175 Additive Behavior During Copper Electrodeposition

Journal of The Electrochemical Society, 150 ~6! C420-C425~2003!0013-4651/2003/150~6!/C420/6/$7.00 © The Electrochemical Society, Inc.

C420

Additive Behavior during Copper Electrodepositionin Solutions Containing ClÀ, PEG, and SPSMin Tan * and John N. Harb** ,z

Department of Chemical Engineering, Brigham Young University, Provo, Utah 84602, USA

This paper documents an experimental investigation of additive behavior in acidic copper plating solutions containing chlorideions, polyethylene glycol~PEG!, and bis-~3-sodiumsulfopropyl disulfide! ~SPS!. These solutions represent a simplified model ofsolutions used industrially for electroplating copper interconnects. Experiments were conducted with use of a rotating diskelectrode under both galvanostatic and potentiostatic conditions and the transient behavior was observed. Linear sweep voltam-metry was also performed. A transition time for the system to reach steady state was observed under both galvanostatic andpotentiostatic conditions and found to be a strong function of SPS concentration. Superfilling is made possible by the existence ofa long transition time that permits establishment of a significant rate difference between the top and bottom of the cavity.Experimental results provide evidence for slow adsorption and desorption of the accelerator and SPS incorporation into thedeposit. Linear sweep voltammetry indicates that behavior of the accelerator is potential dependent. The same mechanism used toexplain superfilling can be applied to explain the leveling/brightening of copper surfaces in these solutions.© 2003 The Electrochemical Society.@DOI: 10.1149/1.1570412# All rights reserved.

Manuscript submitted June 15, 2002; revised manuscript received December 16, 2002. Available electronically April 21, 2003.

r-

utd

-r

l

n

dd

e

i

d

t

la

e

a-

n

he

nthe

ace

c-

Ler

A

s.

as-

o-0,

cesors-

f-

Electrodeposition is used in the fabrication of copper on-chinterconnects to completely fill deep vias and trenches in ordeguarantee good quality interconnects.1 Deposition that meets the requirements for filling high-aspect-ratio~AR, depth-to-width ratio!structures without void formation is called superfilling. During sperfilling, the deposition rate at the bottom of the feature is fasthan that at the top. Various additive systems have been used intrially to achieve superfilling. Of interest to the present study ismodel system consisting of three additives@chloride ions, polyeth-ylene glycol~PEG!, and bis-~3-sodiumsulfopropyl disulfide! ~SPS!#that has been shown to have superfilling properties similar to soof the more complex industrial additive systems.2 Both this modelsystem and a similar system that uses 3-mercaptopropanesulfonate~MPSA! instead of SPS have been used in severecent studies.3,4 Mathematical models have been developed to dscribe the additive behavior and the evolution of surface profiduring superfilling.2,5,6 These models have included a level of empiricism and have utilized experimentally fitted parameters duethe lack of fundamental knowledge regarding additive interactioadsorption, desorption, and incorporation. A more fundamenmodel was developed recently by Hebert7 to describe the current-voltage behavior during linear sweep voltammetry~LSV! of themodel system containing MPSA. However, his study did not incluthe experimental data necessary to validate key assumptions annot address the issue of superfilling. The purpose of the presexperimental study is to investigate the mechanisms of additiveteractions and behavior in the model system (Cl2, PEG, and SPS! inorder to provide the basis for the development/validation of a fudamental model to describe superfilling.

Inhibition in the Cl2-PEG-SPS system is due primarily to thcombination of PEG and chloride ions.8-11 PEG inhibition has beenstudied extensively and is generally attributed to active site blockby PEG directly or by a PEG-containing complex.9 The markedenhancement of the inhibition by the involvement of the chloriions in the solution has also been documented.8-11 Chloride ionsadsorb at the cathode surface in the potential range of interescopper electroplating12-15 and provide strong binding sites for PEGadsorption.8 Acceleration~depolarization! is much less understoodthan inhibition but has been attributed to SPS or DTODSA~4,5-dithiaoctane-1,8-disulfonic acid, a chemical with the molecustructure close to that of SPS!.16-19 A possible explanation for thedepolarization effect is that the accelerator~SPS or derivative

* Electrochemical Society Student Member.** Electrochemical Society Active Member.

z E-mail: [email protected]

Downloaded 21 Jul 2011 to 152.74.222.130. Redistribution subject to

ipto

-erus-a

me

1-ale-es-tos,

tal

edid

entin-

n-

ng

e

to

r

thereof! changes the reaction mechanism by involvement in thelectron-transfer process.20,21Another possibility is that the numberof nucleation sites for copper deposition is increased by the formtion of a sulfide that adsorbs at the surface.19 MPSA and SPS behavesimilarly during copper plating from acid solutions and reach aequilibrium concentration when both additives are in solution.20,22

An interesting phenomenon that has been observed with tmodel additive system, as well as with similar industrial platingbaths, is ‘‘inertial bumping’’ or the formation of bumps due to over-filling of the trenches.2-4,23A simple mechanism related to transportlimitations in the cavity does not explain the continued acceleratioafter the cavity has been filled. Rather, mechanisms presented inliterature include two parts: (i ) preferential concentration of the ac-celerator~SPS or MSPA! in the cavity and (i i ) slow desorption ofthe accelerator. Studies by Westet al. and Moffat et al. attributedthe concentration of the accelerator to the decrease in the surfarea that occurs as a natural consequence of filling the cavity.2,3,5,6

An alternate explanation, offered by Richardet al., was that mass-transport limitations of the inhibitor during the initial plating periodallowed preferential adsorption of the accelerator, catalyzing the acumulation process of accelerator in the cavity.23

Experimental

Experiments were performed in a glass vessel containing 100 mof electrolyte solution. Electrodes consisted of a platinum/copprotating disk working electrode~5 mm diam!, a copper counter elec-trode~Cu 99.99%!, and a copper~99.999%! reference electrode. Thereference electrode was placed far from the working electrode.potentiostat/galvanostat~model 263A, EG&G Princeton AppliedResearch! controlled by a PC was used to perform the experimentThe composition of the basic electrolyte was 0.24 M CuSO4

• 5H2O and 1.8 M H2SO4 , to which one or more of the threeadditives PEG@molecular weight (Mw)5 3400 g/mol, Aldrich#,Cl2 ~added as NaCl, Mallinckrodt!, and/or SPS~Raschig, Germany!were added. At the beginning of each experiment a copper layer wdeposited on the platinum disk in the basic electrolyte without additives at 10 mA/cm2 for 15 min. The electrolyte temperature waskept constant at 2986 1 K. The rotation rate was 400 rpm for allexperiments.

During polarization experiments, copper was deposited galvanstatically onto the working electrode at current densities of 5, 1and 20 mA/cm2. A variety of different additive combinations wereinvestigated in order to elucidate additive interactions. The sequenin which the additives were added to the basic electrolyte was alvaried. In some experiments electrodes were pretreated by immeing them in the solution for a specified time prior to initiation ofelectrodeposition and/or post-treated by rinsing in a variety of di

ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp

pg

ep

o

geo

ng

eio

sn

g

ioece

etee

eebed

lo

aveto

meluemedgedn-ig.ion,Innge 3dentin-

ady-the

statesity

di-s todi-ateon-hen

ged

ingtor.

tionn

atPEGing

si

Journal of The Electrochemical Society, 150 ~6! C420-C425~2003! C421

ferent ways. Deposition experiments were also performed undertentiostatic conditions. LSV was performed by sweeping in a netive direction from the rest potential to2400 mV ~vs. copper! atrates of 0.2, 1, and 5 mV/s. The SPS concentration used in thexperiments was 1 ppm, and several sequential cycles wereformed for each set of experimental conditions.

The change in the surface morphology during galvanostatic cper deposition at a current density of 20 mA/cm2 was studied byatomic force microscopy~AFM!. This work was performed on cop-per disks~99.998%! which were polished in a series of steps endinwith a 0.05mm alumina paste and rinsed thoroughly in deionizwater prior to each copper deposition. The surface resulting frdeposition in the basic electrolyte with PEG and Cl2 was used as thebase case. Experiments examined the influence of SPS concetion, plating time, and pretreatment time on the surface morpholo

Results and Discussion

Behavior of three-additive system: SPS-PEG-Cl2.—As previ-ously mentioned, the behavior of the three additive syst(SPS-PEG-Cl2) was investigated under galvanostatic and potentstatic conditions. Curve ‘‘Cl2 1 PEG1 SPS’’ in Fig. 1 shows thedeposition behavior observed when all three additives were preat the beginning of a galvanostatic experiment at the following cocentrations: 1 ppm SPS, 300 ppm PEG, and 50 ppm Cl2. Underthese conditions, the potential increased gradually, correspondina decrease in the overpotential or polarization necessary to drivecurrent, until it reached a steady value. The transition was quslow, requiring on the order of 2500 s to complete. This typebehavior was not observed in the plating bath with two additiv~Cl-PEG or Cl-SPS!, suggesting that it was the result of the interation of all three additives. In order to gain additional insight into thmechanistic behavior of the system, two additional types of expments were performed. In the first of these, deposition was initiawith SPS and chloride only, followed by the addition of PEG. Thresults ~curve @Cl2 1 SPS# 1 PEG in Fig. 1! show a slight de-crease in the potential at the beginning of the experiment, followby a plateau as the SPS-Cl2 system reached steady state. Subsequaddition of PEG caused a very rapid drop in potential, followeda gradual increase similar to that observed in the previous expment. Note that the time scale for this experiment has been shifteorder to superimpose the transition periods. In the second typeadditional experiment, deposition was initiated with PEG and chride only, followed by the addition of SPS~curve @Cl2 1 PEG#1 SPS in Fig. 1!. In this case, the potential was nearly constant

a low value until the addition of SPS. The same transition obserin the previous two experiments was also observed in this expment after SPS addition~again, the time scale has been shiftedsuperimpose the transition periods!.

Figure 1. Polarization curves of galvanostatic experiments, current den10 mA/cm2.


o-a-

seer-

p-

dm

tra-y.

m-

ent-

tothetefs-

ri-d

dntyri-inof-

tedri-

The transition period can be characterized by the transition ti~defined as the time required to reach 70% of the steady-state va!and the steady-state value of the potential. The system was assuto be at steady state when the average value of the potential chanby less than 1 mV in a 10 min period. The influence of SPS cocentration and current density on the transition time is shown in F2. The transition time was strongly dependent on SPS concentratwith low concentrations corresponding to long transition times.addition, the transition time decreased significantly with increasicurrent density at SPS concentrations less than 1 ppm. Figurshows that the steady-state value of the potential was also depenon the SPS concentration. For each current density tested, ancrease in the SPS concentration caused an increase in the stestate potential, corresponding to a decrease in the magnitude ofoverpotential needed to sustain the deposition rate. The steady-value of the potential also decreased with increasing current denas expected.

Similar experiments were performed under potentiostatic contions and a slow transition in the current was observed, analogouthe transition in the potential observed under galvanostatic contions. Table I shows that both the transition time and steady-stcurrent were strongly dependent on the SPS concentration at cstant potential. The system was assumed to be at steady state wthe average value of the current reached its maximum and chanby less than 0.05 mA in a 5 minperiod. An important observationwas that the current approached a constant value with increasSPS concentration, presumably due to saturation of the accelera

As SPS provides the depolarization effect, a possible explanafor the slow transition is that it is the result of a slow accumulatioof accelerator at the surface. It would then follow from Fig. 1 ththe accelerator accumulates on the surface in the presence ofand chloride ions. In contrast, the rapid drop observed after add

tyFigure 2. Effect of SPS concentration on transition time.

Figure 3. Effect of SPS concentration on steady-state potential.


h

s

m

at

eise

it

ti

a

nfio

th

a

c

du

tha

h

e

u-anngepperat

.andthe

In

p-

al-

Journal of The Electrochemical Society, 150 ~6! C420-C425~2003!C422

PEG to a solution containing SPS and chloride ions~Fig. 1, curve@Cl2 1 SPS# 1 PEG) indicates that relatively little accelerator haaccumulated on the surface in the presence of SPS and chlorideprior to the addition of PEG, or that PEG rapidly displaced taccelerator when added to solution. In either case, it is difficultimagine that the subsequent accumulation of the accelerator waresult of a simple competitive adsorption mechanism. The depdence of the steady-state potential~Fig. 3! on SPS concentrationsuggests a concentration-dependent balance between an accution process and a consumption process. Otherwise, accumulatiothe SPS might be expected to continue until the surface is saturyielding a steady-state potential that is independent of concentra

Morphology associated with transition curves.—AFM was usedto examine the morphology of the copper surface in the differplating solutions. Of particular interest was whether or not therecausal relationship between the changes in morphology and dlarization of the electrode during galvanostatic deposition. It wobserved that plating baths with low concentrations of SPS~0.2 and1 ppm! yielded a smooth bright surface, while plating baths whigher concentrations of SPS~5, 10, and 50 ppm! yielded a roughdull surface. Figure 4 shows the copper surface evolution inplating process at SPS concentrations of 0.2 and 5 ppm. The insurface in both samples was a rough surface~sufficiently rough tomake AFM difficult! that resulted from electrodeposition in the bsic electrolyte solution with PEG and chloride ions~no SPS! at aconstant current density of 20 mA/cm2. The difference in the sur-face morphology between the starting and ending points was sigcant for both samples. However, the average feature size of thesample was much smaller for deposition in the 0.2 ppm SPS stion. Note that no significant change in surface roughness wasserved for a given sample once steady state had been achieveother words, the extent of leveling was indeed much less insolution containing 5 ppm SPS than in the 0.2 ppm solution. Higconcentrations of SPS resulted in surfaces of considerably greroughness, comparable to the initial roughness shown in Fig. 4,thus were difficult to characterize with AFM.

A galvanostatic experiment was performed at 20 mA/cm2 withthe basic electrolyte containing 50 ppm Cl2, 300 ppm PEG, and 1ppm SPS to determine the impact of the morphology change ontransition time required for depolarization. No significant differenin the transition time was observed between experiments initiawith a smooth and with a rough surface, indicating that morpholowas not the dominant factor in the depolarization process. In ation, the roughness change during the transition period was mmore significant for the sample plated in 0.2 ppm SPS thanplated in 5 ppm SPS, while the extent of depolarization was grefor the 5 ppm solution. This again indicates that the morpholochange was not the dominant cause of the depolarization.

Accumulation and consumption of the accelerator at tsurface.—Additional sets of experiments were performed to gainsight into the factors that control the accumulation and consumtion of the accelerator on the electrode surface. The first of thexamined deposition behavior with SPS and chloride~i.e., in the

Table I. Effect of SPS concentration on transition behavior.a

SPS concentration~ppm!

Transition time~s!

Steady-state current~mA!

0.05 3413 0.730.5 2630 3.41 1600 6.45 1164 9.1

50 201 10.0500 45 10.5

1000 10 10.7

a Constant potential20.2 V, contact resistance 3V.


dionsetothe

en-

ula-n ofted,ion.

nta

po-as

h

hetial

-

ifi-nallu-ob-d. Inheeraternd

thee

tedgydi-chatter

gy

einp-se

absence of PEG!. At this point it was postulated that if the accumlation of the accelerator at a low concentration is slow, thenextended exposure should allow additional adsorption and chathe observed deposition behavior. To test this hypothesis, the cocathode was pretreated by immersing it in the plating solutionzero current for 1 h prior to initiation of galvanostatic depositionFigure 5 shows that the behavior observed for deposition withwithout pretreatment was significantly different. The potential ofpretreated sample started at a ‘‘high’’ value~low overpotential! anddecreased with time, presumably due to consumption~desorptionor/and incorporation! of the accelerator in the presence of current.

Figure 4. Ex situAFM images of the surface morphology during RDE coper electrodeposition with 50 ppm Cl2, 300 ppm PEG, and SPS.

Figure 5. Influence of pretreatment on the potential behavior during gvanostatic deposition (10 mA/cm2) in electrolyte containing SPS and Cl2.


lp

i

o-n.ho

npou

ho

dac

hawo

llysti

ntiw

t wc

imwF

itis

trsura

ted

lowtomsel-in

tion

e

PSbe-dhede-

ri-

d in

vi-ple.A

wase-

beis

edrthee-at

es.canig-lso

ans.f 5lts

a

g


contrast, the sample that was not pretreated started at a potentiawas 25 mV lower than that of the pretreated sample. Both samreached the same steady-state potential.

Figure 6 shows the results of similar experiments performed wthe three-additive system containing 50 ppm Cl2, 300 ppm PEG,and 1 ppm SPS. For short pretreatment times~10 min! or no pre-treatment, the potential started low~i.e., high overpotential! and in-creased gradually, similar to the behavior shown in Fig. 1. In ctrast, the sample pretreated for 1 h went immediately to the steadystate value and remained there. The potential of these samplesreached the low values experienced by the untreated samplesparently, the accelerator was adsorbed during the pretreatment pand remained, at least partially, on the surface. Note that this adstion occurred in the presence of PEG.

These experimental results are consistent with slow adsorptiothe accelerator and incorporation of the accelerator into the deduring electrodeposition. Incorporation into the deposit is also sported by impurity measurements in copper deposits which sevidence of the sulfur-containing additive~accelerator!.24

Additional experiments were performed to characterize thesorption behavior of the accelerator. A copper substrate was plin the three-additive solution~50 ppm Cl2, 300 ppm PEG, 1 ppmSPS! at open circuit for 1 h to provide time for SPS adsorption. Tsubstrate was then placed in a different solution that was identicthe first except that it did not contain any SPS. The substratekept in a large excess of this SPS-free solution at open-circuit cditions for an additional hour in order to permit desorption. Finathe substrate was transferred to the test cell where copper waposited galvanostatically on the substrate from an SPS-free solucontaining 50 ppm Cl2 and 300 ppm PEG. For comparison, a simlar copper substrate was pretreated in the three-additive solutionthen moved to an SPS-free solution for immediate plating withtime for desorption. Figure 7 shows the results of these two plaexperiments, along with those from a fresh copper substrate thatplated in an SPS-free solution without pretreatment. Depositionthe sample that was allowed to desorb began at a potential thain between that of the other two samples. This leads to the consion that desorption did occur but was incomplete in the 1 h tallotted. Thus, it appears that desorption does take place sloOther observations support the conclusion of slow desorption.example, samples deposited in the presence of all three addshowed residual depolarization effects, even after they were rinin water or additive-free electrolyte solution. In fact, repeated ulsonic cleaning in an SPS-free solution was required to bring astrate that had been pretreated in 1 ppm SPS back to behaviowas close to that of an untreated substrate. Samples pretreated

Figure 6. Influence of pretreatment on the potential behavior during gvanostatic deposition (10 mA/cm2) in electrolyte containing Cl2, PEG, andSPS.


thatles

th

n-

everAp-aserp-

ofsitp-w

e-ed

el toasn-,de-ion-ando

ngas

onas

lu-ely.orvesed

a-b-

thatt 10

ppm SPS showed residual depolarization effects even after repeaultrasonic cleaning.

Based on these results, desorption is expected to be very srelative to the sharp drop observed in Fig. 1 when PEG was addedthe system containing SPS and chloride ions. Consequently, it seeunlikely that the steep drop was due to displacement of the accerator by PEG. An alternate explanation for the behavior showncurve@Cl2 1 SPS# 1 PEG of Fig. 1 is that the addition of PEG ledto increased adsorption of the accelerator. The increased adsorpmay have resulted from the shift in potential~potential-dependentadsorption! and/or a reduced rate of accelerator incorporation in thpresence of PEG.

In summary, the results presented indicate that at a low Sconcentration, the accelerator adsorbs and desorbs slowly. Thishavior is likely responsible for the long transition times observeand for the persistent acceleration leading to inertial bumping. Tresults also indicate that the accelerator is incorporated into theposit, consistent with the results of previous investigators.7,24

Linear sweep voltammetry.—LSV was also used to examine thedeposition behavior of the three-additive system. Similar expements have been performed by Westet al.2 and Moffatet al.3,5,6andare described theoretically by Hebert.7 Experiments in the presentstudy were performed with 50 ppm Cl2, 300 ppm PEG, and 1 ppmSPS on a rotating disk electrode~RDE, 400 rpm! at scan rates of 0.2,1, and 5 mV/s. Potentials are relative to a Cu21/Cu reference elec-trode. Each scan was initiated at the rest potential and proceedethe cathodic direction to2400 mV before returning to 0 mV. Onedifference between the present study and those mentioned preously is that four scans were recorded in succession for each sam

The results for the three scan rates are shown in Fig. 8a-c.distinct difference between the initial scan and subsequent scansobserved at the high rate as shown in Fig. 8a. This difference bcame less pronounced at the slower scan rates~Fig. 8b and c!. Thedifference between the initial scan and subsequent scans canattributed to the slow accumulation of the accelerator when SPSpresent at a low bulk concentration. The slower scan rates allowadditional time for this accumulation to occur, leading to a smalledifference between the initial scan and subsequent scans. Bytime the initial scan was complete, the sample had finished ‘‘basline’’ adsorption of the additive, resulting in subsequent scans thwere very similar.

A hysteresis loop was observed for all scans at all scan ratThis hysteresis was most pronounced for the initial scan at each srate because of the slow accumulation mentioned previously. A snificant difference between the forward and reverse scans was aobserved at the 1 and 0.2 mV/s scan rates for all subsequent scIn contrast, the degree of hysteresis observed at a scan rate omV/s was small for scans other than the initial scan. These resuare consistent with two types of accumulation: (i ) slow accumula-

l-

Figure 7. Influence of desorption on the potential behavior observed duringalvanostatic copper electrodeposition at 10 mA/cm2.


y

teo

thanof

e to

lesghon

ly aob-rreds areace

n

-

torthe

orc-ov-in-that

e ate

omra-toe-

r-ltsan9,vityetorowuchanbeon

ion

Journal of The Electrochemical Society, 150 ~6! C420-C425~2003!C424

tion to a base level that is manifest in the difference betweeninitial scan and subsequent scans, and (i i ) repetitive accumulationand removal of the accelerator in subsequent scans due to thetential change. The most likely explanation for the reproducible hteresis loops observed for scans other than the initial scan issurface coverage of the accelerator is potential dependent. A swof the potential to more cathodic values allows additional acceletor to accumulate. Because of the finite scan rate and the slow raadsorption, the amount of accelerator on the surface remains bits steady-state value throughout the forward scan. Thus, the am

Figure 8. Results of LSV experiments with electrolyte containing 50 ppCl2, 300 ppm PEG, and 1 ppm SPS at~a, top! 5, ~b, middle! 1, and ~c,bottom! 0.2 mV/s.

Downloaded 21 Jul 2011 to 152.74.222.130. Redistribution subject t

the

po-s-

thateepra-e oflowunt

of accelerator on the surface during the reverse scan is greaterthat in the forward scan at any given potential. The presenceincreased amounts of accelerator during the reverse scan is duaccumulation during the scan, causing the observed hysteresis.

The difference in the extent of hysteresis observed for cycother than the initial scan can be explained as follows. At the hiscan rate of 5 mV/s, the amount of accelerator that accumulatedthe surface was limited due to the fast scan rate. Therefore, onsmall difference between the forward and reverse cycles wasserved. As the scan rate was slowed, more accumulation occuand a greater degree of hysteresis was observed. These resultconsistent with slow adsorption and potential-dependent surfcoverage of the accelerator.

Superfilling of cavities.—Superfilling of small features has beeobserved in systems using a model electrolyte~SPS, Cl2, and PEGor MPSA, Cl2 and PEG! as well as in industrial systems with plating baths of similar composition.2,4,5,23The typical SPS concentra-tion in such systems is in the range 1-10 ppm.25,26 In order to‘‘superfill’’ a cavity without void formation, the electroplating rate athe bottom of the cavity must be higher than that at the top. Fexample, a trench with an aspect ratio of two requires the rate atbottom to be a minimum of about four times that at the top fsuccessful superfilling to occur. The most likely explanation for aceleration inside the cavity is ‘‘curvature-enhanced accelerator cerage,’’ where the surface concentration of the accelerator iscreased due to the decrease of surface area inside the cavityoccurs naturally as the cavity is filled via electrodeposition.2,5 Thesize of the cavity is such that the surface can be assumed to bconstant potential.27 The relationship between the results from thpresent study and superfilling is now examined.

If we assume that the concentration of accelerator at the bottof the cavity approaches saturation to provide maximum acceletion, then the data in Table I provide some interesting insights inthe superfilling process. In particular, the maximum difference btween the current at the bottom~saturated! and top~1 ppm SPS bulkconcentration! at steady state is 1.7, insufficient to result in supefilling. Therefore, superfilling is a transient phenomenon that resufrom the fact that depolarization inside the cavity occurs faster thoutside the cavity. The situation is illustrated conceptually in Fig.where a large difference between the rate at the bottom of the caand the top~or outside! is present at short times relative to thtransition time. It is the large difference between the acceleraconcentration inside and outside of the cavity, combined with a sldesorption rate, that causes the inertial bumping observed for ssystems since the time required to fill the cavity is much less ththe transition time for the electrolyte composition used. It maypossible to minimize inertial bumping by stopping electrodepositiprior to completely filling the cavity and allowing sufficient time~on

m

Figure 9. Conceptual diagram showing the difference between depositrates at the top and bottom of the trench as a function of time.

o ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp

gb

tw

ra

sdelo

fitcve

iob

e

n

t

Ssrp

atortedlysisnsi-ceusedg/

of

R.

, PA

, S.

, PA

M..


the order of the transition time! for the outside of the cavity to‘‘catch up’’ to the bottom before resuming deposition. Use of a hiconcentration of SPS precludes the possibility of superfillinggreatly reducing the transition time~see Table I! and allowing thetop or outside of the cavity to quickly reach the same rate asbottom. In contrast, superfilling will not be observed for very loSPS concentrations~e.g., 0.05 ppm! because the steady-state coveage of the accelerator is so low that ‘‘curvature-enhanced accelecoverage’’ does not provide sufficient acceleration at the bottomthe cavity.3

Relationship to morphological evolution of copper depoits.—The explanation presented for superfilling can also be useexplain the morphological evolution of copper deposits reportedlier in the paper. AFM showed smooth and bright surfaces atSPS concentrations~e.g., 0.2 or 1 ppm! and rough surfaces at highSPS concentrations (.10 ppm). Deposit roughness was also afected by pretreatment in the electrolyte. Long pretreatment timallowed SPS to adsorb to its steady-state value over the entireface, leading to rough deposits. The sample surfaces corresponto the pretreatment times of 1 h were muchrougher than thosecorresponding to the pretreatment times of 0 and 10 min. Whenbulk SPS concentration is low, the accumulation of the acceleratothe surface is a slow process, and the amount of acceleratorsorbed at the cathode surface is small at first. The decrease osurface area in the valleys of the rough surface due to deposleads to accumulation of the accelerator by ‘‘curvature-enhanaccelerator coverage’’ and a faster local deposition rate in theleys. The opposite is true for the peaks of the rough surface. Thfore, valleys have the fastest deposition rate and peaks haveslowest deposition rate, although the difference in the ratescreases with time. If the transition time is long enough for surfaleveling to occur, a smooth layer is formed. If the bulk concentratof SPS is high and the transition time is short, the differencetween the deposition rate of the valleys and peaks is not significand a rough surface results.

Conclusions

An experimental study of additive behavior for model coppplating solutions that exhibit superfilling was performed. The expements show that acidic copper plating solutions containing PECl2, and SPS exhibit a transition time for depolarization that isstrong function of SPS concentration. In galvanostatic experimesamples reached a steady-state value of the potential that was afunction of SPS concentration. In potentiostatic experiments,steady-state current density varied significantly with SPS concention at low concentrations and approached a single value as theconcentration increased. Consequently, relatively small changethe steady-state current at ‘‘high’’ SPS concentrations were obsefor large changes in concentration. Experimental results also


hy

he

r-torof

-to

ar-w

f-es

sur-ding

ther atad-the

ionedal-re-the

de-cene-ant

rri-G,ats,

lso ahetra-

PSin

vedro-

vided evidence for slow adsorption and desorption of the accelerand SPS incorporation into the deposit. In addition, LSV indicathat behavior of the accelerator was potential dependent. Anashowed that superfilling depends on the existence of a long tration time to permit establishment of a significant rate differenbetween the top and bottom of the cavity. The same mechanismto explain superfilling can be used to explain the levelinbrightening of copper surfaces in these solutions.

Brigham Young University assisted in meeting the publication coststhis article.

References

1. P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans, and H. Deligianni,IBM J.Res. Dev.,42, 567 ~1998!.

2. A. C. West, S. Mayer, and J. Reid,Electrochem. Solid-State Lett.,4, C50 ~2001!.3. T. P. Moffat, J. E. Bonevich, W. H. Huber, A. Stanishevsky, D. R. Kelly, G.

Stafford, and D. Josell,J. Electrochem. Soc.,147, 4524~2000!.4. J. Reid and S. Mayer, inAdvanced Metallization Conference 1999, p. 53, M. E.

Gross, T. Gessner, N. Kobayashi, and Y. Yasuda, Editors, MRS, Warrendale~2000!.

5. T. P. Moffat, D. Wheeler, W. H. Huber, and D. Josell,Electrochem. Solid-StateLett., 4, C26 ~2001!.

6. D. Josell, D. Wheeler, W. H. Huber, and T. P. Moffat,Phys. Rev. Lett.,87, 016102~2001!.

7. K. R. Hebert,J. Electrochem. Soc.,148, C726~2001!.8. J. D. Reid and A. P. David,Plat. Surf. Finish.,74, 66 ~1987!.9. D. Stoychev and C. Tsvetanov,J. Appl. Electrochem.,26, 741 ~1996!.

10. J. J. Kelly and A. C. West,J. Electrochem. Soc.,145, 3472~1998!.11. J. J. Kelly and A. C. West,J. Electrochem. Soc.,145, 3477~1998!.12. M. R. H. Hill and G. T. Rogers,J. Electroanal. Chem. Interfacial Electrochem.,68,

149 ~1976!.13. M. Kruft, B. Wohlmann, C. Stuhlmann, and K. Wandelt,Surf. Sci.,377, 601

~1997!.14. W. H. Li, Y. Wang, J. H. Ye, and S. F. Y. Li,J. Phys. Chem. B,105, 1829~2001!.15. G. M. Brown and G. A. Hope,J. Electroanal. Chem.,413, 153 ~1996!.16. E. E. Farndon, F. C. Walsh, and S. A. Campbell,J. Appl. Electrochem.,25, 574

~1995!.17. J. P. Healy and D. Pletcher,J. Electroanal. Chem.,338, 179 ~1992!.18. N. A. Zhukauskaite and A. A. Malinauskas,Zashch. Met.,25, 935 ~1989!.19. D. Stoychev and P. Stefanov,Russ. J. Electrochem.,30, 349 ~1994!; Translated

from Elektrokhimiya,30, 388 ~1994!.20. N. Zukauskaite and A. Malinauskas,Elektrokhimiya,12, 1691~1988!.21. N. Zukauskaite and A. Malinauskas,Elektrokhimiya,12, 1694~1988!.22. L. Yu. Valentelis,Prot. Met.,32, 38 ~1996!; Translated fromZashch. Met.,32, 44

~1996!.23. E. Richard, I. Vervoort, S. H. Brongersma, H. Bender, G. Beyer, R. Palmans

Lagrance, and K. Maex, inAdvanced Metallization Conference 1999, p. 149, M. E.Gross, T. Gessner, N. Kobayashi, and Y. Yasuda, Editors, MRS, Warrendale~2000!.

24. M. E. Gross, R. Drese, D. Golovin, W. L. Brown, C. Lingk, S. Merchant, andOh, inAdvanced Metallization Conference 1999, p. 85, M. E. Gross, T. Gessner, NKobayashi, and Y. Yasuda, Editors, MRS, Warrendale, PA~2000!.

25. P. Taephaisitphongse, Y. Cao, and A. C. West,J. Electrochem. Soc.,148, C492~2001!.

26. H. Deligianni, Personal communication~2001!.27. K. M. Takahashi and M. E. Gross,J. Electrochem. Soc.,146, 4499~1999!.


Documents

175 Additive Behavior During Copper Electrodeposition