A computer-controlled spectroscopic ellipsometer of high accuracy has been designed and constructed. A theta-two-theta goniometer unit and optical rail system allows various ellipsometric methods to be used to measure the parameters A and 4). Three important methods under study for accuracy, precision, and speed of measurement are the conventional null method, the rotating analyzer method, and the principal angle method. All the goniometer angles, including the angle of incidence, can be measured to an accuracy of 0.001 deg. The present light sources are two lasers with fixed wavelengths, 632.8 nm and 441.6 nm, in addition to a monochromator that can be used to scan the wavelength range from 190 to 2600 nm. A unique sample alignment system which utilizes two quadrant detectors has been developed and a simple but very effective nulling scheme is used. This instrument is primarily used for the metrology of semiconductor materials and for the calibration of reference standards for thin film thickness and refractive index.

Ellipsometry is a very useful tool for measuring Si0₂ film thickness and other surface characteristics of silicon wafers; however, the 633-nm source used in most commercial ellipsometers does not permit measurement of subsurface characteristics. Infrared light can be used to measure thickness and complex index of buried layers regardless of the thickness of the overlayers. However, subtle complications in data collection or data analysis can result from the back surface reflection from the silicon wafer. This problem is described with measurement examples of silicon wafers with thin, transparent metal layers. A technique is also discussed for measuring the thickness of a buried layer in silicon.

The yield of integrated circuit devices made from a semiconductor wafer is known to depend critically on the quality of the bulk material and the perfection of the surfaces generated. Two new quality control instruments which measure infrared strain birefringence and quantify the residual defects, such as digs and scratches and contamination on the surface of the wafer, will be described. Typical results obtained on a variety of wafers will be presented.

Energy dispersive x-ray analysis (EDS) is used for non-destructive measurement of photoresist thickness on integrated circuit geometries as small as 2 microns. The technique is based on the principle that the characteristic x-ray emission intensity from the underlying material varies with the photoresist coating thickness. Calibration curves are used to relate this intensity to the photoresist thickness. The method is applied to I. C .structures to acquire data which was inaccessable to optical methods and which previously would have required destructive SEM cross-sections. Data can be collected from any location on a wafer, making this a technique attractive for process development. Mathematical estimates of the electron-solid interactions are used to investigate the ultimate resolution of the technique. The validity of the method is demonstrated by comparison to SEM cross-sections. Both test structures and real integrated circuit device patterns are examined. The test structure results include measurements on small aluminum lines over 1 and 2 micron high steps. Real device results include resist thickness on the severe toography found on the second level metal layer in a double layer metal interconnect structure.

When measurements are made during the production of integrated circuit wafers on wafers that are out of specification, the wafers are considered misprocessed. Measurement accu-racy must be better than the specification to assure that the measurements are not part of the misprocessing problem. Optical tools have been used in the past for these critical dimension (CD) measurements. These tools have evolved from optical filars to TV pattern recognition to stepping a PMT aperture. This has improved the repeatability of measuring a single site multiple times by many operators to ±0.04 pm, 3.0 σ or better. However, with CDs of 1.0 μm or less, optical tools are reaching their limits of usability.

A model is described for the formation of diffraction limited optical images of chrome on glass photomasks using an optical microscope. The effect of the imaging transducer is also considered and a Gaussian model for a television camera is presented. National photomask calibration standards were examined using bright and dark field illumination and the results are compared with the images predicted by the models. From the results it is concluded that the physical edge and the material properties of the chrome coating cannot be ignored when making precise linewidth measurements on photomasks. A dark ground measurement technique is described which makes use of the physical edge of the chrome in order to determine linewidths. It is demonstrated how this technique may be used with bright field illumination in order to measure linewidths on semiconductor wafers precisely. Both the photomask and wafer linewidth measurements may be automated and a method of adapting the technique for automation is described.

An automated system for measuring critical dimensions of hard disk heads in a production environment has been developed and put to use successfully in manufacturing. It measures features ranging in size from less than a micron to more than half a millimeter. The Automated Inspection Microscope (AIM), makes reliable measurements of these dimensions seven to nine times faster and with about twice the precision than the conventional methods which it replaced. A special purpose region building and merging algorithm produces results which are very dependable when contaminants, scratches or chips are present along with the feature to be measured. The system inspects loads of ninety six heads at a time with little or no operator intervention. This paper discusses the hardware components of the system, its programming, and its role as part of a parts tracking and inspection system. Particular attention will be given to the algorithms and techniques used in measuring the head gaps, which are are in the micron size range.

Linewidth measurement of photoresist structures continues to be an integral part of the wafer fabrication process for exposure and development control of all critical process steps. The introduction of 1 pm geometries into a routine production environment, such as the Hewlett-Packard NMOS III VLSI processl, poses formidable measurement problems. Traditional brightfield linewidth measurement techniques, which were adequate for circuit geometries of 3-4 μm and above, have been generally unacceptable for 1 pm complex structures. The principal limitation is that the reflected image profile depends not only on the resist line parameters but also on the optical properties of the substrate and all transmissive sub-layers. As a consequence, even a carefully optimized brightfield measurement system will measure identical resist lines differently due to normal process variations in the substrate and subsurface layers. In-process SEM measurements can be functionally adequate but our experience in routine wafer fabrication has demonstrated severe drawbacks with respect to cycle time, reliability, operator skill and space requirements. To overcome these problems a new technique using fluorescence was investigated. The fluorescence principle was successfully reduced to practice on PMMA photoresist lines, incorporating a bleachable dye (Coumarin 6). The PMMA is the bottom layer resist in a two-layer production photoresist process described by K. Bartlett, et al. 2. The fluorescence technique has all the advantages commonly associated with optical measurements and in addition it virtually eliminates the measurement variations due to the sub-strate and sub-layers. The paper will cover: 1. The parameters of the sample and measurement systems that have an impact on linewidth determination. 2. The experiments and statistical methods to systematically identify problems and optimize critical dimension measurements. 3. The results obtained with brightfield optical metrology. 4. The theory and advantages of fluorescence metrology. 5. The results obtained on production VLSI wafers using fluorescence measurement of critical dimensions

The fabrication of small VLSI circuits (1.25 μ m minimum feature size) requires accurate, routine linewidth measurements of resist structures after exposure and development. The evaluation of several commercial instruments, all of which used a brightfield measurement technique, revealed that they suffered from the same limitations and that none of them were sufficiently accurate. This paper describes experiments demonstrating that thickness variations of films beneath the photoresist are a major source of inaccuracy with these systems. After exploring several potential solutions, UV excitation of a fluorescent dye in the photoresist and measurement of the resulting fluorescence image was selected and enhanced. This technique is insensitive to the optical properties of the subsurface layers and offers improved accuracy. Extensive measurements of 1 to 1.5 ym photoresist structures were made with a commercial measurement system and fluorescent illumination. The desired long term precision of 0.015 μ m (lσ) and accuracy (when compared to SEM measurements) of 0.03 μ m (l σ ) ^were achieved. Fluorescent linewidth measurement systems are now in routine use at Hewlett-Packard's VLSI production facilities.

In a previous paper, 1 a waveguide model was developed for the imaging of micrometer-sized lines patterneu in thick layers of dielectric materials (silicon dioxide) with application to linewidth measurement on integrated-circuit wafers. This paper describes the extension of this work to metals characterized by their complex index of refraction, n + iK, as well as the inclusion of a sublayer such as a silicon dioxide insulating layer. This extension allows the modeling of optical imaging and linewidth measurement on metal-on-silicon (MOS) structures. It is shown that the image structure for metals at and near focus is different from that for dielectrics. Thick and thin layer (less than 200 nm) imaging is compared. Experimental image profiles of metal lines at and near focus are also shown. The experimental data were obtained from a bright-field microscope using a laser source (530 nm) and controlled spatial coherence.

A new method for linewidth measurement on wafers is proposed. Owing to a Fourier transformation this method makes use of a good deal of the information available in the optical image. It applies in the 1 pm to 6 pm range by-passing data such as the thickness and optical index of the material being measured, provided that the line itself is less than 500 nm thick. Only the image sensor has to be calibrated i.e. no standard reference material is needed. Repeatability of linewidth measurement is within 0.01 μm, and accuracy is ± 0.5 % in the whole range of validity of the model presented.

An optical accessory has been devised for doubling the resolution of the Hewlett-Packard linear and plane mirror interferometers. This device, termed the double pass attachment, does not fold one of the two interfering beams, as in previous attempts to extend resolution optically, and hence does not introduce error due to its own motion. The simple addition of a quarter wave plate to the attachment can be used to give a differential version of the plane mirror interferometer. Various configurations of this differential interferometer, and their application to I.C. lithographic and inspection equipment, are discussed.

The trend to smaller feature sizes in the integrated circuit industry has resulted in the need to measure linewidths, periodicities, and registration and overlay errors with greater accuracy and precision. The scanning electron microscope (SEM) has potential as a measurement tool for fine line geometries due to its increased resolution and depth of field compared to optical systems. However, the unique and precise determination of magnification in the SEM is a difficult task without the use of tedious calibration procedures which typically must be performed each time a magnification-determining instrument parameter is altered. A new measurement technique is described which effectively eliminates the need for SEM magnification calibration when performing measurements over a broad range of operating conditions. Examples of measurements made on resist-coated silicon, chrome-on-glass masks, and the NBS 484 standard reference material are presented to illustrate the effectiveness of the new differential measurement technique.

SEM metrology is shown to be the technique of choice for linewidth measurements in cases where high-resolution (0.01 μm, 3-sigma), small sample height (0.005 μm), and large depth-of-focus (DOF>1 μm) are appropriate. Examples are submicron structures, materials with varying surface roughness, and patterned materials with high aspect ratios, respectively. The most consistent measurements are made when compositional contrast is suppressed by gold coating so that the structure in the secondary electron profiles is dominated by topography. Topographic effects due to grain size in aluminum and standing wave structure in photoresist are readily observed.

IC metrology requirements, as feature size reaches one micron or less, are beginning to exceed the capabilities of light optical systems. It is well known that scanning electron microscopy provides sub-micron measurement capabilities. Thus, SEM based techniques are often considered as an alternative to the more conventional metrology techniques. It is possible to modify a commercial, general purpose SEM to perform basic metrology functions. However, in the modification of an instrument designed for other purposes, some trade-offs of performance are inherent. We feel that the most important features of an SEM based metrology system are reliability, reproducibility, throughput and the ability to perform automated sample alignment and that these parameters can only be optimized by the design of a totally dedicated IC metrology SEM. In this paper we will describe the design philosophy, development, and performance of such a dedicated system. The major design objective includes totally automated, non-destructive operation. This required the development of a low keV electron beam column controlled by digital, microprocessor controlled electronics. Fully automated operation, including site location, rotation and height correction, auto-focus, contrast/brightness and astigmatism control required extensive software development; as did the long term accuracy and reproducibility requirements. A hardware description and performance summary of the first generation commercial metrology SEM resulting from this work will be given, along with an outline of current developments towards a second generation system with high throughput and larger wafer handling capabilities.

The progression of Semiconductor device sizes into the submicron region requires inspection and measurement techniques with higher resolution than that presently available with the optical microscope. The increasing need of the semiconductor industry for accurate dimensional measurement and inspection in the submicron range for LSI, VLSI and VSHIC circuitry has led it to the higher resolution and depth of field afforded by the SEM In-process inspection and critical dimension measurement of these devices has been facilitated due to recent developments and applications which reduce,minimize and possibly eliminate sample damage and contamination. This paper will present one method for the acquisition of critical dimension measurements using video profile analysis and will discuss some of the factors that affect these profiles.

The fabrication of microelectronic components calls for efficient inspection instruments for checking the dimensional fidelity of micropatterns on masks and semiconductor wafers. The task, in particular, is to measure feature sizes from some microns down to 0,1μm and distances between line edges up to 150 mm. For solving this tasks electron microscopes are advantageous used. Recently a new precision measuring instrument has been developed by Jenoptik GmbH Jena: the ZRM 20 Electron-Beam Measuring and Inspection System. In the following a short description of the instrument will be given, in particular, of such functional subsystems, which determine the prerequisite to a high measuring accuracy and measuring speed.

The demand for smaller device geometrics and larger wafer size has drastically increased the need for more accurate level-to-level registration. It is more important than ever before to understand and control the overlay parameter of a step-and-repeat mask-making system. Furthermore, such an understanding is necessary for evaluation of mask quality. This paper discusses some important aspects of mask-making systems and gives a mathematical description of the first order and higher-order parameters. Some measurement techniques for obtaining those parameters are presented. Data are presented for special test patterns and for product masks fabricated on both optical and e-beam systems. This paper also includes a discussion of measurement strategy and a scheme for data reduction which minimizes estimating errors.

In order to verify that the optimum performance of step-and-repeat camera lenses has been achieved, the imaging characteristics of the lens must be evaluated during the manufacturing cycle using methods which are independent of the camera or recording medium. At Tropel the image quality of each lens produced is quantified and correlated to design values by measuring 1) the optical transfer function (OTF) and 2) the residual distortion errors. The definition and significance of these parameters as they relate to the photolithographic process are discussed. This paper describes the OTF and distortion measuring equipment developed at Tropel and the process by which it is integrated into our manufacturing and quality control operations. This equipment 1) provides correlation data for tolerance analysis, 2) allows us to monitor final lens correction which might be required to compensate for any adverse build-up of manufacturing tolerances and 3) provides quantitative test results used for customer acceptance. The features and operation of this test equipment are discussed in detail as well as the procedures which we have used to achieve the level of accuracy required by manufacturers of cameras for direct-step-on wafers. Measured test data for several lens types are compared to computed values as well as measurements obtained from photographic test patterns taken after installation in a camera. These test results show OTF performance to be within 5% of design values and OTF measurement accuracy to be better than 2%. The distortion measurement accuracy is within 0.05pm. The effects that partially coherent pupil illumination can have on both OTF and distortion measurements are also described.

To pattern high resolution integrated circuit substrates for multilevel fabrication processes demands exacting alignment and reregistration procedures of a direct-write electron beam exposure system. A laboratory spot-electron-beam direct-write exposure system (LEBES-D, Perkin-Elmer/ETEC) using a laser interferometer controlled stage with a stage error loop to the beam deflection subsystem has been adopted for fine line patterning applications. Single and multiple subfield overlay writing are performed with undesirable errors stemming from both operator and equipment subsystem factors. An optically read vernier metrology method is used to measure overlay errors to below 0.1 microns.

An E-beam lithography machine with a high precision X-Y stage is used to measure a grid plate with grid points at approximately known coordinates. Two interferometer beams, one parallel to the X-axis and one parallel to the Y-axis, measure stage displacements. A calibration procedure is described in which no geometric assumptions other than repeatability are made concerning the stage, e.g. it is not assumed that interferometer mirrors are spherical or parabolic, nor are specific geometric features, such as axis misalignment, modelled. Although the assumptions are very general, it is possible to observe the grid in just three orientations to determine both an inverse distortion function (calibration) and absolute rectangular coordinates for the grid points, providing that only one additional item of information is available, namely the absolute distance between a pair of points on the grid. The mathematical theory is based upon the group of symmetries generated by the three orientations of the grid and an associated lattice of rotational fixpoints. A cali-bration program has been written applying the theory to the practical problem of calibrating an E-beam lithography system. Simulations based on actual E-beam measurements confirm the theory for the practical problem. The calibration program is a powerful tool for observing and measuring the behavior of a working E-beam system. An E-beam machine can be calibrated using the program to an accuracy limited only by the repeatability of the machine, currently on the order of hundredths of a micrometer.

Overlay performance is usually determined by exposing wafers twice, processing the wafer, and then using vernier or electrical probe techniques to measure the overlay error. In this paper, we describe a very different technique where the overlay error is evaluated in situ on the lithographic exposure tool. The basis of this approach is the examination of moire fringes between the projected image of a grating test mask and a grating test wafer. This is a particularly attractive approach to lithographic metrology because the moire fringes contain information about both resolution and overlay error. A measurement instrument has been constructed which attaches to a MicralignTM projection printer. The fringe intensity is measured by a solid state detector array positioned on an image of the illuminated arc-shaped field. Since the overlay measurement can be done many times per second, we can directly observe vibrations between mask and wafer.