Radiation imagery chemistry: process – composition – or product th – Radiation modifying product or process of making – Radiation mask
Reexamination Certificate
2000-03-23
2003-07-22
Huff, Mark F. (Department: 1756)
Radiation imagery chemistry: process, composition, or product th
Radiation modifying product or process of making
Radiation mask
C430S030000
Reexamination Certificate
active
06596442
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to a technique for controlling the size of an image formed by using a template. More particularly, the present invention relates to a sub-grid biasing technique for fabricating photomasks used in VLSI lithography for semiconductor processing, where the photomask has a halftone type pattern so that images formed on semiconductor wafers are related, but not identical, to the sizes and shapes of the photomask features.
BACKGROUND OF THE INVENTION
Photomask manufacturers continue to be challenged by the demands of customers who require smaller and more precise features on wafers. In particular, the need for subtle differences in line width on features in the same reticle is forcing the use of smaller pixel sizes on raster scan e-beam systems to write customer patterns on a photomask.
A smaller pixel size is beneficial when performing, for example, optical proximity correction of line width biasing. On the logic gate level, for instance, optical proximity effects cause lines situated in different environments, which nominally are of the same dimension, to print differently. This problem may be overcome by biasing the mask features as a function of pitch.
The efficiency of this approach is limited by the pixel size or design grid &Dgr;. In general, when an image is printed using a template, the width of a line in the template is limited to integral multiples of &Dgr;. To ensure that an error in the printed image is no larger than &Dgr;/2 requires a design grid of &Dgr;. In the case of photomasks used for printing on wafers, the design grid is forced to be smaller still as the wafer critical dimension (CD) becomes increasingly sensitive to mask dimension error (expressed as the mask error factor MEF) at small k
1
factors (k
1
being CD divided by &lgr;/NA, where &lgr; is the wavelength of the light and NA is the numerical aperture of a corresponding exposure system). A &Dgr;/2 error bound requires a design grid of &Dgr;/MEF.
For example, if an error bound &Dgr;/2 of 2.5 nm is desired (which typically will be required in the near future), and the MEF is 2, the required design grid &Dgr; on the mask is then 2.5 nm. However, the use of a smaller design grid (smaller pixel size) increases the time required to write a pattern on the mask, which in turn reduces throughput and increases production costs. As shown in
FIG. 1
, the writing time for a mask of a given size increases quadratically with decreasing design grid size. Although state-of-the-art mask writers allow a design grid &Dgr; of 2.5 nm at 1×reduction, a 6 inch square reticle with such a design grid would require a prohibitively long 30 hours of write time. A design grid of 2.5 nm is thus too small for efficient mask writing.
Accordingly, it is desirable to design with a grid much larger than 2.5 nm (for example, &Dgr;=25 nm for which the write time is approximately one hour), but still achieve image size increments of 2.5 nm. It will be increasingly important to have this capability when designing future generations of electronic circuits.
A conventional technique for solving this problem is called halftone biasing. The halftone biasing technique incorporates the application of a sub-resolution halftone screen to the edges of features.
FIGS. 2A-2C
illustrate the conventional halftone biasing technique. In
FIG. 2A
, a mask has a shape
20
formed therein of width W. Since the mask is written with a design grid &Dgr;, width W must be an integral multiple of the design grid &Dgr;, so that W=n&Dgr; where n is an integer. Suppose, however, that a critical dimension of W′=W+&Dgr;/2 is desired; that is, the printed image
21
is desired to have a width of W′ (see FIG.
2
B). The halftone biasing technique can be used to achieve this effect. As shown in
FIG. 2C
, an array of protrusions
22
of width &Dgr; is formed on the edge of the mask feature of width W. These protrusions (or “teeth” on the edge of the feature) have a tight enough segmentation period or pitch, P, so that their details are not resolved when the line is imaged (for example, printed on a semiconductor wafer). However, the presence of these protrusions on the mask influences the width of the printed image. The amount of this influence is determined by the “halftone percentage” of the arrangement of the teeth.
This approach is analogous to halftone printing. Since the exposure system acts as a low-pass filter, spatial periods less than &lgr;/NA(1+&sgr;) are not resolved (&sgr; being the partial coherence factor). For mask features having periods beyond this resolution limit, only the average transmittance is captured by the exposure system.
For example, as shown in
FIG. 2C
, protrusions
22
have a width of one design grid &Dgr; and a length D (distance along the edge of the feature) of one design grid &Dgr;; these protrusions are placed on the edges of the feature
20
with a segmentation period or pitch P=4&Dgr;. The halftone percentage, per edge of the feature, is defined by (D/P)×100 (%). Although the protrusions are evident on the mask pattern, they are not replicated in the printed image. Instead, the printed image
23
has straight edges with a critical dimension dependent on the halftone percentage. In this example, with D=&Dgr; and P=4&Dgr;, the halftone percentage is (1/4)×100%=25%. Each edge of the printed image is thus shifted by &Dgr;/4. The critical dimension of the feature, when actually printed, is thus W′=W+2×(&Dgr;/4), or W′=W+&Dgr;/2. The halftone biasing technique thus permits finer control of the printed image without reducing the design grid (or in e-beam mask generation, the address unit size).
In general, an edge may be biased with an increment of &Dgr;
if the edge is segmented into periodic units of n&Dgr;, where n is a positive integer. (Accordingly, &Dgr;
is termed the apparent grid.) However, the increment cannot be made arbitrarily small, because the segmentation period n&Dgr; is limited by the resolution of the exposure system; that is, n has a maximum value n
max
, given by
n
max
&Dgr;≦&lgr;/NA
(1+&sgr;).
For an exposure system where &lgr;=248 nm, NA=0.68, &sgr;=0.8 and &Dgr;=25 nm, this expression yields n
max
=8. In such an exposure system, an edge of a feature line
31
designed with a grid of &Dgr; can be biased in increments of &Dgr;/8, if the edge is segmented into periodic units of 8&Dgr; as shown in
FIG. 3
; the printed line
32
then will have a width W′=W+2×(&Dgr;/8)=W+&Dgr;/4.
There is a need for a sub-grid biasing method which can further reduce the available biasing increment (that is, further reduce the apparent grid relative to the design grid), thereby permitting improved control of the printed feature size while limiting the required writing time for the photomask.
SUMMARY OF THE INVENTION
The present invention provides a technique, based on concepts of halftone printing, for controlling feature dimensions in a printed image at very small increments. The invention permits these increments to be smaller than the smallest addressable unit of the template used to produce that image. Specifically, this technique permits fabrication of photomasks yielding images with sizes differing from a nominal width by increments which are small fractions of the minimum template size or pixel size. Stated another way, the present invention provides a technique for reducing the ratio of the apparent grid to the design grid, without decreasing the size of the design grid.
According to a first aspect of the present invention, a template for forming an image includes a feature having one or more edges, with a first array of shapes and a second array of shapes disposed on the edges. The first and second arrays have a first and a second segmentation period, respectively, and the first and second segmentation periods are different. In a typical arrangement, the feature is a li
Ferguson Richard A.
Liebmann Lars W.
Wong Alfred K.
Anderson Jay H.
Huff Mark F.
International Business Machines - Corporation
Mohamedulla Saleha R.
LandOfFree
Asymmetric halftone biasing for sub-grid pattern adjustment does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Asymmetric halftone biasing for sub-grid pattern adjustment, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Asymmetric halftone biasing for sub-grid pattern adjustment will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3065719