MEMS' World - Sample microtechnology fabrication process

Fabrication of a bilayer thermally actuated cantilever beam

Microfabrication of a component is a stream of basic steps. Each of these steps can be easy or tricky depending on the previous ones. This MEMS actuator example has no other purpose but propose a sample combination. It is definitively not enough to begin a real fabrication. No parameters are supplied, none of the problem you could run in if you try to fabricate MEMS devices are noticed. This is really just a basic sample to have an idea of what a process chart could be. For a bit more details about the technics involved here, check the microtechnology section.
This example can give a good idea of MEMS manufacturing by surface engineering, though other technics exist.


In this example, we will introduce microfabrication of a theoretically very simple system, consisting in a bilayer clampled-free beam thermally actuated. This structure can be used as a basic actuator or directly for material characterization. As for any integrated device, could it be microelectronic or MEMS, microtechnology is the only manufacturing technic able to produce large quantity of these devices at low cost.

Schematics of the cantilever
Schematic of the bilayer cantilever

Preparation of the substrate

The starting point of most of processes is a silicon wafer. It is just a round disk made of silicon. Most of the time, silicon is doped so that it is conductive.

Silicon Wafer
Silicon substrate

Although only the cantilever top layer will be electrically heated, we need wires to access it. Theses wires will be deposit on the top of the silicon substrate. Since the whole wafer is going to be conductive, or in the best case, to be a poor isolant, we need a passivation layer to avoir polarizing all the wafer's surface. In this case, we're going to use Silicon Nitride (SixNy). The step used to have a complete layer that covers the substrate is LPCVD. We put the substrate in an oven with the appropriate gases flows with controlled temperature and pressure. So we cover the complete surface of silicon with nitride.

Passivation layer
Silicon nitride layer deposited on silicon substrate

Sacrificial layer

MEMS often involve mobile parts. For these parts to be free of moving, the fabrication process must include a step allowing the releasing of the structures. What is called releasing is the step consisting in etching the layer underneath the mobile parts. This layer that has no function but be a support for the strucutral mobile parts is called sacrificial layer. The material is chosen so that there exists an etching process able to selectively attack it. In this case, we will use silicon dioxide (SiO2). This is deposited with the LPCVD.

Sacrificial layer
Silicon dioxide layer deposited on passivation layer

If we just go on with the structural parts, we are likely to loose all the structures during the releasing! Structural parts need to be attached to the wafer. To allow anchors to be in contact with the substrate (so for this example the passivation layer), we must open the sacrificial layer in the anchors places. So we need a pattern to be transfered on the substrate, and an etching step to open anchors areas. The pattern is obtained with a lithography. This involve the use of a mask and photosensitive resist. Check the link to learn more about it. Once the mask is transfered, we etch the silicon dioxide using a wet etching with Buffered oxide Etchant (B.O.E.) that is a mix of Florhydric Acid (HF) and Ammonium Fluoride (NH4F).

Opening of anchors
Opening of the silicon dioxide sacrificial layer

Structural layers

Let's deposit the first bimorph material. This could be either a conductive or isolant one. There is no importance for the moment. What will be important is to get sure that the electrical current that will flow through the heater really get across the heater, and not the structural layers. Let's take for this example a polysilicon layer. It is also deposited by LPCVD.

First structural layer
Polysilicon layer deposited on oxide sacrificial layer

Now that we have the material, we need to define the structure geometry. This is done in the same way as for the sacrificial layer opening. We do a lithography to transfer the pattern on the material layer, and then we use a dry etching to remove the polysilicon we don't need.

Cantilever defined
Polysilicon layer etched to form the cantilever

The second structural layer is made of SU8, that is a photoresist. Contrary to most of photoresist, SU8 is commonly used as a structural material.

SU8 pattern
SU8 layer patterned on polysilicon cantilever


Finally, we need a last pattern to allow heating of the structures. This one will be made of metal. In this case, we choose gold. There are several ways to get a metal layer patterned on the substrate. We choose a lift-off process. The metal itself is deposited by electronic-beam evaporation. First, we do the lithography, so that resist doesn't stay where we want the metal at the end. Then we deposit the gold layer. At last, we remove the photoresist, and the metal over it will go away.

Heater defined by lift-off
Gold heater patterned on SU8 by lift-off


As we said above, the last step of the fabrication is the releasing. This step conists in removing completely the sacrificial layer, so that the structural layer above it will be free of moving. Here we have used SiO2 as sacrificial layer. So we use B.O.E. 7.1 to etch oxide. The structure is now completely finished.

Structure released
Complete bilayer cantilever with heater


On the figure, the cantilever is straigth. In real conditions, the difference of residual stress in the materials would result in a strain, and the cantilever would bend up after the releasing. See the introduction to solid mechanics for more informations.

The schematics represent one structure. You have to remember that these processes are made for large scale integration. On a complete substrate, you would find hundreds, if not thousands of structures on the same substrate.