Controlling Copper Electrochemical Deposition (ECD)Michael West, Robert McDonald, Marc Anderson, Skip Kingston

and Rudy MuiMetamlnc., 1225 East ArquesAve., Sunnyvale, CA 94085, USA

Abstract. The implementation of copper processing in semiconductor manufacturing has resulted in majorprocess development and manufacturing challenges. A fundamental understanding of the copper platingprocesses used in manufacturing has been limited by the lack of in-line methods for direct measurement andcontrol of process chemistry. Plating bath chemistry adjustments and change-out frequencies are currentlydetermined using a combination of indirect electrochemical monitoring techniques, off-line analyses of wafermetrology and analytical lab measurements. There have been a number of industry reports of major processstartup delays, yield management problems and reliability issues as a result of these difficulties. A new in-process mass spectrometry (IPMS) approach enables automated, real-time measurement of both the inorganiccomponents and organic additives in the copper electroplating chemistry as they change during production.The tool is not only capable of real time direct quantification of the copper, chloride, pH, and organic additivesin the plating bath, but can also monitor additive breakdown byproducts as they occur during the productionprocess. These breakdown products, as well as changes in the original bath constituent composition can beexpected to have a major impact on process performance. We are now in the process of measuring longer termplating bath stability and chemistry changes in prototype applications in semiconductor fab manufacturingenvironments. The first results demonstrate improved process understanding and the potential for greatlyimproved process control. We will discuss the technical challenges that were successfully addressed indeveloping the IPMS capability for application to the copper plating process and the initial process datasubsequently obtained.

INTRODUCTION

There is a major effort in the semiconductorindustry to migrate from aluminum to copperinterconnects in 1C devices. The commonmethod for depositing the copper conductor filmand filling the narrow vias that interconnectlayers of metallization is electrochemicaldeposition (ECD). The ECD process is anextension of the same process used for platingcopper electrical conductor on printed circuitboards and uses an aqueous copper sulfatesolution containing sulfuric acid and chlorideions.

The properties of the deposited copper suchas its resistivity, ductility and grain size aremodified and improved by the addition oforganic compounds. Historically they have beenassigned functional labels - such as accelerators(brighteners), suppressors, and levelers - thatdescribe their affect on the rate of copperelectrodeposition and its morphology. Howeverthe demands of the 1C industry for successfulbottom-filling of high-aspect ratio features,without defects, and for increasingly shrinkinggeometries, require increasingly tighter controlover the relative concentrations of the organic

additives and the basic inorganic components [1-3].

In spite of the importance of these additives,there is an incomplete understanding at themolecular level of their role in theelectrodeposition process. Current processcharacterization and control methods include acombination of chemical measurements fromoff-line labs [3,4] and bath chemistryreplenishment or changeout at empiricallydetermined frequencies or in response to yieldproblems. Many copper ECD bath users employa "bleed-and-feed" method to control the level oforganic additives. The method consists ofremoval of a portion of the bath (on the order of10%) and replenishment with fresh solution on adaily basis, or other period. Some users thendiscard the bath after several weeks of use.

The most prevalent current technology forautomated in-line control of bath chemistryemploys cyclic voltametric stripping CVS [1-3].This method relies on electrical signals to inferchanges in the major organic components of thebath and has been extended to include"potentiometric titration" for measurement ofinorganic components [2]. Use of these methodsto accurately control process chemistry is

CP683, Characterization and Metrology for VLSI Technology: 2003 International Conference,edited by D. G. Seiler, A. C. Diebold, T. J. Shaffner, R. McDonald, S. Zollner, R. P. Khosla, and E. M. Secula

© 2003 American Institute of Physics 0-7354-0152-7/03/$20.00504

difficult at best and ignores direct measurementof breakdown products. Breakdown productswhich occur during processing and bath agingare expected to be a major factor in the quality ofthe deposited film and are a process variable inneed of direct process control.

There exists a pressing need for a reliabletool that automatically monitors BCD processchemistry directly on-line and providesinformation that can be used to control thereplenishment of the electrodeposition bathcomponents. The same information will alsoaccurately forecast the need for full bathchangeout and warn of unexpected process toolproblems and unforecasted maintenance needs.

In contrast to the above methods, a unique"chemical constituent analyzer" (CCA) variationof a new in-process mass-spectrometry (IPMS)metrology tool [5-6] has now enabled in-situ,live-time measurement of electroplating bathprocess chemistry under actual processingconditions. This capability includes quantitativemeasurement of concentration changes inoriginal bath components as well asidentification and measurement of breakdownproducts as the bath is used in production.

In the following sections we will discuss theCu BCD process in more detail and brieflydescribe the unique instrument design whichmakes this measurement capability possible. Wewill then present results which have beenobtained in measurements of actual BCD processsolutions and some of the new understanding ofcopper plating bath additives and processchemistry that has been enabled by this newcapability. For a more complete description ofthe technology employed in the IPMS instrumentplease see Reference 6 from these proceedings.

THE COPPER ECD PROCESS

Unlike most semiconductor processes,electroplating takes place at atmosphericpressure, room temperature or below and in thepresence of an aqueous electrolyte. The wafer iseffectively immersed in an electrochemical celland acts as the deposition electrode. Prior to theECD process, a thin copper seed layer isdeposited across the wafer surface by eitherCVD or PVD. This seed layer acts as thecathode of the cell. A consumable copper anodeat the other end of the cell completes the circuitand provides the source for the Cu required tomaintain the Cu content in the plating solution.The current in the external circuit is a direct

measure of the copper deposition rate on thewafer. The conductivity and reliability of thecopper is dependent on void-free plating into thevias and trenches of the dual damascene processused in semiconductor manufacturing.

Organic "suppressor" and "accelerator"compounds are added to the copper plating bathto enhance gap filling by locally affecting platingrates. The suppressor adsorbs on the wafersurfaces and slows down Cu deposition in theadsorbed areas. The accelerator competes withsuppressor molecules for adsorption sites andaccelerates Cu deposition in the adsorbed areas.

During electroplating, both the suppressor andaccelerator are consumed at the wafer surface butare being constantly replenished by diffusionfrom the bulk electrolyte. However, differencesin diffusion rates of the suppressor andaccelerator components result in different surfaceconcentrations of suppressor and accelerator atthe top and bottom of wafer features, resulting incorrespondingly different plating rates at theselocations. Ideally plating rates should besignificantly higher at the bottom of thesefeatures for bottom-up fill. Appropriatecomposition of suppressor and accelerator in thebulk electrolyte are required to achieve theplating rate ratios required for the bottom-up fillthat yields a void-free via or trench. In one studythe acceptable process window was in the rangeof 7.5-12.5 ppm for the accelerator and 50-80ppm for the suppressor [1].

It has been found that the concentrations ofsuppressor and accelerator additives decreasewith bath usage and possibly with bath aging.These additives must be replenished in order tomaintain bath composition and the quality of thedeposited copper film. Organics are consumedby 1.) incorporation in the deposited film, 2.)attachment to the wafer as it is removed fromsolution ("drag out") and 3.) chemicalbreakdown into byproducts.

Chloride is also quite important to the reaction.In its absence, the organic additives havereduced or no effect, whether suppressing oraccelerating the reaction. Chloride depletion alsooccurs due to mechanisms similar to thosediscussed above.

In summary, the process window forsuccessful void-free Cu electroplating of sub-0.25 um high aspect ratio structures required incurrent semiconductor manufacturing requiresthat the overall bath chemistry and additives bekept within a very tight range. As criticaldimensions decrease, this range will tighten.Improvement in Cu ECD bath chemistry control

505

using direct in-situ quantitative measurement isexpected to significantly increase the yields andreliability of product which incorporates coppermetallization. Better understanding of processchemistry will also result in a reduction inunexpected process shutdowns, reduce modulequalification and requalification time andquantify chemical replenishment andreplacement needs.

CHEMICAL CONSTITUENTANALYZER (CCA)

As mentioned above the IPMS metrologytool, Figure 1, is described more completely inanother paper in these proceedings, Reference 6.This paper discusses a version of the instrumentthat has been optimized for chemical constituentanalysis (CCA) and applied to the measurementof Cu BCD process chemistry under actual waferprocessing conditions.

FIGURE 1. IPMS CCA Metrology System

Sinks Sample Handling _ Mass_Sp_ectrom_eter_

FIGURE 2. Block Diagram of the IPMS System andInterface with Typical Fab Process Wet Station

3. Full Class-1 clean room compatibilityincluding the use of materials resistant to thecorrosive action of the process chemicalsunder analysis and full compatibility withexisting Fab data systems.

4. Use of an electrospray ionization system,Figure 3, which operates at roomtemperature and enables molecular as wellas elemental analysis by automaticadjustment of operating parameters ("soft"vs. "harsh" ionization mode). Thiscontrasts with the low and high temperatureplasma ionization processes used in ICP-MS(inductively coupled plasma mass spec). Inthese plasma processes the majority of themolecular species are broken down toelemental components or small molecularfragments destroying most if not all thechemistry information.

Briefly recapping the IPMS tool featuresdescribed in Reference 6:

1. Automated sample extraction from theprocess bath and transfer to the metrologytool, Figure 2. The tool can be located up to30 meters away from the process bath whilesampling up to four baths.

2. Automated sample preparation including theaddition of stable enriched isotope solutionfor in-situ quantitative measurementcalibration.

I MHeated Drying Gas

(N2)

FIGURE 3. Electrospray Ionization Interface

5. In order to analyze the masses of the ionizedspecies, the IPMS system is configured withboth hexapole 'Ion-Trap' and 'Time-of-Flight' technologies (Figure 3 and 4). While

506

6.

each of these mass spectrometers have beenused independently as instruments for mass-analysis, and indeed the integration of thetwo in series is rare in commercial systems,this hybridization affords resolution andsensitivity unattainable by either alone.The time-of-flight mass spectrometer (TOF-MS) permits each isotope to be measuredsimultaneously and offers much higherresolution than standard quadrupole massspectrometers.

of -i efih« to

Ihs tons and«t Ifit

Tui?&

FIGURE 4. Time-of-Flight Mass Analyzer withReflection Ion Mirror and Detector

The key modifications required for the CCAapplication include sample preparation moduleswhich enable sample dilution, routine operationin the "soft" ionization mode and software whichsupports the analytical protocols and dataanalysis routines appropriate for chemicalconstituent vs. trace element analysis.

Quantitation of Chemical Species

Classically, each measurement is matchedagainst a standard calibration curve. Undernormal operation, a mass spectrometer will'drift' and become unstable. Constant monitoringand recalibration are absolutely necessary. As aconsequence, samples of varying matrixcharacteristics cannot be analyzed collectively.However, the CCA tool, employing IPMStechnology, applies the principles of isotopedilution mass spectrometry (IDMS) andspeciated isotope dilution mass spectrometry(SIDMS) [7-12] in an automated fashionenabling accurate, unattended quantitativeelemental and molecular species measurements.

All but four naturally occurring elementshave stable isotopes. These isotopes exist inknown ratios, set at the time of earth's formation.A mass spectrometer measures the mass of eachof these isotopes for each element. For example,Cu has two stable isotopes at mass 63 a.m.u. and65 a.m.u., while Cd has eight stable isotopesfrom mass 106 a.m.u. to 116 a.m.u.

By employing on-line mixing, a knownvolume of the 'unknown' sample is mixed withan artificially enriched 'known' isotope 'spike'standard. A composite isotopic ratio can then bemeasured on a mass spectrometer for the elementof interest. The amount of enriched isotopeadded and the concentration of the naturalmaterial present establishes a unique isotopicratio that enables the derivation of the 'unknown'concentration for the element (Figure 5).

Contributions tosignals from the

Contributions tosignals from the

FIGURE 5. Mass spectrum of a sample 'spiked' withan artificially enriched isotope.

Therefore, no longer is quantitationdependent on a calibration curves which areobtained in time consuming separatemeasurement procedures. The final ratio ofnatural elemental isotope to the enriched isotopeprovides an answer that has very few, welldefined possibilities for error. Each of thesepossibilities can be identified and compensatedfor, leaving the final error in the measurement -the uncertainty in ratio determination of the twoisotopes- a function of the mass spectrometer'sability to make this isotopic ratio measurement.

The IPMS technique is effective because foreach element, the element's isotopes respondidentically to the same physical and chemicalinfluences.

507

ANALYSIS OF COPPER ECDPROCESS CHEMISTRY

Sample Preparation and AnalysisProtocols

In this section we describe our first routineuse of the CCA tool for measurement of Cu ECDprocess chemistry. Samples were collected overa period of time from a Cu ECD process modulelocated in a development facility and being usedto electroplate wafers in "production mode."

The samples were automatically diluted withUPW by a factor of between 102 to 106 beforeintroduction into the mass-spectrometer. Eachmass spectrum is the accumulated sum of 60,000scans, requiring a total time for each spectrum of~2 minutes. The images shown are the averageof 5 spectra.

> Positive >

ipper |p! Chloride

SulfateSuppressor

Sodium Accelerator pPS-3H]+

FIGURE 6. The CCA can detect both cations andanions in the ECD solution by reversing the polarity ofappropriate voltages in the mass-spectrometer.

The CCA was operated in both positive andnegative modes for anion and cation detection,Figure 6, and in speciated mode, Figure 7, fordetection of the process chemistry present in theproduction bath.

• 'Soft' mode - the molecular ions are imparted with a lowkinetic energy in the gas-phase.

• Gentle collisions minimize fragmentation and ligand loss.• The integrity of the solution-species is preserved:

Suppressor Leveler

FIGURE 7. CCA Speciation mode provides analysisof ECD solution organics and organic complexes

Results

Figure 8 presents the results obtained for thecommercial copper bath "Virgin MakeupSolution (VMS)" containing accelerator andsuppressor additives. The bath is new at thispoint and the organic additives are present at"normal" operational levels. We will show thatthe organic additives are consumed duringprocessing unless they are replenished.

llllllillli

= Copper sulfate cluster ions

FIGURE 8. Mass Spectrum of a commercial Cu-electrochemical deposition (ECD) solution consistingof "Virgin Makeup Solution (VMS)" which wasspiked with two organic additives (accelerator andsuppressor) to "normal" operational levels.

The speciation capability of the instrumentenables the identification of copper sulfate waterclusters present in the VMS production solution(without organic additives), Figure 9. The softionization or speciation mode enables thedetection of intact clusters of copper ions withsulfate and water by minimizing theirfragmentation in the mass-spectrometer. Thechemical formulae of the species are reliable,although the structural assignments arespeculative.

508

Ionic clusters of copper, sulfate & water (CuSO4 + H2SO4 + H2O)t-,

" 9*z:

o"

oen£-

: 3~

fMX

4,-̂Cfs*X

en

C?V)X

^O ^

<N

w

X

1..4

4,"""en

o]X*

OHL,J£JC?z:

M

ON 331* O "

(M^+ r

e !̂

X""^sfi

c?enX

en

£L^

SM ^^= +

ft?

FIGURE 9. Speciation mode enables the detection of copper sulfate water clusters in the sulfuric acid based VMS CuBCD bath (no organic additives).

Copper Sample 65Cu"Spiked Sample

IllPfl

111111

iiiiiiiiiiiiiiiiii liiiiiiiiiiiiiii

> = organic fragments, such as C5H5+

Sample

VMS

0 hr

13 hr

35 hr

59 hr

71 hr

Analysis Number

i40.9

37.6

37.3

38.9

39.0

37.2

2

40.5

39.0

38.5

38.7

38.8

37.7

3

37.9

37.7

39.5

39.5

37.7

38.7

4

39.6

38.3

36.9

39.1

39.0

37.5

Copper cone.g/L

39.7

38.2

38.0

39.1

38.6

37.8

USD(%)

1,6

3,5

1,7

1,1

1,8

1,7

FIGURE 10. Quantitation of Copper in Cu-ECD Samples. The CCA is operated in elemental mode in this case.

A series of production process bathmeasurements were performed in order todetermine the chemistry changes that occur as afunction of wafer processing and aging time.There was no chemical replenishment in this

case. The quantitative analysis results forchanging copper concentrations as a function ofprocess time are presented in Figure 10. Byoperating in negative ion mode, similar resultswere obtained for chloride species.

509

In order to determine the quantitativeconcentration of accelerator and suppressor inthe production solution it is first necessary tosubtract the underlying spectrum produced bythe'VMS' component of the bath. Figure 11shows the results of such a subtraction.

The accelerator typical used is commonlycalled "SPS" and is more properly identified asbis (3-sulfopropyl) disulfide (CeH^S^e) . In

order to provide quantitative analysis of thiscomponent, each sample is automatically spikedwith enriched isotope standard as shown inFigure 12. A quantitative direct measurement ofthe degradation of the SPS concentration in CuBCD process solution which has been enabled bythis capability is shown as a function of processtime in Figure 13. There has been no attempt toreplenish the process solution as in the earliercases.

FIGURE 11. The 'Accelerator' & the 'Suppressor'spectrum after 'VMS' Subtraction.

FIGURE 12. Spiking of the "accelerator" or ("SPS") organic additive component in the Cu BCD solution withenriched isotope standards permits quantitative analysis of the SPS content.

510

FIGURE 13. Rate of accelerator degradation as a function of process time in Cu BCD measured in a developmentfacility. There has been no replenishment of the process solution

[(HS04)(H20)]

[Q3SS(C3H6)SQ3P~

[Q3S(C3H6)SQ3]2-(ffl/z = 100.980) [G3S(C3H7)]

123.011)03SG(C3H6)SQ3Piiiiiiiiiiiiiiiiiiiiiiiiii [03S(C3H5)]

FIGURE 14. Decomposition Species of the Accelerator SPS in an Aged Sample

The loss of accelerator is due to a combinationof factors including breakdown. Thecomponents of the breakdown processes areshown in Figure 14. The CCA technology isequally applicable for the quantitation of typical

suppressors, and for monitoring their breakdown.The appearance of the suppressor in the mass-spectrum in Figure 11 is typical of polyalkyleneglycol polymers commonly used in commercialCu BCD process solutions. We have chosenpolyethylene glycol of average molecular weight

511

3350 a.m.u. (PEG-3350) to exemplify, Figure15. This figure depicts its mass-spectrum inpositive-ion speciation mode following theaddition of sodium ions (as NaOH) in the samplepreparation module of the tool.

A Cu-ECD Suppressor Sample GPEG-3350):Characterization by formation of Na+ adducts

FIGURE 15. CCA analysis of organic suppressoradditive

The symbol z indicates the number of sodiumions chelated by each polymer chain, and hencethe resulting overall charge. Thus, for example, apolymer unit of mass 3350 a.m.u. that picks uptwo sodium ions gives rise to a peak in the mass-spectrum at m/z = (3350 + (2x23) ) / 2 = 1698,and similarly the adduct with four sodium ionsappears at 860.5.

FIGURE 16. These mass spectra depict the originalsuppressor additive (left), and the breakdown of thesuppressor in well-aged bath sample maintained bydaily bleed-and-feed (right). The latter shows theaccumulation of low-mass polymer fragments thathave occurred as the result of the breakdown of thelonger chain organic additive.

The suppressor components in aged Cu BCDsolution are easily recognized by the CCA tool.Figure 16 shows two mass-spectra; the first of

the suppressor additive, and the second of a well-aged bath sample maintained by daily bleed-and-feed, in which the accumulation of low-masspolymer fragments appears at the expense of thelonger chains in the replenishing solution.These solutions have also been treated withNaOH.

Breakdown of the suppressor in the Cu BCDprocess is characterized by cleavage of thepolymer chains through the ether functionality,yielding two shorter polyalkylene glycolfragment chains. In turn these breakdownspecies undergo further chain shortening.

In summary, the IPMS CCA analyzer hasbeen successfully used to identify and providequantitative measurement of all key elementaland organic components present in Cu BCDsolution. Measurements of bath chemistry as afunction of bath usage and aging and have shownthe expected depletion of initial processcomponents as well as the formation of organicadditive breakdown components.

CONCLUSIONS

The IPMS CCA has been successfullyapplied to measurement of Cu BCD processchemistry as a function of time and usage. Thedata collected in this first applicationdemonstrate that the instrument is capable ofproviding molecular identification andquantitative measurement of the criticalcomponents present in commercial BCD baths(copper, chloride, sodium, and organicadditives). It also provides identification andquantification of the decomposition productsthat occur during processing.

The CCA capability enables a new in-situmetrology tool for Cu ElectrochemicalDeposition process control. It provides fullquantitative characterization and monitoringinformation. Sample extraction andpreparation are fully automated. The tool isoperator unattended and fully compatible withClass 1 cleanroom operation and Fab datasystems. It can be used in-line, real-time in thefab line or off-line in the sub-fab or fab chaseas a "lab-in-a-box". Information provided bythe tool enables automated and advancedprocess control opportunities that are nowbeing explored.

512

REFERENCES

1. Yezdi Dordi and Peter Hey. "AutomatedChemical Management for Production CopperElectroplating," Semiconductor Fabtech, llthEdition, June 1999, pp. 273-276.

2. Peter Bratin, Gene Charyt, Alex Kogan, MichaelPavlov and James Perpich, "Control ofDamascene Copper Processes by CyclicVoltammetric Stripping; Semiconductor Fabtech,12th Edition, July 2000, pp. 275-280.

3. Stafford, G., T. Moffat, V. Jovic, D. Kelley,. J.Bonevich, D. Josell, M. Vaudin, N. Armstrong,W. Huber, and A. Stanishevsky, "CuElectrodeposition for On-chip Interconnections,"Characterization and Metrology for ULSITechnology: 2000 International Conference, D.G. Seiler, A. C. Diebold, T.J. Shaffner, R.McDonald, W.M. Bullis, P.J. Smith, and E.M.Secula, Editors, AIP Conference Proceedings550, Melville, New York, pp. 402-411.

4. Mary Havlicek, Anthony Schleisman and JohnDegenova, Recycling Spent Copper-Plating BathSolutions By Treatment With Activated CharcoalAnd Ozone, Semiconductor Equipment andMaterials International, see web sitewww.semi.org.

5. H. M. Skip Kingston, Ye Han, Rudy W. Mui, andKarin E. Rosen , Real-time, Unattended, TraceContamination and Chemical Species Analysis ofSemiconductor Cleaning Solutions, SemiconSingapore, 2002.

6. Skip Kingston, Robert McDonald, Ye Han, JasonWang, June Wang, Michael West, Larry Stewart,Bob Ormond, and Rudy Mui, "Automated andNear Real-time, Trace Contamination andChemical Species Analysis for the SemiconductorIndustry" Proceedings of the 2003 InternationalConference on Characterization and Metrologyfor ULSI Technology (This Conference, To BePublished by the American Institute of Physics,Melville, New York).

7. Dengwei Huo, H. M. Skip Kingston, and BretLarget, "Application of Isotope Dilution inElemental Speciation: Speciated Isotope DilutionMass Spectrometry (SIDMS)." Chapter 10, InElemental Speciation New Approaches for TraceElement Analysis, Eds. Joe Caruso, K. L. Sutton,K. L. Ackley, Elsevier, NY, pgs 277-310, 2000.

8. US EPA Method 6800, "Elemental and SpeciatedIsotope Dilution Mass Spectrometry", SolidWaste Manual - 846, Update IV, US GovernmentPrinting Office, Washington DC, 1998.

9. Kingston, H. M. "Method of Speciatiated IsotopeDilution Mass Spectrometry," US Patent Number5,414,259(1995).

10. Kingston, H. M., "Speciated Isotope DilutionMass Spectrometry of Reactive Species andRelated Methods," International Patent NumberWO99/39198A1(1999).

11. Kingston, H. M., 'Automated In-Process Isotopeand Mass Spectrometry" International PatentNumber WO 02/060565 Al (2002).

12. Anderson, M. R., Stewart, L. N., Kingston, H.M., "Method and instrument for automatedanalysis of fluid-based processing systems", USPatent and Trademark Office,US20030013199A1, (2003).

513