电镀铜的优化表面预处理

Abstract

We investigated the ability of conventional pretreatment systems to eliminate the surface native oxide from copper sputter-coated TiN/Si (1 00) prior to copper electroplating. The cuprous oxide on copper seed layers was removed by surfacepretreatment processes, which resulted in uniform and smooth electroplated copper. Sheet resistance measurements and AFManalysis revealed that both the wet cleaning method. using 1.200 NH OH solution. and the coulometric reduction method wereeffective at removing the copper native oxide, which are mainly cuprous oxide on a (2 00) copper seed layer, In particular, thecoulometric reduction method offers an in situ pretreatment process, which is self-limiting in terms of etch. Copper filmselectroplated on such pretreated seed layers were thicker, smoother and had finer grains than those formed on artificial copperoxide layers.

Introduction

Copper is being considered as an alternative foraluminum as an interconnection material in highperformance integrated circuits (ICs) metallizationprocess due to its low bulk electrical resistivity[1-5], and high resistance to electromigration andstress migration [6]. Besides the plasma vapor deposi-tion (PVD) and metal organic chemical vapor deposition (MOCVD), copper interconnections can also befabricated by electroplating, and this is believed to bean excellent copper forming method for damascene structures because of its excellent via/trench fillingability, good adhesion, and lower process temperatureand cost [4,7,8].

The electroplating is performed by reducing copperions to metallic copper on a thin copper seed layerdeposited on a TiN-coated Si (1 0 0) substrate. Thisseed layer then provides a conduction path for theelectroplating current. However, pretreatment of theseed layer is necessary to form an electrically uniformsurface, and this is recognized as one of the mostimportant issues in the development of a suitablecopper electroplating method.

Current transport mechanisms through the metaloxide film can be considered to be of two types, whichare dependant upon the oxide film thickness. Heuslerand Yun [9] showed that tunneling is the dominantcurrent transport mechanism through copper oxide films less than 30 A thick, and that current densityfalls exponentially with increasing film thicknessIf the oxide film is thick enough, current transportthrough the oxide layer is achieved by conductioninvolving carriers. Using a simple semiconductormodel, Dogonadze [10] showed that the currentthrough a metal oxide electrode is dependent on theoxide film thickness. Regardless of the mechanism ofcurrent transport, current flow through an oxide film isstrongly decreased with increasing the oxide filmthickness, i.e., both the tunnelingprobability forthe thin oxide layers and the carrier (excess metallicions in CuO and excess holes and vacancies in CuzO[11-14]) densities for thick oxide layers are functionsof oxide thickness. Thus, areas covered by surfaceoxide resist the transfer of surface charge, and there-fore, copper oxide should be removed to ensure con-sistently uniform and densely electroplated copperfilms.

In this study, we investigated the effect of remov-ing the oxide from the copper seed layer on thecharacteristics of the electroplated copper. NH.OHcleaning and coulometric reduction (CR) wereused to remove the copper oxide, and atomic forcemicroscopy (AFM), field emission scanning electronmicroscopy (FESEM),X-ray photoelectron spectro-scopy (XPS) and X-ray diffraction analysis (XRD)were used to verify the effect of the pretreatmentson the electroplated copper.

2.Experimental

CR was conducted in an electrolyte prepared byadding 56 ml of HSO4 (95%) to 1 1 of deionized (DI)water and cooling to room temperature. Sputter depos-ited copper on a silicon wafer with the structure Cu(700 A)/TiN (400 A, MOCVD)/Si was used as thecathode. This wafer was exposed to air for 7 daysto grow the native oxide, and 1 cm’ of its surface wasthen exposed to the electrolyte. For coulometric oxidereduction an electronic grade copper bar (l mm indiameter, 3 cm in length) was used as the anode, and asaturated calomel electrode as the reference electrode.The constant current used for the copper oxide reduction was supplied by an EG&G Princeton AppliedResearch Model 263A potentiostat/galvanostat unitinterfaced with a personal computer. The CR current was varied between 10 and 100 uA, and the responsepotential was measured for 300 s using M270 softwareand the potentiostat unit. Chemical pretreatment of thecopper native oxide was performed by soaking thewafer in a 1:200 mixture of 30.0% ammonium hydro-xide and DI water for half-a-minute. To remove theresidual ammonium hydroxide from the wafer surfaceafter wet cleaning, wafers were dipped in 100 ml of Dlwater for 20 s with stirring. This procedure was carriedout three times and the thoroughness of the rinsingmethod was verified by measuring the amounts ofresidual ammonium hydroxide in the rinsing solution(DI water) by UV spectroscopy. Electroplating wasperformed in a solution containing 13.2 g CuSO4.56 ml H,SO4 and 1 1 DI water. Typical copper elec-troplating solution contents are described in numerouspapers [1,15-21]. The electrode system used for electroplating was identical to that used for CR. Electricalcontact to the wafer during the electroplating wasmade using a copper plate with a 1 cm x 1 cm window connected to the potentiostat unit. The front face of thewafer was set on the window and fixed in position, andthe wafer was exposed to the electrolyte through thiswindow. An insulating coating ofTeflon on the outsideof the copper plate, but excluding the window, pre-vented the plate being exposed to the electrolyte.

Plating wascarried out at constant voltage(-400 mV) for 400 s on three substrates: (i) copperseed layer pretreated with NH4OH; (ii) seed layerpretreated by CR; (ii) artificially oxidized unpretreated seed layer, which was formed by oxidizingthe copper seed layer in a 1:l mixture of 35% HOand DI water to observe the effects of surface copperoxide on plated copper characteristics. CR pretreatedwafers were immediately electroplated after pretreatment, and NH.OH pretreated wafers were rinsed withDI water and dried in a continuous N2 stream beforeplating. After each electroplating run, all the depositedsurfaces were rinsed and dried, and their surfaces werecharacterized by AFM, SEM, XRD and a four-pointprobe station, which was used to determine surfaceresistance. It was evident from the sheet resistancemonitoring results that the electroplated copper layerwhich was free of organic additives, such as levelersand brighteners, showed no self-annealing phenom-enon. Therefore, we believe that negligible morpho.logical change occurred between deposition and thetime spectroscopic measurements were taken.