During 1980, the Photomask Technology Operations activity at the RCA Solid State Division at Somerville, N.J., initiated a program entitled "Metrology Techniques for Linewidth Measurements." The purpose of this program was to improve accuracy, precision, and correlation of linewidth measurements on photomasks. With the cooperation of the NBS (National Bureau of Standards) Dimensional Metrology Semiconductor Materials and Processes Division, this program reached an advanced state of development. Correlation tests between the video image-scanning system used by the RCA Photomask Technology Operations and an optical image-scanning system used by Qualitron, Danbury, CT. indicate the overall linewidth measurement error to be within ± 0.1 micrometer. This error includes the ± 0.05 micrometer uncertainty of the NBS SRM 474 standard. Routine production correlation linewidth measurements between the ITP Mark II and ITP Mark III Video Image-Scanning Systems show the uncertainty (systematic and random error) to be within ± 0.05 micrometers at an 87% acceptance level.

To minimize the risk of damaging expensive high quality linewidth standards, secondary standards can be made for use in a production environment. This paper discribes a simple method of estimating the total uncertainty of secondary standards. The method includes procedures for estimating random and systemetic errors of the linewidth measuring instrument used to characterize the secondary standards. Given the total uncertainty of the NBS SRM474 standard, the total uncertainty of the secondary standards may be estimated.

Both antireflective (AR)-chromium and bright-chromium photomasks are currently used in the production of integrated circuits. Differences in the optical transmittance and reflectance of these photomasks can significantly change the line-image threshold required for accurate edge detection in optical microscope linewidth measurements. The suitability of using a calibration curve based on an AR-chromium optical linewidth-measurement standard (SRM 474) from the National Bureau of Standards (NBS) to correct linewidth measurements on other types of photomasks is discussed. Linewidths on each of three chromium photomasks of different chromium thicknesses were measured on four different types of optical microscope linewidth measurement systems. These measurements were corrected using an SRM 474 and compared with measurements made on the NBS optical linewidth calibration system. For the two bright-chromium specimens with low transmittance, the residual differences between the corrected values and the NBS values as measured on the NBS calibration system are generally less than ±0.05 μm for three of the measurement systems. For the see-through AR-chromium photomask with a higher transmittance, the calibration curve does not correct all systematic errors greater than ±0.05 μm. These results support theoretical studies showing that the degree of correction for systematic linewidth errors varies with the transmittance of the chromium photomask being measured and with the measurement system.

Optical linewidth measurements on patterned wafers are complicated by the wide variety of materials and correspondingly wide variation in optical parameters, complex refractive index and thickness, used in the manufacture of integrated circuits. It has been shown that in addition to linewidth, two key parameters, the normalized local reflectance R and the optical phase difference Φ at the line edge, determine the characteristics of the optical image and, therefore, affect the accuracy and precision of linewidth measurements. Both of these parameters, R and Φ, are dependent upon the illuminating wavelength or spectral bandpass and the coherence parameter of the optical system. To achieve the measurement precision and accuracy required for VLSI dimensions (e.g., 10% tolerance for 1-μm linewidths), it is necessary to control coherence, spectral bandpass, and image integrity as well as to achieve reproducible edge detection and focus criteria. When a system can be operated without further operator intervention despite changes in the materials being mea-sured, it is possible to calibrate the linewidth measurement system using a standard fabri-cated from only a few materials representing a range of image characteristics. The desirable characteristics of such a standard are discussed with respect to durability, edge definition, and equivalence of the image characteristics to materials used in the manufacture of ICs. A prototype design consisting of combinations of SiO2 and chromium layers on a silicon substrate is presented.

While light scattering is widely used for particle detection, this technique has not been applied to automatic inspection of microlithographic reticle surfaces, because the periodic patterns on the surface behave as diffraction gratings producing signals which are hard to distinguish from those produced by particles. Automatic reticle inspection systems have been limited to complex pattern recognition schemes which require lengthy inspection times and a large data base of individual reticle patterns. Light scattering techniques yield simpler systems that have higher throughput and require no apriori knowledge of the reticle pattern. The instrument described in this paper employs a highly focused He-Ne laser beam to alternately scan the reticle surfaces, and signal processing to identify the size and location of contaminants on the surfaces of the reticle. From these data, a quantitative decision as to the reticle's acceptability can be rendered. A system that includes this instrument, along with the means for automatic reticle handling reduces contamination problems associated with manual intervention. This system is presently incorporated in GCA's 4800 DSW wafer stepper when configured with an automatic reticle changer (GCA's 5510 ARC).

The thermal conduction module (TCM) used in the IBM 3081 processor series incorporates a 90mm square multilayer ceramic substrate which requires thin metal masks for thick film personalization. The LSI semiconductors used on the TCM's require similar masks. These masks require accurately located features and precise cavity geometries and surface characteristics. The type of mask encountered in manufacturing is described, and the current inspection, measurement, and data handling techniques are discussed. The need for an improved inspection system is highlighted.

During the next five to ten years we can expect optical lithography to reach its limits, which may be defined as the printability of minimum feature sizes ranging from about 0.4 μm to 0.8 μm.

Characterizing wafer exposure equipment is becoming more important with the increasing use of such equipment. The process of fully characterizing a wafer stepper involves making several hundred measurements to sub-micron tolerances. This paper describes a method of using the automatic reticle-to-wafer aligner to collect the large number of measurements necessary for system characterization. The analysis method involves exposing an array of alignment targets, and then measuring the X-Y coordinates of the targets using the automatic aligner. Using this method, it is possible to obtain precise X-Y coordinate data at the rate of thirty measurements per minute. The coordinate data is recorded on tape cassette and transferred to a desk-top computer for data reduction. The data is analyzed to determine the error contribution of each of a number of systematic error sources. Typical data for a number of systems is presented, and the data reduction method is described. One primary application for this method is to characterize level-dedicated production wafer exposure systems.

We describe an approach to overlay measurement based upon well-known electrical probe techniques. The advantages realized by our approach are process simplicity, measurement precision of ±0.01μm, independence from imagery and process variations, site-by-site accuracy verification, and the capability of performing many successive overlay evaluations on a single reference wafer. We present the results of overlay performance evaluations of state-of-the-art 1:1 projection printers, including alignment and distortion components and the variable magnification capability.

A test structure is a microelectronic device that is fabricated by the same process used to fabricate integrated circuits (ICs) and can be tested electrically to determine important process parameters. Test structures can be used to evaluate semiconductor materials, evaluate and control process uniformity, measure device and circuit parameters, obtain input parameters for circuit simulation programs, and determine the performance of processing equipment. This paper reviews previous work at NBS on the design, measurement, and application of two types of test structures that have been used for evaluating lithographic processes and lithographic equipment performance. First, the cross-bridge sheet-resistor test structure is described. Test results from electrical measurements on this structure can be used to determine the electrical linewidth of a conducting layer. The use of test chips containing arrays of identical cross bridges for determining the uniformity of a lithographic process will be described. Analysis of test results from these arrays has been used to identify and separate the contribution to linewidth nonuniformities introduced by individual equipment and processes. The precision to which linewidth can be determined using this structure is discussed. Also, an electrical alignment test structure for determining the misalignment between two photomask steps is described and an example of its use presented. Finally, an automated dc parametric test system used for measuring these structures is described.

Metal crossbridge electrical test structures have been used to characterize linewidth non-uniformities caused by lithographic systems and processing. Artifacts can complicate the interpretation of electrical linewidth results. Self-heating in crossbridges is an artifact which can cause errors of more than 15%. Self-heating increases as the linewidth decreases. A procedure for eliminating self-heating effects is discussed.

The dependency of wafer flatness on high temperature and Chemical Vapor Deposition (CVD) processes has been quantified for a 400 gate array bipolar process. Experimental data is presented which describes wafer flatness variations at six critical front-end process steps.

A variety of techniques is used for the measurement of thin film thicknesses of the order of one micrometer. These techniques include stylus profilometry, multiple beam interferometry, dual beam interferometry, guided waves, channel spectra and ellipsometry. The principles underlying each of the techniques are discussed and where available, experimental comparisons of the techniques are presented. Advantages, disadvantages, and sources of error are also discussed.

Simultaneous measurement of the refractive index and thickness of a thin film to a sufficient degree of accuracy is usually cumbersome and often requires an elaborate data reduction procedure. In comparison with the conventional techniques that use either a spectrophotometer or an ellipsometer, the prism coupling method, which relies on the coupling of light waves into thin films by a guided wave mechanism, could be more accurate and is easily adaptable for rapid measurements. Prism couplers, though developed mainly to characterize thin-film guided-wave optical devices, are also useful in diverse applications, particularly for the semiconductor industry. The design guidelines of a prism coupler, which involve considerations of parameters such as prism features, light beam, and air gap at the coupling point, have been identified, and the accuracy requirements determined. The results obtained by a prism coupler for the very small absorption due to gap states in hydrogenated amorphous semiconducting thin films are also presented.

Ellipsometry as a technique for the measurement of the optical constants (n,k,d) of two layer films on substrates is presented in a tutorial fashion. The ideal two layer film model is described and departures from ideality that are likely to be encounterd in a real sample are discussed. The ellipsometric measurement of the change in polarization state upon reflection as well as the determination of the optical reflection angles DELTA and PSI are reviewed. Extraction of the optical constants from the DELTA, PSI measurement is discussed at length. Examples are drawn from double film systems of interest to the semi-conductor industry but the principles and procedures of analysis illustrated by these examples have general application.

The effects of improving the accuracy of the angle of incidence on the ellipsometric determination of thickness and refractive index of oxide and nitride films on a silicon substrate are analyzed. It is found that the accuracy of a determination of a film's parameters, thickness and refractive index, depends as much or more on the accuracy of the angle of incidence measurement as on the accuracy of the measurement of the ellipsometric angles Δ and ψ. If measurements of Δ and ψ are made close to the principal angle of incidence, the accuracy of the determined film parameters can be improved by measuring the incident angle to an accuracy better than Δ and ψ. This is especially true for thin films of oxide less than a few tens of nanometers. Because of the higher refractive index of silicon nitride relative to silicon dioxide, a nitride film's thickness can be determined more accurately than an oxide film's thickness. Therefore, silicon nitride may make a good candidate film for a standard thickness sample.