Page 332 - Sami Franssila Introduction to Microfabrication

Tools for Microfabrication 311

(Fed. Std.) cleanroom. In most cases, just the front panel Both academic and industrial labs buy equipment
of the system is in the cleanroom and the rest of the tool for research and development, but what will happen
is in the service area, which has more relaxed particle when a successful new process needs to be scaled
cleanliness requirements. up for production? The popular answer today is that
the basic design of the process chamber (e.g., spinner
bowl geometry, sputter cathode design, etcher gas
30.2.5 MTTF, MTBA, MTBC
manifold, RTA lamp configuration) is fixed. Research
How long will the tool work before failure? Do labs buy the very basic configuration, essentially the
operators need to interfere with its operation? How process chamber only (obviously this works better
often does it have to be cleaned? These questions are for some tools than others and not at all for optical
operationalized by MTTF (mean time to failure), MTBA lithography). Later on, when the process is transferred
(mean time between assists) and MTBC (mean time to manufacturing, productivity features such as cassette-
between cleans). to-cassette automation and advanced software can be
MTBC is process-dependent: particle counts (on test added. This reduces the risk of new equipment purchase
wafers) are checked regularly, and increased counts for the industry, and it allows academic labs to do
indicate a cleaning need. However, the acceptable industrially relevant research without the need to invest
particle count depends on the chip size, sensitivity of in volume manufacturing tools.
the particular process step to particulate contamination
(a subsequent step may be a cleaning step that effectively 30.4 PROCESS REGIMES:
removes particles) or just an engineering judgement TEMPERATURE–PRESSURE
about the acceptable level of particles. Particle counts
in individual process steps cannot easily be correlated Two major process parameters are pressure and temper-
with process yield, and therefore short loop test runs ature. Most microfabrication processes are vacuum/low
with specially designed test structures are used to check pressure processes (CVD, etch, sputter, implant), some
the effects of individual process steps. are room ambient processes (lithography, wet clean-
ing) and high-pressure oxidation is an exception. The
◦
temperature scale extends from 1200 C diffusions to
30.3 TOOL LIFE CYCLES
850 to 1100 C oxidation, 300 to 900 C CVD to
◦
◦
Tool development takes a long time: from the first proof- room-temperature processes (plasma etch, sputtering,
of-concept tool to multiple orders for volume manufac- implant, lithography, wet cleaning). Some etch pro-
◦
turing easily takes 10 years. Proof-of-concept tool is a cesses use cryogenic cooling down to −100 C for
home-built or modified equipment that demonstrates the suppression of spontaneous chemical reactions. Many
key features of a new process. For e-beam lithography, room-temperature processes can be run at higher tem-
◦
it might be a new column design; for a plasma etcher, peratures for special purposes: sputtering at 450 C for
◦
it might be a new RF-coupling scheme. The alpha tool aluminium flow, implant at 800 C for SIMOX wafers
is a built-to-purpose system that has the new key ele- or plasma etching at elevated temperatures to reduce
ments designed in from the beginning. The alpha tool residues. Figure 30.1 shows major processes on a tem-
does not have productivity features such as robotics and perature–pressure chart. High temperature/high vacuum
software, but is designed for the final wafer size. The processes are difficult because of outgassing from vac-
reliability of the alpha tool is not comparable to pro- uum components during high-temperature operation.
duction tools; it is a test-bed for process research, not There are five main methods that are currently in use
for production. Alpha tools are not shipped to outsiders. to heat wafers, but for example microwaves have been
The beta tool is a fully equipped version, with essentially tried (Table 30.2).
all the features that will make the final product distinct. The first three methods are used in high-temperature
Beta tools are shipped to select customers who are will- processes and the latter two in low-temperature pro-
ing to bear part of the burden of testing new equipment cesses. Some degree of heating and/or temperature con-
in order to benefit from new technology. Beta customers trol is desirable in almost all tools. In all plasma equip-
provide productivity-related data that is difficult or even ment, there is plasma heating; in ion implantation, the
impossible to acquire at the tool-manufacturer site: What beam flux can heat the wafer considerably; photore-
is uptime in production-like conditions? Is wafer yield sist baking and UV-assisted stabilization depend on hot
comparable to existing or competing designs? What are plate treatments. Whereas older hot plates had no active
the field servicing requirements? control of wafer-to-plate contact because there was an