Method for coating substrates in inline installations

A method for coating substrates in inline installations, in which a substrate is moved through at least one coating chamber and during this movement is coated. In this method, first, a model of the coating chamber is formed which takes into consideration the changes of the chamber parameters caused by the movement of the substrate through the coating chamber. Subsequently, the particular position of the substrate within the coating chamber is acquired. The chamber parameters are subsequently set based on the position of the substrate according to the model of the coating chambers.

Skip to: Description  ·  Claims  · Patent History  ·  Patent History
Description
BACKGROUND AND SUMMARY OF THE INVENTION

This application claims priority from German Patent Application No. 10 2004 020 466.7 filed Apr. 26, 2004 hereby incorporated by reference in its entirety.

The invention relates to a method for coating substrates in inline installations, in which a substrate is moved through at least one coating chamber and during this movement is coated.

Such inline installations consist of several coating chambers, which are serially arrayed and in which substrates are preferably coated by means of a sputter process.

In this process the substrates are moved past sputter arrangements while being coated continuously, such that a layer of specific thickness is produced.

The sputter arrangements are each located in coating chambers, which are connected with adjacent chambers via connection or transport channels. As a rule, the vacuum conductivity of the transport channels is low compared with the vacuum conductivity of a coating chamber in the transverse direction. The adjacent chamber can be further coating chambers or pump compartments.

The pump compartments establish differential pumping stages between the coating chambers such that between the coating chambers a certain gas separation is ensured, which, however, is not complete. This approach is important to an economical operating process.

During the transport of the substrates through the coating chamber specific chamber parameters, in particular the vacuum conductances of the connection channels and effective evacuation capacity of pump compartments are characteristically changed. Based on the machine parameters of the installation, this is indicated by characteristic fluctuations of the partial gas pressures as well as of the cathode voltage; the latter causes a coating rate which fluctuates over time. Hereby the coating of the substrate is not uniform in the direction of movement. Especially in coating large-area architectural glass, this is of disadvantage since the optical properties of these coatings strongly depend on the thickness of the layer.

In order to be able to predict the changes of the vacuum conductances and of the effective evacuation capacity, it is possible to draw on the Monte Carlo method. This is a stochastic computing method for the numerical determination of approximation solutions of several mathematical problems, especially problems of theoretical probability. The macroscopic values to be calculated for the gas flow simulation, such as, for example, pressure and temperature, are here obtained as mean values from microscopic variables such as particle speeds, which are obtained from a multiplicity of representative gas particles within the scope of a stochastic movement and collision model. The effective application of the Monte Carlo method within this scope only became possible through fast-processing computers.

The Monte Carlo simulation of gas flows by means of stochastic movement and collision models for model molecules is already known (G. A. Bird: Recent Advances and Current Challenges for DSMC, Computers Math. Applic. Vol. 35, No. 1/2, 1998, pp. 1-14).

It is furthermore known to create a macroscopic idealized model for reactive sputtering, for example for the reactive sputtering of Ti in a gas mixture of argon and nitrogen (S. Berg, H.-O. Blom, T. Larsson, and C. Nender: Modeling of reactive sputtering of compound materials, J. Vac. Sci. Technol. A5 (2), March/April 1987, pp. 202-207).

This model describes reactive sputtering by means of the reactive gas flow to the target and to the substrate, by means of sputter yields with pure and oxidized target, by means of adhesion coefficient of the reactive molecules with respect to the substrates and the target, and by means of the ion stream to the target surface. The principal factors of reactive sputtering, such as hysteresis, effects of pumping rate, etc., are herein taken into consideration, however, not the plasma physics and chemistry of the glow discharge.

Plasma chemistry, in contrast, was taken into consideration by Ershov and Pekker (Model of d.c.magnetron reactive sputtering in Ar—O2 gas mixtures (Thin Solid Films 289, 1996, pp. 140-146).

It is also known to obtain with the aid of Monte Carlo simulation methods information about the spatial distribution of flows of sputtered particles in a sputter chamber as well as about their distribution of the kinetic energy when they impinge on the substrate (Karol Macàk, Peter Macàk, Ulf Helmersson: Monte Carlo simulations of the transport of sputtered particles, Computer Physics Communications 120, 1999, pp. 238-254).

The simulation of the kinetics of reactive sputtering in inline chambers is also known (A. Pflug, N. Malkomes, V. Sittinger, and B. Szyszka: Simulation of Reactive Magnetron Sputtering Kinetics in Real In-Line Processing Chambers, Society of Vacuum Coaters, 45th Annual Technical Conference Proceedings, 2002, pp. 16-21). Herein the complex volume of a process chamber is divided into simple volume cells, for example into parallelepipeds. In each cell homogeneous partial pressures for inert and reactive gas as well as homogeneous oxidation states per surface are assumed, and the processes of sputtering and deposition are described in conjunction with simple rate equations corresponding to the model by Berg et al. Through three-dimensional crosslinking of the volume elements, the process dynamics can also be described in complex geometries with the aid of coupled differential equations and can be calculated by means of adaptive Runge-Kutta time step methods.

Gas and sputtered particle transports between the volume elements are first calculated in detail with the aid of parallel Monte Carlo methods. The vacuum conductances obtained herefrom and the particle distribution matrices are later substituted into the kinetic macroscopic model. Hereby a high fidelity to detail in three-dimensional installations is obtained, connected with a calculation efficiency, which makes the system usable for real-time applications.

With the aid of an expanded Berg model, a heuristic dependency among sputter power, target voltage and ion current, a realistic model with respect to the relationship of deposition parameters in metallic mode have also been already developed. (A. Pflug, B. Szyszka, V. Sittinger, and J. Niemann: Process Simulation for Advanced Large Area Optical Coatings, 2003, Society of Vacuum Coaters, 46th Annual.Technical Conference Proceedings, 2003, pp. 241-247). This model permits tying the simulation to machine parameters such as target voltage and sputter power as well as further the simulation of inhomogeneous sputter conditions by means of segmenting long targets.

Several of the above listed simulation methods were also already combined in simulation software (A. Pflug, B. Szyszka, J. Niemann: Simulation des reaktiven Magnetron-Sputterprozesses in Inline-Anlagen {Simulation of the reactive magnetron sputter process in inline installations}, JOT 1, 2003, pp. X-XIII). Here with the aid of Monte Carlo methods on a Linux cluster gas flow kinetics as well as distribution of the sputtered particles on interior chamber surfaces was calculated three-dimensionally. The above cited Bird method for gas flow simulation as well as the method by Macàk et al., also mentioned above, regarding the simulation of particle trajectories were implemented as parallel algorithms in C++ under the parallel environment “parallel virtual machine”.

The Linux cluster consists of several connected PCs, which operate under the operating system Linux. “Cluster” means first that several processors calculate a problem in parallel. In practice, a cluster is often built from commercially available PCs for reasons of cost, and the best cost/benefit ratio is often obtained with dual board PCs. The PCs are networked with one another via a separate network, in order to be able to handle the necessary communication between the subprocesses during the calculation.

The above described macroscopic dynamic model for calculating the time-resolved process dynamics in three-dimensional inline installations was also implemented within the framework of this software system. This model is suitable for real-time applications and imports the data obtained from Monte Carlo calculation in the form of flow conductances for the gas flow kinetics as well as distribution matrices for the sputtered particles.

Further simulation methods for the sputtering process are also known (U.S. Pat. No. 5,751,607, U.S. Pat. No. 6,070,735, KR 2000023859, JP 10294293 A, JP 2003282381 A, U.S. Pat. No. 6,512,998 B1). However, all of them deal with methods for the simulation of sputtered neutral particles. The effect of the neutral particle distribution on the macroscopic sputter process kinetics is, however, vanishingly small and therewith of secondary importance for the present invention. Layer thickness regulations are known from U.S. Pat. Nos. 6,425,988, 6,524,449 and 6,668,207.

The invention has as its aim to provide a method with the application of which the coating of a substrate has a uniform thickness in the direction of movement of the substrate through the coating chamber.

This aim is attained through the present inventin.

With the invention a model of the coating installation is used, which exactly describes the receptacle, including the volume form and the surface form as well as all three-dimensionally resolved partial pressures and partial flows, as they are developed spatially and in time corresponding to the sources and sinks of all chemical products, including the electrons and ions. In addition, a model of the sputter and deposition process is formed.

The advantage achieved with the invention consists in particular therein that during the coating process also parameters can be taken into consideration, which are not accessible to direct measurement during the coating process, for example the sputter rates at the coating targets or the mean deposition rate on the substrate. While the layer thickness can be measured with the aid of in situ spectroscopy, however, for the determination of the coating rate at least two layer thicknesses must be determined at two points in time. Because of the substrate movement, this is nearly impossible unless the optical systems are also moved along in the direction of movement. Through simulated virtual regulation circuits, thus either the sputter rates or the mean coating rate on the substrate can be kept constant. If the sputter rates of, for example, two targets are successfully kept constant through a correction protocol, it would also yield a constant coating rate.

In the application of the regulation or correction protocol from the virtual regulation circuit onto the real sputter process, an improved uniformity results of the coating thickness in the direction of the movement of the substrate. In the simulation the coating rate on the substrate is kept constant by means of a regulation circuit. This yields a correction protocol, for example for the gas flow.

For a uniform layer thickness on the substrate it is sufficient to keep all relevant operating conditions of a coating target and receptacle constant in time. By “operating conditions” are here understood the gas pressures between target and substrate as well as the sputter rates of the targets, which are not directly accessible from the outside. For example, the pressure is not measured directly between target and substrate, but rather at an arbitrary site on a receptacle wall. Tests in coating systems consisting of several targets, for example a double cathode, have shown that, as a rule, constancy in time of the operating conditions of all targets cannot be achieved with a simple regulation protocol. The sputter rates of the individual targets show characteristic time courses, which in the application of the regulation protocol on the substrate, are on average compensated to a uniform coating rate. With more complicated correction protocols, for example two independent correction curves for a right and a left gas inlet, it would also be possible to keep the sputter conditions at two targets constant.

Detailed compensation of all evacuation performances and vacuum conductances changed through the substrate movement would be independent of the sputter process and could, once the corresponding regulation or correction protocol has been calculated for a specific installation geometry, be applied independently of the target material and operating point of the sputter process.

However, in the case of coating sources with complex structure, for example the above cited double cathode, a simple regulation protocol rather causes a further additional fluctuation in specific process parameters, with the aid of which an initially present layer thickness inhomogeneity on the substrate in the direction of movement of all particle flows is on average compensated. In these cases the regulation protocol also depends on the material and operating point of the sputter process. The sputter model in these cases can be calibrated for a target material to a specific parameter range of the sputter process and can output in short time the correspondingly required regulation protocol within this range.

In the present invention are contained the detailed compensation of all operating states within the coating chamber as well as also the indirect compensation of the layer thicknesses inhomogeneity by means of a further time-dependent fluctuation in specific process parameters.

An embodiment example of the invention is shown in the drawing and will be described in the following in further detail.

BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 shows a cross section through a sputter chamber,

FIG. 2 shows a schematic illustration of an employed model of a sputter chamber,

FIG. 3 shows a detailed illustration of a simulated virtual control without in situ process regulation,

FIG. 4 shows a detailed illustration of a simulated virtual control with in situ process regulation.

DETAILED DESCRIPTION

FIG. 1 shows a sputter chamber 1, which comprises the coating chamber 2 proper and two buffer chambers 3, 4. Adjoining this sputter chamber 1 can be on the right and/or the left further sputter chambers, which are not shown here. A substrate 5 is transported from the left to the right via transport rollers 6 supported in a carrier 7. Above each buffer chamber 3, 4 is located a pump chamber 8, 9, and above each pump chamber 8, 9 a pump 10, 11 is disposed. The pumps 10, 11 are turbo pumps with a fixedly specified nominal rotational speed and at this speed have a fixed evacuation capacity. In the model selected here, the fixed evacuation capacity of pumps 10, 11 enters in as a fixed value. Between the pumps 10, 11 is located an installation cover 12, on whose underside a cathode mounting 13 is fastened which bears a cathode 14 with a target 15. An anode 16 located beneath the target 15, is fastened on a mounting 17, which incorporates a cooling system 18 and via an insulation 19 is connected with a wall 20 of the coating chamber 2. Adjacent to the anode 16 are provided supply lines 21 for sputter gases. In a cathode covering hood 22 are provided cathode cooling water tubes 23, 24, which serve for the forward and return flow of cooling water. By 25 is denoted the cathode terminal. A slot lock 26 connects the coating chamber 2 with the buffer chamber 4.

Although FIG. 1 shows only one cathode 14 or one target 15, respectively, the invention can also be applied in installations with two or more targets.

By 37 is denoted a pressure sensor, which is connected across a line 27 with a control 28, which includes a model of the sputter chamber 1. The gas pressure in the coating chamber 2 is correspondingly controlled via control lines 29, 30 and valves 31, 32 as well as the cathode-anode voltage across lines 33, 34. The mean pressure P in coating chamber 2 is calculated from the ratio of gas inflow [rate] F to the effective evacuation capacity Seff of all connected pumps:
P[mbar]=0.01812*F[sccm]/Seff[1/s]

    • where sccm=standard cubic centimeter per minute and
    • 1 sccm=0.01812 mbar 1/s.

The position of the front edge 35 of substrate 5 is either continuously measured or calculated. If it is calculated, it is required to record the point in time of the introduction of the substrate into the sputter chamber 1 or coating chamber 2. The position value of substrate 5 is reported across a control line 36 to the control 28.

Based on the particular entered position of the substrate 5 and with the aid of the correction function determined from the model of the sputter chamber 1 in connection with a software regulation circuit, the control 28 can now be set such that uniform coating takes place in the direction of movement of the substrate 5.

FIG. 2 shows a schematic illustration of the utilized model of a sputter chamber. This model does not refer to FIG. 1, but rather to a chamber with several targets.

The dynamic macroscopic model 40 of the reactive sputtering, on which the invention is based, calculates the interaction between the sputtering process and the glass flow kinetics of the receptacle in a virtual sputter installation, which represents a subregion of the real sputter chamber 1 according to FIG. 1. This subregion comprises a number of M virtual sputter targets 41 to 43 as well as a portion of the receptacle volume in the sputter region, which is represented in the simulation in the form of 44, 45, 46, 47. In the illustration of a real sputter installation according to FIG. 1, the special case of only one sputter target 15 is shown. Model 40, however, permits an arbitrary number of virtual sputter targets 41 to 43, in order to be able to simulate therewith for example also the behavior of double cathodes, etc.

By 55 to 58 are denoted surfaces which are to be sputtered. These can be substrates or chamber walls. The reference numbers 59, 62, 65 indicate the sputtered-off particles, which are subsequently distributed over branches 60, 61 and 63, 64, respectively.

The gas transport between cells 44 to 47 within this subregion is characterized by flow conductances 48 to 50, in contrast, gas inflow and gas outflows are characterized at the boundaries of this subregion by effective evacuation capacity 51 to 54. In contrast to the effect evacuation capacity, a flow conductance S44/45 denotes the ratio of a net gas flow F44/45 between two cells 44, 45 and their pressure difference p44-p45:
S44/45[1/s]=0.01812*F44/45 [sccm]/(p44-p45) [mbar].

The coefficients 48, 49, 50 of the flow conductances and 51, 52, 53, 54, of the effective evacuation capacity are determined with the aid of the so-called “Direct Simulation Monte Carlo” (DSMC) method.

In the case of N cells 44 to 47 the totality of all coefficients 48, 49, 50 forms a symmetric N*N matrix. In the present example for the DSMC method separate software is available, which is embodied as a parallel algorithm and which can calculate on a Linux cluster a realistic three-dimensional pressure and flow profile of receptacles through which flows gas. This method must be repeated for different positions of the substrate 5, such that the coefficients 48 to 50 and 51 to 54 are transferred to the dynamic model 40 in the form of functions of the substrate position.

To the virtual sputter targets 41 to 43 are assigned coefficients which, for example, correspond to the gettering areal fractions of the target surface of target 41 in cell 44. The totality of all coefficients forms an M*N matrix. Further, with each cell 44 to 47 is associated a substrate surface 55 to 58. The fraction of the sputtered material from target 41 to 43 on substrate surface 55 to 58 is determined by the coefficients 60, 61, 63, 64, 65, which in their totality can also be presented as an M*N matrix. To determine the coefficients 60 to 65, a further Monte Carlo method is used for the simulation of trajectories of sputtered neutral particles (Karol Macàk, Peter Macàk, Ulf Helmersson: Monte Carlo simulations of the transport of sputtered particles, Computer Physics Communications 120, 1999, pp. 238-254). This model is also available as an external software model, which is not shown in FIG. 2.

Within each cell 44 to 47 of the dynamic model 40 now one sputter model analogous to Berg et al. (S. Berg, H.-O. Blohm, T. Larsson, and C. Nender: Modeling of reactive sputtering of compound materials, J. Vac. Sci. Technol. A5(2), March/April 1987, pp. 202-207) is utilized. The coupling of cells 44 to 47 is accordingly given by the flow conductances 48 to 50 and targets 41 to 43 with their areal fractions and sputter fractions. Each cell 44 to 47 can be connected with a gas inlet 66 to 69. By means of the coefficients 51 to 54 either directly connected pumps or connections to adjacent sputter chambers, which are not directly calculated in the dynamic model 40, are represented. The entire model 40 can lastly be comprehended as a system of coupled, nonlinear, time-dependent differential equations, for the solution of which standardized numerical methods are available (for example J. R. Cash, A. H. Carp, ACM Transactions on Mathematical Software 16 (1990), p. 201 ff).

The original Berg model contains no mechanism for calculating the target voltage. This is seen as being constant at constant power. In the present invention the secondary electron yield of the target material was introduced as a function of the oxidations state, whereby the target voltage can be calculated as a function of the power and other process parameters.

In order to be able to model correctly the total pressure dependence of process characteristic [line]s, in addition a loss mechanism of argon ions on target-proximate surfaces—for example chamber wall, diaphragms, etc.—was introduced. Especially at low total pressure, consequently a portion of the generated argon ions does not contribute to the sputter process but rather is absorbed by other target-proximate surfaces.

A more precise analysis of the process characteristics showed further that it is unavoidable for a correct modeling to include in the model the voltage or energy dependence of the sputter yield. This has so far not been taken into account in prior art.

In FIG. 3 is depicted a more detailed illustration of the virtual regulation circuit for the stabilization of the coating rate as well as of the connected control for the case without in situ process regulation.

For the correction of layer thickness fluctuations first the case is examined, in which the selected operating point of the sputter process does not require any further in situ process stabilization. In this case in the control 28 of a sputter installation a tabulated or parametrized compensation function 70 is integrated, which depends on the position of the glass substrate 5, the glass position being read out across the control line 36. With the aid of the correction function 70 the sputter process in the sputter installation 1 is modeled as a function of the substrate position. In the embodiment example this takes place either through a variable gas flow, which is transferred across the control lines 29, 30 or via a variable discharge characteristic, which is transferred as total power, current or voltage across lines 33, 34.

In order to obtain a correction or compensation function 70 suitable for the minimization of layer thickness fluctuations, in the preliminary stages the dynamic model 40 of the sputter process of the sputter installation 1 is compiled and this model is connected to a virtual regulation element 72 within a virtual regulation circuit 71. The regulation element 72 receives the modeled mean dynamic coating rate 73 on the substrate as the regulated variable as well as an installation parameter 74, such as for example gas flow or discharge power, as the correcting variable. The virtual regulation element 72 effects first a mean coating rate 73, uniform in time, in the simulation.

With sufficiently good agreement between model 40 and real installation 1, the time course of the correcting variable 74 can be used as correction function 70 for the real model, such that the real coating on the substrate 5 also receives a uniform layer thickness profile.

FIG. 4 depicts a more detailed illustration of a virtual regulation circuit for the stabilization of the coating rate as well as the connected control for the case of an in situ process regulation.

Some case may require that the real sputter installation 1 be equipped with a regulator 76, with which the partial pressure of the reactive gas, measured by means of pressure sensor 37 and transferred across the control line 27, is kept constant. As the correcting variable serves here either the discharge power, transferred across the lines 33, 34, or the inert or reactive gas flow, transferred across the lines 29, 30. Such so-called stabilizations of the operating point are always required if the operating point lies in the so-called unstable transition range of a sputter process characteristic.

In this case a correction function 70 for minimizing layer thickness fluctuations are applied, instead of on installation parameters, on the operating point 77 of the regulator 76. The substrate position x is here, again, read in from the control across line 36. In order to obtain the correction function for this purpose, on the simulation plane a double, virtual regulation circuit 78 is built: to the simulation model 40 of the sputter installation a virtual regulation element 79 for in situ process stabilization is connected. The virtual regulation element 79 ideally incorporates the same regulation algorithm as the real regulation element 76. Herewith the simulated reactive gas partial pressure 81 is regulated to the nominal value 82 from a virtual pressure sensor by variation of a process parameter 64, for example discharge power or gas flow.

The nominal value 82 is now regulated by means of a further regulation element 83 such that the mean dynamic coating rate 73 on the substrate is stabilized to a nominal value 75. The correcting variable of the regulator 83 corresponds to the nominal value 82 of regulator 79. The time course of the correcting variable after the simulation is again transmitted as tabulated correction function 70 into the control 28 of the real installation.

Claims

1-17. (canceled)

18. A method for coating substrates in inline installations, in which a substrate is moved through at least one coating chamber and during this movement is coated, comprising:

a) forming a model of the coating chamber which takes into consideration those changes of chamber parameters, which are caused by the movement of the substrate through the coating chamber,
b) acquiring the position of the substrate within the coating chamber is acquired,
c) setting the chamber parameters are set based on the position of the substrate according to the model.

19. The method as claimed in claim 18, wherein the model determines a correction function for one or several chamber parameters as a function of the position of the substrate.

20. The method as claimed in claim 18, wherein the correction function is transmitted into a control of the coating installation.

21. The method as claimed in claim 18, wherein the dimensions of the substrate are entered into the model.

22. The method as claimed in claim 18, wherein a mean dynamic coating rate is determined on the basis of the model and the determined coating rate is entered as regulated variable into a virtual regulator, which, by variation of gas flow of discharge power, keeps constant the mean dynamic coating rate.

23. The method as claimed in claim 22 wherein the variation of the gas flow or of the discharge power is transmitted as correction function into the control.

24. The method as claimed in claim 18, wherein a mean dynamic coating rate is determined based on the model and the determined coating rate is entered as regulated variable into a virtual regulator, which, by variation of a regulated variable of a further regulator for the in situ stabilization of the operating point, keeps constant the mean dynamic coating rate.

25. The method as claimed in claim 24, wherein the determined variation of the regulated variable of the second regulator is transmitted into the control as correction function.

26. The method as claimed in claim 18, wherein the chamber parameters are the cathode voltage, gas flow, temperature and gas pressure.

27. The method as claimed in claim 18, wherein the substrate is a glass plate.

28. The method as claimed in claim 18, wherein the acquired position of the glass plate is the front edge.

29. The method as claimed in claim 18, wherein for building the model of the coating chamber the Monte Carlo method is utilized.

30. The method as claimed in claim 18, wherein the coating of the substrate takes place by sputtering.

31. The method as claimed in claim 30, wherein the sputtering process is represented by means of Berg's model.

32. The method as claimed in claim 18, wherein the coating chamber is divided into several subvolumes and Berg's model is applied in each of these subvolumes.

33. The method as claimed in claim 18, wherein the model building of the Monte Carlo method and the model building after Berg are mathematically coupled and a coupled system of time-dependent differential equations is formed, from which can be taken the regulated variables for the layer thickness fluctuations minimization.

34. The method as claimed in claim 18, wherein several coating chambers are arrayed in series and the process parameters for all coating chambers are represented through a model.

Patent History
Publication number: 20050236276
Type: Application
Filed: Nov 24, 2004
Publication Date: Oct 27, 2005
Inventors: Albert Kastner (Hanau), Michael Geisler (Wahtersbach), Thomas Leipnitz (Alzenau), Jurgen Bruch (Nidderau), Andreas Pflug (Hannover), Bernd Szyszka (Braunschweig)
Application Number: 10/996,812
Classifications
Current U.S. Class: 205/137.000; 205/138.000