The present patent application claims the priority of the German patent application DE to 2020208883.7 entitled "Verfahren, Vorrichtung und Computerprogramm zur Reparatur einer Maske fur die Lithographie", which was filed on 16 July 2020 at the German Patent and Trade Mark Office. The entirety of the German patent application DE 102020208883.7 is incorporated in the present patent application by reference.

1. Technical field

The present invention relates to a method, a device and a computer program for repair ing a mask for lithography, in particular a mask for EUV lithography.

2. Prior art

As a consequence of the constantly increasing integration density in microelectronics, lithographic masks have to image structure elements that are becoming ever smaller into a photoresist layer (often referred to simply as " photoresist ") of a wafer. In order to meet these requirements, the exposure wavelength is being shifted to ever shorter wavelengths. Currently, it is predominantly argon fluoride excimer lasers that are used for exposure purposes; these emit light at a wavelength of 193 nm, i.e., in the so-called deep ultraviolet (DUV) range of the spectrum. Intensive work is being done with re gards to light sources that emit in the extreme ultraviolet (EUV) wavelength range (i.e., ranging from 10 nm to 15 nm), and corresponding EUV masks.

However, on account of the ever decreasing dimensions of the structure elements, lith ographic masks (often simply referred to as "masks” or else "reticles”) cannot always be produced without defects that are printable or visible on a wafer. Owing to the costly production of masks, defective masks are repaired whenever possible. Two important groups of defects of lithographic masks are, firstly, dark defects and, secondly, clear defects.

Dark defects are locations at which absorber or phase shifting material is present, but which should be free of this material. These defects are repaired by removing the excess material preferably with the aid of a local etching process.

By contrast, clear defects are defects on the mask which, upon optical exposure in, e.g., a wafer stepper or wafer scanner, have a greater light transmissivity than an identical defect-free reference position. In mask repair processes, such clear defects can be elimi nated by depositing a material having suitable optical properties. Ideally, the optical properties of the material used for the repair should correspond to those of the ab sorber or phase shifting material of the mask.

In addition to defect-free masks, the exposure parameters during the exposure of the wafer with the mask also have an influence on the result of the exposure process and hence ultimately have an influence on the component produced, and an interplay may be present between the influence of the exposure parameters on the exposure process on the one hand and, on the other hand, the repair measures which are necessary to at tain acceptable results during the subsequent exposure.

By way of example, US 9,261,775 B2 by the applicant describes a method in which an aerial image stack of a mask or a mask feature is recorded, from which a so-called Bos- sung plot is then ascertained. The Bossung plot can be used to determine at least one parameter which characterizes the mask. A repair parameter which is transmitted to a repair system that determines the process required for the mask repair on the basis of the repair parameter can also be determined from the Bossung plot.

However, a disadvantage here is that only the critical dimension (CD) of the mask fea tures to be examined/ repaired is used as a characteristic. In the case of increasing inte gration depth and modern exposure systems, in particular in the case of EUV masks and corresponding exposure systems, this may be insufficient or problematic since it is only possible to capture changes in the relative distance, which may not sufficiently take account of the complexity of modern mask and exposure methods.

The present invention is therefore based on the object of specifying a method which ac counts for the ongoing advances in respect of integration density and novel exposure methods and, in particular, provides improved options of mask repair. Furthermore, a corresponding device and a computer program with instructions for carrying out such a method should be provided.

3. Summary of the invention

The aforementioned objects are at least partly achieved by the various aspects of the present invention, as described below.

In one embodiment, a method of repairing a mask for lithography, in particular a mask for EUV lithography, comprises the following steps: (la.) ascertaining a first value of an edge position of a structure to be repaired on the mask, (lb.) cariying out a repair pro cess, comprising at least one first repair step on the mask, (tc.) ascertaining a second value of the edge position after carrying out the first repair step and (id.) comparing the second value of the edge position with the first value of the edge position and/or with a target value for the edge position.

The value of the edge position is always denoted by the variable x below and the target value thereof is specified as x ₀ . Where necessaiy, the first and second ascertained value of the edge position is denoted by x and x ₂ , etc.

As mentioned at the outset, modern masks, in particular masks for EUV lithography, are distinguished by an ever greater structure density and, accompanying this, ever smaller structures. Moreover, using light of ever shorter wavelength leads to even fur ther complications which did not need to be taken into account, or at least did not need to be taken into account to the same extent, in conventional masks for longer wave lengths. Firstly, masks for wavelengths at the lower end of the wavelength range currently ac cessible by technology, in particular masks for EUV lithography, are not operated in transmission but are mirroring or reflective optical elements in the current state-of- the-art. This is accompanied by the chief ray angle (CRA) of the light beam exposing the mask generally not being equal to zero (by way of example, 6° is a typical value), and so the incident light cone is separated from the reflected light cone. A chief ray an gle not equal to zero may cause distortion effects or shadowing effects at structure edges of the mask, to name but two possible effects that have an influence on the imag ing performance of the mask.

Further, the mask thickness is typically no longer small in relation to the exposure wavelength employed. On the contrary, an EUV mask for example typically consists of many layers which are layered on top of one another and which have layer thicknesses that are smaller than the wavelength of the actinic EUV radiation. This means that the depth at which the radiation interacts with the mask may vary greatly in units of the ex posure wavelength, leading to so-called 3D effects. By way of example, the aforemen tioned shadowing effect of a structure edge may lead to the maximum interaction depth in the shadowed region behind such an edge varying with the distance from the edge, which may have an effect on the light cast back by the mask and hence on the imaging properties of the latter.

The ever smaller dimensions of the structures also allow or require the alignment of the individual mask structures to generally vary over the mask. Thus, for example, some structures may be arranged horizontally, and other structures may be arranged verti cally or diagonally, etc.

Moreover, the "surroundings" of one structure at one location on the mask may differ significantly from those of another structure at another location on the mask and this generally also leads to different properties of the mask at different portions of the mask since diffraction effects and interference effects are able to influence the imaging prop erties of the mask. Thus, in general, one could say that both the rotational symmetry and the translational symmetry of the mask are broken by some or all of the above-described effects in the case of masks for very short exposure wavelengths, in particular EUV masks.

In respect of a successful repair of the mask, this means that it may no longer be suffi cient, or may be disadvantageous in any case, to only ascertain the relative distance be tween two mask features/structures and take this into account as characteristic for the mask repair. This is a starting point for the method described, by virtue of using the (absolute) edge position x along a structure edge of the mask itself as characteristic. By using the edge position x itself and not the relative distance between two edges (i.e., a critical dimension, CD), it is possible to better monitor and trace the above-described error sources and distortion, shadowing and 3-D effects, etc., as a result of which the quality of the mask repair can be improved.

A person skilled in the art has various options available for ascertaining the edge posi tion before or after the first repair step, all of which can be used in the described method as a matter of principle. One option is that of recording an aerial image or an aerial image stack, for example with the aid of a mask inspection system (e.g., a system by the applicant suitable to this end). By way of example, such a mask inspection sys tem can be substantially embodied like an optical exposure system of a lithography ap paratus, which has a CCD camera instead of a wafer in order thus to measure the aerial image generated by the mask. Further options lie in scanning the mask using a scan ning probe microscope/atomic force microscope (AFM) or a scanning particle micro scope (SEM). The actual exposure of a "test wafer" can also be considered. Further de tails in relation to these options are found, for example, in the document DE 102011 079382 At by the applicant, which is accordingly referred to in the context of analysing a mask defect.

In this case, the (absolute) edge position x can be determined with respect to a suitably defined coordinate system. By way of example, a coordinate system defined by the sen sor of the CCD camera lends itself to this end when using an aforementioned mask in spection system. A coordinate system defined in analogous fashion will be evident to a person skilled in the art when use is made of a different analysis process and/or a dif ferent analysis tool. A coordinate system inherent to the mask itself would also be con ceivable.

To monitor the progress of the repair process, the edge position x is, according to the method described, ascertained at least once before the first repair step is carried out and at least once thereafter in order thus to facilitate a comparison of the two values x and x ₂ among themselves and/ or a comparison with an (absolute) target value x ₀ of the edge position. Naturally, further measurements of the value of the edge position x are also possible and/or the repair process can comprise more than one repair step. Byway of example, the repair process may run in iterative fashion, wherein the value x of the edge position is newly ascertained in each case after carrying out a certain number of repair steps (e.g., after every step, after every second step, etc., or after a varying num ber of steps) and can be compared to one or more preceding values and/or to the target value Xo

By way of example, the target value x ₀ can be specified by a mask design determined in advance, according to which the mask was produced and to which the mask is now compared. By way of example, the mask design can also define an acceptable deviation of the edge position, within which a non-perfect edge may nevertheless still be able to be considered acceptable.

The repair step or steps themselves may comprise an etching process and/or a deposi tion process, for example, as are known per se to a person skilled in the art in relation to the repair of masks, in particular the repair of EUV masks. In this context, reference is made byway of example to documents DE 10338019 At and DE 102013203995 At, the teaching of which in respect of etching or deposition processes for processing and repair of masks is herewith incorporated in the present disclosure.

By monitoring the edge position x (instead of, e.g., a critical dimension CD), the repair process of the mask can consequently be tracked particularly accurately, even and par ticularly in the case of EUV masks which are subject to the aforementioned shadowing, distortion and 3-D effects, etc., and whose rotational and translational symmetry is generally broken. By way of example, the repair of the defect considered can be considered complete if the current value of the (absolute) edge position x is close enough to the (absolute) tar get value _o for example within a specified acceptance value D, i.e., if for example |x ₂ - x ₀ | < D (as already mentioned, this acceptance value D may have been specified to gether with the mask design, for example).

As an alternative or in addition thereto, it is possible to assess whether the value of the edge position x before and after carrying out a repair step has changed by no more than a specified minimum degree of change d, i.e., whether for example | x ₂ - X | < 5, in or der to decide whether new repair steps are necessary or productive. By way of example, if the edge position is determined to have changed by no more than d despite carrying out a further repair step, this may be considered a sign that a further repair step - at least a further repair step relating to the same repair mechanism (e.g., etching, material deposition) - will no longer be expedient.

The method can further comprise the following steps: (2a.) ascertaining a focus-expo- sure matrix for the edge position in respect of an exposure system used for exposure purposes, (2b.) ascertaining a deviation of the edge position from the target value as a function of the focus depth and the exposure dose from the focus-exposure matrix and (2c.) providing the quantity ascertained in step 2b as a quality metric in respect of the repair process.

An important measure relating to the quality of the repair of an edge of a mask struc ture can be the stability of the edge position in respect of (intended or inadvertent) var iations of the exposure parameter with which an exposure system serving to expose a wafer with the mask operates. Important exposure parameters of such an exposure sys tem used to expose a wafer are the focus depth (e.g., measured as a position of a focal plane relative to the wafer surface or to the surface of the photoresist on the wafer) and the exposure dose, which is substantially equivalent to the intensity of the radiation employed. The exposure dose is moreover directly related to the so-called "threshold" and can be converted into the latter. The "threshold" denotes the illuminance on the part of the wafer at which the photoresist just reacts or just does not react. Below, the focus depth is always denoted by the variable z and the exposure dose is de noted by the variable t. The deviation of the edge position x from the target value x ₀ is denoted by dx.

Consequently, the focus-exposure matrix for the edge position x in respect of an expo sure system used for exposure purposes can be understood to be a set of data or meas urement values which specifies the (absolute) edge position x for a set of measurement points in the variables (z, t ), wherein the measurement points will typically be discrete points, i.e., the focus depth z and the exposure dose twill be altered incrementally. By way of example, a mask inspection system can be used to ascertain such a data record, for example a system commercially available from the applicant, by means of which aerial images are recorded with varying focus depth z and varying exposure dose t. From the data obtained thus, it is then possible if desired to obtain a continuous and even a differentiable data record in a manner known per se to a person skilled in the art, for example byway of a suitable interpolation between the discrete data points. Ac cordingly, the ascertained deviation dx, for example, can be a data point, a set of dis crete data points or else a continuous and/or even differentiable data record.

In general, the target will be to bring the edge position as close as possible to the target value X _o following the repair, i.e., to minimize the deviation dx.

In general, the deviation dx may have both a constant component dx and a component that varies with focus depth z and the exposure dose t.

By way of example, the constant component dx can be defined or considered to be a de viation of the edge position from the target value x ₀ in the case of the nominal focus depth Z _o and nominal exposure dose t ₀ , i.e., dx = x | ( _{Z0, to)} - x _{0 º} x - x ₀ .

As described immediately, the non-constant component of the deviation dx can be of particular interest since it may contain further valuable information which can be used as additional quality metrics or "set screws" for monitoring and controlling the repair process. However, it is also possible to monitor and control the repair using only the deviation dx as single quality metric or "set screw", with reference once again being made in this context to the above-described advantages brought by the use of the (abso lute) edge position x in comparison with the use of a critical dimension CD.

The method can further comprise the following steps: (3a.) ascertaining a dependence of the edge position on the focus depth as a function of the focus depth and the expo sure dose from the focus-exposure matrix, (3b.) ascertaining a dependence of the edge position on the exposure dose as a function of the focus depth and the exposure dose from the focus-exposure matrix and (3c.) providing the quantities ascertained in steps 3a and 3b as further quality metrics in respect of the repair process.

This embodiment is based on the discovery that it is not only the offset dx of the edge position from a desired position x ₀ itself that might find use to monitor and control the repair process but that further relevant information about the quality and the design of the repaired edge or edge to be repaired may be contained in the dependencies and, in particular, in the rates of change with which the edge position x will react to variations in the exposure variables, in particular variations in the values of z and/or t.

Thus, it was found that the dependence of the edge position on the focus depth z or the exposure dose t can quantify and describe different and substantially independent properties of the edge. This in turn was found to be advantageous because this, as still described below, may allow influencing of the one or the other aspect of the edge design in a targeted fashion by varying (largely) independent process parameters for the repair process. Moreover, this might render it possible to find those errors or deviation in the edge form which play a particularly critical role in the subsequent exposure process or, conversely, ascertain the aspects in respect of which the requirement profile/ac ceptance limits might be able to be slightly relaxed but still allow an acceptable expo sure result to be obtained.

Metaphorically speaking, the inclusion of the dependencies or the rates of change of the edge position in respect of the focus depth z and the exposure dose t increases the num ber of "set screws" which are available to monitor and control the advance of the repair process such that the quality of the result of the repair process can be increased. Even though the rates of change (e.g., within the first partial derivatives with respect to z or t, as described below) of the deviation of the edge position x from the target value o with variation the focus depth z or exposure dose t can therefore represent an im portant set of information or quality metrics in respect of monitoring and controlling the repair process, it should already be mentioned here that the dependencies of the edge position on the focus depth z or the exposure dose t can also have a more compli cated form and can for example also comprise higher partial derivatives, as likewise de scribed below. Thus, by considering such further variables, it is possible to further in crease the number of quality metrics and consequently also the number of "set screws" with which the repair process can be monitored and controlled.

The method can further comprise the following steps: (4a.) adjusting a set of process parameters for carrying out the first and/or a second repair step of the repair process in order to improve at least one of the quality metrics and (4b.) carrying out the first and/or second repair step using the adjusted process parameters.

In general, attempts will be made in the process to improve all quality metrics, wherein it may be appropriate to consider the duration and the outlay of the repair process on the one hand and the further improvements achieved therewith on the other hand. However, it may also be possible to concentrate on a single quality metric or a subset of the available quality metrics when adjusting the process parameters, for example if ex perimental and / or theoretical results yield that these have a particularly pronounced or critical influence on the ultimate result of the lithography process.

Since, as already described above, the repair process may comprise only the first repair step or else a plurality of repair steps, in particular an iterative sequence of repair steps, this means that the quality metrics can be considered at any point of the repair process and that the process parameters can be adjusted accordingly. However, it would also be conceivable to not consider the quality metrics and adjust the process parameters after each step but only after every second step or after every third step or whenever this may appear necessary. Should the considered quality metric or quality metrics have (signifi cantly) improved in a process step, the subsequent process step for example can also be implemented using the same process parameters and the process parameters might only be adjusted further when "nothing happens" in terms of the quality metrics (e.g., measured in relation to a suitably chosen threshold).

The dependence of the edge position x on the focus depth z might comprise the first partial derivative of the edge position x with respect to the focus depth z. The depend ence of the edge position x on the exposure dose t might comprise the first partial deriv ative of the edge position x with respect to the exposure dose t.

Below, the partial derivatives of the edge position x with respect to one or more varia bles a, b, ... are alternatively presented as follows: dx _ dx _

ΊGa ^{= %a} ’ db ^{= %b} etc. (Eq. l)

The two options specified here therefore represent, possibly in conjunction with the constant component dx already described above, which may also be zero, an expansion of the edge position after the first order in the exposure parameters: dx = dx + ^ dz dz + dt dt º dx + x _z dz + x _t d t (Eq. 3) where even only one of the two first partial derivatives x _z or x ₍ can be used in the method if desired or if required to obtain the desired results.

In this case, the deviations dz and dt are measured from a certain reference point (z ₀ , t _o ), i.e., dz = z - Z _o and dt = t - t ₀ . By way of example, this reference point can be the in tended/nominal work point of the exposure system, with which a wafer should subse quently be exposed with the aid of the mask.

What should be highlighted here is that the relative position of the reference point (z ₀ , t _o ) can itself be considered to be a variable within the scope of the method described herein, i.e., the operations and analysis steps described herein can be implemented once at a first work point (z ₀ ⁽¹⁾ , t ₀ ⁽¹⁾ ) and then at another work point (z ₀ ⁽²⁾ , t ₀ ⁽²⁾ ), for ex ample to ascertain (e.g., within the scope of a simulation of the actual repair carried out before the latter) the work point in respect of which the repair would be carried out more easily or better, or the work point in respect of which the repaired mask would be more stable in respect of variations of the parameters z and t.

At least in the respective vicinity of a certain nominal work point (z ₀ , t ₀ ) of the exposure system, the first partial derivatives represent a measure for the rate of change of the edge position x with the variation of the focus depth z or exposure dose t and conse quently specify the "sensitivity" of the edge position x in relation to the (desired or un desired) variations or deviations of these parameters from the respective nominal/tar get value Z _o or t ₀ -

These rates of change maybe very important for the mask repair because, as described imminently, they directly correspond to relevant physical properties of the edge or edge structure, which consequently can be monitored and "controlled" in a targeted fashion within the scope of the repair process.

Thus, the dependence of the edge position x on the focus depth z, in particular the rate of change thereof within the meaning of the first partial derivative x _z , can represent a quality metric in the form of a measure for a phase error and/or telecentricity error.

In the case of a repair process comprising an etching and/ or deposition process, it is then possible to influence the quality metric in respect of the phase error/telecentricity error in a targeted fashion by adjusting an endpoint of the etching and/or deposition process (e.g., the depth at which the process is ended). By way of example, the quality metric can be optimized (as far as technically possible) by virtue of the etching/deposi tion process being controlled in such a way that the dark/clear defect is removed as far as possible without inadvertently altering the surrounding mask substrate in the pro cess or adversely affecting the latter (e.g., by inadvertent etching away of the substrate near the edge to be repaired or an inadvertent material deposition on the substrate near the edge).

The dependence of the edge position x on the exposure dose t, in particular the rate of change thereof within the meaning of the first partial derivative x ₍ , can represent a quality metric in the form of a measure for an edge slope. The quality metric in respect of the edge slope can be influenced in a targeted manner by adjusting a drift and/or jitter movement during the repair process. In general, the edge slope is increased by minimizing the drift and/or jitter movement, i.e., the corre sponding quality metric is increased. However, this by no means precludes the method for example being operated in such a way that an edge is configured to be flatter, e.g. by intentionally brought about jitter movements, provided this is desired or necessary to obtain a certain repair success.

The method can moreover comprise the following steps: (na.) ascertaining at least one second or higher-order partial derivative of the edge position with respect to the focus depth and/or the exposure dose from the focus-exposure matrix and (nb.) providing the at least one second or higher order partial derivative as a further quality metric in respect of the repair process.

As already mentioned, the dependences of the edge position x on the focus depth z and/or the exposure dose t can also be considered beyond the first partial derivatives, and this allows the derivation of further relative quality metrics or "setting levers" for the repair process.

In particular, the following second or higher-order partial derivatives maybe relevant here to monitoring or controlling the mask repair process:

It is noted here that, at least after suitable preparation of the data of the focus-exposure matrix, it is generally possible to assume that all these derivatives are continuous and even differentiable in turn, and so the sequence of partial derivatives is unimportant ac cording to Schwarz's theorem. Further second or third order partial derivatives possible as a matter of principle were found to be less relevant in the context of the method de scribed here and are therefore not discussed any further below. Naturally, the inclusion thereof is also possible as a matter of principle.

In particular, it is possible for the quality metrics to contain a quality metric in the form of a measure for a depth of focus and for this quality metric to be proportional to x _(zz .

Building on Eq. (3) above, it is possible to specify an advantageous model for describing and controlling the mask repair process within the scope of the method described herein as follows: dx = dx + tee tee

(Eq. 3’) where the last step used the approximation lyl < ¹ (Eq· 7) and the three parameters "tee", "sm" and "zw" were introduced (for clarity and for dis tinguishing these from variables, such parameters are specified below in bold and non cursive font). The combination denoted below as parameter "sme"

1

- sm * zw - º sme (Eq. 8) in this case occurs as a factor of the (dt dz dz)-term and represents a quality metric in the form of a measure for the edge quality. In turn, this has an influence on the size of the range of the depth of focus for which the edge supplies an acceptable image on the wafer and it contains information going beyond that contained in the x ₍ -term. There fore, its inclusion in the present method once again increases the number of parameters or "set screws" that are available for monitoring and controlling the repair process.

The method described herein can further comprise the step of evaluating at least one quality metric at a nominal focus depth z ₀ and a nominal exposure dose t ₀ in order to obtain at least one quality score in respect of the repair process. Expressed differently, to obtain numerical values that can be compared to, e.g., one or more thresholds or acceptance intervals, it is possible to evaluate one or more of the quality metrics (precisely those that should be under current consideration) at a tar geted work point of the exposure device for the subsequent exposure of a wafer with the mask, in each case resulting in a certain numerical value or set of numerical values which are referred to herein as quality score(s). As already indicated above, it is also possible to consider a plurality of work points, e.g. (z ₀ ⁽¹⁾ , t ₀ ⁽¹⁾ ) and (z ₀ ⁽²⁾ , t ₀ ⁽²⁾ ), in the pro cess (e.g., within the scope of a simulation or the like carried out prior to the repair) in order to ascertain whether and, if so, which work point would be better suited.

In line with the embodiments up to here, it will generally be the object of the repair process to process the mask in such a way that this quality score or quality scores come as close as possible to, or at least within a certain acceptance value of, a respective nu merical value. In this case, the acceptance interval may vary by all means for different quality scores. Byway of example, a quality score corresponding to a "critical" quality metric (e.g., within the meaning of a deviation/degradation in respect of this metric would mean a significant error during the exposure of a wafer) could have a stricter measure applied than a quality score that tends to be "non-critical".

In this context, it is emphasized yet again that it is precisely the inclusion of a plurality of quality metrics within the scope of the method described that may allow individual requirement specifications and acceptance intervals to be relaxed since this allows dif ferent error contributions to be (largely) monitored separately from one another and to be also influenced separately from one another If only, e.g., a critical dimension CD is considered as a single characteristic, it will always only be possible to see and track the "interaction" of all error contributions.

Nevertheless, the present invention can also comprise the step of ascertaining a critical dimension (CD) of the structure to be repaired and including the critical dimension in a quantification of the effect of the repair process. Further, it is emphasized that, in general, the edge position x in all what was stated pre viously can always be replaced by a critical dimension CD and that this is also com prised by the present disclosure. However, this will generally be linked to the loss of some of the advantages mentioned at the outset which arise from the use of the (abso lute) edge position x as a characteristic.

At this point, it is further mentioned that although the possible features, options and modification options of the disclosed method were described in a certain sequence herein, this should not necessarily express a certain dependence of the features among themselves - unless this was explicitly presented so or arises from technical, physical and/or mathematical necessity. Rather, the various features and options can also be combined in other sequences and permutations - to the extent this is possible from a physical and technical point of view - and such combinations of features or even sub features are also disclosed by the present disclosure. Individual features or sub-features can also be omitted provided they are not required to obtain the desired technical re sult.

In one embodiment, a device for repairing the lithographic mask contains means for carrying out the steps of an embodiment of the method described herein.

By way of example, such a device can be constructed on measurement apparatuses for analysing lithographic masks and/ or repair devices for lithographic masks, as are made commercially available by the applicant for this purpose. In particular, the device can comprise the following in the process: (17a.) means for carrying out a repair process comprising one or more repair steps on a structure to be repaired of the mask using a set of process parameters and (17b.) means for obtaining at least one quality metric in respect of the repair process.

By way of example, the means 17a and 17b can be realized by the just mentioned de vices or combinations of such devices, as are commercially available by the applicant. The means 17b for obtaining at least one quality metric can determine the quality met- ric(s) themselves, for example, or they can obtain these from an appropriate analysis or evaluation device. However, the device can additionally also comprise means for automatically adjusting the process parameters in order thus to improve the at least one quality metric. Ex pressed differently, the device can be embodied to control the repair process without user input, in such a way that the mask repair leads to the desired result. This is facili tated by virtue of the method described herein and the "setting levers" for the repair process available as a result permitting an (at least largely) automated implementation. The control means necessary to this end can be implemented in software on the one hand. However, it is also possible for a hardware-type control module to be provided to this end. The latter can be disposed, for example, between firstly the means 17a for car rying out the repair step or steps and secondly the means 17b for ascertaining the at least one quality metric and can undertake the necessary conversions and analyses in order to automatically carry out the repair process.

By way of example, the device, in particular the just mentioned control module, can au tomatically adjust a process parameter which influences an endpoint of an etching and/or deposition process so as to improve a quality metric in respect of a phase error and/or a telecentricity error. Or the apparatus of the control module can automatically adjust a process parameter which influences a drift and/or jitter movement during the repair process so as to thus improve the quality metric in respect of an edge slope.

Finally, a computer program can contain instructions which, upon execution, prompt a mask repair device to carry out the steps of an embodiment of the method described herein.

4. Brief description of the figures

The following detailed description describes possible embodiments of the invention, with reference being made to the figures, wherein fig. 1 shows a 3-D effect/shadowing effect which may occur on a mask when

EUV radiation is used and which can influence the imaging properties thereof; figs 2a-c show problems that might occur in respect of the selection of a global isofocal point when using EUV radiation; fig. 3 shows various possible edge errors and the superposition thereof; fig. 4 shows the basis of a 3rd order model for defining suitable quality metrics for monitoring a repair process of an (EUV) mask; figs sa-d show a comparison of the prediction of the 3rd order model with a nu merical simulation of the imaging properties of a mask; and fig. 6 illustrates further options for application of the discussed quality metrics when repairing and analysing mask defects.

5. Detailed description of possible embodiments

Below, embodiments of the present invention are predominantly described with refer ence to repairing a mask for EUV lithography. However, the invention is not restricted thereto and can also be used for other types of masks and mask processing.

Further, reference is made to the fact that only individual embodiments of the inven tion can be described in more detail below. However, a person skilled in the art will ap preciate that the features and modification options described in conjunction with these embodiments can also be modified further and/ or can be combined with one another in other combinations or sub-combination without this leading away from the scope of the present invention. Moreover, individual features or sub-features can also be omitted provided they are dispensable in respect of achieving the desired result. In order to avoid unnecessaiy repetition, reference is therefore made to the remarks and explana tions in the preceding sections, which also retain their validity for the detailed descrip tion which now follows below.

Figure 1 shows an excerpt 100 of an EUV mask 110 exposed with EUV radiation 120, said EUV mask comprising an absorber edge 112 and a substrate 115. The substrate 115 is present as a multilayer structure, as a person skilled in the art knows as a matter of principle for the construction of EUV masks, and, as a matter of principle, the EUV radiation 120 incident on the EUV mask 110 interacts with the substrate 115 at those points where it is not covered by absorber material 112.

Since EUV masks such as the mask 110 shown here are not operated in transmission but are embodied as a reflecting optical component, the chief ray angle Q (CRA), cf. ref erence sign 122, is generally not equal to o°, as also shown in figure 1. This may lead to the following 3-D/shadowing effects: Provided the absorber edge 112 has a sharp embodiment as shown in figure 1, there is a shadow region directly behind the edge 112 in the "inclination direction" of the incident light 120, the width of said shadow re gion in relation to the edge plane 140 depending on the interaction depth of the radia tion 120 with the layers of the substrate 115. Thus, a point of interaction 135 with the uppermost substrate layer as seen from the direction of incidence may be located closer to the edge plane 140 than a point of interaction 136 in the centre of the substrate 115. In turn, the latter can be located closer to the edge plane 140 than a point of interaction 137 with the deepest layer of the substrate 115 as seen from the direction of incidence. This influences the beams 125, 126 and 127 of the EUV radiation respectively reflected at these points of interaction 135, 136 and 137, or the superposition of said beams with reflected radiation from other regions and layers of the mask 110, and therefore has an influence on the imaging properties of the mask 110, in particular since the thickness of the substrate 115 will generally no longer be small in relation to the wavelength of the EUV light 120, and so interference effects between the various reflected beams, e.g., the beams 125, 126, 127, gain importance.

Figures 2a-c show further effects in respect of the imaging behaviour of modern masks, which may be accompanied by ever increasing integration densities and ever smaller exposure wavelengths.

Figure 2a shows an excerpt 200 of an intensity profile (represented by a varying den sity of symbols) depending on the mask position which results from the exposure of a largely defect-free EUV mask with a quite simple structure, specifically a mask with a sequence of vertically extending structures. Masks intended for the production of mod- ern chips or the like often have significantly more complex structuring (by way of exam ple, these masks may contain structures with different orientations, as mentioned at the outset). The details shown in figures 2b and 2c relate to the region 210 of the in tensity profile of figure 2a.

Figure 2b shows a section 201 along the horizontal axis through the centre of the re gion 210 of figure 2a, with the intensity (in relative units) being plotted on the ordi nate in figure 2b. The zero point of the abscissa, which specifies the horizontal posi tion on the mask, has been arbitrarily selected to be in the centre of the figure in figure 2b. Further, the intensity profile was curtailed at high and low values and it is only the region around the critical dose 255 that is shown (in relation to a wafer provided for exposure) since this is precisely the intensity or threshold range which will be relevant in practice during the structure formation in the photoresist of the wafer or which may occur, for example caused by variations in the luminous intensity or (slight) variations in the trigger threshold of the photoresist for polymerisation. Moreover, a whole family of curves of intensity curves for different values of the focus depth z is shown (the vari ous curves are easier to identify in figure 2c) and three regions of intersection 220 of this family of curves with the critical dose 225 are highlighted and labelled by "1", "2" and "3".

Figure 2c shows magnified excerpts 202 of the just aforementioned sectional regions. The excerpt "1" is shown right at the top by reference sign 230, the excerpt "2" is shown in the centre at reference sign 231 and the excerpt "3" is shown below at reference sign 232. Once again, the ordinate plots the intensity (in relative units) and the abscissa plots the horizontal position on the mask, wherein the zero of the abscissa has once again been arbitrarily chosen at the (averaged) point of intersection of the family of curves with the critical dose. The curves denoted by the reference sign 235 represent the curves which have the greatest defocussing (i.e., deviation from the nominal value Z _o ) of all curves of the family of curves.

As arises from the magnified excerpts 230 and 232 in particular (i.e., the sectional re gions "1" and "3"), generally, there is no longer a common point of intersection of all curves of the family of curves with the critical dose in the EUV wavelength range, even for the largely defect-free and simply structured mask considered here. In other words, it is generally no longer possible to find a global isofocal point, even for the "good" and "simple" mask considered here.

By way of example, on account of the circumstances described in conjunction with fig ure l and figures 2a-c, it is therefore highly relevant for the repair of modern masks, in particular EUV masks, not only to know and possibly correct the relative distance of the various mask structures from one another but it is also desirable to have further "set screws" and quality metrics available in order to take account of and include the complicated imaging properties of the mask on the basis and in the case of the devia tion from a selected nominal point (t ₀ , z ₀ ) in the repair process (wherein the nominal values maybe chosen globally for a mask as just described).

This is where the method according to the invention intervenes, some aspects of which should be particularly highlighted in the two subsequent figures, figures 3 and 4.

Figure 3 initially schematically shows an excerpt of an error-free mask 300, compris ing a (possibly multi-layer) substrate 305 and an absorber layer 302 which forms a clearly defined edge at the intended position 308 (which may be specified by the mask design, for example).

Three excerpts of masks 310, 320 and 330 are shown therebelow, said excerpts each having a certain error as maybe monitored and removed during a corresponding repair process using the options described herein.

The edge position 318 is displaced in comparison with the desired position 308 in the mask 310. Since the method described in the present case allows monitoring of the ab solute edge position x, such an error can be rectified, for example by etching away the excess material. In this case, the repair advance can be controlled and monitored con tinuously and/or at certain time intervals by (re-)ascertaining the edge position x. Fur ther, this is also possible under "more difficult conditions", for example if the mask no longer has a rotational and/or translational symmetry, since the absolute edge position x can be used as a characteristic in place of the otherwise utilized critical dimension CD, i.e., a relative position measure. Although the edge positions are at the correct position in the mask 320, the mask 320 has a defect 328 in the substrate 305 near the edge. By way of example, this defect 328 could have been caused by the excess material of the protruding edge 318 in the mask 310 having been etched away but the etching process having been terminated too late, and so the latter continued into the substrate 305 (in a manner analogous thereto, a certain part of the protruding edge 318 could remain on the substrate 305 if the etch ing process had been terminated too early; this would lead to a similar situation to what is described here). To avoid such an error, it is advantageous to take account of infor mation going beyond the edge position x when controlling the repair process. By way of example, the error 328 corresponds to the first partial derivative x _z of the edge position x with respect to the focus depth z, as can be obtained within the scope of the method described herein from the focus-exposure matrix in respect of the edge position x and as can be used to monitor the repair process as a quality metric or "set screw".

The edge slope 338 is not as provided or desired in the mask 330 (it should be men tioned that the drawings in figure 3 are schematic illustrations, i.e., the desired edge slope is shown as a perpendicular edge for reasons of simplicity; in reality, an oblique edge may by all means be desirable, but this case can also be covered by the method de scribed herein). The edge slope 338 (which is often described in the form of the "nor malized image log-slope”, NILS in the literature) corresponds to the first partial deriv ative X of the edge position x with respect to the exposure dose t, as can be obtained within the scope of the method described herein from the focus-exposure matrix in re spect of the edge position x and as can be used to monitor the repair process as a qual ity metric or "set screw".

These errors often occur in combination in a real mask 350, i.e., a superposition of the individual error contributions - as indicated in figure 3 by the arrow 340 - leads to a more complex mask defect 358, the repair of which without suitable control options and "setting levers" possibly being difficult. This is remedied by the quality metrics pro vided by the method described herein since these facilitate a resolution of the com posed error 358 in terms of the individual error contributions (e.g., according to error contributions which are indicated at the reference signs 318, 328 and 338) and more over allow a controlled influence on the respective contribution by way of suitable re pair measures. Thus, for example, the error contribution as indicated at 328 can be avoided or corrected by targeted termination of the etching/deposition process, which can be monitored and controlled largely independently of the other error contributions in the combined error 358 by monitoring the x _z -quality metric. As indicated at 338, the edge slope can be adjusted by adjusting the drift or jitter movement during the re pair process and can be monitored and controlled largely independently of the other er ror contributions by monitoring the x _z -quality metric.

Figure 4 shows the basis for a third order model in terms of the partial derivative with respect to focus depth z and exposure dose t, which goes beyond the options described in figure 3 and which can likewise be used to monitor and control a mask repair pro cess.

Shown is an excerpt 400 of the data of a focus-exposure matrix in respect of the edge position x (possibly after these data were "smoothed" by way of a suitable interpolation or the like), wherein the horizontal axis plots the deviation dx of the edge position x from a target value x ₀ and the vertical axis plots the deviation d t of the exposure dose t from a nominal value t ₀ Shown further are plurality of curves 410, 420, 425, 430, 435, which correspond to different values of the focus depth z. Here, the curve 410 corresponds to the nominal value z ₀ , which represents the "best" focus for the data rec ord illustrated here (at least in a local neighbourhood of the parameter point (z ₀ , t ₀ ))·

It should be noted here that hereinbelow the assumption that the value of the edge po sition x at the nominal value (z ₀ , t ₀ ) corresponds to the target value x ₀ of the edge posi tion (the nominal value (z ₀ , t ₀ ) is also referred to as "isofocal point" below) is made for the sake of simplicity. Using the nomenclature introduced at the outset, the following applies:dx = x |( _{Z0, to)} - x ₀ = o, and hence dx = x - x ₀ . If the value of the edge position at the nominal value (z ₀ , t ₀ ) does not correspond to the target value x ₀ of the edge posi tion, this deviation should be taken into account by an additive constant dx for the de viation dx.

What applies to the curve 410, which is the curve at the "best" focus z ₀ , is that it can be approximated by a tangent in a neighbourhood of the isofocal point (z ₀ , t ₀ )· The slope thereof is referred to below as "sm" (from "slope max"). That is to say, from figure 4: If the focus depth z deviates from the "best value" z ₀ , this can according to the invention be described by a correction term of the following form (to the lowest order): d£ dx Z ¹ ZO = sm * ( 1 - (dz / zw) ² ) (Eq. 10) with the correction parameter "zw" (from "z-width"). This parameter is a measure of the depth of focus (DOF) of the exposure arrangement, i.e., to what extent a deviation from the "best" focus z ₀ is noticeable in the exposure of the mask or, expressed differ ently, how large the tolerable error range is in respect of (unwanted) defocussing.

Further, a telecentricity error/phase error, described by the parameter "tee" (" telecen - tricity error”), is added to these contributions and included in the model according to the invention by the following contribution: (Eq. 11)

Solving Eq. 10 for dx and adding the contributions of Eq. to and Eq. 11 yields dx = tee * dz + [ sm * ( 1 — (dz / zw) ² )] ¹ * dt (Eq. 12) « tee * dz + (1 / sm) * ( 1 + (dz / zw) ² ) * dt (Eq. 13)

Here, the approximation 1 / (1 - y) * 1 + y + ... was used in the last step. By introducing the further parameter sme = (1 / sm) * (1 / zw) ² (Eq. 14)

Eq. 13 can be written as follows: dx « tee * dz + (1 / sm) * dt + sme * dt dz dz (Eq. 15)

In general, this equation applies in the neighbourhood of a certain isofocal point (z ₀ , t ₀ ) (and under the assumption dx = o; see above). If the position of the isofocal point itself is considered a variable and if the definitions dx = x - X _o , dz = z - Z _o and d t= t- t ₀ are inserted into Eq. 15, a model for the edge posi tion x as a function of the focus depth z and the exposure dose t arises with the three parameters x ₀ , z ₀ and t ₀ and the three independent metrics tee, sme and zw (alterna tively, e.g., tee, sme and sm, or tee, sm and zw; cf. Eq. 14, which allows a conver sion), which can be used to describe and control a repair process of a mask as described herein: x — x ₀ ~ — tee * z ₀ — sme * t ₀ * (z ₀ ² + zw ² )

+ t * sme * (z ₀ ² + zw ² )

+ z * (2 sme t ₀ * z ₀ + tee) z ² sme * t ₀

— t * z * 2 * sme * z ₀

+ t * z ² * sme (Eq. 16)

Figures 5a-d show a comparison of the model described in conjunction with figure 4 and Eq. 9 to 16 with simulated data of a mask for DUV lithography under different ex posure conditions. These data were used here since the simulation of a DUV mask is well understood and the simulated data therefore offer a good "comparison measure" in relation to how well the described model describes reality. As a matter of principle, this does not affect the applicability of the model to EUV masks.

The data correspond to a mask with structural lines and interposed clear spaces ("lines and spaces ") in a periodic sequence, which should be imaged on the wafer with a pe riod ("pitch”) of 76 nm in the x-direction (i.e., with a repeated line and clear space width of 38 nm in each case on the wafer). The simulated data show the imaging prop erties of the mask at a different values for the focus depth z and different values for the exposure dose t.

In this case, the focus depth was varied incrementally between a plurality of focal planes, with "Focal plane No. 7" corresponding to the nominal value z ₀ and two adja cent focal planes corresponding to a variation in the variable z of 20 nm in each case. The simulated data comprise "Focal planes No. 1 - No. 13”, i.e., a focus depth range of -120 nm to +120 nm steps of 20 nm in relation to the wafer surface.

The exposure dose t was likewise varied incrementally between seven different, equidis tant doses, with the mean value, i.e., " Dose No. 4", corresponding to the nominal value t ₀ . The range over which the exposure dose was varied corresponds to ±5% of the nomi nal dose to·

Figure 5a shows an excerpt 500 of the (simulated) intensity distribution of the DUV radiation at the wafer surface as a function of the x-position on the wafer. Here, the zero of the x-axis was chosen arbitrarily. A family of curves which corresponds to the aforementioned different values of the focus depth z is shown. The curve labelled by reference sign 520 is the curve at the nominal focus depth z ₀ and has the greatest am plitude of all curves in the family of curves. The curves denoted by the reference sign 525 are the curves with maximum defocus (i.e., at z = z ₀ ± 120 nm) and these have the smallest intensity amplitude. Moreover, a plotted nominal value for the critical dose is indicated by reference sign 530 and, around this critical dose, a plotted range between -5% and +5% is indicated by the reference signs 532 and 535.

In the region 510, the intensity of the curves is below the nominal value denoted by the reference sign 530, and so no photoresist is removed here during the exposure (at least if a perfect contrast of the photoresist is assumed), while outside of this region the in tensity is higher than the nominal value 530 of the critical dose, i.e., the photoresist is removed by the exposure (once again under the simplifying assumption of a perfect contrast). Consequently, structure edges form on the wafer where the intensity curves pass through the critical dose 530, for example in the region denoted by the reference sign 550. In general, the position of this crossing point, i.e., the position of structure formation on the wafer, and hence the edge position x are displaced when the critical dose or the exposure dose t is varied.

Figure 5b shows this relationship in the form of an excerpt 501 of a focus-exposure matrix for the edge position x of the edge forming in the region 550, in comparison with the target value thereof x ₀ . Here, the vertical, dashed line 520 in the centre of the figure corresponds to the nominal focus depth z ₀ , i.e., the curve 520 in figure 5a with the greatest intensity amplitude. To the left there is negative defocussing (up to - 120 nm) and to the right there is positive defocussing (up to +120 nm).

The substantially horizontally extending lines of different symbols correspond to the various exposure doses t, with the central line 540 corresponding to the nominal expo sure dose t _o - The lowermost line 542 corresponds to a deviation of the exposure dose of to by -5% and the uppermost line 545 corresponds to a deviation of the exposure dose of to by +5%.

The same circumstances are also illustrated again in figure 5c, albeit in this case in a different type of illustration 502, in which the focus depth z is plotted on the abscissa and the exposure dose t is plotted on the ordinate, with the deviation dx of the edge po sition x from its target value x ₀ being illustrated for each value pair (z, t ) in accordance with the symbol key shown at the right edge of figure 5c.

Figure sd finally shows a comparison 503 of the (simulated) data illustrated in fig ures 5a-c with the predictions of the third order model as described by Eq. 16. Here, the type of illustration is the same as in figure 5c, i.e., the focus depth z is plotted on the abscissa and the exposure dose t is plotted on the ordinate. However, illustrated by a varying symbol density in this case is the deviation of the data of figures 5a-c from the predictions of Eq. 16 (after fitting the involved parameters to the available data rec ord in a manner known per se to a person skilled in the art). As shown, the correspond ence of the model with the data (after appropriate parameter fitting) is very good, at least in the region indicated by the dashed line 560, and the difference between the model predictions in respect of the edge position x and the data shown in figures 5a-c is less than 0.02 nm throughout the region 560, and usually even less than 0.01 nm.

This means that the model described (cf. Eq. 16) can make very good statements about the influence of possible variations of the exposure parameters on the "effective" edge position x in a significant region around the nominal work point (z ₀ , t ₀ ) and can there fore, within the scope of a repair process which attempts to adjust or correct this posi tion, supply essential information for controlling and influencing this process. Figure 6 finally shows further application options of the method described herein and the models described herein, for example the models described in relation to figures 3 and 4, within the scope of mask repair.

Here, the excerpts 600, 601 and 602 relate to a repair process which contains etching away excess material. Excerpt 600 in this case shows an SEM recording of a mask ex cerpt prior to the etching process, and excerpt 601 shows an SEM recording of the same mask excerpt after the process has been completed. To make things clearer, the three "trenches", for which excess material has been etched away, are labelled by "1",

"2" and "3" in sequence in excerpt 601.

By contrast, excerpts 610, 611 and 612 relate to a repair process containing a material deposition of missing absorber material. Excerpt 610 shows an SEM recording of a mask excerpt prior to the deposition process, and excerpt 611 shows an SEM recording of the same mask excerpt following this process. To make things clearer, the two "ab sorber lines", in which additional absorber material has been deposited, are labelled "1" and "2" in sequence in excerpt 611.

The two excerpts 602 and 612 each show a graphically prepared analysis of the results of these two repair processes, with the numbering of the repaired trenches of "1", "2" and "3" and the numbering of the repaired absorber lines of "1" and "2" having been maintained. In this case, the result of the respective repair process in respect of one or more of the quality metrics disclosed herein (e.g., deviation dx from the target value x ₀ , phase error/telecentricity error ~ x _z and/or edge slope ~ x ₍ ) is illustrated here in colour- coded fashion.

By way of example, a certain symbol type (e.g., small squares, cf. regions 605 and 615) can express that the considered quality metric or quality metrics have already adopted a very good value following the repair or at the currently considered time during the re pair process, i.e., the repair in respect of these metric(s) was successful (e.g., no more than 5% deviation from a respectively predefined target value). A different symbol or symbol type (e.g., hatched, cf. regions 608 and 618) can indicate that the correspond ing quality metric(s) is currently in a mid-range (e.g., between 5% and 10% deviation from the respective predefined target value). Yet another symbol, symbol type, symbol density or colour (e.g., red) or even no symbol or colour could mean that there still is no acceptable result present in respect of the considered quality metric(s) (e.g., more than 10% deviation) and that the repair process should therefore still be continued, for ex ample.

Which quality metric or quality metrics are visually represented here in this manner may be preset or may be selectable by a user, possibly with the option of switching be tween different representations. If a plurality of quality metrics is selected for visualisa tion purposes, the "superposition" (e.g., the sum or mean value thereof) can be pre sented in a single image, or all metrics can be shown in separate images, or mixtures thereof.

In any case, such a visual (e.g., symbol-coded and/or colour-coded) representation al lows accurate and targeted monitoring of the repair process, not only in respect of a single characteristic but a plurality of characteristics in the form of quality metrics pre sented herein can be included.

Finally, it should be mentioned that such a representation of the repair success can also be converted into a numerical or automated method (or vice versa), for example by vir tue of the defining intervals for the respective quality metric that correspond to the var ious symbols and/or a varying symbol density (see above in relation to a specific exam ple in this respect, having deviation ranges of 0-5%, 5-10% and >10%), and so this may allow automated control of a repair process (largely) without human intervention.