JP4495898B2 - Improved pattern generator - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to pattern printing with extremely high accuracy on photosensitive surfaces such as photomasks for semiconductor devices and display devices. Furthermore, the present invention relates to a direct writing operation of a semiconductor device pattern, a display panel, an integrated optical device, and an electronic wiring structure. Further, it can be used for other types of high-precision printing such as security printing (securities printing). The term “printing” is to be understood in a broad sense and means not only the exposure of a photoresist or a photographic emulsion, but also a dry process, by ablation or chemical treatment activated by light or heat. It also means light action on other photosensitive media such as treated paper. Light is not limited to average visible light, but includes a wide range of wavelengths from infrared (IR) to extreme ultraviolet. Of particular importance is the ultraviolet range from 370 nm (UV) through deep ultraviolet (DUV), vacuum ultraviolet (VUV), and extreme ultraviolet (EUV) to wavelengths of a few nanometers. EUV is defined herein as the range from 100 nm to the lower limit where the radiation can be treated as light. The normal wavelength of EUV is 13 nm. IR is defined as 780 nm to about 20 μm.
[0002]
In another sense, the present invention relates to a spatial light modulation device, a technology of a projection display device and a projection printer using such a modulation device, and a scientific research method. In particular, the present invention improves gray scale characteristics, image stability due to focus and image uniformity, and data processing for such modulators utilizing analog modulation techniques. The most important use of analog modulation is an address grid (eg, to locate the edge of a pattern, ie much finer increments than the grid produced by the pixels of a spatial light modulator) The image is generated on a material having high contrast such as a photoresist having
[0003]
[Prior art]
In the latest technology, it is well known to construct a high-precision pattern generator using the projection of a micromirror spatial light modulator (SLM) of the micromirror type (Nelson 1988, Cook 1990). The use of an SLM in a pattern generator has many advantages over the more widely known method that uses a scanning laser spot. That is, the SLM is a large-scale parallel processing device, and the number of pixels that can be written per second is extremely large. This optical system is more simplified in that the illumination of the SLM is non-limitative, whereas in a laser scanner, the entire beam path must be constructed with high accuracy. Compared to several types of scanners (especially electro-optic and acousto-optic scanners), the micromirror SLM can be used at shorter wavelengths because the micromirror SLM is a fully reflective device.
[0004]
In the above two cited documents, the spatial light modulator uses only on / off modulation for each pixel. The input data is converted to a pixel map having a 1-bit depth, for example, a value of 0 and 1 for each pixel. This conversion can be done effectively using a graphics processor (ie, a processing unit) or a custom logic control circuit having area fill instructions.
[0005]
In previous applications by Sandstron (Sandstron et al., 1990) of the same inventor, an intermediate exposure index can be used at the boundaries of pattern elements to fine-tune the edge positions of the elements produced by a laser scanner. Is described.
[0006]
A grayscale image, preferably for projection display and printing of video images, by turning on the pixels with the SLM at varying times, or by turning the pixels on various times by printing the same pixel several times Is also well known in the art. The present invention provides a system for a direct grayscale generator with a spatial light modulator, particularly for the purpose of generating ultra-high precision patterns. The important features of the preferred embodiment are pixel-by-pixel image uniformity and independence with respect to the precise placement of features on the SLM's pixels (whether intentional or inadvertent) during focus changes. It is stability.
[0007]
[Problems to be solved by the invention]
Accordingly, an object of the present invention is to provide an improved pattern generator for printing a high-precision pattern.
[0008]
[Means for Solving the Problems]
This object is achieved by the device set forth in the appended claims which performs analog modulation of the pixels of the SLM.
[0009]
DETAILED DESCRIPTION OF THE INVENTION
The present invention can be understood based on the general arrangement of FIG. 1 showing a generic projection printer having an SLM. Spatial light modulation devices based on reflection are classified into two types: deflection type (Nelson) and phase type (Cook). While the difference between the two seems to be small in a particular example with micromirrors, the phase SLM emits the beam in the direction of reflection due to destructive interference, whereas in the deflection SLM the pixel Causes the specular beam to be geometrically deflected to one side and deviate from the aperture of the imaging lens as shown in FIG. In the case of ultra-high precision printing performed according to the latest invention, the phase modulation system disclosed by Cook in 1990 is superior to the deflection type. First, good contrast is obtained because all parts of the surface (including hinges and struts) are involved in destructive interference and overall light absorption can be achieved. Second, systems that function by deflecting light to one side are difficult to achieve symmetry with respect to the optical axis at intermediate declination points, and features become unstable when the focus changes Danger arises. In the preferred embodiment, a phase type is used, but could be used if it is housed or designed to surround an asymmetric deflection type. This situation is shown schematically in FIGS. In the first FIG. 4 a, the non-deflecting micromirror 401 is illuminated and the reflected light is not directed in the direction of the opening 402, and therefore no light reaches the substrate 403. On the other hand, in FIG. 4b, the mirror is completely deflected and the totally reflected light is directed towards the opening. In the middle position, only part of the reflected light reaches the substrate shown in FIG. 4c. However, in this case, the light does not become symmetric with respect to the optical axis of the lens 404 and enters the substrate obliquely. This makes the distance between the lens and the substrate region extremely important, and a slight change as shown by the broken line position in the region will cause the features on the substrate to be significantly displaced. A method for solving this problem is illustrated by FIGS. Here, the first exposure is performed at the first deflection angle of the micromirror, and then the second exposure is performed against the second deflection angle, preferably with the same amount of light, to supplement the first angle. . Thereby, the combination of the first exposure and the second exposure is symmetric with respect to the optical axis of the lens. As another method for solving this problem, there is a method of using a deformable mirror 401 'as shown in FIG. This last figure can generally show two examples of a phase SLM or a deflection SLM (discussed below), in which case light reflects from different parts of the mirror.
[0010]
Phase-type SLMs can be formed by micromachined mirrors, so-called micromirrors, or continuous mirror surfaces on a support that can be deformed using electrical signals. In the invention disclosed by Cook in 1990, a viscoelastic layer controlled by an electrostatic field is used, especially for very short wavelengths that are sufficiently deformed by a few nanometers, It is equally possible to use a piezoelectric solid disk which is deformed by a reflective surface which is controlled electrically, electromagnetically or thermally. The rest of this document assumes a micromirror matrix (1D or 2D) that can be controlled by static electricity, but as described above, depending on the LCD crystal material or electro-optic material as the modulation mechanism, Other configurations are possible, such as a reflective SLM or a micromachined SLM that utilizes piezoelectric or electrostrictive motion.
[0011]
In the present invention, it is preferable to use a micromirror characterized in that the amount of light reaching the pupil of the light projection lens becomes variable due to the variable phase modulation. 2a-h show several mirrors consisting of several multielements, the tilt of the various parts of each mirror is not critical. In fact, one element directs light in the direction of the lens while another element directs light outside the pupil of the lens. As a method for correctly understanding this function, there is a method in which the complex amplitude reaching the center of the pupil of the lens from each minimum region element of the mirror is examined, and the amplitude of the entire mirror is integrated. The mirror can be properly shaped and deformed so that the total complex amplitude is close to zero, so that no light reaches the lens pupil. This is an off state of the micromirror, and a relaxed state where the mirror surface is flat and the complex amplitude includes a phase is the on state. Between the on state and the off state, the amount of light in the reflection direction is a continuous but non-linear function with respect to deformation.
[0012]
The pattern to be written is usually a binary pattern such as a photomask pattern written in chrome on a glass substrate. In this case, binary means that there is no intermediate region, and a certain point on the photomask surface is black (covered with chrome) or white (no chrome). This pattern is exposed with photoresist by a projection image from the SLM, and the photoresist is developed. Modern resists have a strong contrast, which means that when a slight change in exposure occurs, there is a difference between when the resist is completely removed with developer and when it is almost removed. I mean. Therefore, even if the aerial image gradually transitions from white to black, the photoresist usually has an edge almost perpendicular to the support surface. When chrome etching is performed, the contrast is further enhanced and the resulting image is completely binary and can be divided into either black or white with no intermediate areas.
[0013]
The input data is in digital form and represents the geometrical dimensions of the pattern written on the surface. Input data is often given in minimal address units, eg, 1 nanometer, but if the SLM pixels are set to either on or off, a much coarser pattern is produced. When an SLM pixel is projected onto a 0.1 μm pixel of an image, one line can only have an integer number of pixels wide (n * 0.1 μm, where n is an integer). Until recently, a 0.1 μm address grid was sufficient, but with the advent of so-called optical approximate correction (OPC), a 1-5 nanometer grid is desired. In OPC, when the mask is used, the dimensions of the mask features are slightly modified to compensate for the expected optical image error. As an example, when a 0.8 μm wide mask with four parallel lines is printed on the latest 4X reduction stepper (projection type printer for semiconductor wafers), it is usually 0.187, Printed as 0.200, 0.200 and 0.187 μm wide lines. This can be predicted by simulation of image formation and the mask user compensates the mask using OPC. Therefore, instead of 0.800 μm, we want the first and last lines of the mask to be 4 * 0.213 = 0.852 μm. With an address grid of 0.1 μm, correction is impossible, but with an address grid of 5 nm or thinner, correction is possible.
[0014]
In FIG. 5, a method for providing SLM data is shown in a flowchart. In S1 of the first step, pattern data of a pattern to be written in an individual pattern field is divided. This pattern data is preferably received in digital form. Thereafter, in step S2, each field is rasterized, so that a different exposure index is assigned. Next, in step S3, these values are corrected so as to obtain a non-linear response, and in step S4, pixel-by-pixel deformation is performed. Finally, the pixel value is converted into a drive signal and transmitted to the SLM.
[0015]
The present invention preferably generates a fine-grained address grid (eg, 1/15, 1/25, 1/50 of the pixel size) using an intermediate value between the off and on states. . The printed feature is composed of pixels in the on state, but pixels set to an intermediate value are formed along the edges. This is done by driving the pixel with a voltage other than the on and off voltages. Since there are some cascaded nonlinear effects (exposing pixels on edge positions and boundaries, exposure and deformation, deformation and electric field), nonlinear conversion from input data to electric field is necessary. In addition, this transformation is calibrated empirically at equal time intervals.
[0016]
FIG. 3 shows an arrangement of pixels that generate a phase difference by moving up and down like a piston. This figure shows how the pixel is controlled and the reflectance is generated by this inset (insertion plate). The bright area has pixels with phase 0, while the dark area is generated by pixels with alternating phases of +90 degrees and -90 degrees. The oblique boundary line between the bright area and the dark area is generated by the phase of the intermediate value. This shows how the edges are finely positioned by the phased SLM. However, other types of SLMs with intermediate values can be used as well. The imaging characteristics of the intermediate phase SLM are complex and far from the clarity of edge movement in FIG. However, calculations and experiments based on an enormous theory by the inventor have proved that the described effects are in fact.
[0017]
In order to generate a finer address grid, as shown in FIG. 3, one kind of pixel map in the pattern feature, another kind of pixel map outside the feature, and the boundary Rework the electronic processing system to produce an intermediate pixel map on the line. However, the intermediate pixel map on the boundary line is generated according to the arrangement of the boundary line in the grid that is finer than the pixels of the SLM projected on the processed product. The SLM and projection system generate one exposure level inside the feature, another exposure level between features, and an intermediate exposure level on the boundary. The intermediate exposure level is generated by the function of the SLM and is modulated into a plurality of states. After the characterization from the response by the drive signal to the actual placement of the boundary is made, the correction is applied. Measurements are made empirically and are stored in the data processing and transport system after the calibration function is calculated by the computer.
[0018]
To further improve address resolution, the stage and SLM can be stitched together so that the exposure field is stitched together in a direction that is not parallel to the coordinate system of the SLM (usually 45 degrees). Make it again. In particular, the continuous movement of the stage or optical system occurs in a direction that is not parallel to the SLM (typically 45 degrees from the SLM's coordinate system). It is also possible to have an SLM with a non-perpendicular axis, in which case it is advantageous not to have any axis parallel to the direction of motion. Further, in order to reduce line errors caused by imperfections in the SLM's column and row drivers, the line error of the matrix itself, the row and column lines are at a constant angle with respect to the stitching direction. It is more effective to use (for example, a vector between the centers of connected fields).
Also, a more precise address grid is created by superimposing at least two exposures with correction data so that the combined exposure produces an intermediate value that cannot be obtained with a single exposure.
[0019]
(Design of phase type SLM)
The clover mirror shown in FIG. 2c as used in the prior art can be in an intermediate state between the on state and the off state. However, if the integral complex amplitude is plotted as a deflection function, it will never be completely zero and draws multiple circles around zero, thus having a non-zero minimum reflectivity with varying phase angle Is understood. This state is schematically illustrated by line 701 in FIG. However, 703 shows the position with respect to a fixed deformation value, and (psi) shall show the related phase angle. A thorough analysis of an image with several pixels set to an intermediate state reveals that the edge position of the final image is not stable until it reaches the focus if the integral phase angle of the edge pixel is not zero. it is obvious. This is due to a diffraction effect similar to the reflection effect shown in FIGS. The preferred embodiment of the present invention uses a new type of pixel with a swivel element. Examples of such elements are shown in FIGS. As each element pivots, one end moves in the direction of the light source and the other end faces in another direction, thus maintaining an average phase close to zero. This state is schematically illustrated by the dashed line 702 in FIG. Furthermore, this clover design has the problem that residual internal stresses occur during manufacture. This stress tends to cause partial deformation without using an applied electric field. This internal deformation is due to imperfections during manufacturing and does not always occur in all pixels. In a clover design, this pixel-to-pixel difference causes a primary deviation in reflectivity. Similar results are obtained with pixel cells formed by swiveling elements, but moreover secondary effects occur. Therefore, the uniformity in the projected image is improved.
[0020]
The modulation element design and the exposure method are reworked so that symmetry is obtained in the aperture stop of the projection system with respect to edges of different positions and / or different orientations of the pattern. The inherent asymmetry between edges located at different positions relative to the pixel grid can be mitigated by superimposing at least two images located at different positions of the pixel grid relative to the pattern.
[0021]
In the case of the deflection type SLM, the symmetry is related to the luminance distribution of the aperture stop. Most preferably, it has a modulation element that deflects light symmetrically with respect to the center of the aperture stop, or overlaying exposures supplemented with deflection is used to obtain symmetry. A variable deflection modulation element can create a certain geometric relationship between the deflection at the edge pixel and the edge. For example, the pixels can be oriented in a direction perpendicular to the edge and in the internal direction of the feature.
[0022]
Symmetry is obtained by using a diffractive SLM to superimpose exposures with opposite phase maps. If the complex amplitude is a real number in any part of the SLM, the symmetry can be maintained, and the pixel can be designed with an integral complex amplitude that is always a real number within a range of −1 to 1. In many cases, an amplitude in the range of -0.5 to 1 is sufficient. The above is an example using the square turning micromirror elements of FIGS. 2e, 2f, 2g and 2h.
[0023]
The resolution can be increased by printing the background area using a small negative amplitude. In a more complex manner, each group of adjacent pixels can be combined in the image and the desired actual amplitude can be achieved after being filtered by the imaging system.
[0024]
In order to maintain symmetry, it is advantageous to have at least a 2-fold symmetry, preferably a 4-fold symmetry. Symmetry can be achieved for pixels that do not have inherent rotational symmetry due to multiple superimposed exposures. Furthermore, the pixel design or exposure can be used to increase the resolution by a sequence that provides a controlled actual amplitude. Black lines are particularly strong when placed between regions with opposite phases, and the edge of a feature will cause adjacent pixels in the feature to have a higher positive amplitude or external adjacent pixels. It can be improved by using a negative amplitude.
[0025]
(Higher image quality)
The swivel design has a third advantage. The clover shape does not lead to complete absorption, but the swivel cell can more easily obtain an omnidirectional geometry that achieves complete absorption, or it can pass through zero with a non-zero fraction. Reflection may return, and in this case, the phase is reversed. By performing light absorption well, the degree of freedom to print overlapping exposures is increased, and by designing a low negative value 702, better linearity close to light absorption can be realized. Printing out of phase with a weak exposure of about 5% in the dark area gives a high edge sharpness of 15-30% and can use smaller lenses to print smaller features. This is very similar to the so-called attenuating phase-shifting mask used in the semiconductor industry. As a related method for increasing the edge sharpness, there is a method in which a low value is set for pixels inside the feature and a high value is set for pixels near the edge. Thereby, it is possible to realize a new type of image quality improvement that is impossible by using the pattern projection from the current mask and the projection method by Nelson and Cook. Using a non-zero negative amplitude for the background and increasing the exposure along the edge is consistent with generating a fine address grid with the edge pixels as intermediate values. This is due to each effect being additive or at least calculable. In addition, there are combinations of pixel values where all effects can be obtained simultaneously if the pixel is substantially smaller than the feature to be printed. In order to detect these pixels, not only a fine address grid is generated, but also computational computation is required, but in some applications of the invention, printing of smaller features is not possible. If possible, a high value can be obtained for a great deal of effort.
[0026]
For a continuous mirror on the viscoelastic layer, there is an intrinsic equilibrium of the average phase with respect to zero. Simulations show that a continuous mirror is formed by applying an intermediate value to the fine positioning of feature edges. However, the nonlinearity is lower than when a micromirror is used. Without a fully functioning method, the smallest features must be larger than when using micromirrors, for example, fewer addressed pixels per resolved feature element. must not. As a result, the SLM device becomes large and the amount of data for a certain pattern increases. Therefore, the micromirror is selected in the first and second embodiments.
[0027]
In the present invention, rotational symmetry deformation (at least twice as much symmetry, suitable for giving two symmetrical reasons to the pupil of the projection lens and that the image does not react to rotation. In this embodiment, a pixel with a fourfold symmetry) is used. The latter reason is important in printing a random logic pattern on a semiconductor wafer. If there is an asymmetry with respect to the x-y-axis, each transistor placed along the x-axis will have a different delay than that placed along the y-axis. Such circuits can malfunction or can only be used at slower clock speeds. These two requirements for image invariance due to focus and symmetry between the x and y axes make it very important to generate and maintain symmetry in the optical system. Symmetry may be provided by nature or may be generated by deliberately balancing asymmetry, such as using multiple exposures with complementary asymmetries. However, since a plurality of exposures lead to a decrease in throughput, it is extremely preferable to have a symmetrical layout from the beginning.
[0028]
【Example】
(Preferred embodiment)
The first preferred embodiment is a deep UV pattern generator for photomasks using SLMs consisting of 2048 × 512 micromirrors. The light source is a KrF excimer laser with a pulse output of 248 nanometers, a pulse length of about 10 ns, and a repetition rate of 500 Hz. The SLM also has an aluminum surface that reflects more than 90% of the light. The SLM is irradiated by a laser through a beam-scrambling illumimator, and the reflected light is directed toward the light projecting lens and further directed to the photosensitive surface. The incident beam from the illuminator and the existing beam to the lens are separated by a translucent beam splitter mirror. Preferably, the mirror is polarization selective, the illuminator uses polarized light, and its polarization direction is switched by a quarter wave plate in front of the SLM. In order to be symmetrical with respect to the x and y axes with a high numerical aperture (NA), the image must be symmetrically polarized and a second 1 / b between the beam splitter and the projection lens. An image in which a four-wave plate is circularly polarized is generated. A simpler configuration where possible with the energy of the laser pulse can be achieved by using an unpolarized beam splitter. Even after the second pass of the beam splitter, the quarter wave plate still has advantages, but it affects the design of the beam-splitting coating. This is because it becomes difficult. The simplest configuration as a whole utilizes the oblique incidence in the SLM, where the beam from the illuminator and the beam reaching the projector lens are geometrically separated as shown in FIG.
[0029]
The micromirror pixel is 20 × 20 μm, and the light projection lens has a 200 × reduction magnification, thereby making the SLM pixel correspond to 0.1 μm of the image. The lens is a monochrome DUV lens having a numerical aperture (NA) of 0.8, and the point spread function is 0.17 μm FWHM (full width at half maximum). The minimum line that can be written with good quality is 0.25 μm.
[0030]
A workpiece, such as a photomask, is moved by an interferometer control stage under the lens and an interferometer logic signal to the laser that produces the flash. Since the flash is only 10 ns, the stage is not moved during the exposure and the SLM image is printed to a size of 204.8 × 51.2 μm. After 2 milliseconds, the stage moves 51.2 μm, a new flash is emitted, and a new image of the SLM is printed from the edge to the edge of the first image. Between exposures, the data input system loads a new image into the SLM so that a larger pattern is formed by the stitched flash. Once a column is completely written, the stage proceeds vertically and a new row begins. Any size pattern can be written, but in the first preferred embodiment, a 125 x 125 mm pattern is typically written. Writing a pattern of this dimension takes 50 minutes plus travel time between successive rows.
[0031]
Since each pixel can be controlled to 25 levels (+ zero), a 0.1 μm pixel is interpolated and incremented by 25, each forming 4 nanometers. Data conversion specifies the pattern anatomically and converts the data into a map using pixels set to on, off, or intermediate values. The data path must provide the SLM with 2048 * 512 * 500 words of data per second, in fact, 524 megabytes of pixel data per second. In the preferred embodiment, the writable area is up to 230 × 230 mm, and up to 230 / 0.0512 = 4500 flashes in a row, which is written in 450/500 = 9 seconds. The amount of pixel data required for one column is 9 × 524 = 4800 Mb. A compressed format is used to reduce the amount of data transferred and stored in the buffer. This format is very similar to the invention by Sandstron et al. In 1990, except that the pixel map is compressed instead of segments with a certain length and value. A viable alternative is to generate the pixel map directly and use a commercially available hardware processor for compression and decompression to reduce the amount of data transferred and buffered.
[0032]
However, the data amount of the entire mask is quite large to store the data divided in advance on the disk even if it is compressed, and pixel data must be generated at the time of use. . An array of processors rasterizes the image in parallel with the conversion to a compressed format and transfers this compressed data to an expander circuit that supplies pixel data to the SLM. In the preferred embodiment, the processor also rasterizes different parts of the image, buffers the results and sends them to the input buffer of the expansion circuit.
[0033]
(Second preferred embodiment)
In a second preferred embodiment, the laser is an ArF excimer laser having a wavelength of 193 nm and a frequency of 500 Hz pulses. The SLM has 2072 × 1024 pixels of 20 * 20 μm, and the lens has a 333 × reduction magnification that gives a projected pixel of 0.06 μm. There are 60 intermediate values and the address grid is 1 nanometer. The point spread function is 0.13 μm and the minimum line is 0.2 μm. The data flow is 1572 megabytes / second, and a column of 230 mm long data is 11.8 Gb.
[0034]
A third preferred embodiment is that the pixel matrix is rotated 45 degrees and the pixel grid is 84 μm so that the projected pixels are spaced 0.06 μm along the x and y axes. Is the same as in the second preferred embodiment. The laser is an ArF excimer laser and the lens is 240 scaled magnification. Since the matrix is rotated, the pixel density of the matrix is reduced and the amount of data is half that of the second preferred embodiment, but the address resolution is the same.
[0035]
(Laser flash vs. flash deviation)
Excimer lasers have two disadvantageous characteristics: flash-to-flash 5% energy deviation and flash-to-flash 100 ns temporal variation. In the preferred embodiment, both are compensated by the same method. The initial exposure is formed by the entire pattern at a magnification of 90%. The actual flash energy and the time position of each flash are recorded. The second exposure is formed by a nominal 10% exposure, and analog modulation causes the second exposure to be 5-15%, depending on the actual value of the first exposure. Similarly, the time variation of the first exposure can be corrected by intentionally canceling the time in the second exposure. The second exposure can fully compensate for the error caused by the first exposure, but itself produces a new error of the same type. Since the average of the entire exposure is only 10%, the error between them is effectively reduced by 10. In fact, the laser has a temporal uncertainty much greater than 100 ns. This uncertainty is due to the fact that the pulse of light occurs according to the delay from the trigger pulse, and this delay varies from time to time by a few microseconds. Since the delay becomes more stable over a short period of time, measure the delay continuously, preferably using the filtered final delay value to predict the next pulse delay and position the trigger pulse .
[0036]
Similarly, if a stage error is recorded and the stage is driven by a compensation operation in the second exposure, it is possible to correct the stage imperfection. Any measurable placement error can, in principle, be corrected partially or completely as described above. It is necessary to have a high-speed servo that moves the stage to the point calculated during the second exposure. In the prior art, a method is known in which the SLM itself is mounted on a stage with a small stroke and a short response time and used for precise positioning of an image. Another equally effective method is to use a mirror with piezoelectric control in the optical system between the SLM and the image plane, and which one to select is determined in consideration of the actual situation . Furthermore, it is also possible to add an offset position to the exposure field data and move the image horizontally.
[0037]
The second exposure is preferably performed using an attenuation filter between the laser and the SLM to allow full use of the SLM dynamic range within 0-15% of the nominal exposure. With the 25 intermediate levels, the exposure can be adjusted in steps of 15% * 1/25 = 0.6%.
[0038]
Responses vary slightly from pixel to pixel due to manufacturing imperfections, and such differences may occur due to changes over time. As a result, there arises a disadvantage that the image becomes inhomogeneous. Because the requirements for the image are very high, all pixels may have to be corrected by an increase due to the inverse response of the pixels stored in the lookup memory. It is more preferable to use a polynomial having 2, 3 or more terms for each pixel. This can be done with hardware based logic that drives the SLM.
[0039]
In a more complex and preferred embodiment, the second corrected exposure is performed by combining several corrections. This includes flash-to-flash deviation, flash temporal variation, and well-known pixel-to-pixel response differences. As long as the correction is small, that is, a few percent for each correction, it is added almost linearly, so the correction is added as it is and applied to the SLM. The sum is multiplied by the desired dose value at the relevant pixel.
[0040]
(Alternative light source)
Excimer lasers have a limited pulse repetition frequency (prf) of 500-1000 Hz depending on the laser wavelength and type. Therefore, large fields with stitched edges are used on both the x-axis and the y-axis. In two other preferred embodiments, the SLM is irradiated from a pulsed laser with a much higher prf, such as a Q-switched upconverting solid state laser or a continuous laser source scanned on the surface of the SLM, One part of the SLM is rewritten with new data while another part is printed. In either case, the coherence characteristics of the laser are different from the excimer laser, and larger beam scrambling and coherence control is required, such as multiple parallel optical paths with different optical path lengths. In some means of the present invention, the light output from the flash lamp is sufficient and can be used as a light source. The advantages are low cost and excellent coherence characteristics.
[0041]
In the preferred embodiment in which irradiation is performed by scanning, two problems can be solved. One is the problem of pulse-to-pulse deviation in time and energy, which is preferably due to the scanning being performed under full control by use of an electro-optic scanner such as acousto-optic or electro-optic. This is because using a large number of continuous lasers causes less fluctuation in power than when using a pulsed laser. In addition, the use of a continuous laser allows the selection of different wavelengths, and the continuous laser is less dangerous to the eyes than a pulsed laser. However, the most important point is that the scan is non-limitative and can be performed with a repetition rate of 100 kHz or more, so a much higher data rate can be reached with a matrix of only a few rows. Scanning the illumination beam is also a way to produce a very uniform illumination and is difficult in other ways.
In some embodiments, a flash lamp can be used as the light source, which is a convenient method.
[0042]
(EUV)
The EUV light source is due to particle heating, radiation from a magnetic plasma pinch machine, or heating that is somewhat important to extreme temperatures with high power laser pulses. In either case, the radiation is pulsed. EUV radiation propagates only in vacuum and can only be focused with a reflective optical machine. Typical pattern generators that use SLM have a small exposure field, which is a not so high requirement for optical power. Therefore, since the design of the optical system is gentler than that of the EUV stepper, more mirrors can be used, and a higher numerical aperture (NA) than that of the stepper can be realized. A lens with a high numerical aperture (NA) is expected to have a ring shaped exposure field, and it is well possible to make the SLM shape to match such a field. With a wavelength of 13 nm and a numerical aperture (NA) of 0.25, it is possible to expose a line with a width of only 25 nm. Further, as described above, if high image quality is used, it may be below 20 nm. Is possible. There is no other well-known writing technique that can meet such a resolution, and at the same time, no other technique can achieve the writing speed possible with the same characteristics of the SLM.
[0043]
(Edge overlap)
Stitching is extremely important because a two-dimensional field is printed for each flash and the edges of each field are joined together. Replacing a field of only a few nanometers can cause visible pattern errors along the edges, which can adversely affect the function of the electronic circuit generated by the mask. An effective way to reduce the effects of such unnecessary stitching is to print the same pattern on several paths and replace stitching boundaries between such paths. If the pattern is printed four times, stitching errors are expected to occur at four locations, but only a quarter of the scale. In the preferred embodiment of the present invention, the ability to generate an intermediate exposure is used, along with an overlap band between fields. While the above values are calculated by the computer during rasterization, this calculation can be performed even while decompressing the compressed data. Edge overlap reduces stitching errors and significantly reduces throughput negatives compared to multi-pass printing.
[0044]
(Correction irradiation)
In the first preferred embodiment, the irradiation of the SLM is performed by an optical scrambler such as an excimer laser or a fly-eye lens, and irradiation similar to that from the circular self-emitting surface of the pupil plane of the illuminator Generated. In the case of increasing the resolution during printing by a specific projection system, a correction irradiation method can be used. In the simplest example, there is a method of introducing a pupil filter having, for example, a quadrupole or annular transmission region on the pupil plane of the illuminator. In a more complex example, there is a method of printing the same field several times. Several parameters can be varied between exposures, for example, the focus of the image plane, the illumination pattern, the data applied to the SLM, the pupil filter of the pupil plane of the projection lens, etc. In particular, the resolution can be enhanced by synchronized changes in illumination and the pupil filter, which means that the pupil has a fan-shaped transmission region and that non-diffracted light blocks the absorption patch near the tip of the sector. This is particularly noticeable when the irradiation is aligned in a straight line.
[0045]
(Response linearization)
In order to linearize a transfer function from data to an edge, there are basically the following three methods for edge arrangement performed in this specification.
Taking into account non-linearities in the data converter, generate 8-bit (example) pixel values in the data converter and drive the SLM using a DAC with the same resolution. This situation is shown schematically in FIG. 8a. Here, R is a relay signal, and C is a capacitor provided in each matrix element of the SLM. The SLM is indicated by a broken line.
-Generate a digital value with fewer values (eg 5 bits, ie up to 32 values), convert it to an 8-bit value in a lookup table (LUT) and then feed this 8-bit value to the DAC .
Select a DC voltage generated by one or several high-resolution DACs using a 5-bit value and a semiconductor switch. This situation is shown schematically in FIG. 8b.
[0046]
In either case, when an empirical calibration function is used in the LUT of the data converter or when used with a DC voltage, the empirical calibration function is measured such that the response on the plate is linearized. It is possible.
[0047]
Which linearization scheme to use depends on the data rate, accuracy requirements, and available circuit technologies that can change over time. At present, the data converter is in a deadlock, so linearization by the data converter is not a preferred solution and it is not preferable to generate 8-bit pixel values. Further, the high-speed DAC is expensive and consumes high power. The most appropriate solution is to generate a DC voltage and use a switch. In this method, it is possible to use a resolution higher than 8 bits.
[0048]
(Description of a suitable pattern generator)
FIG. 6 will be described. The pattern generator is a fine positioning system comprising an SLM 601 with single and multiple pixel addressing scheme, a light source 602, an illumination beam scrambling device 603, an imaging optical system 604, and an interferometer position control system 606. It comprises a substrate stage 605 and SLM hardware and software / data processing system 607. Furthermore, in order to provide appropriate functions and simplify operation, the pattern generator includes an ambient chamber with temperature control, a substrate load system, stage movement to achieve optimal pattern placement accuracy, and Software for timing the exposure laser trigger and a software user interface are also provided.
[0049]
Pattern generator irradiation is performed by a KrF excimer laser, which has a band corresponding to the natural linewidth of the excimer laser and emits a long flash light of 10 to 20 nanoseconds in the UV region at a wavelength of 248 nanometers. To do. In order to prevent distortion of the pattern on the substrate, the light from the excimer laser is uniformly assigned to the SLM surface, and the coherence length of the light is made sufficiently short, so that no laser speckle is generated on the substrate. Like that. A beam scrambler is used to achieve these two objectives. The beam scrambler splits the beam from the excimer laser into several beam paths having different optical path lengths, and then combines the beam paths into one to shorten the spatial coherence length. Furthermore, the beam scrambler has a beam homogenizer consisting of a lens system having a set of fly-eye lenses, which beam light from each point of the laser beam from the excimer laser, The light is distributed evenly over the entire SLM surface, and the light is distributed to the “top layer”. The beam scrambling, light uniformity, and reduced coherence described above are advantageous in any SLM printer. Depending on the actual environment, it can be realized using a beam splitter, optical multiplexer, diffractive element, optical fiber, kaleidoscope, lenslet array, prism or prism array, or integrating sphere, as well as others. It is also possible to generate a large number of mutually incoherent optical fields incident on the SLM by combining and combining similar devices in order to split or combine the beams.
[0050]
Light from the SLM is relayed and imaged on the substrate on the substrate stage. This is done using the Schlieren optics disclosed by Cook. Focal width f ₁ Lens ₁ Is the distance f from the SLM ₁ It is arranged at the position. Focal length f ₂ Another lens of ₂ Is a distance 2 × f from the SLM ₁ + F ₂ It is arranged at the position. Next, the substrate is 2 × f away from the SLM. ₁ +2 x f ₂ It is arranged at the position. Distance 2xf from SLM ₁ , There is an opening 608 whose size determines the numerical aperture (NA) of the system and thus the size of the smallest pattern feature that can be written on the substrate. In addition, in order to correct imperfections in the flatness of the optical system and the substrate, the lens l ₂ There is also a focusing system that dynamically positions the lens in the z-direction, and a 50 micron position span provides optimum focus characteristics. Further, this lens system is wavelength-corrected so that the wavelength of the irradiated light is 248 nanometers, and the bandwidth tolerance of the irradiated light is at least ± 1 nanometer. Lens l ₁ Irradiation light is reflected into the imaging optical system by a beam splitter 609 positioned directly above. In the case of a reduction ratio of 250 and a numerical aperture (NA) of 0.62, a high quality pattern can be obtained by exposing a pattern feature whose size is reduced to 0.2 micrometers. At 32 levels from each SLM pixel, the minimum grid dimension is 2 nanometers.
[0051]
The pattern generator has a finely positioned substrate stage with an interferometer position control system and consists of a movable air bearing xy table 605 made of Zerodur for minimum thermal expansion. A servo system including an interferometer position feedback measurement system 606 controls stage positioning in each direction. In one direction y-axis, the servo system keeps the stage in a fixed position, and in the other direction x-axis, the stage moves at a continuous speed. The interferometer position measurement system is used in the x-axis direction to trigger the exposure laser and make the position between each image of the SLM on the substrate uniform. When the entire row of the SLM image is exposed on the substrate, the stage returns to its original position in the x-axis direction and moves in the y-axis direction by one increment of the SLM image so that another row of SLM on the substrate. Expose the image. This procedure is repeated until the entire substrate is exposed.
[0052]
The SLM image overlaps many pixels in both the x-axis and y-axis directions, and the exposure data pattern is modified locally with overlapping pixels, resulting in many such overlapping regions. The exposure is compensated.
[0053]
The pulse-to-pulse intensity deviation from the excimer laser is compensated by the use of a two-pass exposure of the pattern, where it is performed with a nominal 90% intensity where the first pass is the correct intensity. In the first pass, the actual intensity of each laser flash is measured and stored. In the second pass, the correct intensity for each SLM image exposure is used based on the measured intensity value from the first pass. In this way, the scale of the influence due to the deviation of the pulse-to-pulse intensity from the excimer laser can be suppressed.
[0054]
(correction)
The functionality of the SLM is described in more detail elsewhere in this document. The SLM has 2048 × 256 pixels with a pixel size of 16 micrometers and can address all pixels within 1 millisecond. The SLM is firmly attached to an elaborate stage. This sophisticated stage can move 100 microns in the x and y directions with an accuracy of more than 100 nanometers between flash exposures. By performing fine positioning of the SLM, inaccuracy of the position of the substrate positioning stage is corrected and pattern stitching errors are further reduced. In addition to positioning in the xy direction, the SLM stage can also be rotated to expose the pattern on the substrate at an angle other than that specified in the substrate stage coordinate system. The purpose of such rotation is to allow the substrate adjustability to be incorporated for a substrate having an existing pattern to which supplemental features are added. Using an off-axis optical channel and / or a CCD camera, after loading, the exact position of the substrate on the stage is measured and viewed through the lens so that multiple alignment marks on the substrate are Coordinates in the system can be determined. During exposure, the stage position is corrected in the x and y directions based on the measurement position of the alignment mark. While using a stage servo system that follows the rotating coordinate system, the SLM's sophisticated stage can be rotated as described above to perform rotational adjustment. By being able to rotate the SLM, it is possible to write to a distorted coordinate system and compensate for subsequent pattern distortion.
[0055]
Further, the means for rotating the coordinate system of the pattern to be written may include means for rotating the digital description supplied to the electronic data processing and delivery system. This converts the data into a rotated digital description before sending it to the electronic data processing and delivery system. Alternatively, the modulator signal provided by the electronic data processing and delivery system can be rotated. In this case, the modulator signals are converted after rasterization before they reach the modulator. The conversion is preferably performed in real time.
[0056]
The same method for rotating the coordinate system can be used for other types of pattern writer systems, such as raster scan laser beams as well. In particular, it can be used in systems that use acousto-optic deflectors, such as the system described in EP 046776 (Sandstron et al.). This is incorporated herein. However, in such a scanning writing apparatus, it is preferable to rotate it by rotating the deflector. An example of a mechanism for rotating the image of the modulator and adjusting the azimuth angle to compensate for the rotation angle between the first layer and the second layer relative thereto is shown in FIG. The deflector is attached to the azimuth mechanism 1006 in this example. The azimuth mechanism is an actuator 1007, such as a stepping motor, and a connecting device 1008, such as a lever arm coupled to the deflector to control the orientation of the deflector and thereby control this azimuth angle. including. A similar mechanism may be used for the orientation of the modulator. The stage then operates as previously discussed.
[0057]
However, other means for rotating the coordinate system of the pattern to be written are also feasible. For example, the means for rotating the coordinate system of the pattern to be written can include an optical element for rotating the image projected from the modulator to the workpiece. For this, a Dove prism can be used. However, it is also possible to use an analog device configuration of a reflecting surface as shown in FIG. This can impose a controlled rotation on the image projection from the modulator before it reaches the workpiece.
[0058]
The above rotation is particularly useful when the second pattern is to be written on a workpiece that already contains the first write pattern. In this case, the workpiece may need to be replaced in the same writing device, or the second pattern may even be written by a writing device other than that used to write the first pattern. In this case, it is most important to reduce the rotational deviation between patterns.
[0059]
The above correction means can be used for the production of masks and reticles, but can also be applied to direct writing devices.
[0060]
(Alignment)
In order to control the rotation of the second pattern in an appropriate manner, it is preferable to detect the rotation deviation between the first existing pattern and the second pattern to be written by the detection system. However, in other applications, it is also useful to appropriately control the rotational deviation. Such a detection system preferably includes means for detecting the position of at least one alignment mark on the workpiece, and further relates this measured position to the coordinate system of the writing device.
[0061]
With reference to FIG. 9, the alignment mark preferably includes at least two lines 901 in different predetermined non-parallel directions. In the preferred embodiment, the alignment marks include snake marks 902, i.e. several lines joined in a zigzag form. For example, the snake mark can be composed of six interconnected lines that are at least slightly acute between adjacent lines. The advantage of such a snake mark is that it can be used to determine both the rotation and position of the mark.
[0062]
The alignment mark preferably also includes a search mark 903, which allows for quicker and more efficient positioning of the alignment mark in an overview search. The search mark can be composed of a line 904 that is longer and thicker than the snake mark. The search mark preferably includes two non-parallel lines 904 arranged relatively close to each other.
The total length of the snake mark and the search mark may be about 500 to 600 nm.
[0063]
Several alignment marks can be placed at different locations on the workpiece. For example, four alignment marks can be distributed over the periphery, such as the corners of the workpiece or, preferably, the sides between these corners.
[0064]
The mark detection system includes an off-axis light channel for detecting and measuring the position of the alignment mark on the workpiece, and can further relate this measured position and rotation to the coordinate system of the writing device. . However, in the preferred embodiment of the present invention, the detection and alignment system at least partially uses the same beam path as the writing beam, and preferably also uses the same lens used for the writing operation.
[0065]
With reference to FIG. 10, the detection and alignment system includes a laser that emits light of a wavelength that can be transmitted through a light source 1001, preferably a resist layer or the like disposed over an alignment mark on a workpiece. Furthermore, the emitted light preferably has a wavelength that does not affect the photosensitive layer of the workpiece. In the preferred embodiment, a wavelength of 532 nm is used.
[0066]
The reflected light is then examined for detection and alignment. It is preferable to examine the dark field signal of the reflected light because it increases the visibility of details. The specularly reflected light is then filtered out and only the scattered light is focused on a detector, such as a PIN detector. However, other alternative solutions are possible to achieve contrast enhancement, such as any one of the types used in light microscopy. For example, you may use one of the following contrast enhancement methods:
-Bright field to detect reflected light;
A dark field that reduces the light intensity but achieves improved edge discrimination;
Confocal imaging to achieve good lateral and depth resolution;
-Phase contrast imaging to achieve good depth resolution;
・ Nomarski; and
-Lateral interference.
[0067]
Contrast enhancement improves the signal to noise ratio. It is particularly useful in direct write applications where improved accuracy and precision are required and when using chemical mechanical polishing (CMP).
[0068]
A mechanical shutter 1002 is further placed in front of the laser 1001 to block the beam when not in use. The shutter may be vacuum-operated and is controlled by a vacuum control unit 1003.
[0069]
The laser beam is preferably focused on a narrow spot of about 1 μm diameter on the substrate using the same focusing lens 1004 as the writing beam. This can be arranged by adding the beam splitter cube 1005 to the beam path for the writing beam in front of the final lens 1004.
[0070]
In order to calculate the displacement and rotation of the first pattern, the exact positioning of the alignment mark must be determined. To detect the position of the alignment mark, the substrate is driven by a mechanical positioning system as described above to scan the alignment beam on the substrate in a continuous back and forth motion. Values from the alignment detector 1009, such as from an x- and y-interferometer, from a stage position detector, and such as one or more PIN detectors are thereby collected in the step during the scan. This point-by-point measurement forms a matrix from which these alignment marks can be located. However, in the preferred embodiment only changes are stored, i.e. only positive and negative edges or edges. However, other types of detectors, such as charge coupled device (CCD) detectors, can be implemented as well. In order to detect the position of the alignment mark, it is preferable to perform a first scan of the overview with a CCD or the like, and then perform a more detailed measurement using a PIN detector. It is particularly preferred to use a detector with a detection area divider, such as a detection area divided into four detection fields. This makes it possible to detect the edge more easily.
[0071]
These marks are preferably detected accurately so as not to affect or expose the resist. For example, light with a wavelength of 248 nm typically exposes the resist, whereas light with a green light wavelength (eg, 532 nm) has no photochemical action, ie does not expose the resist. However, it turns out that it is difficult to optimize the final lens for 532 nm, which may result in poor performance. A way to solve this problem is to use only a small part of the available exposure field, preferably less than 20%. Within such a small field, for example, a wavefront error at 532 nm can be corrected by a special optical element. Furthermore, problems associated with field aberration distortion, such as small field distortion, astigmatism and focal plane curvature, are also reduced.
[0072]
Such field reduction can be achieved with a system as shown in FIGS. Both systems can be incorporated into the system described above, in particular the system described with reference to FIG. In a system such as that shown schematically in FIG. 13, the final lens assembly 1004 is optimized for writing light, for example at a wavelength of 248 nm. Further, the laser 1001 emits light having a wavelength of, for example, 532 nm. The emitted light is focused on the modulation device 1301. The modulation device may be an acousto-optic modulation device, but may be another type of modulation device such as a mechanical shutter 1002 shown in FIG. Thereafter, this light is projected onto the substrate via the final lens assembly 1004 and reflected back onto the detector 1009. This detector is preferably synchronized with the modulator. A correction optical element 1302 is preferably disposed between the detector 1009 and the final lens assembly 1004.
[0073]
In the alternative embodiment shown schematically in FIG. 14, the configuration is substantially the same as the system described above. However, in this case, a small angle scanner 1401 is used instead of the modulator. This small angle scanner scans the beam within a small area of the reduced field. The scanner may be an acousto-optic xy deflector, such as NEOS, Florida, USA, or a small galvanometer mirror that can tilt in two directions (x and y directions).
[0074]
When these mark locations are detected, they should be accurately defined. Next, a more complete scan is performed on the detected mark. The values from the x and y interferometers and from the PIN detector are then collected in more detail. This result is then used to determine the exact rotation and position of this mark.
[0075]
The alignment method according to the invention is schematically illustrated in FIG. The method begins at 1101 and first determines at 1102 whether the alignment parameters are properly defined. If not defined, parameters are read from the file at 1103. These parameters indicate what to search and how to perform this search. The item to be searched can be defined by a parameter specifying the alignment mark. The alignment mark may be of the form discussed above. The alignment marks are preferably grouped according to geometric shapes, such as circles, lines or more complex combinations. Each such group is preferably pre-defined in code form in a parameter file and comprises a set of parameters indicating the dimensions of the mark and / or group of marks. The way to do the search is: the placement of the mask in the system, the tolerance for translation as well as the rotation, the type of mask used, the maximum number of scan lines for the mask, the signal format at the edge, polarity and dynamic effects 1 There may be more than one specification. These parameters preferably specify whether this signal should be searched for negative or positive edges, how long these edges should be, and the minimum amplitude to be detected. This makes it possible to cancel the difference for exactly the same edge, making this measurement independent of the exact edge position on the signal edge.
[0076]
Most parameters may be set automatically by a computer program or may be adjusted manually. The parameter file may also include measured signal statistics.
[0077]
Subsequently, it is determined whether or not the statistical value of the measurement signal is properly defined. This statistic can include information about noise, background level, maximum and minimum levels, signal edge length, and the like. If sufficient statistical data is not available, such information may be read from a file or measured. When measuring, the line on the mark is scanned at 1105. Thereafter, at 1106, a statistical value for the line is calculated. Next, it is determined in step 1107 whether this statistical data is sufficient. If not, steps 1105 to 1107 are repeated.
[0078]
Thereafter, in 1109, the mark is searched by scanning the substrate with a line. After each scan, the signal is processed at 1110 to identify the edge at the scan line. After each scan, 1111 determines whether enough data has been collected and repeats steps 1109-1111 until enough data is collected. Further, the next scan line to be performed can thereby be determined depending on the result from the previous scan line. This keeps the required number of scan lines low and makes this search quick and effective. In scanning line processing, it is preferable to search for edges only where there are likely edges. This speeds up the search process and reduces the problems associated with perturbation noise between edges.
[0079]
When the appropriate number of edges has been identified and calculated, at 1112 the position of the alignment mark is calculated. Different search and certification methods may be used depending on the form of the alignment mark. For example, for an alignment mark that includes two intersecting lines, linear interpolation of the edges may be performed, thereby calculating the intersection. At 1113, the output from the search process is the position of this mark. It is then determined at 1114 whether these parameter and statistics files should be terminated, and if so, the parameters and statistics specified at 1115 are saved to the file. At 1116, the alignment method is finished.
[0080]
The measured positions of these alignment marks can be used to determine the difference in rotation, translation, magnification difference and / or orthogonality of the first already written layer relative to the second layer. Differences in magnification and orthogonality are particularly important when writing different layers using different writing devices. Magnification differences may also occur when writing conditions for writing different layers are different in different ways, such as temperature differences.
[0081]
In addition, it may be necessary to measure the offset between the light beam and write beam used during this alignment. This positioning difference usually must be compensated during writing of the different layers. Furthermore, the misalignment is likely to vary over time and therefore should preferably be measured periodically.
[0082]
This misalignment can be determined in a manner similar to alignment. Thus, the alignment mark discussed above or another beam measurement mark can be used. Such a measurement mark may include four lines that intersect at a common intersection, as shown in FIG. These beams are preferably evenly distributed with an angle between each adjacent line pair of 45 °. Since the position of the search mark is known in advance, the measurement mark does not need to include the search mark. Therefore, this measurement can be performed in a limited area of the raster, that is, at a fixed position on the stage.
[0083]
A special measurement light beam may be used for measuring the displacement. Instead, writing light may be used. For measurement, a detector 1009 ′, such as a PIN detector, can be used (see FIG. 10).
[0084]
To make the measurement, the stage, i.e. the substrate table, can be arranged so that a micro sweep starts from this mark at a distance corresponding to approximately the width of the line in this mark. Next, stage movement in the first direction (X) is started, and preferably the beam is scanned in the second direction (Y) at the same time. Edges are identified with reflected light, as in the alignment method discussed above. Every line of passing marks produces a pair of edges. The position in the X direction can be given by the x positioning means of the stage, while the position in the Y direction can be determined by the pixel size times the start of the sweep and the number of pulses between these edges. The position of the line can then be determined by calculating the average value between the line edges.
[0085]
After sweeping the entire mark, many positions of these lines gather. This position data can then be used for line interpolation calculations to determine a common intersection. This determines the position of the light with the mark as a reference.
[0086]
The methods and means for alignment and positioning discussed above can be used for mask and reticle production, but can also be applied to direct writing systems.
[0087]
In the pattern rasterizer 610, any data pattern having any format is converted to a compressed rasterized pixel map of 32 (5 bits) gray levels per pixel. In response to the voltage applied to the pixel electrode, the gray scale stage of the exposed pixel is not linear, so each of the 32 gray levels corresponds to a uniform increase in the next level of irradiation dose. As described above, the input data is linearized by the pixel linearizer 611. This operation is performed using an 8-bit digital-to-analog converter (DAC) 612, and from the DAC by each gray level from the pixel map, according to a pre-empirically calibrated linearization function. Select the voltage. In selecting an analog level from the DAC, each value corresponds to an SLM pixel, and each such value provides additional correction using a look-up table that corrects the irregularity of the corresponding pixel. It is. The lookup table calibration values are generated by an empirical calibration procedure, in which a continuous test pattern is sent to the SLM, the resulting exposure pattern is measured, and the measured pattern is an individual pixel correction. Used for. The above means that an analog voltage is selected according to each gray level of the pixel map, and that a correct irradiation dose is supplied by performing pixel modification on all corresponding SLM pixels.
[0088]
(Cited document)
Nelson 1988: US Pat. No. 5,148,157
Cook 1990: European Patent 0 610 183
Sandstron et al. 1990: European Patent 0 467 076
[Brief description of the drawings]
FIG. 1 is a diagram showing a conventional printer. The SLM consists of a micromirror that deflects light from the lens pupil.
FIGS. 2a to 2h are diagrams showing some pixel designs in which four upper pixels are turned off and the remaining five pixels are turned on. FIGS.
FIG. 3 is a diagram illustrating an arrangement of pixels that generate a phase difference by moving up and down like a piston. Thus, it is shown how the edge position can be finely adjusted by the phase SLM.
FIGS. 4a to 4g are diagrams showing a schematic comparison between an SLM with a deflection mirror and an SLM with a deformation mirror; FIGS.
FIG. 5 is a flowchart illustrating a method for translating data and sending it to the SLM.
FIG. 6 shows a preferred embodiment of a pattern generator according to the present invention.
FIG. 7 shows the generally expected complex amplitude for different types of SLMs.
FIGS. 8a and 8b are diagrams schematically showing different types of SLM mirror control. FIGS.
FIG. 9 shows an example of alignment marks to be used in the present invention.
FIG. 10 schematically illustrates an example of a detection and alignment system according to an embodiment of the present invention.
FIG. 11 schematically shows an example of an alignment method according to the invention.
FIG. 12 schematically shows an example of a beam measurement mark to be used in the present invention.
FIG. 13 schematically shows a first example of an embodiment of a detection and alignment system according to the present invention including field reduction means.
FIG. 14 schematically shows a second example of an embodiment of a detection and alignment system according to the invention including field reduction means.
FIG. 15 schematically shows an analog device configuration of a reflecting surface to be used in an embodiment of the present invention.
[Explanation of symbols]
601 Modulator
602 Light source
604 projection system
607 Electronic data processing and transmission system
902 snake mark
903 Search mark
1001 Light source
1004 lens
1009 Detector
1009 'detector
1301 (Second) Modulator
1401 Small angle scanner

Claims (4)

An apparatus for generating a pattern on a workpiece,
A light source that emits light in the wavelength range from extreme ultraviolet (EUV) to infrared (IR);
A spatial light modulator (SLM) mounted on an elaborate stage, the spatial light modulator (SLM) having multiple modulator elements (pixels) adapted to be illuminated by the radiation;
A projection system for generating an image of the modulation device on the workpiece;
An electronic data processing and transmission system for receiving a digital description of the pattern to be written, converting the pattern into a signal for a modulator, and transmitting the signal to the modulator (601);
A precision mechanical system for positioning the workpiece and / or the projection system relative to each other;
An electronic control system that controls the position of the workpiece on a positioning stage, the transmission of the signal to the modulator, and the intensity of the radiation so that the pattern can be printed on the workpiece;
Said apparatus comprising means for rotating the coordinate system of the pattern to be written by rotating the digital description supplied to the electronic data processing and transmission system,
The means for rotating the coordinate system of the pattern to be written by rotating the digital description comprises means for rotating the modulator signal provided by the electronic data processing and transmission system;
The apparatus, wherein the modulator signal is converted in real time after rasterization and before reaching the modulator .   The apparatus of claim 1, wherein the SLM is a two-dimensional micromirror matrix.   The apparatus of claim 1, wherein the SLM is a one-dimensional micromirror matrix. A method for generating a pattern on a workpiece:
Emitting light in the wavelength range from EUV to IR;
Modulating the emitted light by a spatial light modulator (SLM) having multiple modulator elements (pixels);
Projecting an image of the modulator onto the workpiece;
Receiving a digital description of a pattern to be written in an electronic data processing and transmission system;
Further comprising rotating the coordinate system of the pattern to be written by rotating the digital description supplied to the electronic data processing and transmission system, comprising:
Rotating the coordinate system of the pattern to be written by rotating the digital description comprises rotating the modulator signal provided by the electronic data processing and transmission system;
The method, wherein the modulator signal is converted in real time after rasterization and before reaching the modulator .

Images