Techniques for identifying and mitigating effects of process variation on the electrical performance of integrated circuits
are described. These results are from multi-discipline, collaborative university-industry research and emphasize
anticipating sources of variation up-stream early in the circuit design phase. The lithography physics research includes
design and testing electronic monitors in silicon at 45 nm and
fast-CAD tools to identify systematic variations for entire
chip layouts. The device research includes the use of a spacer (sidewall transfer) gate fabrication process to suppress
random variability components. The Design-for-Manufacturing research includes double pattern decomposition in the
presence of bimodal CD behavior, process-aware reticle inspection, tool-aware dose trade-off between leakage and
speed, the extension of timing analysis methodology to capture across process-window effects and electrical processwindow
We present an improved method of phase retrieval from through-focus image series with higher precision and reduced
sensitivity to noise. The previous method, developed for EUV, actinic mask measurements, was based on the Gerchberg-Saxton algorithm and made use of two aerial images recorded in different focal planes. The new technique improves the
reconstruction uncertainty and increases the convergence speed by integrating information contained in multiple images
from a through focus series. Simulations characterize the new technique in terms of convergence speed, accuracy and
stability in presence of photon noise. We have demonstrated the phase-reconstruction method on native, mask-blank
phase defects and compared the results with phase predictions made from AFM data collected after the multilayer
deposition. Measurements show that a defect's top-surface height profile is not a reliable predictor of phase change in all
cases. The method and the current results can be applied to improve defect modeling and to enhance our understanding
of the detectability and printability of native phase defects.
This paper presents kernel convolution with pattern matching (KCPM), which is an updated version of fast-CAD
pattern matching for assessing lithography process variations. With KCPM, kernels that capture lateral feature
interaction between features due to process variations are convolved with a mask layout to calculate a match
factor, which indicates approximate change in intensity at the target location. The algorithm incorporates
a custom source, a mask with electromagnetic effects, and an arbitrary pupil function. For further accuracy
improvement, we introduce a source splitting technique. Though the evaluation speed is decreased, R2 correlation
of the match factor and change in intensity is increased. Results are shown with R2 correlation as high as 0.99 for
nearly coherent and annular illumination. Additionally, with a numerical aperture of 1.35, unbalanced quadrapole
illumination, 10mλ RMS random aberration in projection optics and complex mask with EMF effects included,
R2 correlation of more than 0.87 is achieved. This process is extremely fast (40μs per location) making it valuable
for a wide range of applications, most commonly hot spot detection and optimization.
Partially coherent imaging is formulated using two positive semi-definite matrices that include the mask as
well as the source and pupil. One matrix E is obtained by shifting the pupil function as in Hopkins
transmission cross coefficient (TCC) approach while the other matrix Z is obtained by shifting the mask
diffraction. Although the aerial images obtained by the matrices are identical, it is shown rank(Z) ≤ rank(E)
= N, where N is the number of point sources in the illumination. Therefore, less than N FFTs are required to
obtain the complete aerial image. Since the matrix Z describes the signal as partitioned into eigenfunctions
orthonormal in the pupil, its eigenvalues can be used to quantify the coherence through the von Neumann
entropy. The entropy shows the degree of coherence in the image, which is dependent on the source, mask
and pupil but independent of aberration.
In lithography for the 45nm node and beyond, phase errors introduced through electromagnetic field (EMF)
effects at photomask openings are significant sources of error in calculating on-wafer images. These edge effects
create distortion in both real and imaginary field transmission, which leads to a tilt in the process window, and
must be addressed in mask design to avoid loss of process latitude. This study presents a new formulation for
pattern matching, which allows EMF effects to be included via boundary layer modeling to facilitate extremely
fast assessment of EMF impact on imaging. Boundary layers are first used to model these edge effects, by
adding additional transmission features to a layout to represent the error transmissions caused by edges. Pattern
matching is then used to determine susceptibly to various pre-existing perturbations, in the presence of defocus.
This process can be extremely fast and hotspot detection can be run on an entire chip in hours, compared to
days for aerial imaging. Correlation between pattern matching and full aerial imaging can be as high as 0.97
for coherent imaging, and ≈ 0.75 for off-axis dipole illumination. This pattern matching framework is extremely
flexible and can be used for fast assessment of any series of effects which can be described as a path difference
in the pupil or as a transmission on the mask.
We present the basic concept of a fast mask optimization method that utilizes target-intensity back propagation. This method decomposes the target-intensity using a two-dimensional (2-D) transmission cross coefficient. After applying normal incidence approximation to the decomposed target intensity, the spectrum of the mask is optimized in the pupil plane. Since the optimization is performed in the pupil plane, one can use relatively small sampling points, leading to a fast optimization. By setting a high-contrast target intensity, we can obtain an approximation of a phase shifting mask, whereas a low-contrast target intensity approximates a binary intensity mask or attenuated phase shifting mask.
The extendibility of 2D-TCC technique to an isolated line of 45 nm width is investigated in this paper. The 2D-TCC
technique optimizes mask patterns placing assist pattern automatically. For 45 nm line patterns, the assist pattern width
generally becomes much smaller than the exposure wavelength of 193 nm. Thus, the impact of the topography of a mask
is examined using an electro-magnetic field (EMF) simulation. This simulation indicates that unwanted assist pattern
printings are brought about by assist patterns with a smaller size than expected by the Kirchhoff's approximation. The
difference, however, can be easily solved by giving a bias to the main pattern in the optimized mask. The main pattern
bias decreases DOF very little. Furthermore, DOF simulated with a thick mask model is roughly the same as that
simulated with a thin mask model. Therefore the topography of the optimized mask does not have an influence on the
assist pattern position of the optimized mask. From these results, we have confirmed that the 2D-TCC technique can be
extended to the optimization of 45 nm line patterns. As one of the notable features, the optimized aperiodic assist pattern
greatly reduces MEEF compared with the conventional periodic assist pattern. To verify the feasibility of the 2D-TCC
technique for 45 nm line, we performed experiment with an optimized mask. Experimental results showed that DOF
increased with the number of assist pattern as simulation indicated. In addition, a defect whose length was twice that of
the assist pattern did not have an influence on CD. From these results we have confirmed that the 2D-TCC technique can
enhance the resolution of 45 nm line and has practical feasibility.
In this paper, a new resolution enhancement technique named 2D-TCC technique is proposed. This method can
enhance resolution of line patterns as well as that of contact hole patterns by the use of an approximate aerial image.
The aerial image, which is obtained by 2D-TCC calculation, expresses the degree of coherence at the image plane of a
projection optic considering mask transmission at the object plane. OPC of desired patterns and placement of assist
patterns can be simultaneously performed according to an approximate aerial image called a 2D-TCC map. Fast
calculation due to truncation of a series in calculating an aerial image is another advantage. Results of mask
optimization for various line patterns and the validity of the 2D-TCC technique by simulations and experiments are
A newly developed sub-resolution assist feature (SRAF) placement technique with two-dimensional
transmission cross coefficient (2D-TCC) is described in this paper. In SRAF placement with 2D-TCC, Hopkins'
aerial image equation with four-dimensional TCC is decomposed into the sum of Fourier transforms of diffracted
light weighted by 2D-TCC, introducing an approximated aerial image so as to place SRAFs into a given reticle
layout. SRAFs are placed at peak positions of the approximated aerial image for enhanced resolution. Since the
approximated aerial image can handle the full optical model, SRAFs can be automatically optimized to the given
optical condition to generate the optimized reticle. The validity of this technique was confirmed by experiment
using a Canon FPA6000-ES6a, 248 nm with a numerical aperture (NA) of 0.86. A binary reticle optimized by this
technique with mild off-axis illumination was used in the experiment. Both isolated and dense 100 nm contacts (k1
= 0.35) were simultaneously resolved with the aid of this technique.
We have built a visible light point-diffraction interferometer with the purpose to characterize EUVL projection optics. The interferometer operates at the wavelength of 532 nm and utilizes two identical pinhole wavefront reference sources for generation of both signal and reference wavefronts. In the simple configuration of our interferometer, the main source of system error is the pinhole reference wavefronts. It is important that the reference wavefronts are calibrated and the calibration is stable. The calibration using our refractive test optic is reproducible to better than 0.1 nm RMS. The interferometer measured the wavefront of our refractive test optic with the repeatability of 0.1nm RMS. This paper will discuss the error sources and removal of the errors with experimental results.
As imaging properties of ArF Immersion optics are evaluated in a hyper-NA region, the polarization of illumination systems and vectorial mask diffraction play an important role. We investigate the effectiveness of polarized illumination for practical patterns including the border of dense line-and-space (L/S) patterns, semi-dense L/S patterns, isolated lines, and contact holes. The results show that polarized illumination is effective in projecting many patterns except semi-dense L/S patterns and relatively large contact holes. Secondly, we examine how bias settings of alternating phase-shift masks (AltPSMs) are affected by vectorial mask diffraction, which depends on the polarization of incident light and feature size on the mask. Although a reduction ratio of 8x facilitates bias settings compared with that of 4x, it is necessary to take into account the effect of vectorial mask diffraction even in the case of 8x. Since polarized illumination also simplifies bias settings, the illumination is useful for 4x projection optics.
High-index fluids have recently attracted considerable attention because they are capable of extending the numerical aperture of projection optics beyond the refractive index of water (n=1.44). We study imaging properties of 1.50NA projection optics with an immersion fluid of n=1.64 and the preliminary requirements of fundamental optical characteristics of the fluid.
According to sizes dictated by ITRS road map, contact holes are one of the most challenging features to be printed in the semiconductor manufacturing process. The development of 90[nm] technology FLASH memories requires a robust solution for printing contact holes down to 100[nm] on 200[nm] pitch. The delay of NGL development as well as open issues related to 157[nm] scanner introduction pushes the industry to find a solution for printing such tight features using existing ArF scanner. IDEALSmile technology from Canon was proven to be a good candidate for achieving such high resolution with sufficiently large through pitch process window using a binary mask, relatively simple to be manufactured, with a modified illumination and single exposure, with no impact on throughput and without any increase of cost of ownership. This paper analyses main issues related to the introduction of this new resolution enhancement technology on a real FLASH memory device, highlighting advantages as well as known problems still under investigation.
IDEALSmile is introduced as a new exposure technique that realizes k1 equals 0.29. In this paper IDEALSmile is targeted for contact hole patterns (C/H). The results validate that it is possible to simultaneously expose not only k1 equals 0.32 half-pitch dense and isolated C/H patterns, but also different pitches using Canon FPA- 5000ES3, which is impossible by conventional methods. Since these results are obtained using a binary mask and modified illumination with single exposure, there are no concerns with regards to a decease in throughput and an increase in cost of ownership. However, one of the issues in fabricating C/H patterns is the mask error enhancement factor (MEEF). Our simulation ha shown that IDEALSmile exhibits good MEEF. Although there are questions regarding optical microlithography for critical C/H patterning, the IDEALSmile exposure method has the potential to be the solution. By attaining k1 equals 0.32, printing 100nm C/H patterns can be achieved with a single exposure using KrF lithography, such as the Canon FPA-5000ES4. Furthermore the IDEALSmile technique using ArF or F2 lithography will be effective for C/H patterns below the 100nm node. There is no doubt that optical microlithography will continue for some time.
IDEALSmile is introduced as a new exposure technique. Since we have realized k1 equals 0.29, k1 equals 0.32 optical lithography is now achievable. In this paper IDEALSmile is targeted for contact hole patterns. The results validate that it is possible to simultaneously fabricate 110 nm (k1 0.32) half-pitch dense and isolated contact hole patterns using Canon FPA-5000ES3 (KrF, NA equals 0.73). Furthermore, our experimental results also show that it is possible to fabricate different half-pitch patterns at the same exposure dose, which is impossible by conventional methods. Since these results are obtained using binary mask and the modified illumination with single exposure, there are no concerns with regards to decrease in throughput and increase in cost of ownership. By attaining k1 equals 0.32 for contact hole patterns using binary mask with single exposure, printing 100 nm contact hole patterns can be achieved with single exposure using KrF lithography, such as the Canon FPA-5000ES4 (KrF, NA equals 0.80) scanner which will soon make its market debut. ArF or F2 lithography is effective as for contact hole patterns below the 100 nm node. There is no doubt that optical microlithography will continue for some time.