2. BACKGROUND

The basic approach to photomask fabrication is photolithography with optical instruments operating in the visible spectrum. In addition, non-optical techniques such as electron-beam [9] and X-ray lithography [10] are currently being used, on a limited scale, for mask fabrication. A master mask with an array of primary patterns is shown in figure 1. This mask includes a wide range of test, continuity, and alignment patterns used during the fabrication process.

Figure 1. An integrated-circuit photomask and its various patterns. (Reprinted with permission from The Bell System Technical Journal. Copyright 1970, The American Telephone and Telegraph Company.)

BACKGROUND

The procedures used to manufacture photomasks are not widely standardized throughout the IC industry, although there are basic steps common to mask fabrication for silicon devices. These steps are illustrated schematically in figure 2 and include the following: (1) pattern generation; (2) stepping the pattern over the photoplate; and (3) photoprinting the master onto the working mask.

The first step, pattern generation, is the transfer of the design information to a glass plate or master reticle with a pattern size typically ten times that to be printed on the silicon wafer. This transfer may be accomplished directly by a computer-controlled pattern generator or, as shown in figure 2, by photographically reducing artwork that previously has been cut from plastic film. A photographic reduction of 20X to 100X is generally used and for larger reductions involves two or more reductions. In the second step, the master reticle is placed in a step-and-repeat camera which produces multiple exposures of the same pattern on another glass plate. The pattern reduction for this step is generally 10X and, depending upon the size of the production wafer and. the chip area, several hundred to several thousand repeat patterns may be printed on the resulting master photomask. In the third step, the master photomask is contact printed to form a submaster, and the sub-master, in turn, is contact printed to give a working mask. All masters are generally inspected for defects, and for hard-surface masks, most of the sub-masters and working masks are also inspected. In addition, criticaldimension measurements are made on several selected areas of the mask patterns; all masters and from 10 to 25 percent of the working masks are measured.

The master photomask is usually photographic emulsion on glass. Although emulsion on glass is also used frequently for the sub-master and working masks, the use of hard-surface masks consisting of chromium on glass and iron oxide on glass is increasing significantly. The working masks are commonly 6.3 cm x 6.3 cm (2 1/2 in. x 2 1/2 in.) and 7.6 cm x 7.6 cm (3 in. x 3 in.); the use of masks measuring 10 cm x 10 cm (4 in. x 4 in.) is increasing with adoption of the 7.6-cm (3 in.) diameter silicon wafer.

The working masks are used as negatives, or stencils, to expose a thin layer of photoresist previously coated on a semiconductor wafer. After a sequence of development, chemical etching, and sometimes impurity diffusion, this entire procedure is repeated with as many as twenty different masks for the same wafer. Therefore, each mask must register precisely with the wafer patterns produced by the previous masks. The resulting wafer contains many repetitions of the desired IC and is subsequently diced.

3. APPROACH

The major effort to identify the measurement and optical problems associated with current photomask technology consisted of a literature search and field visits to photomask, IC, and instrument manufacturers and research organizations. The additional technical investigations of measurement and optical problems discussed in appendices A and B were selected on the basis of information gathered during the early phase of the study. Attendance by NBS personnel at four meetings of the American Society for the Testing of Materials (ASTM)/Committee F-1 on Electronics* afforded additional contact with the IC community and the opportunity to participate in discussions 'of measurement and optical problems.

The sources for the literature survey included the NBS library and computeraided literature searches. The computerized searches were obtained from the following organizations: (1) National Technical Information Service (NTIS), Springfield, Va.; (2) Smithsonian Scientific Information Exchange (SSIE), Washington, D. C.,; and (3) Defense Documentation Center (DDC), Alexandria, Va. Selected literature was obtained and reviewed. A bibliography of publications pertinent to the present study is given in appendix C.

The field visits by NBS personnel included four photomask suppliers, three research organizations, seven equipment manufacturers, and eight IC producers. Most of the IC producers had in-house mask-making facilities. On-site technical discussions were held with industry personnel. Generally, a plant tour of maskfabrication and mask-inspection facilities was included.

*Meetings at Palo Alto, Ca. (Sept. 6, 1973), New Orleans, La. (Jan. 15-16, 1974), Gaithersburg, Md. (June 12, 1974), and Scottsdale, Ariz. (Sept. 5, 1974).

4. RESULTS AND DISCUSSION

The present study reveals that significant dimensional-measurement and optical problems exist throughout the IC industry in the manufacture and use of photomasks. The relative importance of these problems to the various segments of the IC community depends primarily on the function and services of the particular facility, i.e., photomask supplier, IC producer, or instrument maker. The present discussion of these problems does not reflect any particular ordering of importance; this discussion treats in depth only those problems of widespread significance in the IC industry.

4.1. Optics

Although the present study includes the broad use of optics in photomask technology, the major optical problems were found to be unrelated directly to the photolithographic processing of photomasks. Instead, these optical problems are related to the existing measurement problems through the optical systems used for photomask measurements. Therefore, these optical problems will be discussed under the following section on measurement problems.

Several current observations that may not be regarded specifically as major optical problems do warrant a separate discussion in this section. First, the basic limitations imposed by diffraction and partial coherence on the formation of optical images have obviously not changed despite improvements in such areas as photolithographic process controls, photoresist resolution, and mechanical precision of optical instruments. These limitations have been acknowledged repeatedly in the technical literature by workers in the IC field [11-14] and were known to optical and microphotographic specialists prior to IC development [15-18]. A study of image formation by visible radiation passing through simple geometrical apertures, without the additionally degrading effects of lenses, mirrors, and windows, will show the following: photolithographic techniques cannot provide photomasks and IC devices with good dimensional fidelity for geometrical elements comparable in size to the wavelength of light (about 0.4 um to 0.7 μm). Linewidths of these dimensions have been produced with uv-visible radiation and deformable photomasks which apparently reduce diffraction effects [19]; this processing technique is currently not adaptable to production methods. For geometry smaller than about 1 μm, the effects of illumination coherence [20] on image formation become significant and must be considered by users of optical instruments as discussed later. In any event, a technology using radiation of shorter wavelengths than visible radiation appears necessary to provide high quality photomasks and IC devices with pattern dimensions well below about 1 μm. Electron-beam and x-ray lithography are possible considerations that presently exist on a developmental basis.