Trends in semiconductor technology

1
Trends in semiconductor technology

• Jurriaan Schmitz
• Chairholder of Semiconductor Components
• MESA institute
• University of Twente

2
The Microstrip Gas Counter and its application in
the ATLAS inner tracker

• Fragment of my introductory talk, October 14,
  1994
• We want to use the MSGC in an experiment named
  ATLAS. Unfortunately this will only be conducted
  from the year 2002 onwards.
• DISCLAIMER Consider my upcoming statements on
  the future of CMOS as predictive as the above

3
Contents

• MOSFET basics
• The start of MOS technology
• Moores Law
• The ITRS roadmap
• Modern CMOS technology
• The challenges ahead
• The role of academia

4
Semiconductor essentials
n-type doped semiconductor e.g. silicon with
phosphorus impurity electrons determine
conductivity
p-type doped semiconductor e.g. silicon with Al
impurity holes determine conductivity
5
The field effect

accumulation
6
The field effect transistor
gate
source
drain
Gate voltage controls the current between source
and drain
7
(No Transcript)
8
The first transistor (re-created)
9
Kilbys first IC 1.5 mm x 1 mm
10
Fairchilds flip-flop 1961 4 transistors, 5
resistors Notice metal interconnect
1.5 mm
11
RCA, 1962 Logic chip, 16 transistors First MOSFET
IC
12
5
10
Gordon
Moore
1965
Fairchild
4
10
3
10
Number of components per chip
Moores Law (1965) Progress in technology At
the same cost, one can add more and more
components on a chip. The number of components
doubles each 1.5 years.
2
10
1
10
0
10
1960
1965
1970
1975
Year
13
(No Transcript)
14
1976 Apple I motherboard 1981 The first PC
IBMs 5150 PC Intel microprocessor DOS operating
system
15
INTEL microprocessors
Number of transistors
Year
16
Reflections on Moores Law

• Exponential growth with time is
  universalpassenger airplanes, cargo ships, hard
  disk drives, nuclear fusion, Collider energy?
  Luminosity?
• but only for a while!
• So its not particularly Moores and its not a
  law.

17
Technology driven exponential progress
18
Impact of Moores Law

• Device dimensions shrink (scaling)
• Cost per function decreases ( 35 per year)
• Power per function decreases
• Speed increases
• application field of semiconductors increases!
• (e.g. personal computers, handheld telephones,
  solid state audio, speech recognition)

19
(No Transcript)
20
You might still consider this big
Modern transistor
Influenza virus
21
What does CMOS scaling bring us?
Higher frequency operation
Cheaper integrated circuits (25 p.y.)
Lower power operation
1950, 6
2000, 145
22
But also
Reduced static noise margin
Increased gate leakage

• Lower supply voltage
• Smaller devices, larger fluctuations

Higher price for small quantities

• Masks increases
• Mask cost increases
• Fab COO increases

23
(No Transcript)
24
Transistor Technology
Well Technology
25
Transistor downscaling

• Reduction of gate length (lithography)
• Increase of impurity concentrations
• Decrease of gate dielectric
• Reduction of source and drain dimensions
• Brews Law
• Lmin 0.4 xj tox (Ws Wd)2 1/3
• Lmin minimum gate length with normal behaviour
• xj source and drain depth
• tox gate dielectric thickness
• Ws, Wd depletion widths of source and drain
  junctions

26
Lithography
EUV prototype
27
The interconnect shrink
0.5 µm technology
0.1 µm technology
Al
Cu
SiO2
Low-K
W
28
The Red Brick Wall(s)

• Further scaling of the circuit
• Atomic dimensions are in sight
• Gate dielectric needs replacement
• Gate electrode needs replacement
• Interconnect becomes a speed and power bottleneck
• The economy
• Fabrication plants get too expensive to build (3
  B)
• Semiconductor market is too big to grow much
  further

29
The power problem
Power per transistor decreases but not the power
density!
Fortunately, most ICs do better than Pentiums
30
Atomic dimensions and the loss of information
Quantum fluctuations
Dissipation problems
31
Semiconductor economy

• Traditional scaling can no longer facilitate the
  strong market growth seen in the past

1) The semiconductor industry has acquired a
strong position in the total electronics
market 2) New technology generations show
progressively less benefits over their predecessor
32
The design and verification gaps
Do we want nanotechnology?
33
Semiconductor market development
2000
Annual turnover (G)
2001
Actual (Dataquest) 2002 Forecast (6 growth)
No clear trend - a mature market?
34
Research at MESA

• MESA 18 participating chairs from TN, CT, and
  EL
• Nanotechnology, microsystems, materials science
  and microelectronics
• 400 people, including over 200 PhDs and
  postdocs
• Yearly turnover 31M
• 1250 m2 fully equipped clean room
• A materials analysis laboratory
• Several satellite laboratories

35
Running projects
IC-technology
Devices
Reliability
High-k ALCVD
ESD in CMOS
Micro Gas sensors
STW
EU
Philips
Cool dielectrics
Deuterium dielectrics
E-T-M in interconnect
Philips
FOM
FOM
Cu barriers ALCVD
Plasma damage
1/f noise
STW
STW
Philips
Reliable RF
Light from Silicon
STW
STW
NEW
NEW
Ends soon
36
Submitted new projects
IC-technology
Devices
Reliability
NBTI
SmartOxides
Philips
EU
Planned new projects
Low Temp devices
High K reliability
Vulcano
STW
STW
STW
37
Outlook

• There is still plenty of room at the bottom
• Standard CMOS scaling will end soon
• New technologies will emerge NOT for ordinary
  computing
• Light-silicon interaction huge potential,
  physics?
• Novel devices may well include particle detectors