EUV光刻中的随机效应

This paper focusses onstochastic printing failures,suchas microbrid es in spaces or randomly mis s ing contacts. Wequantify such failures with a metric we call NOK (not OK), a metric that essentially rep resents the failure pro bab ilityWe measure this NOK-quantity from SEM image analys is, using an in-house software package called Stochalis . We wilargue that the mo st fundamentaldependency ofthis failure probability is its CD dependency: NOK/CD). Us ing currentlavailable CD-SEM ore-beam inspection tools, it is now possible to measure this NOK(CD)de pen de ncy down to theppm-ppb level This is not suficient to prove or disprove yield, but the NOK(CD) function is an excellent tooltocompare materials and conditions, ie. to quantify sensitivities to pattering conditions and show directions forimprovement. We will illustrate this with examples on the impact ofdose, pitch,resist, etch and llumination mo de. Wwill also showhow CD non-uniformiies. from very localto global, further aflect the local failure probabilities. Finallwe will argue that stochastic failure probabilities and CD-non-unifomities together place practical re sol ution limitsboth on CD and on pitch. These limits are not absolute (whch is whv we call them 'practical lmits) as thev depend orthe patteming settings and materials used, but they nevertheless need to be considered very carefully wh en se ting upEUVl processes.

1.INTRODUCTION

The term "stochastic effects in lithography' refers to random, localvarability that occurs between structures that shoulin principle print identically.The "trad itiona"  WR and LCDU metrics guantify the local CD-variability that is a resultofthis. But far more serious than local-CD Variability are local Printing Failures such as microbridges in spaces ormiss ing contacts, as they directly affect yield. Note that we prefer to use the term "stochastic printing failuresratherthan “stochastic defectsto emp hasize the fact that mo st ofthese failures are generate d in and by the pa tern-formationprocess itse!",and are not related to e.g, particle-contaminations, ie. to emphasize the diference with "traditionalsources ofdefectivityeven though these ofcourse contnueto exist). Such stochastic printig lailures start to occurwhen pattern dimensions become very small. Micro bridges.forexample. start occurring when the width ofthespaceinres ist is small: localline breaking when the width ofthe resistline is small This type offailures has always exis ted inithography,but untilrecently they occu rred sufficiently faraway from the target dimension. and s ufliciently far from thenominalprinting conditions, notto be a majorconcern. With the introduction ofEUV, target dimensions havemade abig jump to smaller values, which means that stochastic failures now occurmuch closer to or even at the CD-target andat nominal conditions. So the study ofsto chastic failures. at whch CDs thev start to occur. what they depend on. anchence howthey can be minimized, then becomes a matter ofmajor importance. This is the topic ofthis paper.

Stochastic failures are notexclus ive to EUV: e.g. miss ing contacts and microbrid ges are kno wnin immers ionlithography as well, and we expect that the same typeofstudies as the ones described in this paper will need to be donesoon at anywavelength when CD-limits are being pushed to the limit. But the large EUV-photon energy , and thelowab sorbed) photon density, is one (thoughby farnot the only)reason why stochastic failures are more prominent in EUvlithography. Theroot cause for stochastic effects is the particle nature ofmatter and light (photons) and the fact that onthe atomic/molecular scale interact ions havea probabilis tic nature'.4.5, The lowabsorbed pho ton density is then thechronologically ) first- but by no means only -cause for randomn ess in the developed pattern. All this should be welknown and acceptedby now.

In this paper we wil first briefly remind the metric we have introduced to guantify stochastic failure, a metric calledNOK (not OK). Then we will study what we considerto be the most importantdependency, being the CD-dependency oiNOK. This NOK(CD) dependency is what deterines which CDs or CD-range can be printed with yield and is thereforeofutmost importance. Afterwards we illustrate the impact ofsome ofthe imaging or processing conditions on thisNOK(CD) dependency,and then discuss the importance ofCD-non-uniformities. Finally , we wll return to question hovstochastic failures impose a practical limit to what are the minimum CD and pitch that can be printed with yieldsomething which we call the “practical resolution limits”(CD and pitch), at the given imaging- and processingconditions. These resolution limits are not absolute, as they can be improved ifimaging- or process-conditions areimproved, but fora given process, they are realenough, and its essentialto know where they are.