EP3720985A1

EP3720985A1 - Method and device for automatically determining values of adjustments to operating parameters of a deposition line

Info

Publication number: EP3720985A1
Application number: EP18833275.3A
Authority: EP
Inventors: Yohan FAUCILLON; Vojislav Gajic; Thierry KAUFFMANN; Etienne Sandre-Chardonnal
Original assignee: Saint Gobain Glass France SAS; Compagnie de Saint Gobain SA
Current assignee: Saint Gobain Glass France SAS; Compagnie de Saint Gobain SA
Priority date: 2017-12-07
Filing date: 2018-12-07
Publication date: 2020-10-14
Also published as: FR3074906B1; RU2020121298A3; CN111670264A; RU2020121298A; BR112020011322A2; MX2020005919A; US11739417B2; WO2019110948A1; TW201936958A; US20200392617A1; FR3074906A1

Abstract

The determining method according to the invention comprises: — obtaining (E10) a mathematical model relating at least one operating parameter of the deposition line to at least one quality function defined on the basis of at least one quality measurement characterising the quality of a stack of thin layers deposited by the deposition line on a transparent substrate; — obtaining (E50) a value of said at least one quality function from a value of said at least one quality measurement measured at the end of the deposition line on a stack of thin layers deposited by the deposition line on a substrate with the deposition line parameterised with a so-called current value of said at least one operating parameter; and — automatically determining (E80) by means of the mathematical model a value of an adjustment of the current value of said at least one operating parameter allowing a discrepancy existing between the obtained value of said at least one quality function and a target value chosen for said at least one quality function for the stack of thin layers to be decreased.

Description

Method and apparatus for automatically determining operating parameter adjustment values of a deposit line

Background of the invention

The invention relates to the manufacture of transparent substrates coated on at least one face of a stack of thin layers, in particular transparent substrates made of glass or polymeric organic material.

It is conventional to provide such transparent substrates with coatings conferring on them particular properties, especially particular optical properties, for example reflection or absorption of radiation of a given wavelength range, electrical conduction properties, or properties related to ease of cleaning or the possibility for the substrate to self-clean. These coatings are generally stacks of thin layers based on inorganic compounds, especially metals, oxides, nitrides or carbides. For the purposes of the invention, the term "thin layer" refers to a layer whose thickness is less than one micrometer and generally ranges from a few tenths of nanometers to a few hundred nanometers (hence the term "thin").

A stack of thin layers is generally manufactured via a succession of thin film deposits made in a plurality of compartments of a deposition line (typically 20 to 30 compartments), these deposits being made in the different compartments using one or more methods. of deposition such as, in particular, magnetic field assisted sputtering (also known as magnetron sputtering), ion-assisted deposition (or IBAD for ion-assisted waste deposition), evaporation, chemical vapor deposition ( or CVD for Chemical Vapor Deposition), plasma enhanced chemical vapor deposition (PECVD for Plasma-Enhanced CVD), low pressure chemical vapor deposition (LPCVD for Low-Pressure CVD).

In order to optimize the quality of the final product manufactured, it may be necessary to adjust the thicknesses of the thin layers deposited on the transparent substrate by the deposition line with a high accuracy, typically less than 1%, and this, on the entire width of the stack of thin layers deposited on the substrate. This can be achieved by acting on various parameters of the manufacturing process, and more particularly by adjusting the values of different operating parameters of the deposition line. Such operating parameters are, for example, for a magnetron sputtering deposition method, the powers applied to the cathodes of the different compartments of the deposition line, the speed of travel of the transparent substrate opposite each cathode, the pressure and / or the composition of the gases used in each compartment of the deposit line, etc. They can therefore be multiple and varied, and concern all or part of the compartments of the deposit line. It is understandable that in certain circumstances (for example when the deposit line is large and includes a large number of compartments), such an adjustment can be complex to implement and needs to be optimized.

Various solutions are proposed today in the state of the art to make such an adjustment.

A first solution is to perform a layer-by-layer adjustment. The deposition line is parameterized and used to deposit at each passage of the substrate a single layer of material. Optical measurement devices placed at the output of the deposition line then measure different spectrometric values (eg in transmission and / or in reflection) at a plurality of points of the substrate on which the layer of material, distributed over the substrate, has been deposited. entire width of the board. Then the measured spectrometric values are compared to target values, determined over the entire width of the plate for the material layer considered by observing a stack of thin layers deemed satisfactory. The operating parameters of the deposition line are then adjusted by trial and error to attempt to obtain the target spectrometric values determined for the layer. This procedure (repository, measurements, comparison with target values and adjustment of the deposition line) is repeated for the same layer until target values are obtained for this layer (within a tolerance factor) and then repeated for each layer to be deposited on the substrate.

This first solution can therefore be extremely time-consuming and time-consuming, even in the case where it is automated, since each layer is individually processed to adjust the operating parameters of the deposition line.

A second solution is to use an optical simulation software that relies on physical modeling of the thin film stack deposited by the deposition line on the transparent substrate. This physical modeling takes into account, in particular, the thickness of the layers, possible gradients, the physical characteristics of the materials used for the layers (and in particular their refractive and absorption indices), etc. The optical simulation software is used to determine the "real" thicknesses of the thin film stack, and the operating parameters of the deposition line are then adjusted to make these thicknesses determined by means of the optical simulation software to desired target thickness values for thin film stacking.

The physical modeling of a stack of thin layers is a difficult task, especially when the number of layers is large and important accuracy is expected (this is the case for example when one wishes to achieve high-end products ). Such modeling requires a very large number of measurements and is extremely expensive and time-consuming.

A third solution proposed in the prior art is to rely on in-situ instrumentation: more specifically, various optical measuring devices (eg ellipsometry, reflection spectrometry and / or transmission) are placed in various locations. the deposit line (in the actual deposit rack) so that it can perform optical measurements on each layer deposited on the substrate. Target values for these optical measurements are moreover determined from a stack of thin layers considered satisfactory. Adjustments of the operating parameters of the deposition line are identified to make the values measured by the different optical measuring devices positioned in the deposition line tend towards these target values.

This approach using in-situ instrumentation has several disadvantages.

First, depending on the technology used by the drop line, it can be difficult to implement. In the case for example of a deposition line using a magnetic field assisted sputtering (magnetron), it is extremely difficult to protect the optical measuring devices from spurious deposits. It follows that the optical measuring devices clog up quickly, which significantly alters the measurements delivered by these devices and does not provide the required accuracy for the thicknesses of the layers of the stack deposited on the transparent substrate. Moreover, this solution requires a large number of measuring devices (devices must be arranged after each compartment of the deposit line and over the entire width of the tray), and is therefore particularly expensive and difficult to implement.

There is therefore a need for a method for adjusting the operating parameters of a deposition line of a stack of thin layers on a transparent substrate that does not have the disadvantages of the state of the art.

Object and summary of the invention

The invention responds to this need by proposing a method for automatically determining an adjustment value of at least one operating parameter of a deposition line for depositing a stack of thin layers on a transparent substrate. this process comprising:

A step of obtaining a mathematical model connecting at least one operating parameter of the deposition line with at least one predetermined quality function defined from at least one quality measurement of a deposited thin-film stack; by the deposit line;

A step of obtaining a value of said at least one quality function from a value of said at least one quality measurement measured at the output of the deposition line on a stack of thin layers deposited by the line deposition on a transparent substrate while the deposition line was parameterized with a current value said at least one operating parameter; and

A step of automatic determination by means of the mathematical model obtained of an adjustment value of the current value of said at least one operating parameter making it possible to reduce a difference existing between the value obtained from said at least one quality function and a target value chosen for said at least one quality function for thin film stacking.

Correlatively, the invention also relates to a device for determining an adjustment value of at least one operating parameter of a deposition line for depositing a stack of thin layers on a transparent substrate, this device comprising :

A first obtaining module, configured to obtain a mathematical model connecting at least one operating parameter of the deposit line with at least one predetermined quality function defined from at least one quality measurement of a stack of thin layers deposited by the deposit line;

A second obtaining module, configured to obtain a value of said at least one quality function from a value of said at least one quality measurement measured at the output of the deposition line on a deposited thin-film stack; on a transparent substrate, said deposition line having been configured by means of a so-called current value of said at least one operating parameter; and

A determination module configured to automatically determine, by means of the mathematical model obtained, an adjustment value of the current value of said at least one operating parameter making it possible to reduce a difference existing between the value obtained from said at least one quality function; and a target value chosen for this at least one quality function for thin film stacking.

No limitation is attached to the technique used by the deposition line to deposit the stack of thin layers on the transparent substrate. Thus, for example, the deposition line may implement a magnetic field assisted sputtering technique. Alternatively, it may use an ion beam assisted deposition (IBAD) technique, an evaporation deposition technique, a chemical vapor deposition (CVD) technique, a plasma enhanced chemical vapor deposition (PECVD for Plasma-Enhanced CVD), a low-pressure chemical vapor deposition (LPCVD) technique, etc.

It will be noted that within the meaning of the invention a value may consist of a single component or designate a vector or matrix comprising a plurality of components (for example when the values of several operating parameters have to be adjusted, typically one or several operating parameters for several compartments of the deposition line, or when several quality measurements are typically considered at different points of the plateau).

Similarly, according to the invention, one or more quality measures can be considered in the mathematical model envisaged. These measurements can be considered individually, in the state or in a modified form, or be linked together if necessary via one or more mathematical functions (for example via an average). These various possible forms are taken into account in the sense of the invention by the concept of quality function.

The invention thus proposes a simple and innovative solution compared to the state of the art for automatically adjusting the operating parameters of a deposition line of a stack of thin layers on a transparent substrate, from measurements. of quality performed at the output of the deposition line on the complete stack of thin layers deposited by it. These quality measurements make it possible to determine to what extent the stack of thin layers deposited by the deposit line meets the target characteristics that it is desired for it to respect. The solution proposed by the invention is based on a simplified mathematical model connecting the operating parameters of the deposition line to the quality measurements made at the output thereof on the stack of thin layers deposited on the substrate. Such a mathematical model is in practice much more robust and easy to determine than a physical model of the stack as proposed in the state of the art, which, on the contrary, is a nonlinear model requiring many physical measurements to be established. On the contrary, in the invention, the model considered is a mathematical model that does not seek to rely strictly on physical realities and is therefore more robust to errors (due to the very nature of the model). This mathematical model can also, according to the choice made, have much fewer coefficients to estimate than a physical model, and its nature (eg linear model or more generally polynomial, etc.) is a priori independent of the number of layers and / or materials used for stacking. Such a mathematical model can also be easily determined by learning, from a set of reference data established considering a reasonable number of previously deposited stacks (according to a dedicated experiment plan, or during previous productions); it can also be easily updated.

An advantage of the invention is the adjustment of manufacturing parameters of several sensitive thin layers.

Those skilled in the art know that the characteristics of a layer, in particular its thickness or its stoichiometry, may vary due to the variation of non-adjustable parameters, for example due to the decrease in the thickness of the target. Thin films whose tolerances are less than the variability of the characteristics of the layer due to these non-adjustable parameters can in particular be considered as a sensitive layer.

The invention can notably be implemented as follows:

a first stack of thin layers is produced with parameters which are well known to lead to the obtaining of a stack whose properties are unsatisfactory;

one or more measurements are taken, but only at the output of the deposition line, and the manufacturing parameters of several layers, in particular of sensitive layers, are adjusted according to the result of these measurements, to obtain, at the next iteration of the process, or even after several iterations, a stack whose properties are satisfactory.

The invention therefore offers a considerable time saving compared to the state of the art for the adjustment of operating parameters of a deposition line. This is also permitted by the invention without requiring the realization of partial stacks, or resort to expensive in situ instrumentation and difficult to maintain. Nevertheless, the adjustment values of the operating parameters of the deposition line provided by the invention are precise and make it possible to quickly reach the desired specifications for the thin film stack to be produced.

The adjustment method proposed by the invention is advantageously based on measurement devices already present at the output of a deposition line, and particularly robust. Such measuring devices are for example spectrometers and / or ellipsometers making it possible to obtain quality measurements of the complete thin-film stack, and more specifically measurements of optical spectra (eg in transmission and / or in reflection , at an angle of 0 ° or other angles (eg 60 °)) and / or ellipsometric values.

In addition to optical measurements, said at least one quality measure taken into account by the mathematical model may also comprise at least one of:

A measurement of an electrical property of the stack of thin layers (for example of its electrical resistance);

A measurement of a mechanical property of the stack of thin layers (for example of its mechanical strength);

A measurement of a chemical property of the stack of thin layers (for example of its chemical resistance); and

A measurement of a mass property of the stack of thin layers.

Of course, these examples are given for illustrative purposes only and other measures may be considered in the mathematical model in addition to the aforementioned quality measures. In particular, it is possible to consider not only quality measurements made on the complete stack but also other measurements made on a partial stack by means, for example, of in situ instrumentation provided in one or more compartments of the deposit line. .

Likewise, the operating parameters of the deposit line that can be adjusted by means of the invention can be of different natures. Thus, in a particular embodiment in which the deposition line comprises a plurality of compartments in which said thin layers are deposited on the transparent substrate, said at least one operating parameter can comprise:

At least one power or voltage applied to a cathode of a said compartment of the deposition line; and or At least one rate of movement of the substrate opposite a cathode of a said compartment of the deposit line; and or

At least one pressure and / or an amount of a gas used in a said compartment of the deposit line; and or

At least one magnetic field or a magnetic field distribution applied in a said compartment of the deposition line; and or

At least one adjustment of a mechanical element of a said compartment of the deposit line.

Of course this list is not exhaustive and other operating parameters can be considered.

No limitation is attached to the type of mathematical model used by the invention. This can be a linear model that is very simple to implement, such as a more complex polynomial model (eg of order 2 or 3), or a model determined via a learning algorithm. statistic such as a linear regression algorithm (eg regression of Ridge, Lasso or carrier vector type (SVR), an algorithm using a neural network or decision trees (eg Random Forest algorithm), an algorithm learning by reinforcement or interpolation (eg Gaussian Process algorithm or nearest neighbors), etc.

In a particular embodiment, the mathematical model used by the invention relates at least one variation of said at least one operating parameter with at least one variation of said at least one quality function.

Such a model, which seeks to consider variations in the operating parameters and variations in the quality function or functions, makes it possible to overcome any noise or bias that may result from various "hidden parameters", such as for example acquisition conditions of the different quality measurements that may differ between the stack considered during the adjustment and the reference stacks considered in the learning game of the mathematical model, or errors or drifts presented by the measurement used to acquire the quality measurements at the output of the deposition line (for example by optical sensors), or slow drifts affecting the deposition line (related, for example, to the age of the cathodes used in the compartments, etc. .).

To determine the mathematical model used by the invention, the obtaining step (in other words the learning step of the mathematical model) may comprise, for example:

Obtaining a so-called reference value of said at least one operating parameter corresponding to the parameterization of the deposition line during a so-called reference deposition of a stack of thin layers on a substrate;

Obtaining, for a plurality of stacks of so-called learning thin layers deposited by the deposition line on a substrate when said deposition line is parameterized with values of said at least one operating parameter exhibiting determined variations. with respect to said reference value, a plurality of values of said at least one quality function; and

A step of determining the mathematical model from the plurality of values of said at least one quality function obtained for the plurality of stacks of thin learning layers.

Different strategies can be used to select stacks of thin learning layers.

For example, it is possible to define an experiment plan, specifically dedicated to the determination of the mathematical model. In this experimental design, for example, variations of the values of the operating parameters are chosen to guarantee the linear hypothesis envisaged for the mathematical model (for example + 10% or -10% with respect to the reference values), and the deposition line is parameterized as many times as necessary with operating parameters adjusted according to the selected variations. In this way, the stacks of training layers are limited to a reasonable and limited number of stacks of relevant learning layers (typically fifteen) to obtain a reliable model. It is noted that the training stacks can be made following each other by the line of deposit, or at a more spaced way in time. Such an experimental design is particularly relevant when it is desired to determine the mathematical model via a deterministic calculus, the complexity of which is directly related to the number of learning stacks considered.

Alternatively, another approach may be to consider various learning stacks already made by the drop line in a "random" manner when using the drop line, and to exploit the information available for these stacks (measurements quality and operating parameters of the associated deposit line) to determine the mathematical model. There is therefore much more information to determine the mathematical model. However, since the stacks in question have not been made specifically within the framework of a pre-established experiment plan and dedicated to the determination of the model, we preferentially filter to constitute the model among all available data, those which correspond to outliers (related to errors from sensors for example) or to stacks with anomalies (eg stacks that do not have the right composition). It is also possible to consider only the data corresponding to stacks having, relative to reference values of the operating parameters, variations within a given range of variations (for example in a range of variations of +/- 15% around a reference value). It can also be ensured that the retained training stacks have a uniform distribution over this range of variations (elimination of outliers); in this way, fewer learning stacks are needed to establish the mathematical model (eg, 40). The selection of the training stacks may be accompanied by other conditions aimed at ensuring the compatibility of the training stacks with each other, such as, for example, conditions on the date of deposit of the stacks, etc.

This approach, without experimental design, has a privileged application when the mathematical model is determined via a statistical algorithm. It also makes it possible to update more easily and more often the mathematical model used.

As mentioned above, different mathematical models can be envisaged within the scope of the invention for connecting the operating parameters to the quality function. These mathematical models may furthermore provide values of the quality function as a function of the values of the operating parameters (then referred to as "direct models" in the remainder of the description), or vice versa, values of the operating parameters depending on values of the quality function (then referred to as "indirect models" in the rest of the description).

In one embodiment, the stack of thin layers is intended to undergo another treatment, called "post-treatment" after the series of measurements carried out at the outlet of the deposition line (for example thermal quenching, laminating, or double glazing). . A goal sought by the invention can be to obtain a final product whose properties are satisfactory after this treatment and thus to adjust the operating parameters so that the final product (and not the thin film stack) at the output of the deposition line) has desired properties. This embodiment of the invention thus aims to minimize the difference between the value of the quality function measured at the output of the deposition line for a product whose properties have been judged satisfactory after further processing and the current value of the quality function.

In practice, properties of the final product can be related to measurements made at the output of the deposition line, for example by an automatic statistical learning mechanism. In one embodiment, the quality function is determined to account for this relationship.

In one embodiment of the invention, the quality function is chosen to represent certain characteristics of the stack of thin layers that are not easily measurable at the line output. This may be the case of thermal characteristics, the measurement of which takes a long time. Therefore, and more generally, in one embodiment of the invention, the vector or matrix representing the quality function may have one or more components that are not measurements but transformations of these measurements. Such a transformation can in particular be a transformation making it possible to predict a property that is difficult to measure according to an easily measurable property;

a transformation making it possible to predict a property after further processing.

The use of a direct mathematical model requires an inversion of the model and can thus require, to converge towards the chosen target value for the quality function, several iterations in other words, several successive adjustments of the operating parameters. However, this convergence is quite fast, and only a few iterations are enough in practice. In addition, the use of such a direct model is more intuitive since it models the behavior of the deposited thin-film stack according to the setting of the deposition line. Its interpretation is therefore simpler from a physical point of view since it makes it possible to identify how the setting of the deposition line influences the properties of the thin film stack deposited.

On the contrary, the use of an indirect model makes it possible to directly obtain the adjustment values to be applied to the deposit line in order to obtain the target values of the quality function without carrying out an optimization as such and therefore without requiring inversion of the model. Such a model is therefore simpler to use in determining the adjustment values.

In a particular embodiment:

The mathematical model is a polynomial model ("direct") providing variations in the value of said at least one quality function as a function of variations in the value of said at least one operating parameter; and

The determination step comprises a step of inverting the polynomial model to determine said adjustment value.

The inversion step may further comprise a regularization of the inversion of the polynomial model according to known techniques in order to improve the robustness of the model, such as for example using Tikhonov or Lasso regularization. Such regularization makes it possible to better distinguish the different layers of the stack of thin layers.

The inventors have found that a simple polynomial model such as a linear model is sufficient to lead to good results in terms of fit. This is the case, for example, when it comes to determining variations in optical spectra as a function of power variations to be applied to the cathodes of a deposition line using a magnetron sputtering technique.

Such a polynomial model (and a / brf or / linear) has the advantage of being very simple to determine; it includes few coefficients to estimate and is particularly robust. In addition, when the model chosen is a linear model, it can be represented in matrix form, easy to manipulate, which allows to derive the adjustment values by a deterministic calculation.

Thus, more particularly, in a particular embodiment, the deposition line comprises N compartments, N denoting an integer greater than 1, each compartment being respectively associated with an operating parameter P (n), n = 1, ... , N, and the mathematical model is a linear model using a matrix denoted M comprising N columns u (n), each column being defined by: var (n)

where var (n) denotes the difference between the values of the at least one quality function obtained for two stacks of thin learning layers deposited by the deposition line while the latter was parameterized with a value of the operating parameter P (n) having a variation +5 (n) with respect to the reference value of the operating parameter P (n) and with a value of the operating parameter P (n) having a variation - (n) with respect to the reference value of the operating parameter P (n).

In this embodiment, the adjustment value, denoted a, of said at least one operating parameter making it possible to reduce the difference existing between the value obtained from said at least one quality function and the target value chosen for it. , is given for example by:

a = (M ^T M) ~ ¹ M ^T v

where v denotes the difference between the value obtained from the quality function and the target value chosen for the said quality function for the stack of thin layers and M ^T denotes the transposed matrix of the matrix M.

As mentioned above, in a particular embodiment, the mathematical model used by the invention is obtained by using an automatic statistical learning algorithm.

The advantage of using a statistical learning algorithm is, as mentioned above, that it offers the possibility of establishing the mathematical model from stacks of "random" thin layers made by the deposit line, and for which quality measures are available in association with the operating parameters with which the deposit line was configured to perform these stacks. One can dispense with resorting to a dedicated experience plan to build the mathematical model, and advantageously exploit information already available, which has not necessarily been generated for this purpose.

In a particular embodiment, the determination method further comprises a validation step of the mathematical model obtained.

Such validation can be performed automatically or manually by an operator, for example by comparing, on a set of thin film stacks selected for validation, the layers that need to be adjusted by observing these stacks of validation against to the layers identified by the mathematical model (ie for which the mathematical model indicates non-zero adjustment values of the operating parameters associated with these layers).

Of course, if the mathematical model is, at the end of this validation step, considered invalid, a new mathematical model is determined for example from another set of learning stacks, or from a filtered set of stacks in which the data relating to stacks with anomalies or outliers has been deleted.

In a particular embodiment, the various steps of the determination method are determined by computer program instructions.

Consequently, the invention also relates to a computer program on an information carrier, this program being able to be implemented in a determination device or more generally in a computer, this program comprising instructions adapted to the implementing the steps of a determination method as described above.

This program can use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in a partially compiled form, or in any other form desirable shape.

The invention also relates to a computer readable information or recording medium, and comprising instructions of a computer program as mentioned above.

The information or recording medium may be any entity or device capable of storing the program. For example, the medium may comprise storage means, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or a magnetic recording medium, for example a floppy disk or a disk. hard.

On the other hand, the information or recording medium may be a transmissible medium such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by other means. The program according to the invention can be downloaded in particular on an Internet type network.

Alternatively, the information or recording medium may be an integrated circuit in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of the method in question.

The invention also relates to a method of parameterizing a deposition line capable of depositing a stack of thin layers on a substrate, this method comprising:

A first step of setting up the deposit line with a current value of at least one operation parameter of the deposit line;

A step of deposition by the deposition line parameterized with said current value of a stack of thin layers on a substrate;

A step of determining an adjustment value of the current value of said at least one operating parameter implementing a determination method according to the invention and using a value of said at least one measured quality measurement at the output of the deposition line on the stack of thin layers deposited during the deposition step; A step of estimating an adjusted value of said at least one operating parameter from the current value and the determined adjustment value; and

A second step of setting the deposition line with the adjusted value of said at least one estimated operating parameter during the estimation step.

Correlatively, the invention also provides a system for depositing a stack of thin layers on a transparent substrate comprising:

A deposition line capable of depositing a stack of thin layers on a transparent substrate;

A deposition line configuration module configured to parameterize the deposition line with a current value of at least one operation parameter of the deposition line;

A determination device according to the invention of an adjustment value of the current value of said at least one operating parameter configured to use a value of at least one quality measurement measured at the output of the deposit line on a stack of thin layers deposited on a transparent substrate by the deposition line;

An estimation module, configured to estimate an adjusted value of said at least one operating parameter from the current value and the adjustment value determined by the determination device;

the parameterization module being configured to parameterize the deposition line with the adjusted value of said at least one operating parameter estimated by the estimation module.

The parameterization method and the deposition system according to the invention have the same advantages as previously described, as the method and the determination device.

In a particular embodiment, the steps of deposition, determination, adjustment and the second parameterizing step of the parameterization process are implemented as long as:

The difference between the value of said at least one quality function estimated from said measured value at the output of the deposition line and the target value chosen for said at least one quality function is greater than a predetermined threshold, or

The adjustment value determined for said at least one operating parameter is less than a predetermined threshold.

The value of a quality function can be represented by a vector or a matrix.

In a first embodiment, the difference above is a distance (between two vectors or two matrices) and the threshold a real value.

In another embodiment, the difference between two vectors is a vector and the difference between two matrices is a matrix and the threshold is an element of the same nature. In the case of vectors for example, and equivalently for the case of the matrices, it is considered that the difference between two vectors is lower than the threshold vector, if the absolute value of each component of the difference vector is less than the threshold vector component. This embodiment allows more flexibility because it is possible to set the components of the threshold vector differently to give importance to certain characteristics, in particular to those related to the product specifications.

The first condition makes it possible to ensure convergence towards the target value of said at least one quality function, while avoiding implementing unnecessary iterations. The second condition makes it possible to ensure that the determined adjustment value is not aberrant given expert criteria (according to which, in particular, the predetermined threshold is set). This second condition can also make it possible to force the adjustment value to meet certain criteria (eg to be for example in a certain range of values to guarantee certain properties to the substrate covered with the stack of thin layers).

It may also be envisaged, in other embodiments, that the determination method, the determination device, the parameterization method and the deposition system according to the invention present in combination all or some of the aforementioned characteristics.

Brief description of the drawings

Other features and advantages of the present invention will emerge from the description given below, with reference to the accompanying drawings which illustrate an embodiment having no limiting character. In the figures:

Figure 1 shows, schematically, in its environment, a deposition system according to the invention, in a particular embodiment;

Figure 2 shows a deposition line of the deposition system of Figure 1;

FIG. 3 schematically represents the hardware architecture of an estimation device of the deposition system of FIG. 1, in accordance with the invention, and configured to determine operating parameter adjustment values of FIG. depot line of Figure 2;

FIG. 4 shows in a particular embodiment, in the form of a flow chart, the main steps of a parameterization method according to the invention, as implemented by the deposition system of FIG. that the main steps of a determination method according to the invention as implemented by the estimation device of Figure 3; and

FIGS. 5A-5C illustrate the impact on different optical spectra measured on a thin film stack produced by a line of variation deposition applied on operating parameters of this deposition line.

Detailed description of the invention

FIG. 1 represents, in its environment and in a particular embodiment, a deposition system 1 according to the invention, intended to be used for carrying out the depositing a stack of thin layers on a transparent substrate 2 by means of a deposition line 3. The deposition system 1 advantageously allows the deposition line 3 to be properly parameterized in order to ensure that the stack of layers thin meets predetermined target characteristics.

By way of illustration, in the example envisaged in FIG. 1, the transparent substrate 2 is a 6 mm thick glass substrate on which is deposited a stack of thin layers comprising a plurality of thin functional layers that can act on the solar radiation (namely here, two silver layers denoted by Agi and Ag2), and coatings formed of one or more thin layers located on either side of each functional layer. In the following, the term "module" (M1, M2, M3) denotes each of the coatings which surround the functional silver layers, it being understood that a module may consist of a single thin layer or a thin layer. plurality of thin layers. For the purposes of the invention, the term "thin layer" refers to a layer whose thickness is less than one micrometer and generally ranges from a few tenths of nanometers to a few hundred nanometers (hence the term "thin").

In the example envisaged in FIG. 1 (hereinafter referred to as EX1), the stack of thin layers consists, successively from substrate 2:

A module Ml consisting of a thin layer of silicon nitride (Si3N4) with a thickness of 14.5 nm and a thin layer of zinc oxide (ZnO) with a thickness of 6 nm;

- a thin layer of silver Agi thickness 5.5nm;

A module M2 consisting of a thin layer of blocking alloy of nickel and chromium (NiCr) with a thickness of 0.6 nm, a thin layer of zinc oxide (ZnO) with a thickness of 6 nm, a thin layer of silicon nitride (Si3N4) with a thickness of 34 nm, a thin layer of niobium nitride (NbN) of thickness lnm, a thin layer of silicon nitride (Si3N4) with a thickness of 34 nm and a thin layer of zinc oxide (ZnO) with a thickness of 5 nm;

A thin layer of silver Ag2 17nm thick; and

A module M3 consisting of a nickel-chromium alloy thin-walling layer (NiCr) of thickness lnm, a thin layer of zinc oxide (ZnO) with a thickness of 5 nm, a thin layer of silicon nitride (Si3N4) 23nm thick and a thin layer of zinc oxide and tin (SnZnO) 5nm thick.

Of course, this EX1 example is in no way limiting and other stacks of thin layers with other thicknesses as well as other transparent substrates (eg substrates of polymeric organic material, flexible or rigid) can be envisaged. .

In the embodiment illustrated in FIG. 1, the deposition line 3 used to deposit the stack of thin layers on the transparent substrate 2 implements a magnetic field assisted sputtering technique also known as magnetron sputtering. The glass substrate coated with the stack of thin layers deposited by the filing line 3 (also sometimes referred to as a tray in this description) is hereinafter referred to as 4.

In known manner, a sputtering technique relies on the condensation within a rarefied atmosphere of a vapor of a target material from a source of spray on a substrate. More precisely, the atoms of the source (also referred to as a target) are ejected into an ionized gas, such as, for example, argon, in a vacuum chamber maintained at a certain pressure. An electric field is created leading to the ionization of the gas thus forming a plasma. The target is brought to a negative potential (cathode) so that the ions present in the plasma are attracted to the target and eject atoms from it. The particles thus sprayed are diffused in the chamber and some of them are collected in particular on the substrate on which they form a thin layer. In the case of a magnetron cathode, a magnetic field oriented perpendicular to the electric field is also created by magnets placed near the cathode so as to confine the electrons in the vicinity of the cathode. This makes it possible to increase the ionization rate of the gas, and thus to significantly improve the deposition efficiency compared to a conventional sputtering technique. Since sputtering techniques are known to those skilled in the art, they are not described in more detail here.

FIG. 2 diagrammatically represents the deposition line 3 used for depositing thin layers by sputtering on the glass substrate 2. It comprises here an inlet chamber 5, a first buffer chamber 6, a magnetron sputtering chamber 7 comprising a first transfer section 8 and a second transfer section 9, a second buffer chamber 10 and an exit chamber 11.

The sputtering chamber 7 comprises, in addition to the two transfer sections 8 and 9, a succession of elements Ei, i = 1, ..., N where N denotes an integer greater than 1. Each element Ei comprises a compartment or deposition chamber 12-i containing a cathode used as a target during magnetron sputtering, and optionally one or two compartment (s) or pump chamber (s) equipped with a pump, and located (e) (s) where appropriate on both sides of the deposit chamber to create a vacuum therein. The glass substrate 2 circulates in the different successive compartments of the sputtering chamber 7, driven by a conveyor or a conveyor belt 13.

The succession of the different elements Ei, i = 1,..., N makes it possible to deposit a stack of thin layers on the glass substrate 2, in other words in the example EX1 previously described of the two silver functional layers Agi and Ag2, and thin layers constituting the three modules M1, M2 and M3. It should be noted that several successive compartments 12-i can participate in the deposition of a thin layer of the same material, in predefined proportions. However, in the example envisaged here, for the sake of simplification, it is assumed that a thin layer of material is deposited by a single compartment of the deposition line. The parameterization of the various compartments 12-i in which the thin-film depositions take place can be carried out by regulating various parameters of the magnetron cathode sputtering, and in particular, the pressure of the gas, its quantity and its composition, the power applied on the cathode of the compartment in question, the angle of incidence of the bombardment particles and the magnetic field applied in the compartment, the speed of travel of the glass substrate with respect to the cathode, etc. These adjustments are made so as to obtain at the output of the deposition line 3 a stack of thin layers respecting one or more target properties (for example mechanical and / or optical), desired for this stack. The respect of these properties is ensured as soon as the thin films deposited respect a certain thickness gauge with a predetermined precision.

The various aforementioned parameters constitute operating parameters of the deposition line 3 within the meaning of the invention, which can be adjusted in whole or in part by means of the invention, by means of a parameterization module 14 of the deposition system 1 , provided for this purpose and using adjustment values of the operating parameters determined by the invention. The term "currents" denotes the operating parameters with which the deposition line 3 is configured by the parameterization module 14 and which it applies to deposit, during the passage of a transparent substrate through its compartments, a stack of thin layers on this transparent substrate.

It should be noted that other operating parameters of the deposition line 3 may be adjusted when the deposition line 3 is used, such as, for example, mechanical element settings of the compartments 12. the deposition line 3 such as the position of metal masks provided in the compartments for limiting the deposition of thin layers made on the substrate, the position of actuators for moving the magnets placed near the cathode, etc. The invention can also apply to this type of operating parameters of the deposit line.

No limitation is therefore attached to the nature of the operating parameters of the deposit line to which the invention can be applied. In the remainder of the description, however, in order to simplify the disclosure of the invention, attention is given to the adjustment and adjustment of the powers denoted Pi, i = 1,..., N of the cathodes of each compartment 12 -i, i = l, ..., N of the deposition line 3.

To enable this adjustment, the deposition system 1 here comprises one or more systems 15 for controlling the quality of the stacks of thin layers deposited by the deposition line 3. This or these control systems 15 are placed at the output of the line of deposition. deposit 3 so as to provide quality measurements of the stacks of complete thin layers deposited by the deposit line 3.

In the embodiment illustrated in FIG. 1, for the sake of simplification, we consider a single optical control system 15 such as, for example, a spectrometer, capable of obtaining and providing optical spectral measurements made on the stack of layers. thin whole deposited by the deposition line 3 on the glass substrate 2 at different points of measured over the entire width of the substrate. Optical spectra measurements are referenced in a general manner in the following description by SP. Each optical spectrum SPj, j = 1,..., J (where J denotes an integer greater than 1), measured by the optical control system 15 at a given measurement point, provides several measured spectral values for several lengths of time. wave (for example for K wavelengths, K denoting an integer greater than 1): each optical spectrum SPj can thus be represented in the form of a vector of spectral values, each spectral value corresponding to a given wavelength .

In the remainder of the description, a representation in columns is adopted as a vector representation convention.

No limitation is attached to the types of optical spectra that can be measured by the optical control system 15: the optical control system 15 can measure one or more optical spectra typically at various points of the thin-film stack deposited by the line deposition 3 (ie over its entire width and over its entire length), in transmission and / or in reflection, on the side of the stack and / or on the side of the substrate, at different angles (eg angle of 0 ° and / or or 60 °), etc. Such an optical control system is known per se and conventionally used at the output of deposit lines. It is not described in more detail here. The integer J here takes into account, where appropriate, the different types of measured optical spectra and / or the different measurement point (s) at which these optical spectra are measured.

Alternatively, other types of control systems 15 may be placed at the output of the deposition line 3 to acquire measurements of the quality of the stack of thin layers deposited by the deposition line 3 on the glass substrate 2 These may be other optical control systems such as ellipsometers, or control systems making it possible to acquire quality measurements of other types, such as, for example, measurements of an electrical property of the stack. thin layers (eg of its electrical resistance), measurements of a mechanical property of the stack of thin layers (eg of its mechanical strength), measurements of a chemical property of the thin film stack ( (eg chemical resistance), or measurements of a mass property of the thin film stack.

It is also possible, in the context of the invention, to consider, in addition to the quality measurements provided by the control system placed at the output of the deposition line 3, measurements carried out on a partial stack and collected via a in situ instrumentation planned at one or more compartments of the deposit line 3.

The quality measurements made by the control system 15 on the complete stack deposited on the glass substrate 2 are supplied to a system determination device 16 according to the invention and configured to determine adjustment values of the common operating parameters with which the deposition line 3 is configured in view of the quality of the thin-film stack having been the subject of the quality measurements. These adjustment values are then supplied to the parameterization module 14 to modify, if necessary, all or part of the operating parameters of the deposit line 3 in order to achieve or any at least as close as possible to the desired target characteristics for the stacks of thin layers made by the deposition line.

In the embodiment described here, the determination device 16 is a computer whose hardware architecture is shown schematically in FIG. 3. It comprises a processor 17, a random access memory 18, a read-only memory 19, a non-volatile memory 20 , a communication module 21.

The communication module 21 enables the determination device 16 to obtain the quality measurements made by the control system 15 (measurements of optical spectra in the example envisaged here) on a stack of thin layers deposited by the deposition line 3. and to communicate with the parameterization module 14 of the deposition line 3 to provide it with the adjustment values to be applied to the current operating parameters with which the deposition line 3 is configured. This communication module 21 may comprise in particular a digital data bus, and / or communication means on a network (local or remote) such as, for example, a network card, etc., depending on the way in which the optical control system 15 , the parameterization module 14 and the determination device 16 are interconnected. It is assumed, in the embodiment described here, that the parameterization module 14 also integrates an estimation module 22, configured to estimate adjusted values of the current operating parameters of the deposition line 3 from the current values with which the deposition line 3 is configured and adjustment values determined and provided by the determination device 16.

The read-only memory 19 of the determination device 16 constitutes a recording medium in accordance with the invention, readable by the processor 17 and on which is recorded a computer program PROG according to the invention, comprising instructions for execution. steps of a localization method according to the invention.

This computer program PROG equivalently defines functional and software modules here configured to implement the steps of the determination method according to the invention. These functional modules support or control the hardware elements 17 to 21 mentioned above. They include in particular here:

A first obtaining module 16A, configured to obtain a mathematical model, denoted MOD, connecting at least one operating parameter of the deposition line 3 with at least one predetermined quality function defined from at least one measurement of quality provided by the control system 15 placed at the output of the deposit line. By quality function, we mean here any mathematical function of one or more quality measures;

A second obtaining module 16B, configured to obtain a value of said at least one quality function from a value of said at least one quality measurement measured at the output of the deposition line 3 on a stack of layers thin deposited on a transparent substrate, while the deposition line 3 is configured by means of a current value of said at least one operating parameter; and An automatic determination module 16C configured to determine, by means of the mathematical model MOD, an adjustment value of the current value of said at least one operating parameter making it possible to reduce a gap existing between the value obtained from said at least one function of quality and a target value chosen for this at least one quality function for thin film stacking.

Note that in the description a value can refer to either a single value or a vector of values. Typically when several operating parameters (eg the same operating parameter for each compartment of the deposit line or a plurality of operating parameters for all or part of the compartments of the deposit line) are considered, "the value of said at least an operating parameter "denotes the vector whose components are the values of the various operating parameters considered. The same is true when considering several quality measures or several quality functions.

The functions of the modules 16A-16C of the determination device 16 are described in more detail now with reference to FIG. 4.

FIG. 4 represents, in the form of a flow chart, the main steps of a method of parameterizing the deposition line 3 according to the invention as implemented by the deposition system 1, in a particular embodiment of FIG. production. This parameterization method is based on the steps of a determination method according to the invention implemented by the determination device 16 of the system 1 to determine operating parameter adjustment values of the deposition line 3. These adjustment values are intended to make it possible to obtain, by means of the deposition line 3, a stack of thin layers satisfying predetermined characteristics (eg layers of predefined thicknesses with a given precision, etc.).

As mentioned above, the invention is particularly advantageous in that it relies for determining the adjustment values, on a simplified MOD mathematical model. To facilitate the understanding of the invention, we are interested here more particularly in the parameterization of the powers of the N cathodes of the compartments of the deposition line 3. For this purpose, it is envisaged, by way of illustration, a mathematical model MOD connecting the powers Pi , i = l, ..., N of the N cathodes of the compartments of the deposition line 3 (N operating parameters in the sense of the invention) to the optical spectra SPj, j = 1, ..., J (measurements of quality within the meaning of the invention) measured by the control device 15 at the output of the deposition line on the stack of thin layers deposited by the latter (possibly at various measurement points distributed over the width of the substrate) when it is parameterized with current values Pi_curr, i = l, ..., N of the powers of the cathodes.

In the example considered here, we consider J = 3 optical spectra SPj measured by the control device 15, each optical spectrum SPj comprising K measurements taken at K different wavelengths: SP1 is an optical spectrum in transmission, SP2 is a optical spectrum in reflection measured on the stack side, and SP3 is an optical spectrum in measured reflection transparent substrate side. Each optical spectrum SPj can be represented by a vector of K components.

It is assumed here that only one measurement point is considered for the optical spectra SPj, this measurement point being situated for example at the center of the substrate taken in the width direction. SP is the optical spectrum obtained by concatenating the optical spectra SPj, j = 1, ..., J. The optical spectrum SP can be represented by a column vector having Kx] components corresponding to K distinct wavelengths considered for each optical spectrum SPj, j = 1, ..., J. The spectrum SP corresponds to a quality function according to the invention defined from the optical spectra SP1, SP2,..., SPJ, which constitute quality measurements within the meaning of the invention.

This quality function considers the quality measurements SP1, SP2, ..., SPJ as they are (ie they are just concatenated without being modified). As a variant, if several measurement points of the spectra SP1,..., SPJ are envisaged (ie for example measured at the output of the deposition line J spectrums in L distinct measurement points distributed on the plateau, resulting in JxL measured spectra SP1 (1), ..., SP1 (L), SP2 (1), ..., SP2 (L) ..., SPJ (1), ..., SPJ (L), each spectrum SPj ( l) constituting a quality measure within the meaning of the invention), it is possible for each type of spectra indexed by j, j = 1,..., J, to average the measurements available at these different measurement points (or extrapolated values from such measurements), so as to obtain J averaged spectra SPlmoy = average (SPl (l), ..., SPl (L)), SP2moy = average (SP2 (1), ..., SP2 (L)), ..., SPJmoy = mean (SPJ (l), ..., SPJ (L)). In other words, according to this example, the quality function corresponds to a concatenation of the different types of measured spectra combined with an average performed for each type of spectrum on all the measurement points considered.

As a variant, all the quality measurements carried out can be kept and the SP optical spectrum can be obtained by concatenating the measured optical spectral LxJ SP1 (1), ..., SP1 (L), SP2 (1), ..., SP2 (L), SPJ (1), ..., SPJ (L).

In the example considered here, the mathematical model MOD connecting:

Variations, with respect to reference powers denoted Pi_ref, i = 1,..., N, of the powers Pi of the N cathodes of the compartments of the deposition line 3, these variations being denoted DRΐ, i = 1 ,. .., N, and supposed identical for the sake of simplification for each cathode and equal to x, x denoting a real number;

The variations of optical spectra denoted ASPj, j = 1,..., J, resulting from the power variations applied to the N cathodes of the compartments of the deposition line 3.

The advantage of considering a mathematical model establishing a relation between variations in operating parameters and variations in quality measurements is that it makes it possible to overcome various hidden factors that may have an impact on this model, such as errors or biases that may affect the optical control system 15, etc. It should be noted, however, that this example is given for illustrative purposes only and is not limiting of the invention. As mentioned above, the invention applies to other operating parameters of the deposition line 3, to other quality measurements performed at the output of the deposition line 3 on the deposited thin-film stack, other quality functions, etc. For example, it is possible to consider in the quality function, in addition to the above-mentioned optical spectral measurements, integrated values (ie obtained) from optical spectral measurements, for example color values derived in the known color space L * a * b * (L * denotes clarity, and the components a * and b * characterize the deviation of the color of a given point from that of a gray surface of the same clarity).

The mathematical model MOD can be obtained in various ways by the determination device 16, and more particularly by its obtaining module 16A (step E10). It can be derived, as is the case in the embodiment described here, upstream of any adjustment of the parameters of the deposition line 3, by the obtaining module 16A of the determination device 16 or by another device and stored in a memory accessible by the obtaining module 16A (for example in the non-volatile memory 20 of the determining device 16) for future use. The obtaining module 16A can then obtain the MOD model when it needs it by accessing the non-volatile memory 20.

It can alternatively be derived on the fly by the obtaining module 16A when the adjustment of the parameters of the deposition line 3 is envisaged and required, or when the first adjustment is made, then stored in the non-volatile memory 20. As discussed in more detail later, the mathematical model MOD may be updated at each required adjustment.

Different types of mathematical model can be envisaged by the obtaining module 16A according to the invention. Thus, for example, the mathematical model MOD can be a linear, polynomial, statistical model, etc. The relation established by the mathematical model MOD between the operating parameters and the quality measurements can be of different types: the mathematical model MOD can actually be a so-called "direct" model, providing values of the quality functions according to the values of the values. considered operating parameters, or a so-called "indirect" model providing values of the operating parameters as a function of the values of the quality functions.

In a first exemplary embodiment, it is envisaged that the obtaining module 16A establishes a linear and direct MOD1 mathematical model. Other types of mathematical models are described later in alternative embodiments.

In order to determine (ie to learn) the mathematical model MOD1, the obtaining module 16A uses a set of stacks of so-called thin learning layers deposited by the deposition line 3. This set of stacks of thin learning layers has been obtained according to the following experimental plan: For each operating parameter considered (here each power Pi, i = 1, ..., N), a reference value denoted Pi_ref corresponding to the setting of the deposition line 3 is determined during a reference depot. These reference values are determined, for example experimentally or by experts, from a previously produced product whose quality has been judged satisfactory. One possibility for obtaining such a product is for example to use the methodology of the state of the art described above and of validating each layer separately;

The power Pi, i = 1,..., N of each cathode of the deposition line 3 is parameterized, via the parameterization module 14, with the corresponding reference value Pi_ref;

Then, for each operating parameter considered Pi, i = 1, ..., N:

o a +/- x% variation is determined to be applied to the reference value Pi_ref (eg x = + / - 10%);

o parameter, via the parameterization module 14, the power Pi of the ith cathode of the deposition line 3 with the value Pi_ref + x%, the powers Pk, k = l, ..., N, k¹i of the other cathodes being maintained at their reference values Pk_ref; the deposition line 3 is used to deposit a stack of thin learning layers corresponding to this setting of the ith cathode; and the optical control system 15 is used to measure, at the output of the deposition line 3, the optical spectra denoted by SPj (Pi_ref + x%), j = 1, ..., J of the stack of thin learning layers. realized ;

o is then parameterized, via the parameterization module 14, the power Pi of the ith cathode of the deposition line 3 with the value Pi_ref-x%, the powers Pk, k = 1, ..., N, k¹i of the others cathodes being maintained at their reference values Pk_ref; the deposition line 3 is used to deposit a stack of thin learning layers corresponding to this setting of the ith cathode; and the optical control system 15 is used to measure at the output of the deposition line 3 the SPj optical spectra (Pi_ref-x%) of the learning thin film stacks made;

the SP (Pi_ref + x%) and SP (Pi_ref- x%) spectra obtained by concatenating respectively the optical spectra SPj (Pi_ref + x%) and SPj (Pi_ref-x%) are stored in the non-volatile memory 20; ), j = l, ..., J in association with the corresponding power variations of the ith cathode (ie +/- x% with respect to Pi_ref).

Note that in the procedure described above, each training stack corresponds to the variation of a single operating parameter at a time, the same for each compartment (ie, it is considered to apply the variations of +/- x % the operating parameters individually and successively so as to individually identify the contribution of each power variation to the optical spectra measured by the optical control system 15). However, it is possible, under another procedure, to simultaneously varying during the deposition of a stack of thin learning layers, several operating parameters of the deposit line of a predetermined quantity. In addition, it has been chosen here, for the sake of simplicity, to vary by the same amount x% upward or downward, the reference powers of the different cathodes of the deposit line. However, it is possible alternatively, to apply a different quantity upward and downward, and to vary this quantity from one cathode to another (x is in this case dependent on n, n = 1, ...,NOT).

Preferably, the variations of +/- x% considered around each reference value Pi_ref of the operating parameters are chosen so as to maintain the hypothesis of linearity which is made, in the first embodiment, of the variations of the measurements. of quality depending on the variations of the operating parameters (ie x is chosen sufficiently important to be distinguished from noise while retaining the hypothesis of linearity made for the mathematical model). This is illustrated in Figures 5A-5C.

FIGS. 5A, 5B, 5C respectively represent observed variations on optical spectra measured in transmission (FIG. 5A) and in reflection (stack side for FIG. 5B and substrate side for FIG. 5C) on a stack of thin layers deposited on a EX1-type glass substrate introduced previously, when the reference power of the first silver layer Agi is increased or reduced by x = 5%. It can be seen from these figures that the spectra observed at the reference power are in the middle of the spectra respectively obtained by varying this reference power by +/- 5%.

From the optical spectra stored in the non-volatile memory 20, the obtaining module 16A determines the linear mathematical model MOD1. More specifically, in the first embodiment described here, the obtaining module 16A constructed from the N optical spectrums SP (Pi_ref + x%) and SP (Pi_ref-x%), i = 1, ..., N stored a matrix M defining the mathematical model MOD1. To this end, the obtaining module 16A evaluates, for each variation of +/- x% of the power Pi_ref of a cathode i of the deposition line, the column vector VARi (x%), i = 1,. .., N including KxJ components and defined by:

VARi (x%) = [SP (Pi _ref + x%) - SP (Pi _ref - x%)}

Then, the obtaining module 16A determines the vector Ui, i = 1, ..., N defined by:

VARI (x%)

Ui = - - -

2x

The obtaining module 16A then constructs the matrix M by concatenating the N vectors Ui, i = 1, ..., N obtained according to:

M = [U1 U2 U3 ... UN]

The matrix M thus obtained is of dimensions (KxJ) × N.

This matrix, in relation to the reference powers and the power variations applied to the reference powers, defines the linear mathematical model MOD1. It should be noted that the person skilled in the art would have no difficulty in deriving such a matrix in the case where different variations +/- x (i)% are applied to the reference powers Pi_ref of each cathode, but also if a different amount is applied upwards and downwards relative to the reference power on each cathode.

The obtaining module 16A stores the matrix M in the non-volatile memory 20 as a mathematical model MOD1.

It is noted that the mathematical model MOD1 can be updated, for example from a new set of training stacks, to take account of the evolution of various factors that can affect the quality of thin film stacks made by the deposit line.

It is now assumed that the deposition line 3 is used to deposit a thin film stack of EX1 type on the glass substrate 2. For this purpose, the operating parameters of the deposition line 3, and more precisely here, the powers Pi, i = l, ..., N of the N cathodes of the deposition line, are set via the parameterization module 14, with so-called current values Pi_curr, i = 1, ..., N (step E20) . The current values Pi_curr can be chosen by an expert or correspond for example to the values used during the last production carried out by the deposit line 3.

Once this setting has been made, the deposition line 3 is activated to deposit a stack of thin layers on the glass substrate 2 (step E30).

Following this deposition, the optical control system 15 measures at the output of the deposition line 3, various quality measurements on the complete thin film stack deposited (step E40). In the example envisaged here, the quality measurements made are J optical spectra (so-called "currents") SP1_curr, ..., SPJ_curr, as already detailed previously (including spectrum measurements for K distinct wavelengths). These optical spectra are provided by the optical control system 15 to the determination device 16 (and more precisely to its second obtaining module 16B) (step E50).

In the example envisaged here, as mentioned above, the quality function considered corresponds to the concatenation of the optical spectra SP1_curr,..., SPJ_curr (quality measurements within the meaning of the invention) in the state. The concatenated quality measurements SPl_curr, ..., SPJ_curr to form the current optical spectrum SP_curr thus constitutes a value of a quality function obtained from a current value (corresponding to J spectra SPl_curr, ..., SPJ_curr) quality measurements within the meaning of the invention.

The determination device 16, via its automatic determination module 16C, then evaluates the difference between the value obtained from the quality function and a target value chosen for this quality function for the thin-film stack produced by the line of measurement. 3. In the example considered here, this difference noted v is a column vector evaluated as follows:

v = SPtarg - SP_curr where SP_targ denotes the target optical spectrum targeted for the thin film stack produced by the deposition line corresponding to the concatenation of the target optical spectra SPj_targ determined for each of the types of optical spectra considered (ie in the example of J = 3 considered here, for the transmission spectrum, for the stack-side reflection spectrum and for the substrate-side reflection spectrum). These target optical spectra SPj_targ can be determined experimentally from a stack of reference thin films having expected characteristics (eg in terms of mechanical and / or optical properties, or thicknesses of the layers, etc.) for this stack. .

The determination device 16 seeks to minimize or at least reduce the difference v found between the current spectrum and the target spectrum to obtain a product at the output of the deposition line according to certain predetermined properties. For this purpose, the automatic determination module 16C determines, using the mathematical numerical model MOD stored in the non-volatile memory 20, the variations of the operating parameters to be applied to the current values of the operating parameters with which the deposit line 3 is set to minimize (if not cancel) the deviation v evaluated (step E60). This determination requires, when the mathematical model MOD is a direct model, an inversion of the mathematical model MOD (this one giving variations of the quality function according to the values of the operating parameters).

In the first exemplary embodiment in which the mathematical model MOD is the linear model MOD1, this inversion consists more particularly for the determination module 16C in determining a vector AAJ of N adjustment values of the current powers Pi_curr, i = l ,. .., N cathodes of the deposit line checking the relation:

AAJ = argmin _ÛP (|| MAP - v ||)

where DR denotes the N-dimensional column vector of the power variations to be applied to the N cathodes (M denoting a matrix having LxJ lines and N columns, DR is therefore an N-dimensional column vector). The quantity || MDR - v || to minimize for determining the vector AAJ of the power adjustment values of the cathodes constitutes a cost function that must be minimized in order to try to correct the residual difference v between the current optical spectrum SP_curr measured on the stack at the output of the deposition line and the target optical spectrum SPtarg desired for this stack.

The automatic determination module 16C here determines the vector AAJ by performing the following deterministic calculation (step E70):

DA] = (M ^r M) ^_1 . M ^T .v

In another embodiment, the automatic determination module 16C can implement a regularization of the inversion. This regularization can be, for example, a regularization of Tikhonov which consists in determining a vector AAJ of N adjustment values of the current powers Pi_curr, i = 1, ..., N of the cathodes of the deposit line satisfying the relation:

DA] = argmin _AP (|| MAP - v || ² + a || DR || ² ) where a denotes a regularization parameter. The vector DA] verifying this relation can be deterministically calculated by the following automatic determination module 16C:

DA] = (M ^G, M + ai) ^-1 . M ^r . v

where I designates the identity matrix.

In an alternative embodiment, the automatic determination module 16C implements a Lasso regularization which consists in determining a column vector AAJ of N adjustment values of the current powers Pi_curr, i = 1, ..., N of the cathodes of the deposit line verifying the relationship:

where a denotes a regulation parameter and DRί the adjustment value applied to the current power of the cathode i.

Note that the regularization parameter applied by the automatic determination module 16C may be a constant or vary over time.

Whatever the cost function minimized by the automatic determination module 16C (including or not a regularization term), the vector of the adjustment values AAJ determined by the determination device 16 is then supplied by the latter to the module parameterization 14 of the deposition line, and more particularly to its estimation module 22. The estimation module 22 estimates, from the adjustment values supplied to it, adjusted values of the operating parameters of the line deposition 3 (step E80). In the example envisaged here, it adds for each operating parameter concerned (ie for which a non-zero adjustment value has been determined by the determining device 16), the adjustment value supplied to the current value of the parameter of operation.

Then the deposition line 3 is parameterized by the parameterization module 14, with the adjusted values of the operating parameters estimated by the estimation module 22 (step E90).

Once this parameterization has been performed, the deposition line 3 is activated to re-deposit a stack of thin layers on the glass substrate 2 (step E100).

Following this new deposit, the deposition system 1 can be activated to repeat the steps E40 to E100 until a difference is reached between the value of the quality function estimated from the quality measurement values obtained by the control system. optical 15 output of the deposition line 3 and the target value chosen for it less than a predetermined threshold.

Alternatively, other stopping criteria may be considered. Thus, the steps of deposition, determination, adjustment and parameterization of the deposition line 3 can be implemented as long as the adjustment values determined for the operating parameters are below a predetermined threshold (that is, as long as the adjustment values are sufficiently small, the values of the operating parameters are adjusted, but if they reach incoherent values greater than the threshold considered, we stop adjustment of operating parameters of the deposit line). One can also consider stopping the adjustment of the operating parameters of the deposit line as soon as it is detected that the product 4 resulting from the deposit line is in accordance with the desired properties, or with properties expected by the user. final product.

Note that in the embodiment described above, the setting of the deposition line 3 is carried out in steps E20 and E90 by means of the parameterization module 14. However, this assumption is not limiting in itself and the The invention also applies when this parameterization is performed for example manually by an operator.

Moreover, before using the mathematical model MOD1, it is possible, in a particular embodiment, to validate the mathematical model MOD1 obtained by the obtaining module 16A, using dedicated tools. Such a tool can for example proceed in a manner similar or identical to that performed by the determination device 16 to determine the adjustment values of the operating parameters from a set of observed test stacks for which we know the actual adjustment values to be applied, and compare the adjustment values obtained using the MOD1 model with these actual values. If the values coincide with a certain reliability, the model MOD1 is validated. In case of invalidating the model, the obtaining module 16A is configured to determine a new mathematical model, for example from a new set of training stacks: this new set of learning stacks can correspond to in particular to take into account new values of variations x to be applied to the operating parameters around the reference values.

In the first embodiment envisaged here, the mathematical model MOD used by the determination device 16 to determine the adjustments to be made to the operating parameters of the deposition line 3 is a direct linear mathematical model MOD1, connecting variations of the spectra. optics measured at the output of the deposition line on the stack of thin layers to the power variations of the cathodes of the deposition line.

As mentioned above, other mathematical models can be envisaged during the implementation of the invention. Typically, other types of models than a linear model can be considered, such as for example a polynomial model of degree greater than 1, or models obtained by using an automatic statistical learning algorithm. The mathematical model considered can also be direct like the MOD1 or indirect model, that is to say, provide values of the operating parameters (eg variations with respect to a reference value) as a function of the values of the quality measurements. (eg variations in quality measures).

By way of illustration, a second exemplary embodiment is envisaged in which the mathematical model MOD used by the determination device 16 is a direct mathematical model MOD2 obtained by using an automatic statistical learning algorithm. Such an algorithm, in a manner known per se, makes it possible to determine the best a function that makes it possible to predict a variable Y from an X variable. As for the first exemplary embodiment, the focus here is more particularly on a mathematical model MOD2 providing variations of the quality measures or quality functions considered ( optical spectral variations provided by the optical control system with respect to reference spectra as a function of variations of the operating parameters considered (eg power variations of the cathodes of the deposition line 3 with respect to power reference of these cathodes). Determining the direct mathematical model MOD2 thus consists in determining here via the automatic learning algorithm considered, the function (noted for example f) which makes it possible to predict the variations of optical spectra from variations in the power of the cathodes.

No limitation is attached to the automatic statistical learning algorithm used to determine the direct mathematical MOD2 model. Given the mathematical model sought, however, preference is given here to the choice of a supervised statistical learning algorithm of the regression type. It can be for example a regression of RIDGE, a regression of LASSO, a regression of the type SVR (Support Vector Regression) or a Random Forest Regressor type algorithm. Of course, other types of statistical learning algorithms may alternatively be considered (eg neural network, etc.). The invention is not limited to linear or polynomial models.

In a manner known per se, this type of algorithm is based on two phases: a learning phase of the mathematical model MOD2 followed by a test phase which aims to validate the model learned during the learning phase. These two phases are implemented from a set of thin film stacks observed (stacks of thin learning layers in the sense of the invention) which are divided into two subsets (for example in a proportion 70% / 30%), one for use during the learning phase, when intended to be used during the test phase. It should be noted that advantageously, the observed thin film stacks used to learn the mathematical model MOD2 have not necessarily been produced for this purpose, that is to say according to a predetermined experimental design as for the model mathematical MOD1. They may have been produced in another context, during previous uses of the deposit line 3 to make thin film stacks (ie they belong to the production history of the deposit line 3), since for these stacks, information is available on the values of the operating parameters applied by the deposition line 3 during their production and the quality measurements measured by the optical control system 15 on these stacks at their output from the deposit line. 3. This advantageously makes it possible to have more easily a large number of stacks observed to learn the mathematical model MOD2. In addition, new stacking observations can be considered at any time to update the MOD2 mathematical model. However, the inventors have found that in order to obtain a more relevant (less noisy) mathematical model MOD2, it can be ensured that the observed thin-film stacks used to determine the MOD2 model are selected to respect a given distribution of variations around values. reference of the operating parameters, for example a uniform distribution on +/- 15%.

Alternatively, other methods for "cleaning" the set of thin film stacks observed to keep only stacks relevant for the determination of the mathematical model MOD2 can be envisaged, such as the elimination of stacks corresponding to obviously aberrant values, the elimination of redundant or generated data during malfunctions of the optical control system 15 (such malfunctions can be detected via expertise tests from the spectral values provided by the optical control system 15, by example if they exceed 100%, or via statistical tests for example by checking if the distribution of the spectral values provided does not conform to a given distribution), etc. In another example, it is possible to keep only the stacks generated on the same day by the deposit line in order to maximize the compatibility between the stacks considered, etc.

The way in which the mathematical model MOD2 is determined from a set of stacks observed is not described in detail here, it is indeed known to those skilled in the art, and depends of course on the algorithm of statistical learning considered.

It should be noted that the use of the direct mathematical model MOD2 by the determination device 16 to determine the adjustment values of the operating parameters of the deposit line requires, as for the direct mathematical model MOD1, a model inversion. It is indeed during steps E60 and E70 to determine, thanks to the mathematical model MOD2, the variations of the operating parameters to be applied to the current values of the latter which minimize the difference v. This is done by the determination module 16C by means of the statistical learning algorithm from the MOD2 model previously learned, in a manner known per se (in the example previously considered of N operating parameters to be adjusted, by reusing the Notations previously introduced for the mathematical model MOD1, the vector DA] of the adjustment values of these N operating parameters is the vector DR minimizing the cost function vf (AP), where f denotes the function modeled by the model MOD2.

As mentioned above, it is also possible for the determination device 16 to use an indirect mathematical model providing values of the operating parameters (eg variations with respect to a reference value) as a function of the values of the quality measurements (e.g. variations in quality measures).

By way of example, a third exemplary embodiment is envisaged in which the mathematical model MOD used by the determination device 16 is an indirect mathematical model MOD3 obtained by using a statistical learning algorithm. automatic. We are interested here more particularly in a MOD3 mathematical model providing variations of the considered operating parameters (eg powers of the cathodes of the deposition line 3) as a function of the variations of the quality measurements or of the quality functions considered (ex. optical spectra provided by the optical control system 15).

No limitation is attached to the automatic statistical learning algorithm used to determine the indirect mathematical model MOD3. The same learning algorithms as for the MOD2 direct mathematical model can be used. The same remarks also apply to the set of observed stacks considered to learn the MOD3 model.

However, it should be noted that, unlike the direct mathematical models MOD1 and MOD2, the indirect mathematical model MOD3 can be used by the determination device 16 to determine the adjustment values of the operating parameters of the deposit line without requiring inversion. The knowledge of the difference v is sufficient to determine directly, from the mathematical model MOD3, the adjustment values of the operating parameters with respect to their current values.

In the example of N operating parameters to be adjusted RI, RN, RN previously envisaged, and L quality measurements available for the stacking made in L measurement points, this can be done typically by determining the adjustment values. operating parameters P1 ... PN which, under constraints H, minimize the cost function fcost defined by:

- q (l) denotes the considered quality measurement measured at the position indexed by I (q (1) can be a vector typically in the case of an optical spectrum measurement, having as many components as measured wavelengths ;

- qtarg is the desired target quality measure;

- dist (a, b) is a selected distance between two quality measures a and b;

- reg (P1, ..., PN) is a regularization of the operating parameters possibly considered;

- f (Pl, ... PN, I) is the effect of the adjustment of all parameters measured at the position indexed by I (it can be assumed that the effect is the same over the entire width or substrate surface or not, for example when the operating parameters considered are the powers of the N cathodes.)

- H are the constraints imposed on the optimization (for example sum of the quantities of gas for each cathode is constant).

Whatever the MOD mathematical model envisaged (MOD1, MOD2, MOD3 or another model), the determination of the adjustment values of the operating parameters of the deposition line 3 by the determination device 16 is based on the following steps : Obtaining the values of the quality function or functions from the quality measurements made at the output of the deposition line 3 on a stack of thin layers deposited by the deposition line while the latter is parameterized with a current value of operating parameters;

Determination (automatic) using the mathematical model considered of the adjustment values of the current values of the operating parameters so as to reduce (ideally to minimize) the difference v existing between the values of the quality function (s) and the values selected targets for these quality functions for the stack of thin layers, this determination may or may not include an inversion of the mathematical model MOD considered.

In a particular embodiment, the MOD2 or MOD3 model will first be realized with the given production history.

Then it will be strengthened to improve its accuracy. It can be strengthened both in the next production of these stacks, and by stacks specially made to strengthen the model. The choice of stacks to be produced to reinforce the model can be chosen either by an expert or by an algorithm. These new stacks can be produced in one or more times.

Similarly, in the case of the MOD1 model, the new x variation values to be applied can be determined by an expert or an algorithm. Stacks of thin layers generated by these variations can also be produced in one or more times.

Whatever the model, the deposit line can be configured or reconfigured either by a human or automatically by software.

Note that the function, denoted here by f, which connects said at least one operating parameter of the deposition line and said at least one quality function defined from said at least one quality measure, can be composed of a plurality of functions or combine a plurality of functions. To determine a numerical model of the function f, it is possible to determine separately for each of the plurality of functions of which it is composed a numerical model, for example by statistical learning, then to determine by means of an ad hoc algorithm the model digital function f.

In the embodiment described here and the various examples considered for the mathematical model MOD, it has been envisaged to adjust a value of a thin-layer operating parameter (in other words by compartment of the deposit line, a compartment single being responsible for the deposition of a given thin layer of the stack), namely the power of the cathode used by this compartment to proceed to the deposition of a thin layer on the glass substrate 2. However, this assumption n ' is not limiting in itself. The invention applies to other operating parameters of the deposition line 3, as already mentioned above (eg composition of gases, gas pressures, etc.). Moreover, only a part of the layers of the stack deposited on the transparent substrate 2 by the deposition line 3 may alternatively be considered through the mathematical model MOD (the other layers being ignored by the model so that the device of FIG. determination does not provide adjustment values of the operating parameters for these layers, such "ignored" layers are for example very thin layers or not having a measurable impact on the stack made, for example layers of type "blockers", etc.), or some layers can be grouped via the MOD mathematical model. Thus, for example, if several cathodes contribute to the deposition of the same layer, it is possible to adjust the sum or the average of the operating parameters corresponding to this plurality of cathodes according to the invention, and then to adjust the parameters of operation of each of the cathodes so as to verify that the sum of the adjustments or the average of the adjustments made correspond to the determined adjustment values.

In one embodiment of FIG. 2, all the compartments of the deposition line are in the same surface treatment machine known under the name of "coater".

Alternatively, the compartments of the deposit line may be distributed in several surface treatment machines, the method thus comprising a passage to the air (or return to the atmosphere) between two machines. In one embodiment, the measurement or measurements made are all performed at the output of the deposition line, that is to say, in the case where several surface treatment machines are used, after the exit of the last machine.

In another embodiment, at least one measurement, called additional measurement, can be performed, in situ, in other words, in the deposition line, before the last thin layer is produced by the last machine. However, in this embodiment of the invention, no further measurement is taken after each thin film production, it being said that a thin film may be obtained in several times, in other words by several compartments depositing the same material.

In the context of the invention, one can use a model specific to each surface treatment machine or a model for several surface treatment machines.

In the same way, you can use a product-specific model or a model for several different products.

In a particular embodiment, the value of the quality function is obtained from quality measurements taken at different points over the entire width or surface of the substrate. This embodiment makes it possible to know the parameter or parameters to be modified (for example to adjust the thickness of at least one layer), for all or some of the thin layers, over the entire width or surface of these layers so as to respect the desired properties at any point of the plateau.

In the case where it is desired to obtain a homogeneous thickness for a given layer, it is possible to control the valves to produce a different gas flow over localized areas of the substrate.

Claims

1. A method for automatically determining an adjustment value of at least one operating parameter of a deposition line (3) for depositing a stack of thin layers on a transparent substrate (2), this method being implemented by computer and comprising:

A step of obtaining (E10) a mathematical model (MOD) connecting at least one operation parameter of the deposit line with at least one predetermined quality function defined from at least one quality measurement of a stack of thin layers deposited by the deposit line;

A step of obtaining (E50) a value of said at least one quality function from a value of said at least one quality measurement measured at the output of the deposition line on a deposited thin-film stack; by the deposition line on a transparent substrate while the deposition line was parameterized with a current value of said at least one operating parameter; and

An automatic determination step (E80) by means of the mathematical model obtained of an adjustment value of the current value of said at least one operating parameter making it possible to reduce a difference existing between the value obtained from said at least one function of quality and a target value chosen for the at least one quality function for thin film stacking.

2. The method of claim 1 wherein the stack comprises a plurality of thin layers, in which one adjusts several operating parameters associated with several thin layers.

3. Determination method according to claims 1 or 2, wherein the value of the quality function is obtained from quality measurements taken at different points over the entire width or surface of the substrate.

4. Determination method according to any one of claims 1 to 3, wherein the mathematical model (MOD) relates at least one variation of said at least one operating parameter with at least one variation of said quality function.

5. Determination method according to any one of claims 1 to 4 wherein said deposition line (3) implements a sputtering technique for depositing the stack of thin layers on the transparent substrate.

6. Determination method according to claim 5 wherein said cathode sputtering is assisted by magnetic field.

7. Determination method according to any one of claims 1 to 6 wherein said at least one quality measure comprises at least one optical spectrum measurement of the thin film stack.

The method of determining according to claim 7 wherein said at least one quality measure further comprises at least one of:

A measurement of an electrical property of the stack of thin layers;

A measurement of a mechanical property of the stack of thin layers;

A measurement of a chemical property of the stack of thin layers; and

A measurement of a mass property of the stack of thin layers.

9. Determination method according to any one of claims 1 to 8 wherein said deposition line comprises a plurality of compartments in which said thin layers are deposited on the transparent substrate and said at least one operating parameter comprises:

At least one power or voltage applied to a cathode of a said compartment of the deposition line; and or

At least one rate of movement of the substrate opposite a cathode of a said compartment of the deposit line; and or

10. Determination method according to any one of claims 1 to 9 wherein the step of obtaining the mathematical model comprises:

Obtaining, for a plurality of stacks of so-called learning thin films deposited by the deposition line on a substrate when said deposition line is parameterized with values of said at least one operating parameter having variations determined with respect to at said reference value, a plurality of values of said at least one quality function; A step of determining the mathematical model from the plurality of values of said at least one quality function obtained for the plurality of stacks of thin learning layers.

11. Determination method according to any one of claims 1 to 10 wherein:

The mathematical model is a polynomial model providing variations in the value of said at least one quality function as a function of variations in the value of said at least one operating parameter; and

12. Determination method according to claim 11 wherein the step of inverting the polynomial model comprises a regularization of the inversion.

13. Determination method according to claim 10 wherein the deposition line comprises N compartments, N designating an integer greater than 1, each compartment being respectively associated with an operating parameter P (n), n = 1, ..., N, and the mathematical model is a linear model using a matrix denoted M comprising N columns u (n), each column being defined by:

var (n)

where var (n) denotes the difference between the values of the at least one quality function obtained for two stacks of thin learning layers deposited by the deposition line while the latter was parameterized with a value of the operating parameter P (n) having a variation + _< i (n) with respect to the reference value of the operating parameter P (n) and with a value of the operating parameter P (n) having a variation - (n) with respect to the reference value of the operating parameter P (n).

14. The method of determining according to claim 13 wherein the adjustment value a of said at least one operating parameter is given by:

a = (M ^T MY ^x M ^T v

where v denotes the difference between the value obtained from the quality function and the target value chosen for the quality function for the stack of thin layers and where M ^T denotes the transposed matrix of the matrix M.

15. Determination method according to any one of claims 1 to 9 wherein the mathematical model is a polynomial model providing variations in value. said at least one operating parameter as a function of variations in the value of the quality function.

16. A method of determining according to any one of claims 1 to 11 or 15 wherein said mathematical model is obtained using an automatic statistical learning algorithm.

17. Determination method according to any one of claims 1 to 16 further comprising a step of validating the mathematical model obtained.

Computer program (PROG) comprising instructions for performing the steps of the method of determining according to any one of claims 1 to 17 when said program is executed by a computer.

19. A computer-readable recording medium on which a computer program is recorded according to claim 18.

20. A method of parameterizing a deposition line (3) capable of depositing a stack of thin layers on a transparent substrate (2), this method being implemented by computer and comprising:

A first parametering step (E20) of the deposit line with a current value of at least one operating parameter of the deposit line;

A deposition step (E30) by the deposition line parameterized with said current value of a stack of thin layers on a substrate;

A step of automatically determining an adjustment value of the current value of said at least one operating parameter implementing an automatic determination method according to any one of claims 1 to 14 and using a value of said at least one a measurement of quality measured at the output of the deposition line on the stack of thin layers deposited during the deposition step;

A step of estimating (E80) an adjusted value of said at least one operating parameter from the current value and the determined adjustment value; and

A second parameterization step (E90) of the deposition line with the adjusted value of said at least one estimated operating parameter during the estimation step.

21. Parametrization method according to claim 20 wherein the steps of deposition, determination, adjustment and the second parametering step are implemented as long as: The difference between the value of the quality function estimated from said measured value at the output of the deposit line and the target value chosen for said quality function is greater than a predetermined threshold, or

Apparatus for automatically determining (16) an adjustment value of at least one operating parameter of a deposition line for depositing a stack of thin layers on a transparent substrate, said device comprising:

A first obtaining module (16A), configured to obtain a mathematical model connecting at least one operation parameter of the deposit line with at least one predetermined quality function defined from at least one quality measurement of a stack of thin layers deposited by the deposit line;

A second obtaining module (16B), configured to obtain a value of said at least one quality function from a value of said at least one measured quality measurement at the output of the deposition line on a stack of thin layers deposited on a transparent substrate, said deposition line having been configured by means of a so-called current value of said at least one operating parameter; and

A determination module (16C), configured to automatically determine, by means of the mathematical model obtained, an adjustment value of the current value of said at least one operating parameter making it possible to reduce a difference existing between the value obtained from said at least one quality function and a target value chosen for said at least one quality function for thin film stacking.

23. A system for depositing (1) a stack of thin layers on a transparent substrate (2) comprising:

A deposition line (3) capable of depositing a stack of thin layers on a transparent substrate;

- a parameterization module (14) of the deposit line, configured to set the deposit line with a current value of at least one operating parameter of the deposit line;

An automatic determination device (16) according to claim 19 of an adjustment value of the current value of said at least one operating parameter configured to use a value of at least one measured quality measurement at the output of the line deposition on a stack of thin layers deposited on a transparent substrate by the deposition line; An estimation module (22) configured to estimate an adjusted value of said at least one operating parameter from the current value and the adjustment value determined by the determination device;

the parameterization module (14) being configured to parameterize the deposition line with the adjusted value of said at least one operating parameter estimated by the estimation module.