fig. 7b. The averagemodulus for films cured at 200�C was 2.6 GPa witha range of 3.8% for all values from 200 mJ/cm2 to1575 mJ/cm2. The decrease in modulus for filmscured at 200�C, compared with the values at 160�C,is attributed to the slight degradation of the poly-ether cross-linking linkages between multifunc-tional epoxy additives at 200�C. It should be notedthat, at 200�C, the minimum dose required for PAGactivation to obtain full cross-linking is lower thanthat for samples cured at 160�C. This is due tothermal activation of the PAG.

Thus far, it has been shown that the method ofthe postexposure bake, exposure to the TMAHdeveloper, and concentration of the acid catalystaffect the degree of epoxide ring opening and

Soft Bake

Exposure

Post Exposure Bake

Cure

2nd Cure

3rd Cure

0.2

0.4

0.6

0.8

1.2

1.4

Ab

sorb

ance

(A

.U.)

1

Wavenumbers (cm -1)

700 750 800 850 900 950

Fig. 6. Effect of exposure dose on cross-linking degree of Avatrel 8000P (soft bake: 100�C for 5 min; exposure dose: 450 mJ/cm2; postexposurebake: KBr disc inverted onto the hotplate, 100�C for 5 min; cure at 160�C for 1 h).

0.2

0.4

0.6

0.8

1

1.2

1.4

Ab

sorb

ance

(A

.U.)

Wavenumbers (cm -1)

Soft Bake

Equivalent to Post Exposure Bake

Cure

700 750 800 850 900 950

Fig. 5. Effect of thermal cure on cross-linking of unexposed Avatrel 8000P (soft bake: 100�C for 5 min; postexposure bake: KBr disc invertedonto the hotplate, 100�C for 5 min; cure at 160�C for 1 h).


cross-linking. FT-IR spectra were taken after eachstep for an Avatrel 8000P sample exposed to a450 mJ/cm2 (365 nm) UV dose, postexposure bakedon an inverted KBr disk on a hotplate, and cured at160�C for 1 h after development for 3 min. As shownin Fig. 8, full epoxide ring opening was achievedafter cure. The samples cured for extended periodsof time at 160�C showed no detectable change in theheight of the epoxy peaks in the IR spectra. Thisconfirms that the epoxy ring-opening reaction wascomplete after curing at 160�C for 1 h. Comparedwith Fig. 2, which showed a small change in epoxidering opening (normalized absorbance 0.18) after thefirst cure at 160�C for 1 h, Fig. 8 shows a fully curedpolymer (normalized absorbance 0.11) after the first1 h cure.

Lowering the cure temperature is important insome packaging applications due to the presence oftemperature-sensitive materials and as a way tolower the overall thermal profile of the process. Toinvestigate the lowest cure temperature to achievefull cross-linking of Avatrel 8000P after 1 h cure,samples cured at 140�C and 150�C for 1 h wereprepared. All other processing variables were held

constant. The FT-IR analysis of the samples curedat 140�C for 1 h showed incomplete epoxy ringopening, suggesting a reduced degree of cross-link-ing. Thus, a higher cure temperature or longer curetime at 140�C is required for full curing of Avatrel8000P. A sample cured at 150�C (Fig. 9) showedthat very little epoxy remained after being cured for1 h, suggesting that a 150�C cure for 1 h may too beadequate. There was only a slight change in theepoxy peak after the second cure at 150�C.

The extent of cross-linking plays an importantrole in establishing the properties of the curedpolymer.23,24 To evaluate the effect of cure temper-ature on the mechanical stability of the cured Ava-trel 8000P, films cured at temperatures between140�C and 240�C were evaluated using nanoinden-tation. Nanoindentation was performed on two tofour samples prepared at each cure temperature.The nanoindentation results were repeated on eachsample two or three times. Table I shows the num-ber of samples for each cure temperature and thenumber of indentation measurements performed oneach sample. It was found that it is especiallyimportant to calibrate the nanoindentation tool

0

0.5

1

1.5

2

2.5

3

Red

uce

d M

od

ulu

s (G

Pa)

0

0.5

1

1.5

2

2.5

3

Red

uce

d M

od

ulu

s (G

Pa)

0 500 1000

Dose (mJ/cm

(a)

(b)

2)

1500

0 500 1000

Dose (mJ/cm2)

1500

2000 1728 (mJ/cm2)

100 (mJ/cm2)

200 (mJ/cm2)

400 (mJ/cm2)

600 (mJ/cm2)

800 (mJ/cm2)

1000 (mJ/cm2)

1200 (mJ/cm2)

1400 (mJ/cm2)

100 (mJ/cm2)

200 (mJ/cm2)

400 (mJ/cm2)

600 (mJ/cm2)

800 (mJ/cm2)

1000 (mJ/cm2)

1200 (mJ/cm2)

1400 (mJ/cm2)

1575 (mJ/cm2)

Fig. 7. Reduced modulus dependence of cured Avatrel 8000P on exposure dose: (a) cure at 160�C for 1 h, and (b) cure at 200�C for 1 h.


2133

before, during, and after a set of experiments tomake sure no drift occurred in the calibration. Ineach set of experiments, multiple calibrations wereperformed to ensure accuracy. In addition, each setof measurements was performed during a single96 h period.

Figure 10a shows the reduced modulus of devel-oped Avatrel 8000P as a function of cure tempera-ture using a 1 h cure. Each data point is the average

of all measurements performed on all samples curedat each temperature. The reduced modulus valueincreased nearly 10% for the samples cured between140�C and 160�C. The modulus reached a maximumvalue of 2.8 GPa. The increase in modulus is likelydue to a higher cross-link density as the cure tem-perature increased. When the cure temperature wasincreased from 160�C to 240�C, the reduced modu-lus decreased from 2.8 GPa to 2.4 GPa due to ahigher degree of degradation of the polyether cross-links between the multifunctional epoxy additivesat the higher cure temperature. The maximumreduced modulus value, 2.8 GPa, was obtained aftercuring at 160�C for 1 h. The same trend wasobserved for the samples that were not exposed tothe TMAH developer (Fig. 10b). The maximumreduced modulus value, 2.8 GPa, was obtained forthe sample cured at 160�C for 1 h. However, thereduced modulus dropped by 15% when the temper-ature was increased from 160�C to 240�C comparedwith a 12% drop for the developed samples. The dif-ference between the developed and nondeveloped

Soft Bake

Exposure

Post Exposure Bake

Develop

Cure

2nd Cure

3rd Cure

Wavenumbers (cm -1)

700 750 800 850 9000.2

0.4

0.6

0.8

1

1.2

1.4

Ab

sorb

ance

(A

.U.)

Fig. 8. Changes in spectrum of Avatrel 8000P from 700 cm�1 to 950�1 as a function of processing conditions (soft bake: 100�C for 5 min;exposure dose: 450 mJ/cm2; postexposure bake: KBr disc inverted onto the hotplate, 100�C for 5 min; develop for 3 min; cure at 160�C for 1 h).

0.2

0.4

0.6

0.8

1

1.2

1.4

Ab

sorb

ance

(A

.U.)

Soft Bake

Post Exposure Bake

Develop

Cure

2nd Cure

Wavenumbers (cm -1)

700 750 800 850 900 950

Fig. 9. Changes in spectrum of Avatrel 8000P from 700 cm�1 to 950�1 as a function of processing conditions (soft bake: 100�C for 5 min;exposure dose: 450 mJ/cm2; postexposure bake: KBr disc inverted onto the hotplate, 100�C for 5 min; develop for 3 min; cure at 150�C for 1 h).

Table I. Summary of nanoindentation samples

CureTemperature (�C)

No. ofSamples

No. ofMeasurements

per Sample

140 2 3160 4 3180 3 2200 4 2220 4 3240 2 2


samples could be linked to the contribution ofthe TMAH developer to curing, as noted above. Thehardness values showed the same trend, and thehighest hardness value, 0.13 GPa, was achievedafter cure at 160�C for both developed and nondev-eloped samples (results not shown).

Swelling experiments were performed to investi-gate the degree of Avatrel 8000P cross-linking. Thesolvent swelling in a polymer film is inversely

related to the degree of cross-linking between poly-mer chains, other factors remaining constant.23,37–39

Polymer films (25 lm thick) made by spin-coating onsilicon wafers were UV-exposed at 450 mJ/cm2 andcured for 1 h. Swelling experiments were performedin PGMEA. The mass of each sample was measuredat seven different times during the 25 h swellingperiod. The percentage increase in weight was cal-culated from Eq. (9). Each data point is the averageof four measurements. The average standard devi-ation was less than 0.0003.

S ¼Wt �W0

W0; (9)

where S is the swelling, Wt is the weight of thesample swollen with solvent at time t, and W0 is thesample weight in the dry state. Figure 11 shows anincrease in weight with swelling time. The swellingratio was less for samples cured at 160�C comparedwith those cured at 140�C. This shows that curing at160�C for 1 h leads to higher cross-link density thancuring at 140�C. Raising the cure temperature from160�C to 240�C increased the swelling from 3.1% to4.1%. The decrease in the cross-link density, as no-ted above, for films cured at higher temperature islikely due to degradation of polyether cross-linksbetween the multifunctional epoxy additives athigher cure temperatures. These data are consistentwith the hardness and reduced modulus results.

The cure temperature and quality of the three-dimensional cross-link network can have a dramaticeffect on the residual stress of the polymer film.Residual stress in a polymer-coated substrate isgenerated as a result of solvent evaporation,shrinkage during the cross-linking of the polymerfilm, and mismatch of the thermal expansion coef-ficient between the polymer film and the substrate.The magnitude of the film stress is related to thecoefficient of thermal expansion (CTE), Poisson’sratio, biaxial modulus of the polymer film, andstress neutral temperature of the polymer film, all

(a)

(b)

2

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3

130 150 170 190 210 230 250

Red

uce

d M

od

ulu

s (G

Pa)

2

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3

Red

uce

d M

od

ulu

s (G

Pa)

Cure Temperature (°C)

130 150 170 190 210 230 250


Fig. 10. Reduced modulus of Avatrel 8000P as a function of curetemperature from 140�C to 240�C: (a) developed for 3 min, and (b)not developed.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 5 10 15 20 25 30

Sw

ellin

g R

atio

(%

)

Time (h)

Cure @ 140°C

Cure @ 160°C

Cure @ 180°C

Cure @ 200°C

Cure @ 220°C

Cure @ 240°C

Fig. 11. Effect of cure temperature on swelling ratio of Avatrel 8000P.


2135

of which can change with cure temperature, espe-cially the stress neutral temperature.40

To evaluate the effect of cure temperature, theresidual stress of Avatrel 8000P films was charac-terized for samples after each process step and curetemperature. The measurements were repeated onfour different samples (10 measurements each) foreach processing step, and the standard deviation forall measurements was 0.78. As shown in Fig. 12, theresidual stress of a 45-lm-thick Avatrel 8000P filmsranged from 13 MPa after soft-bake to 20 MPa aftercure at 160�C for 1 h. The creation of residual stressin the soft-baked film was primarily because of sol-vent removal and subsequent shrinkage of thepolymer film. The CTE mismatch during the ther-mal process also contributes to the stress build-up inthe soft-baked film. After postexposure bake, themodulus of the polymer film increased due to theacid-catalyzed cross-linking within the film, result-ing in an increase in stress of the postexposurebaked film. The decrease in the Avatrel 8000P filmstress after developing can be the result of filmsoftening and swelling. The high residual stress of20 MPa after cure was due to the high cross-linkdensity of the polymer and the CTE mismatch.

The residual stress of Avatrel 8000P was inves-tigated for cure temperatures ranging from 140�C to220�C. The measurements were performed on twodifferent samples at each cure temperature, and 10values were taken on each sample. The results arepresented in Fig. 13, where each data point is theaverage of all measurements performed on allsamples at each cure temperature. The residualstress of the polymer ranged from 18 MPa to25 MPa with increasing cure temperature (each for1 h), as shown in Fig. 13. The increase of residualstress values for samples cured from 140�C to 160�Cis likely due to higher cross-link density andincreased rigidity of the polymer. The drop in thefilm stress for the film cured at 180�C, comparedwith those cured at slightly lower temperatures, canbe attributed to degradation of polymer at high curetemperature. The increase in stress at even highertemperatures may be due to a higher stress neutral

temperature, although the trend is not as obvious asfor the other physical properties discussed above.Overall, the film stress increased about 30% fromthe lowest to highest value for cures from 180�C to220�C. The individual contributions of the differentfactors (i.e., stress neutral point, CTE mismatch,modulus of the film, and rate of the degradationreaction) were not resolved.

The extent of cross-linking can change the dielec-tric properties of Avatrel 8000P due to the polariz-ability of the structures and products. The impact ofcure temperature on the dielectric constant ofpolymer films cured at temperatures from 160�C to220�C was evaluated. Parallel-plate capacitors werefabricated, as described in the ‘‘Experimental Pro-cedures’’ section. Measurements were repeated fourtimes for each cure temperature. The standarddeviation for all cure temperatures was 0.072. Thedielectric constant values for the samples curedfrom 160�C to 220�C are shown in Fig. 14. Thedielectric constant of the cured Avatrel 8000Pincreased from 3.9 to 4.4 with increasing cure tem-perature. The increase in the dielectric constantwith cure temperature is due to the higher polariz-ability of the structures in the cured film at eachtemperature; for example, the degradation reactionat high cure temperature could result in C=O bondproduction with a dipole moment of 2.3, which ishigher than the dipole moment of the C–O (0.7)bonds.41 As a result, degradation of the fully cross-linked polymer results in higher electron and

0

5

10

15

20

25

Soft Bake Post Exposure Bake Develop Cure @ 160°C

Res

idu

al S

tres

s (M

Pa)

Processing Condition

Fig. 12. Changes in internal film stress of Avatrel 8000P as afunction of processing conditions (soft bake: 100�C for 5 min; post-exposure bake: 100�C for 5 min; develop for 3 min; cure at 160�C for1 h).

0

5

10

15

20

25

30

130 150 170 190 210 230

Res

idu

al S

tres

s (M

Pa)

Temperature (°C)

Fig. 13. Residual stress of Avatrel 8000P as a function of curetemperature from 140�C to 240�C.

3

4

5

140 160 180 200 220 240

Die

lect

ric

con

stan

t (ε

r)


Fig. 14. Dielectric constant of Avatrel 8000P at cure temperatures of160�C, 180�C, 200�C, and 220�C.


induced dipole polarization. In summary, curingAvatrel 8000P at 160�C for 1 h resulted in thelowest relative dielectric constant, er = 3.9. Thiscure temperature agrees well with the cross-linkingand modulus results shown above.

DISCUSSION

The cure temperature for optimal mechanical andelectrical properties of Avatrel 8000P was experi-mentally identified. The optimal cure temperatureis a compromise between formation of cross-linkswithin the polymer and degradation of structures.The highest cross-link density and optimum prop-erties were achieved for samples exposed to 365-nmradiation, exposed to TMAH developer, and cured atthe relatively low temperature of 160�C for 1 h.Several important results were found. The optimalcure temperature for this polymer was found to belower than for other PNB-based systems studiedpreviously.9,17,18,25 The addition of multifunctionalcross-linkers to this polymer mixture results inhigher mobility and easier access of the cross-linkable sites to each other, compared with epoxy-functionalized PNB polymers (i.e., PNB withpendant epoxy moieties). The high cross-link den-sity of Avatrel 8000P at low cure temperatures isalso attributed to the efficient activation of themultifunctional epoxy additive. As discussed ear-lier, cross-linking can be achieved by both photo andthermally activation of the PAG. However, as shownin Fig. 5, if one were to rely only upon thermalactivation of the PAG to initiate cross-linking of theepoxy compounds, a higher temperature and/orlonger curing cycle would be required to achieve fullepoxy ring opening and cross-linking within thepolymer.

It was also found that thermal activation of theacid catalyst during postexposure bake, whichassists in development of the latent image, alsoassists in achieving full epoxy ring opening duringthermal curing at low temperature, as shown inFig. 3. The extent of photocross-linking can beimproved by excess UV exposure, resulting in ahigher degree of PAG activation and greater epoxyconversion (Fig. 6).

The increase in the rate of the epoxy ring-openingreaction as a result of TMAH exposure prior to cur-ing is an unexpected benefit of developing in aqueousbase. This could be due to an increase in mobility ofthe multifunctional cross-linkers, or catalysis of thecross-linking reaction due to absorbed TMAH,resulting in improved network interconnectivity.This corroborates with the reduced modulus mea-surements presented in Fig. 10. The reduced mod-ulus of the developed samples decreased less (12%)when the cure temperature increased from 180�C to240�C compared with the reduced modulus of theundeveloped sample, which decreased 15%. Theaction of the developer also results in a lower resid-ual stress value for the developed sample, as shown

in Fig. 12, possibly due to plasticization of film bythe TMAH. These factors contribute to full cross-linking of the polymer at a relatively low cure tem-perature of 160�C, resulting in the highest measuredhardness and modulus, and lowest dielectric con-stant and residual stress.

CONCLUSIONS

The acid-catalyzed epoxy ring-opening reaction ofa photodefinable polynorbornene-based dielectricformulation was studied using FT-IR spectroscopy.The photo and thermal acid generation initiatesepoxy ring-opening reactions and subsequent cross-linking of polymer. This reaction is limited by thediffusive mobility of reactive groups (epoxy ringopening and acid catalyst). The rate and ultimateconversion of the epoxy ring-opening reactionincrease with increasing cure temperature, result-ing in a higher degree of cross-linking at cure tem-peratures above 140�C. Degradation reactions likelyoccur at temperatures above 160�C, indicating lossof epoxide cross-linking groups and linkages. Thesehypotheses were supported by electrical andmechanical property studies. It was shown thatcuring Avatrel 8000P at 160�C for 1 h after developresulted in full epoxy ring opening and highestcross-link density. This sample showed lowerdielectric constant (3.9), residual stress (20 MPa),and solvent swelling (3.1%).

ACKNOWLEDGEMENTS

The authors gratefully acknowledge the intellec-tual and material contribution of Ed Elce and BrianKnapp of Promerus LLC.

REFERENCES

1. Y.Q. Bai, P. Chiniwalla, E. Elce, R.A. Shick, J. Sperk, S.A.B.Allen, and P.A. Kohl, J. Appl. Polym. Sci. 91, 3023–3030(2004).

2. R. Tummala, E. Rytmaszewski, and A. Klopfenstein,Microelectronics Packaging Handbook: Technology Drivers(Boston, MA: Kluwer Academic, 1997).

3. P. Singer, Semicond. Int. 19, 88 (1996).4. L. Peters, Industry Divides on Low-k Dielectric Choices.

Semiconductor International, May (2001).5. G. Maier, Prog. Polym. Sci. 26, 3–65 (2001).6. K. Maex, M. Baklanov, D. Shamiryan, S. Brongersma, and

Z. Yanovitskaya, J. Appl. Phys. 93, 8793 (2003).7. N. Grove, P. Kohl, S. Allen, S. Jayaraman, and R. Shick,

J. Polym. Sci. B: Polym. Phys. 37, 3003–3010 (1999).8. N. Grove, P. Kohl, S. Bidstrup-Allen, R. Shick, B. Goodall,

and S. Jayaraman, Polynorbornene for Low K Interconnec-tion (Materials Research Society, 1997), p. 3.

9. V. Rajarathinam, C.H. Lightsey, T. Osborn, B. Knapp, E.Elce, S.A.B. Allen, and P.A. Kohl, J. Electron. Mater. 38,778–786 (2009).

10. M. Raeis Zadeh, E. Elce, B. Knapp, and P.A. Kohl, J. Appl.Polym. Sci. 120, 1916–1925 (2011).

11. P. Plesch, The Chemistry of Cationic Polymerization(New York: Pergamon, 1963).

12. J. Fouassier and J. Rabek, Radiation Curing in PolymerScience and Technology: Fundamentals and Methods(New York: Springer, 1993).

13. L. Matejka, K. Dusek, and I. Dobas, Polym. Bull. 14, 309–315 (1985).

14. C. Decker, Prog. Polym. Sci. 21, 593–650 (1996).


2137

15. G. Cai and W.P. Weber, Polymer 45, 2941–2948 (2004).16. M. Moniruzzaman, F. Du, N. Romero, and K.I. Winey,

Polymer 47, 293–298 (2006).17. P. Chiniwalla, Y. Bai, E. Elce, R. Shick, W. McDougall, S.

Allen, and P. Kohl, J. Appl. Polym. Sci. 89, 568–577(2003).

18. P. Chiniwalla, Y. Bai, R. Elce, S. Allen, and P. Kohl, J. Appl.Polym. Sci. 91, 1020–1029 (2003).

19. Y.Q. Bai, P. Chiniwalla, E. Elce, S.A.B. Allen, and P.A. Kohl,J. Appl. Polym. Sci. 91, 3031–3039 (2004).

20. J. Crivello, T. Lockhart, and J. Lee, J. Polym. Sci. Polym.Chem. Ed. 21, 97–109 (1983).

21. H.S. Yu, T. Yamashita, and K. Horie, Macromolecules 29,1144–1150 (1996).

22. B. Plage and H. Schulten, Macromolecules 21, 2018–2027(1988).

23. M. Soh and A. Yap, J. Dent. 32, 321–326 (2004).24. J. Mark, Rubber Chem. Technol. 55, 762–779 (1982).25. M. Raeis Zadeh, E. Elce, B. Knapp, and P. Kohl, J. Appl.

Polym. Sci. 120, 1916–1925 (2011).26. ASTM Materials, Standard Test Methods for AC Loss

Characteristics and Permittivity (Dielectric Constant) ofSolid Electrical Insulating Materials (1992).

27. Thin Films Stress Measurement System, OperationalManual, Model F2320 (San Jose, CA: Flexus Inc.).

28. C.A. Klein, J. Appl. Phys. 88, 5487–5489 (2000).

29. G. Shaw, J. Trethewey, A. Johnson, W. Drugan, and W.Crone, Adv. Mater. 17, 1123–1127 (2005).

30. J. Loubet, J. Georges, O. Marchesini, and G. Meille, J.Tribol. 106, 43 (1984).

31. B.J. Briscoe, L. Fiori, and E. Pelillo, J. Phys. D Appl. Phys.31, 2395–2405 (1998).

32. M. VanLandingham, J. Villarrubia, W. Guthrie, and G.Meyers, Nanoindentation of Polymers: An Overview(New York: Wiley, 2001), pp. 15–44.

33. L. Shen, I.Y. Phang, T.X. Liu, and K.Y. Zeng, Polymer 45,8221–8229 (2004).

34. W. Oliver and G. Pharr, J. Mater. Res. 7, 1564–1583 (1992).35. E. Wornyo, K. Gall, F. Yang, and W. King, Polymer 48,

3213–3225 (2007).36. P. Chiniwalla, Chemical Engineering (Georgia Institue of

Technology, 2001).37. Y. Zang, R. Muller, and D. Froelich, Polymer 30, 2060–2062

(1989).38. P. Flory and Y. Tatara, J. Polym. Sci. Polym. Phys. Ed. 13,

683–702 (1975).39. S. Rosen, Fundamental Principles of Polymeric Materials

(New York: Wiley, 1982).40. J. Crivello and J. Lee, J. Polym. Sci. Polym. Chem. Ed. 21,

1097–1110 (1983).41. J. Brydson, Plastic Materials (Woburn: Reed International

Books, 1989).


aqueous-develop, photosensitive polynorbornene dielectric...

Documents