Integration of Single Damascene 85/85 nm L/S copper trenches in Black Diamond

using 193 nm optical lithography with dipole illumination

J. Van Olmen, W. Wu*, M. Van Hove, Y . Travaly, S. H. Brongersma, B. Eyckens, M. Maenhoudt,
J. Van Aelst, H. Struyf, S. Demuynck, Zs. Tokei, I. Vervoort, B. Sijmus, I. Vos, I. Ciofi, M. Stucchi,
K. Maex*, F. Iacopi

IMEC, Kapeldreef 75, B-3001 Leuven, Belgium

*Also at E.E. Department, K.U. Leuven, Leuven, Belgium

Abstract The k, factor determines the need for resolution enhance-

ment techniques (RET). Resolution enhancement can he
This paper describes the integration of Single Damascene achieved by on-reticle techniques like Phase Shift
85/85 nm U S copper trenches in Black Diamond (Applied Masking (PSM) and Optical Proximity Corrections (OPC),
Materials) dielectric (k=2.85). Optical lithography (193 or by off-axis illumination conditions. From figure 1 it
nm) with off-axis illumination was used to print the becomes clear that in order to print 85 nm lines on a
trenches. Integration issues are discussed, and resistance stepper with A = 193 nm and NA = 0.75, the k , factor is
and RC delay data are presented. The method is applied to 0.33. In this study dipole-x illumination was used to
study the resistivity for sub 100 nm copper lines. achieve this. It is very important to note that this technique
favours only one orientation (perpendicular to the dipole
Introduction axis). Trenches along the other direction will have to be
wide enough on the reticle, or else will not print. The
For the 65 nm technology node minimal local wiring method is also very prone to proximity, as is shown in
linewidths/spacings will be of the order of 85 nm. Most figure 2 . From these aerial image simulation results one
results on these very narrow copper lines have been can observe the strong pitch dependence. Dense 85 nm
accomplished by means of electron beam writing. Major trenches will only be printed as 85 nm trenches for the
drawback of this approach is the long writing time needed optimised pitch of 170 nm. At other pitches the lines will
to expose a rather limited set of structures. In this paper we be printed smaller, or not at all (so-called forbidden
will demonstrate that 193 nm lithography with off-axis pitches). Also the simulated behaviour for isolated lines is
illumination can be used to produce 85/85 nm L/S copper depicted. Both for the orientation along and perpendicular
trenches in Black Diamond. to the dipole axis, the dimension on the reticle has to he
larger than a certain cutoff value in order to print.
The Rayleigh criterion for the resolution (at % pitch) links Therefore assist lines have to be used on the reticle to
the resolution R to the wavelength A of the source, the enable printing of (semi-) isolated features.
numerical aperture NA of the lens system, and the k,
factor: R=k,.A/NA (1)

0.6 . . , . . . . , . . . . ,. . . . . . . , ..
I Reticle CD is0 (nm)
0 55 k1>0.5

0.5
easy pimessing
,. NA 0.75

,*f
-b
-0
o'45

0.4
0.25<k1<0.50
RET needed
,a'
i'
~

Pitch (85 nm trench)

 Experimental

The following dielectric stack was deposited in a Centura

CVD system from Applied Materials:

30 nm SiCN as etch stop layer

150 nm Black Diamond (BD) with O2
plasma pre-treatment and He plasma post-
treatment (thickness of BD film was 4, h,

optimized for minimal substrate reflectivity) SEM crors-recriox 0/85/85 n m USfeotsres <$er etch (lop)
Figure 4 U)
10 nm oxide cap layer to prevent resist and @er strip ( ~ o I I o , ?hj? ~SEM
: c r o ~ s - ~ ~ c roi /oon(nontird)
poisoning of the 193nm photoresist 85/85 ,m U S inrerline Cnpacirnnce slnrcfiire wilh h i g h AR
t r m d e s in ED. 1We nrrirr liner a m nor well riefltred nr
erpecred.)
Litho was performed with 250 nm GARS8105 resist from
Arch Chemicals on an ASML PAS1 100 stepper with NA
Wafers with regular AR trenches continued processing as
= 0.75 and h = 193 nm. Dipole-illumination was optimised
follows:
for printing 85 nm trenches at I70 nm pitch. On the reticle,
manual OPC was performed and assist lines were drawn 3 min. degas @ 350 "C + 10 sec. Ar Preclean
next to (semi)-isolated features. The U setting of the dipole before 10/15 nm Ta(N)Ka SIP bilayer barrier and
illumination was optimized for reduced risk of scumming. 100 nm SIP Cu seed in Electra (AMAT)
Figure 3 shows some top down SEM micrographs of a 85 450 nm electrochemical plating in AMAT I-ECP
nm LIS meander fork structure after dipole-x illumination. with in-situ anneal for 30 sec. @ 250 "C under
For the most aggressive (resulting in the narrowest lines) forming gas
dipole setting (a = 0.89/0.65) a lot of scumming can he 0 2 step Cu and barrier C M P on a LAM Teres
observed. For a more "relaxed" dipole setting (o= system with Hitachi low abrasive slurry (@3 Psi
0.8910.40) all features are well defined. This setting will down force).
he used throughout the remainder of this paper.
Resistance and RC delay data

For the electrical evaluation 46 dies were measured

distributed a11 over the wafer. Figure 5 shows the results
for continuity of the 85/85 nm L/S meanders with a length
of I cm, 10 cm and 50 cm. Good yield is observed even
for the 50 cm meander. Fig. 6 shows the RC delay
distribution for a (nominal) 85/85 nm WS meander-fork
with a length of 1 mm. A median RC delay of 624 ps is
obtained.

Fi,qrrre 3 SEM piosrrr o/ 85 iim U S rrtrmtdrr fork .sIn~Iares'$le,

c i i p o l r ~ . ~illunzinnrio,! n,irh a = O.X9/0.65 ( l e / ) and a =
O.X9/0.40 (righr).

Trench etching was performed in a Lam Exelan chamber

with a mixture of Ar/O2/CFdCH2F2.In-situ resist strip was
done with 02/CF4. Figure 4a shows a SEM cross section
and top down picture of the obtained profiles after etch
and strip. It is observed that etch and strip result in nice
trench profiles. Some CD gain (-10 nm) is introduced by
etch and strip. After etching 150 nm BD still a lot of the
250 nm resist is remaining. Therefore it was also tried to
etch high aspect ratio (AR) trenches in BD. By changing
the trench etch chemistry to Ar/O2/N&Fs nice trenches
with AR -
8 could be achieved. Figure 4b shows an
example of nominal 85/85 nm US trenches in a BD stack
-
of 600 nm.

172
 700 pulse plated
I V DClow(l.5A)

50 100 150 200 250 300 350

Linewidth (nm)

Figure 7 Effective resistivity as (Ifunctio,, of linrwidrh /or difereni

Figare 6 RC delay disrributiuri fiir U I mn! l m g 85/85 nm (,io,niml) ploririg currenr cundirion.5.
US mcnndrr-fork structure.

Resistivity of narrow lines

The effective resistivity of 300 pm long Kelvin lines as a

function of their width is depicted in fig 7. The cross-
sectional area, required to obtain this data, is determined
from SEM images (fig. 8). taken on cleaved samples that
were locally smoothed in an FIB microscope. In
agreement with earlier publications, a strong increase with
decreasing linewidth is observed (1). The y-axis error bars
represent the resistive spread over the wafer, while the Figure 8 SEM cross-sections of Kelviri 1Ltu.r njlcr C M P lclenverl
uncertainty in determining the cross-sectional area gives samples were l o c o l l ~x,riooihrd in ii FIB).
the x-axis error. It should be noted that the small thickness
of the lines (150 nm) causes the resistivity values to be Conclusions
well above those predicted in the ITRS roadmap, even for
the larger linewidths. Optical lithography (193 nm) with off-axis dipole-x
illumination has enabled successful integration of single
Nevertheless, an influence due to plating conditions is damascene 85/85 nm LIS copper trenches in Black
clearly visible in fig. 7 when comparing three different Diamond. Perfect yield was obtained for 50 cm meanders.
recipes (pulsed plating, low (1.5 A) and high DC (6.0 A) Trenches filled with pulsed plating have a systematically
current density), each followed by a 30 sec. 250 “C anneal. lower resistivity, arguably related to a lower incorporation
Trenches tilled with pulsed plating have a systematically of impurities due to the removal of additives and their
lower resistivity, arguably related to a lower incorporation breakdown products from the surface in the de-plating
of impurities (scattering sites) due to the removal of phase of each pulse.
additives and their breakdown products from the surface in
References
the de-plating phase of each pulse. Comparing the two DC
recipes, slightly lower values are measured for the higher . , Kuan. C. Inoki. G. Oehrlein. K. Rose. Y.P.2hao.G.C. Wan&
(IiT. -
current density while the curves merge when the linewidth S. Ronmagel, and C. Cabrill. Mat. Res. Soc. Symp. Proc.(2000).
decreases. For wide lines this corresponds well with data Vo1.612. D7.1.1
(2) S.H. Brangerrmu. E. Kerr. I. Vervoon. A. Sacrens and K. Maex
showing lower impurity incorporation at high current
J. Mater. Res.. Vol. 17, No. 3, 2002
densities (2). For narrow lines the brightener concentration
(an additive that reduces the as-deposited grain size and Acknowledgements
counteracts the plating deceleration related to the
suppressor), driving the bottom-up filling, is expected to The authors wish to thank A. Vandervorst and M.
locally reach its maximum concentration for all global Moelants for extensive SEM work. H. Bender and C.
current densities in the bottom of the trenches. Drijboom are acknowledged for their assistance in sample
Consequently the resulting copper deposit during bottom- preparation (FIB smoothing). The IMEC Pilot Line is
up till will also be the same. thanked for their assistance in processing.

173