DEFECT REDUCTION AND YIELD ENHANCEMENT S …

1 As the semiconductor industrycontinues to expand design and processboundaries, especially at cutting-edgedevice nodes below m, IC manu-facturers and capital equipment vendorshave been under considerable pressureto develop new and cost-effective pro-cess methods to meet the new transition from 200- to 300-mmwafer processing has compounded thisproblem. New tool sets must be devel-oped, and either 200-mm processes mustbe adapted to the demands of 300-mmmanufacturing or completely new pro-cesses must be designed. Innovative ap-plications that offer IC manufacturersvalue-added solutions must be devel-oped as the demands of semiconductorprocessing the many challenges facing thesemiconductor industry are those asso-ciated with postash poly-mer residue removal inboth front-end-of-line(FEOL) and back-end-of-line (BEOL) process-ing. In BEOL processing,the removal of residuefrom both metal linesand contact/via struc-tures is critical. This arti-cle discusses research conducted by SEZon the development of a novel inorgan-ic chemistry for removing postash poly-mer from BEOL Chemical CostsThe main impetus for exploringwhether inorganic chemicals can beused to remove polymer residue fromBEOL structures came from a major ICRemoving postashpolymer residue fromBEOL structures usinginorganic chemicalsLeo Archer and Sally-Ann Henry, SEZ America; andDave Nachreiner, KantoA research project investigates the use of a sulfuricacid/peroxide/hydrogen fluoride mixture in placeof organic chemicals to remove postash polymerresidue from BEOL REDUCTION AND YIELD ENHANCEMENT STRATEGIES FOR SEMICONDUCTORS AND ADVANCED MICROELECTRONICS manufacturer in the process of shifting from 200- to 300-mmDRAM production.

2 The manufacturer s primary concern washow to make the transition to 300-mm production whilemaintaining costs at or close to existing 200-mm levels. Thatis no easy feat, given the initial capital construction and equip-ment costs of shifting to 300-mm production. As part of anoverall cost- REDUCTION strategy, the company has evaluatedhow to decrease manufacturing costs, which requires a re-duction in both FEOL and BEOL costs per die and, whereverpossible, an increase in effective cost breakdown in Figure 1 reveals that fabs spendmore money on chemicals than on any other items of dailyuse. Although several successful chemicals are available forpostash residue removal, all of them have similar problems:First, the use of organic compounds in residue-strippingchemicals increases production and disposal costs. Second,organic chemicals are environmentally unfriendly, prompt-ing many IC manufacturers to find ways to limit their , the vast majority of residue-stripping chemicals onthe market are proprietary, which can lead to higher costsand supply bottlenecks.

3 Because many such products areavailable from only one vendor, fabs can be crippled whensupplies run short, as occurred with hydroxylamine last issues have spurred the search for a complete postashresidue application combining spin processing and a non-proprietary dilute inorganic mixture containing chemicalcomponents readily available in any a DSP Spin-Processing Technique toRemove Postash ResidueSome companies have attempted to use dilute inorganicacid mixtures to remove polymer ,2In one case, amanufacturer tried to use a conventional wet bench to re-move polymer from both metal and contact/via structureswith a dilute sulfuric acid/hydrogen peroxide (DSP) mixturecontaining parts-per-million concentrations of hydrogenfluoride (HF). It was discovered that the effectiveness of thismixture depends on HF concentration. When used in a wetbench, there is only a 2-ppm process window for HF con-centration. While insufficient cleaning takes place below thelower control limit, excessive etching of the underlying ma-terial takes place above the upper control limit.

4 Both of theseeffects result in poor electrical the difficulties of using DSP on a wet bench, the re-sults of these experimentsshowed promise. After furtherinvestigation, it was decidedto try DSP on a spin processorto determine whether spin-processing fluid dynamicsavoid the problems associatedwith the classical wet work performedat SEZ s Phoenix research laband at customer sites demon-strated that a special DSP mix-ture known as DSP+ fromKanto (Portland, OR) can suc-cessfully remove postash poly-mer residue from wafers andthat when used in a spin pro-cessor, the process window ofthat mixture is significantlylarger than when used on a wetbench. When HF concentra-tions of between 5 and 1000ppm were investigated duringpreliminary work, it was de-termined that the allowableSURFACECONDITIONING/CHEMISTRIES F igure 1: Breakdown of running costs associated withsemiconductor manufacturing ULTRAPURE WATER-5%BULK GASES-4%POWER-27%OTHER-2% SPECIALTY GASES-6%56%2%4%5%6%27%Figure 2: Schematic drawing of the spin-processor CHAMBERETCHANT 3 ETCHANT 2 ETCHANT 1 DRAINN2N2DI WATEREXHAUSTLEVEL 4 LEVEL 3 LEVEL 2 LEVEL 1HF concentration on the spinprocessor is about two ordersof magnitude larger than thaton a wet andProcess DevelopmentThe process chuck of thespin processor used to conductthe HF tests relies on Bernoul-li s principle to fix the wafer ata constant distance from thechuck surface on a bed of ni-trogen (N2).

5 The wafer is heldin place by six edge-contact-only pins that make contact atthe wafer bevel with sufficientforce to center the wafer on theN2bed and hold it in placewhile the chuck rotates. Thechuck rests in a process cham-ber, as depicted in the sche-matic drawing in Figure process chamber canhave up to four independentprocess levels, three of whichdispense different processchemicals (or chemical blends)and one of which is dedicatedto DI-water rinsing and nitro-gen drying. The process chuckrotates clockwise or counter-clockwise within the processchamber while the medium isdispensed. The different chemistries are dispensed onto aspinning wafer at three dedicated process levels, allowing fortight process control and eliminating the risk of on the use of inorganic chemicals for postash poly-mer removal has been conducted at facilities in Europe, whileongoing process development, including experiments withthe DSP mixture, has been carried out at SEZ (Phoenix) andIC fabs in the United States.

6 Process development has in-volved a variety of different BEOL device structures fromseveral manufacturing partners. Wafer samples were provid-ed by major and European tests performed on a 200-mm spin processor oper-ating at room temperature focused on the cleaning of metallines. These tests quickly revealed that it is possible to cleansuch structures in 30 seconds in a single-step process. Figures3a and 3b are scanning electron microscope (SEM) images ofan unidentified standard metal structure before and aftercleaning with the DSP mixture. The large amounts of postashresidue polymer evident in Figure 3a, especially on the largeexposed areas off the metal line, are no longer present inFigure cleaning results were observed on other aluminum/copper structures. Figures 4a and 4b are SEM images of ametal 6 stack before and after exposure to the DSP this example, the residue was removed in 25 seconds with-out the aluminum lines being DSP mixture also has proven successful in cleaningcontact/via structures.

7 However, the process protocols usedon such structures are very different from those used on metalstructures. Initially, contact/via structures were processed forup to 120 seconds, with less than conclusive results. Conse-quently, a series of design-of-experiment (DOE) process runswere performed to determine the optimum process condi-tions for cleaning such structures. The SEM images in Figures5a through 5d present a synopsis of some of these experi-ments conducted on contact/via 6 structures, the most diffi-cult structures to clean. The image in Figure 5a shows such astructure before residue cleaning, while the images in Fig-ures 5b, 5c, and 5d depict different structures processed underdiffering conditions after the initiation of residue on the knowledge gained from these experiments,further optimizations were performed in facilities in Europeand Asia. Figures 6a and 6b are images of a contact/viaFigure 3: SEM images of a standard metal structure (a) before and (b) after cleaningwith the DSP 4: SEM images of a metal 6 stack (a) before and (b) after exposure to the DSPmixture.

8 (a)(b)(a)(b)structure cleaned at SEZ s European lab in Villach, structure was processed in 90 seconds. Although theprocess layer is unidentified, the conditions under which itwas processed are typical for contacts/vias 1 through experiments demonstrated that the application ofthe DSP mixture on a spin processor removes postash residuefrom wafers. However, because a variety of commerciallyavailable products also can be used on spin processors to re-move polymer residues, it was necessary to compare the ef-fectiveness of the DSP mixturewith that of existing residuestrippers containing organicchemicals. Although furthertesting is being conducted,some preliminary findings determine the superiormethod for cleaning metalstructures, first the DSP mix-ture and then a proprietaryammonium fluoride basedchemistry were used to stripwafers in a spin processor attemperatures ranging from 20 to 40 C. All other process con-ditions remained essentiallythe same throughout the ex-periment.

9 The results of thetest are shown in Figure 7. Fig-ure 7a is an SEM image of ametal structure before clean-ing, Figure 7b is an image of ametal structure after cleaningwith the DSP mixture, andFigure 7c is an image of ametal structure after beingcleaned with the ammoniumfluoride based the test demonstratedthat both chemistries can beused successfully on a spin-processing system to cleanmetal structures, the chem-istries processing times differmarkedly. The DSP mixturecan remove surface residues inhalf the time it takes for theproprietary ammonium fluo-ride based chemistry to re-move electron mi-croscopy is typically used as afirst-pass metrology test to de-termine the effectiveness of techniques for removing postashresidues. However, the only truly reliable method for deter-mining the effectiveness of cleaning techniques is to test theelectrical characteristics of processed structures. Electricalcharacterization has been used extensively throughout thedevelopment of the DSP+ application.

10 The devices shown in Figure 7 underwent early electricaltests, the results of which are presented in Figure 8. The snakecontinuity (resistivity) measurements shown in that figureSURFACECONDITIONING/CHEMISTRIESF igure 5: SEM images of contact/via 6 structures: (a) before cleaning; (b, c, and d)different structures processed under differing conditions after the initiation of 6: SEM images of a contact/via 6 structure cleaned with the DSP mixture for 90seconds.(a)(b)(c)(d)(a)(b)demonstrated that the use of the DSP mixture produces struc-tures with lower resistivity than the hydroxylamine-basedprocess of record (POR). While the DSP mixture seemed toperform slightly better than the proprietary ammoniumfluoride based chemistry, the results were not yet statistical-ly significant. However, the process throughput of the DSPmixture was greater than that of the proprietary ammoniumfluoride based (resistivity) data from subsequent, more com-prehensive electrical tests on contacts/vias 1, 2, and 6 andmetal 6 structures cleaned with the DSP mixture are typical-ly comparable to or better than those from similar structurescleaned with the POR hydroxylamine-based chemistries(within 3 error).