Hendrik Rothe

Generic detrending of surface profiles

A commonly used estimator for the microtopography of an optical surface is its rms-roughness. Raw surface profile data may contain trending components. Therefore they should be subjected to a detrending procedure before estimating the rms value. This procedure is limited in most cases to the removal of piston, slope, and curvature. Consequently, undesired artifacts may arise, which negatively influence the precision of rms-roughness estimation. In scanning surface metrology, the eigenvalues and eigenvectors of the covariance matrix of the surface can be used for a robust and precise multivariate estimation of rms-roughness.

Discrimination of surface properties using BRDF-variance estimators as feature variables

By means of image processing it is often difficult to detect minor defects of smooth surfaces such as glass and polished metal or ceramics. It was found that certain statistical moments of the bidirectional reflectance density function (BRDF) contain consistent information regarding different kinds of defects. Using Multivariate Discriminant Analysis with BRDF moments as feature variables the defect recognition and classification can be carried out in a straightforward way. Because the classification algorithm is very simple it is suitable for real- time application.

Investigations of smooth surfaces by measuring the BRDF with a stray light sensor in comparison with PSD curves evaluated from topography of large AFM scans

For quality inspection of polished surfaces as applied in semiconductor and optical industry, various methods are used for a fast detection of microroughness, defects, and contaminations. With the aid of stray light sensors the intensity distribution of the reflected and scattered light, i.e. the BRDF, is measured. The probability distribution of values of a BRDF is parametrized to obtain a measure for roughness and for classes of defects. There is still need for justifying the choice of statistical moments to characterize and finally to classify different surfaces. Of course, a basic quantitative, i.e. metrological understanding of stray light sensors is necessary. The power spectrum of surface topographies sufficiently smooth to obey Rayleigh-Rice approximation is proportional to the BRDF. Therefore a comparison was only carried out with sample surfaces obeying this approximation. Defects and contaminations with lateral sizes smaller than the wavelength of the illuminating light employed in the stray light sensor, however, could not be analyzed within this investigation. We have measured the topography of large areas up to 600 micrometer X 100 micrometer with an AFM by patching several scans (up to 8) with overlap. BRDFs evaluated from AFM measurements agree well with BRDFs measured with a stray light sensor.

Real-time detection of surface damage by direct assessment of the BRDF

An approach is presented which allows the development of an in-process fiber optic stray light sensor and adapted real-time signal processing procedures for quality control of polished optical glass surfaces. The power of this new method is illustrated by presenting an instructive example in great detail.

Characterization of surfaces using neural pattern recognition methods based on BRDF and AFM measurements

A major problem of in-situ surface characterization by using angle-resolved light scattering is the fast and accurate surface parameter identification. This paper will deal with surface parameter identification methods from BRDF measurements of rough surfaces with stochastical height topographies. First, neural classification methods will be discussed. Second, the discussed classification method will be applied to BRDF data taken by an ARS silicon sensor with 8013 polar photodiodes. The classification results will be compared to topography data taken from AFM measurements. Finally, neural self-organized networks will be applied to classify in unsupervised manner rough surfaces based on BRDF measurements.

Evaluation of in-situ ARS sensors for characterizing smooth and rough surfaces

A major problem of in-situ surface characterization by using angle resolved light scattering (ARS) is the contradiction of speed and accurate detection of the scatter signal. During the last years several fast and compact ARS sensors have been developed, namely a fiber optic stray light sensor (FOSSIL) with 516 fibers in 3 azimuths, an integrated optics stray light sensor with 120 waveguides in one azimuth and a planar silicon scatter sensor (PSS) with 8013 detector elements. This paper deals with the evaluation of these sensors and their employment to characterize smooth and rough surfaces. After introducing the sensor setups the theoretical performance of the various sensors is compared. This is done by modeling the properties of the radiation detectors and the arrangements of the sampling points. The real performance is obtained by applying the sensors to surfaces with a known BRDF and comparing the measurements with the expected scatter distributions, i.e. from Lambertian reflection standards and preliminary scatter standards of NIST. Furthermore a set of smooth surfaces (polished silicon and steel) was scanned by an atomic force microscope (AFM) and the computed surface statistics is compared to the values obtained from scatter measurements. Finally, the sensors ability to characterize rough surfaces is shown by using pattern recognition methods.

Modeling of an ARS sensor system in spatial and time domain

Scatterometry is a powerful and fast measurement method to measure surfaces and its properties. The backscattered light from a coherently illuminated surface contains information about integral surface topography constants, material properties, surface defects and contamination. In this paper a generic system model of an angle resolved light scatter sensor (ARS sensor) will be discussed and exemplified by using the sensor system LARISSA (Large Dynamic Range Intelligent Scatter Sensor Approach). The system model consists of two parts - firstly in spatial domain and secondly in time (or frequency) domain. The part of the system model in spatial domain contains the characteristics of the optical map of the scattered light onto a detector system. The optical map will be discussed by using an elliptical mirror optics with regard to aberration effects. The part of the system model in time (or frequency) domain contains the characteristics of the conversion of scattered light into quantized signals. The basic steps of the conversion process will be considered. Furthermore, the characteristics in time domain of a single CMOS detector (photo diode) with logarithmical intensity characteristics will be discussed to estimate the opto electronic bandwidth limitation and the minimal exposure time for different applications. Based on the system model basic performance limits will be summarized and further design steps of the sensor system LARISSA will be outlined. This paper is a continuation of a previous paper.

Performance limits of ARS sensors based on CMOS photodiode array

Measuring the light scatter back from a coherently illuminated surface is a powerful and fast measurement method to observe surfaces and its properties. It opens the possibility to measure integral surface topography constants, material properties, surface defects and contamination. In this paper the performance limits of the angle resolved light scatter sensor (ARS sensor) LARISSA (Large Dynamic Range Intelligent Scatter Sensor Approach) will be discussed and exemplified by using a technical application. In the first part of the paper the experimental setup of the ARS sensor LARISSA will be considered. In the second part of the paper the performance limitation of the ARS sensor LARISSA concerning particle detection will be derived based on simulation and measurement results. Finally a short overview is given about further development of the ARS sensor.

Design problems of a calibrated BRDF sensor with respect to dynamic range, speed, and large angle of view

This paper deals with applications of angle resolved light scatter (ARS) measurements as well as with the discussion of design and application problems of ARS sensors. The first section gives a description of the experimental sensor setup. In the second section of the paper two applications will be outlined, firstly particle detection on smooth Si surfaces, and secondly defect detection in small Si v-grooves. In the third section of the paper principal drawbacks of the experimental ARS sensor and their elimination will be discussed.

Fast and accurate roughness characterization techniques for wafers and hard disks

Especially for wafers, hard disks and flat panel displays fast and accurate technical means for roughness characterization are needed. However, speed and accuracy are contradictory. Generally speaking, fast roughness sensors are not accurate, and precise instruments are slow. It turned out in the last years that with multi aperture fiber optic sensors which acquire ARS/TIS data a very fast estimation of surface roughness is possible. But it is rather difficult to convince e.g. chip manufacturers that the results of such sensors are reliable, because there are no accepted international standards for these kinds of optical measurements. Therefore we decided to establish a setup of our ARS/TIS sensor for roughness characterization and an instrument for roughness measurement in a cleanroom consisting of the following parts: (1) 200 X 200 mm stages, speed 0.4 ms-1, +/- 1 micron accuracy, acceleration 1 g; (2) visual inspection head consisting of 50 X objective and CCD camera; (3) AFM scan head; (4) ARS/TIS fiber optic sensor; and (5) laminar box. Topics of the paper are measurement philosophy, specs of the setup, architecture of the fiber optic ARS/TIS head, as well as data processing algorithms and software.