The combination of a reflective photomask with the non-telecentric illumination and arc shaped slit of the EUV scanner introduces what are known as shadowing effects. The compensation of these effects requires proper biasing of the photomask to generate the intended image on the wafer. Thus, the physical pattern on the mask ends up being noticeably different from the desired pattern to be written on the wafer. This difference has a strong dependence on both the illumination settings and the features to be printed. In this work, the impact of shadowing effects from line and space patterns with a nominal CD of 16nm at wafer was investigated with particular focus on the influence of pattern orientation and pitch, illumination pupil shape and fill (coherence) and absorber height. CD, best focus shift and contrast at best focus are utilized in detail in order to study the impact of the shadowing effects. All the simulation cases presented employ a complete scanner arc emulation, i.e. describe the impact of the azimuthal angle component of the illumination arc as in the NXE:3300 scanner and as it can be emulated by the AIMS™ EUV. KEYWORDS: EUV photomask, shadowing effect, EUV scanner, illumination slit, compensational repair, EUV absorber, photomask bias, defect inspection, AIMSTM EUV.
INTRODUCTION
The technological step required by the introduction of EUV lithography into high volume manufacturing can be considered the most complex within the development of photolithography over the last several decades. Waiting for the whole infrastructure to be available and make EUV a reality, some of the most challenging characteristics can be modelled and studied with the support of simulation platforms in order to achieve the best understanding of the EUV lithographic process and the interplay between the several parameters which describe it. One of the most critical of these aspects is related to the manufacturability of the EUV photomasks, a complex reflective optical component, whose reflectivity is based on the in-phase addition of reflections coming from each boundary of a Molybdenum (Mo) Silicon (Si) multi-layer (ML) structure. The fact that the EUV photomask is a reflective optical component, as are all optics in an EUV system, requires the illumination of the features printed on the mask to be non-telecentric. In this scheme EUV illumination of the photomask is not normally incident onto the reticle, but has an offset inclination described by a chief ray angle (CRA) of 6 degrees. Depending on the shape of the pupil and upon its coherence, the mask is illuminated with a variety of angles distributed around the CRA. The reason behind this oblique illumination is that the incident ray bundle must be physically decoupled from the one which is reflected from the mask and collected into the projection optics system. Oblique illumination of three-dimensional mask structures introduces a new aspect with respect to the EUV lithographic process, known as shadowing effects. The overall effect produced by the absorber onto the mask reflective structure and back to the projection optics is a complex combination of angular illumination efficiency and mask reflectivity, meaning that the illumination conditions strongly influence the final printed results. To make the picture even more complex, exposure of EUV photomasks in a high volume manufacturing (HVM)
scanning system (i.e. ASML NXE:3300) is performed through an arc shaped slit which spans a linear size perpendicular to the scanning direction of approximately 100 mm, in which the pupil is kept on axis with the scanning direction throughout the whole illumination field. The three-dimensional direction of an EUV photon can therefore be described by two angles: the CRA θ, which remains constant through the whole illumination domain, and the azimuthal angle φ, which can vary in the range ±18.6 degrees. Targeting the complete emulation of the scanner imaging process, the AIMSTM EUV platform matches this sophisticated exposure scheme, although engineered to target a much smaller field1. The arc shaped illumination setup has to be taken into account during mask design. Clear and opaque features are illuminated with different sets of angles depending on their x coordinate on the photomask; as a result, the aerial image of the same structure varies across the x direction. Heavy optical proximity corrections (OPC) must be studied and applied to the mask structures in order to print at target across the whole exposure field; as a consequence, the physical pattern on the mask ends up being noticeably different from the desired pattern to be written on the wafer. This is an application where AIMSTM EUV will play a leading role as actinic review of EUV photomasks is able to thoroughly determine the overall printing behaviour of biased EUV photomasks. In this work, the printing process of EUV scanners is discussed with respect to shadowing effects. Special attention is devoted to the impact that the arc shaped illumination has on the shadows produced by the EUV photomask. The structures which have been investigated are lines and spaces with vertical (V), horizontal (H) and 45 degrees inclined orientation, with a 16 nm (at wafer level) minimum half pitch; this can be considered as the 7 nm node, which is also the target for AIMSTM EUV. The main goal of this work is to highlight the complexity of EUV imaging in terms of mask biasing; more than for DUV lithography, the physical size of the structures on the EUV photomask will differ from a 4x replica of the wafer target. Actinic aerial image inspection will therefore assume a key role within the photomask production line. The following sections include the study of the shadowing effects and their dependence on structure pitch, pupil shape and coherence and absorber height. In order to best simulate the imaging process, the shift of the best focal position and contrast through focus are taken into account in order to emulate the workflow as employed in AIMSTM EUV.
DESCRIPTION OF SHADOWS Figure 1 presents a conceptual diagram of the shadowing effects introduced by non-telecentric illumination of EUV photomasks. The reflective portion of the EUV reticle is composed of ~40 bilayers of Mo-Si which provide a peak reflectivity at 13.5 nm wavelength of about 69%2,3. Each boundary between two subsequent layers contributes only to a small portion (~ 0.1-0.3%) of the whole reflectivity of the ML. This means that two EUV photons with the same energy and incidence angle can be reflected at different locations within the ML structure, one being reflected and channelled towards the projection optics, the other being absorbed by the back side of the absorbing structures on its way back out of the mask surface (see Figure 1). This simplified visualization of the reflection process gives a basic idea of the high complexity of the EUV reflectivity process. Additionally, due to the finite area of the pupil, the mask is illuminated with a distribution of angles centered around the CRA, all of which contribute in shaping the aerial image or the wafer print. EUV ML reflectivity is strongly dependent on the incident angle of the photon, and therefore different regions within the same pupil may give a different contribution to the imaging performance4. This short introduction is necessary to justify the use of simulation platforms as the only investigation method of shadowing effects and their effect on scanner imaging performance, before actinic imaging with AIMSTM EUV will be available. In order to give a quantitative description of the difference between a geometric calculation and an actual simulation result, the right panel of Figure 1 presents the formulas for the shadows which are produced on a EUV photomask: this quantity depends on the absorber height, the penetration depth into the ML structure (which together with the absorber height gives the effective height heff to consider for the shadow calculation), the CRA, and the magnification of the projection system. In order to obtain the results shown in the left panel of Figure 2, azimuth angles of 0±18.6 degrees are considered for H lines, 90±18.6 deg for V lines and 45±18.6 deg for 45 deg oblique lines.
Proc. of SPIE Vol. 9231 923109-2
Figure 1. Screated ondepths of Ethe effectivsimple ske In the pro
1. Ptt
2. Gpsfbt
3. Anca
The thirdthe CD vanm.
The resulposition a16 nm (atthe left pacase, withtarget CDimaging,
Sketch of the strn EUV photomaEUV photons. Tve absorber heietch of the lines
ocessing of an
Process bias, kthe target featto produce theGlobal bias. Tphotomask, spstructure depefeatures at tarbias is appliedthe center of tAzimuthal bianon-telentricitcomponent ofand structure
d component oalues measure
S
lting CD derivalong the imagt wafer level) anel of Figureh a disar 0.6-0
D of 16 nm of i.e. what in th
ructure of a typasks during imaThe parametersight heff (absorband spaces stru
n EUV photom
known from Dtures on the we mask (e-beaThis bias is nepecifically its endent, and it rget in the cend to the every the imaging fias. This bias isty of EUV sysf the CRA intrdependent.
of the photomaed from the sim
HADOW D
ved from the gging slit (azimand an absorb
e 2. The result0.9 as illuminathe plot repre
his paper is ca
pical EUV photoaging. The greens shown in the fober height d + puctures conside
mask, three dif
DUV lithograpwafer. This proam writing, etceded in order dependence ocan be thoughter of the scansimulation se
ield (φ=0), regs also needed stems at maskroduced by the
ask bias as desmulations pre
DEPENDENC
geometrical apmuthal φ angleber height of 5t of a first simuation pupil is sesents the impaalled azimutha
omask (left) togn and black pri
formulas (top rigpenetration depered throughou
fferent compo
phy. The physoduction procech…) and it wto compensat
on the θ compht as the ΔCD nner illuminatetup so that a lgardless for itsto take into a
k level. Howeve arc shape of
scribed abovesented in this
CE ON FEA
pproximation e within ±18.650 nm (heff=d ulation run ofshown in the ract of the azim
al bias. V lines
gether with the gnted space widght) are: the ab
pth), the magnifit the whole pap
onents are requ
sical mask feaess driven bias
will not be conte for the non-
ponent of the Cthat must be p
tion field. In thline of 16 nms orientation occount the shaver, the local Δf the illuminat
e will be the fowork at φ=0 w
ATURE OR
introduced in 6 degrees) for for the geome
f the same seturight panel of muthal angle cs at the edge o
geometrical desdths correspond bsorber height dfication M, and per is also show
uired to create
atures are not as takes into ac
nsidered in this-telecentric illCRA. This biaprinted on mahis work, an o(wafer level)
or the process adowing effecΔCD is depention field; this
ocus of this wowill always be
RIENTATIO
the last sectiolines and spac
etrical approxiup as describeFigure 2. The
component of of the slit print
scription of sha
d to different pend, the CRA θ (6the azimuthal a
wn at the bottom
e a proper bias
an exact 4x reccount the steps work. umination of
as is process aask in order toorientation depis printed at taused to image
cts introduced ndent on the φ
bias is also p
ork; as a conse the target CD
ON
on as a functioces with nomiimation) is sh
ed for the geome deviation frof the scanner at larger than ta
adows netration 6 degrees), angle φ. A
m right.
s process:
eplica of ps needed
the and o print the pendent arget in e it. by the
rocess
sequence, D of 16
on of its nal CD of own in metric om the rc arget CD,
Proc. of SPIE Vol. 9231 923109-3
whereas Htherefore 18.6 deg The overageometricrepresentwhereas fazimuthalarger thalevel). Thprint all obeen applneeded innegative blines needamongst t
Figure 2. LRight: sha This first proper bidesired pplatform verificatiolithographmanufactlithograph
A furtherdense andnm (duty lithographdiffer fromstructuresfirst been
H lines print smust be subjeand φ=18.6 deall behaviour cal calculationting V lines is for H lines it i
al angle for thean for the V linhis graph can aorientations aclied in order ton order to combias must be ad a negative bthe three cons
Left: Calculatioadowing effects
simulation seas and OPC stattern. Given will become eon. In the nexhic parameterturing of the Ehic process th
r lithography pd isolated linecycle 1:7) at
hy processes im each other.s with differen
n carried out.
smaller and thect to the moseg). within the samn, although dirounded and
is increased we 45 deg oriennes, with a difalso be read frcross the illumo print all thre
mpensate for thapplied to V li
bias at the φ=-sidered orienta
on of shadows presults for the s
erves as a remtructures to bethe importanc
even more indxt sections it wrs. The importEUV photomahat will be use
SH
parameter cones and spaces wwafer level, ait is known th In order to gint pitches, a st
he 45 deg obliqst severe bias (
me orientationfferences are nthe impact of
while keeping tnted lines is sufference betw
from the perspmination slit atee orientationshe scanner arcines across the18.6 deg edgeations.
produced for 1:same structures
minder of the ime applied to thce of bias strudispensable fowill be shown ttant consequenask must be dod to print the
HADOW DE
nsidered in thiswill be considand the three oat dense and iive a thoroughtudy of the shi
que lines print(up to 8 nm di
n (H-V-45 degnoticeable thr
f the azimuthalthe overall tre
ubject to the laeen the two ed
pective of the bt the target CDs separately at
c shaped illume mask, where
e, and a positiv
:1 V, H and 45 ds as given by a
mportance of he features prictures and OP
or proper scannthat the bias tonce, and fundaone with the pmask itself.
EPENDENC
s work is the pdered, with pitorientations deisolated structh description oift of the best
t larger to smaifference at m
g) is similar torough a closerl angle on the
end unaltered.argest differendges of the exbias to be appD of 16 nm. At 16 nm in the
mination mentieas H lines reqve bias at the φ
deg oriented lin0th order simula
simulation capnted on a EUV
PC in EUV lithner emulationo be applied tamental differreceding know
CE ON PITC
pitch of the strtch values betwescribed in thetures have besof the impact ofocal plane in
aller across thmask level betw
o that derived r inspection. T
printed CD isThe behaviou
nce: the CD prxposure field olied to the ma
Assuming a gloe center of the oned above. Iquire a positivφ=18.6 deg, b
nes via the geomation.
pabilities in thV photomask hography, the
n and OPC corto all features rence to DUVwledge of all t
CH
ructures printeween 32 nm (de last section. t focal planes of the arc shap
n response to a
he entire field, ween structure
through the The shape of ths noticeably reur of the CD thrinted at the edof about 2 nm ask features inobal bias has aslit5, a second
In this contextve bias. 45 degbeing the large
metric approxim
he calculationin order to pr
e AIMSTM EUVrrection and is a function o
V lithography, the parameter
ed on the phoduty cycle 1:1From current which notice
ped illuminatia pitch variati
and es at φ=-
he curve educed, hrough dges is (at wafer
n order to already d bias in t, a g inclined est in size
mation.
of the int the V
of several is that the s of the
tomask; 1) and 128
ably ion on on has
Proc. of SPIE Vol. 9231 923109-4
In order tthe centerprocess wfor each s
Figure 3
The left pthree oriefocus planconsidereabsolute mbest focushift (awaemployedisolated) This relatthe photowafer. Thstack is bdefocus. Iimpact ofwindow pnm6, the d~25% of process wAs a furthfocal pospitches ar
to calculate thr of the slit (φ
window and prsimulation set
1. Fit obe anthe f
2. Analcoorddefoc
3. Left: Shift of Right: A
panel of Figurentations consne towards need as differentmeasure of ths plane shift aay from the md in this work duty cycles. tionship is par
omask and, as his case does nby default centIn the case thaf the defocus aparameters. Adisplacement the entire bud
window may bher measure oition for the sire imaged wit
he shift of the b=0) have beenrovides the matup undertakin
f the contrast n even numbefirst even degrlytically find tdinate of this cus corrected
the best focal pAerial image co
re 3 shows theidered in this
egative defocutial measureme best focal pl
as a function omask) of the co
is measured t
rticularly releva consequenc
not however atered in the beat multiple difas it will be on
As the depth ofof the best foc
dget. Althoughbe, as a result, of the influencimulated test h a lower con
best focal plann considered.aximum imagng the followin
vs. defocus cuer in order to pree above whicthe maximum point is the besimulation se
plane (nm at retontrast through
e shift of the bwork. As the
us is observedment, having fu
lane. All line of the structureorresponding bto be 25 nm (m
vant to logic dce, scanning siapply to the Aest focal planefferent featuren the scanner f focus for thecal plane in reh within the deseverely redue pitch has oncases. As kno
ntrast. This hol
ne within the Assuming tha
ging contrast, ng steps:
urves with a pprovide symmch no improvepoint of the d
est focus shiftetup.
ticle level) throuh pitch for the sa
est focal planpitch of the st. The values f
ull meaning onorientations a
e pitch: an incbest focus posmask level) an
designs, whereingle exposureIMSTM EUV p
e, although maes are containewill be measu first generatio
esponse to theepth of focus,uced. n imaging, theown also from lds for all orie
simulations ruat the best focathe shift of th
polynomial funmetry with resp
ement of the fderivative of tht value to emp
ugh pitch for V,ame structures,
e for lines andtructure increafor the best focnly with respeare subject to creasing pitch sition. The largnd applies to V
e different pitce can not printplatform, wheanual operatioed within the Aured, giving a on EUV lithog
e change of the this is a subst
e right panel oDUV lithogra
entations, with
un, contrast veal plane is in t
his plane can b
nction of degrpect to the besfit residuals cahis best fit fun
ploy as input p
V, H and 45 deg also measured
d spaces as a fases, a noticeacus shift here
ect to one anotthe same genealways transl
gest shift for tV lines with 1
ches and orient all the featur
ere the throughon also allows AIMSTM EUVtighter contrography scannee structure pittantial portion
f Figure 3 shoaphy, patternsh H lines prov
ersus defocus the center of tbe calculated s
ree 6. The degst focal positioan be noticed. nction. The abparameter of th
lines, measuredd at φ=0.
function of pitable shift of threported are t
ther and not aeral behaviourlates into a negthe simulation:1 to 1:7 (dens
ntations are prres at best foch focus aerial to emulate th
V field of viewol on the proceer is larger thach can be as hn of the budge
ows the contras with increasividing in gener
curves at the separately
gree must on. Six is
bscissa he
d at φ=0.
tch for all he best to be s an r of the gative
n cases se-
rinted on us on image
he scanner w, the ess an 100 high as et and the
ast at best ing ral a
Proc. of SPIE Vol. 9231 923109-5
higher coorientatiolines with The defocimpact oforiented lsemi-isol
Figure 4. D
Some obsoverall shpanel of Secondlylarger biavariation structuresare signifnoticeablvariation different 0.1). In order tfield, a pe(1:1) and and 1,5%CD resultfor the meven with
ontrast with reon change are h a 1:1 duty cy
cus correctionf the azimuthalines derived aated and isola
Dependence of
servations canhape of the CDFigure 2. The
y, regardless oas in order to cof the CD me
s (1:1 to 1:3 dficantly smallee variation to within the CDpitches is mea
to quantify theercentage variisolated (1:7)
% for 45 degreeting at the edg
measured CD, ah no correct pi
spect to the othowever a rat
ycle and the lo
ns as displayedal angle of theapplying the dated patterns, c
f the shadowingorder to give
n be drawn onD curves as fue pitch impact f the orientaticompensate foeasured at the duty cycles). Ter. Looking ata variable pit
D values measasured within
e impact of theiation can be c) lines. The ree oriented linege of the slit wall structures aitch based bia
ther orientatiother small effeowest contrast
d in the left pe EUV illumindefocus correccorresponding
g effects across te a CD=16 nm
the basis of thunction of the
can be roughon of the patteor the arc-shapedge of the sl
The variation t the dashed retch (higher thasured for a sinV lines (STD
e pitch variaticalculated whsulting valueses: if, for examwould differ 4%at φ=±18.6 de
as applied to th
ons consideredect, with a dift of the V line
panel of Figurenation on the sction to all simg to pitch valu
the illuminationat the center of
he set of simuazimuth angle
hly described aern, dense pitcped illuminatilit occurs durinfrom semi isoed lines in Figan 1:3) with rengle pattern orDEV 0.4), whe
ion on the shaich considers s for this quanmple, 1:1 V lin% from the tag which werehe structure pr
d here. Contrafference betwees at pitch 1:7
e 3 were applishadowing effmulations setuues of 32 nm (
n field on the stf the scanner ex
ulation results e supports theas a fine tuninches need to bon of the EUVng the transiti
olated to isolatgure 4 , the shaespect to the arientation at thereas the small
dowing effectthe CD printe
ntity are 4% fones would be b
arget. Assumin considered hrinted on the p
ast differenceseen the highesof about 10%
ied to the newfects for V, H up is shown in 1:1), 64 nm (1
tructure pitch. Axposure field.
presented in Fe trend first deg of this gene
be printed on tV scanners. Tion from densted structures adowing effec
azimuthal anglhe edge of the lest is observe
ts measured aced at the edge or the V lines, biased as for
ng a 5% specifere would be photomask.
s due to pitch ast contrast for
%.
w simulations sand 45 degreeFigure 4 for d
1:3) and 128 n
All test cases ar
Figure 4. Firstscribed in the
eral behaviourthe photomask
Third, the largee to semi isola(1:3 to 1:7 du
cts on H lines le. Fourth, theillumination f
ed for H lines
cross the illumof the field fo0.6% for hori
1:7 V lines infication requirprinted in spe
and 45 deg
setup. The es dense, nm (1:7).
re biased in
t, the e right . k with a est ated
uty cycles) suffer no e highest field for (STDEV
mination or dense izontal stead, the rement
ecification
Proc. of SPIE Vol. 9231 923109-6
A further issue to be addressed in the context of the characterization of the shadowing effects dependence on the structure pitch is the impact of the pitch driven defocus correction (see left panel of Figure 3) on the measured CD across the field. For this purpose, the CD values at the edge of the slit for V, H and 45 deg oriented lines and spaces with 128 nm pitch (1:7, the value which nominally has the largest focus shift for all three orientations) have been compared before and after correction for the best focal plane shift. The difference between these two values, which quantifies the way the azimuthal angle component impacts the shadowing effects, is measured to be within 0.1% for all test cases. Based on the previous results, it can be concluded that the impact of the azimuthal angle of the illumination on the shadowing effects with respect to pitch variation is a minor one. The results presented in this and following sections are based on simulations which consider only a few of the several parameters which determine the entire lithographic process. They serve as guidelines to achieve understanding of the interplay of the different parameters within the EUV lithographic process, as well as to gain the confidence that shadowing effects can be tightly predicted and controlled with currently available technologies. Other works have addressed the best focus shift dependence on other process parameters; for example, the displacement of the best focal plane through pitch also has a strong dependence on the illumination settings7 , of which this section considered only one setup configuration.
SHADOWS DEPENDENCE ON ILLUMINATION PUPIL
The third parameter considered is the pupil shape and its coherence. The pupil used to illuminate the structures on the photomask has an impact on the shadows produced by the lithography process. The way shadows change depending on the azimuthal angle of the EUV illumination is also dependent on the pupil shape and coherence, and must be carefully investigated for the final modelling of the structure across the whole photomask. In this section, a description of the dependence of the shadows introduced by the arc shaped illumination on the pupil is given. In these simulations of 16 nm line and space patterns with a 32 nm pitch (wafer level), an absorber height of 50 nm has been considered. Annular and disar pupils are employed with a variable coherence, which spans the range 0.2-0.9 in steps of 0.1 while keeping the outer radius of 0.9 constant. Fundamental quantities like best focus shift in response to mask and system parameter change and contrast at best focus are considered as input and control parameters of the simulations. Values for the best focus shift and contrast at best focus were calculated as in the last section and they are shown in Figure 5, in which the disar and annular pupil cases are respectively shown in the left and right panels. The blue data points in each plot show the shift of the best focal plane (at mask level) as a function of the inner sigma used for the illumination pupil. Ranging from the most to the least coherent illumination settings, the shift of the best focal plane for both pupil shapes can be as high as 50 nm. Once the displacement of the best focal plane is taken into account the imaging performance is also improved; contrast enhancement is measured to be on the order of 14% for the disar pupil and 7% for the annular pupil. As already stated for the pitch section, the measurements of the best focus shift have to be considered with respect to one another and not as an absolute estimation of the best focal plane vertical position.
Proc. of SPIE Vol. 9231 923109-7
Figure 5. S1:1 vertic
Once the simulatioangle comspaces pathe arc-shon the maimpact ofshape is s
Figure 6
Second, tFocusingilluminatitranslatesof bias ththe centercenter of Third, shacoherencewhole do
Shift of the bestcal lines and sp
defocus correon setup, simumponent of theattern with a 1haped slit is plask). Some obf the azimuthaslightly modif
6. Impact of the coh
the annular pu on the extremion slit betwes into a ΔCD ahat must be appr of the field rthe field, in oadowing effece only in the o
omain, since th
t focal plane (nmpaces. The pupi
ection describlations were re scanner illum:1 duty cycle lotted against
bservations canal angle on thefied by lithogr
scanner arc-shherence. In this
upil leads to lame values for σen the annularat mask level oplied to V linereflects the gloorder to highligcts within the outermost halfhe dashed and
m at reticle levl symbols on th
ann
ed in the precrun in order tomination slit f(pitch 32 nm)the azimuthaln be drawn: fie shadowing eraphic parame
haped illuminatgraph, only σin
arger shadowinσin (σin=0.2 anr and disar puof 1.2 and 0.4es at this nomobal bias applght the impactannular pupilf of the field.
d solid blue lin
el) and aerial imhe top corners anular (right) ca
eding paragrao describe howfor different il) is presentedl component o
first, for all simeffects for V lieters such as th
tion on the shadn=0.2 and 0.8 h
ng effects thannd 0.8, σout fix
upils with the s4 nm respectivminal position o
ied to every st of the arc sh (red curves inFor the disar p
nes in Figure 6
mage contrast iare only used toases.
aph and plottedw the shadowinllumination sein Figure 6. In
of the CRA (dmulated cases,ines structureshe pupil shape
dowing effects ohave been plotte
n the disar pupxed at 0.9), wesame σin of 0.3
vely, which repon the mask. Timulation sett
haped illuminan Figure 6) repupil, this diff6 are never ov
improvement afo distinguish bet
d in Figure 3 hng effects resp
etup. A set of rn the plot, the
directly conver, the typical “Us has been foue and coherenc
of V lines with ded for reading e
pil for all cohe measured a Δ3 nm and 0.1 present the difThe fact that ating all featureation alone. spond differenference is noti
verlapping eac
fter defocus cortween the disar
has been inpupond to the azresults for V l
e CD measuredrtible to an X U” shape descund. The indivce.
different pupil sease.
herence settingΔCD at the ednm at wafer lfference in theall curves givees to 16 nm in
ntly to a varyiiceable acrossh other.
rrection for r (left) and
ut to the zimuthal lines and d across position
cribing the vidual
shape and
gs. dge of the evel. This e amount e 16 nm in n the
ing s the
Proc. of SPIE Vol. 9231 923109-8
Fourth, Coherence has an opposite impact on the shadowing effects for the two pupil shapes employed in this example. A smaller coherence (larger σin) produces a larger shadow at the edge of the slit for the disar pupil, whereas for the annular pupil the effect is reversed. Although the shadow is larger for a disar pupil with σin =0.8, the image contrast is the highest registered (94%). The dependence of the shadowing effects across the azimuthal angle illumination due to the pupil shape and coherence is a small effect with the largest differences calculated on the order of 1.2 nm at mask level. It is however important to describe the dependence of shadowing effects on the illumination setup in order to correctly bias a photomask across its whole patterned surface, in this way maximizing the process window. The results from this section have also been considered as base to select proper illumination settings for the study of the impact on shadowing effects across the illumination field of quantity such as the structure pitch (previous section) and the absorber height (next section). Disar and annular pupils with a σ=0.6-0.9 have been selected for this purposes because they provide average shadow effects and good contrast for the 1:1 pitch case with 50 nm absorber height.
SHADOW DEPENDENCE ON ABSORBER HEIGHT The last parameter that has been considered within this work is the height of the absorber. The goal of this section is to describe the impact of the azimuthal angle component of the CRA on the imaging of L/S structures in relation to its dependence on the absorber height. Two application relevant topics have been considered: the impact of a clear absorber repair (deposition) with an incorrect height and the overlay error introduced by a non-uniform absorber height deposition. The CD printed at a certain X location (or azimuthal angle φ) CDφ can be written as the target CD plus a structure and process dependent shadow ΔCDφ. In formulas:
φφ CDCDCD ett Δ+= arg , where the φ index identifies a certain azimuthal angle within the range ±18.6 deg, and the factor ΔCDφ represents the shadow produced on the aerial image. This can also be written in an explicit form as
φθφ sin)(tan2 ××=Δ MhCD eff , with θ=6 deg CRA and M=4 magnification factor. The quantity heff can be thought as an effective height which takes into account both the height of the absorber on top of the photomask, and the penetration depth of the EUV photons within the multi-layer reflective structure of the photomask itself. Whereas a linear relation exists between heff and the absorber height, the relation between heff and the penetration depth cannot be easily put into a geometric formula, due to the complex nature of EUV reflectivity. The dependence of the CD variation in response to an absorber height variation has been investigated in the literature8. Rigorous simulations show an increase (decrease) of the measured CD at wafer level for increasing (decreasing) absorber thickness, with embedded oscillations showing a period of about half the exposure wavelength9. This effect has also been confirmed experimentally via wafer print studies10. As a conclusion, a direct proportionality between CD and absorber thickness variation can be expected, which will affect the overall dependence of the shadowing effects on the absorber height across the scanner illumination field. The results of simulations run for the two different pupil shapes are shown in Figure 7. As reported in Reference 7, the absorber height also has an impact on the best focal plane: different height values are focused at different vertical planes. This has been taken into account within the simulation setup, with the best focus shift being calculated from the contrast vs. defocus curves as explained in the previous sections. Typical values are within 25 nm at reticle level.
Proc. of SPIE Vol. 9231 923109-9
Figure 7absorb
This workpresentedannular peffects achave beentarget CDdetail thethe previoprocedureproduced
Figure 8. and 45 d
7. Shadowing efber height valu
k considers vad in this sectiopupils with σ=cross the maskn considered.
D value of 16 ne behaviour of ously presentee are presente
d on the EUV p
CD measured adegrees oriente
effects through tes (40 to 70 nm
illumin
alues for the aon have been r=0.6-0.9. As itk illumination As expected, nm, are largerf the printed Ced data on thed in Figure 8 photomask.
at the edge of thed lines (right p
the illuminationm). Panels in thenated with a dis
absorber heighrun for all thret is clear fromfield remainsthe shadows p
r for higher abCD at the azim
edge of the ilin order to dis
he scanner slit fanel). L and R coordinates φ=
n field due to the left (middle, rsar (top) and an
ht between 40 ee orientations
m the plots shos the same throproduced at th
bsorber height muthal angle co
llumination fiesplay the depe
for disar and anindexes in the r
=-18.6 deg (L) a
e azimuthal angright) column shnnular (bottom)
and 70 nm, ins of 1:1 L/S paown in Figure ough the diffehe edge of thefor all orienta
oordinate φ=±eld has been pendence on the
nnular illuminaright panel refeand φ=18.6 (R)
gle component how the V (H, 4) pupil.
n 10 nm stepsatterns illumin7, the overall
erent values of slit, i.e. the dations. It is in
±18.6 deg. Forperformed ande absorber hei
ation pupils for er to the left and).
of the CRA for 45 deg) lines tes
. The simulatinated by disartrend of the s
f absorber heigdeviation fromnteresting to dr this purpose d the results ofight of the sha
V and H lines d right of the sli
different st cases
ions r and shadowing ght which
m the describe in
a cut of f this adows
(left panel) it, i.e. the
Proc. of SPIE Vol. 9231 923109-10
As previothicknessthe proceQuantitatderived brange of aright pane A fundamdefect repthe impaccapabiliticompensa To quantiilluminatebeen takedeg). All wafer levmask havthe final pheight paacross thethe left pa(over-depintroducerespect totypical 5%both the o
Figure 9. cases are
ously discusses cannot be expess. However, tively, a first oby dividing theabsorber heighel of Figure 8)
mental applicapair. The simuct of an absorbies of the MeRational repair
ify the impacted with a disaen into accouncombinations
vel. The fundave also been suproduct mask
arameter withie whole field. anel) and 70 nposition) and 4ed in these simo the target CD% ΔCD specifover-depositio
Left: Impact ofshown in the to
in def
ed, the proportpected to be pa global trend
order estimatioe difference Cht used in the ) and 1 nm (4
ation within phulations presenber defect repRiT® repair toof multi-layer
t of an absorbear 0.6-0.9 pupint and simulates of structuresamental differeubject to azim. This is essenn a repair proFigure 9 show
nm (bottom pa40 (under-dep
mulations is thD, ΔCD = CDfication for abon and under-d
f incorrect absoop and bottom pfect repair. Righ
tionality betwperfectly linead can be obseron of the varia
CDφ=±18.6 (70 nmsimulated cas5 deg lines at
hotomask procnted in Figureair performed
ool which presr defects11,12.
er repair with il have been ced for: the cen and positionsence with all s
muthal bias, anntial in order tcess, where thws the results art of the left pposition) nm herefore 30 nm
D – 16 nm, is pbsorber repairdeposition cas
orber height ontpanels respectivht: Repair num
ween the magnar, because of rved within allation of the shm) – CDφ=±18.6
ses. This indicφ=±18.6 deg
cessing whiche 7 and Figured with an inaccsently offers p
incorrect heigconsidered. Fonter (φ=0 deg)s have been bisimulations rend therefore fuo quantify thehe features shoof this investi
panel) nm heigheight, respectm. The value oplotted for all tprocess is ploses.
to the repair prvely. The black ber for the diffe
itude of the shinherent oscill test cases wh
hadow ΔCD±18
6 (40 nm) by 3cates values bewith annular p
h is very sensite 8 can be furthcurate absorbeprocesses for E
ght, V – H andor each structu) and the two iased in order eported above ully resemble te difference inould be biasedigation. Structght have beentively; the erroof the ΔCD inttest cases in th
otted as a dash
rocess of an absk dashed line repferent test cases
hadowing effellations of thishich were sim8.6 per nm abso30 nm, which etween 0.2 nmpupil).
tive to the heiher extended ier height, and EUV absorber
d 45 deg linesure, three azimedges of the imto print at theis that structuthe on-target f
n shadows intrd ahead of timtures biased fo
n simulated to or in the absortroduced by thhe left panel ohed black line
sorber defect. Opresents a 5% Cdescribed in th
ect and the abss dependence t
mulated in this orber height crepresents the
m (green curve
ight of the absin order to invrelate it to the
r defect repair
s with a 1:1 dumuthal angles maging slit (φ
e target CD ofures at the edgfeatures patterroduced by the
me to print at tafor 40 nm (top
be repaired wrber depositionhe erroneous rof Figure 9. Heat ΔCD=±0.8
Over and under CD specificatiohe text.
sorber typical of work.
can be e total e in the
sorber is vestigate e r and
uty cycle have
φ=±18.6 f 16 nm at ge of the rned on e absorber arget part of
with 70 n process
repair with ere, the nm for
deposition on common
Proc. of SPIE Vol. 9231 923109-11
From the plot it is noticeable that H lines have the highest ΔCD at the center, and this quantity diminishes towards the edge of the slit; the impact of a repair with a height deviating from the nominal on H lines is therefore higher in the center of the imaging field than at the edges. This property can be expected, since it has already been shown that H lines need the highest compensation for the azimuthal bias at φ=0 deg. For V lines this effect is the opposite, whereas for the 45 deg lines the impact is variable across the entire field. In order to better quantify the process sensitivity to a non-nominal repair height, the repair number shall be introduced. This is a typical quantity used in the process of defect repair, defined as CD variation per nm absorber height deposited (or etched). For the test cases presented in Figure 9, this number can be derived by dividing the ΔCD values by the 30 nm difference in absorber height (for nominal and repair). Throughout the calculations, a linear trend of the ΔCD vs absorber height has been assumed; as it has been shown in Figure 8, this assumption can be thought as a valid first order approximation. Repair numbers for the over and under deposition cases described above are reported in the right panel of Figure 9. Interestingly, it is possible to notice that the repair number is different for over and under deposition cases. The maximum absorber height deviation which would still provide an acceptable repair can be found dividing the ΔCD specification of 0.8 nm (at wafer level) by the correspondent repair number shown in the Table. As a result for the 16 nm half-pitch node, V lines at the center of the illumination field can be repaired with a ≤ 20 nm over-deposition or ≤ 26nm under-deposition processes. For H lines the process must be more tightly controlled: a repair with 8nm over-deposition or 10 nm under-deposition is the limit for printing within the target CD of 16 nm ± 5%. Some conclusions can be drawn on the basis of the findings described above. First, H lines are the most sensitive to the height of the absorber deposition or etch within a repair process, contributing to higher height control requirements for this orientation. Second, the repair numbers for the over and under deposition cases are different from each other. Third, a height control of ±8 nm, as well as all the other values reported in this work, is well within the MeRiT® capabilities of repairing absorber materials. The absorber height control capabilities by the current MeRiT® platform is within nm precision13,14, and therefore the tight control of the repair process is not an issue with respect to shadowing effects and bias. One more aspect related to the impact of the absorber height onto the imaging performance relates to the overlay or image placement error. With the simulations it is possible to measure the shift of the central coordinate of the aerial image in response to a variation in the height of the absorber. This is shown in Figure 10, where the aerial images of H and 45 deg oriented lines with a 1:1 duty cycle and illuminated by a disar pupil with σ=0.6-0.9 are presented. A first glance at the plots shows that the center of the images, identified as the point of lowest intensity in the different colored curves, is subject to a drift towards X values lower than 16 nm (the center of the simulation domain) as the absorber thickness is increased changed. The behavior of this shift is displayed in the insets within each panel of Figure 10, where the center of the aerial image for H (left panel) and 45 deg (right panel) lines is plotted against the absorber height. V lines are not shown because no shift has been measured for this orientation in response to an absorber height variation. The shift of the aerial image center has been measured for both H and 45 deg oriented lines towards lower X coordinates. A good linear trend between this shift and the absorber height was found and quantified by R2 values higher than 0.9 for both cases. Between the absorber thickness values of 40 and 70 nm, the largest shift was measured for H lines to be 1.3 nm; the highest gradient measured within this trend is about 0.5 nm for 10 nm variation in the height of the absorber. According to the 2013 release of the international technology roadmap for semiconductor (ITRS)15, overlay specification for production EUV photomasks will be 3% of the pitch size. Considering the structures which have been simulated in this last section, a 1:1 lines and space pattern with CD of 16 nm is described by a pitch of 64 nm. Therefore, overlay specifications must be within 2 nm. A 0.5 nm overlay error (at wafer level) introduced by a 10 nm absorber height variation from nominal can absorb up to 100% of the 2 nm (at mask level) overlay specification, i.e. the entire budget for overlay errors. This argument shows the fundamental importance of height control within the EUV photomask production and repair process.
Proc. of SPIE Vol. 9231 923109-12
Figure values of t
As shownsuch as oinfluencecapabiliti
Shadowintelecentriproductioon the phwhich areparametestructure these effeuniformitThe AIMcost-effecapplied mthicknessuniformitphotomasthe develthe heigh
1 HellwegHarnisch79690H (
10. Aerial imagthe absorber (4
n by the simulverlay and rep
e the positive oies.
ng effects are ic illuminationon process of Ehotomask in ore directly connrs, i.e. are proorientation, p
ects must be tity targets.
MSTM EUV plactive way to fu
mask bias calcus of the absorbty, and overlaysks within theopment of EU
ht of the absorb
g, D., Ruoff, J, W., "AIMS™(2011).
ges of 1:1 H (le40 to 70 nm in 1
lation work prpair success. Houtcome of a r
an inherent chn of the mask,EUV photomarder to print atnected to the b
ocess dependepitch, absorberightly controll
atform, in addifully emulate tulated previou
ber must also by errors. With
e nm regime, thUV lithographyber.
J., Herkommer™ EUV - the a
eft panel) and 410 nm steps). Th
position as a fu
resented here, However, all srepair process
CO
haracteristic o, finite absorbeasks is made mt target acrossbias needed onnt. The depenr height and illed in order to
ition to a prodthe scanner imusly meets allbe tightly conh the ability tohe Carl Zeiss y and able to
R
r, A., Stühler,actinic aerial i
45 deg (right pahe insets within function of the a
the absorber
specifications s due to absorb
ONCLUSIO
of EUV lithoger height and more complicas the whole fien the reticle, d
ndency of shadlumination co
o maximize pr
duction tool fomaging conditi
target specifitrolled in ordeo control the dMeRiT® platsuccessfully a
EFERENCE
, J., Ihl, T., Feimage review
anel) lines and sthe two panels
absorber height
thickness has needed in ord
ber misplacem
ONS
raphy, arisingthe reflective ated by the neeld. This workdepend on a grdowing effectsonditions haverocess window
or photomask dions and experications. As deer to reduce erdeposition or etform for photaddress all the
ES
eldmann, H., Rplatform for E
space patterns ws show the drift t.
an influence der to reduce oment are well w
g from the comnature of the
eed of properlyk has shown threat variety ofs across the ill
e been investigw and meet CD
defect reviewrimentally valemonstrated inrrors in CD metch of absorbtomask repair e afore mentio
Ringel, M., StrEUVmasks",
with different thof the aerial im
on imaging paoverlay errorswithin the Me
mbination of nEUV optics. Ty biasing the fhat shadowingf lithographic lumination arcgated. AlthougD requirement
, is an efficienlidate whethern the last secti
measurements aber material on
is already on oned issues rel
rößner, U., PeProc. SPIE 79
hickness mage center
arameters s or to eRiT®
non-The features g effects,
c on gh small, ts and
nt and r the ion, the and n EUV track for
lated to
erlitz, S., 969,
Proc. of SPIE Vol. 9231 923109-13
2 Folta, J. A., Bajt, S., Barbee, T. W., Grabner, R. F., Mirkarimi, P. B., Nguyen, T., Schmidt, M. A., Spiller, E., Walton, C. C., Wedowski, M., Montcalm, C., "Advances in multilayer reflective coatings for extreme ultraviolet lithography", Proc. SPIE 3676, 702 (1999). 3 Braun, S., Mai, H., Moss, M., Scholz, R., Leson, A. "Mo/Si Multilayers with Different Barrier Layers for Applications as Extreme Ultraviolet Mirrors", Jpn. J. Appl. Phys. 41 (2002) 4074B. 4 Neumann, J.T., Gräupner, P., Kaiser, W., Garreis, R., Geh, B., "Mask effects for high-NA EUV: impact of NA, chief-ray-angle, and reduction ratio", Proc. SPIE 8679, 867915-1 (2013). 5 Garetto, A., Capelli, R., Blumrich, F., Magnusson, K., Waiblinger, M., Scherübl, T., Peters, J.H., Goldstein, M. " Defect Mitigation Considerations for EUV Photomasks ", submitted to Journal of Micro/Nanolithography, MEMS, and MOEMS (2014). 6 Peters, R. et al., "ASML’s NXE platform performance and volume introduction", Proc. SPIE 8679, 8679-50 (2013). 7 Davydova, N., de Kruif, R., van Setten, E., Lammers, A., Oorschot, D., Schiffelers, G., van Dijk, J., Connolly, B., Fukugami, N., Kodera, Y., Morimoto, H., Sakata, Y., Kotani, J., Kondo, S., Imoto, T., Rolff, H., Ullrich, A., "Achievements and challenges of EUV mask imaging", PMJ (2014). 8 Yan, P.Y., "The impact of EUVL Mask Buffer and Absorber Material Properties on Mask Quality and Performance", Emerging Lithographiy Technologies,VI, Proc. SPIE 4688 (2002). 9 van Setten, E., Man, C.-W., Murillo, R., Lok, S., van Ingen Schenau, K., Feenstra, K., Wagner, C.,"Impact of mask absorber on EUV imaging performance ", Proc. SPIE 7545, 754503 (2010). 10 van Setten, E., Oorschot, D., Man, C.-W., Dusa, M., de Kruif, R., Davydova, N., Feenstra, K., Wagner, C., Spies, P., Wiese, N., Waiblinger, M.," EUV mask stack optimization for enhanced imaging performance ", Proc. SPIE 7823, 78231O-1 (2010). 11 Edinger, K. et al., "A novel electron-beam-based photomask repair tool", Proc. SPIE 5256, 1222 (2003). 12 Waiblinger, M., Kornilov, K., Hoffman, T., Edinger, K., "e-beam induced EUV photomask repair: a perfect match", Proc. SPIE 7823, 782304 (2010). 13 Waiblinger, M., Bret, T., Jonckheere, R. and Van den Heuvel, D., "Ebeam based mask repair as door opener for defect free EUV masks", Proc. SPIE 8522, 85221M (2012). 14 Bret, T., Jonckheere, R. and Van den Heuvel, D., Baur, C., Waiblinger, M., Baralia, G., "Closing the gap for EUV mask repair", Proc. SPIE 8322, 83220C-1 (2012). 15 The International Technology Roadmap for Semiconductors 2013, http://www.itrs.net/Links/2013ITRS/ 2013Chapters/2013Litho.pdf.
