This paper discusses a novel pattern based standalone process verification technique that meets with current and future needs for semiconductor manufacturing of memory and logic devices. The choosing the right process verification technique is essential to bridge the discrepancy between the intended and the printed pattern. As the industry moving to very low k1 patterning solutions at each technology node, the challenges for process verification are becoming nightmare for lithography engineers, such as large number of possible verification defects and defect disposition. In low k1 lithography, demand for full-chip process verification is increasing. Full-chip process verification is applied post to process and optical proximity correction (OPC) step. The current challenges in process verification are large number of defects reported, disposition difficulties, long defect review times, and no feedback provided to OPC. The technique presented here is based on pattern based verification where each reported defects are classified in terms of patterns and these patterns are saved to a database. Later this database is used for screening incoming new design prior to OPC step.
Extreme ultra-violet (EUV) lithography has been planned for high-volume manufacturing (HVM) in 2014 for critical
layers of advanced nodes in the semiconductor industry. Process and proximity correction (PPC) and verification is
necessary in order to compensate various optical and other process effects in EUV lithography. Since the long-range
flare, mask shadowing effect, and lens characteristics all vary throughout the whole mask range, position dependent PPC and verification may be needed for accurate mask pattern synthesis. In this paper, we will study the PPC accuracy. The PPC flow uses a single PPC kernel set and a full-mask flare map for long-range flare correction. The lithography model is calibrated in accordance with this PPC flow. The lithography model is used to perform full-mask correction for the 10nm node test chip mask for BEOL/FEOL short loop flow development. The optimized full-mask patterns were placed on the mask and printed using a 0.25 NA EUV scanner at various focus and dose conditions. Printed wafers were measured by a CD-SEM and compared to post-PPC verification results.
As the semiconductor industry is moving to very low-k1 patterning solutions, the metrology problems facing process
engineers are becoming much more complex. Choosing the right optical critical dimension (OCD) metrology technique
is essential for bridging the metrology gap and achieving the required manufacturing volume throughput. The critical
dimension scanning electron microscope (CD-SEM) measurement is usually distorted by the high aspect ratio of the
photoresist and hard mask layers. CD-SEM measurements cease to correlate with complex three-dimensional profiles,
such as the cases for double patterning and FinFETs, thus necessitating sophisticated, accurate and fast computational
methods to bridge the gap. In this work, a suite of computational methods that complement advanced OCD equipment,
and enabling them to operate at higher accuracies, are developed. In this article, a novel method for accurately modeling
OCD profiles is presented. A finite element formulation in primal form is used to discretize the equations. The
implementation uses specialized finite element spaces to solve Maxwell equations in two dimensions.
In recent years, mask critical dimension (CD) linearity and uniformity has become increasingly important. The ITRS roadmap shows the mask CD control requirements exceeding those of the wafer side beyond the 45nm node. Measurements show that there are systematic, uncorrected proximity effects even when a state-of-the-art proximity effect correction (PEC) algorithm is used. The uncorrected proximity effect is predictable with a computational model. The model for e-beam lithography and etch process contains terms to model short-range pattern density effects and plasma shadowing effect in Cr-etch. The model is calibrated using CD measurements on a test mask. The model is valid for arbitrary 2-D patterns. We present a model-based mask process compensation (MPC) method which applies geometric changes to polygons as in OPC. We discuss the goodness of model fit to the calibration data; verification of the calibrated model by SEM images; and the improvement obtained by MPC. The mask writing error, i.e. final inspection CD minus incoming database CD, was reduced by a factor of 2 through the use of MPC.
We present a methodology for building through-process, physics-based litho and etch models which result in accurate and predictive models. The litho model parameters are inverted using resist SEM data collected on a set of test-structures for a set of exposure dose and defocus conditions. The litho model includes effects such as resist diffusion, chromatic aberration, defocus bias, lens aberrations, and flare. The etch model, which includes pattern density and particle collision effects, is calibrated independently of the litho model, using DI and FI SEM measurements. Before being used for mask optimization, the litho and etch models are signed-off using a set of verification structures. These verification structures, having highly two-dimensional geometries, are placed on the test-reticle in close vicinity to the calibration test-structures. Using through-process DI and FI measurement and images from verification structures, model prediction is compared to wafer results, and model performance both in terms of accuracy and predictability is thus evaluated.
A technique for measuring the profile of the illumination in the pupil of a lithography projector is presented. The technique is based on exposing pinhole patterns on a wafer at different dose and defocus settings, and processing the scanning electron microscopy (SEM) images of the printed pinholes. The latent image intensity at the edges of the resist patterns equals the dose-to-clear. This establishes a multitude of equations, each of which states that the latent image intensity at a particular field location, dose, and defocus is known. The intensity distribution in the pupil of the illuminator is obtained by solving a large system of such equations, subject to the constraint that the intensity distribution is non-negative. An image processing algorithm based on nonlinear diffusion is used for finding coordinates of points on the edges of resist in SEM images. The results of the inversion for 193-nm stepper with 0.55/0.85 annular illumination and numerical aperture of 0.75 at five exposure field locations are presented.
Computational models used in process proximity correction require accurate description of lithography and etch processes. We present inversion of stepper and photoresist parameters from printed test structures. The technique is based on printing a set of test structures at different dose and defocus settings, and processing the CD-SEM measurements of the printed test structures. The model of image formation includes: an arbitrary pupil illumination profile, defocus bias, flare, chromatic aberrations, wavefront errors and apodization of the lens pupil; interaction of vector EM waves with the stack of materials on the wafer; and molecular diffusion in photoresist. The inversion is done by minimizing a norm of the differences between CDs calculated by the model and CD-SEM measurements. The corresponding non-linear least square problem is solved using Gauss-Newton and Levenberg-Marquardt algorithms. Differences between the CD measurements and the best fitting model have an RMS error of 1.63 nm. An etch model, separate from the lithography model, is fitted to measurements of etch skew.
Computational models used in process proximity correction require accurate description of the pupil illumination function of the lithography projector. Traditional top-hat approximation for pupil illumination function is no longer sufficient to meet stringent CD control requirements of low-k1 applications. The pupil illumination profile can change across the exposure field, contributing to across-field linewidth variation. We present a measurement of the pupil illumination based on exposing pinhole patterns on a wafer at different dose and defocus settings, and processing SEM images of patterns printed in photoresist. The fundamental principle of the method is Abbe's formulation of image formation: the intensity-image formed in resist is an incoherent, linear superposition of images each one of which is formed by illuminating the photomask by a single plane-wave. A single plane-wave that is incident on the photomask maps to a single point in the Fourier-transform aperture of the illuminator. The pupil-fill of the illuminator is obtained from SEM images by a model-based method consisting of these steps: First, resist edges in the SEM images are detected by an edge detection algorithm based on Perona-Malik diffusion. Coordinates of the points on the resist edge are obtained with respect to a reference ruler. The image intensity at any resist edge is equal to the dose-to-clear. This provides an equation for the image intensity at each point on the edge of a pinhole image. Multiple values of dose and defocus, and multiple points on each resist edge provide a large system of equations. The result of the inversion for a 193nm 0.75 NA stepper with σ = 0.55/0.85 annular illumination at five exposure field locations is presented. The CD difference between the nominal top-hat illumination and the inverted illumination was up to 1.8 nm for 1:1 line and space features ranging from 100nm to 300nm. Variation of the illumination along the long-dimension of the slit of the scanner caused 0.6 nm of CD variation for the same 1:1 dense lines.
As minimum groundrules for chipmanufacturing continue to shrink the lithography process is pushed further and further into the low k1 domain. One of the key characteristics of low k1 lithography is the fact that process variations are increasingly more difficult to manage and the resulting CD variations are significant relative to the nominal dimensions. As a result it is quite common for process engineers to define process budgets, mostly dose and focus budgets. These budgets summarize the effects of various exposure and process contributors and provide the range within which the process is expected to fluctuate. An important task of process design is to ensure that within these budgets no catastrophic patterning failures occur, but even more importantly that the CD variations remain within the allowed design tolerances. Various techniques have been developed to reduce the sensitivity of the lithography process to process variations, among those one of the more prominent and quite widely adopted techniques are subresolution enhancements. Traditionally subresolution assist features are placed in the design using rules based approaches. This work presents a model based approach to assist feature placement. In this approach assist features are placed such that the resulting mask exhibits the minimum sensitivity to the specific process variations encountered. The type of process variation may be defined by the user as serious of worst case conditions, for example in dose and focus. The technique however is general enough to allow a variety of process variations to be included. This work focuses on demonstrating the key concept and show it's validity. The approach demonstrated in this work is fully integrated with the process budget concept and therefore allows a "process aware" mask optimization.
In this article, we present micromachined 2D array flextensional transducers that can be used to generate sound in air or water. Individual array elements consist of a thin piezoelectric ring and a thin, fully supported, circular membrane. We manufacture the transducer in 2D arrays using planar silicon micromachining combined with reactive ion etching of bulk silicon to provide back access holes. Such an array could be combined with an on-board driving and an addressing circuitry.
This article presents a technique for resist deposition using a novel fluid ejection method. An ejector has been developed to deposit photoresist on silicon wafers without spinning. Drop-on-demand coating of the wafer reduces waste and the cost of coating wafers. The novel piezoelectric fluid ejector is based on a variation of the design of a flextensional transducer that excites axisymmetric resonant modes in a clamped circular membrane. The ejector is made by bonding a thin piezoelectric ring to a thin, fully supported, circular membrane. The ejector design is optimized for maximum flexure at the lowest order resonant frequency using finite element modeling. The resist is placed behind one face of the membrane which has a small orifice (50 - 150 micrometers diameter) in its center. By applying an ac signal across the piezoelectric element, continuous or drop-on-demand ejection of the resist is achieved. Shipley 1400-21, 1400-27, 1805, and 1813 resists were used to coat sample 3' wafers. Later, these wafers were exposed and developed. The deposited resist films was 3.5 micrometers thick and had a surface roughness of about 0.2 micrometers . The ultimate goal is to deposit resist films with a thickness of the order of 0.5 micrometers , and a surface roughness of the order of 30 Ao. Such goals can be attained by using micromachined multiple ejectors presently under development, or with better control over the deposition environment. In the micromachined configuration, thousands of ejectors will be made into a silicon die and thus allow for full coating of a wafer in a few seconds. Coating in a clean environment will allow the lithography of circuits for microelectronic applications.