c292 journal of the electrochemical society c292-c301 0013 … · 2013-03-18 · atomic force...

Journal of The Electrochemical Society, 150 ~5! C292-C301~2003!0013-4651/2003/150~5!/C292/10/$7.00 © The Electrochemical Society, Inc.

C292

Atomic Force Microscopy Examination of Cu Electrodepositionin TrenchesMyungchan Kang,a Mihal E. Gross,b and Andrew A. Gewirth a,* ,z

aDepartment of Chemistry and Fredrick Seitz Materials Research Laboratory, University of Illinois, Urbana,Illinois 61801, USAbAgere Systems (formerly Bell Labs, Lucent Technologies), Electronic and Photonic Materials PhysicsResearch, Murray Hill, New Jersey 07974, USA

Electrodeposition of copper in trenches with several acid cupric sulfate electrolytes containing benzotriazole~BTA!, 4,5-dithiaoctane-1,8-disulfonic acid~SPS!, polyethylene glycol~PEG!, or chloride ~Cl! has been probed byin situ atomic forcemicroscopy~AFM! andex situscanning electron microscopy~SEM!. The two techniques provide complementary information. Aseries of AFM images reveals dynamic morphology changes of deposits and trenches during open-circuit dissolution and elec-trodepostion. Analysis of these AFM images shows vertical and lateral growth of deposits in addition to cross-sectional profileevolution. The grain size analysis obtained from AFM images has been correlated with structures observed in cross-sectional SEMimages. While there is a good correlation between grain size and overpotential observed with AFM, this correlation does notextend to the interior structures observed with SEM. Different additives are found to influence growth with behavior ranging fromclear conformal growth with BTA to nascent superfilling with SPS1 PEG1 Cl2.© 2003 The Electrochemical Society.@DOI: 10.1149/1.1562597# All rights reserved.

Manuscript Submitted June 3, 2002; revised manuscript received November 16, 2002. Available electronically March 18, 2003.

dedeos

sity,ceis

V-om

fe-

nt

rfilela

ucav

reansirdd

o

ionto

rolte-

red as

elec-chevi-

Cu

chare

ith-

ition

of

to

ndtheiallyput-i-rr,

35

The recent development of electrodeposition technology toposit Cu for use as the interconnect material in microelectronicvices has led to dramatically increased interest in Cu electrodeption. Cu has a number of advantages over physical vapor depo~PVD! Al that it replaces. In particular, Cu exhibits low resistivithigh electromigration resistance, and amenability to dual damasprocessing.1-3 The electromigration resistance of the plated Cuorders of magnitude better than that of the Ti/Al~Cu!/Ti system itreplaces.3

Cu electrodeposition exhibits a number of advantages over Por chemical vapor deposition~CVD! processes. First, Cu electrodeposition is a very inexpensive process, estimated to be s60% cheaper than a putative CVD method.2 Second, electroplatedCu exhibits large grains, resulting in longer electromigration litimes relative to CVD Cu.4,5 Finally, a well-tuned electroplatingprocess can fill the high-aspect-ratio structures that are importadeep submicrometer dual damascene architectures.2 The high-aspect-ratio structures are filled via a process known as ‘‘supeing’’ wherein deposition at the bottom of a trench is enhanced rtive to that occurring at the top.6 This enables filling ofsubmicrometer damascene structures without defects and prodvoid-free and seamless deposits inside lithographically defined cties with vertical walls.1,3,7-9

The advantages attendant to super-fill Cu electrodepositionquire the use of additives, trace amounts of selected inorganicorganic molecules, in the plating bath. In particular, the very deable superfilling effect occurs only in the presence of certain ative combinations. The effects of additives on electrodepositionmany materials have been examined for years.10 Additives may beinvolved in ~i! grain refinement of the deposit, (i i ) polarization ofthe cathode, (i i i ) incorporation of additive in the deposit, and (iv)change of the orientation of electrodeposited crystals. In additadditives may interact synergistically, an effect of relevancesuperfilling.7,11-13

Because of their importance in the superfilling process, theof additives has been examined in detail. Many workers have inrogated the Cu electrodepositex situ using scanning electron microscopy~SEM! or focused ion beam~FIB! microscopy.8,12-20Thesestudies have shown that certain additive combinations7,21,22 do in-

* Electrochemical Society Active Member.z E-mail: [email protected]

--i-

ed

ne

D

e

in

l--

esi-

-d

-i-f

,

er-

deed provoke the superfilling response and have further uncoverecrystallization phenomena,8,20 apparently promoted by additivedirectly incorporated in the deposit.

These studies raise questions about the initial stages of Cutrodeposition in trenches. While inhibition of deposition at trenedges has been postulated in superfilling mechanisms, visualdence is lacking. The advantages of thein situ experiment to inter-rogate processes at the solid-liquid interface are well known.23 Wehave previously used atomic force microscopy~AFM! to develop amechanistic understanding of the role of certain additives inelectrodeposition on planar surfaces.24,25 In this paper, we use AFMto examine the electrodeposition of Cu onto well-defined trenstructures both with and without plating bath additives and compthese results with SEM images.

Experimental

1H-Benzotriazole~BTA!, polyethylene glycol~PEG,ca. averagemolecular weight 3400!, and KCl were purchased from Aldrichwhile 4,5-dithiaoctane-1,8-disulfonic acid~SPS! was obtained fromRaschig. These chemicals were employed as plating additives wout further purification.

Sulfuric acid~ultrapure!, CuSO4•5H2O ~Aldrich, 99.999%!, andwater purified in a Milli-Q system~18.2 MV cm21! were used in allsolutions. Unless noted otherwise, the solution for copper deposin all the experiments was either a solution of 0.5 M H2SO4 and0.05 M CuSO4 ~additive-free solution! or a solution of 0.5 MH2SO4 , 0.05 M CuSO4 , and the additives. The compositionsadditives were~i! 500 mM BTA, ( i i ) 1 mM SPS, (i i i ) 1 mM SPS1 1 mM PEG~SPS1 PEG!, (iv) 1 mM SPS1 1 mM PEG1 1mM KCl (SPS1 PEG1 Cl2). The relatively high BTA concen-tration was used to ameliorate additive depletion effects knownoccur with this additive.25

The working electrode consisted of a Si wafer with 2mm wideby 0.5 mm deep trench grating arrays in 0.5mm plasma-enhancedtetraethyl orthosilicate~PETEOS!-SiO2 . 500 Å PVD tantalum ni-tride was first deposited on the oxide as a diffusion barrier aadhesion layer, followed by 1000 Å PVD copper to serve ascathode for electroplating. These films were deposited sequenton 150 mm wafers, with no vacuum break, by dc magnetron stering in a Novellus M2i cluster deposition tool. Deposition condtions were as follows: tantalum nitride: 3 kW, no bias, 2.9 mTo1:1 collimator, 150°C wafer temperature, 35 sccm Ar, 35 sccm N2 ,Cu: 3.1 kW, no bias, 1:1 collimator, 50°C wafer temperature,sccm Ar.

-

Journal of The Electrochemical Society, 150 ~5! C292-C301~2003! C293

Figure 1. Images of copper deposits obtained from~a! 500mM BTA solutions atthe beginning of adding solution,~b! 500mM BTA solution after 50 min immer-sion time,~c! 1 mM SPS solution at thebeginning of adding solution, and~d! 1mM SPS solution after the immersiontime of 50 min without applying anycharge.

f tht

nt

tid.sit

at

Pi-l

rdeag

waan

nof

tingth

thaapthesrvre

hoC

per

theddi-s ofeenngnoBTA

c-di-hedis-

mins aies.S issamealsoive-

be-

M

tionof

hich

The geometric area for electrodeposition wasca. 0.64 cm2, anumber that does not include the area of the vertical sidewalls otrenches. The sidewalls are estimated to add another 10% togeometric area at the start of the deposition process. Both couand reference electrodes were copper wires~99.999%, 0.25 mmdiam! cleaned in 50% sulfuric acid followed by rinsing with hoMilli-Q water. The counter electrode was wrapped around the insof the cell in order to obviate any potential distribution problems

Deposition was carried out galvanostatically at a current denof either 0.8 or 5 mA/cm2 using a galvanostat~EG&G PrincetonApplied Research, model 273!. Hydrogen evolution was observedcurrent densities significantly higher than 5 mA/cm2. Potentials arereported with respect to the Cu21/Cu redox couple.

AFM studies were performed using a Molecular Imaging,coSPM 300, equipped with a 20mm scanning head. A pyramidaSi3N4 tip mounted on a gold-coated 100mm v-shaped cantileve~manufacturer’s force constant 0.58 N/m, manufacturer’s incluangle 35°! was used. Images were acquired in contact mode. Imsize was 203 20 mm or 10 3 10 mm. Before each depositionstep, the AFM scanner was withdrawn by 40mm to eliminate pos-sible tip effects on the deposition process. The applied forceminimized so as to reduce any disruption of the surface while scning.

In these measurements, we utilized 2mm wide trenches with a 4mm center-center spacing between adjacent features. Trenchesrower than 2mm could not be interrogated due to the finite radiusthe tips used here. Narrower trenches yielded images represenconvolution of the feature shape with that of the tip becauseincluded angle of the tip makes the tip blind to angles steeperitself. For 2mm wide trenches, the images clearly reveal the shand depth of the trench, but with slanting sidewalls showingeffect of the tip. We note that 2mm wide and 500 nm deep trencheexhibit aspect ratios smaller than that deemed optimal for obsetion of the superfilling effect, and that higher aspect ratio featucould in concept be imaged with sharper tips.26-28

After Cu deposition, the wafer was cleaned in isopropyl alcoand cleaved perpendicular to the trench direction to expose thestructure within the trench. Scanning electron microscopy~SEM!

eheer

e

y

de

s-

ar-

aene

a-s

lu

characterization of the microstructure of the electroplated copfilm was made using a Hitachi S4700 instrument.

Results

Dissolution of PVD Cu at open circuit.—Prior to performingelectrodeposition measurements, we examined the stability ofPVD copper seed layer in electroplating baths containing the ative combinations examined in this paper. Figure 1 shows a seriein situ AFM images taken immediately following immersion of thsample into solution and then 50 min following immersion at opcircuit. The additives fall into two categories. In solutions containi500 mM BTA comparison of before and after images revealedchange in the shape of the trench. This behavior is expected, asis a well-known corrosion inhibitor for Cu29,30 due to its ability toform a passivating polymeric barrier layer.25,31-36 Not surprisingly,BTA exhibits an overpotential nearly 200 mV greater for Cu eletrodeposition under similar conditions than any of the other adtives considered. All the other additive combinations fall into tsecond category wherein the PVD copper layer is unstable andsolves after immersion. This effect has been noted previously.6 Fig-ure 1c and d show the change in the PVD Cu seed layer after 50immersion into a solution containing 1 mM SPS. Figure 1d showmuch rougher surface, exhibiting pits and other inhomogeneitThat corrosion occurs at open circuit in solutions containing SPan indication that SPS does not passivate the surface to theextent as does BTA. The dissolution of copper seed layer wasobserved in the other solutions examined here, including additfree, 1 mM SPS1 PEG, and 1 mM SPS1 PEG1 Cl2 solutions.The rate of dissolution was not observed to vary appreciablytween these different additives.

Morphology changes during electrodeposition-AFimages.—We followed the change in morphology of a 2mm wide3 0.5 mm deep trench during the course of the electrodeposiprocess. Figure 2 is a series of AFM images showing the fillingtrenches with Cu in the presence of 500mM BTA solution as afunction of charge using a current density of 5 mA/cm2. Cu deposi-tion was continued until a charge of 12 C had been passed, at w

-

Journal of The Electrochemical Society, 150 ~5! C292-C301~2003!C294

Figure 2. Images of copper deposits obtained from 500mM BTA solution ~a! be-fore deposition, and~b! after 2.0,~c! 4.0,~d! 6.0, and~e! 8.0 C.

lecick

silecthib

ro-difsys

ssing

ainthehowre-

nch. Then

point the images did not reveal any further development of etrodeposit microstructure. 12 C corresponds to an equivalent thness ofca. 6 mm.

Figure 2a is an AFM image of the trench prior to electrodepotion. As charge is applied to the sample, Fig. 2b-e show that etrodeposition inside the trench catches up with that growing onterrace next to the trench. After passage of 8 C, the trench is visas only a slight indentation in the Cu-covered surface.

In order to elucidate the influence of current density on micstructure evolution in the trench, we examined deposition usingferent current densities. Figure 3 shows a comparison of two

--

--

ele

--

tems ~additive-free and 500mM BTA ! as a function of currentdensity. Figure 3a and b are the images of the deposits after pa2.0 C with current densities of 0.8 and 5 mA/cm2, respectively. Theimages obtained by applying low current density display larger grsize and exhibit an apparently lower nucleation density thanimages obtained using high current density. The images also sthat for the additive-free solution, considerably less charge isquired at low current density before the Cu deposited in the trecatches up with that deposited on the terrace above the trenchtrench filled using low current density in BTA-containing solutio

-
Figure 3. Images of copper deposits obtained from~a! additive-free solution atthe current density of 0.8 mA/cm2, ~b!additive-free solution at 5.0 mA/cm2, ~c!500 mM BTA solution at 0.8 mA/cm2,and ~d! 500 mM BTA solution at 5.0mA/cm2.

-


Figure 4. Images of copper deposits obtained from~a! additive-free,~b! 500 mMBTA, ~c! 1 mM SPS,~d! 1 mM SPS1PEG, and ~e! 1 mM SPS1 PEG1 Cl2 solutions after passing 2.0 C witha current density of 5.0 mA/cm2.

bocurres.witen

thromhethehi

s eion

ghamum

ion

y,thnnties

r-

Mc-

withgeson-rro-gese-

thatCum

lliza-

own

ely.par-

y is-of

also appears to have caught up with deposition on the terrace athe trench to a greater extent than that accomplished with highrent density for the same charge passed. However, higher cudensities are likely more relevant to the actual industrial proces

Figure 4 shows a series of images of Cu deposits obtainedfive different additive systems after passing 2.0 C with a currdensity of 5.0 mA/cm2. All of the deposits exhibit filling in of thetrench relative to the bare surface shown in Fig. 2a. However,deposits exhibit differing degrees of roughness. For deposition fthe BTA-containing solution, copper grains evident both in ttrench and on the terrace are small and numerous. We note thatfeatures might not necessarily be grains in the crystallograpsense. In the case of additive-free, SPS, and SPS1PEG solutions,the surface is rougher and the grains are larger. These deposithibit minor differences relative to each other. However, depositfrom the SPS1 PEG1 Cl2-containing solution evinces a fillingaction very different from the others. The deposits are quite rouand the trenches are almost completely filled after passing the samount of charge. This illustrates that the presence, type, and nber of additives significantly affects deposit leveling. The additof chloride ion into the solution containing SPS1 PEG markedlyinfluences the deposition process.

In addition to monitoring the evolution of deposit morphologwe also examined the overpotential developed in the plating baTable I shows that the overpotential depends on the compositiothe plating solution and the current density. The largest overpotewas developed in the BTA-containing solution, while the smallwas observed in that containing SPS1 PEG1 Cl2. BTA is knownto increase the overpotential for Cu electrodeposition.37 Interest-ingly, addition of Cl2 to SPS1 PEG induces a significant depolaization effect, the origin of which may reside in the Cu~I! coordina-tion chemistry of SPS and Cl2. We showed that Cl2 could changethe coordination of BTA to Cu~I! in a previous study.38

Morphology evaluation after electrodeposition-SEimages.—We performed cross-sectional SEM imaging of Cu ele

ver-nt

ht

e

sec

x-

,e-

s.ofalt

trodeposits grown in trenches in order to compare these resultsthose from AFM, described previously. Figure 5 shows SEM imaof the morphology of Cu electrodeposits grown under the same cditions as those used for AFM analysis. The sample areas integated in these images were not subject to AFM analysis. The imashow widely differing habits for the electrodeposited Cu that dpends on additive composition and current density. We notethere is considerable literature describing recrystallization ofelectrodeposits, especially those obtained froSPS1 PEG1 Cl2-containing solution.20 We obtain SEM imagesat time scales short enough after Cu deposition so that recrystation was not an issue.

Figure 5a and b shows SEM images obtained on deposits grfrom solutions containing 500mM BTA at a current density of 0.8and 5.0 mA/cm2 and at 6.5 and 12 C current passed, respectivThere is an important difference between the two deposits. Inticular, the grain size in the trench changes fromca.0.23 to 0.13mmat a coverage corresponding to 6.0 C as the current densitchanged from 0.8 to 5.0 mA/cm2. This means that the deposit obtained with higher current density appears with a higher density

Table I. The results of the reduction overpotentials of differentsystems with respect to the Cu2¿ÕCu redox couple.

SystemCurrent density

~mA/cm2!Overpotentiala

~mV!

Additive-free 5.0 2173With 500 mM BTA 5.0 2419With 1 mM SPS 5.0 2216With 1 mM SPS1 PEG 5.0 2227With 1 mM SPS1 PEG1 Cl2 5.0 264Additive-free 0.8 234With 500 mM BTA 0.8 2256

a The data were reported at 2.0 C.


Figure 5. SEM images of trenches obtained from~a! 500mM BTA, ~b! 500mM BTA, ~c! additive-free,~d! 1 mM SPS,~e! 1 mM SPS1 PEG, and~f! 1 mMSPS1 PEG1 Cl2 solutions with current density of 0.8 or 5.0 mA/cm2. The scale bar represents 2mm.

hena

by

e6 Cornto

ols

ow

bye oesitEMuri

telol-itan-d

o

ecf a

heionple,

e.

nalfor5.0age

shol-ing

pesalthean-oms isas a

ar.in-t

th.feen

eicu-

nuclei in the trench relative to the low-current-density deposit. Tdeposits obtained with low current density evince a columgrowth pattern which is known as a field-oriented texture~FT! type,in which crystals grow by two-dimensional nucleation followedgrowth with constant grain diameter in all deposition areas.39 How-ever, with high current density, the morphology appears to maktransition in type at a depth corresponding to a charge of aroundBelow the transition point, the deposit appears to be of the unented dispersion~UD! type, in which three-dimensional nucleatiooccurs. Above the transition point, UD-type texture is changedFT-type texture with larger grain size.

The SEM images show that additive identity strongly contrthe morphology of the deposit. In the additive-free case~Fig. 5c!obtained after passage of 8 C, the image shows a plate-like grknown as a basis reproduction~BR! structure39 in which the Cucrystallites are obtained by two-dimensional nucleation followedgrowth in all dimensions. The image also shows the presencvoid areas, especially in the vicinity of the trench. Some of thvoid structures may be due to imperfect adhesion of the deposthe substrate following the cleavage required to make the Ssample. The void structure may also have been generated delectrodepostion and be intrinsic to the Cu deposit.

Incorporation of SPS into the plating bath leads to a compledifferent morphology, as shown in Fig. 5d, which was obtained flowing passage of 10 C at 5.0 mA/cm2. The image shows a deposwith characteristics intermediate between the field-orientedbasis-reproduction~BR-FT! type. Addition of PEG to the SPS solution gives rise to a deposit~Fig. 5e! that is changed again anevinces a more nearly FT structure. Finally, inclusion of Cl2 in thebath changes the deposit yet again~Fig. 5f!, leading to a BR-typestructure. In this case, the grains are much bigger than thosetained from the other solutions.

The BTA, SPS, and SPS1 PEG-containing solutions all giverise to electrodeposit structures that feature voids and other defas observed by SEM images. These voids are possible sites o

r

a.

i-

th

feto

ng

y

d

b-

ts,d-

ditive inclusion.40 The size of these defects is correlated with toverpotential that, in turn, is related to the solution composit~Table I!. Higher overpotential reduces the defect size. For examcomparing Fig. 5b~overpotential2419 mV! with Fig. 5f ~overpo-tential 264 mV!, the apparent grains are larger in the latter imag

Analysis

Section analysis.—Figure 6 shows the change in cross-sectioprofiles derived from the AFM images as a function of chargefive different systems obtained using a current density ofmA/cm2. Each cross-sectional profile shown in Fig. 6 is the averof all of the 512 scan lines of the 103 10 mm image. For clarifi-cation, thez scale value atx 5 1 mm is set to 0 nm for each crossection. The profiles show that the development of trench morpogy depends strongly on the additive with different rates of platat the bottom and along the sidewall of the trenches.

This section analysis provides considerable insight into the tyof profile evolution occurring in Damascene plating. In conformplating, copper is deposited equally at all points. Therefore,depth of trench will not change until the flat bottom becomes trigular in shape. In superfilling conditions, the deposit in the bottof the trench is thicker and the deposit at the top of the trencherelatively thinner. The depth of the trench, therefore, decreasesfunction of coverage before the cross section becomes triangul

The cross-sectional profiles obtained from the solution containg BTA displays typical conformal plating evolution withouchanges of depths, whereas the profiles from the SPS1 PEG1 Cl2 solution shows superfilling plating with decrease in depFor additive-free, SPS, and SPS1 PEG solutions, the changes othe profiles seem to indicate a plating action intermediate betwsuperconformal and conformal.

Roughness analysis.—Quantitative measurement of surfacroughness is obtained from the standard deviation of the perpend

f


Figure 6. Cross-sectional analysis otrenches obtained from~a! additive-free,~b! 500 mM BTA, ~c! 1 mM SPS,~d! 1mM SPS 1 PEG, and~e! 1 mM SPS1 PEG1 Cl2 solutions at 5.0 mA/cm2

current density as a function of charge.

sohd

nrveslueugh

ese

tottese

ainis-

.nanri-t twtoe

cine tf th

lar surface variation,j, also known as the root mean square~rms!height of the surface.j is obtained over the entire image andinitially reflects the roughness due to the presence of the trencthe image. As the deposit fills the trench, the roughness shouldcrease.

Figure 7 shows the change ofj for the five systems as a functioof charge. At the early stages, the value of roughness obsereflects the trenches. The value ofj saturates after the trencheare filled in and flat surfaces are formed. The saturated vaindicate the roughness of deposits. The order of the value of roness is SPS1 PEG1 Cl2 . additive-free. SPS1 PEG. SPS. BTA.

Grain size analysis.—Electrodeposits featuring large grain sizimply the presence of a substantial lateral growth mechanism duthe small grain size of the seed layer. The grain sizes andnumber of grains in the images shown in Fig. 2 were investigaquantitatively in the following manner. We note again that thefeatures are not necessarily crystallographic grains but are certnuclei resulting from electrodeposition. The morphology is dplayed by measuringz height at eachx andy coordinate in the AFMimage. The number of points of the samez height was determinedAt different amounts of charge passed, a histogram as a functioheight was obtained. The histograms for the additive-free, BTA,SPS1 PEG1 Cl2 systems are shown in Fig. 8. For clear compason, the depth of the largest peaks was set to 0 nm. The largespeaks at the charge of 0 C correspond to height points from theand bottom layers of the bare copper trenches. The region betwthe two peaks reflects the sidewalls of the trenches. The spabetween the two peaks indicates the depth of the trenches, whilwidth of the peaks shows the roughness of deposits in and out o

ine-

d

s-

toald

ly

ofd

openg

hee Figure 7. Log-log plot of j vs. the amount of charge passed.

vove

ssinsn b

easflaf thgetiotio

ms t

tf thow

ed toaxi-ctsafter

r ofns

ityting

he

ameand

Fore-e ortingr

e20

2D

5

ms.aren-lcu-e inwth

ithr-Wendse-anor-nots iserialgesreteigin


trenches. The roughness obtained by scaling analysis is a contion of the roughness intrinsic to the trenches alone and that deoping from electrodeposit. From these histograms, the roughnethe copper electrodeposits can be investigated excluding the intrtrench contribution. The roughness of top and bottom layers caexplored separately by measuring the width of the two peaks.

In Fig. 8b, obtained with BTA as the additive, the width of thpeak associated with the top of the trench does not changefunction of coverage. This indicates the electrodeposits are veryThe disappearance of the peak associated with the bottom otrench in this plot implies that the bottom of the trench is no lonflat and reflects the evolving trigonal shape seen in the secanalysis. The behavior of the histogram obtained from the solucontaining SPS1 PEG1 Cl2 shown in Fig. 8c is different fromthat obtained with BTA. After passing 2.0 C, the two peaks becomuch broader and the maximum of the bottom layer peak shiftsmaller values. This means that the electrodeposits on both theand bottom layer of the trenches are very rough and the depth otrenches decreases. The peaks for the additive-free solution shpattern intermediate between those with BTA and SPS1 PEG1 Cl2 as shown in Fig. 8a.

Figure 8. Histograms as a function of charge passed obtained from~a!additive-free solution,~b! 500 mM BTA solutions, and~c! 1 mM SPS1 PEG1 Cl2 at a current density of 5 mA/cm2. ~ !, ~n!, and ~.!denote histograms obtained at 0, 2.0, and 4.0 C, respectively.

lu-l-ofice

at.e

rnn

eoopea

The full width at half maximum~fwhm! from the peak in thehistogram corresponding to the area above the trenches was usset a threshold to visualize the grains. The height below half mmum is displayed in black as shown in Fig. 9. Figure 9 reflegrains of electrodeposits on the top of trenches for five systemspassing 2.0 C with a current density of 5.0 mA/cm2. Different grainsizes and numbers of grains are shown clearly in Fig. 9. The ordethe average grain size obtained from those solutiois SPS1 PEG1 Cl2 . additive-free. SPS. SPS1 PEG. BTA, as summarized in Table II. The number of nuclei densand the deposit size depends on the composition of electroplasolution. By correlating the roughnessj values obtained previouslyand the grain size, the electrodeposits containing SPS1 PEG1 Cl2 grow laterally and vertically to a greater extent than tdeposits obtained from the other additive mixtures.

Power spectral density analysis.—Unfilled large trenches andsmall electrodeposits on the large features can produce the sroughness value as that obtained from large electrodepositssmooth filled trenches. Power spectral density~PSD! is a method forcharacterizing each surface’s feature frequency or wavelength.more complex surfaces, it would be helpful to know which wavlengths occur most often and which impart the greatest influencpower to the surface topography. The PSD is obtained by calculathe fast Fourier transform~FFT! of the autocovariance function oby calculating the square of its Fourier transform~FFT!. The fre-quency distribution for a digitized profile of lengthL, consisting ofN points,zj , sampled at intervals ofd0 is approximated by

PSD~ f ! 5d0

N (j 51

N

uzje2i2p f ~ j 21!d0 /Nd0u2 @1#

where i 5 A21, and frequencies,f, range from 1/L to N/2/L. Inour case,L 5 20 mm andN 5 512. The wavelength is the inversof frequency. The maximum of the wavelength in the images ismm/cycle and the minimum is 0.078mm/cycle. The power,P, whichis obtained from squaring the FFT of the image, is used to deriveisotropic PSD as follows

2D isotropic PSD5P

2p f ~D f !@2#

Here, D f is the difference in frequencies which in this case is3 1025 nm21.

Figure 10 shows power spectral density plots for three systeIn the case of a BTA-containing solution, many distinct peaksshown. The peaks at 4 and 2mm/cycle arise from the trenches. Ithe solution containing SPS1 PEG1 Cl2, the trench-induced periodicity of the surface is lost, suggesting that the roughness calated previously is not associated with the trenches as it might bthe BTA case. In an additive-free solution, an intermediate gropattern results as discussed previously.

Comparison between SEM and AFM cross sections.—We usedSEM to examine the effect of scanning the Cu electrodeposit wthe AFM. In contact mode AFM, the tip is in contact with the suface, and may remove or alter soft layers in contact with Cu.grew the electrodeposit with the tip withdrawn from the surface athen stopped deposition to interrogate with the AFM. The conquences of this interrogation are shown in Fig. 11a, which isSEM image for an area scanned by the AFM tip. The overall mphology of this cross section is similar to that found in areasscanned by the AFM. However, the presence of discrete layerclearly noted in the image. These layers correspond to the matdeposited between AFM interrogation steps. Note that SEM imaobtained from areas absent AFM scanning do not show this discpattern, even though they were grown in the same way. The or

f


Figure 9. Images of grains on the top otrenches obtained from~a! additive-free,~b! 500 mM BTA, ~c! 1 mM SPS,~d! 1mM SPS 1 PEG, and~e! 1 mM SPS1 PEG1 Cl2 solutions after passing 2.0C at a current density of 5.0 mA/cm2.

anthe

on.0o

lcu11foEte

ivelsis

.ing

of the discrete layering likely arises from interference by the scning AFM with adsorbed layers of additive on the surface duringinterrogation step.

Analysis of the image provides some additional informatiabout the AFM imaging. AFM images were obtained 0.0, 0.5, 1and then at 1.0 C increments. Assuming a current efficiency100% and a uniform Cu layer, each 0.5 C of charge is then calated to correspond to a Cu deposit layer 287 nm thick. Figureshows the result of offsetting the AFM cross-sectional images0.5, 1.0, and 2.0 C by 574 nm/C. The correspondence with the Simage shown in Fig. 11a is quite good. In particular, the calculadepth of the deposit at 2.0 C is 1148 nm, while the SEM image ga depth of1150 6 20 nm. The SEM/AFM comparison also reveaissues at low Cu coverage with sidewall imaging by AFM, as

Table II. The grain size and the number of grains.

SystemAverage grain

size ~nm2!Number of

grains

Additive-free 31,480 117With 500 mM BTA 2877 1928With 1 mM SPS 17,140 414With 1 mM SPS1 PEG 9726 561With 1 mM SPS1 PEG1 Cl2 36,870 119With 500 mM BTAa 9999 627

a The data were reported at 12.0 C, other data were reported at 2.0 C

-

,f-brMds

Figure 10. Power spectral density is plotted for each solution after pass4.0 C.

breheicao-ithind

odnove

alstho

a-ul

ata

lu

eandesdior-liethe

witlyE

heiveofila

als

izeur-eW

e,

atss.the

osit

e UDthersec-

isat

e

ntial-num-

age

-th

ce aThe

ertalac-

tiond in

durespo-hertothe

tialthatnm

sore

s-

gestsaretialion

a

dif-aceityres.s int, asnt-po-

f


evident in a comparison of the circled areas in Fig. 11a andAlternatively, the top corners of the trenches and the terrace aboth in and out of the trench appear to be well imaged. At higcoverages, the SEM and AFM cross sections are nearly identIntermittant rastering with the AFM tip during electrodeposition prvides a way to induce structuring into the Cu structure that wappropriate development may lead to a way to introduce other kof striation into the deposit.

Discussion

In this paper we used AFM, SEM, and electrochemical methto address a number of issues relevant to electrodepositiotrenched microstructures. The results and analysis presented prinsight into the way in which additives control the filling of thestrenches by the Cu electrodeposit. The additive makeupstrongly influences the overpotential at which plating occurs andnumber and distribution of nuclei in the process. From analysisAFM and SEM images, insight into the particular growth mechnism directed by the additives can be obtained. Finally, the resspeak to the correlation of AFM and SEM methodologies.

Perhaps the major result of this study is the demonstration thsurface analytical method, contact mode AFM, can provide informtion about the fill characteristics of Cu electrodeposited from sotions of different additive combinations.

Comparison of AFM and SEM-derived information.—One ad-vantage accruing to AFM over the more conventional SEM msurement is that the AFM can provide high-resolution lateral avertical information in one image, whereas a multitude of imagare required to obtain the same information with SEM. This adtional information is useful here, because the AFM provides infmation about the development of grains as a function of apppotential. Use of the SEM has previously been limited only toevolution of the Cu electrodeposit in the trenches.7,8,13 However,there are clearly different topographical features associatedtrench filling that are seen with the AFM. Grains with sizes onseveral tens of nanometers in width are hard to ascertain in Simages.

The AFM can also be used to monitor the evolution of trencand deposit as a function of charge. The histogram analysis ginsight into the vertical growth of the deposit and the evolutiontrenches. The AFM also distinguishes between cases having simcross sections but different grain sizes, as is the case in Fig. 3cd. The equivalent cross-sectional SEM images, Fig. 5a and b,evince differences, but these are somewhat harder to see.

Correlation between electrochemical, SEM, and grain smeasurements.—The analysis shows clearly that for a constant crent density, the overpotential can be related to both the grain sizdetermined by AFM and inner structures as determined by SEM.showed that higher overpotential makes for smaller grain siz

Figure 11. ~a! SEM trench cross sections and~b! cross-sectional analysis oAFM images of trenches for BTA-containing plating solution.

.asrl.

s

sinide

oef

ts

a--

-

-

d

h

M

ss

arndo

asea

result described previously only for the additive BTA on flsurfaces.37 The overpotential does not depend on deposit thickneHowever, there is also a correlation between overpotential andinner structure of the deposit as determined by SEM. The depstructure evolves as FT-UD. FT . BR-FT . BR as the overpo-tential decreases. This suggests that smaller grains may producor FT structures. These FT or UD structures feature voids and oinhomogeneities that may be sites of additive inclusion. Indeed,ondary ion mass spectroscopy~SIMS! results25 suggested the inclu-sion of BTA in what we now know are FT-UD or FT structures. Thinclusion may be the origin of the high resistivity found by Moffand co-workers in this system.8

BTA.—The deposits grown from BTA-containing solution yield thhighest overpotential~Table I!, lowest roughness~Fig. 7!, highestnuclei density~Fig. 9!, smallest average grain size~Table II!, butpoor ability to fill the trenches~Fig. 10!. The inhibiting organic layerformed on copper surfaces by this additive makes the overpotehigh. In the initial stage~below 6 C!, the high overpotential suppresses lateral and vertical growth and consequently creates aber of nucleation sites.37 The lateral diffusion length of copper ionon the surface suggested by grain size analysis of the AFM im~Fig. 9b! is small.

The cross-sectional SEM image~Fig. 5b! shows consistent results with numerous small grains showing UD texture-type growstructure. In this stage, surface relaxation, in which particles~Cuadatoms! on the surface move and fill up recesses so as to produflat surface, is supposed to contribute to the growth mechanism.flat surface gives a typical conformal plating action~Fig. 6b!.

The cross-sectional SEM image~Fig. 5b! shows the transition ofthe types of growth structure from UD type to FT type with larggrains. The incorporation or inclusion of molecules into the mematrix during deposition is a process that is believed to furthercount for additive consumption.40 In our previous results, SIMSmeasurements of the copper deposits for BTA-containing solushow that a significant amount of organic material is incorporatethe deposits, implying a consumption of BTA.25

In our previous results,25 big, pyramidal features were observefollowing passage of a certain charge. These pyramidal featwere associated with depletion of additive, due to additive incorration in the deposit. In this experimental system, we used a higconcentration of BTA so that depletion was not significant enoughpromote the pyramidal features. Over the course of deposition,overpotential decreased from20.419 V ~at 2.0 C! to 20.204 V ~at12.0 C! due to a decreased inhibiting layer. Lower overpotenmakes grains wider and rougher. Grain size analysis revealsaverage grain size and number of grains are changed from 28772

and 1928 at 2.0 C to 9999 nm2 and 627 at 12 C, respectively. Amore charge is applied to the sample, lateral growth makes mcontribution at higher charges.

SPS1 PEG 1 Cl2.—The cross-sectional SEM images of depoits grown from SPS1 PEG1 Cl2 show large grains with a BR-type structure. The presence of these large grain boundaries sugthat diffusive mechanisms transporting Cu onto growing grainsoperative and that lateral growth is a factor. The lower overpotenfound here relative to the other solutions may admit longer diffusof particles on the surface before reduction and incorporation.

SPS and SPS1 PEG.—The deposit grown from SPS revealsBR-FT structure while the SPS1 PEG-containing solution givesrise to a FT-type deposit. These different structures must reflectferent 3D growth modes for the deposit. Even though the surfmorphologies are similar with similar overpotentials, the identand combination of additives alters the internal deposit structuWe suggest that different internal structures arise from differencethe mode and mechanism of additive incorporation in the deposideposits obtained from SPS1 PEG feature enhanced C conterelative to those formed from SPS alone.41 The SEM reveals abundant voids and defects, which may be sites of this additive incor

om. Itre

ogeessso

Cuse

EMce

iorv-ertonPSi-

. Wit

ThE

t th

this

nd T

i,

el-

R.

om-

i,


ration. The origin of these voids may be imperfections arising frthe lack of relaxation, possibly inhibited by incorporated additiveis apparent that overpotential alone does not control the structuthe Cu electrodeposit but is certainly a major contributor.

Conclusion

This work shows that some insight into the growth mechanismCu electrodeposits in trenches can be obtained from AFM imaThe AFM can follow the deposition of Cu into trench structurunder potential control. Analysis of these images in terms of crosectional profile evolution, roughness, and change in depthtrenches and grain size provides insight into the way in whichgrows under conditions of different additive compositions. A clocorrelation of results is found in cases where the AFM and Scross sections can be measured independently at the same plathe surface.

Appropriate analysis of these images shows different behavof Cu deposit growth from different additive solutions. This behaior ranges from clear conformal growth with BTA to nascent supfilling with SPS1 PEG1 Cl2. The AFM-derived insight abougrowth mechanism can be combined with that from cross-sectiSEM. In some cases, notably those involving BTA and S1 PEG1 Cl2, there is a good correlation between the two mcroscopies. However, in others, this correlation appears lackingsuggest that the degree of additive incorporation in the deposthen important.

Acknowledgments

The assistance of Vania Petrova is gratefully acknowledged.work was funded by a U.S. Department of Energy grant no. DFG02-91ER45349 through the Materials Research Laboratory aUniversity of Illinois.

The University of Illinois assisted in meeting the publication costs ofarticle.

References

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

2. R. L. Jackson, E. Broadbent, T. Cacouris, A. Harrus, M. Biberger, E. Patton, aWalsh,Solid State Technol.,41, 49 ~1998!.

3. R. Rosenberg, D. C. Edelstein, C. K. Hu, and K. P. Rodbell,Annu. Rev. Mater. Sci.,30, 229 ~2000!.

4. C. Ryu, K. W. Kwon, A. L. S. Loke, H. Lee, T. Nogami, V. M. Dubin, R. A. KavarG. W. Ray, and S. S. Wong,IEEE Trans. Electron Devices,ED-46, 1113~1999!.

5. B. Chin, P. J. Ding, B. X. Sun, T. Chiang, D. Angelo, I. Hashim, Z. Xu, S. Edstein, and F. S. Chen,Solid State Technol.,41, 141 ~1998!.

of

fs.

-f

on

s

-

al

eis

is-e

.

6. J. Reid, S. Mayer, E. Broadbent, E. Klawuhn, and K. Ashtiani,Solid State Technol.,43, 86 ~2000!.

7. Y. Cao, P. Taephaisitphongse, R. Chalupa, and A. C. West,J. Electrochem. Soc.,148, C466~2001!.

8. 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!.

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

10. M. Datta and D. Landolt,Electrochim. Acta,45, 2535~2000!.11. L. Oniciu and L. Muresan,J. Appl. Electrochem.,21, 565 ~1991!.12. A. C. West, S. Mayer, and J. Reid,Electrochem. Solid-State Lett.,4, C50 ~2001!.13. P. Taephaisitphongse, Y. Cao, and A. C. West,J. Electrochem. Soc.,148, C492

~2001!.14. S. Subbanna,Tech. Dig. - Int. Electron Devices Meet.,1996, 275.15. J. J. Kelly, C. Y. Tian, and A. C. West,J. Electrochem. Soc.,146, 2540~1999!.16. Q. T. Jiang, R. Mikkola, and B. Carpenter,J. Vac. Sci. Technol. B,17, 2361~1999!.17. E. C. Cooney, D. C. Strippe, and J. W. Korejwa,J. Vac. Sci. Technol. A,18, 1550

54 ~1950!.18. J. J. Kelly and A. C. West,Electrochem. Solid-State Lett.,2, 561 ~1999!.19. K. M. Takahashi and M. E. Gross,J. Electrochem. Soc.,146, 4499~1999!.20. C. Lingk and M. E. Gross,J. Appl. Phys.,84, 5547~1998!.21. D. Josell, D. Wheeler, W. H. Huber, J. E. Bonevich, and T. P. Moffat,J. Electro-

chem. Soc.,148, C767~2001!.22. D. Josell, D. Wheeler, W. H. Huber, and T. P. Moffat,Phys. Rev. Lett.,87, 016102

~2001!.23. A. A. Gewirth and B. K. Niece,Chem. Rev.,97, 1129~1997!.24. W. U. Schmidt, R. C. Alkire, and A. A. Gewirth,J. Electrochem. Soc.,143, 3122

~1996!.25. T. Y. B. Leung, M. C. Kang, B. F. Corry, and A. A. Gewirth,J. Electrochem. Soc.,

147, 3326~2000!.26. J. E. Griffith, L. C. Hopkins, C. E. Bryson, A. Berghaus, E. J. Snyder, J. J. Pl

bon, L. A. Vasilyev, M. Hecht, and J. B. Bindell,J. Vac. Sci. Technol. B,15, 2189~1997!.

27. J. E. Griffith, H. M. Marchman, and L. C. Hopkins,J. Vac. Sci. Technol. B,12,3567 ~1994!.

28. J. E. Griffith, H. M. Marchman, G. L. Miller, and L. C. Hopkins,J. Vac. Sci.Technol. B,13, 1100~1995!.

29. J. B. Cotton and I. R. Scholes,Br. Corros. J., London,2, 1 ~1967!.30. I. C. G. Ogle and G. W. Poling,Can. Metall. Q.,14, 37 ~1975!.31. G. W. Poling,Corros. Sci.,10, 359 ~1970!.32. Y. Ling, Y. Guan, and K. N. Han,Corros. Sci.,51, 367 ~1995!.33. P. G. Fox, G. Lewis, and P. J. Boden,Corros. Sci.,19, 457 ~1979!.34. J. Rubim, I. G. R. Gutz, O. Sala, and W. J. Orville-Thomas,J. Mol. Struct.,100,

571 ~1983!.35. D. Thierry and C. Leygraf,J. Electrochem. Soc.,132, 1009~1985!.36. H. Y. H. Chan and M. J. Weaver,Langmuir,15, 3348~1999!.37. M. J. Armstrong and R. H. Muller,J. Electrochem. Soc.,138, 2303~1991!.38. M. E. Biggin and A. A. Gewirth,J. Electrochem. Soc.,148, C339~2001!.39. R. Winand, P. Vanham, R. Colin, and D. Milojevic,J. Electrochem. Soc.,144, 428

~1997!.40. G. A. Hope, G. M. Brown, D. P. Schweinsberg, K. Shimizu, and K. KobayashJ.

Appl. Electrochem.,25, 890 ~1995!.41. M. Kang and A. A. Gewirth,J. Electrochem. Soc., In press.

c292 journal of the electrochemical society c292-c301 0013 … · 2013-03-18 · atomic force...

Documents