52

Chapter 3 Photomask Writing 3.0 Introduction :

Pattern generator (PG) a large tool set within the mask shop, which is a module within the IC design to manufacturing process. Two key factors, Lithography capability and the integrated circuit (IC) design determine the quality requirements of these PGs. Today, a variety of PGs are commercially available to meet the changing needs of the industry in a timely and cost effective manner.

3.1 Technological evolution of mask pattern generators: The patterning and use of masks in photolithography was adopted from the very

beginning of the invention of the IC and prediction of the LASER in 1958 by a new company Geophysics Corporation of America (GCA). One of the first mask writing equipment known as 'Photo repeater' was introduced in 1961 by D.W. Mann and metal foil masks were made at Siemens (2000). This machine was a shaped beam, light-optic pattern generator. Siemens first production of masks for transistors and diodes occurred in 1966. A five-stage mask technique was developed subsequently in 1965 by one of the first European institutes for microelectronics. The artwork original was cut with a scale of about 200:1. A double layer Rubylith film was used for the artwork originals, and 2-inch. photographic plates were used for the reticles and masks. Using a reduction camera the original was reduced about 20 times, to get the reticle. This reticle was then further reduced about 10 times using a step and repeat projection microscope, to get the master mask whose scale was then 1:1 to the final image. Working masks were obtained by direct photocopying of the master mask. Finally, an aligned direct exposure of the working mask onto a wafer created, layer by layer, the micro pattern of the chip array. By 1967 bipolar integrated circuits, NOR gates with eight transistors, were made with four masking levels on 25-mm Si wafers with 1.5-mm chip size and 20 µm critical dimensions (CDs). By the early 1970s D.W. Mann reticle pattern generators became available which had a high pressure Hg arc lamp to illuminate the source and a reduction system with 10X demagnification, which printed 10 flashes per second at ~1 megapixel per second, writing primitive shapes on a photosensitive mask in a step and repeat process with 1 µm resolution.

53

Pattern designs were created in AutoCAD and this file was then converted into binary format, which could be fractured into data read by the pattern generator and then transferred to the mask plates. The early optical pattern generators were however, superseded by leading edge mask pattern generators which remained useful tools for another 20 years [1-3]. M/S Planar of Belarus is one of these companies which developed LASER pattern generators 1983-84 [4].

Laser PGs first became commercially available in 1987, at the beginning of a protracted recession in the mask making industry. In early to mid-1990s, Etec introduced major architecture and platform changes for both e-beam and laser mask pattern generators, MEBES IV and the ALTA 3000. These were followed with various platform extensions throughout the 1990s. JEOL, Hitachi, and Toshiba also had active continuous engineering developments to support periodic system and product introductions.

Reticle enhancement technologies (RETs) including phase-shift mask (PSM) and optical proximity correction (OPC) began in the mid-1990s and grew into accepted manufacturing technologies. These applications required resolution and tighter process control. The feasibility of improved process latitude using 50-kV e-beam printing was documented throughout the decade [5-7].

Between 1984-1990 a new company came into existence in Germany, called Heidelberg Instruments Mikrotechnik GmbH / Heidelberg Instruments [8]. Between 1991-1995 it developed its first Direct Write LASER pattern generator and, by 2002, the Company had sold/installed 100 LASER pattern generators. In 2006 Heidelberg Instruments introduced the VPG 200/400 and the VPG 800/1100/1400 for high volume production of photomasks used in the advanced electronic packaging industry and in 2007 a table top maskless lithography system, the µPG 101 system.

Presently, there are only three manufacturers worldwide at present viz Micronic (now renamed as Micronic Mydata AB), Sweden; Heidelberg Instruments, Germany and Planar Concern, Belarus, who manufacture LASER Pattern generators for fabrication of photomasks [4,8,9].

3.2 Comparison of LASER and electron beam capabilities: Laser systems are optimally designed to function in a fixed optical configuration,

which results in a fixed exposure rate or throughput. To provide laser systems with

54

significant user adjustments in the form of beam size or spacing is technically difficult and would defeat the benefits arising from the simplicity of laser technology. On the other hand, electron beam systems are designed to exploit their adjustability. Such systems provide user control over a range of conditions of beam size and spacing and of exposure modes. The result is a variable exposure rate or throughput. Typically, laser systems provide behavior that is highly predictable and uniformly robust over a broad range of requirements. These attributes are ideal for volume commercial manufacturing.

Electron beam systems provide extended capability at the cost of lower throughput. These attributes are suited for pilot, pre-production, and developmental requirements. Laser PGs can however, deliver multiple independently modulated beams through a single optical apparatus to produce exposure parallelism. Throughput is directly proportional to the number of beams used. Similar means of multiple beam exposure has not yet been achieved in a commercial electron beam system. In the applications of economic interest, on average for similarly demanding exposures, laser systems have roughly 2 to 3 times the throughput of charged particle systems making them today the most preferred tools for fabrication of photomasks by both research institutions and commercial establishments.

In general LASER systems provide higher throughput, better reliability and lower cost of ownership as compared to e-beam systems, while e-beam systems offer better resolution. Moreover, LASER pattern generators can be used to fabricate large size photomasks for display industry and for custom applications.

3.3 Mask – Writing Principles: Prior knowledge of the writing tool is essential to optimize the quality of a mask.

Apart from differences in the actual data format, there are big differences in the exposure system used in the various mask-writing systems. Data optimized for one writing tool may work differently for another tool, apart from the actual data format. The overall mask quality is getting very critical in the total imaging process of the wafer, which is why mask quality is a subject of increasing importance. Knowledge of critical data details can favour a certain writing tool and/or writing strategy over another. Different writing techniques [10] that are used in the various mask-writing systems are discussed below:

55

3.3.1 Raster Scanning Writing Principle: A raster scanning mask writer scans the whole mask area and switches a beam

“on” and “off” to image the pixelized pattern, similar to the TV principle. The beam can be either a laser or an e-beam. Since, the pattern is converted into pixels, the way the data is divided into blocks does not affect the final pixel configuration. Therefore, the way the patterns are sliced does not affect the quality of the mask. The mask quality in terms of line-width control or CD control is purely a function of machine properties, such as the speed and reproducibility of the beam switch.

In most raster scanning tools, the mask area is written using a combination of stage travel and beam deflection, whereby a small rectangular block of data is written using the raster scanning principle. These rectangular blocks are usually referred to as scan fields or main-fields. The way these scan fields are overlaid, the actual data differs from machine to machine. In some systems, the scan field positions can be controlled, whereas in others these are hardware determined. It is possible that a pattern falls right on the interface of scan fields. In that case, this pattern is written as two individual parts separated into two different scan fields and thus separated in time. This implies that the pattern fidelity, such as line-width accuracy (CD control), is a strong function of this scan field-stitching accuracy.

This is hardly a practical approach in mask writing to improve mask quality, but it can be useful in certain cases. Looking at the writing time of a pattern using the raster scanning approach, the pixel size is by far the quantity with the biggest impact.

3.3.2 Vector Scanning Writer Principle: A vector scanning writing tool has a lot of similarities with a raster scanning tool.

The total mask-writing area is still partitioned in rectangular scanfields. But, instead of raster scanning the whole scanfield area, only the actual data blocks (rectangles and trapezoids) are exposed.

During mask writing, pixelizing still takes place in the end. Similar to the raster scanning principle, the way the data is sliced in the scanfields does not have an impact on the final pixel assignment and hence on the line-width control. Also from a scanfield stitching standpoint, the vector scanning and raster scanning techniques are equivalent. In both tool types this is a point of concern.

56

Compared with raster scanning tools, vector scanning tools do not usually show as high a writing time increase as a function of smaller pixel size. Where the writing time on a raster scanning tool is inversely proportional to the pixel size, the writing time on a vector machine is proportional to the area that has to be written and inversely proportional to the area that has to be written and inversely proportional to the pixel size. Because of the scanning nature of the vector scanning tool, the way data is sliced has little or no effect on the final result after writing and processing of the mask.

3.3.3 Variable Shape Beam Writer Principle: A variable shape machine differs from a vector scanning machine in the way the

final data elements are exposed on the mask. Vector scanning and raster scanning machines scan the element, whereas the variable shape machine is able to expose a complete data element in a single shot. These elementary blocks are usually rectangles and triangles or trapezoids. The whole data pattern is divided into these elementary blocks. For tools using this principle it does make a difference on how data is organized and sliced. For example, there is usually a direct relation between the number of elementary data blocks (rectangles and trapezoids) and the writing time. The way data is sliced affects not only the writing time but also the dimension accuracy. The size of the elementary blocks is subject to a certain nonlinearity, which can have an adverse effect on the CD uniformity. Narrow shots should be avoided when such a mask writer is used, because of these nonlinearity issues.

3.4 LASER Pattern generators : LPGs use single or multiple LASER beams and utilize raster mechanism for pattern writing on masks. In this respect they are similar to e-beam raster scan systems, but unlike e-beam systems LPGs do not need vacuum for their operations. Moreover, there is no need of proximity correction associated with scattering of electrons in EBL and for grounding the substrate to dissipate the charge. The LPG mask fabrication process benefits from advances in high resolution, high contrast photoresist originally developed for wafer production. The earliest laser pattern generators exposed the photomask with a single focused laser beam that moved over the surface of the photomask. Bell Labs and

57

what later became Micronic Systems AB made early developments in the 1970s [11-12]. The trend of ever smaller features was followed by a smaller size of the focused spot. The decreased spot size meant increase exposure time to print the same area. To counter the loss of throughput in the systems, the use of multiple beams writing different parts of the pattern in parallel was introduced. A pattern generator, very advanced for its time, was built by TRE Semiconductor Equipment Corporation writing with 16 beams in parallel using an acousto-optic deflector (AOD) to create a crosswise motion of the beams [13]. The TRE system was the first pattern generator of the raster scan type. Today the majority of laser pattern generators are of this type. Later developments of this technology have been made Texas Instruments, Heidelberg Instruments, Ateq/Etec Systems [14]. The first commercial industry system bringing the raster scan technology to the commercial mask making market was the CORE system from Etec systems, Inc.

Some modern LASER based photo mask writing platforms are based on the lithographic exposure tool, a stepper or scanner. The first production system for writing photomasks with dynamic Spatial Light Modular (SLM) imaging is the Sigma product family from Micronic LASER Systems. The Sigma uses an SLM with pivot micro-mirrors operating in diffraction mode developed by the Fraunhofer IPMS institutes [15-16].

The semiconductor industry has followed the pace outlined by Moore’s law so far. Significant part of this growth has been due to improvements in performance and increased resolution of the lithographic exposure tools (stepper or scanner), making smaller and smaller image features possible. The current trend is that image feature shrinkage outpaces the increase of resolution capabilities in the lithographic exposure tool. This has the consequence that the lithographic exposure systems must operate closer to the resolution limit, where the performance of the tool is limited by diffraction. A diffraction limited optical system does not image perfectly due to information loss. This information loss is deterministic and can to some extent be compensated for in the photomask. This puts higher demands on the photomasks when resolution enhancement techniques (RET), such as sub-resolution assist features (SRAF), phase-shift mask (PSM), and optical proximity correction (OPC), are introduced. Further, adding SRAF and OPC to pattern data can make the data seizes grow significantly, and the data path must be dimensioned accordingly,

58

increasing the cost of the pattern generator. Due to the scaling factor of 4 between photomasks and wafer, one can expect that the resolution demands on a photomask pattern generator are four times lesser than in a stepper. However, this is not true for advanced photomasks, where RET complicates the simple scaling relationship between photomasks and wafer.

The mask error enhancement factor (MEEF) or mask error factor (MEF) is a metric on how errors in the photomasks scale when imaged in an exposure tool. The MEEF is defined as the line width error at wafer level divided by the line width error at photomasks level multiplied by the exposure tool magnification. The MEEF is as high as 5 - 10 putting tight requirements on photomasks.

3.4.1 Raster scan LASER Pattern Generators:

Raster-scan pattern generators are the workhorses of the photomasks shops. The technology is mature and known to be both stable and fast. Raster scan LPGs use a continuous wave LASER beam where the stage moves at a constant speed and beam scans in a direction perpendicular to the stage motion. Scanning is provided by a rotating polygonal mirror [17] or by accousto-optical deflector (AOD). In this manner, the scanned angle is converted to a spatial displacement at the mask blank, and a stripe of pattern is written. The modulation of beam is controlled by varying the radio frequency power to accousto-optical modulator (AOM). The stripes are lined up so that the entire mask pattern is generated. All the systems today are multibeam. Earlier, multibeam LPGs were ALTA and Custom Optical Reticle Engraver (CORE) from Etec Systems and Omega from Micronic LASER Systems. Five parallel individually blanked LASER beams are used in Omega. Combining digital and analogue modulation at AOM, writing on a fine address grid is achieved. Omega uses AOD for all beams. Data path flexibility, including opportunity to write directly from GDSII file, and real time corrections are attractive features of Omega systems. Eight beams write in parallel in Etec System’s CORE LPG. The system was later redesigned to ALTA system that uses large number of beams and enhanced platform and data path. In ALTA systems [18], 32 parallel beams are generated by a beam splitter. Beams pass through an array of AOMs and are deflected by a rotating 24-facet polygonal mirror that created scanning of the beams as by a brush. Four pass and eight pass writings are used. In each

59

Fig. 3.1 : Construction of raster LPG (a) AOD (b) Polygon mirror [1]

pass, features are printed with a different polygon facet and different portion of lens field to average out systematic errors. Gray levels of exposure in each pixel, as well as an offset of writing grids in multiple passes, ensure a fine pattern address grid and edge placement. Humidity-compensated focus system keeps constant the level of humidity in the photo resist during writing to ensure better CD control. Typical time to write a photo mask is 2h. ALTA systems of ETEC and Omega systems of Micronic are the most widely used laser systems in mask making industry. The raster-scan pattern generator writes the pattern with several laser beams of an approximate Gaussian shape. The beams scan over the mask surface while being amplitude – modulated according to pattern data. The writing time for a pattern is essentially proportional to the sizes of the pattern area and not to the complexity or number of features on the pattern. Above a limit, set by the data handling capabilities, the relation between pattern areas and writing time breaks down. The data handling capabilities of the pattern generator is designed to ensure that the system operated below this limit.

Fig. 3.1 shows the Micronic Omega pattern generator. It consists of LASER, beam splitters, modulator, acousto-optic deflector (AOD) creating a crosswise scanning motion of the beams, reduction lens and X-Y stage. The beam splitter divides the light from the laser into several beams for increased capacity by writing different parts of the pattern in parallel. Each beam is individually amplitude modulated in the acousto-optic modulator (AOM). The modulation is controlled by input from the data path. The area to be exposed is divided into stripes of equal width. The width of a stripe, or a scan stripe, is typically a few hundred micrometers. During exposure the X-Y stage moves the photomask at a constant speed along the stripe. At the same time the focused spots scan in a direction

(a)

(b)

60

Fig. 3.2 : Construction of Micronics SIGMA LPG (a) Control & beam delivery (b) SLM concept [28]

perpendicular to the movement of the X-Y stage. An AOD creates the crosswise scanning motion of the spot. The spot is moved continuously in the scan direction, and discrete scans are added in the stripe direction. After reaching the end of a stripe and before the start of the next stripe, the X-Y stage moves one increment in the direction of the scan and returns to the start position in the stripe direction. In some pattern generators, a rotating polygonal mirror performs the job of AOD as shown in Fig: 3.1.

LASER source, in newer generations of LPGs, is gradually being replaced by those with a shorter wavelength to address the need for higher resolution [19-25]. Excimer lasers (pulses :1-10 kHz) developed for optical steppers are incompatible with raster pattern generators because of the use of continuous-wave lasers in raster LPG architecture. Significant improvements in resolution have been achieved by working with shorter wavelengths. For example , an argon-ion ultraviolet (UV) laser with a wavelength of 363.8 nm is used in ALTA 3500, which results in 270 nm diameter of Gaussian spot on mask blank. In ALTA 4000, an Argon ion LASER with wavelength doubling delivers a deep UV 257 nm beam, the final spot size is about 50% smaller. This is small enough to write sub quarter micron assist features and serifs on the mask.

3.4.2 Matrix exposure LPG (SLM technology) : A new approach for exposure has been developed by Micronic [19]. In its sigma systems, a spatial light modulator (SLM) is used. This is shown in Fig: 3.2. It is based on an array of micromirrors, each mirror is tilted individually modulating the beam according to pattern data; the scattered light is blocked by an aperture. A single block of mirrors is capable of modulating simultaneously about 1 million beams. This SLM works actually as a computer controlled reticle in a microstepper. The light from a DUV laser is reflected from a SLM and is focused on a mask blank through a high numerical aperture objective lens. This system uses pulsed

(a)

(b)

61

laser illumination. The SLM is reloaded with a new pattern data during delay between laser pulses. The exposures or arrays are synchronized with continuously moving stage.

The system provides potentially high throughput. It can also achieve high resolution by incorporating some of the techniques developed for optical steppers because, in principle, matrix exposure LPG is similar to optical steppers. Resolution enhancement techniques like OPC and phase shift, high-power pulses short wavelength laser sources, and resist technologies developed for wafer steppers can be directly applied to this type of mask writing system.

3.5 Components of a LASER pattern generator: All laser pattern generators are composed of the same basic components, a

laser sources an image formation system, a focus control system, and an interferometer controlled X-Y stage. All contamination- and temperature –sensitive components are enclosed in a climate chamber. The functionality of the data path, the focus control mechanism, the X-Y stage, and the climate chamber are similar for all image formation architectures.

Data Path: The data path accepts treated CAD data usually in a vector format with

hierarchic structure. This input is converted to a data stream that is tailored to control the image formation subsystem. The vector pattern data is rasterized into a gray-scale bitmap. The bitmap is often rasterized with area coverage algorithms. A pixel that is partly covered by a feature will receive a gray – scale value that is proportional to the area covered.

Address Grid: In a high throughput laser pattern generator, there is a need to place edges with

accuracy better than the pixel grid. This is possible with the use of gray –scaling and multiple exposures technique.

In a pattern generator three grids coexist; the pattern data is expressed on a virtual data grid or design grid; the pixels are located on the pixel grid; the address grid is the pixed grid extended through techniques, such as multiple exposure and gray-scaling. The address grid limits the resolution of the edge placement, and edges

62

in the data will round to the address grid-grid snapping. For example, a pattern generator with a 10 nm address grid would not be able to print a 505-nm wide line, since it would automatically be truncated to either a 500-nm or a 510-nm wide line.

Multiple Exposure Techniques: Multiple exposures, or multipass writing, decrease the sensitivity to both

stochastic and systematic deviations. Statistically, several consecutive exposures of the same pattern will average out noise in the imaging process.

It is possible to apply an offset to the placement of the pixel grid before dividing the data into pixels. If multiple exposures are done with an offset of a part of a pixel between data into pixels. If multiple exposures are done with an offset of a part of a pixel between each exposure, the address grid will be extended. Further, the impact of systematic stitching error can be decreased if the locations of the stripe borders are displaced between each exposure.

Focus control: The focus control mechanism maintains focus by keeping a fixed distance

between the photomask and the objective lens. The topology of the photomask deviates from an ideal flat surface, and the focus control mechanism must be able to compensate and keep a fixed distance to the photomask surface.

The principle used to measure the distance between lens and photomask is airflow measurement. A pod is placed close proximity to the photomask and the airflow through a small nozzle in the pod is measured. The flow is function of the distance between the pod and the photomask. To remove the correlation with ambient air pressure, a reference is used. The reference pod is placed at a fixed distance to a reference surface. Either electronically controlled piezo-electric materials or magnetic coils can be used to make the mechanical adjustment of the distance between the lens and the substrate.

Climate chamber: The pattern generator is self-contained in a climate chamber, where

environmental variables are controlled and monitored. The air in the climate chamber is flowing as closely to a laminar flow as possible at a pressure slightly higher than the ambient pressure to prevent contaminants from entering the pattern generator.

63

Fig. 3.3 : DWL 200 LPG with E-rack

Fig. 3.4 : DWL 200 LPG Stage, LASER & Optical path

Before the air is fed into the climate chamber it passes through an environmental module where it is filtered and adjusted to the right temperature.

3.6 LASER PATTERN GENERATOR (Heidelberg DWL-200): The DWL 200 LPG (Figs. 3.3 & 3.4)

system used in the Study is a high-resolution imaging system made to expose chrome, or standard chrome blanks. It is a four beam system, which can be operated in two beam or four beam mode. Achieving a resolution of 25-nanometer writable address grid, the DWL 200 can accommodate media up to 200 x 200mm, or 8“ x 8”. Design data is generated on any program using DXF, HPGL, Gerber, GDSII, or CIF files and is converted into a LIC format that can be processed by the DWL system through “XCONVERT“ software on a workstation having Linux OS [29].

The DWL 200 is an evolutionary high resolution direct write laser pattern generator designed to meet the needs of advanced lithography. Excellent placement and overlay performance are ensured by rigid temperature control and acousto-optic beam steering elements. The main system design utilizes a heavy granite for its weight and stability. These properties ensure effective vibration isolation in connection with the air buffer system and a low thermal expansion coefficient. In order to maintain an even expansion coefficient, all other stage parts are constructed from granite as well. Although this material has certain advantages granite is not completely insensitive to external influences. Temperature and humidity play an important role in the stability of the machine, therefore, its external environment must be kept extremely stable. The flow-box provides a stable environment for the DWL in terms of temperature

(±0.1oC) airflow and clean air, guaranteeing constant exposure conditions and thus eliminating variation of exposufollowing features: • Resolution: 25nm, 50nm, 125nm• Minimum feature: 0.5 µm• Laser source: 413 nm • Substrates: Glass, silicon or any other flat materials• Exposure: Photoresist

3.6.1 The optics system The optics system of DWL200 LPG comprises of following seven major elements:

LASER Unit: The unit is mounted directly on

the granite base. The power supply is mounted away from the system, as it generates heat. A 300 mW, nm Kr-ion is used for its extreme reliability and long life. In addition, the runs at a very low noise level, which is necessary when exposing high end photo masks or direct writing on the substrate. The optical path is shown in Fig: 3.5.

Intensity Modulator: The intensity modulator is a relatively low frequency acousto

which is used for intensity correction as a function of scan angle and to fine adjust the energy dose of an exposure.

According to the applied electrical signal, thleast 2 beams leaving the AOM under a different angle (diffraction). There are 2 main

Fig. 3.5 : DWL 200 LPG Optical Path

C) airflow and clean air, guaranteeing constant exposure conditions and thus eliminating variation of exposure parameters. DWL 200 LPG system has the

Resolution: 25nm, 50nm, 125nm Minimum feature: 0.5 µm Laser source: 413 nm Substrates: Glass, silicon or any other flat materials Exposure: Photoresist

3.6.1 The optics system : cs system of DWL200 LPG comprises of following seven major

The unit is mounted directly on the granite base. The power supply is mounted away from the system, as it generates heat. A 300 mW, 413

ion is used for its extreme reliability and long life. In addition, the runs at a very low noise level, which is necessary when exposing high end photo masks or direct writing on the substrate. The optical path is shown in Fig: 3.5.

The intensity modulator is a relatively low frequency acoustowhich is used for intensity correction as a function of scan angle and to fine adjust the energy dose of an exposure.

According to the applied electrical signal, the AOM can split the beam into at least 2 beams leaving the AOM under a different angle (diffraction). There are 2 main

64

Fig. 3.5 : DWL 200 LPG Optical Path

C) airflow and clean air, guaranteeing constant exposure conditions and thus re parameters. DWL 200 LPG system has the

cs system of DWL200 LPG comprises of following seven major

The intensity modulator is a relatively low frequency acousto-optic modulator, which is used for intensity correction as a function of scan angle and to fine adjust the

e AOM can split the beam into at least 2 beams leaving the AOM under a different angle (diffraction). There are 2 main

65

beams: 0 order and 1st diffraction order. The amplitude of the applied electrical signal controls directly the amount of light diffracted in the first order.

Beam Shaping Modulator: The beam shaping modulator, controlled by digital electronics for stability and

reliability, swiftly regulates the light, generating up to 30 differential beam shapes according to the design data.

Fast Scan Deflector: The acousto-optic deflector device generates a swift scan with an approximate

sweep rate of more than 12 kHz. The scan angle is imaged into the back focal plane of the write lens resulting in a telecentric scan arrangement, so that the beam is always perpendicular to the substrate. The wide beam coming from right passes the AOD. The sound wave traveling through the crystal in the AOD splits the beam into 2 main orders: 0th order and 1st order. The 0th order is not used. It leaves the crystal in some angle and is blocked by an aperture in the 4f-tube. The 1st order is scanned. That means the beam is moving from left to right, jumping back and again moving from left to right.

Write Lens System: Three write lenses (2mm, 4 mm and 10 mm) can be used for writing the

requisite features on the Mask plate. The details of the achievable feature sizes with corresponding process parameters is given in Table 3.1. The write heads are connected to an air gauge auto-focus system to correct variations in substrate thickness by employing a piezo. The piezo features a Z range of 120 microns.

Table 3.1 : Write lenses for DWL 200

Focal length

NA Spot size Depth of focus

Pixel size

Micropixel size

10 mm 0.32 1.7 µm 8 µm 500 nm 125 nm

4 mm 0.6 0.8 µm 2 µm 200 nm 50 nm

2 mm 0.67 0.7 µm 1 µm 100 nm 25 nm

66

Camera Unit:

The camera unit comprises of a micro camera, a macro camera, and a white light illumination. Such a unit can be used to inspect and measure plates and films. For chrome mask applications the cameras are employed for alignment. The cameras are connected to a video image processing system helping in metrology functions.

Interferometer: DWL uses an interferometer system to measure the XY-position of the stage,

which carries the substrate being exposed. Interferometer uses a 20 mW He-Ne LASER (λ=632 nm) for measurement and control of precision movement of the stage.

Convert Workstation: Design files in GERBER, DXF, CIF or GDSII layout have to be converted into a

group of internal machine files in LIC format (LASER Internal Code) in order to be read by the Mask Writer to enable the user to perform a fast and user friendly file translation. The conversion software offers many options like positive / negative, rotation or scaling to customize the conversion to the LIC format. In addition to changing the data format, the conversion software also splits up the original file into a set of LIC files, which are smaller and thus easier to handle than one single file. Once data is in the LIC format it can be converted real-time into the final pixel data set.

Autofocus System: To get a good exposure result, the write lens must keep constant distance from

the substrate. DWL use an auto focus based on airflow through a small gap (100 - 200 microns) between the write head and the substrate. It provides a working range of some 70 microns with accuracy better than 200 nanometers.

Robotic Handler: Robotic handler is used for automatic loading of mask plates. It picks the plate

from the specified slot of the mask/wafer holder, put it on the stage and after the

67

Fig. 3.6 : Thickness profiles of Chromium/ Chromium oxide layer

exposure is finished, it puts the exposed plate back in the same slot of mask/wafer holder. The number of masks can be sequenced in a batch. CD Measurement System:

The LASER Pattern generator (DWL 200) has a built in CD measurement system which is used to measure the line widths and edge roughness using a computer software. 3.7 Chrome blanks :

Complete photomask writing process development and its optimization was done on Chrome Blanks, using DWL200 LASER pattern generator. These chrome blanks used were low reflective (12 %) Sodalime (3”X3” and 4”X4”) chrome blanks of two grades (Print and Master grades) comprising of Sodalime as base followed by Chrome/Chrome oxide and Photoresist layers. The plates were pre baked at 90 deg C for 30 minutes. The optical density of Chrome blanks was 2.8±0.2. 3.7.1 Sodalime plate :

The specified thickness of the soldalime glass in the chrome blanks was 0.060" (1.42-1.62 mm) with parallelism < 5 µm, flatness of 5 µm/total surface for Master grade and 10 µm/total surface Print grade plates. 3.7.2 Cr/CrO thickness : The Chrome/Chrome oxide thickness specified by manufacturer was 1000±60Å. This thickness was precisely determined by Dektak measurements and was found to be in the range of 1072.2 to 1246.3 Å over complete surface of the chrome blank. Minimum layer thickness profile is shown above in Fig. 3.6. 3.7.3 Photoresist :

The specified thickness of the photoresist layer as per the manufacturers data sheet was 5300±100 Å, which was again measured using thin film measurement

68

system Mikropack Nanocalc 2000 and was found to be 5605 ± 812 Å, highly non-uniform thickness. 3.8 Pattern generation :

The fabrication of photomask is basically the transfer of a two-dimensional geometrical pattern, defined in a data base (CAD file) file to Chrome Blank. The etched image is formed on a thin chrome film which is opaque material. The photosensitive material (photoresist), is exposed to LASER beam according to a two-dimensional binary geometric pattern defined in the CAD file converted to a LIC file. The photons from the LASER beam introduce chemical changes in the photoresist making up a latent image. The latent image is then developed to create a three –dimensional relief image in the resist. Etching the chrome, partly covered by photoresist, transfers the two- dimensional image on to the chrome.

To characterize a pattern generator, it is in most of the cases sufficient to analyze the irradiance distribution - the aerial image. A further study takes into account the effects related to the latent image, such as focus change along the depth of the photoresist film, standing waves in the photoresist, and also transmittance change of photoresist with exposure.

The steepness of the photoresist wall depends on the gradient in the aerial image. Due to the resist threshold even sloping photoresist walls produce a sharp edge on the final photomask. Despite this, steep photoresist wall in the aerial image are important because they are less sensitive to process variations and decrease the statistical deviation of the chrome edges in the final image. Further, due to the resist threshold fine irradiance structures in the aerial image not located on the edges will not be present in the image in the chrome film. To place an edge at the correct position several factors must be controlled. The irradiance in the aerial image is not allowed to fluctuate randomly from position to position; the system must be dose stable. A system that is out of focus will produce an aerial image with low gradients making the system sensitive to process variations. Furthermore, distortion in the aerial image will displace the edges as well.

Therefore, keeping in mind various criticalities of the device fabrication process, it becomes necessary to optimize various writing parameters to develop the requisite

writing process which can be used to fabricate the photomask as per desired design tolerances.

3.9 Photomask Writing for process development : Heidelberg Test structure (Fig. 2.6) shown in Chapter 2 was used for development of a base process of photomask writing. The data was verified and converted to machine (LIC) as shown in Fig. 3.7 and transferred to the system controller for exposure on to the Chrome blank loaded on the LASER pattern generator. This structure comprises of test pattern and Process control monitor (PCM) features. An exposure map was prepared (Fig. 3.8) and exposure done.

3.9.1 Exposure optimization : A proper combination of exposure energy and defocus is necessary for fabricating patterns on chrome blanks. Both these parameters were optimized for BF and DF test patterns written on print grade chrome blanks, using 2 mm Write head of LASER PG in 2 Beam modes by varying exposure energy and defocus values. During exposure 30%+10% filters were used. Methodology of process development used was as

Figs. 3.7 : HIMT Test pattern

Fig. 3.8 : Exposure map of the test pattern

Fig. 3.9 : Energy-Defocus optimization for DF mask

1.5

2

2.5

3

-25 -50 -85

CD (m

ic)

DefocusE180 mW E190 mWE210 mW E220 mW

writing process which can be used to fabricate the photomask as per desired design

3.9 Photomask Writing for process development : Heidelberg Test structure

(Fig. 2.6) shown in Chapter 2 was used for development of a base process of photomask writing. The data was verified and converted to machine format (LIC) as shown in Fig. 3.7 and transferred to the system controller for exposure on to the Chrome blank loaded on the LASER pattern generator. This structure comprises of test pattern and Process control monitor (PCM) features. An

prepared (Fig. 3.8) and exposure done.

3.9.1 Exposure optimization : A proper combination of exposure energy and defocus is necessary for

fabricating patterns on chrome blanks. Both these parameters were optimized for BF and DF test patterns written on print grade chrome blanks, using 2 mm Write head of LASER PG in

m modes by varying exposure energy and defocus values. During exposure 30%+10% filters were used. Methodology of process development used was as

69

Figs. 3.7 : HIMT Test pattern

. 3.8 : Exposure map of the test pattern

Defocus optimization for DF

85 -105 -115Defocus

E190 mW E200 mWE220 mW

writing process which can be used to fabricate the photomask as per desired design

A proper combination of exposure energy and defocus is necessary for fabricating patterns on chrome blanks. Both these parameters were optimized for BF

70

0.5

1

1.5

2

-25 -50 -85 -105 -115CD

(mic)

DefocusE190 mW E200 mW E210 mWE220 mW E230 mW

Fig. 3.11 : (a) DF & (b) BF patterns

Fig. 3.10 : Energy-Defocus optimization for BF mask

followed and recommended by the manufacturer of the equipment. Post exposures, patterns on photoresist layer of the exposed masks (BF & DF) were developed by using developer solution (MicroChrome PPD 455) for a period of 60 secs. Optimum values of Energy and corresponding defocus values were determined from the CD measurement data plots shown in (Figs. 3.9 & 3.10). The exposure energy for DF and BF masks was varied from 180 mW to 220 mW and 190 mW to 220 mW, respectively. The defocus value variation was however, kept same for both DF & BF masks at -25 to -115. As can be seen from Fig.’s 3.9 & 3.10, there is no direct correlation evident between the defocus value and exposure energy for either DF or BF masks. The CD that could be achieved in either of DF & BF is limited to about 1.0 µm, which doesn’t make the developed process suitable for any useful mask fabrication. The observed limitations of the process developed were found primarily due to large non-uniformity in the photoresist and chrome thickness, determined from the optical characterization of the photoresist and chrome layers of the chrome blanks. The best CD’s achieved in the developed DF & BF test patterns ( Fig. 3.11), were about 2.0 µm for DF mask obtained at 210 mW of Exposure energy & -50 Defocus value and about 1.0 µm for BF mask obtained at 190 mW Exposure energy & -85 Defocus value, a major limitation of the developed process as the intended objective of fabrication of SAW device requiring a versatile writing process with complete process control cannot be achieved.

Another set of experiments for development of a suitable writing process was therefore, done using a different lot of Chrome Blanks (Master grade) and fine tuning the process parameters. Prior to mask writing, Chromium and photoresist

(a) (b)

71

10101040107011001130116011901220

0 4 8 12 16 20 24 28 32 36 40

Cr thi

ckness

(Å)

Location (Point)

Cr profile

Figs. 3.13 : Thickness profile of Photoresist

565057005750580058505900

0 4 8 12 16

Resist

thick

ness (Å

)

Location (point)

Resist thickness (Å)

Figs. 3.14 : Energy-Defocus optimization for DF

-0.1

0.4

0.9

1.4

-90 -60 -30 0 25

CD (m

icron)

DefocusE85 mW

Figs. 3.12 : Thickness profile of Chromium

0

0.5

1

-90 -60 -30 0 25

CD (m

icron)

DefocusE80 mW

Figs. 3.15 : Energy-Defocus optimization for BF

thicknesses were independently characterized. The measured thickness profiles of Chromium and photoresist over entire chrome blank are shown in Fig. 3.12 and Fig. 3.13.

As can be seen in these Fig's., both Chromium and photoresist profiles are uniform as compared to those of the chrome blanks used in the initial experiments. Thickness of Chromium varies from 1040 Å to 1192 Å; a variation of 152 Å , very close to the supplier's specification of 1000 ± 60 Å. Thickness of photoresist is found to vary from 5730 Å to 5820 Å , a variation of 90 Å over complete layer, indicating a highly uniform chromium layer, although thickness being higher than specified (5300 ± 100 Å) , it is not of much significance. The test pattern was exposed on new set of chrome blanks, using 2 mm Write head of LASER PG in 2 Beam mode by varying exposure energy and defocus values as done in the initial experiments. During exposure 30%+10% filters were used. The final process was optimized for various combinations of exposure energy and defocus values for both DF and BF.

72

Table 3.2 : Optimized Writing process

Fig. 3.16 : (a) DF & (b) BF patterns

Range of exposure energy used for the study was from 70 mW to 105 mW; defocus values varying from -90 to 25. The exposed patterns were developed using developer solution (MicroChrome PPD 455) for a period of 60 secs. The optimum values of Energy and corresponding defocus values, best achieved in both DF and BF masks were determined from the CD measurement plots shown in (Figs. 3.14 & 3.15). As can be seen from these plots of the developed process for these optimum values of exposure energy and defocus values best achieved CD is of 0.5µm size; a value depicting an excellent process variability. The Spot size correction (SSC) appropriate for this process was also identified and incorporated in the data conversion and exposure. Developed Images of the both DF and BF test patterns are shown in Fig. 3.16. Smallest developed CD of 0.5 µm is marked in both the images, which confirm the suitability of the developed process. The optimized writing process parameters are summarized in the table below.

3.10 Conclusion:

Complete photomask writing process using DWL 200 LASER pattern generator including the results of writing process and optimized writing process are presented. The developed process is found to be suitable for the intended objective.

Parameter DF BF

Energy (mW) 85 80 Defocus -90 -90 Filter 10%+30% 10%+30% SSC(x,y) 300,500 300,500

(a) (b)

73

3.11 References :

1. M. Gesley, Pattern generation. Photomask Fabrication Technology (Eynon, B.G. and Wu, B. (Eds)), 2005, McGraw-Hill, p. 55-175.

2. K. Levy, The history and future of mask making, Proc. SPIE, 2884, 1996, pp 2–5.

3. S. Babin. Mask Writers: An Overview, Handbook of Photomask Manufacturing Technology (S. Rizvi (Ed)), 2005, CRC Press, p. 43-57.

4. Planar Corporation: https://www.planar.by. 5. K. Hattori, R. Yoshikawa, H. Wada, H. Kusakabe, T. Yamaguchi, S. Magoshi, A.

Miyagaki, S. Yamasaki, T. Takigawa, M. Kanoh, S. Nishimura, H. Housai, and S. Hashimoto, Electron-beam direct writing system EX-8D employing character projection exposure method, J. Vac. Sci. Technol, B11(6), 1993, 2346.

6. M. Gesley, T. Mulera, C. Nurmi, J. Radley, A. Sagle, K. Standiford, Z. Tan, J. Thomas and L. Veneklasen, 0.25 um lithography using a 50 kV shaped electron-beam vector scan system, Proc. SPIE 2437, 1995, p. 168.

7. B. Cha, J. Park, Y. Choi, J. Kim, W. Han, H. Yoon and J. Sohn, Improved process control of photomask fabrication in e-beam lithography, Proc. SPIE 4186, 2001, p. 508.

8. http://www.himt.de 9. Micronic Mydata: http://www.mycronic.com/ 10. P.J.M.V. Adrichem and C.K. Kalus, Data preparation. Handbook of Photomask

Manufacturing Technology (Rizvi, S. (Ed), 2005, Taylor & Francis, pp 28-30. 11. F. L. Howland and K. M. Poole, An overview of the new mask-making system,

Bell System. Tech. J., 1970 (1997). 12. G. Westerberg, Device for generating masks for microcircuits, US patent no.

3903536, 1975, 2 Sep 1975. 13. D. B. MacDonald, M. Nagler, C. Van, Peski and T. R. Whitney, 160 Mpx/sec

laser pattern generator for mask and reticle production, Proc. SPIE 470, 1984, p. 212-220.

74

14. P.A. Warkentin, and J.A. Schoeffel, Scanning LASER technology applied to high speed reticle writing, Proc. SPIE, 633, 1986, pp 286-291.

15. H. Lanker, P. Durr, U. Dauderstaedt, W. Doleschal and J. Amelung, Design and fabrication of micromirror arrays for UV lithography, Proc. SPIE, 4561, 2001, pp 255-264.

16. P. Durr, A. Gehner, and U. Dauderstadt, Micromirror spatial light modulators, Proc MOEMS, 99, 1999, pp 6065.

17. Allen, P. C. and Warkentin, P. A. 1989. Laser pattern generation apparatus. US patent no. US 4796038 A.

18. M.J. Bohan, H.C. Hamaker and W. Montgomery, Implementation and characterization of a DUV raster scanned mask pattern generation system. Proc. SPIE, 4562, 2002, pp 16-37.

19. P. Liden, T. Vipholm, L. Kjellberg, M. Bjuggren, K. Edgren, J. Larsson, S. Haddelton and P. Askebjer, CD performance of a new high resolution LASER pattern generator, Proc SPIE Vol 3873, 1999, pp 28-35.

20. H.K. OH, Process study of a 200 nm LASER pattern generator, J. Kor. Phy. Soc., 41(6), 2002, pp 893-842.

21. U. Ljungblad, U. Dauderstadt, P. Durr, T. Sandstrom, H. Buhre, and H. Lakner, New LASER pattern generator for DUV using a spatial light modulator, Microelectronic Engg., 57-58, 2001, pp 23-29.

22. H.A., Fosshaug, A. Bajramovic, J. Karlsson, K. Xing, A. Rosendahl, A. Dhalberg, C. Bjoernberg, M. Bjuggren and T. Sandstorm, Resist process optimization for a DUV LASER pattern generator, Proc. SPIE 5256, 2003, pp 355-365.

23. C. Jackson, P. Buck, S. Cohen, V. Garg, C. Howard, R. Kiefer, J. Manfreto and J. Tsou, DUV LASER lithography for photomask fabrication, Proc. SPIE 5256, 2004.

24. R. Keifer, P. Buck, V. Garg, Hickethier, C. Jackson, J. Manfredo, C. Morgante, P. Allen and M. White, Pattern fidelity performance from next-generation DUV LASER lithography on 65-nm masks and wafers, Proc. SPIE 5992, 2005.

25. B. Olshausen, M. Chandramouli, D. Wall, B. Auches and C. Damon, Production performance of a Sigma 7300 DUV writer, Proc. SPIE 5992, 2005.

75

26. H. Sjoberg, T. Karlin, M. Rosling, T. Osrom, J. Mahlen and T. Newman, Sigma 7300: an improved DUV LASER pattern generator addressing sub 100 nm photomask accuracy and productivity requirements, Proc. SPIE 6283, 2006.

27. T. Ostrom, J. Mahlen, A. Karawajczyk, M. Rosling, P. Carlqvist, P. Askebjer, T. Karlin, J. Sallander and A. Osterberg, Improved photomask accuracy with a high productivity DUV LASER pattern generator, Proc. SPIE 6349, 2006.

28. H.G. Esser, Excimer LASERS speed microlithography mask writing, Euro Photonics, Aug/Sep, 2006.

29. Operation Manual for DWL 200 LASER Pattern Generator, Heidelberg Instruments Mikrotechnik GmbH, Germany, 2004.