Too often, metrologists and designers position themselves into two different categories. There might be a cultural lack of awareness on both sides being manifested in the designer's lack of considering metrology early enough or the metrologist's lack of integrating measurement better in the design cycles. This article, therefore, will explain the importance of design tools in metrology and vice versa.

Challenges

During the last two years, two roadmap planning meetings have been organised by the UK based company CEMMNT (The Centre of Excellence in Metrology for Micro and Nano Technology). Experts in the field have been invited to these workshops in order to identify the different challenges related to metrology in the micro nano domain. We like here to pick out only a few of these identified challenges, which are relevant in view of MEMS modelling software.

When developing a new micro-nano-product, very often, the focus is on developing the technology. Metrology is often seen as an afterthought, and in many cases, applied only when the first prototypes have been fabricated and unexpected device behaviour observed. Often, there is no direct link between design and metrology. The overall development process could be greatly improved by considering metrology earlier into the development cycle and integrating it within the design process. In other words, metrology should be 'built-in' to the early development phases, making it a standard part in MEMS product development methodology.

Another challenge here is related to the specific importance of the three-dimensional (3D) aspect of micro-nano devices such as MEMS. Although huge progression has been made in 3D measurements, major issues still remain related to true 3D measurement capability, such as the measurement of side-walls or membrane deformation, especially in the case of in-line measurements. The progress of 3D modelling capabilities is often not exploited.

Finally, how can we better predict the possible failure modes of a device before engaging into an expensive fabrication and prototyping run? In the end, the important factor is to have effective, working devices with sufficient characterisation of functions - not individual properties.

Technology Parameters

There are two categories of parameters that are relevant for MEMS design. The first, consists of the material properties and geometric parameters that are determined by the manufacturing process. The second, consists of the dimensional parameters of the MEMS building blocks that are determined by the engineer (such as length, width, number of comb fingers, etc.). Before the first wafer has been taken out of the box to begin a prototyping run in the clean room, the development team can make efficient use of design and modelling software not only to predict the performance of the device but also to investigate manufacturability. During the last two decades, dedicated MEMS design and modelling tools have been developed and currently cover the complete range of modelling levels, ranging from the physical and process level to system level, where MEMS and electronics can be simulated together. However, in many cases, the designer is not yet aware of all material and process properties. Whilst recent design tools are very powerful and high in accuracy, taking into account coupled multi-physics behaviour, the software relies on the fundamental technology related to the input parameters mentioned above. In many cases, the manufacturing related parameters are not sufficiently known. If the designer's input parameters scatter and are uncertain, the simulations result in form of output parameters will also scatter and be even more uncertain. In order to design a device from the beginning, the development process should start early with characterisation of materials and processes. Device modelling based on empirically obtained technology parameters is a key factor to achieving successfully accelerated MEMS product development.

Modelling and Measurement of Test structures

One of the most common methods to obtain technology and geometry parameters is the use of so-called 'test structures'. Significant amounts of valuable manufacturing information/data may be obtained from such test structures through metrology - film thickness, gaps, line-widths and line-spacing capabilities, alignment capabilities, and so on. MEMS are predominantly mechanical and their design requires accurate characterisation of all relevant material properties, such as Young's modulus or mechanical layer stress. Along with process characterisation structures, material property characterisation structures can be placed at specific locations on test wafer lots. A sub-set of selected structures is typically used to monitor run-run variations and obtain statistical process data.

An effective combination of modelling and measurement can be used to extract parameters, which are often difficult to determine to examine. For example, electrostatically driven laterally resonant comb-drive test structures with changes in spring width are used to ascertain systematic variations in process offsets (edge biases) and sidewall angles. The techniques is both in situ and non-destructive. A 3D behavioural model for the resonant frequency includes effects of a distributed mass, residual stress, and compliant supports. Application of this model to experimental data can determine the off-set and sidewall angle of MEMS devices. In practice, all detailed material and process property information that has been previously correlated with simulation models of the test structures can be implemented into a process design kit.

Behaviour Modelling and Metrology

The key to developing several viable concepts is the ability to create and evaluate concepts rapidly, whilst being precise at the same time. For MEMS, this can be done through the availability of parameterised component libraries where designers can piece together different library elements to create more complex designs. Each element of the library has an underlying model that captures the behaviour of the element accurately across all possible displacements. All parameters of the element model are available to the designers as variables to be defined and set. A model assembled from parameterised library components immediately enables very rapid and accurate characterisation of the behaviour of a specific conceptual design. This modelling approach has two significant further advantages over conventional Finite-Element-Methods (FEM).

The first, is the ability to set process restraints. A process design kit is built into the conceptual design phase through process constraints on certain parameters. For example, parameters such as layer thickness, Young's modulus, Poisson's ratio and stress gradient may be well defined within the tolerances of the specific manufacturing technology. The process design kit would automatically set these design constraints for the designer, allowing other parameters to be varied during the evaluation of different designs. This also includes statistical distributions (SPC data). Importantly, simulations can be carried out in dependence of the statistical variation due to technology parameters, and as a result, more sophisticated analysis  becomes possible, such as a Monte-Carlo simulation in dependence of variations of the underlying material and process parameters.

The second major advantage, is the ability to include the system level electronic design during the evaluation of the conceptual designs. Each conceptual design can be hierarchically contained within a schematic of the entire system that is available to the designer in the form of re-usable blocks in the library. By using available circuit components, such as transistors and passives, it is possible to create virtual "test" beds for the product within a single environment. The significant advantage of this is in providing immediate feedback as to the viability of certain concepts over others. This can literally save years of extra effort and investment by further filtering the set of available design concepts to only those that meet the customer requirement. In addition, tremendous advancements have been made in visualising the behaviour of modelling simulation results directly in 3D, or very recently, even creating multi-physics parameterised behaviour models directly into a 3D environment that are even compatible with standard EDA tools.

Process Modelling and Metrology

On the other end of the spectrum of modelling tools is the physical level - capturing the geometric details of the technology. Engineers can create realistic looking virtual 3D representations of a MEMS or IC device in a very short time frame by combining the actual 2D layout file with a technology file. The technology template file includes a detailed description of the manufacturing process and is based on the parameterised and experimentally calibrated individual process steps. Software using voxel algorithms are able to furnish the accuracy of physical process simulation and the speed and capacity of process emulation techniques. A voxel is a 3D element comparable to a 3D pixel. The user specifies the number of voxels per micron. A higher number results in a higher resolution. This software not only enables the user to create highly realistic 3D models of micro-fabricated devices, but it also allows the user to illustrate the precise impacts of process changes. The results are realistic virtual prototypes of the final MEMS device, allowing process specific design control before both the first mask creation and the first lab-run. The software can be applied to all devices fabricated with IC-style manufacturing techniques.

Visualisation in 3D, is therefore, a main enabler in anticipating the fabrication output before entering into the clean room. Errors or unwanted design effects can be prohibited providing that the technology is described effectively and has been accurately measured. The process emulation software can also be used for failure analysis in conjunction with analytical  techniques (FIB, SEM, TEM, etc.), and for the documentation and training of lab personnel.

It is crucial to consider the impact of changes or variation in technology throughout the complete product development phase and at different levels. Materials and process parameters need to be shared, not only when modelling on the physical level, but also when entering into the system level. Preferably, there is a seamless connection between design tools for the different modelling levels sharing all relevant metrology parameters. Combining metrology and design tools provides a powerful methodology to develop MEMS products with less risk.