This paper describes the effects of grazing angle illumination on the particle sensitivity for small particles resting on bare silicon wafers and on patterned wafers. A mathematical model of particle illumination is used to support data gathered from the Inspex EX2000 Patterned Wafer Inspection System. A methodology is given for setting up the instrument in order to achieve maximum particle sensitivity on patterned wafers.
Because particulate contamination is the largest yield detractor in semiconductor fabrication, a support technology is evolving for inspecting product (patterned) wafers to identify sources of contamination. We will describe a variety of optical phenomena that are used to detect the presence of particles and some of their performance limitations. The Lloyd's mirror configuration can be used in a coincidence geometry to suppress the pattern signals; we will show performance data from a prototype of such a system. Most of the techniques we will describe rely on the particles protruding above the surrounding pattern. There are two dominant limitations to this approach. One limitation are the classes of pattern features that will be incorrectly identified as particles. The other is that semiconductor devices are becoming increasingly sensitive to particles smaller than the vertical pattern height.
Although laser scanning for contamination has become an accepted and crucial tool in the semiconductor industry, several misunderstandings still exist with many users with regard to its capabilities and limitations. These properties with respect to defect sizing, substrate influence, count accuracy and repeatability are discussed. A criterion is suggested for wafer acceptance based on the whole histogram of the contamination, rather than the present partial criteria. Projections are given for future evolution of the field.
This paper demonstrates numerical solutions of Maxwell's equations for laser diffraction and scattering from submicron objects on silicon wafers. Discrete numerical solvers using time domain, explicit finite element (FE) and finite difference (FD) methods have been implemented and vectorized for supercomputers. Both 2-D and 3-D FE and FD simulations have been performed for fibers and spheres (.5 to 1.6 microns) in free-space and on bare silicon, illuminated at normal incidence by a He-Ne laser beam. The wave solvers were verified against exact solutions for cylinders and spheres in free-space. Laboratory scattering experiments have been conducted with latex spheres on bare silicon, and the results compared to the corresponding numerical simulations. Simulations of glass and silicon fibers on a layer of photoresist over silicon examined differences in exposure patterns as well as the implementation of bleaching in nonlinear calculations. The results indicate that there is no inherent limitation to rigorously simulating many of the optical problems associated with IC fabrication, inspection, and design. The idea here is not to obtain just "exact" solutions for unsolved problems, but rather to provide the analyst with a broader capability for numerical experiments and simulations. In this way computers and mathematical algorithms can augment ad hoc approximations and laboratory experiments. The ultimate goal of this research is to simulate imaging systems and resist exposure patterns, and incorporate some of these results in process simulation codes like SAMPLE.
A surface analysis system is described using an unconventional illumination approach. The optical system utilized incorporates two orthogonally polarized He-Ne lasers operating at 633nm. It is capable of sizing PSL (latex) particles on bare silicon wafers smaller than 0.2μm and provides first order differentiation between particles and defects. The scanning mechanisms are most similar to that used in optical disk drives. The optical system employs a dark-field microscope objective enabling the analysis of particles on transparent materials, such as glass, as well as on highly reflective materials such as silicon and dielectric mirrors. In addition to semiconductor wafer inspection, the new instrument can be configured to analyze polished memory disks, optical elements, and ring laser gyro mirrors.
The performance of micron and submicron devices and chips for the coming decade is advancing at a rapid pace. The requirements to test these ultrafast, small, and dense circuits give rise to great challenges for high speed testing. Methodology to meet these challenges and yet unsolved forthcoming problems are discussed.
The increasing complexities of Integrated Circuits (I. C.) have driven the package pin counts into the hundreds and the number of internal gates into the hundres of thousands. Verification of device performance is no longer an easy task. The conventional methods of internal functional testing using a wire probe is becoming more and more default because of decreasing cell geometries, 2um or smaller. I will present a non-contact, non-destructive laser beam probing technique for extracting the functional state of internal nodes of an IC. The effect of the laser injecting energy on silicon structures and the detection of the energy is explained. A system to automate the test process is outlined.
All current commercial test systems are limited to data and clock rates of less than 200 MHz. However, many devices being built today operate at data rates greater than 1 GHz (1000 MHz). The inability to make high-speed tests stems from the many problems of measuring a device's high-speed input and output waveforms electrically. The electrical measurement methods used in the current test systems can no longer meet time accuracy and waveform fidelity requirements and are unlikely to do so in the foreseeable future. The problems associated with high-speed measurement include high pin capacitance, long device pin to receiver distances, limited receiver bandwidths, and situating numerous complex electronic assemblies within a small radius of the device. Many of these difficulties can be substantially reduced or eliminated using an electro-optic means of sensing the high-speed test waveforms. The advantages of extracting high-speed device voltage information using electro-optic techniques include non-invasiveness (low capacitance) and very high bandwidth (greater than 10 GHz). Current work in this field is reviewed showing that although electro-optic test techniques have been demonstrated to work well into the hundreds of gigahertz, the techniques are not as yet suitable for the production testing of high-speed integrated circuits.
Electron beams as a viable technique for contactless testing of electrical functions and electrical integrity of different active devices in VLSI-chips has been demonstrated over the past years. This method of testing electronic networks, most widely used in the laboratory environment, is based on an electron probe which is deflected from point to point in the network. A current of secondary electrons emitted in response to the impingement of the electron probe is converted to a signal indicating the presence of a voltage or varying potential at the different points. Voltage contrast, electron beam induced current, dual potential approach, stroboscopic techniques and other methods have been developed and are used to detect different functional failures in devices. Besides the VLSI application, the contactless testing of three dimensional conductor networks of a 10cm x 10cm x .8cm multilayer ceramic module poses a different and new application for the electron beam test technique. A dual potential electron beam test system allows to generate electron beam induced voltage contrast. The same system at a different potential is used to detect this voltage contrast over the large area without moving the substrate and thus test for the electrical integrity of the networks. Less attention in most of the applications has been paid to the electron optical environment, mostly SEM's were upgraded or converted to do the job of a "voltage contrast" machine. This by no means will satisfy all requirements and more thoughts have to be given to aspects such as: low voltage electron guns: thermal emitter, Schottky emitter, field emitter, low voltage electron optics, two lens systems, different means of detection, signal processing - storage and others. This paper will review available E-beam test techniques, specific applications and some critical components.
The early work of Brattain and Bardeen demonstrated that under illumination, a semi-conductor surface develops a potential which can be measured with non-contact capacitive coupling. Since then, the surface photo-voltage (PV) effect has received sporadic experimental as well as theoretical considerations. A number of experiments have been reported (with and without contacts) which make use of both DC and AC surface PV. In the latter form, the incident light is modulated at frequencies ranging from a few Hz to a few MHz. Both the DC and AC embodiments of the surface photo-voltage effect have been used to study various minority and majority carrier transport mechanisms in the presence of various degrees of inversion (or accumulation) of the semiconductor surface. Theoretical treatment of the PV effect has also been sporadic and lacking in completeness. Existing models for the PV effect are either too simplified to accurately represent a physical semiconductor surface, or are too general to be useful for the interpretation of experimental data. It is nevertheless clear from theoretical considerations and from existing experimental results (using incident photon energies above and below the semiconductor bandgap) that the PV effect contains valuable quantitative information which is specific to the space charge region of a semiconductor surface or junction. It is the object of this paper to demonstrate that the PV effect affords the opportunity to quantitatively measure many of the technologically important properties of a semicon-ductor surface or junction without contacting the sample and with a spatial resolution that is consistent with very large scale integration (VLSI) technology. It is also shown that non-contact PV is very sensitive to junction bias and can therefore be used to detect the on-off state of an operating transistor.
High-speed electrical signals on any semiconductor device or circuit are probed by generation of short (80 fs) pulses of photoelectrons at the point of measurement; and performing an energy analysis of the emitted electrons. A temporal resolution of 5.3 ps has been achieved on a photoconductive signal generated on a 5 μm co-planar strip line.
A turn-key system for rapidly producing application specific integrated circuit devices (ASICs) has been developed. A design program is translated from the compiler work station directly to the interconnection metallization of a pre-processed but unpersonalized wafer by a direct write laser pattern generator (DWLPG). This instrument efficiently and precisely exposes photoresist and allows interconnection metal not pertinent to the circuit operation to be removed by etching.
ASIC devices can be programmed by lasers using two techniques - a photochemical approach to direct-write connections or localized heating of the circuits by a laser spot to cut or form link structures. The technique to cut links is a mature process based on cutting similar link structures used to repair memory devices with redundant circuitry. The link-cutting process described here provides a 90% or higher programming yield on a 100,000 link ASIC based on specific process recommendations including optical verification of link cutting performance.
A laser linking technology has been developed which allows the formation of connections after fabrication in order to circumvent defects for yield enhancement in wafer-scale silicon circuits. This technique also facilitates rapid-turnaround customization. A number of digital signal processing systems have now been implemented with applications to speech, radar and image processing. In order to expedite technology transfer a new link structure has been developed which can be fabricated using a standard MOS process and has been demonstrated using the MOSIS silicon foundry.
Cutting of conductors using laser-induced chemical etching as well as deposition of conducting links by laser photolysis have been demonstrated. These operations form the basis for a laser microchemistry tool useful for repair or modification of microcircuits. With our prototype system, we have been able to photolytically deposit conducting lines with resistivities of 250 μohm-cm. Because the process is photolytic, deposition can be initiated even over transparent areas, and film growth is well controlled.
In recent years deep UV lithography systems have generated a great deal of interest because of their potential for a greater depth-of-focus for a given resolution. Recent experimental results have shown that 0.5 micron imaging is possible with a number of different deep UV designs. However, this does not guarantee that the next generation of optical lithography systems will operate in the deep UV. Adequate resolution is only one of many criteria that will determine the most practical approach to 0.5 micron lithography. The inherent advantages of deep UV systems may be compromised by a wide variety of difficult problems, while the performance of near UV systems may he extended further than is generally expected.
This paper reviews the rapidly emerging field of excimer laser lithography. Beginning with the first sub-micron exposures, developments in excimer laser projection printing on various commercial lithographic machines are reviewed. Recent results obtained with full-field scanning projection systems as well as step-and-repeat tools are summarized. Future directions in optical lithography are examined in view of these advances. Finally, a discussion of various key excimer laser parameters is presented from two points of view: availability, and requirements for various practical lithographic systems.
A deep UV projection system, capable of imaging lines and spaces of A 0.5 m width over a 14.5 mm diameter field has been developed. The system utilizes a single glass (fused silica) projection lens and a KrF excimer laser source at 248.4 nm wavelength. The fact that the lens has no chromatic correction imposes certain requirements on the laser source, such as the spectral bandwidth, wavelength stability and broad band background. The results of a study (theoretical and experimental) on how these laser characteristics affect the projection lens imaging are presented in this paper. We show the dependence of image contrast on laser bandwidth and wavelength shift, and the effect of injection locking efficiency on the size of printed lines and spaces. We also discuss the various approaches to laser line narrowing and some problems associated with a coherent light source.
Photoetching with excimer lasers has been studied in a variety of polymeric materials. Photoetching rates of polymers irradiated were measured at various laser wavelengths and fluences. The relationship of these results to the polymer absorption coefficient is examined. We propose that different mechanisms of photoetching may prevail, which depend on the absorption coefficient of the polymer. Potential use of this dry-etching process for lithography applications is evaluated.
Half-micron patterns have been fabricated using a newly developed high-speed KrF excimer laser stepper system with new resist, NOEL (Novolak based resist for Excimer Laser), and water-soluble contrast enhanced material, WSP-EX.
To investigate the way to half micron photolithography, experiments have been performed with a high numerical aperture lens, with multilayer and contrast enhancement layer resist processes, and with an excimer laser deep UV exposure system. The 0.6 N.A. lens is for the g-line and has a 5 mm by 5 mm field size. Single layer resist exposures show good profiles at 0.6 μm line/space with no effect of highly oblique illumination, and a depth of focus of 1.25 μm. Multilayer resists using spin-on-glass and contrast enhancement layers improve the resolution to 0.375 μm with the large N.A. lens. This lens, which proves the practicality of achieving better resolution through larger N.A. and improved resist, has been made available as a first generation small field half micron stepper. As a more advanced experiment, a KrF excimer laser stepper with an achromatic quartz/fluorite lens of N.A. = 0.37 shows no effect of speckle and has produced 0.35 μm L/S in PMMA which proves the usefulness of achieving higher resolution through shorter wavelength. The resolution with MP2400 photoresist is only 0.4 μm because of the high deep UV absorption, and points out the need for more work on practical deep UV resists. In addition, much more work remains on alignment, lasers, and illuminators to make possible a production excimer stepper.
Optical steppers are expected by many workers to be applicable to design rules below 0.5 micrometers. In 1986 workers at Bell Labs reported exposure of features below 0.5 micrometers using a GCA optical stepper, incorporating an excimer laser source and a refractive Tropel lens designed to operate at 248nm. In addition to a light source, lens, wafer chuck and stages, a practical wafer stepper must incorporate alignment systems, reticle and wafer handling and control systems. If the stepper is illuminated by an excimer laser, safety systems must also be provided for protection from laser light, toxic gasses, and high voltage. The system should also be configured to enhance usability and maintainablity. We will discuss the concept and performance of such a stepper system, with emphasis on the performance of lens and illuminator.
As integrated circuit feature sizes have continued to decrease, plasma etching has become the method of choice for achieving the resolution and process control required for device fabrication. To maintain tight control, the precise detection of etch endpoint is very important. A variety of endpoint detection schemes are available. These include the optical emission spectroscopy, mass spectroscopy, monitoring of chamber pressure and dc bias, and laser interferometry.
This paper deals with the application of laser inteferometry for endpoint detection in plasma or dry etching, which is the removal of a film in a plasma or low pressure gaseous discharge. Etching consists of the following steps. The exposed film, i.e. not covered by resist, is removed by chemical and/or physical processes which are determined by the type of etch. In plasma etching, the etching is done by ions and highly reactive chemical species called free radicals. There are many different types of dry etch methods. In this case, the process was RIE, reactive ion etching, where chemical and physical effects interact to produce the etch.
The reduction in VLSI feature dimensions has progressed at a rapid rate. Dynamic memory circuit complexity has doubled every year and a half over the past twenty years. A large percentage of these circuits' complexity improvement can be directly attributed to improved optical exposure tools and photoresist processing.. Optical exposure tools have extended microlithography well into the submicrometer regime by reducing the exposing wavelength and increasing the numerical aperture. Another key factor in extending resolution capabilities and proess latitude is the introduction of higher gamma, resist processes. For example, contrast enhancing materials (CEM) can effectively increase the resist gamma by a factor of two to three.
Recent advances in laser processing have reduced the capital investment required to deposit and to etch metal interconnections. Now, it is attractive to develop prototypes in the field by stocking a prefabricated CMOS circuit, and to customize it at the metallization level. The homogeneous prefabricated medium is mounted on a motorized X-Y table and moved with respect to a fixed laser beam. The pattern of metal illuminated by the focused souare beam scatters light. This scattered light conveys the form of the metal pattern in the illuminated neighborhood. Identical information is incident simultaneously on 256 different templates. Each one of these matched filters has 4x4 pixels. These 256 matched filters provide one prioritized input at 25 kHz. to a finite state controller for the X-Y table. The design of this controller will be discussed in detail. It can track polygon edges, wires which branch, and can locate open connections. Particular attention will be paid to the decomposition of the finite state machine, and to flow control within it. As each feature is recognized, its coded nature and its location are stored. The encoded table position maintains a direct registration of the laser beam to the existing metal pattern. Tracking polygons and wires generates a very compact representation for stored patterns. After further development, this recognizer will be able to support laser pattern generation in real time.