• material: silicon, III–V, wide band gap (SiC, dia- mond), polymer, glass;
• integration: monolithic integration, hybrid integration, discrete devices;
• active vs passive: transistor vs resistor; valve vs sieve;
• interfacing: externally (e.g., sensor) vs internally (e.g., processor).
The above classifications are based on device func- tionality. In this book, we are concentrating on fabrica- tion technologies, and then the following classification is more useful:
• volume (or bulk) devices;
• surface devices;
• thin film devices;
• stacked devices.
1.8.1 Volume devices
Power transistors, thyristors, radiation detectors and solar cells are volume devices: currents are generated
Introduction 9
Finger
Rear contact Oxide
‘Inverted’ pyramids
p+ p+ p+
p+
n+ n
p-silicon
Oxide
(a)
Drain
n+ n+ n+
p+
p+ p p
n−
Source Gate Source
Cell space (Ls) Halfcell
Width (Lw)
RCH
RCH RACC
RJFET
Repl RACC
(b)
Figure 1.6 Volume devices: (a) passivated emitter, rear-locally diffused solar cell. Reproduced from Green, A.M.:
(1995), by permission of University of New South Wales. (b) n-channel power MOSFET cross section. Reproduced from Yilmaz, H.et al. (1991), by permission of IEEE
and transported (vertically) through the wafer (Figure 1.6), or alternatively, device structures extend through the wafer, like in many bulk micromechanical devices. The starting wafers for volume devices need to be uniform throughout. Patterns are often made on both sides of the wafer, and it is important to note that some processes affect both sides of the wafer and some are one sided.
1.8.2 Surface devices
Surface devices make use of the materials properties of the substrate but generally only a fraction of wafer thickness is utilized in making the devices. However, device structure or operation is connected with the properties of the substrate. Most ICs fall under this category: metal oxide semiconductor (MOS) and bipolar transistors, photodiodes and CCD image sensors.
10 Introduction to Microfabrication
Figure 1.7 Surface devices: a 0.5àm CMOS in a scan- ning electron microscope view
In silicon CMOS (Figure 1.7), only the top 5àm layer of the wafer is used in making the active device, and the remaining 500àm of wafer thickness is for support: mechanical strength and impurity control. Sur- face devices can have very elaborate three-dimensional structures, like multilevel metallization in logic circuits, which can be 10àm thick but this is still only a frac- tion of wafer thickness; therefore the term surface device applies.
1.8.3 Thin-film devices
Devices can be built by depositing and patterning thin films on the wafers, and the wafer has no role in device operation. Wafer properties like thermal conductivity or transparency may be important (Figure 1.8), but
the substrate is not machined or modified. Thin-film transistors (TFTs) are most often fabricated on non- semiconductor substrates: glass, plastic or steel. Surface micromechanical devices like switches, relays, DNA arrays, fluidic channels and gas sensors are often fabricated on silicon wafers for convenience but they could be fabricated on glass substrates as well.
1.8.4 Membrane devices
Membrane devices are a sub-class of thin-film devices:
again, all functionality is in the thin top layer, but instead of full wafer mechanical support, only a thin membrane supports the structures. Many thermal devices are membrane devices for thermal isolation: thermopiles, bolometers, chemical microreactors and mass flow meters (Figure 1.9). Many acoustic devices also utilize bulk removal. Optical paths can be opened by removing the bulk semiconductor. X-ray lithography masks are gold or tungsten microstructures on a micrometre- thick membrane.
1.8.5 Stacked devices
Stacked devices are made by layer transfer and bonding techniques. Two or more wafers are joined together per- manently. Devices with vacuum cavities, for example, absolute pressure sensors, accelerometers and gyro- scopes are stacked devices made of bonded sili- con/glass wafer pairs. Micropumps and valves, and
Si wafer
Doped
polysilicon Undoped
polysilicon Oxide Nitride anti-reflective coating Metal
Tunable air gap
Figure 1.8 Surface micromachined Fabry–Perot interferometer: thick oxide has been etched away to create a tunable air gap. Silicon is transparent at infrared wavelengths, and radiation can enter the device through the wafer. Redrawn from Blomberg, M.et al. (1997), by permission of Royal Swedish Academy of Sciences
Introduction 11
Figure 1.9 Mass flow sensor: a resonating bridge over an etched channel. Reproduced from Bouwstra, S. et al.
(1990), by permission of Elsevier
Figure 1.10 A microturbine by silicon-to-silicon bonding.
Reproduced from Lin, C.-C.et al. (1999), by permission of IEEE
many micropower devices like turbines and thrusters are stacked devices with up to six wafers bonded together (Figure 1.10). More and more layer transfer and wafer bonding techniques are being developed, and stacked devices of various sorts are expected to appear; for example, GaAs optical devices bonded to Si-based elec- tronics, or MEMS devices bonded to ICs.