MIXED SYSTEM AND COMPONENT LEVEL T-CAD FOR MICRO FABRICATION. TU Braunschweig, Germany

Transcription

1 Cannes-Mandelieu, 5-7 May 2003 MIXED SYSTEM AND COMPONENT LEVEL T-CAD FOR MICRO FABRICATION U. Hansen 1, C. Germer 2, S. Büttgenbach 1, H.J. Franke 2 1 Institute for Microtechnology, 2 Institute for Engineering Design, TU Braunschweig, Germany ABSTRACT In this paper two T-CAD systems are presented, one dealing with the design of micro devices on the system level and allowing checking of process sequences using rules comprising expert knowledge. The other one is a workflow-based system for the layout of micro components fabricated by wet chemical etching of silicon, thus providing a component level view to the design process. It is shown how both systems are integrated providing the designer with concurrent access to system and component design. 1. INTRODUCTION The design of a micro device is excessively more complex than common in conventional engineering or even micro electronics, the latter being the origin of many technologies still used for MEMS. The reason for this is the possibility to produce three-dimensional structures of almost arbitrary shape. These structures are not standardized and therefore the design of a new micro device usually demands a completely new layout for the components used to build it. In micro electronics on the other hand the designer can choose from a set of standard components to assemble the device. These standardized components for e.g. integrated circuits are commonly silicon based and of a planar buildup. Well-understood and compatible fabrication processes allow for a multitude of applications by combining a rather small amount of standard components. In MEMS design, however, the fabrication processes and materials to be used are not predetermined and differ considerably with varying applications. The process definition and choice of materials is therefore an essential part to the MEMS design process [1,2]. The majority of T-CAD applications available today can be classified by two major characteristics. The first group focuses primarily on the behavioral simulation thus pursuing a so-called top-down approach to the design of the micro device. Opposed to dealing with the complete micro system at once the second group of T-CAD software concentrates on the simulation of individual processing technologies. This is a differing approach to the design, since single detached fabrication steps are considered ( bottom-up ). While engaged with system design the engineer has to bear in mind both views switching iteratively between them in order to obtain a micro system, which fulfills the intended functions as well as being feasible in fabrication. Hence, systems allowing a straightforward data interchange are desirable. The following chapters will introduce two T-CAD softwares developed at our institute - one providing system level, the other component level view of the micro device - and show their integration into a workflow based design environment. 2. VALIDATION OF PROCESSING SEQUENCES As stated not only the actual layout of the micro device needs to be designed, but also the requirements for its fabrication like the choice of processing and the used materials have to be defined. Each fabrication process and each material is equipped with a number of constraints concerning their combination, which have to be regarded when setting up a processing sequence. Hence, in the field of micro technology engineers need to possess a considerable amount of knowledge about the different processing technologies in order to define an accurate processing sequence flawlessly. Attempts have been made to facilitate this task using inference based computer software, whereupon the main concept is to formulate design rules comprising expert knowledge gained in process development. These rules facilitate evaluating the feasibility of the planned processing. The common incompatibilities found to occur can be roughly grouped into three categories [3,4]: Firstly, typical incompatibilities may arise from the use of materials, which may be affected in their chemical or mechanical properties by the following processing - or even not withstand it at all. As an example of a case, where the chemical properties of a material are insufficient, may serve the surface roughness of a microtribological layer (fig. 1a). Its quality might be affected by an etch process for structuring the substrate, which is carried out subsequently. A slight susceptibility of the tribological material of being etched may therefore make the process sequence impractical. A solution to this problem might be the insertion of an additive process for the creation of a protective coating on the tribological layer, /03/$ IEEE

2 a. b. Fig. 1: Examples for process incompatibilities. which needs to be removed afterwards in an additional step. However, maybe the simple rearrangement of the process sequence placing the etch step prior to the production of the tribological layer may be a possible remedy as well. As a second kind of incompatibility certain manufacturing processes may not be carried out successively without inserting proper treatment of the micromechanical part in between the processing. These incompatibilities may occur because of e.g. chemical residues caused by the preceding processing (fig. 1b). Such problems usually will be easily solvable by inserting an appropriate cleaning step. Other difficulties belonging to this kind of incompatibility may arise from insufficient adhesion of adjacent material layers, etc., which may be overcome by inserting additional treatment. Lastly, the third main kind of constraints to be evaluated is the feasibility of generating the intended geometry using the specified fabrication process. This rather classical demand is already handled by a range of stand-alone software tools for process simulation. The described incompatibility types are not always clearly separable and it can sometimes not be stated explicitly, what the cause for the incompatibility is at all. Frequently mixtures of several cases may take place Rules - main principles and examples Considering the above, rules only allow for a thorough validation of the defined sequence, if they contain expert knowledge on all the processing chosen and are above that highly flexible in their usage. To obtain this desired flexibility they have to be phrased generally applicable. This means, they should on the one hand apply to as many fabrication processes as possible, on the other hand their evaluation should rely on as little additional data as achievable. The following examples will clarify these preconditions and describe, how incompatibilities are handled in our process sequence checker RUMTOPF [5]. In the case of depositing a material layer on the micro component using electroplating it is essential to be certain reactions 2Cu + O 2 2CuO CuO + 2H + Cu 2+ + H 2 O G [kj/mol] -255,17-948,92 4Cu + O 2 2Cu 2 O 2H + + Cu 2 O H 2 O + 2Cu ,60-826,10 Fig. 2: Detected reactions of a copper starting layer in an electrolyte containing H 2 SO 4. that none of the component materials will undergo an unwanted reaction with the electrolyte. This can be monitored by surveying all possible reactions of the device materials and the medium composition. According to a calculation of the free enthalpy ( G) of each reaction using the Gibbs-Helmholtz equation (eq. 1) it can be evaluated, whether a reaction is likely to take place (fig. 2). H = G + T S (eq. 1) The process temperature (T) and the enthalpies of formation ( H) as well as the entropies ( S) of all educts and products have to be known. Our tool relies on an additional software specifically developed for doing these calculations and furnished with the necessary table data, so that the only information needed to be supplied is the process temperature and the device and medium composition. However, this inspection does not give specifics on the reaction rate. Next to the device and electrolyte composition only the process temperature needs to be known making this a very generic rule. Furthermore, it not only applies to electroplating, since - adding some modifications - it can also be generally applied to processes using liquid media. Contrary to this adhesion of deposited materials can be taken as an example for a very non-generic rule. The factors influencing the adhesion are not known well enough to allow an analytical description of the problem. Due to testing methods a relative value for the quality of adhesion may be formulated for specified material combinations and this information accessed in the form of tables. This, of course, lacks the generics described above, since specific values have to be declared for all possible layer combinations. Additionally it has also to be regarded, that the way how material layers were deposited might also have an influence on the adhesion. However, using definite values is the only way to handle the task in this matter. As has become evident, generics in our perception means, that the rule should need as little supplementary data as possible and rely basically on process, material and medium parameters. This is very important, because commonly, the rules will only be able to reproduce the expert knowledge that was put into them when checking the sequence. In the case of highly generic rules, however, it is possible that incompatibilities in the processing may

3 Fig. 3: Flow of work within the CAD environment. be found, that was not thought of when creating the rule set. The rule examples shown only apply for the first two categories of incompatibilities mentioned in the beginning of section 2. The third kind treats rather geometric dependencies important in fabrication, which cannot be covered by mere design rules. However, the geometric data model used in RUMTOPF is a rather simple one giving only a schematic 2D view of the device as known from process plans. It does not allow more thorough analysis of geometric feasibility of the chosen processing. This is perfectly acceptable considering that the task of identifying geometric incompatibilities is already well-covered by a range of commercial and non-commercial design tools. Integrating process simulation into the tool would go beyond of the reasonable scope and create a large and complex system. Rather a modular approach integrating process rule checking with available process simulation tools seems adequate. 3. CAD ENVIRONMENT FOR SILICON MICRO COMPONENTS Due to the strong dependency between product design and manufacturing technology the development process of micro mechanical systems has to keep a balance between function and the feasibility of fabrication. In order to minimize time and cost consumption of the design phase of micro mechanical components software tools for simulation and analysis have been developed and partly integrated into MST-CAD environments. The MST-CAD environment BICEPS is based on wet chemical etching of silicon in KOH, which is a wide spread technology for the manufacturing of micro system components. Crucial to this technology is finding a suitable design for the 2D mask layout, which has the decisive influence on the 3D geometry of the resulting silicon structure. Because of the anisotropy of etching silicon in KOH the designer needs to possess a certain amount of knowledge about the etch process and experience with the creation of appropriate mask layouts. The more complicated the intended silicon structure has to become with regard to growing system integration and miniaturization, the harder an acceptable solution can be found. BICEPS is based on a workflow system, which integrates the appropriate software tools and provides an overall mechanism for the common structure of a design task. Furthermore the workflow system manages the information flow inside of BICEPS. Next to administering all in- and output data issues it guides through the design process offering the necessary software tools when needed - thus allowing the designer to concentrate on the design task [6-8]. A general overview of BICEPS is given in figure 3. A commercial CAD software is employed to create a 3D reference model resembling the intended silicon geometry. Its functional properties can be checked against given specifications using an FEM software. In most cases the reference model will need some modification and again further analysis until an optimal design is found. Having achieved this, the mask layout for the etch process is generated by the tool OMEA, which iteratively searches for an optimal mask layout utilizing a genetic algorithm (GA). The final mask design is verified using the etch simulator SUZANA to reveal the resulting 3D structure. Lastly again finite element analysis is carried out to make sure its functional properties are as desired. By accessing the attached central project database available at any point in the flow of work, data of already completed projects can be imported into the current project supplying access to already acquired design knowledge and experience from other projects. Thus the BICEPS environment accomplishes the component view, being originally an environment of the bottom-up type. Design optimization for a silicon wetetching processes in KOH was fully implemented making use of several specialized tools as described Simulation of wet-chemical anisotropic etching The known approaches for the simulation of wet-chemical anisotropic etching can be subsumed according to their internal representation of the etched crystal into two main categories: a.) Atomistic models directly image the atomic structure of the silicon crystal as cells and try to emulate the modi-

4 U. Hansen, C. Germer, S. Büttgenbach, H.J. Franke Fig. 4: Simplified illustration of the macro cell model. fications to the shape according to the energy of the bindings of single atoms within the crystal structure. Depending on these energies an etching probability of individual atoms or clusters of atoms is determined. The advantage of this procedure is a high accuracy and the consideration of not defined, broken crystal planes. Thus arbitrarily complex geometric shapes may be simulated. b.) Geometric models describe the etching on a more abstract level, since they do not focus on atoms but regard the change of geometric forms like planes, edges and vertices during the etching process [9-11]. Diagrams for the etch rates of silicon for as many plane orientations as possible are required. Advantageous compared to atomistic models is its fast computation. However, especially when used for more complex shapes geometric models lack precision. The etch simulator SUZANA is based on an atomistic model using a cellular automaton. Cellular automata may be considered as discrete systems containing large numbers of simple identical components with local interactions. They consist of a lattice of cells, each with a finite number of possible states. The cells evolve synchronously in discrete time steps according to identical transition rules. The rules can be regarded as a function, whose arguments are the states of the cell under consideration and Fig. 5: Anisotropic etching of a comb-shaped corner compensation comparing technology (left) and simulation (right) at KOH 30%, 80 C. of its neighbors. Its value is the new state of the considered cell [12]. In the present case the silicon crystal structure was chosen as the lattice of the cellular automaton, meaning each silicon atom corresponds to a cell. However, in order to reduce the large amount of data that has to be handled, it is necessary to define macrocells consisting of a cluster of atoms (see fig. 4). Therefore, the range of application can be widened by changing the resolution of the simulation. A simulation task is defined by the model dimensions, its resolution and process particulars like etch time, intended etch depth as well as concentration and temperature of the etchant. Etch rates for the three main planes need to be provided serving as a basis for the calculation of the etching probabilities. Double-sided etching of both Fig. 6: Main workflow of BICEPS and sub-workflow for etch simulation.

5 Fig. 7: Main window of the interface of RUMTOPF and SUZANA. (100)- and (110)-oriented silicon can be simulated. The definition of etch stop layers is possible. For visualization of the 3D geometry of the etched microstructure, the 3D solid model is translated into a 3D surface model [13]. An example of the capabilities of SUZANA is shown in figure 5. The etching of a comb-shaped compensation structure after different time intervals showing simulation and actual etching results is displayed. Figure 6 shows how SUZANA is integrated into BICEPS. The main workflow is shown, from which a subworkflow is spawned giving access to all tasks needed when simulating, i.e. tools for importing, exporting and conversion of the needed configuration data and achieved results. Every feasible task is represented by a box - its color indicating the status of the tool, which may change according to available input or output data. 4. COMBINING THE TOOLS As stated the system designer has to consider both system and component level views in order to define a feasible processing sequence. So practically he has to switch back and forth between the different tools needed to make sure that the intended component geometries are feasible with the processing technologies chosen while the processing sequence for the device as a whole is still practicable. I.e. the constraints of component and system level validation need to be satisfied. In order to provide a link between the system view to the fabrication offered by RUMTOPF and the component view offered by SUZANA the process sequence checker has been integrated into the workflow system used by BICEPS. As already mentioned, the validation tool has only limited information on the precise geometry being processed since it is designed to detect flaws in the processing sequence. But on the other hand it holds precise information on the process configuration along with the Fig. 8: Workflow for the integration of RUMTOPF into BICEPS. used etchant. Therefore the available project data has to be filtered and can then be enhanced with additional information needed for process simulation. This task is handled by the interface software shown in figure 7. With the import of RUMTOPF project data all usable information is extracted and filled into the process definition for the simulation job. This data can be altered or enhanced with tool-specific particulars by the user also offering the possibility for defining the needed mask layouts. In the presented case the lack of precise geometric representation in the sequence checker is not much of a disadvantage, considering the fact, that most simulation jobs will focus on certain parts of the mask design needing further specification of the layout. Holding this data in RUMTOPF would not be reasonable from the computational point of view and create an unnecessary overhead enlarging the systems footprint. Seamless integration of both systems is provided by the workflow shown in figure 8. The same system is used as in BICEPS offering all the tools expedient while checking a process sequence. If a project is not to be created from scratch a project file maybe imported. The task Validate Process Sequence wraps the sequence checker described in section 2. Viewing the process sequence or the output of the rule checking is possible. Convert Project Data for Etch Simulation encloses the presented interface to the etch simulation tool SUZANA. The simulation configuration produced maybe exported and imported into the workflow shown in figure 6. A major advantage of this concept is the simplicity of enlarging the system, since it is very modular. Practically all available tools can be embedded and equipped with an interface to the sequence checker. Due to the workflowbased structure of the system, the different tasks can be fulfilled concurrently by different people.

6 5. SUMMARY AND PERSPECTIVE A modular software environment has been presented, that is capable of dealing with specialized tasks for micro components as well as considering the whole system in form of a process sequence. The rule based validation of process chains using RUMTOPF has been illustrated and the iterative design of components by silicon wet-etching using BICEPS has been discussed. To make this system usable in practice, work has to be done in adding more simulation tools capable of dealing with other process technologies. Next to widening the systems abilities on the component level, another interesting enhancement would be the extension of the system level by an integration of behavioral simulation. A suitable interface to RUMTOPF needs to be developed here. Last but not least, with growing integration of other tools, the data exchange will become an important issue. Finding a common data representation (e.g. based on STEP) is another task, that needs to be tackled. 6. ACKNOWLEDGEMENT This work has been financially supported by the Deutsche Forschungsgemeinschaft (DFG) within a collaborative Research Center (Sonderforschungsbereich 516) titled, Design and Fabrication of Active Microsystems. 7. REFERENCES [1] S. Büttgenbach, Mikromechanik, Teubner, Stuttgart, [2] K. Hahn, Methoden und Werkzeuge zur fertigungsnahen Entwurfsverifikation in der Mikrotechnik, Ph.D. Thesis, Universität Dortmund, [3] U. Hansen and S. Büttgenbach, A Data Model for the Representation of Fabrication Dependencies concerning Micromechanical Devices, Proc. MSM 2001, Hilton Head, USA, pp , [4] U. Hansen, S. Büttgenbach, C. Germer and H.-J. Franke, Rule Based Validation of Processing Sequences, Proc. MSM 2002, Puerto Rico, USA, pp , [5] U. Hansen, S. Büttgenbach, C. Germer and H.-J. Franke, Analysis and Verification of Processing Sequences, Proc. MSM 2003, San Francisco, USA, [6] L. Steffensen, O. Than and S. Büttgenbach, BICEPS: A modular environment for the design of micromachined silicon devices, Sensors and Actuators A 79 (2000) [7] U. Hansen, L. Steffensen and S. Büttgenbach, MST-CAD Environment for Anisotropic Etching of Silicon, Proc. of the VDE World Microtechnologies Congress (MICRO.tec 2000), Vol. 2, Hannover, pp , [8] L. Steffensen, Entwicklungsumgebung für den rechnerunterstützten Entwurf von Mikrokomponenten, Ph.D. Thesis, Universität Braunschweig, [9] J. Frühauf, Simulation of Orientation Dependent Etching of Masked Silicon Substrates, Proc. Micromechanics Europe, Berlin, Germany, pp , [10] J.S. Danel and G. Delapierre, Anisotropic Silicon Etching. A Simulation Program, Sensors and Actuators A 31 (1992) [11] U. Heim, Untersuchungen zur Simulation des nasschemischen anisotropen Ätzens von Einkristallen, Ph.D. Thesis, Technische Universität Ilmenau, [12] S. Wolfram, Theory and Application of Cellular Automata, Physica D 10 (1984) [13] S. Büttgenbach and O. Than, SUZANA: A 3D CAD Tool for Anisotropically Etched Silicon Microstructures, Proc. European Design & Test Conference, pp , 1996.