Publication date: 05 January 2010
MEMS release processing has historically been performed using wet chemistry. However, just as CMOS technology evolved from wet chemistry to dry etching, MEMS technology must also now adopt dry etch processing in order to become the ubiquitous technology it promises to be.
Release etching is the final stage of fabrication for surface micro-machined MEMS devices, to release the structure from the surrounding sacrificial material. Used with anhydrous HF as the release agent, vapour-phase etching of an oxide sacrificial layer achieves enhanced control and selectivity and allows MEMS devices to be built using materials that would not be compatible with a wet-chemical process. These advantages allow device designers to create new structures and achieve finer features. For the production engineer these advantages provide greater repeatability, reliability and ultimately yield.
Although CMOS technology has evolved in the direction of a standardised set of processes, MEMS will depend on the extra flexibility and diversity made possible by vapour-phase processing in order for market growth to continue accelerating. Because MEMS devices are mechanical structures, new functions are achieved by creating innovative mechanisms, which in turn demands extra flexibility to build new structures and to choose from a wide range of materials. Fortunately, emerging vapour-phase release processing provides this flexibility for device designers, and can be tuned to deliver best results for a given process/device combination.
Vapour-phase etching of an oxide sacrificial layer is performed using anhydrous HF. The key to successful oxide release is the choice of, and accurate control over, the catalyst for the reaction between the anhydrous HF and the oxide material, see equation 1.
4HF + SiO2 → SiF4 + 2H2O
Controlling the catalyst enables process engineers to maintain an optimal etch rate for the structure to be released; the optimum etch rate depends on factors such as its mechanical properties and the structural material.
If we expand on the reaction above and look at what happens at the oxide surface we have the following reaction as proposed by Miki et al, [1].
2HF + H2O → H3O++ HF-2
SiO2 + 2H3O+ + 2 HF-2 → SiF4 + 4H2O
H2O is the natural catalyst for the reaction between the oxide and the anhydrous HF. However, liquid H2O has traditionally been regarded as undesirable when used with HF to etch silicon dioxide. It is believed to be a cause of stiction and corrosion on exposed metal surfaces, which has driven most vapour-phase process development towards the use of ethanol or methanol as the preferred catalyst. Although wet etching silicon dioxide using HF acid can be problematic, vapour-phase etching using anhydrous HF is rather different. In this case controlling the quantity of H2O, present in its vapour phase, provides greater freedom to adjust the etch rate and to achieve faster etch rates than can be achieved using an alcohol catalyst.
The etch mechanism of HF vapour with SiO2 as the sacrificial material, in the presence of H2O, can be explained by adsorption of H2O and HF at the oxide surface. Where condensation of H2O and HF at the oxide surface occurs, the etch rate of silicon dioxide is high due to the high ionization efficiency of HF in the condensation layer.
Controlling the process to optimise the quantity of H2O condensing on the surface of the oxide being etched allows the etch rate to be reduced or increased to achieve the best possible throughput, without compromising selectivity or other factors that influence yield. At all times it is necessary to control the level of H2O, balancing this against the rate of gas flow, to catalyse the reaction without allowing excessive H2O vapour to accumulate. Excessive H2O impairs control of the etch rate and also risks the creation of liquid water within the process leading to poorer selectivity and control. Combining liquid water with anhydrous HF creates HF acid, which will attack many types of structural materials including metals.
The quantity of H2O present also depends on the type of oxide to be etched. Oxide films have water absorbed into the film, and the quantity varies depending on the method of oxide deposition. Less dense films such as spin-on-glass films have high water adsorption compared to dense thermal oxides, which tend to have lower water content. The water absorbed in the film is released during the etch adding to the level of H2O present during the process. In addition the structure being etched influences the H2O present in the process; with a larger exposed area the rate at which the absorbed water is released also increases.
Taking these factors into consideration in order to optimise the quantity of H2O present, allows oxide-release processes to be tailored for maximum productivity for any given device design, and avoids extra steps completely such as rinsing and critical point drying.
The SVR – vHF sacrificial vapour release system, a proprietary system developed by memsstar®, continually monitors the reaction in real time. This allows the process to be set up to maximise the reaction efficiency whilst avoiding a ‘runaway’ process condition. The graph of figure 1 illustrates real time data from a controlled process in which the reaction rate is kept constant by balancing the chemical reactions at the oxide surface to prevent excess H2O from being produced.
This allows for a fully released, free MEMS structure with no stiction. An example is shown in figure 2.
Ultimately, processes and equipment that allow MEMS fabrication specialists to optimise those process parameters that influence the natural reaction catalyst of H2O afford greater flexibility in the design and manufacture of MEMS devices. These will allow device designers to improve established devices, to experiment with structural materials that are incompatible with wet etching, and to discover innovative new structures that would be considered too fragile for a conventional wet-release process.
MEMS foundries, which depend on the ability to build many different types of structures for a variety of customers, will value the flexibility to optimise processes individually through selection of HF release agent and by tailoring process settings according to the composition and mechanical design of the structure. Equipment vendors, process experts and foundries must take these variables into account, in order to work with device designers to create new types of MEMS products performing functions that even now remain difficult to imagine or visualise.
Reference1. N. Miki, H. Kikuyama, I. Kawanabe, M. Miyashita and T. Ohmi, IEEE Transactions on Electron Devices, Vol 37, No. 1, pp 107-115, January 1990
