Counting
pennies
As
the end of conventional planar CMOS scaling nears, chipmakers will
need alternatives to traditional silicon substrates. Production of
advanced logic ICs in particular will require new wafer technologies
such as strained silicon, silicon-on-insulator (SOI), and multigate
FETs. Naturally, the winning technology will depend to a great extent
on cost.
 |
|
SOURCE: INFINEON
TECHBNOLOGIES, ILLUSTRATION BY JAMES SCHLESINGER
|
In
a technical presentation at Semicon Europa 2003, Helmut Tews and three
colleagues from Infineon Technologies focused on cost-benefit factors
and other issues that will determine the successful new wafer types.
The Munich-based Infineon team believes chipmakers will need to manufacture
next-generation logic devices on something other than standard wafers.The
pie chart here depicts the distribution of processing costs for a
logic device using triple-oxide technology and six BEOL levels of
copper. The metal levels make up a good portion of the costs, which
include capital investment, cleanroom operations, cleanroom technicians,
and materials. Mask and wafer costs are not included.
The
chart divides processes into their FEOL and BEOL constituents. Tews
says the team separated FEOL processing into "building blocks" beginning
with shallow-trench isolation, shown as segment one. Typically, each
block consists of a lithography step, an implantation step, cleans,
and strips, he says. The number of implant levels depends on the number
of different types of transistors on the chip. Contact holes, shown
as number nine, mark the transition between FEOL and BEOL, notes Tews,
adding "for many products BEOL is more expensive than FEOL."
Regarding
production costs, Tews says chipmakers try to minimize the use of
metal levels because they require very expensive lithography. A device
manufacturer will switch to 2X metalsmeaning the width of the metal
layers is twice as large as 1X metalsas soon as possible. For the
sake of their cost calculations, Tews and crew assumed two metal levels
with 2X metals. Determining the number of 1, 2, and 4X levels requires
striking a difficult balance between area request and processing costs,
Tews says.
The
winning wafer technology will need to be available when the industry
is ready to jump to the next process node. The market ultimately will
determine which device concepts and wafer types will emerge to keep
the industry on the schedule dictated by the ITRS, the Infineon team
concludes.