Improved Mask Technology for
X-Ray Lithography
Project Staff: Michael H. Lim, James M. Daley, Ken-ichi
Murooka, Shingo Uchiyama, M. Feldman (LSU) and Professor Henry I. Smith.
Sponsors: Defense Advanced Research Projects
Agency/Naval Air Systems Command Contract N00019-98-K-0110 and Louisiana State
University, Contract R110030
At feature sizes of 100 nm and below the mask-to-substrate gap in x-ray lithography,
G, must be less than ~10 µm. Thus, for nanolithography the mask membrane should
be considerably flatter than 1 µm, preferably ~100 nm. Our mask technology is
based on low-stress, Si-rich silicon nitride, SiNx. This material is produced
in a vertical LPCVD reactor. Membranes of SiNx can be cleaned and processed
in conventional ways. For absorber patterns we electroplate gold onto the membrane,
using a specially designed apparatus, after resist exposure and development.
A Ti/Au plating base is deposited on the membrane prior to resist coating. To
pattern periodic structures on the x-ray masks, we use interferometric lithography
(IL), and for patterns of arbitrary geometry we use e-beam lithography. A high-resolution
Leo SEM and a Digital Instruments STM/AFM are used to inspect x-ray masks for
defects. Radiation hardness for SiNx membranes remains a problem at dose levels
corresponding to production (i.e., millions of exposures). For research purposes,
however, the material is entirely acceptable. Currently we are investigating
the problem of x-ray mask distortion, which is a potential problem in x-ray
lithography.
X-ray mask distortion is rooted in the flexibility of its membrane. The membrane
responds to stress in the absorber patterns by flexing both in-plane and out-of
plane. Distortion caused by this motion, especially in-plane, must be overcome
if x-ray lithography is to meet the overlay requirements of future electronic
and optical devices. Thus far, the x-ray lithography community has attacked
this problem by trying to achieve very low-stress absorbers and, when necessary,
compensating for absorber induced stress by modifying the pattern written by
the electron beam. We are pursuing a new approach in which we first measure
the membrane distortion and then correct it.
Figure 6: A schematic of the Holographic Phase-Shifting Interferometer (HPSI)
based on the interferometric lithography system that we use to generate highly-coherent
gratings.
To measure distortion, we have developed a broadly applicable, nondestructive,
global, membrane-distortion measurement technique called Holographic-Phase-Shifting
Interferometry (HPSI). The HPSI system is based on the interferometric lithography
(IL) system we use to generate large-area, highly-coherent gratings. Figure
6 is a schematic of the IL apparatus, configured as a HPSI system. The IL system
splits a laser beam (l=351nm) and forms two mutually coherent spherical waves,
which interfere at the substrate at a half-angle q. The standing wave created
at the substrate surface is used to expose photoresist. After development, the
grating is present on the substrate surface or can be etched into it. The IL
system is used as a holographic interferometer by mounting the IL-generated
grating on the substrate platform, and placing a fluorescent screen in front
of one or both of the spatial filters, as depicted in Figure 6. A fringe pattern
appears on the screen, which is due to the superposition of two wave fronts:
one reflected from the substrate surface and the other back-diffracted from
the grating. If the grating has suffered no distortion between exposure and
reinsertion, the reflected and back-diffracted beams will be identical and no
fringes will be observed on the screen. Any in-plane distortion of the grating
will result in a fringe pattern. A CCD camera is used to capture the fringes.
To increase the precision, a phase-shifting measurement is implemented, by changing
the phase of one of the arms and acquiring several images, Fig. 7. In order
to use this apparatus to measure the in-plane distortion, we etch shallow IL-generated
gratings into the membrane.
Figure 7: The HPSI uses diffractive metrology to make a rapid, global measurement
of the distortion of a membrane. (a) Contour plot of phase distortion (nonlinear
component) obtained by the HPSI. Successive contours are separated by Õ/2 radians.,
(b) conversion of the phase map of (a) into a distortion map.
We have developed an analytical technique that predicts both in-plane distortion
(IPD) and out-of-plane distortion (OPD) arising from arbitrary stress distributions
in 2D. Moreover, we can also solve the inverse problem; i.e., we can predict
the stress distribution which, when applied to any existing distortion, eliminates
it. The calculational techniques are based on the variational method. It is
relatively straightforward to formulate the total energy due to membrane distortion,
even for a very complicated stress distribution. We calculate the true distortion
by minimizing the total membrane energy due to the placement of the stressed
absorber; the total energy is straightforward to formulate for even complicated
absorber distributions. Figure 8 shows the results of a calculation where half
the SiNx membrane is covered with an absorber under tensile stress.
Figure 8: Our calculational technique can solve both in-plane and out-of plane
membrane distortions in 2-dimensions. For example, covering half-plane of a 10
mm square SiN-x membrane with an absorber results in both IPD and OPD displacements.
(a) shows a 2-dimensional map of the IPD, which clearly indicates that a 1-D approximation
would be inadequate. This is more clearly shown in (b), as the cross-sectional
slices of the x component of the distortion along the x-axis vary as one moves
closer to the boundary of the membrane. (c) and (d) show that the OPD also
has a significant 2-D component.
Figure 9: The calculated distortion of a 50 mm square SiN membrane, with a thickness
of 1 µm, in the presence of a 7.5 mm spot with 1žC rise in temperature. (a)
The vector map showing the 2D distortion; (b) the cross-section of the x-component
along the x-axis.
A correcting stress distribution can be introduced by means of local heating.
Figure 9 shows the analytical result of a 7.5 mm spot, with a 1žC elevation
in temperature, centered on a 50 mm square SiNx membrane that is 1 µm thick.
There is a peak displacement of 6.3 nm that occurs around the edges of the spot.
This is equivalent to a circularly shaped absorber with 1.3 MPa of stress. The
analysis indicates the time constant of the heating is less than one second.
Moreover the calculational process also requires less than one second to complete.
Figure 10 shows a comparison of a self-consistent calculation of the distortion
from a locally heated spot and the measured distortion.
We hope to build on this work by developing a system that can actively introduce
a heat distribution into the x-ray mask in order to correct for membrane distortion
in real-time. If successful, this should enable x-ray nanolithography to be
used in applications such as integrated optics which demand the highest accuracy,
precision and coherence in the placement of pattern elements.
Figure 10: A comparison of the calculated distortion and the measured distortion.