Structure manufactured on a photosensitively layered substrate by means of focussed laser radiation and method and apparatus for its fabrication.

p0001The invention concerns a manufactured on a light-sensitive coated substrate by focused laser radiation surface pattern, in particular structure and a method and a device for its or their preparation, in particular structures of chromium on glass which is used as masks or reticles for photolithography in semiconductor production will.

p0002Masks and reticles called with extremely precise structures of chromium on glass substrates are used for the production of photolithographic structures in semiconductor manufacturing. A 5X reticle, ie a structure which is reduced fivefold photographically on the semiconductor wafer and in the near future embodies probably the most common type of mask, may consist of a quartz plate having a dimension of 150 mm consist x 150 mm, the structure of an out having an opaque chrome. The structure is formed by exposure to light or an electron beam of a light or electron-sensitive coating which covers a chromium film. The light- or electron-sensitive coating is then chemically developed and the exposed portion removed. In a subsequent etching operation, the chromium is etched away at the locations of the removed coating, and the remaining chromium film forming the structure.

p0003The smallest line widths known 5X reticle be about 2 microns. The required accuracy is considerably higher. Permissible overlay error, ie, the permissible difference between the position of the chrome edges in two successive steps prepared identical reticle is in the order of 0.05 microns.

p0004Reticles are in the semiconductor manufacturing its main application for the production of more accurate structures. However, there are also many other applications exist, for example in the field of integrated optics, diffractive optics, computer-generated holograms, the micro-machining of miniaturized sensors, optical information storage and superconducting devices. The high cost of existing Herstelleinrichtungen for surface patterns or structures with sufficient accuracy, particularly electron-Reticleschreibern, represents a hurdle for the development of new devices on these less developed areas, especially in universities and smaller companies.

p0005All known pattern generators is common that the mask or the reticle is described in a digital database. This database contains a list of all structural elements with their geometrical data.

p0006Before writing the structures the geometric data is converted to the format used by the writing hardware. During this reaction, the geometric coordinates on the addressing resolution of the hardware, that is, tailored to the writing Adreßraster.

p0007A case occurring approximation of 0.25 microns, which is generated by a Adreßraster of 0.5 microns, is acceptable if the structure with this address grid is formed and will never be rescaled. However, effect operations, such as a uniform scaling of a structure by, for example, 93%, a predetermined process margin of error (for example, an increase in the size of all the elements to compensate for low etching during the process sequence), for example, 0.15 microns or a displacement of the structure or surface pattern to an additional distance, unpredictable approximation that appear in the structure.

p0008An address grid that is fine enough to make the approximation meaningless is required for these operations and process steps are possible. A finer grid also allows the structure regardless of the used machine, with which the structure is written, can be produced. This is desirable because the design and fabrication of the mask takes place in different production sites. Preferably, the approximation should be microns to 5-fold reticles less than 0.05.

p0009Most modern pattern generators use a raster scan principle with a scanning beam which is either an electron beam or laser beam along parallel lines and rows on the substrate, which is covered with a radiation-sensitive coating is deflected. The beam is turned on and off in dependence on a bitmap (bitmap) for the structure, which is stored in the control system. Another possibility is that the beam is produced during a writing time which is derived from data that are stored in a compressed intermediate format.

p0010For a structure (surface pattern) with an address grid of 0.5 microns, a bit map with one bit for each dot can be produced. Normal write speeds are from 5 to 10 square millimeters per second. This corresponds to a reasonable data rate of 20 to 40 megabits per second with a data volume of 10 to 100 gigabytes per mask. Using an appropriate Datenverdichtungsalgorithmusses it is possible to store the compressed data on a hard disk and to expand to the full volume of data to the write time. Further, the data rate is compatible with commonly used families of integrated circuits and are commercially available electro-optical and acousto-optical modulators.

p0011In principle a pattern with an addressability of 0.05 micron with scan lines at a spacing of 0.05 micron and a pixel spacing of 0.05 micron along the scan line could be written. However, corresponding to a bitmap with a 0.05 micron grid and an above-mentioned areal throughput of a data speed 2-4 Gbit per second. The modulation of a single writing beam at that rate is not possible. In addition, the data volume is a hundred times larger than the 0.5 micron grid and is fairly unhandbar. Even a Realzeitexpandierung from the compressed data format is impracticable as the flow of data would clog the data bus of most processors. Technological limitations in the modulation rate and the data flow would reduce the write speed considerably and make a reticle writer not applicable to a full pixel map with 0.05 micron address grid.

p0012It is known to produce masks or reticles by means of a laser scanning device (U.S. Patent No. 4,060,816). This known device is, however, too slow for example, a practical application in mask manufacturing.

p0013The simple mechanical scanning in the X and Y direction results in an inappropriate flow rate, and it is known that a sub-scan principle can be used to increase the writing speed (US-PS 4,455,485).

p0014It is also known (GB-PS 2,215,553) that one may obtain a finer Adressierraster with a fixed write coarse grid if the surface is scanned multiple times. Successive viewing screen can be overlaid with less displacement, so that the density of the resulting grid is increased. several times is written with a modified each bitmap, so that some pixels to be written twice, and other pixels are written only once as another way to improve the Adressierauflösung. Since the Bronnpunktfleck is greater than the distance between the picture elements, the exposure is a smooth function, which covers the individual picture elements. A structural element has edges from which the exposure of zero increases smoothly up to full exposure. Adding a number of pixels with half intensity, the effect of the displacement of the intensity profile by half a pixel pitch. Four times to post a 0.5 micron grid with twice the original position and twice a 0.25 micron offset positioning is chosen will result in an effective Adressierraster of 0.125 microns according to the method of GB-PS 2 215 553. In this case, results in an effective Adressierraster that is less than four times the hardware grid; However, one needs four over the surface guided writing passages.

p0015It is further known that when playing video picture elements can be positioned at pixel fractions using the so-called "anti-aliasing" technique (IEEG CG + A, in January 1981, pages 40 to 48). The image which can be generated from a digital database is scanned to fictitious sub-pixel positions. The sampled subpixel are smoothed over several real pixels out and smoothed, ie spread data of each sub-pixel are added to the actual pixel positions. The eye interprets the resulting blurred transitions as sharp edges, and the coming into appearance position of edges can be changed in small fragmentary pixel increases by modifying the dose to the pixel locations.

p0016Further, it is known that the run-length encoding (RLE), a suitable data compression algorithm for pixel data containing long sequences of individual values ​​is. The volume of the compressed data is a function primarily of the number of transitions from one value to another and back and depends to a small extent from the bitmap resolution.

p0017The object of the invention is to provide a structure of the aforementioned type and a method and an apparatus for their production by means of a focused laser beam, in which a considerable increase of the raster resolution without sacrificing the recording speed and good edge quality is achieved with manageable data volume.

p0018This object is procedurally solved by the subject matter of claim 34, by the subject matter of claim 1 and device standpoint by the subject of claim 26th The dependent claims characterize further developments of the invention.

p0019Advantageously, can be by the invention is an improvement of the grid resolution by a factor of 4 or more without affecting the write speed can reach. The edge variation is smaller than in the naked "anti-aliasing" technique, and the data volume has a manageable size.

p0020With regard to increasing the density of Adressierpunkten has been found that an increase in the density of the scan lines requires an unacceptable amount of write time. In practice, both the scan speed and the maximum modulation frequency of the laser beam is limited. Doubling the scan line density would require a doubling of the scanning speed and a doubling of the number of per second to be written pixels required. In direction along the scan line is to reduce the pixel pitch at least four times, especially five- and eight times, not too serious, since a denser Adressierraster means only a small time increment between grid points. With an unchanged minimum feature size in the surface pattern, the switching frequency of the modulator is unaffected. The difficulty lies in the greater volume of data to specify a structure with a denser grid.

p0021The figures serve to illustrate the invention.

p0022Show it:<ul><li>FIG. 1a to 1 d</li><li>Representations of known methods;</li><li>FIGS. 2a and 2b</li><li>Representations of edge regions of scan lines;</li><li>Fig. 3</li><li>an apparatus for manufacturing a structure or a surface pattern which is an embodiment of the invention;</li><li>FIGS. 4a and 4b</li><li>Representations of edge regions of scan lines;</li><li>Fig. 5a</li><li>a fixed pixel grid;</li><li>Fig. 5b</li><li>a pixel arrangement according to one embodiment of the invention; and</li><li>Fig. 5</li><li>an embodiment of a modulator driver logic that can occur in the embodiment of Fig. 3 for use.</li></ul>

p0023Prior to a detailed explanation of the invention first is the asymmetry between edges in two directions of a catch-scan structure are explained. Figure 2a shows that in a cross section perpendicular to the scan line, the entire energy to the scan lines 201 is centered with a scattering around each scan line 201. The slope of the edge 202 of a resulting exposure 203 by the course of the scattering function Only determined. By appropriate selection of the scanning line spacing depending on the line width of a litter-filled structure with almost level exposure can be achieved. The need for solid, smooth filling up between the scan lines determines the irradiation spot size. If the irradiation spot size is determined once, there is no further deterioration of the edge sharpness.

p0024As Figure 2b shows two effects are effective in the direction along the scan lines that make the edge less sharp. The first effect is due to the fact that the modulator is not an ideal switch, but has a finite rise time, which causes a certain amount of imprecision in the structure. The second effect is more fundamental. The radiation energy is not supplied to the pixels in a flash points but one of the pixel dots on or off. From Figure 2b it can be seen that a large number is scattered on weak line spread functions 204 through the range over which the beam is switched on. This is a newly rising integrated exposure is achieved 205th In a typical raster scan structure, the edges 205 are less steep than transverse to the scan line in the direction along the scan lines by 30%.

p0025The edge sharpness is an important quality parameter, and it should be as high as possible. However, the symmetry of the edge sharpness in the X and Y directions is equally important. The photosensitive coating and the Chromätzverfahren a photomask having a threshold characteristic. This has an effect such that the chromium film remains intact at the location where the exposure dose is below the threshold, and is removed at the location where the exposure dose is above this threshold value. A different edge steepness in different directions complicate the process and exposure depending on the dimensions in machined patterns more and require greater calibration effort and increased accuracy in the process control.

p0026The YY asymmetry could possibly compensate by an elliptical focal spot. A detailed investigation, however, shows that in a given system, the number of resolution points along the scan line is a system parameter and an elliptical focal spot, which is shortened along the scan line, leads to a shortening of the scan line itself, and thus a reduced writing speed causes.

p0027In the invention, a refined Adressierraster is achieved in a manner which is adapted to the practical and theoretical limitations above optimal, so that with minimal effort at writing time, data volume, and edges defining a compressed grid of pixels is reached.

p0028In particular, in the present invention is a circular focal spot for use, which ensures a higher than required edge definition in the direction transverse to the scan lines. In this direction antialiasing interpolation between the scan line is inserted, which ensures a sub scanning line addressing. The existing loss of edge definition is taken from the excessive edge sharpness and provides an approximate symmetry between Y and Y direction restores.

p0029In direction along the scan lines, the increased address dot density is ensured in such a way that no deterioration of the edge sharpness is caused. This results in an increased density of pixels (pixel dots). The resulting, highly asymmetric grid having at least four times higher pixel density in the direction along the scan lines, together with a one-dimensional interpolation transverse to the scan lines permits compared to conventional techniques, a finer address grid in two directions, the edge definition in the worst direction is hardly affected.

p0030The requirement of a higher modulation frequency does not occur, since the minimum feature size that must be written, does not shrink with the address grid. Under this approach, the data still contain strains of multiple address points with constant energy and a suitable compressed data format, that is, the run length coding prevents an increase in the volume of data is proportional to the density of the address points.

p0031In the figures 4a and 4b are shown cross-sections perpendicular to the longitudinal extent of the scan lines by an edge portion. 4a shows the cross section of an edge area without interpolation and 4b the edge region interpolation. Modern photosensitive coatings have an extremely hard gradation. They possess a dose threshold value above which the photosensitive coating film is removed in the developing solution and below which the photosensitive coating film remains unaffected. Figures 4a and 4b respectively show two lines, an upper line 401 and a lower line 402. The upper line 401 represents the lower limit line is for the complete elimination of the photosensitive coating film during the developing treatment, and the lower line 402 represents the upper boundary line for the highest light dose at which the coating film is not attacked. For light doses between the two lines, the result is uncertain and depends on differences in the developer solution and starch - and aggressiveness of the coating layer thickness and sensitivity. This uncertainty zone transmits a zone of edge roughness and feature dimensional change 403. Of course, the exposure profile should be as steep as possible. In Figure 4b, a first scan line 404 is written with a lower dosage, and the edge 405 is shifted in comparison to the edge in the Fig. 4a. Further, the edge is less steep, and the zone of the edge roughness 406 is slightly wider. For clarity, the roughness is exaggerated, and in reality, the difference of the slope at the edges between 4a and 4b about the slope difference between 2a and 2b.

p0032Typically, the interpolation is not at the edges of the structural elements in the input data. However, it may be added during the data preparation, although the reverse is possible.

p0033Hereinafter, the data delivery will be explained. The changes in the equivalent pixel data stream are low, which is known from the assumption that a minimum feature size is considerably larger than a pixel in the scanning direction. This gives rise to two consequences, first, that the run-length encoding (RLE) is effective as a storage format, and secondly that the hardware does not have to be configured to generate any Bitstream. Preferably used in the invention, a hardware (Figure 3) are used, which run length-encoded data processed directly, ie a part of the energy value and of a permanent or switch position value.

p0034Repeated charge of RLE data items in a modulator driver logic 6 occurs at a rate that is usually lower than the pixel rate. The loading of data is in time with a data delivery clock which has a clock speed that is compatible with standard IC families, for example FAST-TTL. Two pieces of data can be combined for an effective application of the available bits to a data word. The modulator drive logic 6 separates power and change position fields at the input and supplies a modulation means 5, the power control signal after a time which is derived from one or both of the last position fields. If the data delivery clock having a frequency which is lower than the pixel rate, which is usually the case, the delay of the modulator control signal is not generated by a counter but preferably, a selection among the various clock signals with time delays between them.

p0035Loading a RLE ingredient for each clock cycle or for each clock phase multiphase clocks of data Returns Act prohibits the writing of structural elements, which are shorter than the equivalent of a clock cycle or a clock phase.

p0036The radiation power may be modulated either by modulation of the light source or by using a modulator with a continuous light source . It is also possible to provide a nonlinear relationship between the exposure of the FEU ßersten pixel at the edge of a structure element and the displacement of that edge, by using a table look-up procedure or other correction processes. Further, it is possible to modify the exposure of more than one row of pixels along the edge, to obtain an exposure gradient which is independent of the position with respect to the scan lines.

p0037In contrast to the multi-pass principle, which is used in the GB-PS 2,215,553, is used in the invention, a variable exposure by modulation of the beam at different power levels during a single writing pass on the application. The result is an unlimited fine grid without performance problems.

p0038It is possible to apply the invention in such a way that successive pixels with an unchanged power can be written without more than one control word is generated.

p0039In the Fig. 3 shows an embodiment for a device is shown with which structures (surface patterns) may be manufactured in photosensitive coated substrates. In the embodiment shown, a substrate 3 is used in the form of a glass plate with a photosensitive coating, for example a photoresist coated. The glass plate 3 is located on an object table 19 which can be moved in X and Y direction. A writing head with a focusing lens 15 and a deflector 14, which is preferably constructed as acousto-optical deflection device is arranged stationary in the X- and Y-directions. The focusing lens 15, however, can be moved freely in the vertical direction (Z-direction). It is located a few microns above the surface of the glass plate 3 on an air cushion. Since the air cushion is loaded only with the weight of the focusing lens 15, the height of the air cushion can be kept constant due to the independence of the Z-coordinate, so that the surface of the glass plate 3 is always located in the focal plane, even if the surface is uneven ,

p0040A laser beam source 13 generates a laser beam which is, in view of its performance, in particular intensity, modulated by a modulator device 5, in particular acousto-optic Modular device. The focusing lens 15 focuses the laser beam and forms a focused laser beam 1 which is directed at the surface of the formed glass plate as substrate. 3

p0041To monitor the position of the stage 19 are provided 18x and 18y position monitoring devices. These monitor the positioning of the object table 19 in the X- and Y-direction relatively with respect to the writing head (deflector device 14 and focusing lens 15). They, together with electric motors 17x and 17y a servo mechanism that causes the precisely controlled movement of the stage 19th

p0042The pressure acting in the X-direction servo mechanism moves the stage 19 by means of the electric motor 17x which is preferably designed as a linear motor, in the X direction at a constant speed for producing stripes 30 of a certain width in the scanning of the laser beam 1 in the scanning rows 2. When a respective stripe 30 is completed, the X-servo mechanism performs a return stroke into the starting position. Subsequently, the object table 19 by the electric motor 17y which is preferably formed as a stepping motor to move a strip width in the Y direction.

p0043A scanning circuit 26 generates a high frequency deflection with linear ramp function. The scanning circuit 26 is connected to the deflector fourteenth In this way, a linear deflection of the laser beam 1 in the width of the respective strip 30 is effected. The focal point of the focused laser beam 1 is deflected along the scan lines 2 extending perpendicularly to the longitudinal extension of the strip 30 on the surface of the glass plate 3rd With 8 the positions of the pixels in the scan lines 2 are merely schematically, ie not to scale, respectively.

p0044The laser beam source (HeCd) 13 generates the laser beam 1 with a preferred wavelength of 442 nm. The laser beam is directed through the acousto-optical modulator 5. This is controlled by a modulation drive signal hochfrequenzten 4th The modulation drive signal 4 is supplied from a modulator driver logic 6th The power of the modulation drive signal 4 is controlled by an analog modulation drive signal of a digital-analog converter. The digital to analog converter, which exist in the modulator driver stage 6 (Figure 6), is controlled by a power control signal 7 via a preferably designed as a register memory. This stores the digital power control word required for the control.

p0045Figure 6 shows a preferred embodiment of the modulator drive logic used in FIG. 3 This has two trained as registers 605 and 608 store bit words. The register 605 is with performance 7, coming from the data delivery device 24 (Figure 3), loaded. The register 608 is delayed data 25, which also come from the data delivery device 24, loaded. This is due to a caused by the clock signal 31 transition. A digital to analog converter 606 is thereby still operated by previously in the memory 607, stored performance data. This may also be designed as a register. A digitally controlled Verzö-ge tion circuit 609 transmits the clock signal 31 with delay on its output side. The delay depends on the data stored in the register 608th After this delay, the active edge of the clock signal at the output 610 of the delay circuit 609 is supplied. The value stored in the register 605 is loaded into the register 607 and supplied to the input side of the digital-analog converter 606th After the internal delay in the digital to analog converter 606, a new analogue Leistungsseuersignal 7 is supplied at the output of the modulator driver stage of the modulator device. 5

p0046The embodiment of Figure 3, in which the thick solid arrow lines represent the flow of data structure and the thin lines, the lines for the control signals further includes the write control unit 29, which can be configured as a processor. The write control unit 29 initiates the reading of the data from a memory 23 and sends instruction or command signals to an XY-servo control unit 27, 19 for the movement control of the stage A clock generator 28 generates the clock signal 31, the operation of the data delivery logic or data delivery device 24, the modulator driver logic 6 and the deflection circuit 26 synchronized. From the servo control device 27, a separate readiness signal 32 is further supplied to the deflection circuit 26th This readiness signal 32 ensures that operation of the deflector 14 rests until the object table 19 reaches its correct X-position. In this manner an accurate positioning of the object table 19 is achieved in view of the deflector fourteenth Since the modulator device 5 and the deflector 14 operate free of inertia and are driven by the same clock signal 31 will result in a high position accuracy.

p0047The XY-servo control unit 27 thus operates time coordinated with the servo system (17x, 17y, 18x, 18y) and the deflection circuit 26, together, so that it is ensured that the scan line stripes 30, scanning in the respective stripe width by the deflector 14 and modulation are performed synchronously by the modulator means. 5 This ensures that the average position error is less than 0.05 micron, including environmental influences on the laser interferometer (position monitoring devices 18x and 18y). In the illustrated embodiment, the position word has eleven bits, and the power word has four bits. One can achieve a grid of 0.5 x 0.037 microns without antialiasing. From this approach errors that are compatible with the requirements described above result. The clock frequency preferably is 25 MHz, and the scanning length corresponding to the width of the strip 30, preferably is 160 microns. Considering the scanning and Streifenrücklaufhubzeiten the write speed is four square millimeters per second with a single write beam.

p0048In the invention, also a plurality of writing beams can be used instead of the single writing beam. It then requires a corresponding number of modulators, lenses and parallel data paths, etc ..

p0049The position field contains sufficient data to compute an absolute position for a change in performance. The location field is not necessarily encoded as absolute position. Particularly in the case of paired power and run length data, the Runlängecodierung may be embodied either as a spacing in order to retain the current value before the next change, or as a distance to retain the next current value, or as an absolute position at which the next value begins, or as an absolute position at which the next value ends, or as any other adequate encoding. Furthermore, the position field may include two subfields have a completely trained subfield and another relatively trained subfield. Instead of the modulator device 5 can also be a modulated light source, for example a semiconductor laser or a plurality of semiconductor lasers, are used.

p0050A data input device 20 includes digital input data for a produced on the substrate 3 structure 9. These data are in the form of a list of structural elements 12 and 16 or regular structure forms or structure constant before, so that the structure elements can be computed. The specification of the exposure may be normalized to a maximum exposure or implicitly assumed in the case of a single exposure value for all patterns. In all cases X- and Y-length scales, the angle between the axes and the absolute exposure dosage can be modified by the operator of what is specified in the input data. There may be other processing of data is carried out, for example, Reflection, Invsersion, gray scale corrections and Vorkompensierungen for undercutting or proximity effects. The delivery cycle may be a Einzelphase- or multiphase clock. Loading the RLE (Runlängecodierung) -Dateninhalte can be performed in a register or in two or more registers. The decode logic may include a prefetch or other buffers to increase the operating speed.

p0051To increase the operation speed, it is also possible to multiply a part of the data paths, several times to provide, for example, the modulator driver logic, even if only one laser beam is used.

p0052The relative movement between the print head and the substrate can be obtained such that the write head moves, or vice versa with respect to the stationary substrate. It is also possible that the print head in one direction and the substrate in a perpendicular direction to move.

p0053In the embodiment shown in FIG. 3 in the data input device 20, the geometrical characteristics of the structure to be fabricated 9 or structural elements 12, 16 together with their exposure in a format 10 which is passed to a data processing device 21, specified. In the data processing device 21, the geometric data in a compressed intermediate format 22 is converted. This compacted intermediate format 22 is input to a digital storage device 23rd Further, the compressed intermediate format of a data delivery device 24 is supplied, in which the stored data are processed in an internal data format. This internal data format consists of a pair of data contents 7 and 25, namely a data content 7 for the beam power and a data content 25 for the position. This data content 25 may, as has been explained already in connection with Figure 6, include delay data. The data content 7 for the beam power is supplied in the form of a power control signal 7 and the data content for the position in the form of a position signal 25, the modulator driver stage 6th The modulator driver stage, which is formed as a logic circuit, retrieves data from the data delivery device 24th The modulator driver stage 6 supplies the modulation drive signal 4 which is variable at the arranged at regular intervals along the extending perpendicular to the stripes 30 scan lines two pixels times 8. The distance between two neighboring pixel points 8 along the scanning lines 2 by a factor of at least 4 less than the distance between two adjacent scan lines.

p0054The invention can be a fine structure 9, which is composed of the structural elements 12, 16, produced by the focused laser beam 1, whose accuracy is comparable to those structures which are produced by means of electron beam machines. However, in the device according to the invention requires a significantly lower cost than with electron beam equipment. Further, one achieves a considerable improvement of the address grid, characterized in that in the longitudinal direction of the scan lines, the distance between adjacent pixel dots is dimensioned smaller than the distance between adjacent scan lines. This leads to an improvement of the addressing grid or resolution lower edges disorder than in an anti-aliasing scheme while maintaining a manageable volume of data.

p0055The Fig. 5a shows a fixed grid with small circles showing the diffraction-limited extent of each pixel. The pixels are centered on a pixel grid 504 and a scan line stripe 503 is formed by successive scan lines 502nd The pixels are centered on the scan line 502 and arranged at equal distances from one another. The pixel pitch and the distance between successive scan lines 507 is sized equal. The figure shows an exposed scan line 502 and a Stuffed with pixel area, which extends over three stripes 503rd This area has a relation to the vertical sloping edge 5 and a relative to the horizontal sloping edge 506. In the illustrated structure, it is not possible to position the edge of fractional pixel points. The sloped edge profile is therefore formed jagged.

p0056Figure 5b shows a pixel arrangement of the invention. The pixel density along the scan lines 502 is higher, namely by at least four times higher than in the direction of the longitudinal extent of the stripes 503 or perpendicular to the scan line 502 in the figure 5b this is still illustrated by the spacings A1 and A2. A1 is the distance of the respective pixel addressing in a scan line 502, and the distance A2 identifies the spacing of the respective scan lines 502 from each other. 5b as can be seen from the FIG., is obtained in this way an extremely fine addressing at the vertical edge. It can therefore be a relatively smooth flow even with a reach from the vertical sloping edge 508th

p0057The case of a preferred embodiment of the invention coming advantageously still apply power modulation also improves relative to the horizontal inclined edge 509, as shown in Fig. 5b is clearly seen. It is thus that one can be a very fine Addressing achieved in the invention and also achieve a smooth edge course also the tapering section. These advantages are erriecht, although the writing time. 5b does not differ in the embodiment of Fig. 5a and the inventive embodiment of Fig. The invention is achieved to increase the versatility of writing different surface patterns and structures and an additional freedom of movement relative to the discrete screen specified by the scan lines and the pixel clock.