This paper describes the work conducted at IBM to study the feasibility of x-ray lithography for production of high-density silicon chips. The system approach to x-ray lighography adopted at IBM which considers the interaction of all the components is presented. In particular, the following areas are described in some detail: x-ray sources, masks, resists, exposure tools, prototype devices fabricated with x-ray lighography, and the resolution of x- ray lighography. In addition, the status of the Advanced Lithography Facility which house the compact electron storage ring x-ray source, procured by IBM from Oxford Instruments, is presented.

The synchrotron radiation (SR) soft x-ray photolithography experimental beam line (3B1 beam line) at Beijing National Synchrotron Radiation Laboratory, was completed and tested in June 1990. A soft x-ray photolithography experiment was successfully completed, and the width of linear etch on a silicon chip by the device with a 3B1 beam line is up to 0.5 micrometers . This SR soft x-ray photolithography experiment was done successfully for the first time in China. This paper describes the design of the beam line and the fabrication of the most important optical element--the cylindrical scanning mirror in the beam line. The 3B1 beam line consists of the shielding light plate with water-cooling, laser simulation light source system, 3-D adjustable scanning mirror, high pass-band filer (beryllium window), acoustic sensor, helium gas chamber, and vacuum system. The main specifications of the 3B1 beam line are as follows: spectral range 0.4-2 nm; horizontal acceptance angle 7.5 mrad; vertical acceptance angle 0.4 mrad; grazing incidence angle 1.5 deg; light spot size 35 nm X 12 nm; vacuum degree of the mirror box 5 X 10-10 torr (static). The cylindrical scanning mirror in grazing incidence is used in the beam line for photolithography to obtain uniform distributed intensity of illumination of the SR source in the vertical direction (Gaussian distribution) and sufficiently concentrated energy. It is made of aluminum alloy LD2 with a supersmooth optical surface. The curvature radius of the cylindrical surface is 527.5 mm; surface figure error is less than (lambda) /10; surface roughness is better than 1 nm RMS, and fold coating on the surface of the mirror under UHV of 109 torr. The laser simulation light source system is used for adjusting the optical system in the beam line instead of the SR source. The cylindrical mirror was polished supersmoothly using Al2O3 ultra micropower grinding material made in TOMAS in Japan on modified traditional machine tools, and surface roughness is better than 1 nm RMS.

An overview of current activities in Canada is reported, including x-ray lithography studies based on laser plasma sources and x-ray mask development. In particular, the application of laser plasma sources for x-ray lithography is discussed, taking into account the industrial requirement and the present state of laser technology. The authors describe the development of silicon carbide membranes for x-ray lithography application. SiC films were prepared using either a 100 kHz plasma-enhanced chemical vapor deposition (PECVD) system or a laser ablation technique. These membranes have a relatively large diameter (> 1 in.) and a high optical transparency (> 50%). Experimental studies on stresses in tungsten films deposited with triode sputtering are reported.

Focused ion beam induced deposition complements the process of material removal by focused ion beam milling. Together, these two techniques are being used commercially for photomask repair and for repair or restructuring of integrated circuits and are being developed for the repair of x-ray lithography masks. This microsurgery of masks and circuits can be carried out with a precision determined by the minimum diameters of the ion beams which are now approaching 0.05 micrometers . In ion-induced deposition, a local gas ambient in the millitorr range is created on the surface around the point of ion incidence, usually by aiming a miniature gas nozzle at the surface. Incident ions break up the gas molecules that are adsorbed on the surface. The precursor gas is usually an organometallic or a metal halide. Deposits of W, Au, Al, Cr, Ta, and Pt have been produced. Often these deposits have high concentrations of impurities, particularly carbon if organometallics are used, and sometimes also oxygen. The resistivities of the 'metal' films fall in the 70-1000 (mu) $OMEGAcm range that one would expect for pure metals. Nevertheless, even at these resistivities, useful conducting connections can be made for integrated circuit repair. Under special circumstances resistivities approaching the pure metal values have also been demonstrated. For x-ray mask repair, high aspect ratio (e.g., 0.25 micrometers wide by 0.4 to 0.7 micrometers high) deposits of a high-Z material such as Au, W, or Pt are needed. The considerable body of experience in this field is revealed, and the theoretical models of the process are examined.

The use of focused ion beam (FIB) circuit repair systems has drastically affected the entire design cycle of modern integrated circuits. Product analysis is changing as well with the ability of a focussed ion beam system to make very accurate cuts on a die and then image the layers using differentiation by the primary ion beam. Transmission electron microscopy is also being positively affected by this new technique due to its ability to produce thinner samples with more positive location than were available by previous techniques. In order to maximize the return, this type of system is best utilized by many operators rather than a select dedicated group. The applications and the organization of use of one such system is described in the deign debug phase of the advanced microprocessor, the Motorola MC68040, along with some suggestions for future directions of focused ion beam machines.

The vacuum lithography technique has been extended to the nanometer spatial resolution regime by incorporating a 100 keV focused ion beam system which has a beam diameter of 50 nm or less. A portable vacuum transfer chamber is used so that the sample is maintained on vacuum throughout processing without the processing chambers being directly connected. This process employs focused ion beam pattern writing on an ultrathin surface oxide layer which acts as a mask in a subsequent dry etching step. Molecular beam epitaxy is used to overgrow new epitaxial material on the patterned substrates. Examples of patterning and overgrowth applied to InP/InGaAs heterostructures are described.

X-ray mask repair is a critical element in the commercialization of x-ray lithography. The purpose of this paper is to demonstrate routine repair of conventional defects on an x-ray mask and to test the repairs by exposure to an x-ray stepper.

The current state-of-the-art in multilayer x-ray optics is presented, and the specific problems posed by projection lithography are considered. To make soft x-ray projection lithography a viable technology, the multilayer optics must demonstrate both high normal incidence reflectivity and long-term stability in the potential hostile lithography environment.

As electron storage ring (ESR) based x-ray lithography technology moves closer to becoming an industrial reality, more and more attention has been devoted to studying problem areas related to its application in the production environment. A principle component is the x-ray lithography beamline (XLBL) and its associated design requirements. XLBL, an x-ray radiation transport system, is one of the three major subunits in the ESR-based x-ray lithography system (XLS) and has a pivotal role in defining performance characteristics of the entire XLS. Its major functions are to transport the synchrotron orbital radiation (SOR) to the lithography target area with defined efficiency and to modify SOR into the spectral distribution defined by the lithography process window. These functions must be performed reliably in order to satisfy the required high production rate and ensure 0.25 micron resolution lithography conditions. In this paper the authors attempt to answer some specific questions that arise during the formulation of an XLBL system design. Three principle issues that are essential to formulating a design are (1) Radiation transport efficiency, (2) X-ray optical configurations in the beamline, (3) Beamline system configurations. Some practical solutions to thee problem areas are presented, and the effects of these parameters on lithography production rate are examined.

The fundamental aspects and implementation of the SUSS ALX alignment system for x-ray lithography are described. This paper details the algorithm, the logic behind the alignment target design, and the real-world overlay results achieved. The system reliably achieves an overlay accuracy of 150 nanometers at 3(sigma) being the norm. Focus repeatability is +/- 0.2 microns, including zero level. Alignment calculations are performed in less than 200 msecs. Focus finding is complete in 400 msec. The ALX is a three-camera video system imaging specified mask and wafer alignment patterns to perform coarse alignment of mask and wafer to the imaging field of view, fine alignment of mask to wafer, and automatic focus. The illumination is brightfield, white light. As an auxiliary function, the ALX is used to set planarity and gap of the mask and wafer, by means of fast autofocus. This allows for completely passive (non-contact) gap setting. Both alignment and autofocus on a high SNR image profiling technique for image preprocessing and data compression, here termed 'summed projection.' This vital operation is performed in hardware in true real-time (frame rate). An optical edge detection signal processing algorithm is applied to extract position information from the compressed video input signal.

An unique lithography beamline is described that delivers x-rays using a flexible arrangement of two toroidal mirrors providing high spectral and spatial uniformity at the wafer plane. The image produced is a thin, scannable, horizontal line suitable for exposing a 25 nm X 50 nm field, compatible with submicron ULSI. The authors also present the current status of such a beamline currently under construction at CXrL for use with the Silicon Valley Group Lithography, Inc.'s 0.25 micron x-ray stepper.

The techniques used in the experimental characterization of thin membranes are considered for their potential use as mask blanks for x-ray lithography. Among the parameters of interest for this evaluation are the film's stress, fracture strength, uniformity of thickness, absorption in the x-ray and visible spectral regions and the modulus and grain structure of the material. The experimental techniques used for measuring these properties are described. The accuracy and applicability of the assumptions used to derive the formulas that relate the experimental measurements to the parameters of interest are considered. Experimental results for silicon carbide and diamond films are provided. Another characteristic needed for an x-ray mask carrier is radiation stability. The number of x-ray exposures expected to be performed in the lifetime of an x-ray mask on a production line is on the order of 107. The dimensional stability requirements placed on the membranes during this period are discussed. Interferometric techniques that provide sufficient sensitivity for these stability measurements are described. A comparison is made between the different techniques that have been developed in term of the information that each technique provides, the accuracy of the various techniques, and the implementation issues that are involved with each technique.

Proximity low energy x-ray lithography using a diamond membrane, in which the wavelength is around 5 nm, is evaluated in order to avoid the difficulties of mask fabrication, inspection, and defect repairs. The resolution is estimated based on a simulation in a SR exposure system optimized considering mask contrast and absorbed power in a resist. The simulated data show that 0.2 micrometers lines and spaces of a 0.1 micrometers tungsten absorber on a 1 micrometers diamond membrane are replicated in a 1 micrometers resist at a mask-to-wafer gap of 10 micrometers.

Kinematically mounted x-ray lithography masks are investigated to optimize various design parameters. Given the limited error budget for x-ray mask mounting, it is essential to minimize the mechanical distortions in the exposure area. Three-dimensional finite element models of the support ring with a membrane are used to analyze the gravitational effects for both the horizontal (e-beam patterning) and vertical mounting (synchrotron exposure). In-plane and out-of-plane distortions of the membrane are computed and the nodes in the patterned area are uniquely mapped. Results of the finite element calculations show that the mask distortions can be minimized by optimizing the design of the support ring in conjunction with the holding mechanism. The actual cross section of the ring is designed in correlation with the specifications on the position of the mount. Several design rules are developed from the analyses, relating the axis or rotation of the cross section with the radial position of the mount. Results of this study offer guidelines in choosing the optimum mask parameters considering the parametric designs presented.

The effect of electron beam acceleration voltage and beam sharpness upon process latitudes of 0.2 micrometers line fabrication is estimated by computer simulation. Process latitude refers to dose and development time latitudes whereby the proper resist profiles are obtained. The latitudes are compared for acceleration voltages of 30 and 50 keV; beam blurs of 0.0, 0.05, and 0.1 micrometers ; resist patterns on bare Si and on Si covered with W layer; and three categorized exposed pattern with different pattern densities. in the case of the bare Si substrate, even beams at 30 keV acceleration with 0.1 micrometers beam blur give proper process latitudes. Thus, the 0.2 micrometers lines can be fabricated at 30 keV acceleration with 0.1 micrometers beam blur. On the contrary, for the resist patterns on W layer, 50 keV is necessary. Moreover, in the case of the bare Si substrate, the higher acceleration voltage made the process latitude larger at each categorized pattern. However, for reducing proximity effects between different categorized patters, a sharper beam blur is more effective than a higher acceleration voltage.

Half-micron lithography for a production environment is not considered realistic with currently available lithography tools. While optical steppers have high wafer throughputs, they do not have sufficient process latitude at half-micron geometries. In contrast, advanced technologies with sufficient capabilities for half-micron processing such as direct-write e-beam and x-ray lithography are extremely expensive and have low effective throughputs. A mix-and- match lithography approach can take advantage of the best features of both types of systems by sing an optical stepper for noncritical levels and an advanced lithography system for critical levels. In order to facilitate processing of a triple level metal half-micron CMOS technology, a mix-and-match scheme has been developed between a Hitachi HL-700 D e-beam direct write system and an Ultratech 1500 wide-field 1x stepper. The Hitachi is used to pattern an accurate zero or registration level. All critical levels are exposed on the Hitachi and aligned back to this zero level. The Ultratech is used to align all other process levels which do not have critical targets that are placed on subsequent process levels. The mix-and-match approach is discussed, and optical to e-beam as well as e-beam to optical alignment results from seven production lots are presented. The linear alignment error components X translation, Y translation, rotation and magnification are extracted and analyzed to determine their source. It was found that a simple adjustment improved the registration capabilities of these two lithography tools by reducing the X translation, Y translation and rotation standard deviations by a factor of two or more, while greatly reducing the magnification errors between the two tools.

A high-speed proximity effect correction system with two-level cell hierarchy processing has been developed to realize an accuracy-assuring electron beam (EB) direct-writing for high density VLSI. The system has two distinct advantages. First, a new hierarchial zoning algorithm is introduced to realize a data compaction for the total pattern transactions. Zone data or assemblies or patterns to be proximity-corrected are created by the zoning procedure. Frame region is associated with each zone in order to incorporate the effect of back-scattered electrons into the zone data. Second, a fast iterative technique is introduced for the proximity effect correction calculation based on a dos modulation method. A double Gaussian proximity function is used for describing the electron scattering. The present correction system was applied to 64 Mbit DRAM pattern with a 0.4 micrometers design rule. The total correction processing for the layer with maximum data volume was completed within four hours in CPU time. The patterns after delineation and development were successfully obtained by combining the present proximity effect correction with tri-layer resist process.

The multiple electron beam scattering effect is studied experimentally in an electron optical column. This effect causes a serious problem on the critical dimension of LSI pattern when the whole area of a wafer is exposed to an electron beam. This paper discusses the quantitative analysis and a method of reducing this effect.

Proximity correction schemes as used in electron beam lithography require the determination of electron scattering within the exposed resist and underlying substrate material. Scattering of an electron beam in a solid can be described by a double Gaussian function with coefficients (alpha) (forward scattering), (beta) (backward scattering), and (eta) E (ratio of energy deposition due to backscattering and forward scattering). These three coefficients are also referred to as 'proximity parameters.' This paper discussed proximity parameters mainly as a function of resist thickness. New experimental results are reported.

Sub-0.1 micrometers mushroom-shaped gates (T-gates) have been realized with a three-layer resist technique using e-beam exposure. The exposure was carried out on a Philips EPBG-3 system operating at 50 kV. The resist system and writing strategy were investigated. Test exposures on SiN-capped GaAs wafers with ohmic contacts having the same topography as active devices were carried out. Using this T-gate lithography, pseudomorphic AlGaAs/InGaAs/GaAs HEMTs were fabricated. These devices have transit frequencies of 120 GHz.

The development of both negative and positive resists with high sensitivity and high resolution capability is critical to the production of devices using x-ray exposure technology. This paper describes a class of aqueous developable acid hardened negative and positive resists which produce crosslinked images under x-ray exposure and subsequent processing steps. The resists are chemically amplified for high sensitivity. Linewidths down to 0.1 micron have been printed with a negative resist using e-beam exposure, and 0.4 micron mask-limited featurers have been printed with an x-ray dose of > 60 mJ/cm2. The feasibility of a high resolution positive acid hardened resist has been demonstrated and remains to be optimized.

In the fabrication of sub-quarter micron devices, the key components needed to realize this task must be fully integrated. The lithography capability and subsequent device processing control must be understood by the device designers and the device technology area as well. One such ingredient in this integrated operation is a high-resolution electron beam lithography sector which encompasses both resist process control and exposure tool capability. A high- throughput variable shape beam system operating at 50 keV provides both the resolution and overlay required at 0.25 micrometers . However, for 0.1 micrometers structures and below, a thermal field emission Gaussian beam system is used. This paper reports on the application of various resist systems in the fabrication of these experimental devices. Conventional resists like diazonaphthoquinone novolac based resists have been successfully applied a single-layer resist systems in the patterning of the contact and deep trench levels down to 0.20 micrometers . For negative tone imaging, an epoxy crosslinking resist based on chemical amplification has been successfully used to pattern poly gates down to 0.10 micrometers with vertical walls. This negative resist was also used as a thick single-layer resist system in implant levels where vertical resist walls are essential. Furthermore, 0.25 micrometers lines in single layer resist over 0.4 micrometers steps were resolved with no evidence on linewidth distortions.

A simulation package has been developed to address a wide variety of process latitude issues. The authors demonstrate its versatility by studying three examples: (i) process latitude degradation due to notching, (ii) line width dependence of a newly developed negative I-line chemically amplified resist on baking time, and (iii) linewidth control of an exposed x-ray 0.5 (mu) line as a function of baking time. A powerful all-purpose 3-D dissolution algorithm has been developed for this purpose. It is the only dissolution algorithm capable of handling changes to topology of the dissolution surface.

Image intensity profiles and resist profile calculations using the XMAS simulation program are presented for storage ring x-ray lithography proximity printing under several illumination conditions. The calculations indicate the existence of a wide process window for the simultaneous replication of several kinds of subquarter-micron features.

High packing density integrated circuits such as 4 MB DRAM require many interconnect layers. The resulting topography can be very challenging at each layer, particularly for patterning contacts. This paper provides practical techniques for imaging submicron contact holes on topographical substrates. Patterning is done using a 0.48 NA i-line stepper, and methods for achieving 0.5 micrometers contact holes and other features on topographical substrates are described. Results for process latitudes, depth of focus, and feature size dependencies are reported.

In order to enhance the contrast of the e-beam negative resist, Suga has combined DUV and e- beam exposures. This application has been further extended to the high sensitive three component SAL 601 resist. As the absorption phenomenon of UV light by resist layers obeys Lambert's law, the optically deposited energy density in a photolithographic process decreases through the photoresist layer from top to bottom. In contrast, for a 20 KeV imaging process of an electron-sensitive resist, the energy deposition profile corresponds to that of the particles implanted in the resist, i.e., a Gaussian profile centered at 2 to 3 micrometers in depth. Therefore, the deposited energy density throughout a 1 micrometers thick resist layer is an increasing function from the surface to the resist to the substrate. The combination of both exposure modes provides an almost constant density of deposited energy throughout the layer depth, giving rise to better profiles. This paper discussed the experimental characteristics of the DUV flood exposure which are required for the process to work correctly. As DUV flood exposure aims to treat the top part of the resist, making it more sensitive to the electron beam than the bottom part, it is of prime important to adjust the absorption coefficient of the resist, and consequently the wavelength spectrum of the light, so that light is principally absorbed near the surface and does not insolate the bottom of the resist.

As feature sizes of semiconductors grow smaller, a resist having dry etching durability and high sensitivity is required for electron beam lithography. However, the positive type electron beam resist having both high sensitivity and high dry etching durability, which suits for practical use, has not been developed yet. In order to solve this problem, a homologous series of poly(alkyl 2-cyanoacrylate) has been investigated. As a result, the new positive type electron beam resist having high sensitivity, high dry etching durability, and high thermal resistance has been developed. This new type of resist consists of poly(cyclohexyl 2- cyanoacrylate), and these features of this resist are due to the cyano and the cyclohexyl groups. The dry etching durability of this resist is 2.19 times as high as that of poly(mthyl methacrylate) (PMMA). The sensitivity is 1.7 (mu) C/cm2 at accelerating voltage of 20 kV, which is about the same as that of poly(butene-1-sulfone) (PBS). Moreover, poly(cyclohexyl 2-cyanoacrylate) has the glass transition of 152 degree(s)C, and then it is thermally stable. Using this resist in photomask fabrication by dry etching, the chrome linewidth uniformity of 0.034 micrometers 3 (sigma) can be obtained.

The paper deals with the application of DESIRE techniques on e-beam lithography. The authors propose the basic statements for designing e-beam resists in order to apply properly the silylation method. The preliminary experiments with the designed resist prove the silylation capabilities of PMMA/P(tert-Butyl)MA copolymer. A mathematical model is built which emphasizes the role of Tg in diffusion. The simulation results are presented.

The push towards faster, denser VLSI device structures and eventually to ULSI devices means ever-decreasing design rules for IC manufacturers. In order to define patterns on silicon and gallium arsenide substrates with feature sizes of 0.25 microns, lithography, metallization, and electronic materials processing techniques will be pushed beyond current limitations. Of these technologies, lithography in the sub-0.5 micron region appears to be the main obstacle yet to be overcome. As deep-UV optical systems become more expensive and the useful field sized decrease in the attempt to achieve finer resolutions, the question of whether to switch to an alternate lithographic method becomes imminent. X-ray lithography is the leading candidate. In this paper, the question of whether x-ray lithography is economically superior to optical lithography and the cost-effectiveness of x-ray lithography are addressed. Also, the question of how x-ray lithography can be performed in a production environment is considered. First shown is that more elaborate optical systems are simply not going to match x-ray proximity system in terms of resolution because of the need to use exotic lens materials or complicated and ever finer reflection systems, none of which can correct for diffraction effects, yet must be corrected for every other aberration. The economic superiority of a synchrotron-based x- ray lithography beamline is demonstrated in a production facility using a processing-cost model based on Shinji Okazaki's cost-per-bit model. Considered, as well, is the strong possibility that exists for the use of an optically based production line which would use an anode or plasma x-ray stepper to define only the smallest geometries, such as the gate level on a DRAM chip. It is shown that it is unlikely, even pushing the limits of materials and optics, that deep-UV systems will be able to define patterns below 0.35 microns in a production environment. X-ray lithography systems could define 0.20 micron patterns in a production environment with a yield that promised to be better than that of optical systems.

A new type of plasma x-ray source has been developed by applying the spherical pinch concept of imploding shock waves to produce a dense plasma hot enough to emit soft x-rays. Very intense x-rays in a few keV energy (5-10 angstroms) is radiated from a small plasma volume as microsecond pulses. SPX II is a prototype machine designed to demonstrate the engineering feasibility of the spherical pinch scheme. It is expected to generate at least 10 mJ/cm2 of usable soft x-rays for microlithography.

EXCON is a software package developed to control and define the modeling of experiments. In this case, EXCON is used to integrate various individual modeling systems, in a user- definable manner, to simulate the semiconductor manufacturing process. These numerous systems model specific aspects of the integrated circuit fabrication process. Each can be a large complex software program requiring many system resources to reliably emulate the physical processes, in many cases at the atomic level, in an analytical manner. There are many different program data formats and user interfaces within the modeling systems used. EXCON addresses the automatic insertion of configuration and process data, the conversion of data formats between modeling systems, and the sequence of model execution. EXCON also has a mechanism to re-run the sequence of models with variations in one or several configuration or data parameters, thereby creating an environment to do controlled experiments. EXCON assists in the visualization of the data in the experimental data sequence. EXCON allows for coarse-grained parallelism by connecting processes with an interprocess and inter-machine communication mechanism, thereby allowing for concurrent execution of processes on multiple machines. System performance enhancement is done with a incremental directed graph analysis technique. EXCON will, when appropriate, transparently convert file formats between modeling systems.

Lift-off process is proved to be one of the simplest and most reliable technique for fabrication of microstructures [1], including those with submicrometer sizes [2]. The control of resist image profiles becomes of increasing importance if desired image size is decreased. The easiest way to control the profile is to involve multilayer resist systems with a bottom planarizing layer more soluble by the developer than a top image recording layer. In trilayer systems intermediate layer usually of inorganic matter serves as a separator preventing the two resist layers mixing. The intermediate layer's drawback is necessity of its removal in openings before the development of the bottom layer. Bilayered resist system for submicrometer E-beam lithography based on PMMA as the top imaging layer and P(MMA-MAA)-copolymer as planarizing underlayer has been recently reported by Kuzmin et al [3]. Earlier Dolan [4] proposed offset mask technique for lift-off photoprocessing in order to obtain metal lines as narrow as 0.25 pm. To get appropriate amount of undercut in his trilayer system Dolan applied blanket exposure of the bottom layer of photoresist to provide its solubility by the developer. Image transfer has been made into top layer of the photoresist spun on an opaque separating metal film. In E-beam lithography similar approach has been applied for fabrication of nanometer scale structures [3]. The authors of [3] used two selective developers for the top and the bottom layers to control the undercut. In order to improve E- beam resist profile control we combined methodic from [3] and the Dolan's blanket exposure of the copolymer bottom layer before spinning on of the top PMMA image layer in the bilayer system.

The characteristics of sources for x-ray lithography are analyzed in terms of image formation. In particular, laser-induced plasma and synchrotron radiation sources are compared. New design considerations are presented for both types of sources, and we show that both can be good x-ray lithography sources in terms of image formation.