DE102009033206A1 - Polishing method and polishing apparatus for correcting geometric deviation errors on precision surfaces - Google Patents

Abstract

A method and a device for polishing and shaping workpiece surfaces are described, in which the dimensional deviations from the desired geometry are determined, an elastic surface (19) is placed on the workpiece surface (14) and the workpiece surface is created by generating a pressure on the elastic surface on the workpiece surface and moving the elastic surface and the workpiece surface relative to each other. The material of the elastic surface is selected so that it has a dynamic dependency in such a way that the force which is necessary to generate a time-varying deformation in the material is greater for rapid deformations than for slow deformations.

Description

Technical field of the invention

In Many technical applications become one of a kind well-defined shape needed. The degree of deviation the actual shape of one in the technical process It is crucial for applicability.

Examples of applications where the accuracy of such well-defined surface shapes play a significant role:
Surfaces of optical elements such as lenses and mirrors.

In optical systems become the desired imaging properties achieved by the fact that the electromagnetic field of the used Light with the help of refraction (lenses) and the reflection (mirror) well-defined interfaces of materials of different optical properties in the desired manner being affected. The tolerable deviations from the ideal Surface shape are in the range of fractions the wavelength. The smaller the deviation the larger is the goodness of the optical picture. Especially for optical systems with high numerical apertures like those for example in the projection systems for semiconductor lithography to be used, aggravating that the shape of the optical Interfaces no longer by a simple spherical surface can be approximated, but aspherical geometries necessary.

Planarity of surfaces to be structured in semiconductor and microsystem production:
In semiconductor manufacturing, electronic components are built up on silicon substrates using lithographic patterning techniques. For reasons of economy, attempts are made to make the dimensions of the components (in particular of the transistors) smaller and smaller in order to increase the number of components per silicon area unit. The optical imaging methods used in this process also become more sensitive to defocus errors with increasing resolution, ie in particular also deviations from the surface to be structured from the focal plane of the optical system. Therefore, it is necessary to get the surfaces to be structured as planar as possible. This applies both to the surface of the starting substrates and to the surface of intermediate layers. For example, the surface of dielectric layers insulating the electronic components from the overlying metallization level or successive metallization levels must be very flat.

Planarity of Mask Substrates in EUV (Extreme Ultraviolet) Lithography In order to pursue the above-mentioned trend towards ever smaller feature sizes in semiconductor fabrication, it is necessary to use smaller wavelengths than those currently used. A common approach is to use soft X-radiation with a vacuum wavelength of about 13.5 nm. For this radiation, no sufficiently transparent material is known for use in refractive optical elements. Therefore, optical systems are constructed exclusively of reflective elements. In order to separate the incident from the reflected beam spatially, is used with angles of incidence, which differ slightly from the vertical incidence (typically 6 degrees). For the photomask, which carries the circuit pattern to be imaged, this in turn results in a very strict requirement for the planarity of the surface. As in 1 As shown, each local vertical deviation translates into a lateral offset of the imaging rays. An incident beam ( 2 ) is determined by the ideal planar surface ( 3 ) (ray ( 6 )). A surface deviating from the ideal surface ( 4 ) reflects a beam incident at the same location ( 1 ) at a different altitude. This will result in the reflected beam ( 5 ) an offset from the beam reflected from the ideal planar surface ( 6 ) generated.

Hard drive disks:

at the today common magnetic mass storage, the so-called Hard disk drives, the binary information is called Polarization direction of a small magnetic domain stored. The domains are characterized by a magnetic Layer formed on a fast rotating disc is. For read access, a magnetic sensor in very low Distance over the magnetic layer, to unequivocally measure the polarization direction of the domains to be able to. To the necessary small distances between read head and magnetic layer in the required To enable high reading speeds, the magnetic must Layer can be applied to a very flat substrate.

typically, may in the described applications, the deviations of the nominal geometry only a few nanometers,

to Simplification is hereafter with the term "workpiece surface" only that part of the surface of the workpiece for which the requirements in terms of shape deviation and roughness should apply.

It is curved in the technical optics usual Surfaces by the spherical geometry deviate into so-called aspheres and free-form surfaces divide, the first term designating such areas, which still have a rotational symmetry. Because this distinction for the application of the method disclosed here not is essential, in the following the terms "freeform surface" and "aspherical" surface or in short "Asphere" as synonymous term for a curved surface deviating from the spherical shape used.

Around the above described form deviation requirements of To achieve surfaces are according to the prior art Polishing method used. The need for such Polishing arises from this, the previous process steps leave an intolerable surface roughness to have.

For optical components and wafer substrates, these are mechanical Machining processes such as sawing, grinding or milling with which the workpiece surface in the desired Shape is brought.

in the Case of semiconductor manufacturing and in the production of microsystems The roughness is created by having an integrated circuit or a microsystem of multiple layers of locally separated components finite height is built. The components are usually produced by means of lithographic processes, which is a finite Have image depth. The component height in one layer intrinsically produces a surface roughness that is detrimental to the lithographic patterning of the following Component level.

polishing process have the task left over from the previous process steps remaining surface roughness to the required extent reduce, but without the surface shape over to change the tolerable level.

Surface polishing process

To For this purpose, moldings of non-deformable Materials used whose shape is the negative impression of the workpiece surface is, in the following also called Polierkissenträger. On These moldings are polishing pads made of elastic materials attached.

Of the The resulting polishing head is characterized by a suitable mechanical Device under a defined pressure on the workpiece surface pressed. At the same time the workpiece and the polishing head is set in motion so that their surfaces perform a relative movement to each other.

In the gap between the polishing pad and the workpiece becomes a So-called polish, this is typically a suspension from very small polishing particles of a sufficiently hard material. Furthermore, in the liquid optionally suitable Reagents for a chemical attack of the to be polished Material solved.

The Polishing particles in the suspension are from the polishing pad against the Workpiece surface pressed and simultaneously by the relative movement over the surface emotional. This combination of pressure and movement causes a material removal similar the macroscopic scratching.

This Mechanical material removal can be due to the chemical attack of the Surface to be supported, in which the in the abrasive contained reagents the chemical nature of the surface change so that the mechanical removal easier or at all only possible.

In such polishing methods, the macroscopic material removal can be described by the so-called Preston equation ( FW Preston, "The theory and design of plate glass polishing machines," J. Soc. Glass Technol. 11, 214-256 (1927) ), That is, the removal rate is proportional to the product of contact pressure and relative speed between the two surfaces. The proportionality constant is like that called Preston coefficient. It depends in particular on the material of the workpiece surface and on the properties of the polishing suspension.

Out the pressure dependence of the material removal results in the smoothing effect of the process. At local surveys The surface of the elastic polishing pad becomes stronger pressed. Therefore it sets the impressing Surface against a stronger resistance. This local pressure boost causes after the above-mentioned Preston's equation a larger material removal as in the lower environment. This will be the local survey removed faster. This effect continues until the compression of the polishing pad over the entire contact surface is equal to. Then there is the same contact pressure everywhere and the surface of the workpiece only becomes evenly removed.

It it is immediately clear that in this process the surface shape of the workpiece gradually the shape of the polishing head will adapt, and the more so, the longer the polishing process takes. For flat workpieces This means that the polishing pad holder and the polishing pad even worked very flat, so that the workpiece just stays.

For the polishing of aspherical surfaces and free-form surfaces This fact constitutes a principle problem Relative movement between polishing pad and workpiece too Both surfaces have to be realized in certain ways Borders are constantly being moved against each other. The condition of a perfect match of the surface shape But in aspirations at best only in one Point to be guaranteed. In slightly shifted positions Adaptation deviations occur.

About that In addition, it is not for reasons of economy possible for each freeform surface a dedicated To provide polishing head. Instead, you usually hold a more or less large set of polishing heads different curvature forms before and selects depending on the application the most appropriate. These only approximately Increased adaptation of the surface geometry the formal error in addition.

to The solution to this problem are methods and devices have become known in which the shape of the polishing head directly The tool to be ground can be adjusted, for example by sliding Metal pins or thermally deformable materials. In these Method remains the problem described above that the Adjustment can only be made exactly in one point. Because of the required for mechanical removal relative movement is a Surface deviation in the other points unavoidable. In addition, one can through the previous work steps no longer correct the form deviation caused, since it is in the polishing tool is impressed.

Local polishing process

to Avoid these problems when using large area Polishing heads or for the correction of the planar polishing processes generated errors, become local Polishing methods are used, which allow local corrections of the To attach surface shape while the roughness does not have a increase tolerable level.

A Group of such methods is characterized in that the Polishing tool is designed so that the polishing pad either in itself has only a small extent (eg in the case of so-called polishing pins) or so shaped and guided will that at any one time only a small area is in contact with the workpiece.

Exactly as in the planar polishing process is by a suitable Movement of the polishing tool a relative movement between workpiece and Polierkopfoberfläche generated. This relative movement produced with the addition of a polishing suspension and under application a suitable pressure limited to the contact surface Material removal.

The ablated height is controlled by how long the Polishing head at a specific point on the workpiece surface buffing lingers. For correction of surface deviations from the target geometry, known from previous measurements are, the corresponding local polishing times are determined.

adversely in these methods, that is compared to the surface polishing method small contact surface between polishing head and workpiece, too long process times and thus high wear of the Polishing head leads.

adversely is also that the correction of target deviations on such deviations is limited, their lateral dimensions are at least as large as the contact surface. These are typically in the range of a few millimeters.

A Another group of local polishing methods form the so-called magneto-rheological method. These are characterized that the polishing slurry in addition to the polishing grains also contains magnetizable particles. The polishing suspension is applied to a support body and can then with the help solidified by a strong magnetic field. Due to the shape of the supporting body and the magnetic lines can turn a small contact surface between workpiece and polishing head (here the support body with the magnetically solidified suspension).

Of the further polishing process is analogous to those described above Method.

adversely It is a feature of this process that only relatively low removal rates are achieved can be. That's because of that the polishing slurry is not very strongly solidified by the magnetic field can be.

Another method that is used in the state of the art for local post-processing is not to press the particles dissolved in the polishing suspension by a solid body on the workpiece, but solely by a fast-flowing liquid flow. Such a method is for example in DE 101 13 599 A1 described. Here, the liquid flow flows through an annular gap, which is formed between an outlet nozzle and the workpiece surface.

Due to the lack of direct mechanical contact pressure of the polishing particles, the achievable removal rates are very low. In addition, at least in the case of DE 101 13 599 A1 described device that the dimension of the outlet nozzle limits the lateral extent of the correctable deviations downward.

Of the The present invention is therefore based on the object described Disadvantages of existing polishing process to improve. Especially allows the invention to correct shape deviations the lateral extent of which is smaller than the area with existing procedures (ie usually a few millimeters) can be polished. In addition, the required polishing times shortens, which helps on the one hand, the surface fidelity (i.e., the accuracy with the desired surface shape achieved) and on the other hand the consumption or Reduce wear on process agents. Another The task is to use as much as possible of the process parameters accurate and simple mathematical modeling of the surface geometry to get after polishing. Such modeling allows it, the geometry of any remaining deviations after polishing predict their influence on subsequent process steps and also in the end use estimate and if necessary appropriate To apply countermeasures.

As described above, the smoothing effect of polishing with elastic polishing pads due to the different contact pressures the polishing particles on the workpiece surface caused. These differences result from having a local survey of the workpiece surface relative to undeformed surface of the polishing pad deformation in the polishing pad causes. This local deformation causes an elastic counterforce in the elastic medium with which the Polishing pad surface the polishing particles on the workpiece surface suppressed.

For Deformations are much smaller than the extent of the Polishing pad in the direction of deformation is the elastic force proportional to the deformation itself. The proportionality constant is the so-called spring hardness. Will the spring hardness based on the expansion of the pillow so obtained a characteristic Material property, the so-called modulus of elasticity (im English also called Young's modulus).

The Smoothing effect of the polishing principle described above The faster the harder the material of the used Polishing pad is. A hard material (that is, a large one) Young's modulus) becomes a given indentation by a local unevenness of the workpiece surface to oppose a greater force than a soft one Material. This is also the local removal of the local unevenness larger and the unevenness is leveled faster.

In many elastic materials, especially in polymers, the modulus of elasticity depends on it Whether the deformation is of a static or dynamic nature, that is to say more precisely at what speed a certain deformation is achieved. This behavior is usually described by a frequency-dependent, complex-valued modulus of elasticity. A force that varies sinusoidally over time will also cause an equally varying deformation. The maximum amplitude of the deformation is proportional to the maximum amplitude of the applied force, the proportionality constant now being the amount of the (complex) Young's modulus. The modulus of elasticity phase describes a possible time delay between the force and the deformation.

In many polymers, the modulus of elasticity increases with increasing frequency. This increase is due to a so-called glass transition and is particularly strong in the vicinity of a so-called glass transition frequency, while it has a saturation for very small and very high frequencies. The elastic modulus above and below the glass transition frequency may differ by a factor of 10 or more. A typical frequency dependence is in 2 showing the glass transition of a polyurethane (PTMG2000 / MDI3 / BDO / DMPD). The values for the shear modulus (curve ( 8th ) characterized by o) and for the loss factor (curve ( 7 ) characterized by □) at room temperature are against the decadic logarithm of the frequency on the x-axis ( 9 ). The curve for the shear modulus refers to the left y-axis ( 10 ) in units of Pascal, while the curve for the loss factor on the right y-axis ( 11 ) in dimensionless units. (Out: JV Duffy, GF Lee, JD Lee, and B. Hartmann, in RD Corsaro, and LH Sperling, eds., Sound and Vibration Damping with Polymers, (ACS Symposium Series 424), ACS Press, Washington, DC, 1990, pp. 281-300 ),

That in the 2 Shear modulus shown is identical to the modulus of elasticity up to a factor of 2 (1-μ). Here μ denotes the Poisson number, which is independent of the frequency at about 0.5 for most polymers.

This frequency dependence can be described to a good approximation by the following function:

Here, E _{0 is} the static modulus of elasticity and E _{α is} the elastic modulus in the limit of infinite frequencies. ω _tr is the glass transition frequency. A more detailed description of the glass transition succeeds with the Havriliak-Negami Model. In both models the frequency dependence is a function of the ratio between the frequency and the glass transition frequency ω _tr .

Of the Increase in the frequency-dependent modulus of elasticity As a result, slow changes in the deformation require a smaller force than fast.

Indeed is the deformation of a polishing pad during the polishing process dynamic. The relative movement between the workpiece surface and the polishing pad guides each point of the polishing pad one certain distance over the workpiece surface. Looking at the track which a particular point of the polishing pad surface travels on the workpiece surface, This causes any unevenness in the workpiece surface along this route a temporally varying local deformation of the polishing pad at the point under consideration.

Of the temporal course of the deformation is given a given relative movement between polishing pad and the workpiece surface determined by the lateral shape of the unevenness. A far-reaching one Unevenness gives a slower at identical height Deformation in the polishing pad as a little extended rub. Due to the dynamic dependence of the modulus of elasticity is then in points of identical height of both bumps elastic force in case of widespread roughness always smaller than in the corresponding points of less extensive Unevenness. This difference results from the increase in Young's modulus (more precisely its amount) with the speed of deformation. In particular, in the case of a resting polishing pad for both bumps the elastic force in points same level, since in the static case alone the local extent of the elastic deformation the elastic force certainly. Because of the overall smaller forces in the case The extended unevenness is also the difference between the Strength at the foot and at the top of the rub less than in the case of less extensive. Therefore, the latter becomes faster leveled.

This Difference does not stem from that with a more extensive one Rub more material must be removed to a flat surface but is an independent effect. The easiest way to do this is by means of two sinusoidal waves Surface topographies of different spatial Frequency clear. In this case, the amount of material used to complete planarization must be removed in both cases equal, because the volume of the trenches and balancing surveys exactly. A point of the polishing pad, which perpendicular to the wave crests over one of the two Surfaces is experienced in In this case, a sinusoidally varying deformation over time. Because of the elastic behavior described above, the resulting Force difference between Wellenberg and Grabenboden in the case of short-wave Obeflächentopographie larger than in the long-wave Case. Thus, the Abtragsdifferenz is greater in the short-wave case with the result that the unevenness is removed faster.

For the further discussion is the Fourier decomposition of a given one Surface topography appropriate. there is the surface topography as a linear combination of plane waves in two dimensions of different directions and Describe wavelength. In the following we refer to the Inverse of the wavelength as a spatial frequency.

The in Claim 1 formulated property of the dynamic behavior the modulus of elasticity of the polishing pad makes you in the process according to the invention, that the existing deviations by impressing a appropriate shortwave surface structure corrected so that the long-wave deviations (with a small spatial Frequency) are suppressed as much as possible and advantageously only short-wave deviations remain, which then removed much faster in the subsequent polishing step become, as the original long-wave deviations.

to Determination of a suitable shortwave surface structure becomes from the described elastic properties of the polishing pad in combination with the other process conditions a mathematical Derived model, which is the change of a given Surface topography by the polishing process at least approximately describes.

The in Claim 1 formulated property of the dynamic behavior the modulus of elasticity is equivalent to having a such mathematical process model the character of a low pass for the spatial frequencies. The quantitative The characteristics of such a process model depend exclusively from the process conditions and materials used, but especially not on the shape of the workpiece surface.

Polyurethane polishing pads, which are typically used during polishing, do not lie completely on the workpiece surface in the typical contact pressures used during polishing. Instead, because of their rough surface, they only have contact with the workpiece surface with the most prominent tips. The effective contact area accounts for only a fraction in the single-digit percentage range of the nominal contact area (see CL Elmufdi and GP Muldowney, "A novel optical technique to measure pad-wafer contact area in chemical mechanical planarization," in Mater. Res. Soc. Symp. Proc., TY Tsui, Y.-C. Joo, L. Michaelson, M. Lane, and AA Volinsky, Eds., Vol. 914, 2006, pp. paper 0914-F12-06 .).

Under certain assumptions (in particular, that the frequency of the polishing pad tips decreases exponentially with their height, the tips in the contact area are spherical and Hertzian contact is present), it can be shown that the elastic contact can be described by an effective exponential spring law, wherein the prefactor the modulus of elasticity of the polishing pad material is linear (see J. Vlassak, "A model for chemical-mechanical polishing of a material surface based on contact mechanics," Journal of the Mechanics and Physics of solids, vol. 52, pp. 847-873, 2004 ),

For small deformations can be the exponential law by the linear Approach Term of Taylor Development. In this case follows from the Preston equation that leveling a sinoidal surface topography during polishing runs exponentially with time, wherein the time constant proportional to the value of the modulus of elasticity at the resulting frequency, which by the relative speed and the spatial frequency is determined. This behavior is equivalent to a convolution in space with a convolution function, which due to the frequency dependence of the modulus of elasticity is bell-shaped and whose characteristic width with the polishing time increases

For larger deformations, a linear approximation of the effective spring law is no longer sufficient accordingly. The Taylor development must then be extended to higher terms. This complicates the mathematical description of the process, because convolution in space is then no longer sufficient. The mathematical description is advantageously carried out in the Fourier space of the spatial domain. The quadratic term of the Taylor expansion can then be calculated, for example, by the autocorrelation function of the surface profile.

The basic behavior of the polishing process, that due to the in Claim 1 described course of the dynamic elastic modulus, Short-wave surface structures are faster as longwave, higher orders remain even when added consist. This property is the basis of the invention Process.

It It is obvious that a process model is always a statistical one Approximation for the sum of the individual removals can be through the individual polishing tips in the course of the Polishing process are caused. Therefore, for use advantageous for the process according to the invention, ensure that all areas of the workpiece surface, whose shape is to be processed in the polishing process, with a sufficient large number of polishing pad tips come into contact. This also offers the advantage of having the desired Shaping also in the low-lying areas of the workpiece surface is reached quickly enough.

This can be, for example, by a sufficient to achieve great roughness of the polishing pad surface. This will ensure that even in the lower-lying areas the workpiece surface a large number rests of polishing pad tips.

In front This background is the use of foamed materials for the polishing pad advantageous because by the setting the pore size the roughness of the polishing pad can be influenced. According to the prior art, polishing pads roughened with suitable diamond tools before the polishing process (Conditioned). The surface is removed so far, that an open pore structure develops on the surface. The pore walls then form the tips with which the polishing pad during the polishing process on the workpiece surface rests. Larger pores result after roughening a greater roughness.

It is also advantageous between the polishing pad, which directly rests on workpiece surface and the polishing pad carrier an additional intermediate layer of softer elastic Apply material. By suitable choice of elasticity modules and the layer thicknesses of the two elastic layers, one can advantageously achieve long-range height differences in the workpiece surface by the softer Intermediate layer are compensated. For this softer Layer is also a material with a high viscosity such as certain pitches or resins. Such materials offer the advantage of having their static modulus of elasticity is zero. Therefore, the dynamic modulus of elasticity is also very low at low frequencies. Therefore, the Leveling of long-wave deviations between polishing head surface and workpiece surface, such as when used only approximately adapted polishing head surfaces occur very slowly.

As in the current state of the art it is assumed that the deviations from the desired surface shape at a sufficiently large number of points on the Workpiece surface is known.

For the fabrication of optical elements typically becomes this information obtained with interferometric measuring methods.

at the manufacture of semiconductor devices or microsystems can the surface shape usually from the placement data of the components (i.e., the layout data) in the previous levels and the process conditions are derived.

Of Another is required for the polishing process used and whose process conditions a process model has been determined, which as stated above, based on the in claim 1 formulated property of elastic modulus of elasticity Surface a low-pass character for has the surface geometry. This can be done either by empirical measurements on the basis of suitable test structures or directly from the process conditions and material properties or be derived from a combination of both methods.

The workpiece surface is a kurzwel by suitable methods (see examples below) imprinted surface structure. This surface structure is selected with the aid of the process model in such a way that, in combination with the existing surface geometry (that is to say in particular with the known deviations), it corresponds as well as possible to the desired target surface after the polishing process. The aim is to correct the long-wave deviations with the help of short-wave structures.

The lateral extensions of the impressed surface structure should be chosen as small as possible to the Flattening duration for the impressed surface structure in the subsequent polishing step as short as possible to keep. Here, the possible resolution limits of the structuring process used.

The so prepared workpiece is now with the, the Polished process model underlying polishing process. According to the effect described above becomes the impressed short-wave Surface structure leveled much faster than the long-wave deviations of the untreated surface.

Micromachining techniques

The Method for imprinting the surface structure must make it possible with a sufficient lateral and vertical resolution to impose the required structure. The experiments that led to the described invention, show that advantageously minimal feature sizes in the range of about 10 microns are suitable to combine with commercially available polishing pads and typical Polishing speeds (eg 0.5 m / s) sufficient correction options to offer.

Such feature sizes may, for. Example, with the aid of laser-based lithographic processes, such as are customary in the production of photomasks for semiconductor fabrication, are made without technical difficulties on flat substrates. On curved substrates, lithographic structuring is also possible (see, eg "Laser lithography on non-planar surfaces" by D. Radtke and UD Zeitner in Optics Express, Vol. 15, Issue 3, pp. 1167-1174 ),

at This procedure can be used for any workpiece customize the required surface structure. It is therefore very suitable for the correction of individual pieces and small series, as well as for the correction of unsystematic deviations, d. H. Deviations from workpiece to workpiece are different.

to Correction of systematic errors, especially during production of larger numbers appears the Using a photomask makes sense, with the help of an identical Surface structure on a variety of workpieces can expose. The exposure can then be either as contact exposure or done with the help of a projection optics.

A Combining both methods is also beneficial to both to correct systematic as well as random deviations.

Farther The varnish mask can also by high-resolution printing techniques be applied to the workpiece surface. Advantageous are so-called inkjet processes, which take advantage of of piezoelectric or thermal processes, very small amounts of liquid at predefined positions of the workpiece surface can muster. Such methods are well known to print images in electronic form on paper. Accordingly, with such procedures, which can be used for correction necessary Lackmaskenstrukturen on the workpiece surface apply and then analogously to photolithography etch into the workpiece surface.

Analogous to the photomask, can also in the printing technology based Lackmaskenstrukturierung use a printmask to avoid systematic deviations on one large number of workpieces easier to correct. This method is well known in the offset printing art. Here One uses printing plates, which are illustrated with the picture and text data and then with those in a kind of stamping process a large number of prints can be made.

9 shows an apparatus for carrying out the method according to the invention using a lithographic patterning method. A robotic arm ( 43 ) with a workpiece receiving device ( 44 ) places a workpiece lacquered with photoresist first on the object table ( 41 ) of an exposure unit. There, the photoresist with the aid of a laser beam ( 40 ). The laser beam is generated by a suitable optic ( 39 ) focused on the workpiece surface. In order to expose the appropriate correction pattern in the paint at each point of the workpiece surface, the focusing optics ( 39 ) (or alternatively the stage ( 41 )) can be moved in all three spatial axes. Then the robot arm ( 43 ) first place the exposed workpiece in a developer station ( 36 ), in which the paint develops, ie, depending on the type of paint, the exposed or unexposed paint is advantageously removed wet-chemically. Subsequently, in the etching chamber ( 37 ) etching the areas not masked by the paint so that the desired microstructure is created in the workpiece surface. In a cleaning station ( 38 ), the remaining paint residues are removed. Now the workpiece is transferred to the polishing station ( 42 ) and polished.

Other structuring methods are also conceivable for the method according to the invention, such as micrograving, laser ablation or microetching. Such methods have the advantage over microlithography that not only the lateral dimensions but also the local structuring depth for correcting the surface deviation can be varied within one process step. In addition, the material removal takes place directly (ie without detour via a masked etching). Due to the simpler process management, a direct integration into a polishing device is possible. As a result, structuring and polishing can take place directly in the same workpiece fixture as in 10 shown.

To simplify the illustration, a planar workpiece ( 50 ). The workpiece is on a holder ( 49 ), which by the shaft ( 48 ) is set in rotation. On the polishing pad ( 52 ) is the polishing pad ( 51 ) appropriate. Both are going through the wave ( 54 ) also rotated and also pressed against the workpiece. About the polishing suspension feed ( 58 ) the polishing suspension ( 45 ) is applied to the workpiece surface and distributed (by ( 57 ) indicated) there in particular also in the space between polishing pads ( 51 ) and workpiece ( 50 ), For impressing the microstructure according to the method, a device for laser ablation is integrated into the polishing machine. A suitable optic ( 46 ) focuses a laser beam ( 47 ) on the workpiece surface. The optics can be moved in the plane parallel to the surface in order to impart the appropriate microstructure at each point of the workpiece surface. For curved surfaces, the vertical position of the laser structure can be imposed. For curved surfaces, the vertical position of the laser focus can be varied by appropriate optics or by optics itself.

Next the described material-removing Strukturierverfahren can also material-applying method can be used. So can For example, layering processes using a appropriately structured shadow mask are applied. Also high resolution Printing techniques can, with a suitable choice of process parameters be used. Such an application appears in particular advantageous for the use of the invention Method for reflective optical elements. In this case are the optical properties of the substrate unimportant, since the reflection on a subsequently vapor-deposited thin layer he follows.

at the material-applying patterning process may be inevitable that the polishing process applied to structuring Material removes faster or slower than the material the workpiece surface. This difference can in the process model by introducing a location-dependent Preston coefficient be taken into account. Alternatively, after the material applying Structuring a sufficiently thick homogeneous intermediate layer be deposited, which then the new workpiece surface forms, which is processed in the subsequent polishing step.

Naturally It is also possible, material-removing and -aufbringende Combine method.

This may be particularly advantageous if the height deviations extend only over a small proportion of the entire workpiece surface, but point both into the surface and out of the surface. In this case, the different deviations can be corrected by the respective complementary structuring method, so that the area to be structured remains limited to the deviations themselves. An example of such a process sequence is in 11 shown. In those areas where the deviation of the material surface ( 60 ) to the outside, material is removed locally by an ablative patterning process ( 61 ), In the areas where the deviations point inwards, a material-applying process locally converts material ( 62 ) applied. Then a homogeneous layer ( 63 ) which is so thick that it is not completely removed during the polishing time necessary to level the remaining height differences. The polishing process then creates a flat surface ( 64 ) of the layer ( 63 ),

In order to achieve a high dynamic modulus of elasticity for the impressed microstructure, despite a limited resolving power of a structuring method used, it is possible to choose an elastic material whose glass transition frequency depends on the resolution of the structure used coordination procedure.

The elastic properties of many polymers can be deduced from the contributions of the individual monomers via so-called group contribution or fragment methods. This is especially true for the frequency dependence of the modulus of elasticity. Thus, Davis and Szabo were able to determine the frequency dependence of 14 polyurethanes very precisely from the monomer contributions. Noteworthy in this context is that the glass transition frequency in the materials studied varied by about 10 orders of magnitude (see WM Davis and JP Szabo, "Group contribution analysis applied to the Havriliak-Negami model for polyurethanes", Corp. And theor. Polymer Science, 11 (2001), pp. 9-15 ), This means that the glass transition frequency of the polishing pad material used can be adapted to the resolution and the relative speed used within a wide range by suitable material synthesis

Also the sizes of static and high frequency Young's modulus can be achieved by a suitable composition the material used for the inventive Optimize procedure. To the influence of long-range deviation errors between polishing head and setpoint geometry should minimize that static elastic modulus as small as possible be. At the same time, the high-frequency elastic modulus should be as large as possible, so that the imprinted Surface structure is leveled as quickly as possible.

About that In addition, two process parameters can be suitably selected, within certain limits, the tuning between the glass transition frequency and the resolution limits of the patterning process optimize.

To the one shifts an increase in the relative speed between workpiece and polishing pad surface the temporal course of the effected by a given microstructure Deformation to higher frequencies and increases with it the effective modulus of elasticity.

To the another is the frequency dependence of the elastic modulus temperature dependent. Shifts at lower temperatures The glass transition frequency to smaller values. With a decrease in temperature is therefore achieved for a given microstructure also higher values the acting Modulus of elasticity.

It it is obvious that the invention Process cycle, consisting of initiation of surface deviation, Determination and imprinting of the correction pattern and subsequent Polishing, also be carried out several times in succession can. This is particularly advantageous if the achievable Structuring depths of the selected structuring processes insufficient to cover the existing deviations in a process cycle to correct. Also can with repeated application in the different Steps different structuring widths, polishing pad materials and process conditions are used to detect deviations on different To correct length and height scales

Advantages over the state of the art

Out The method according to the invention results in a Number of advantages over the prior art.

The actual correction of the long-wave deviations is no more achieved by polishing but by microstructuring. Micromachining techniques have orders of magnitude better (ie higher) Spatial resolution on, as the local polishing method after the state of the art. This is a much more accurate correction the deviations possible, especially if the lateral Expansion is small.

By the utilization of the dynamic characteristic of suitable elastic Materials for the polishing pad is embossed according to the invention Surface topography leveled much faster as the original deviations. This reduces the required polishing times compared to the state of Technology. With the process shortening also decreases the consumption of process agents (especially less energy and energy) Abrasives consumption) as well as the wear of the polishing tools.

In addition, a shortened polishing time also offers a principal advantage. It is relatively expensive and even impossible for aspheric surfaces to adapt the surface of the polishing pad exactly to the desired surface. For planar and spherical surfaces the Polierkissenträ must ger and the polishing pad itself are worked very precisely. For aspherical surfaces, an exact adaptation is generally possible in principle only in one point. As a result of the relative movement required for the polishing process, deviations from the ideal adaptation result at almost all times because of the relative displacement of the polishing head from the ideal position. With otherwise identical process conditions, such adaptation errors are transmitted to the workpiece the more the longer the polishing time. A shorter polishing time, as achieved by the process according to the invention, therefore allows a greater dimensional accuracy of the polishing process.

Farther can also be the adjustment deviations described in the previous section the polishing head of the desired workpiece surface also corrected by the described microstructure correction become.

Especially for the mentioned local polishing with Polishing pins and similar small polishing heads, allows the described correction of the adaptation deviations the Use of larger polish surfaces. This will be the polishing time for a given workpiece additionally reduced.

Farther the method according to the invention requires only one additional process step, namely microstructuring, to integrate it into existing manufacturing processes. Especially it can be applied to surface polishing processes and already allows the more accurate correction in this process step from local deviations. It also reduces in aspherical Areas due to the shorter process time due to the principle-related mismatch caused geometry deviations. This eliminates the need for at least, however, subsequent local correction polishing steps the necessary corrections are greatly reduced. Furthermore These advantages allow aspherical surfaces same surface fidelity at a reduced rate To manufacture polishing head shapes, thereby also process costs saved.

Further Advantages arise when used in semiconductor manufacturing or the production of microsystems. There are the components produced by layered construction, wherein in each layer the necessary Microstructuring by lithographic processes takes place. The used lithographic processes usually have a planar Focus plane and a small depth of field (also depth of field called). Any unevenness in the surface to be exposed causes a local focus error and thus an error in the width the pictured structure. To minimize this error to hold the intermediate surfaces with above chemically polished mechanical polishing process.

The planarization mechanism of the method according to the invention is essentially based on the property of the elastic modulus of the elastic surface (19) formulated in claim (1). This makes it possible to predict the temporal change of the surface structure with relatively simple mathematical means very precisely. Since known process and material properties are incorporated into the process model, this model can be set up for any process conditions. This eliminates or reduces the hitherto complex process characterization with the aid of topographical measurements on defined test structures, as described, for example, in US Pat DE000010065380B4 to be discribed.

It is customary in the prior art, the arrangement of necessary for the function of the integrated circuit To choose structures so that the surface after the polishing is as even as possible. It is also common additional structures (so-called filling structures) insert or the form of the functionally necessary structures to choose that surface after polishing as even as possible. The rules according to which these changes and insertions are empirical and little exactly.

A The widespread approach is to use the local layout density (ie the proportion of structured area in one place, averaged within a certain environment) to a certain range of values narrow. The size of the averaging area is determined by empirical measurements and as a characteristic Length of the CMP process viewed. Typical values for these planarization lengths are in the range of several 100 μm up to a few millimeters. It is immediately clear that such an approach the influence of the small-scale Layout density variations are not considered. It has but shown that these small-scale layout density variations a substantial part of the height variation after polishing in particular, not by an automatic Descendants of the focal plane can be corrected during the exposure.

About that In addition, such empirical process models need to change every time the process conditions are redetermined.

By Application of the method according to the invention can these rules, determined without renewed empirical characterization Be as the process model in a defined manner of material properties and process parameters. In addition, the Predictive force of the inventive Given process model much more accurate, since the influence modeled the small-scale layout density variations correctly becomes.

Out The same reason is due to the use of the invention Method possible according to the surface geometry to predict the polishing process very accurately. By this prediction is it possible to influence any non-correctable Compensate geometry deviations by appropriate measures. So it is possible, for example, an existing defocus error by a suitable Strukturbreitenvorhalt in the exposed on it Compensate for levels. Furthermore, it is possible, for example, the influence of a local deviation of the dielectric layer thickness simulatively examining the nominal value between two metal levels, to check if the deviation at this point is critical for the function of the circuit is and their influence if appropriate by a suitable interpretation of the electronic Compensate components.

at Lens elements it is with knowledge of the systematic deviations possible from the target geometry, the negative effects through suitable design of other or additional optical elements to alleviate.

The following embodiments of the inventive Methods are exemplified by the planar grinding of silicon wafers explained. This is partly because of that This method provides a quantitative process characterization Is available, based on which the benefits of the process can be quantified by way of example. The other is the illustration of planar workpieces in drawings easier and clearer reach as for curved surfaces.

All Application examples can be achieved by suitable polishing devices also generalize to curved surfaces. Of the The main difference is a curved one To use polishing pad head, which has a negative impression of the workpiece surface is.

Of the significant difference when polishing workpieces other materials consists in an adjustment of the polishing times and optionally the use of other polishing suspensions. Both Modifications have no influence on the essential properties the method according to the invention.

3 shows the schematic representation of a chemical mechanical polishing machine for wafer substrates. In this application, the wafer surface should be polished planar. The polishing pad carrier ( 15 ) is therefore designed as a disc and the polishing pad ( 19 ) as a thin polishing cloth. The polishing pad carrier ( 15 ) is replaced by a wave ( 16 ) rotated. On the carrier is a polishing pad ( 19 ) attached. Through a liquid outlet ( 17 ) the polishing suspension ( 18 ) is applied continuously to the polishing pad so that it rests over the polishing pad and in particular also between the wafers ( fourteen ) and the polishing pad ( 19 ). The wafer carrier ( 13 ) presses the wafer ( fourteen ) with an adjustable contact pressure on the polishing pad ( 19 ), About the wave ( 12 ), the wafer carrier ( 13 ) and with it the wafer ( fourteen ) rotated.

wafer and polishing head are each in rotation about the respective center added. The angular velocity of these rotations is set the same amount on the entire wafer surface Relative speed between wafer and polishing pad surface prevails. The direction of the relative velocity rotates in the frame of reference of the wafer evenly over time. This will achieved an isotropic removal, which in addition in each Point of the wafer is identical.

The relative speed between wafer and polishing cloth surface was set to about 0.5 m / s. The polishing pad ( 16 ) is a double-layer polyurethane towel. The bottom layer which is in direct contact with the wafer surface is a foamed polyurethane cloth about 1270 μm thick. The pores have a typical size of about 10 microns to 80 microns. The second layer is another, thicker polishing cloth, especially with a lower modulus of elasticity. The nominal contact pressure between wafer and polishing pad is about 41.4 kPa. The polishing suspension advantageously contains 30% polishing particles of silicon oxide, which have an average size of 50 nm. A preferred polishing time is 90 seconds.

Empirical studies on the smoothing of the surface topography as a function of the spatial frequency have the in 4 Relationship under these process conditions. The curve ( 21 ) shows the attenuation of the amplitude (plotted on the y-axis ( 22 in dimensionless units) of sinoidal Surface topographies as a function of their wavelength (plotted on the upper x-axis ( 20 ) in μm), or the spatial frequency (lower x-axis ( 23 ) in μm ^-1 ). Significantly, the low-pass behavior of the polishing process can be seen. This process model results from the curve of the dynamic modulus of elasticity approximated by formula (F1) with a glass transition frequency of about 18000 Hz.

As an example of a possible surface deviation, we assume that the otherwise ideally planar 300 mm diameter wafer has a Gaussian elevation anywhere that is 200 nm high and has a half width of about 500 μm. The surface geometry of the wafer is known from an interferometric measurement, which has a lateral resolution in the range of 10 microns. In 5 is the intepolated height profile ( 24 ) of such a deviation as a function of the location (in μm on the x-axis ( 26 )). The zero point of the elevation axis ( 25 ) is chosen so that cancel all deviations on average. The direction of the elevation axis points out of the wafer surface. The height on the elevation axis are given in units of nm. Because of the very small extent of the survey, the average height of the wafer surface is almost identical to the height at long distances from the survey.

After application of the above-described polishing method of the prior art (that is, in particular, without imparting a short-wave surface structure) remains in 6 shown residual topography ( 27 ) (Axis definition and scaling as in 5 ), The assumed deviation is only reduced by a factor of about 4. The remaining maximum height difference after the polishing process is about 54 nm.

Tabelle 1: Zuordnungstabelle zwischen lokaler Höhenabweichung und lokalem Korrekturätzmuster For an embodiment of the method according to the invention is from the height profile in 5 now determined a suitable microstructure. In this example we assume that a lithography method is available which has a lateral resolution of better than 2 μm and an etching depth of d = 400 nm. Then, a suitable microstructure can be found in the following way. The available etch depth is divided into 16 equidistant classes. Each of these height classes is assigned a 4 × 4 pixel pattern. Filled (black) pixels denote 2 × 2 μm ² large fields which are etched, while no etching takes place on blank (white) pixels on 2 × 2 μm ² large fields. The average height of the resulting local etching patterns are identical to the center of the associated height interval except for an offset of d / 2 and the factor "-1". Such a set of local etch patterns along with the associated height classes is listed in the following table.

Table 1: Mapping table between local height deviation and local correction etching pattern

From the interferometric height data of the workpiece surface, the average heights are determined in flush adjacent square 8 × 8 μm ² large areas. According to the definition of the height scale, this local average height is equivalent to the difference between this local and the global average height. The local average height falls into one of the 16 equidistant altitude classes. The local microstructure needed to correct this difference is then determined from Table 1.

This relatively simple method with the aid of a fixed assignment between height class and local microstructure serves at this point only to clarify the method. It is obvious that the Be mood of the microstructure can be significantly improved. For example, a finer height discretization can be used in combination with a larger set of local patterns.

Furthermore, in the method described by way of example, the process model for the polishing step has only been used to specify the size of the local correction etching patterns. One recognizes in 3 in that the size of the local correction patterns of 8 μm falls within the structural width range in which the attenuation of the surface topography is very strong. The exact shape of the attenuation curve, however, was not exploited to determine the correction pattern. Completely computer-aided procedures, which optimize the correction pattern taking into account the entire process model, allow an even better adaptation to the existing deviations.

This microstructure is now impressed on the original height profile. This creates the in 7 shown height profile ( 30 ) (Axis definition and scaling as in 5 ), It can clearly be seen that almost the entire surface is etched off at the maximum. Here, the etching lowers the local height level by 400 nm. Outside the local deviation, the etch pattern is used with 50% area coverage, resulting in a local mean height reduction of only 200 nm. This difference corresponds exactly to the original amount of deviation. On the flanks of the deviation, the area occupancy of the correction patterns towards the center increases to compensate for the increasing altitude profile.

In 8th the surface profile of the thus modified surface is shown according to the polishing method adopted here (axis definition and scaling as in 5 ), The residual topography ( 33 ) is only about 5 nm and has thus been reduced by an order of magnitude compared to the pure polishing process. In order to achieve a comparable planarity without the surface structuring, the polishing time would have to be increased from the 90 s assumed here to about 300 s.

In this exemplary method of determining the microstructure, areas in which the workpiece surface differs only slightly from the desired geometry are etched with a correction pattern of 50% area occupation. This is necessary to also (in 5 not shown) to be able to correct negative height deviations. The disadvantage, however, is that basically the entire workpiece surface must be etched. Depending on the requirement and the output topography, it may be advantageous to appropriately shift the reference level for the height deviation and to use the entire height correction range made possible by the etching process only to correct the positive deviations. Alternatively, a combination of material removing and applying microstructuring method can be used.

In In another application example we use a laser ablation method for surface structuring.

laser ablation typically achieve a minimum patterning width of 10-50 μm. Using the same procedure as in the previous example for determining the correction etching pattern, this results a maximum structure width of about 40 μm-200 μm. With this structure width, the modulus of elasticity is below the above process conditions are not yet in the high-frequency Saturation.

Therefore, in this application example, a relative speed of 2.5-12.5 m / s is used, which is a factor of 5-25 higher than in the previous application example. Should such relative speeds with an in 2 can not be reached, for example, can be used with linear polishing machines in which the polishing cloth is guided over the workpiece only in a translational movement.

The increase in the relative speed has the consequence that the spatial frequency of the impressed microstructure is converted into a higher frequency of the deformation generated in the polishing pad. In particular, the context applies: ω _t = ν _r ω _s (F 2) where ω _t denotes the temporal frequency of the deformation, ω _s the spatial frequency and v _r the relative velocity.

In another application example, we assume that the achievable maximum relative speed is only 2 m / s and the structuring method has a minimum resolution of 20 μm. Thus, the temporal deformation frequencies generated by the microstructuring are only a factor 0.4 lower than in the first application example.

Around To compensate for this factor is in this application example the Temperature dependence glass transition exploited.

wobei ΔH eine materialspezifische Aktivierungsenergie, R die Gaskonstante, T die Temperatur und ω₀ eine Materialkonstante ist. Da die Frequenzabhängigkeit des Elastizitätsmodul nur vom Verhältnis ω/ω_tr abhängt, ist eine Verschiebung der Glasübergangsfrequenz von ω_tr,1 zu ω_tr,2 equivalent zu einer Skalierung der Frequenzachse um den Faktor ω_tr,2/ω_tr,1.The temperature behavior of the glass transition frequency follows, in the case of most polymers, to a good approximation an Arrhenius law, that is to say:

where ΔH is a material-specific activation energy, R is the gas constant, T is the temperature and ω _{0 is} a material constant. Since the frequency dependence of the modulus of elasticity depends only on the ratio ω / ω _tr , a shift of the glass transition frequency from ω _{tr, 1} to ω _{tr, 2 is} equivalent to a scaling of the frequency axis by the factor ω _{tr, 2} / ω _{tr, 1} .

In order to reduce the glass transition frequency ω _tr by a factor of 0.4 compared to the temperature T ₁ from the first example of application and thus to scale the entire frequency dependence by the same factor, the following relationship arises:

Typical activation energies are in the range of 10-100 kJ / mol and can be adjusted by suitable methods of preparation of the polymer used. Polishing processes like the one in 4 characterized run off at room temperature and above. Depending on the activation energy, a temperature reduction of 10 K to 30 K can therefore be used to achieve scaling of the frequency dependence by the factor 0.4 and above assumed here.

Therefore, in this application example, the polishing pad carrier is cooled by means of a suitable temperature control to a temperature which is about 10 K to 30 K below the process temperature from the first application example. This can be done for example by liquid cooling with a suitable coolant, as in 10 is indicated. A coolant flows over the inflow ( 55 ), a pipeline ( 52 ) inside the polishing pad carrier and the drain ( 56 ) through the polishing pad carrier. Other cooling devices such as with Peltier elements are possible.

In order to you reach again despite the larger structure widths and the lower relative velocity that is due to the microstructuring induced deformation frequencies in the range are in the dynamic modulus of elasticity the desired high Takes values.

In Another application example is the inventive Process for the preparation of a microlens matrix used. In contrast to the previous application examples, here the shape of the optically active surface directly through the suitable application of the method according to the invention generated.

each Single lens should have a diameter in the range of 1 mm and the arrangement is intended, for example, in a square matrix take place, wherein the distance of the lenses identical to their diameter be. The lens profile should be spherical with a spherical radius of 500 mm. This results in a lens height of 1 micron results.

you starts from a planar substrate of suitable material (e.g. Quartz glass or silicon). In this substrate is by photolithographic Microstructuring embossed a suitable microstructure. The Calculation of the microstructure may be analogous to that in the first Application example, where as the target geometry in this example the surface of the microlens matrix is used. By the subsequent polishing step is again the short-wave Microstructure leveled, but the long-wave components of the applied microstructure are retained and form the desired spherical microlenses.

In order to achieve smaller ball radii, larger etching depths can be used in this application, or several process cycles, if appropriate with different process parameters (in particular polishing pad roughness), can be used. In addition, to calculate the microstructure, it may be necessary to work with process models that take into account the higher orders of the Effective Spring Law into account. It is also advantageous to use a softer elastic intermediate layer between the upper polishing pad and the polishing pad carrier to allow the polishing pad surface to adapt to the long range height differences.

In another application example, we use the dynamic elastic modulus and other process parameter process model to place dummy patterns to enhance the planarity of a dielectric isolation layer between two metal layers of an integrated circuit. We assume that the process model by a simple convolution with the in 4 That is, the height differences in the dielectric layer are small compared to the roughness of the polishing pad.

First becomes an approximation of the trace layout of the lower metal layer for the surface geometry of the workpiece surface calculated after the deposition of the dielectric layer. In the simplest Case, this is done by having all surfaces on those a trace is drawn, the height of the metal layer thickness is assigned, and all other areas to the height set to zero. It is more advantageous at this point more accurate Use models for the deposition process. Then it will be the surface is divided into small plots, and in each Parcel calculated the mean height. The parcel size is advantageously chosen so that they clearly is smaller than the typical radius of the convolution function of the process model.

With the thus determined surface geometry and the process model the surface geometry is determined after the polishing process. For the case assumed here, this takes place advantageously by multiplication of the Fourier-transformed surface geometry with the Fourier-transformed folding function of the process model.

In order to achieve the smoothest possible surface geometry of the workpiece surface after polishing, one now looks for a height correction function by which the original surface geometry must be corrected. Due to the low-pass behavior of the process model, such a height correction function can be found by taking into account only the long-wave components up to a certain spatial frequency from the original surface geometry and setting all shorter-wave contributions to zero. The spatial cutoff frequency is determined from the convolution function such that the area below the cutoff frequency has a very large proportion (eg 95%) of the total integral of the convolution function. The location of a corresponding selected cutoff frequency is exemplary in 4 by ( 66 ) and designated there by ω _thr . By inverse transformation of the height correction function thus determined into the spatial space, the required height is obtained at each location by which the mean height must be changed by adding corresponding filling structures. Since the altitude correction function is different from zero only up to the spatial limit frequency ω _thr , the determined altitude correction applies to a larger area than the originally selected pitch size. This accommodates the expansion of the convolution function. The larger this area is, the larger the area that can be placed on the fill structures and the easier it is to achieve the height correction.

in the Generally, this procedure will be both positive and negative Determine height corrections. As usually only adding additional filling structures is allowed, this results in the problem that only positive Altitude corrections (i.e., an increase in the mean local Height) are possible. This problem leaves solve themselves by the fact that the altitude correction function is extended by a constant offset, which is chosen as large will that the negative altitude corrections into positive values being transformed. Because it is a constant, so location independent Offset, it changes the surface geometry not after polishing.

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• - DE 10113599 A1 [0034, 0035]
• DE 000010065380 B4 [0098]

Cited non-patent literature

• FW Preston, "The theory and design of plate glass polishing machines," J. Soc. Glass Technol. 11, 214-256 (1927) [0019]
• - JV Duffy, GF Lee, JD Lee, and B. Hartmann, in RD Corsaro, and LH Sperling, eds., Sound and Vibration Damping with Polymers, (ACS Symposium Series 424), ACS Press, Washington, DC, 1990, pp , 281-300 [0041]
• - CL Elmufdi and GP Muldowney, "A novel optical technique to measure pad-wafer contact area in chemical mechanical planarization," in Mater. Res. Soc. Symp. Proc., TY Tsui, Y.-C. Joo, L. Michaelson, M. Lane, and AA Volinsky, Eds., Vol. 914, 2006, pp. paper 0914-F12-06 [0053]
• J. Vlassak, "A model for chemical-mechanical polishing of a material surface based on contact mechanics," Journal of the Mechanics and Physics of solids, vol. 52, pp. 847-873, 2004 [0054]
• - "Laser lithography on non-planar surfaces" by D. Radtke and UD Zeitner in Optics Express, Vol. 15, Issue 3, pp. 1167-1174 [0070]
• - WM Davis and JP Szabo, "Group contribution analysis applied to the Havriliak-Negami model for polyurethanes", Corp. And theor. Polymer Science, 11 (2001), pp 9-15 [0084]

Claims (15)

Method for polishing and shaping workpiece surfaces ( fourteen ) comprising the steps: - determining the dimensional deviations from the desired geometry - placing an elastic surface ( 19 ) on the workpiece surface ( fourteen ) - polishing the workpiece surface ( fourteen ) by generating a pressure of the elastic surface ( 19 ) on the workpiece surface ( fourteen ) and moving the elastic surface ( 19 ) and the workpiece surface ( fourteen ) relative to each other characterized by the use of a material for the elastic surface ( 19 ) whose modulus of elasticity has a dynamic dependence in a certain frequency range such that the force necessary to produce a time-varying deformation in the material is greater for rapid deformations than for slow deformations. The polishing method according to claim 1, comprising the step Structuring the workpiece surface by means of a microstructuring method before the polishing step, in such a way that long-wave deviations from the desired geometry under acceptance a defined short-wave surface topography reduced become. Polishing method according to one of the claims 1 to 2 with the step selecting the process conditions, in particular the temperature and the amount of relative velocity between the workpiece surface and the elastic surface, such that the short-wave deviations generated by the microstructuring essential be leveled faster than uncorrected long-wave deviations. Polishing method according to one of the claims 1 to 3 wherein the method steps are carried out several times in succession become. Polishing method according to one of the claims 2 to 4, characterized in that the structuring of the workpiece surface with the help of a photolithographic method under assistance a photomask is made. Polishing method according to one of the claims 2 to 4, characterized in that the structuring of the workpiece surface using a photolithographic process using a computer-controlled laser beam is made. Polishing method according to one of the claims 2 to 4, characterized in that the structuring of the workpiece surface with the aid of lithographic processes known from printing technology for patterning a resist mask and a subsequent etching step For molding the Lackmaskengeometrie in the workpiece surface is made. Polishing method according to one of the claims 2 to 4, characterized in that the structuring of the workpiece surface made with the help of a computer-controlled laser ablation process becomes. Polishing method according to one of the claims 2 to 4, characterized in that the structuring of the workpiece surface with the help of a computer-controlled mechanical removal process made by mechanical scribing with suitably hard tips becomes. Polishing method according to one of the claims 2 to 4, characterized in that the microstructure already by a suitable process management of formative preliminary processes is produced. Polishing method according to one of the claims 1 to 3, characterized in that the surface structure, by a placement of components and the manufacturing processes is generated by a placement of other non-functional Components is influenced so that long-wave deviations from the nominal geometry at the expense of short-wave deviations be reduced. Polishing method according to one of the claims 1 to 11 characterized in that from the process and material properties a process model for the prediction of surface geometry is derived after the polishing process. Polishing method according to claim 12, characterized in that the predicted from the process model surface geometry is used, the aberration caused by a deviation from the Target geometry arises to correct by shaping the workpiece itself or other optical elements. Polishing method according to claim 12, characterized that predicted from the process model surface geometry is used, the aberration caused by a deviation from the desired geometry, by adjusting the geometry of components to compensate, which with a subsequent lithography process be structured. Apparatus for carrying out the method according to one of claims 1 to 10, comprising a wafer carrier ( 13 ) for carrying the workpiece and a polishing pad carrier ( 15 ) for carrying the elastic surface, wherein the polishing of the workpiece surface is effected by generating a pressure of the elastic surface on the workpiece surface and moving the elastic surface and the workpiece surface relative to each other by a suitable mechanism and a device, which microstructures by one of in claim 5 to 8 described method can produce.

Images