Introduction to Micro and Nano Sensing

1
Introduction to Micro and Nano Sensing
Heikki Seppä VTT
2
Why sensing is made using micro devices

1. Price, price, and size
2. Compatible with silicon IC processes
3. Good performances in some cases (silicon MEMS)

Why using nano devices

• In general, no reason what so ever, poor coupling
  to the quantity to be measured (acceleration,
  pressure,..) and to the preamplifier
• 2) Quantum effect can be utilized
• 3) Low thermal and mechanical mass

Why using nano particles and printing
1) Price, price, price 2) Good coupling and
quantum effects can be combined 3) Good
performances in some cases (silicon MEMS)
3
About sensing
1) Sensitivity does not matter, resolution
does. 2) Coupling to the object as important as
the intrinsic resolution of the device. 3)
Sensor and readout electronics has to be
optimized together, do not develop any sensor
without asking how its output signal will be
monitored. 4) You do not understand how your
sensor is operating if you cannot predict its
noise performance precisely 1/f included. 5) Do
not believe what you are reading calculate your
self e.g, what is the best way to measure
displacement Most people believe that the laser
interferometer is the best because of the short
wavelength, some have claimed that the tunnelling
tip because of its extremely high sensitivity -
all wrong, the capacitive reading with maximized
energy and optimized noise matching via high
Q-resonator.
4
Trend in Microelectronics
2040
Linewidth
100 nm
CMOS
10 nm
Size of the Molecyle
Puolijohdemarkkinat x transistorin koko 20
000 m
5
Future in Electronics
Silicon (or new semiconductor) based electronic
nanopattering nanoparticles
Nanotechnology
Printable electronics
Discrete components Circuit board Packaging
RFID EPC
2010
Reel-to-reel fabrication
Now
6
Roadmap in Electronics
7
3th Economical Revolution
8
Langaton maailma
Ihmiseltä ihmiselle
Koneelta ihmiselle
Koneelta koneelle
Esineeltä koneelle
Esineeltä Ihmiselle
Esineeltä esineelle?
Broadcasting Zone
Wide Area Zone
Personal Zone
Local Area Zone
RFID on radio radioiden joukossa !
9
VISION
Networks and the recent progress in
microelectronics and telecommunications leads to
increasing number of microsensors
A network will end up with a sensor as it does in
a human nervous system
In sensing, intelligent measurement systems
will be replaced by stupid but un-expensive
microsensors
We will end up with Wireless Functional
Environment Silicon functionality Printed
functionality (nanoparticle based electronics)
Mobile terminal and physical interface (browsing)
will be used as a link to the data
10
Physical browsing will be based on the following
user interfaces SweepMe, information or sensor
data is detected by sweeping the surface
TouchMe, sensor is touched by the mobile
device PointMe, pointing of the object activates
the reading
The role of the network in sensing will have a
bigger role than the sensors itself
Wired or wireless network?, power managament,
user interface, ... become more and more
important issues
Due to demand for un-expensive sensors, the
trends in sensors will be to replace a single
sensing element with multiple ones (from3 x1D to
3D accelerometers, magnetometers,.., to replace
silicon MEMS with plastic MEMS, to replace
solid state sensors with printable ones.
11
From Instruments to Microsensors
Measurement instruments is usually composed of
one or more sensing elements, tens of
microcircuits or components, which are
independently optimized. Mosts cases the design
is made by several people.
Microsensors is usually composed of one or only
few microchips which are optimized as a whole.
Usually one person is more or less responsible
for the whole design one person should be an
expert in physics, electronics, ....
12
Why small small is beatiful or is it.
In field effect transistors, the narrow gate
leads to high bandwidth and low noise In
micromechanics a narrow gap leads to high force.
In a capacitance, energy is proportional to 1/d
but force is proportional to 1/d In mesoscopic
devices an energy related to a single electron is
proportional to e /C, this is a dominant term
only in small devices.
2
2
13
Sensor
Sensor
Amplifier
Object
Dissipation
Power Gain
Dissipation
Fluctuation
Fluctuation
Signal
T
T
T
a
Sensor device making power or energy coupling
between two quantities
good sensor strong coupling and low dissipations
Strong coupling enables us to make noise
matching between an object and an amplifier
without loosing information bandwidth
Low dissipations
sensors does not add noise
14
Energy Resolution
Thermal energy
Thermal noise
Intrinsic energy resolution
Bandwidth
Dissipation
Bandwidth
Reactance/Frequency
Reactance/Frequency size, L uoD or 1/C 1/A
15
Why small devices? (cont.)
If
(quantum noise dominates)
Energy resolution per unit band
then
decreased coupling to the object
resolution
increased 1/f noise
Typically the compromize with the coupling and
the noise
quantum noise
thermal noise limit
size
small
large
optimal size of the sensor
16
Oscillator as an example
How to make a good local oscillator for GSM
If multiplying the frequency f up nf the phase
noise is also n times higher. Why not to make
directly a smal size high frequency oscillator
instead of a low frequency one. If we have a
given process to produce MEMS device. It turns
out that there is an optimum frequency for the
local oscillator. E.g., a low phase noise local
oscillator for GSM ( 1 GHz) leads a 10 MHz 100
MHz reference oscillator. In the practical
applications, the size of the device is a
compromise. In our production, the gap size below
100 nm is needed other dimensions tends to be
from 2 um up to 1 mm. In 19992 1994 when we
started to activate MEMS in VTT, we did some
plans for the future within 5 10 years, we
will enter to NEMS. Now we have 40 50 people
working with MEMS. Only the gap size is reduced
in some devices below 100 nm.
17
Different kind of sensors
Sensor an element converting quantity under the
measurement into to an electrical form

• Active sensor provides power or energy gain
• SQUID, temperature sensor based on transistor,
  bolometer, FET- gas sensor
• Resolution related to readout electronics has no
  problem
• Problems are related to dynamic range, bandwidth,
  etc
• Enables effective multiplexing
• Passive sensors no power neither energy
  amplificatios
• Passive sensors can be split into the two
  categories
• Dissipative sensors an element where
  dissipations depends on the quantity
• Reactive sensors capability to store energy
  depends on the quantity

18
Passive sensors (dissipative)
Resolution when dc bias
Resolution when ac bias with tuning
19
Passive sensors (dissipative)

• sensitivity determines the resolution
• higher the measurement power the better the
  resolution
• amplifier not important since Ta lt T
• usually high power and small size lead to high
  1/f-noise, need to make compromise
• owing to high dissipation noise is high
• AC-readout usually better than dc readout
  eliminates 1/f-noise of an amplifier

20
If a reactive sensor is detected by without
biasing it (x-ray detector) or with a dc bias,
resolution is limited by a preamplifier usually
optimization leads to the matching of reactances
i.e., capacitance of the sensor equals to the
capacitance of the amplifier, x-ray detector,
micromechanical microphone Reason for this is
that when an object is reactive and very high or
very low, noise matching cannot be done. Optimal
source impedance is very close to real. If a
reactive sensor is tuned and ac biased, its
resolution is independent of the noise of the
preamplifier (if noise matched). Resolution is
determined by the sensitivity of the sensor,
power or energy used for its readout and its
temperature.
Note that there is always some limits for energy
or power, the sensor can stand
21
Reactive or dissipative sensor?
Use always, if you can, a reactive sensor. Tune
it mechanically or electrically (or both). The
point is to make the impedance of the sensor real
in order to make noise matching. Use always an ac
bias since it enables impedance transformation
this is usually neede for noise matching.
After this, decrease power of the sensor and
electronics, use high noise but un-expensive
amplifiers until the resolution becomes dominated
by the electronics. Optimization of the
resolution, or sensitivity is not the key point,
The aim is to make a microsensor having low power
consumption, small size, and low price.
22
Passive sensor (dissipative)
If 1/f- noise is included then

• contribution of the electronics decreases
• most sensitive sensor not nessescarily the best

usually 1/f limits the uncertainty of the sensor
Noise temperature
Frequency
23
Nonlinear Sensors

• Nonlinear sensor usually indicates instability
  (negative resistance) in same point of operation.
  Setting the sensor purposely in the unstable
  point high gain can be achieved. Adding the
  strong negative feedback (positive resistance)
  the whole system can be stabilized.
• SQUID, noise cancellation electronics
• Bolometer, thermal feedback based electronics
• MEMS microphone, mechanical amplifier
• This actually means that the sensor itself is
  used a parametric amplifier.

24
Optimizing the sensor

• use reactive sensor if possible
• minimize the dissipations
• use feedback to damp the system electrical
  dissipation

Dc SQUID resistors has to be used to
make system stable (McCumber 1968)
25
Block Diagram of the SQUID Based on the Unshunted
Josephson Junctions, High-Gain SQUID (hg
SQUID) and Unshunted SQUID (un SQUID)
Voltage bias
Energy resolution can be near tens times better
26
Reactive Sensor
Reactive sensor is better than a resistive one
Capasitive sensor tuned with an inductance
limited by the sensor
Resolution is independent of the amplifier
MEMS
Since
then
If then mechanical
fluctuations dominate and thus neither the
amplifier nor the sensor has any influence to
the resolution
and thus high gap and (high voltage) leads to the
better resolution
27
Each measurement problem is solved with
a different measurement method
and/or with a different sensor
fabricating technology
value of the unit, dynamic range, resolution,
uncertainty, stability
MEMS is the most generic fabrication technology
ever MEMS for Sensing Transistor for
Information
28
Different MEMS Technologies
Surface Micromachining Low mass Spring constant
is determined by tensile stress Made of poly
silicon or metal Main application is pressure
sensor Bulk Micromachining High mass Spring is
formed by silicon not in tensile nor in
compressive mode Main applications are
accelerometers and gyroscopes SOI
Micromachining Low mass (insensitive to
movement) Spring formed by single crystal
silicon stable spring constant Not yet common
in market places
29
Readout methods

• Optical (displacement) - expensive, low
  resolution
• Piezoresistive (force) - high noise, drift
• Tunnelling Tip -Readout (displacement)
• - low dynamic range, high noise
• Capacitive (displacement)
• dc bias
• ac readout
• mechanical resonance based ac readout
• electrically tuned resonant rf readout
• both mechanical and electrical tuning
• the use of the unstable pull-in point to
  increase
• force to displacement conversion gain

30
Noise in FET
Noise voltage proportional to channel resistance
Noise current is arising from the thermal noise
from the channel resistance converting into a
current in a a gate capacitance gate leakage
current
Transconductance gain is inversely
proportional to the gate capacitance
In general dispersive sensor and FET preamplifier
leads to
Why, show it
31
Measurement of Capacitance
Maximum readout voltage is set by the pull-in
voltage
Mechanical displacement fluctuation
Best displacement resolution
constat x temperature x frequency
mechanical fluctuations dominates
but if the readout frequency has to be much
higher than the bath temperature, the
noise temperature of the preamplifier tends to
exceed the bath temperature.
32
Resolution of the MEMS
MEMS
Since
then
If then mechanical
fluctuations dominate and thus neither the
amplifier nor the sensor has any influence to
the resolution
and thus high gap and (high voltage) leads to the
better resolution
Since in the small device the mechanical resonant
frequency becomes high, the readout frequency
should be high as well. It is very difficult to
detect thermal or quantum fluctuations directly
using capacitive readout. It may be done by using
a SET or parametric amplifications without an
extra inductance or we can soften the spring
constant and decrease the mechanical
resonant frequency (we have use that i.e., in a
microphone)
33
Dynamics and Forces
can be modified by a feedback
x-dependence
F
2
d
x
d
x
m

m



k
x

F

F

2
s
m
d
t
d
t
F
s
Inertia
Dissipation
Spring Force
Mechanical
Electrical
Random
Force fluctuations
Electrical Force
Mechanical fluctuations
DC Voltage Drive
AC Voltage Drive
Electrical fluctuations
AC Current Drive
Charge Drive
34
Fluctuations
If then
If then force fluctuations close
to the mechanical resonance are dominated by
electrical dissipations
If then
If then effective force
fluctuations are dominated by mechanical
dissipations