Documente Academic
Documente Profesional
Documente Cultură
|
.
|
\
|
+ +
(
|
.
|
\
|
+
= =
kT
E E
n p
kT
E E
n n
N n n p V
T U
A x
De
t i
i p
i t
i n
t i th p n
pix dep
dep
exp exp
) (
2
o o
o o
, (1)
where x
dep
is the width of the depletion region, A
pix
is the area of the pixel, and n
i
is the intrinsic carrier concentration. Given being that besides the numerous
physical parameters involved by expression (1), the temperature dependence of
the depletion dark current due to the semiconductor lattice defects or impurities
includes also the parameters describing the temperature dependence of all
physical quantities from relation (1), it results that the use of this (too intricate)
expression requires to consider some particular cases.
The most important such particular case refers to the region depleted of carriers,
where n and p n
i
. In this case, the expression of the depletion dark current
becomes:
kT
E E
kT
E E
N n V
A x q I
t i
p
i t
n
t i th p n
pix dep dep
+
=
exp exp o o
o o
. (2)
From relation (2), one finds [4] that the depletion dark current reaches a sharp
maximum for the traps with energy:
p
n
i t
E
t
dep kT
E E
dE
dI
dep
I t
o
o
ln
2
0
max ,
= =
|
|
.
|
\
|
. (3)
Taking into account that the trap energy level E
t
and its capture cross-sections of
free electrons
n
and holes
p
, respectively, are independent quantities, the
quantities
kT
E E
i t
and
p
n
o
o
ln
2
1
are not strictly equal, but they have to be of the
same magnitude order for the detectable traps by means of the Dark Current
Spectroscopy Method (DCS) [4, 5].
For this reason, the usual procedure (see e.g. [4, 6], relation (7.8), page 610) is to
take into consideration
kT
E E
i t
and neglect
p
n
o
o
ln :
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Definition and Use of a New Parameter
Characterizing the Depletion Dark Current in Semiconductors 39
kT
E E
kT
E E
N n V
T U
t i i t
t i th
=
exp exp
) (
o
(where:
p n
o o o = = ) (4)
is not rigorous and cannot lead usually to accurate evaluations.
2. Check of the classical approximation (4) relative to the experimental results
concerning the generation rate
The specialty literature provides (e.g. the works [5]) a sufficient number of
experimental data about the generation rates of different deep level traps in
silicon, pointing out even the existence of some still unexplained irregularities:
there is actually more Au present on the Mn scribe line than there is Mn,
although the stronger Mn trap dominates the dark current [5c], p. 478.
For this reason, we studied the predicted by means of approximation (4) ratios
o / cosh ) / (
exp (
ave
i t
kT
E E
s e U , which should be [according to relation (4)] almost
equal for all traps. The obtained results were synthesized by Table 1.
Table 1. Calculated values (in the frame of this work) of the invariant [according to relation (4)]
o / cosh ) / (
exp (
ave
i t
kT
E E
s e U , proportional to the ratio
exp theor
U U
Trap
Generation rate
U
exp
(e/s at 55
o
C)
Trap level
|E
t
E
i
|
(meV) [5c]
Average
cross-section
(cm
2
) [5c]
o / cosh ) / (
exp (
ave
i t
kT
E E
s e U
Mn 6400 ~ < 50 ~ 110
-15
31074.0
Ni; Co 3700 < 30 6.610
-15
1161.5
Pt 970 60 710
-15
1051.0
Au
s
565 < 30 110
-15
1170.6
Fe
i
195 120 150 310
-14
1461.5
Trap 1
(acceptor
E-center)
70 100 210
-15
1615.2
Trap 2 8 200 810
-15
4258.3
Trap 3
(donor
E-center)
1.8 270 210
-14
9100.6
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
40 Ionel Tunaru, Ralf Widenhorn, Erik Bodegom, Dan Iordache
One finds that the approximation (4) leads to evaluations of the generation rate:
(i) of the same magnitude order for Ni, Co and Au (especially), as well as
(though sensibly different) for the Pt, Fe
i
, and the acceptor E-center traps,
(ii) considerably smaller (almost 30 times) than the experimental results
for the Mn traps,
(iii) sensibly smaller, but not so low as for Mn, for the traps
2 (unidentified) and 3 (donor E-center).
As it results from the more recent [7] studies of the Mn traps in Si, the most important
causes of their striking strange overall behavior (see Table 1 and [5c]) are:
a) the large number of Mn defects types in Si: at least 4 different interstitial
charge states Mn
i
, other 3 different substitutional charge states Mn
s
, those of the
cluster Mn
4
, those of the Mn-metal (B, Al, Ga, Sn, Au) pairs (at least 2 charge
states Mn-B, 2 charge states of Mn-Au, etc ; see fig. 3),
b) the huge values interval of the capture cross-sections corresponding to
the different charge states of Mn defects: starting from
p
= 210
-18
cm
2
for the
E
v
+ 0.27 eV state of Mn
i
, passing through the value o
n
= 3.110
-15
cm
2
for the
E
c
0.12 eV and E
c
0.43 eV states of Mn
i
[8], up to o
n
= 2.110
-12
cm
2
for the
level located in proximity of the Fermi level in Si [7a], the value ~ 10
-15
cm
2
indicated by McColgin [5c] being in fact a geometric average of the Mn defects
capture cross-sections over this extremely broad interval.
According to the conclusions of the [7a] work: Mn shows a far more complex
defect structure than Fe or Cr for instant, which do not normally exist in the
substitutional form and also exhibit one single interstitial energy level in the Si band
gap. A consequence of these additional energy levels for Mn is that is difficult to state
a priori which level dominates recombination, as the populations of the various
charge states will depend on the Fermi level, i.e. on the doping type and
concentration. It results that the approximation (4) should be replaced by a
considerably more accurate one, which will be of special interest for the other
deep-level traps (Ni, Co, Pt, Au
s
, and even for the Fe
i
and the acceptor E-center trap).
3. Definition of the new parameter polarization degree of capture
cross-sections
Instead of the above approximation, this work has written (without any additional
approximation) the expression (2) as:
(
= pdg
kT
E E
h
N n V
A x q I
i t
t i th p n
pix dep dep
sec
2
o o
, (5)
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Definition and Use of a New Parameter
Characterizing the Depletion Dark Current in Semiconductors 41
where ) ( sec x h is the hyperbolic secant function:
x
x h
cosh
1
) ( sec = . (6)
The new parameter polarization degree of the capture cross-sections of free
electrons and holes, respectively (pdg), describing the asymmetry of the pair of
capture cross-sections
p n
o o , was defined by this work as:
p
n
p n
p n
pdg p
o
o
o o
o o
ln
2
1
tanh arg
+
=
1
. (7)
Figure 1 below presents the plots of the function ) ( sec p x h + for different values
of the variable x and of the parameter p of the magnitude order of 1. Taking into
account the considerable deviations from the values of the previous factor
(
kT
E E
h
i t
sec of the function
(
pdg
kT
E E
h
i t
sec , even for the very-deep
level traps detectable by the DCS method, significant corrections of the depletion
dark current evaluations are expected to be brought by the expression (5).
Fig. 1. Plots of the hyperbolic secant function sech(x
+
p) for different values of the parameters
) ( ) ( kT E E x
i t
=
and pdg p (polarization degree of the capture cross-sections).
The thick vertical line through x = +2 indicates the descent of the sech(x
+ p) function
from the value +1 for p = 2, up to less than + 0.05 for p = + 2.5.
1
The traps of the crystalline lattice (embedded nano-particles/systems, nano-defects) capture both
free electrons and holes, but with different probabilities. Given being the capture probability is
proportional to the corresponding cross-section (
n
or
p
), the capture probability asymmetry can
be given by the expression (7) [somewhat similar to those from Optics, Nuclear physics, etc.],
because: pdg > 0 corresponds to prevalent free electrons capture, pdg = 0 to equal capture
probabilities, etc.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
42 Ionel Tunaru, Ralf Widenhorn, Erik Bodegom, Dan Iordache
4. Existing experimental results about both capture cross-sections (of free
electrons and holes, respectively) and their polarization degree (pdg)
Despite: a) the high technical interest presented by the identification of
contaminants (see e.g. [5], [6]), b) the possibility to determine experimentally
both capture rates for a given trap (by means of the DLTS method [9], especially),
in practice there were evaluated: (i) usually up only the (geometrical) averages
p n
o o o =
, (ii) both capture cross-sections only for a part of the different studied
traps. Table 2 synthesizes some known values of both capture cross-sections of
free electrons and holes, respectively, by certain defects (traps) from silicon, as
well as of their polarization degree (pdg), implicitly. In order to understand easier
the location of the different defects (traps) in Si, instead to indicate their positions
relative to the upper limit of the valence band (Ev) or to the bottom limit of the
conduction band (E
c
), we give in Table 2 the evaluated traps energies relative to
the intrinsic Fermi level, considered as E
i
0.54 eV [4].
Table 2. Some known values of both capture cross-sections of free electrons and holes,
respectively, in silicon, as well as of their polarization degree (pdg), implicitly
Trap Group
Energy
(eV)
n
(cm
2
)
p
(cm
2
)
k
= o
n
/o
p
pdg Ref.
U (55C)
e/s [5c]
Tii
+
4 E
i
+ 0.27 3.110
-14
1.410
-15
22.14 +1.549 [7d]
Tii
++
4 E
i
0.28 1.310
-14
2.810
-17
464.3 +3.070 [7d]
Vi
++
5 E
i
0.18 5.010
-14
3.010
-18
16667 +4.86 [7d]
Cri
++
6
E
i
+ 0.32 2.310
-13
1.110
-13
2.091 +0.369 [7d]
E
i
+ 0.30 2.010
-14
4.010
-15
5 +0.805 [10a]
(Cri
+
Bs
-
)
-
6; 13 E
i
0.26 5.010
-15
1.010
-14
0.5 0.3466 [10a]
Moi
+
6
E
i
0.26 1.610
-14
6.010
-16
26.67 +1.642 [10b]
E
i
0.223 7.810
-15
6.010
-16
13 +1.282 [8b]
Mni
+
7 E
i
+ 0.09 9.4 +1.1204 [7a]
(Mni
+
Bs
-
)
+
7; 13 E
i
+ 0.01 2.110
-12
3.510
-13
6.0 +1.282 [7a]
Mni
++
7 E
i
0.21
23.1
(18.528.3)
+1.57 [7a]
all Mn
Traps
7 o = (o
min
o
max
)
1/2
1.010
-15
; o
max
/o
min
~ 10
6
[5c],
[7a]
6400
Fei
+
8 E
i
0.16 5.010
-14
7.010
-17
714.3 +3.286 [7d]
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Definition and Use of a New Parameter
Characterizing the Depletion Dark Current in Semiconductors 43
(FeI
+
Bs
-
)
+
8; 13
E
i
+ 0.28
( 0.02)
1.410
-14
( 0.02)
1.110
-15
(0.52.5)
13 +1.282 [7c]
Coi
+
9
E
i
0.02 o = (o
n
o
p
)
1/2
6.610
-15
[5c] 3700
Nii
+
10
Pts
-
10 E
i
+ 0.31 3.410
-15
[11]
Pti
-
10 E
i
+ 0.02 4.510
-15
1.0910
-14
2.42 +0.442 [11] 970
Pts
+
10 E
i
0.18 5.410
-14
[11]
Aus
-
11
E
i
0.01 1.410
-16
7.610
-15
0.01842 - 1.997 [7d]
5.010
-16
1.010
-15
T
- 4
0.5 -0.3466 [12c] 565
Zns
- -
12 E
i
+ 0.07 1.310
-19
6.610
-15
1.9710
- 5
-5.417 [7d]
Zns
-
12 E
i
0.21 1.510
-15
4.410
-15
0.3409 - 0.538 [7d]
PVp
-
E-center
14
E
i
+ 0.10 o = (o
n
o
p
)
1/2
6.610
-15
[13a] 70
E
i
+ 0.084 3.710
-16
[13b]
PV
+
14 E
i
0.27 [13a] 1.8
From Table 2 (see e.g. the results referring to the traps Cr
i
++
, Mo
i
+
, Au
s
-
, PV-pair
-
acceptor E-center), there result also the accuracies (not too good) of the capture cross-
sections evaluations, and of their polarization degree. One finds also that the
recombination rate U was not still measured systematically. Given being the capture
cross-sections and their polarization degree pdg are often temperature dependent,
table 3 synthesizes the main types of the temperature dependence of the capture
cross-sections.
Table 3. Main types of the temperature dependence of the capture cross sections of free electrons
and/or holes, respectively in different semiconductors
The type of the
temperature dependence
Typical expression of
temperature dependence
Some examples for the
traps in the semiconductor
Reference
Arrhenius law o
n
(T) ~ o
n
exp(-E
B
/kT) EL2, HL1, HL10 in GaAs [14a]
Power laws o(T) T
-n
n ~ 0.21 (o
n, Au
in n-Si) [12d]
n ~ 2.5 (o
n, Au
in p-Si) [12c]
n ~ 4.0 (o
p, Au
in n-Si) [12c]
Temperature
independence
o(T) ~ constant
o
p
in GaAs [14a]
o
n, Au
in n-Si [12c]
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
44 Ionel Tunaru, Ralf Widenhorn, Erik Bodegom, Dan Iordache
5. Experimental evaluation of the difference E
t
E
i
In order to interpret the obtained (calculated) values of E
t
E
i
corresponding
sometimes to a mixture of defects inside each studied pixel, an experimental
evaluation of the difference E
t
E
i
is necessary. Given being that the most
efficient experimental method intended to the defects location inside the studied
semiconductor forbidden band is the deep-level transient spectroscopy (DLTS)
method, we have to evaluate the E
t
E
i
difference starting from the defects (traps)
positions relative to the upper limit E
v
of the valence band and the bottom limit
(edge) E
c
of the conduction one (see fig. 2).
Fig. 2. Positions of the defects (traps) energy levels and of the Fermi intrinsic level
according to the DLTS results.
Given being for a certain temperature T the energy gap is (approximately) known,
it is necessary to evaluate also the intrinsic Fermi energy E
i
(T) at this temperature.
According to the basic Condensed Matter Physics treatises [2], the temperature
dependence of the intrinsic Fermi energy E
i
(T) is given by the expression:
*
*
ln
4
3
) (
2
1
) (
dn
dp
g i
m
m
kT
T E T E + = , (8)
where:
3 / 2
2 / 3
*
2 / 3
* *
(
+ =
ph pl dp
m m m and:
3 *
2
* *
6
l t dn
m m m = (9)
are the effective masses corresponding to the states densities of holes and free
electrons, respectively, while
* * *
, ,
l ph pl
m m m and
*
t
m are the effective masses
associated to the light and heavy holes, respectively, and to the longitudinal and
transverse electrons effective masses, relative to the major axis of the energy
ellipsoid E( k ). For free electrons in silicon, there are used the values
o l
m m 98 . 0
*
= ,
o t
m m 19 . 0
*
= , and
o l
m m 916 . 0
*
= ,
o t
m m 225 . 0
*
= , which lead to
the values
o dn
m m 5965 . 0
*
1
~ and
o dn
m m 6528 . 0
*
2
~ of the effective mass
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Definition and Use of a New Parameter
Characterizing the Depletion Dark Current in Semiconductors 45
associated to the electrons state density in the conduction band of Si, while for the
holes from Si there are usually used the values
o pl
m m 16 . 0
*
= ,
o ph
m m 49 . 0
*
= ,
which lead to the value
o dp
m m 5492 . 0
*
~ of the effective mass associated to the
holes state density in the silicon valence band. As we use the first or the second
pair of electron effective mass in Si, we obtain 06196 . 0 ln
4
3
*
*
~
dn
dp
m
m
, or 0.1296,
hence the temperature coefficient of the silicon intrinsic Fermi energy correction:
1 5
*
10 116 . 1 ln
4
3
s K
dn
m
k dp
is at least one magnitude order less than that of the
energy gap corresponding to silicon: ) ( 10 2 . 4 21 . 1 ) (
4
,
eV T T E
Si g
~ [2].
For an average use temperature T 250 K (this work, [12a], etc.), one finds that
the intrinsic Fermi energy correction
*
*
ln
4
3
dn
dp
m
m
kT
has a value inside the interval
(1.34; 2.8) meV, remaining so negligible relative to the usual defects (traps)
energies in Si, even for the very-deep level defects, detectable by the studied dark
current spectroscopy method.
6. On the evaluation of the polarization degree of capture cross-sections of
free electrons and holes, respectively
Taking into account that the power law expressions of the temperature
dependence of capture cross-sections can be approximated by Arrhenius type
expressions (with negative activation energy E
B
) and substituting the Arrhenius
expression [see Table 3] in the polarization degree definition (7), one obtains:
kT
E
pdg
kT
E
T
T
T pdg
B B
p
n
p
n
2 2
ln
2
1
) (
) (
ln
2
1
) ( = = =
o
o
o
o
, hence the temperature
dependence of the hyperbolic secant argument can be expressed as:
( )
+
= +
= +
= pdg
kT
E E
pdg
kT
E E E
T pdg
kT
E E
T h
eff
i t
B i t i t
2 /
) ( ) ( sec arg , (10)
by means of the effective energies difference ( )
eff
i t
E E and of the pdg(T)
asymptotic value
pdg .
The DCS method allows however an indirect evaluation of the polarization
degree pdg (in fact of its asymptotic value
|
.
|
\
|
+
1
sec arg . (11)
In order to provide correct interpretations of the results of our procedure (10), we
studied the existing experimental data for 20 randomly chosen pixels of a Spectra
Video CCD camera (model SV512V1) manufactured by Pixelvision, Inc. (see
[9a]). The obtained (in the frame of this work) types of qualitative results and
their interpretations are synthesized by Table 4.
Table 4. Main types of results obtained by means of the regression line
|arg[sech(pdg+(E
t
-E
i
)/kT]|= i + s/kT study and their interpretation
Slope s
sign
Intercept i
sign
Interpretation
Examples of
pixels [12a]
> 0 > 0
Both (E
t
E
i
)
eff
and pdg
= i and: (E
t
E
i
)
eff
/pdg
> 0
188; 471
> 0 < 0
(E
t
E
i
)
eff
/kT > - pdg
> 0 or
(E
t
E
i
)
eff
/kT < - pdg
< 0
121; 200
< 0 > 0
pdg
> - (E
t
E
i
)
eff
/kT = (E
i
E
t
)
eff
/kT > 0 or
pdg
< - (E
t
E
i
)
eff
/kT = (E
i
E
t
)
eff
/kT < 0
321; 400
In order to estimate the impact of the polarization degree (pdg) values on the
depletion dark current the obtained numerical results for the 17 pixels whose data
sets lead to physically convergent evaluations were synthesized by Table 5. As
one can find easily, the accuracy of the obtained results decreases very much for
the pixels emitting weak depletion dark current (low values of the depletion pre-
exponential factor, Dep).
Table 5. Interpretation of the obtained quantitative (numerical) results [s = sign(E
t
-E
i
)
eff
]
Pixel
Dep (Mcps K
-3/2
),
10
6
counts/s K
-3/2
Obtained information about
the effective parameters
Calc. Depletion
Dark Current
Accuracy (%)
sech[pdg+
(E
t
-E
i
)/kT]
|E
t
E
i
|
eff
, meV s pdg
321; 400 2.541 32.125 3.075 17.47% 0.3808
181; 260 5.433 24.39 0.0187 2.975% 0.5887
121; 200 6.022 34.44 0.3274 5.65% 0.5381
141; 220 6.478 8.49 0.773 6.158% 0.5872
101; 180 6.623 24.66 0.0455 2.454% 0.5702
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Definition and Use of a New Parameter
Characterizing the Depletion Dark Current in Semiconductors 47
301; 380 7.224 30.85 0.3114 3.002% 0.6068
188; 471 9.530 28.04 0.790 6.988% 0.2511
81; 160 11.710 33.59 0.0154 2.294% 0.4117
221; 300 15.998 28.03 0.438 3.797% 0.2176
201; 280 21.944 31.21 0.117 1.77% 0.4140
29; 88 32.120 51.860 0.155 2.473% 0.2207
61; 140 40.792 27.352 1.369 7.055% 0.1467
281; 360 66.386 45.697 0.397 3.197% 0.1688
261; 340 105.688 57.225 0.135 2.2798% 0.1306
341; 420 254.039 55.479 0.259 2.277% 0.1249
161; 240 268.671 76.223 0.632 3.73% 0.0338
31; 247 421.361 2.16 1.3896 5.379% 0.5109
Figures 3 and 4 present the diagrams of the traps levels for GaAs and Mn in the Si
lattice, respectively. From Table 2, one finds that the majority of donor /acceptor
states are located in the upper/lower half of the forbidden band, respectively,
while the other states named here trans-Fermi level donor/ acceptor states (as
the levels EL2, E5, HL10, HL16 in GaAs, (Mn
+
Au
-
)
+
and Mn
i
++
in the Si lattice)
have negative values of the product sign( )
t i
pdg E E .
Fig. 3. Traps levels diagram for GaAs [14a].
Fig. 4. Traps levels diagram for Mn in the silicon lattice [7, 8, 15].
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
48 Ionel Tunaru, Ralf Widenhorn, Erik Bodegom, Dan Iordache
Taking into account that the number of the trans-Fermi level states is sensibly
less than the number of the other electronic states, this finding represents a useful
tool in the assignment of these states (traps).
6. Study of the anomalies of the generation rate U values
This work achieved the least-squares fit (using the gradient method [17], [16b]) of
the improved approximation (5):
(
|
|
.
|
\
|
+
|
|
.
|
\
|
=
pdg
kT
E E
h
kT
E
T De
kT
E
T De De
eff i t g
dep
g
diff
) (
sec
2
exp exp
2 / 3
, 0
3
, 0
of the experimental results concerning the dark current emitted by some pixels of
the studied Spectra Video CCD camera (model SV512V1) [12a].
There were obtained evaluations of all parameters of the above expression, particularly
of the depletion pre-exponential factor Dep De
dep
, 0
, effective energy gap E
g
and of
the trap location
eff i t
E E ) ( inside the silicon forbidden band, as well as of the
asymptotic value pdg
|
|
.
|
\
|
=
pdg
kT
E E
h
kT
E
T De T U
gr
eff i t
gr
g
gr dep gr
) (
sec
2
exp ) (
2 / 3
, 0
. (12)
The obtained results are presented in Table 6, where the pixels were written in the
monotonic order of their increasing depletion pre-exponential factors Dep. While
according to the classical approximation (4) it was expected to find the
proportionality of the generation rate U(T) with the depletion pre-exponential factor
dep
De Dep
, 0
, the examination of Table 6 points out several discontinuities, that
can be explained only by the different values of polarization degree pdg
and of the
effective energies difference ( )
eff
i t
E E .
Table 6. Comparison of the evaluated values of the depletion pre-exponential factor Dep and of the
generation rate (corresponding hyperbolic secant function values in brackets) for different
pixels of a SpectraVideo CCD camera (model SV512V1)
Pixel
Dep (Mcps K
3/2
),
10
6
counts/s K
3/2
Evaluated
generation rate
U(e
-
/s) at 55C
Pixel
Dep (Mcps K
-3/2
),
10
6
counts/s K
3/2
Evaluated
generation rate
U(e
-
/s) at 55C
321; 400 2.541 124.6 (0.3808) 201; 280 21.944 2243 (0.4140)
181; 260 5.433 680 (0.5887) 29; 88 32.120 1981 (0.2207)
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Definition and Use of a New Parameter
Characterizing the Depletion Dark Current in Semiconductors 49
121; 200 6.022 866 (0.5381) 61; 140 40.792 1464 (0.1467)
141; 220 6.478 713.5 (0.5872) 281; 360 66.386 3184 (0.1688)
101; 180 6.623 751 (0.5702) 261; 340 105.688 3276 (0.1306)
301; 380 7.224 1564 (0.6068) 341; 420 254.039 3899 (0.1249)
188; 471 9.530 588 (0.2511) 161; 240 268.671 1996 (0.0338)
81; 160 11.710 1176 (0.4117) 31; 247 421.361 4923 (0.5109)
221; 300 15.998 1654 (0.2176)
The considerable impact of the polarization degree (pdg) on the values of the
hyperbolic secant
( )
(
(
pdg
kT
E E
h
eff
i t
sec and of the generation rate, is
indicated both by the values of the last column of Table 5 (see the value for the
pixel 161; 240, particularly) and of Fig. 1.
According to the work [5c], the even more striking large generation rates of the Mn,
Ni, Co traps (see e.g. Table 1) could be explained by means of a complex character
(like Mn
4
) of these traps and/or of a possible Poole-Frenkel [18] effect. The
accomplished analysis pointed out that even the capture cross-sections of the Ni and
Co traps are considerably larger than that of the substitutional gold Au
s
, their
generation rates agree very well [according to the classical approximation (4), see
Table 1, with that of Au
s
, while even the use of the improved approximation (5) does
not lead to a quantitative justification of the experimental value of the Mn trap
generation rate.
Given being the: a) considerably more complex defect structure (larger number of
charge states) of Mn than those of Cr, Fe, Ni, Co, etc, b) still incompletely
characterized (mainly by means of the capture cross sections of free electrons and
holes, respectively) of many Mn defects (e.g. of the substitutional states, of the
Mn
4
cluster, of the Mn pairs with Al, Ga, Sn, etc), c) intentional contamination
with Mn atoms which lead to the McColgins results [5c], our analysis lead to the
conclusion that the striking strange pair of values referring to the average cross-
section and to the generation rate/Mn atom can be explained (more than by the
Poole-Frenkel effect) by an averaging of the generation rates of Mn defects in
different charge states with very different capture cross sections (many of them
considerably higher than the average value indicated by [5c]).
The obtainment of new experimental information about both free electrons o
n
and
holes o
p
capture cross-sections of the main deep-level traps in silicon will
contribute of course to a more accurate description of the temperature dependence
of the depletion dark current in semiconductors.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
50 Ionel Tunaru, Ralf Widenhorn, Erik Bodegom, Dan Iordache
7. Use of the polarization degree pdg as an assignment criterion of the
deep-level traps
The accomplished study pointed out that the newly defined physical parameter
polarization degree of the capture cross-sections of free electrons and holes,
respectively is a basic feature of the deep-level traps, allowing considerably more
accurate descriptions of the temperature dependence of the depletion dark current
in charge-coupled devices.
For this reason, we consider that besides the classical assignment criteria as the:
a) depletion pre-exponential factor
dep
De
, 0
, b) generation rate U(T
g.r.
) at a given
temperature, c) the location of the studied trap inside the semiconductor forbidden
band, d) average capture cross-section
p n ave
o o o o = , the: e) polarization
degree
p n
p n
p
n
pdg
o o
o o
o
o
+
= = tanh arg ln
2
1
of capture cross-sections, as well as the:
f) average value | |
+ pdg kT E E h
ave eff i t
/ ) ( sec
of the hyperbolic secant function,
are also basic assignment criteria for the deep-level traps in a semiconductor,
starting from the observed temperature dependence of the dark current in each
CCD pixel (the Dark Current Spectroscopy method).
Conclusions
This work pointed out the necessity to introduce the new parameter polarization
degree of the capture cross-sections of free electrons and holes, respectively in
order to ensure a sufficiently accurate description of the temperature dependence
of the depletion dark current in semiconductors. It was found also the ability of
this new parameter to: a) analyze successfully the anomalies of some reported
generation rate values [e.g., no anomaly for the reported generation rates of Ni and
Co [5c], but probably wrong value (1210
14
cm
2
, instead of ~ 110
15
cm
2
) for
the average cross-section of the electronic states of the embedded Mn traps which
produce a 6400 e
-
/s generation rate], b) discriminate among the contributions of
the different deep-level traps to the depletion dark current in each CCD pixel,
allowing so certain assignments of the deep-level traps in the CCDs pixels. In this
aim, both newly introduced notions of polarization degree and trans-Fermi
level donor/acceptor states are useful.
Acknowledgements
The authors thank very much for the fruitful discussions and important received
information to Dr. G. Moagar-Paladian, National Institute for Research and
Development in Micro-technology, Bucharest.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Definition and Use of a New Parameter
Characterizing the Depletion Dark Current in Semiconductors 51
R E F E R E N C E S
[1] C. Sah, R. Noyce, W. Shockley, Carrier Generation and Recombination in p-n Junction and p-n
Junction Characteristics, in Proc. IRE, vol. 45, 1957, p. 1228.
[2] a) A. S. Grove, Physics and Technology of Semiconductor Devices, New York, Wiley, 1967;
b) S. M. Sze, Physics of Semiconductor Devices, New York, Wiley, 1981; c) I. Munteanu, Fizica
strii condensate, Editura Hyperion, Bucureti, 1995.
[3] R. Widenhorn, M. M. Blouke, A. Weber, A. Rest, E. Bodegom, Temperature dependence of
dark current in a CCD, in Proc. SPIE Int. Soc. Opt. Eng., vol. 4669, 2002, pp. 193-201.
[4] R. D. McGrath, J. Doty, G. Lupino, G. Ricker, J. Vallerga, Counting of Deep-Level Traps
using a Charge-Coupled Device, in IEEE Trans. on Electron Devices, vol. ED-34, no. 12, 1987,
pp. 2555-2557.
[5] a) W. C. McColgin, J. P. Lavine, J. Kyan, D. N. Nichols, C. V. Stancampiano, Dark Current
Quantization in CCD Image sensors, in Proc. Int. Electron Device Meeting, pp. 113-116,
13-16 December 1992; b) W. C. McColgin, J. P. Lavine, C. V. Stancampiano, Probing metal
defects in CCD image sensors, in Proc. Mat. Res. Soc. Symp., vol. 378, pp. 713-724, 1995;
c) W. C. McColgin, J. P. Lavine, C. V. Stancampiano, J. R. Russell, Deep-level traps in CCD
image sensors, in Proc. Mat. Res. Soc. Symp., vol. 510, pp. 475-480, 1998.
[6] J. R. Janesick, Scientific Charge-Coupled Devices, SPIE Press, The International Society for
Optical Engineering, Bellingham, Washington, 2001.
[7] a) T. Roth, P. Rosenitz, S. Diez, S. W. Glunz, D. Mcdonald, S. Beljakowa, G. Pensl, Electronic
properties and dopant pairing behavior of manganese in boron-doped silicon, in J. Appl. Physics,
102, 103716-8, November 2007; b) P. Rosenitz, T. Roth, S. Diez, D. Mcdonald, S. W. Glunz,
Detailed studies of manganese in silicon using lifetime spectroscopy and deep-level transient
spectroscopy, Proc. 22
nd
Eur. Conf. Photo-voltaic Solar Energy, EU-PVSEC 2007, Milan, Italy,
3-7 September 2007, pp. 1480-1483(2007); c) J. E. Birkholz, K. Bothe, D. Mcdonald, J. Schmidt,
Electronic properties of iron-boron pairs in crystalline silicon by temperature- and injection-level-
dependent lifetime measurements, in J. Appl. Phys., vol. 97, 2005, 103708; d) D. Mcdonald,
L. J. Geerligs, Recombination activity of interstitial iron and other transition metal point defects in
p- and n-type crystalline silicon, in Appl. Phys. Lett., vol. 85, no. 18, November 2004,
pp. 4061-4063.
[8] a) K. Graff, Metal impurities in silicon-device fabrication, vol. 24, Springer, Berlin, 1999;
b) S. Rein, Lifetime spectroscopy: a method of defect characterization in Si for photo-voltaic
applications, Springer, Berlin, 2005.
[9] D. V. Lang, Deep-level transient spectroscopy: A new method to characterize traps in semi-
conductors, in Journ. Appl. Physics, vol. 45, no. 7, July 1974, pp. 3023-3032.
[10] a) J. Schmidt, R. Krain, K. Bothe, G. Pensl, S. Beljakava, Recombination activity of
interstitial chromium and chromium-boron pairs in Si, in J. Appl. Phys., vol. 102, 2007, 123701;
b) B. B. Pondyal, K. R. McIntosh, D. H. Mcdonald, G. Coletti, Temperature dependent e
-
and h
capture cross sections of Mo in silicon, Proc. 24
th
European Photo-voltaic Solar Energy Conf.,
21-25 September 2009, Germany.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
52 Ionel Tunaru, Ralf Widenhorn, Erik Bodegom, Dan Iordache
[11] Y. K. Kwon, T. Ishikawa, H. Kuwano, Properties of platinum-associated deep levels in silicon,
in J. Appl. Phys., vol. 63, February 1987, pp. 1055-1058.
[12] a) A. G. Nassibian, L. Faraone, Capture cross section of gold in silicon, in Appl. Phys.
Letters, vol. 33, 1978, p. 451; b) W. D. Davis, Lifetimes and capture cross sections in gold-doped
silicon, in Phys. Rev., vol. 114, 1959, pp. 1006-1008; c) G. Bemski, Recombination properties of
gold in silicon, in Phys. Rev., vol. 111, 1958, pp. 1515-1518; d) J. Barbolla, M. Pugnet,
J. C. Brabant, M. Brousseau, Thermal capture cross section of free electrons at neutral gold
centres in n-type silicon, in Phys. stat. sol. (a), vol. 36, no. 2, 1976, pp. 495-498.
[13] a) A. N. Larsen, A. Mesli, The hidden secrets of the E-centre in Si and Ge, in Physica B, vol.
401-402, 2007, pp. 85-90; b) J. B. Spratt, B. C. Passenheim, R. E. Leadon, The effects of nuclear
radiation on P-channel CCD imagers, Proc. Radiation Effects Data Workshop, IEEE, 1997,
pp. 116-121.
[14] a) A. Mitonneau, A. Mircea, G. M. Martin, D. Pons, Electron and hole capture cross-sections
at deep centres in gallium arsenide, Rev. Phys. Applique, vol. 14, October 1979, pp. 853-861;
b) D. V. Lang, R. A. Logan, in J. Electron. Mat., vol. 4, 1975, p. 1053; c) C. H. Henry, D. V. Lang,
in Phys. Rev. B, vol. 13, 1977, p. 989; d) P. E. Sterian, R. Rogojan, A. R. Sterian, Alumino-
phosphate semiconductor, Proc. SPIE Laser Physics and Applications, 4394, 2001, pp. 358-361.
[15] a) H. Lemke, Energieniveaus und Bindungsenergien von Ionenpaaren in Silizium, in Phys.
Stat. Sol. (a) vol. 76, 223-234(1983); b) H. Lemke, Eigenschaften einiger Strstellen-komplexe
von Gold in Silizium, in Phys. Stat. Sol. (a) vol. 75, 473-482(1983); c) E. R. Weber, Transition
Metals in Silicon, in Appl. Phys., vol. A30, 1-22(1983).
[16] a) R. Widenhorn, E. Bodegom, D. Iordache, I. Tunaru, Computational Approach to Dark
Current Spectroscopy in CCDs as Complex Systems. I. Experimental Part and Choice of the
Uniqueness Parameters", in the Scientific Bulletin of "Politehnica" University Bucharest, Series A:
Applied Mathematics and Physics, vol. 72, no. 4, 2010, pp. 197-208; b) I. Tunaru, R. Widenhorn,
D. Iordache, E. Bodegom, ibid. II. Numerical Analysis of the Uniqueness Parameters Evaluation,
ibid., 73(1) 149-162(2011).
[17] a) K. Levenberg, A method for the solution of certain problems in least squares, Quarterly of
Applied Mathematics, vol. 2, 1944, pp. 164168; b) D. W. Marquardt, An algorithm for least-
squares estimation of nonlinear parameters, Journal of the Society for Industrial and Applied
Mathematics, vol. 11, no. 2, 1963, pp. 431441; c) N. Andrei, Another Conjugate Gradient
Algorithm for Unconstrained Optimization, Annals of the Academy of Romanian Scientists, Series
on Science and Technology of Information, vol. 1, no. 1, 2008, pp. 7-20.
[18] A. Sevcenco, Gh. Brezeanu, An accurate analytical model with a reduced parameter set for
short-channel MOS transistors, Annals of the Academy of Romanian Scientists, Series on Science
and Technology of Information, vol. 2, no. 1, 2009, pp. 107-120.
[19] a) J. Frenkel, On pre-breakdown phenomena in insulators and semiconductors, in Phys. Rev.
vol. 54, 1938, p. 647; b) J. van der Spiegel, G. J. Declerk, Characterization of Dark Current Non-
Uniformities in CCDs, in Solid-State Electronics, vol. 27(2) 1984, pp. 147-154; c) W. E. Meyer,
Digital DLTS studies on radiation induced defects in Si, GaAs and GaN, PhD Diss., Univ. of
Pretoria (South-Africa), Nov. 2006; d) E. A. G. Webster, R. Nicol, L. Grant, D. Renshaw,
Validated Dark Spectroscopy on a per-pixel basis in a CMOS image sensors, Proc. International
Image Sensor Workshop, Bergen (Norway), 26-28 July 2009.
[20] A. Danchiv, M. Bodea, Cl. Dan, Chopper Stabilization Techniques, part I: Chopper Amplifier
Topologies Overview, Annals of the Academy of Romanian Scientists, Series on Science and
Technology of Information, vol. 1, no. 1, 2008, pp. 31-42.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Annals of the Academy of Romanian Scientists
Series on Science and Technology of Information
ISSN 2066-8562 Volume 5, Number 1/2012 53
AUTOMATIC COMPUTER
MUSIC CLASSIFICATION AND SEGMENTATION
Adrian SIMION
1
, Stefan TRAUSAN-MATU
2
Rezumat. Lucrarea de fa descrie i aplic diferite metode pentru segmentarea
automat a muzicii realizat cu ajutorul unui calculator. Pe baza rezultatelor i a
tehnicilor de extragere a caracteristicilor folosite, se ncearc de asemenea o
clasificare/recunoatere a fragmentelor folosite. Algoritmii au fost testai pe seturile de
date Magnatune i MARSYAS, dar instrumentele software implementate pot fi folosite pe
o gam variat de surse. Instrumentele descrise vor fi integrate ntr-un framework /
sistem software numit ADAMS (Advanced Dynamic Analysis of Music Software -
Software pentru Analiza Dinamic Avansat a Muzicii) cu ajutorul cruia se vor putea
evalua i mbunti diferitele sarcini de analiz i compoziie a muzicii. Acest sistem are
la baz biblioteca de programe MARSYAS i conine un modul similar cu WEKA pentru
sarcini de procesare a datelor i nvare automat.
Abstract. This paper describes and applies various methods for automatic computer
music segmentation. Based on these results and on the feature extraction techniques used,
is tried also a genre classification/recognition of the excerpts used. The algorithms were
tested on the Magnatune and MARSYAS datasets, but the implemented software tools can
also be used on a variety of sources. The tools described here will be subject to a
framework/software system called ADAMS (Advanced Dynamic Analysis of Music
Software) that will help evaluate and enhance the various music analysis/composition
tasks. This system is based on the MARSYAS open source software framework and
contains a module similar to WEKA for data-mining and machine learning tasks.
Keywords: automatic segmentation, audio classification, music information retrieval, music
content analysis, chord detection, vocal and instrumental regions
1. Music Information Retrieval
The number of digital music recordings has a continuous growth, promoted by the
users interest as well as the advances of the new technologies that support the
pleasure of listening to music. There are a few reasons that explain this trend, first
of all, the existential characteristic of the musical language. Music is a form of art
which can be shared by people that belong to different cultures because it
surpasses the borders of the national language and of the cultural background. As
an example the West American music has many enthusiasts in Japan, and many
persons in Europe appreciate the classical Indian music. These forms of
1
Eng., Ph.D. student, Faculty of Automatic Control and Computers, University Politehnica of
Bucharest, Bucharest, Romania, (simion.adrian@gmail.com).
2
Corresponding member of AOSR. Prof., Ph.D., Faculty of Automatic Control and Computers,
University Politehnica of Bucharest, Bucharest, Romania, (stefan.trausan@cs.pub.ro).
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
54 Adrian Simion, tefan Truan-Matu
expression can be appreciated without the need of a translation that is in most of
the cases necessary for accessing foreign textual papers.
Another reason is the fact that technology for recording music, digital
transformation and playback allows the users access to information that is almost
comparable to live performances, at least at audio quality level.
Last, music is an art form that is cult and popular at the same time and sometimes
is impossible to draw a line between the two, like jazz and traditional music.
The high availability and demand for music content induced new requirements
about its management, advertisement and distribution. This required a more in-
depth and direct analysis of the content than that provided by simple human
driven meta-data cataloguing.
The new techniques allowed approaches that were only encountered in theoretical
musical analysis. One of these problems was stated by Frank Howes [1]: There is
thus a vast corpus of music material available for comparative study. It would be
fascinating to discover and work out a correlation between music and social
phenomena. With the current processing power and advancements we can answer
questions such as: What is the ethnic background of a particular piece of music or
what cultures it spawns.
In light of these possibilities and technological advances we needed a new
discipline that would try to cover and answer the various problems. Music
Information Retrieval (MIR) is an interdisciplinary science that retrieves its
information from music. The origins of MIR are domains like: musicology,
cognitive psychology, linguistic and computer science.
An active research area is composed of new methods and tools for pattern finding
as well as the comparison of musical content. The International Society for Music
Information Retrieval [2] is coupled with the annual Music Information Retrieval
Evaluation eXchange (MIREX) [3]. The evaluated tasks include Automatic Genre
Identification, Chord Detection, Segmentation, Melody Extraction, Query by
Humming, to name a few. This paper will focus mostly on Automatic
Segmentation and Genre Identification.
2. Former studies and related work on Automatic Music Segmentation
The topic of speech/music classification was studied by many researchers. While
the applications can be very different, many studies use similar sets of acoustic
features, such as short time energy, zero-crossing rate, cepstrum coefficients,
spectral roll off, spectrum centroid and loudness, alongside some unique
features, such as dynamism. However, the exact combinations of features used
can vary greatly, as well as the size of the feature set.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Automatic Computer Music Classification and Segmentation 55
Typically some long term statistics, such as the mean or the variance, and not the
features themselves, are used for the discrimination.
The major differences between the different studies lie in the exact classification
algorithm, even though some popular classifiers (K-nearest neighbor, Gaussian
multivariate, neural network) are often used as a basis.
For the studies, mostly, different databases are used for training and testing the
algorithm. It is worth noting that in these studies, especially the early ones, these
databases are fairly small. The following table describes some of the former
studies:
Table 1. Some of the former studies
Author Application Features Classification method
Saunders,
1996 [4]
Automatic real-time FM
radio monitoring
Short-time energy, statistical parameters of
the ZCR
Multivariate Gaussian
classifier
Scheirer and
Slaney, 1997
[5]
Speech/music
discrimination for
automatic speech
recognition
13 temporal, spectral and cepstral features
(e.g., 4Hz modulation energy, % of low
energy frames,
spectral roll off, spectral centroid, spectral
flux, ZCR, cepstrum-based feature,
rhythmicness),
variance of features across 1 sec.
Gaussian mixture model
(GMM), K nearest
neighbour (KNN), K-D
trees, multidimensional
Gaussian MAP estimator
Foote, 1997
[6]
Retrieving audio
documents by acoustic
similarity
12 MFCC, Short-time energy
Template matching of
histograms, a tree-based
vector quantizer,
trained to maximize mutual
information
Liu et al.,
1997 [7]
Analysis of audio for
scene classification of
TV programs
Silence ratio, volume std, volume dynamic
range, 4Hz freq, mean and std of pitch
difference,
speech, noise ratios, freq. centroid,
bandwidth, energy in 4 sub-bands
A neural network using the
one-class-in-one network
(OCON) structure
Zhang and
Kuo, 1999 [8]
Audio
segmentation/retrieval
for video scene
classification, indexing
of raw audio visual
recordings, database
browsing
Features based on short-time energy,
average ZCR, short-time fundamental
frequency
A rule-based heuristic
procedure for the coarse
stage, HMM for the second
stage
Williams and
Ellis, 1999
[9]
Segmentation of speech
versus non speech in
automatic speech
recognition tasks
Mean per-frame entropy and average
probability dynamism, background-label
energy ratio, phone distribution match
all derived from posterior probabilities of
phones in hybrid connectionist-HMM
framework
Gaussian likelihood ratio
test
El-Malehet
al., 2000 [10]
Automatic coding and
content based
audio/video retrieval
LSF, differential LSF, measures based on
the ZCR of high-pass filtered signal
KNN classifier and
quadratic Gaussian
classifier (QCG)
Buggati et al.,
2002 [11]
Table of Content
description of a
multimedia document
ZCR-based features, spectral flux,
shorttime energy, cepstrum coefficients,
spectral centroids, ratio of the high-
frequency power spectrum, a measure
based on syllabic frequency
Multivariate Gaussian
classifier, neural network
(MLP)
Lu, Zhang,
and Jiang,
Audio content analysis
in video parsing
High zero-crossing rate ratio (HZCRR),
low short-time energy ratio (LSTER),
3-step classification:
1. KNN and linear spectral
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
56 Adrian Simion, tefan Truan-Matu
2.1. Digital Audio Signals
When music is recorded, the continuous pressure from the sound wave is
measured using a microphone. These measurements are taken at a regular time
and each measurement is quantized.
Fig. 1. Digital sound representation (time domain):
a. Music is a b. that is sampled c. and Quantized
continuous signal; .
Sound can be represented as a sum of sinusoids. A signal of N samples can be
written as:
2002 [12] linear spectral
pairs, band periodicity, noise-frame ratio
(NFR)
pairs-vector quantization
(LSP-VQ)for
speech/nonspeech
discrimination.
2. Heuristic rules for
nonspeech classification
into music/background
noise/silence.
3. Speaker segmentation
Ajmera et al.,
2003 [13]
Automatic transcription
of broadcast news
Averaged entropy measure and
dynamism estimated at the output of a
multilayer perceptron (MLP) trained to
emit posterior probabilities of phones.
MLP input: 13 first cepstra of a 12th-order
perceptual linear prediction filter.
2-state HMM with
minimum duration
constraints (threshold free,
unsupervised, no training).
Burred and
Lerch, 2004
[14]
Audio classification
(speech/
music/background
noise), music
classification into genres
Statistical measures of short-time frame
features: ZCR, spectral centroid/roll
off/flux,
first 5 MFCCs, audio spectrum
centroid/flatness, harmonic ratio, beat
strength, rhythmic regularity, RMS
energy, time envelope, low energy rate,
loudness
KNN classifier, 3-
component GMM classifier
Barbedo and
Lopes, 2006
[15]
Automatic segmentation
for real-time
applications
Features based on ZCR, spectral roll off,
loudness and fundamental frequencies
KNN, self-organizing
maps, MLP neural
networks, linear
combinations
Munoz- Exp
osito et al.,
2006 [16]
Intelligent audio coding
system
Warped LPC-based spectral centroid
3-component GMM, with
or without fuzzy rules-
based system
Alexandre et
al, 2006 [17]
Speech/music
classification for
musical genre
classification
Spectral centroid/roll off, ZCR, short-time
energy, low short time energy ratio
(LSTER), MFCC, voice to-white
Fisher linear discriminant,
K nearest neighbor
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Automatic Computer Music Classification and Segmentation 57
. )) ( 2 sin( )) ( 2 cos(
2 /
0
) ( ) (
N
k
i
k
r
k
N
k
a
N
k
a x
(1)
The signal can be represented in the frequency domain using the
coefficients )} , ( ),..., , {(
2 /
) (
2 /
) (
1
) (
1
i
N
y
N
i y
a a a a .
The magnitude and phase of the k
th
frequency component are given by:
2 ) ( 2 ) (
) ( ) ( ] [
i
k
r
k M
a a k X (2)
) arctan( ] [
) (
) (
r
k
i
k
p
a
a
k X (3)
Perceptual studies on human hearing show that the phase information is relatively
unimportant when compared to magnitude information, thus the phase component
during feature extraction is usually ignored. [19]
The Spectral Centroid is another spectral-shape feature that is useful in the
extraction and analysis process. We can see form Table 1 its various uses. The
Spectral Centroid is the center of gravity of the spectrum and is given by:
2 /
1
2 /
1
] [
* ] [
N
k
M
N
k
M
k X
k k X
C
(4)
The Spectral Centroid can be thought of as a measure of brightness since songs
are consider brighter when they have more high frequency components.
2.2. Time-Frequency Domain Transforms
In MIR and sound analysis in general it is common to do transformation between
the time and frequency domains. For this the mathematical apparatus gives us the
real discrete Fourier transform (DFT), the real short-time Fourier transform
(STFT), discrete cosine transform (DCT), discrete wavelet transform (DWT) and
also the gammatone transform (GT).
Music analysis is not concerned with complex transforms, since music is always a
real-valued time series and has only positive frequencies.
Given a signal x with N samples, the basis functions for the DFT will be N/2 sine
waves and N/2 cosine waves that correspond to the previous coefficients.
The projection operator is correlation, which is a measure of how similar two time
series are to one another. The coefficients are found by:
1
0
) (
) 2 cos( ] [
2
N
i
r
k
i
N
k
i x
N
a (5)
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
58 Adrian Simion, tefan Truan-Matu
1
0
) (
) 2 cos( ] [
2
N
i
i
k
i
N
k
i x
N
a
(6)
The DFT is computed in an efficient manner by the fast Fourier transform FFT.
One drawback of both the time series representation and the spectrum
representation is that neither simultaneously represents both time and frequency
information. A time-frequency representation is found using the short-time
Fourier transform (STFT): First, the audio signal is broken up into a series of
(overlapping) segments. Each segment is multiplied by a window function. The
length of the window is called the window size.
Fig. 2. Magnatune apa_ya-apa_ya-14-maani-59-88.wav (time domain).
Fig. 3. Magnatune apa_ya-apa_ya-14-maani-59-88.wav (spectrogram).
Fig 2 and 3 were obtained with a tweaked version of the MARSYASs tool
sound2png with the following commands:
./sound2png -m waveform ../audio/magnatune/0/apa_ya-apa_ya-14-maani-59-88.wav
../saveres/magnatunewav.png -ff Adventure.ttf
./sound2png -m spectogram ../audio/magnatune/0/apa_ya-apa_ya-14-maani-59-88.wav
../saveres/magnatunespec.png -ff Adventure.ttf
Another useful transformation is the wavelet transform.
2.3. Mel-Frequency Cepstral Coefficients (MFCC)
The most common set of features used in speech recognition and music annotation
systems are the Mel-Frequency Cepstral Coefficients (MFCC). MFCC are short-
time features that characterize the magnitude spectrum of an audio signal. For
each short-time (25 ms) segment, the feature vector is found using the five step
algorithm given in Algorithm 1. The first step is to obtain the magnitude of each
frequency component in the frequency domain using the DCT We then take the
log of the magnitude since perceptual loudness has been shown to be
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Automatic Computer Music Classification and Segmentation 59
approximately logarithmic. The frequency components are then merged into
40 bins that have been space according the Mel-scale.
The Mel-scale is mapping between true frequency and a model of perceived
frequency that is approximately logarithmic.
Since a time-series of these 40-dimensional Mel-frequency vectors will have
highly redundant, we could reduce dimension using PCA.
Instead, the speech community has adopted the discrete cosine transform (DCT),
which approximates PCA but does not require training data, to reduce the
dimensionality to a vector of 13 MFCCs. [20]
Algorithm 1. Calculating MFCC Feature Vector
1: Calculate the spectrum using the DFT
2: Take the log of the spectrum
3: Apply Mel-scaling and smoothing
4: Decorrelate using the DCT.
3. Problem description
A common feature that aids record producers to meet the demands of the target
audiences, musicologists to study musical influences and music enthusiasts to
summarize their collections is the musical genre identification.
The genre concept is inherently subjective because the influences, hierarchy or the
intersection of a song to a specific genre isnt universally agreed upon.
This point is backed up by a comparison of three Internet music providers that
found very big differences in the number of genres, the words that describe that
genre, and the structure of the genre hierarchies. [18]
Although there are some inconsistencies caused by its subjective nature, the genre
concept has shown interest from the MIR community.
The various papers and works on this subject reflect the authors assumptions
about the genres. Copyright laws prevented authors from establishing a common
database of songs, making it difficult to directly compare the results.
4. Experiments description
The datasets used for training and testing were MAGNATUNE [21] and two
collections that were built in the early stages of the MARSYAS [22] framework.
As the ADAMS system is built in a modular form the various tasks (described
below) can be automatized and the sound can flow through these modules until
the complete analysis is made.
The ADAMS main directory structure can be seen in the following picture:
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
60 Adrian Simion, tefan Truan-Matu
Fig. 4. ADAMS Main Directory Structure.
The machine learning tasks are done with the WEKA [23] tool, loading the
compatible arff files produced with the aid of MARSYAS.
The chosen OS for these experiments was Mandriva Linux 2011, the compiler
version being gcc (GCC) 4.6.1 20110627 (Mandriva).
Extractors that were used:
- BEAT: Beat histogram features
- LPCC: LPC derived Cepstral coefficients
- LSP: Linear Spectral Pairs
- MFCC: Mel-Frequency Cepstral Coefficients
- SCF: Spectral Crest Factor (MPEG-7)
- SFM: Spectral Flatness Measure (MPEG-7)
- SFMSCF: SCF and SFM features
- STFT: Centroid, Rolloff, Flux, ZeroCrossings
- STFTMFCC: Centroid, Rolloff Flux, ZeroCrossings, Mel-Frequency
Cepstral Coefficients
On every experiment for the specified extractors are also presented the confusion
matrices [24] in order to have an idea about the actual and the predicted
classifications done by the classification system.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Automatic Computer Music Classification and Segmentation 61
4.1. Experiment 1: Classification using Timbral Features
This experiment uses the following extractors: Time ZeroCrossings, Spectral
Centroid, Flux and Rolloff, and Mel-Frequency Cepstral Coefficients (MFCC).
We extract these features with the option timbral and we also create the file that
will be loaded with the WEKA environment for analysis with the following
command:
./adamsfeature -sv -timbral ../col/all.mf -w ../analysis/alltimbral.arff
Based on experiment the following classifiers were chosen: Bayes Network,
Naive Bayes, Decision Table, Filtered Classifier and NNGE.
The results are shown in the following table:
Table 2. Timbral Features - Classifier Results
Table 2 was build loading the file alltimbral.arff in WEKA and training the built-
in classifiers
Fig. 5. WEKA Prediction Errors Graph.
Classifier
Model
Build
Time(s)
Coorectly
Classified
Incorrectly
Classified
Mean
absolut
error
Root
mean
squared
error
Relative
absolute
error
Root
relative
squared
error
Bayes Network 1.78 62.5% 37.5% 0.0753 0.2648 41.82% 88.28%
Naive Bayes 0.04 55% 45% 0.0902 0.2925 50.09% 97.51%
Decision Table 15.49 51.6% 48.4% 0.1467 0.2599 81.53% 86.64%
Filtered Classifier 4.55 87.8% 12.2% 0.0348 0.1318 19.31% 43.94%
NNGE 10.69 100% 0% 0 0 0 0
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
62 Adrian Simion, tefan Truan-Matu
Fig. 6. Confusion Matrices for Timbral Features Classification
4.2. Experiment 2: Classification using Spectral Features
This experiment uses the following extractors: Spectral Centroid, Flux and Roll
off. The feature extraction was done with the following command:
./adamsfeature -sv -spfe ../col/all.mf -w ../analysis/allspectral.arff
Using the same classifiers the results are:
Table 3. Spectral Features - Classifier Results
Classifier
Model
Build
Time(s)
Correctly
Classified
Incorrectly
Classified
Mean
absolute
error
Root
mean
squared
error
Relative
absolute
error
Root
relative
squared
error
Bayes Network 1.78 46.5% 53.5% 0.1192 0.2742 66.21% 91.41%
Naive Bayes 0.23 42.5% 57.5% 0.1205 0.2924 66.92% 97.47%
Decision Table 0.72 46.1% 53.9% 0.1491 0.2655 82.82% 88.49%
Filtered Classifier 0.41 63.6% 36.4% 0.099 0.2225 54.98% 74.15%
NNGE 2.02 100% 0% 0 0 0 0
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Automatic Computer Music Classification and Segmentation 63
Fig. 7. Confusion Matrices for Spectral Features Classification
4.3 Experiment 2: Classification using MFCC
This experiment uses the Mel-Frequency Cepstral Coefficients extractors. The
feature extraction was done with the following command:
./adamsfeature -sv -mfcc ../col/all.mf -w ../analysis/allmfcc.arff
Table 4. MFCC Features - Classifier Results
Classifier
Model
Build
Time(s)
Correctly
Classified
Incorrectly
Classified
Mean
absolute
error
Root
mean
squared
error
Relative
absolute
error
Root
relative
squared
error
Bayes Network 1.23 63.3% 36.7% 0.0764 0.2475 42.42% 82.50%
Naive Bayes 0.22 58.5% 41.5% 0.0847 0.2694 47.07% 89.80%
Decision Table 6.4 49.1% 50.9% 0.1481 0.2638 82.27% 87.94%
Filtered Classifier 0.81 87.1% 12.9% 0.0363 0.1348 20.18% 44.92%
NNGE 3.74 99.8% 0.2% 0.0004 0.02 0.22% 6.66%
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
64 Adrian Simion, tefan Truan-Matu
Fig. 8. Confusion Matrices for MFCC Features Classification
4.4 Experiment 4: Classification using Zero Crossings
The feature extraction was done with the following command:
./adamsfeature -sv -zcrs ../col/all.mf -w ../analysis/allzcrs.arff
Table 5. Zero Crossings Features - Classifier Results
Classifier
Model
Build
Time(s)
Correctly
Classified
Incorrectly
Classified
Mean
absolute
error
Root
mean
squared
error
Relative
absolute
error
Root
relative
squared
error
Bayes Network 0.09 34.7% 65.3% 0.1437 0.2789 79.83% 92.97%
Naive Bayes 0.01 34.5% 65.5% 0.1441 0.2869 80.06% 95.63%
Decision Table 0.22 42.4% 57.6% 0.1511 0.2691 83.95% 89.71%
Filtered Classifier 0.15 44% 56% 0.1403 0.2649 77.94% 88.24%
NNGE 0.52 99.8% 0.2% 0.0004 0.02 0.22% 6.66%
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Automatic Computer Music Classification and Segmentation 65
Fig. 9. Confusion Matrices for Zero Crossings Features Classification.
4.5 Experiment 5: Classification using Spectral Flatness Measure
The feature extraction was done with the following command:
./adamsfeature -sv -sfm ../col/all.mf -w ../analysis/allsfm.arff
Table 6. SFM Features - Classifier Results
Classifier
Model
Build
Time(s)
Correctly
Classified
Incorrectly
Classified
Mean
absolute
error
Root
mean
squared
error
Relative
absolute
error
Root
relative
squared
error
Bayes Network 1.78 58.4% 41.6% 0.0838 0.2738 46.53% 91.28%
Naive Bayes 0.15 53.2% 46.8% 0.0935 0.294 51.96% 97.99%
Decision Table 12.35 50.4% 49.6% 0.1472 0.2621 81.78% 87.37%
Filtered Classifier 2.1 83.8% 16.2% 0.045 0.15 25.01% 50.12%
NNGE 9.24 99.8% 0.2% 0.0004 0.02 0.22% 6.66%
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
66 Adrian Simion, tefan Truan-Matu
Fig. 10. Confusion Matrices for Spectral Flatness Measure Features Classification.
Conclusions
Five experiments were conducted for determining the music genre of a specific
audio file. The extracted features varied in each experiment in order to determine
which one was more suited to the dataset used. The five classifiers provided
different results based on the extracted features and these were put to test with
well known machine learning tools and music analysis frameworks like WEKA
and MARSYAS, and also with an analysis system developed on top of the
MARSYAS framework.
The results show that satisfactory results can be obtained even from the simplistic
approaches as Nave Bayes classification, but better results were obtained using
more advanced techniques. The fact that the nearest neighbor produced very good
results doesnt mean that it will have the same behavior on another dataset.
Improvements on the presented methods can be obtained by testing these methods on
a broader dataset and determining the intrinsic influences of each genre on another.
The conclusions of these influences can have a more meaningful sense from the
social point of view like blues and its derivatives and we can find very unlikely
results like death metal having roots in jazz music.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Automatic Computer Music Classification and Segmentation 67
R E F E R E N C E S
[1] Howes, F. Man Mind and Music. Marin Secker & Warbug LTD., 1948.
[2] Ismir. http://www.ismir.net/ (Visited on 2012/01/23)
[3] Mirex. http://www.music-ir.org/mirex/wiki/MIREX_HOME (Visited on 2012/01/23)
[4] J. Saunders, Real-time discrimination of broadcast speech/music, in Proceedings of IEEE
International Conference on Acoustics, Speech, and Signal Processing (ICASSP 96), vol. 2, pp. 993
996, Atlanta, Ga, USA, May 1996.
[5] E. Scheirer and M. Slaney, Construction and evaluation of a robust multifeature speech/music
discriminator, in Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal
Processing (ICASSP 97), vol. 2, pp. 13311334, Munich, Germany, April 1997.
[6] J. T. Foote, A similarity measure for automatic audio classification, in Proceedings of the AAAI
Spring Symposium on Intelligent Integration and Use of Text, Image, Video, and Audio Corpora,
Stanford, Calif, USA, March 1997.
[7] Z. Liu, J. Huang, Y. Wang, and I. T. Chen, Audio feature extraction and analysis for scene
classification, in Proceedings of the 1st IEEE Workshop on Multimedia Signal Processing (MMSP
97), pp. 343348, Princeton, NJ, USA, June 1997.
[8] T. Zhang and C.-C. J. Kuo, Hierarchical classification of audio data for archiving and
retrieving, in Proceedings of IEEE International Conference on Acoustics, Speech, and Signal
Processing (ICASSP 99), vol. 6, pp. 30013004, Phoenix, Ariz, USA, March 1999.
[9] G. Williams and D. P. W. Ellis, Speech/music discrimination based on posterior probability
features, in Proceedings of the 6th European Conference on Speech Communication and Technology
(EUROSPEECH 99), pp. 687690, Budapest, Hungary, September 1999.
[10] K. El-Maleh, M. Klein, G. Petrucci, and P. Kabal, Speech/music discrimination for multimedia
applications, in Proceedings of IEEE International Conference on Acoustics, Speech, and Signal
Processing (ICASSP 00), vol. 6, pp. 24452448, Istanbul, Turkey, June 2000.
[11] A. Bugatti, A. Flammini, and P. Migliorati, Audio classification in speech and music: a
comparison between a statistical and a neural approach, EURASIP Journal on Applied Signal
Processing, vol. 2002, no. 4, pp. 372378, 2002.
[12] L. Lu, H.-J. Zhang, and H. Jiang, Content analysis for audio classification and segmentation,
IEEE Transactions on Speech and Audio Processing, vol. 10, no. 7, pp. 504516, 2002.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
68 Adrian Simion, tefan Truan-Matu
[13] J. Ajmera, I. McCowan, and H. Bourlard, Speech/music segmentation using entropy and
dynamism features in a HMM classification framework, Speech Communication, vol. 40, no. 3, pp.
351-363, 2003.
[14] J. J. Burred and A. Lerch, Hierarchical automatic audio signal classification, Journal of the
Audio Engineering Society, vol. 52, no. 7-8, pp. 724739, 2004.
[15] J. G. A. Barbedo and A. Lopes, A robust and computationally efficient speech/music
discriminator, Journal of the Audio Engineering Society, vol. 54, no. 7-8, pp. 571588, 2006.
[16] J. E. Munoz-Exp osito, S. G. Galan, N. R. Reyes, P. V. Candeas, and F. R. Pena, A fuzzy
rules-based speech/music discrimination approach for intelligent audio coding over the Internet, in
Proceedings of the 120th Audio Engineering Society Convention (AES 06), Paris, France, May 2006,
paper number 6676.
[17] E. Alexandre, M. Rosa, L. Caudra, and R. Gil-Pita, Application of Fisher linear discriminant
analysis to speech/music classification, in Proceedings of the 120th Audio Engineering Society
Convention (AES 06), Paris, France, May 2006, paper number 6678
[18] F. Pachet and D. Cazaly, A taxonomy of musical genres, RIAO 00: Content-Based Multimedia
Information Access, 2000.
[19] B. Logan, Mel-Frequency Cepstral Coefficients for music modeling, ISMIR 00: International
Symposium on Music Information Retrieval, 2000.
[20] D. Turnbull, Automatic music annotation, Department of Computer Science, UC San Diego,
2005.
[21] Mangatune. http://tagatune.org/Magnatagatune.html (Visited on 2012/01/23).
[22] MARSYAS. http://marsyas.info/ (Visited on 2012/01/23).
[23] WEKA. http://www.cs.waikato.ac.nz/ml/weka/ (Visited on 2012/01/23).
[24] http://www2.cs.uregina.ca/~hamilton/courses/831/notes/confusion_matrix/confusion_matrix.
html.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Annals of the Academy of Romanian Scientists
Series on Science and Technology of Information
ISSN 2066-8562 Volume 5, Number 1/2012 69
POLYMERIC PRESSURE SENSORS: A CONCEPTUAL VIEW
Cornel COBIANU
1*
and Bogdan SERBAN
1
Abstract. In the first part of this paper, we present a review of the piezoresistive pressure
sensors based on polymeric thick films deposited on rigid and flexible diaphragm. The
study of the state of the art has shown the performances of this technology, where
maximum sensitivity is obtained on thin flexible diaphragm for a gauge factor of about
10, in a pressure range of 05 kPa. The present challenges come from the high
temperature coefficient of the resistance of about 500 ppm/C, and the long temperature
drifts of about (0.52)%, which may require improved repeatability of fabrication
technology and advanced differential signal processing techniques for the market
acceptance. In the second part of the paper, we present our novel concepts for the
realization of the piezoresistive pressure sensors. The first concept consists in the surface
modification of the organic substrate by ion implantation of nitrogen and phosphorus
species for creating piezoresistive behavior and high electrical conductivity of organic
piezoresistors. The second concept consists in the novel chemical synthesis route of
organic thin film by doping the polyaniline with large molecules of p-sulfonated
calix[n]arene (n =4, 6, 8), sulfonated crown ethers, in the liquid state. Addition of the
metal nanoparticles to the previous homogeneous solution can further increase the
piezoresistive factor. Other new features of our second concept come from the direct
printing from solution of the above piezoresistive organic thin films, as well as metallic
films interconnecting the piezoresistors, and finally the monolithic fabrication of the
sensor rim and diaphragm by plastic injection molding, where the pressure diaphragm
could be as thin as 75 micrometers.
Keywords: piezoresistive organic films, polymeric thick film, pressure sensors, metal
Nanoparticples, p-sulfonated calix[n]arene, sulfonated crown ethers
1. Introduction
Pressure monitoring is an important parameter in the control of a large diversity of
industrial processes and medical applications. Pressure can be measured by
mechanical devices, as well as electro-mechanical and electro-optical instruments.
The measurement of the pressure of a fluid by pure mechanical principle is based
on the presence of an elastic diaphragm fixed at one end, which is moving its free
end as a result of pressure variation, and its position change is indicated by a
needle connected to the free end, and which is thus rotating with respect to its zero
position (Fig. 1).
This principle is used for the pressure measurement on gas/liquids pipelines,
where pressure manometers based on Bourdon tubes are still in place, today.
1
Honeywell Romania, Sensors and Wireless Laboratory, Bucharest, Romania.
*Member of Academy of Romanian Scientists, cornel.cobianu@honeywell.com.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
70 Cornel Cobianu, Bogdan Serban
In other cases, the elastic diaphragm is fixed at both ends, and this is deflecting in
the central part as a result of pressure change (Fig. 2). Such deflection is creating
strain in the diaphragm, and its value can be also used for the pressure
measurement, in the so called strain gauges. These deflecting diaphragms have
opened the way for the electro-mechanical principles of pressure measurement.
Fig. 1. Schematic picture of a Bourdon Fig. 2. Pressure diaphragm flexing
tube used for a pressure manometer. under applied external pressure.
Historically speaking, in 1856 Lord Kelvin has discovered that the electrical
conductor can change its electrical resistance, when it is strained due to an
external applied force. Thus, he can be credited with the discovery of the
piezoresistive effect in metals, which was then used in metallic strain gauge
devices for multiple applications (strain, torque, force) including pressure sensing.
Later, the piezoresistive effect was defined as a change in the electrical resistivity
of a material as a function of the externally applied stress on it. Now, it is
generally agreed that the discovery of piezoresistive effect is at the origin of the
most of electric sensors for mechanical measurands defined as devices able to
convert a non-electrical signal (like fluid) pressure into an electrical signal (like
voltage). The silicon technology followed by the micro-electro-mechanical system
(MEMS) technology have used all these principles for the miniaturization of the
existing macroscopic devices and sensors.
The impetuous development of the MEMS technology was founded on two
important technical pillars, consisting of well-established integrated circuit (IC)
infrastructure and excellent operation of the macroscopic principles at the
micrometer scale. A convincing demonstration of this successful approach is
coming from microsystems for pressure measurement, where the well-known
principle of macroscopic diaphragm movement as a function of pressure has been
transferred to microtechnology scale with very good results.
The era of miniaturized pressure sensors has been triggered by the discovery of
the piezoresistive effect in silicon and germanium, in 1954 by Charles Smith [1].
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Polymeric Pressure Sensors:
A Conceptual View 71
The time interval from piezoresistance discovery to the first associated
commercial product was short, as in 1959, Kulite was already delivering the first
silicon strain-gauges bonded on a metal diaphragm. During sensor operation, the
metal diaphragm is elastically deflecting due to the applied pressure, and a tensile
stress is developed in the central region of the diaphragm, while a compressive
stress is developed at the periphery of the diaphragm (Fig. 2.).
Such a stress is transmitted to the strain gauge, which is feeling it by the well-
known stress-strain correlation, and a change in the resistance is obtained due to
the piezoresistive effect, which is thus an indication of the pressure to be
measured. In the presence of an external pressure, a piezoresistor located on the
central part of the diaphragm from Fig. 2. is exposed to tensile stress and its
resistance value is increasing, while a piezoresistor located at the periphery of the
diaphragm, near the edge, is exposed to compressive stress and its resistance value
is decreasing with respect to the value specific to zero stress. For an accurate
pressure measurement, where the ageing effects in the piezoresistors, as well as
external temperature variations to be compensated, four piezoresistors are located
in the arms of a Wheatstone bridge, and thus obtaining a differential sensing
configuration for the signal conditioning.
Using such an approach, at the beginning of 1970s, IBM has proven the
operation of the first piezoresistive pressure sensor with silicon diaphragm, while
the first commercial all-silicon pressure sensor was delivered in 1974 [2].
The piezoresistive effect was measured in doped silicon resistors, which were
located in well-defined regions of the diaphragm, as described above. The
excellent elastic properties of silicon diaphragm, combined with the fact that the
silicon piezoresistor is obtained intrinsically in the diaphragm, without any need
of strain gauge bonding to silicon diaphragm can explain the excellent
performance of silicon MEMS pressure sensors. Thus, the MEMS silicon
technology has proven its capability to generate commercial products, where thin
silicon diaphragms have been used from very beginning in medical application for
measuring blood pressure.
In parallel with silicon MEMS technology, which has brought to the market the
first miniaturized microsystems, sensors and actuators, like pressure sensors,
accelerometers, inkjet nozzles for thermal inkjet technologies and thus predicting
its long-term innovation capabilities, other sensing technologies were emerging,
which were targeting mechanical sensing applications (pressure torque, force)
based on other than silicon materials and substrates. It is the case of thick film
technologies which have found their niche sensing applications.
In this paper, we shall briefly present polymeric film technologies for pressure
sensing applications. At the beginning of the study, the state of the art for the
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
72 Cornel Cobianu, Bogdan Serban
piezoresistive polymeric thick films will be shown, where such materials are
deposited on diaphragms made of glass, alumina, or even flexible substrates, and
their pressure sensing properties are described [3-6].
Finally, our concepts for new polymeric pressure sensors, including all-organic
technologies will be shown. Here, firstly, we present a novel pressure sensing
concept, where the plastic substrate will receive electrical conductivity and
piezoresistivity properties on well-defined regions of the organic diaphragm by
using IC specific technology like ion implantation [7]. Then, another novel
concept for all-plastic pressure sensor is described where chemical synthesis for
the preparation of new organic thin films with enhanced electrical conductivity is
shown. This piezoresistive pressure sensor is made by additive, maskless direct
printing of the organic films in well-defined positions of plastic diaphragm, which
is obtained by injection molding [8].
2. Piezoresistive pressure sensors based on thick film resistors
The operation principle of the piezoresistive mechanical sensors based on thick
films is similar to that described for silicon sensors, but the diaphragm and the
piezoresistor are made of different materials. For the evaluation of different
piezoresistive materials, the gauge factor is used, and this is equal to the ratio
between the relative variation of the resistance (R/R) and the relative variation of
the resistor length (or the strain (/)). The piezoresistive behavior in thick
resistive layers was systematically investigated by using commercial ruthenate
thick films from Dupont [9]. The thick films were deposited on alumina substrate
as a gel of high viscosity by screen printing technology, and then fired at high
temperatures, around 850-950
o
C, for obtaining thick solid layers. The gauge
factor for these ruthenate thick films was in the range of 11-14, being weakly
dependent on the strain direction.
Such piezoresistive thick films have a gauge factor which is higher than that of the
metal strain gauge (1.8-4.5), and much smaller than that of semiconductors (40-
200) [10]. The temperature coefficient of these thick ruthenate film piezoresistor
(TCR) was about 100 ppm/
o
C, while the temperature coefficient of the gauge
factor (TCGF) was smaller than 500 ppm/
o
C.
These values can be compared with the similar ones of the metal wire strain
gauges, where the TCR has a large range of variation (20-4000) ppm/
o
C, and
TCGF is in the range 20-100 ppm/
o
C [9]. Unfortunately, such high temperature
technology is restricting the type of substrate to be used to only ceramics, and is
increasing the cost of processing.
As a low temperature alternative to the well-established ruthenate thick film
piezoresistive technology presented above, the pressure sensor based on
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Polymeric Pressure Sensors:
A Conceptual View 73
polymeric thick film resistors was proposed at the beginning of the 90 exploiting
the piezoresistive effect in such organic layers, which can tolerate processing
temperatures not higher than 150-300
o
C [ 3].
The thick film composition consists of an organic polymer matrix, like a
polyimide, which is loaded with carbon for obtaining the resistive and
piezoresistive behavior. The sensing diaphragm is made of different rigid
materials like, alumina, glass-reinforced epoxy laminate (FR 4), or even a flexible
substrate [4].
For the realization of the planar piezoresistor, initially, the metallic electrodes
were deposited and patterned on the substrate. Then, the deposition of the
polymeric thick films on the rigid substrate, like alumina, glass, or FR 4 is made
by screen printing technology, followed by thermal consolidation process at low
temperatures, below 300
o
C. These planar polymeric thick films based polymeric
piezoresistors have a gauge factor of about 10, which is rather similar to the value
obtained for ruthenate thick films, but the TCR is equal to +/ 500 ppm/
o
C, which
is much higher with respect to the value obtained for ruthenate material
(100 ppm/
o
C).
In the case of a sandwich piezoresistor configuration, when the polymeric thick
film is sandwiched (on the z direction) between two electrodes, with the first
electrode deposited on the substrate and the second electrode deposited on the
polymeric thick film, the gauge factor has reached a much higher value of about
80 [5]. In addition, these sandwich polymeric piezoresistive pressure sensors have
shown a very high value of the TCR, of about 2200 ppm/
o
C, which may be
more difficult to compensate, even in Wheatstone configuration, due to existing
mismatches between the TCR of different piezoresistors.
Also, poorer reproducibility of sandwich sensor with respect to planar ones was
obtained.
When such polymeric thick films were deposited on flexible substrates, having
much lower thickness with respect to the rigid one, the sensitivity of the planar
polymeric thick film pressure sensors has increased of about 6 times, but, in our
opinion, this result should be correlated only with the pressure diaphragm
properties (thickness and Young module) and not to sensing layer, itself. Such
thick films polymeric planar pressure sensors on rigid substrate are working in the
temperature range from 0 to 75
o
C, and their linear response was obtained for a
pressure range from 0 to 10
6
Pa.
Typical linearity plots of the planar polymeric thick film deposited on alumina
substrates have shown a non-linearity of about 3% in the range [0; +500]
microstrains (1 microstrain means a dilation/shrinkage of 1 micrometer of
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
74 Cornel Cobianu, Bogdan Serban
material having a length of 1 meter), while the sandwiched piezoresistor have had
a non-linearity of 14% in the same strain range [5]. One of the biggest drawbacks
of such polymeric pressure sensors is the long-term drift, equal to about 0.5-2 %,
and which was estimated by the sensor output drift after an accelerated ageing test
at 1000
o
C and 85
o
C.
This behavior should be correlated to the specific ageing mechanisms of the
organic materials. The challenging temperature behavior of these polymeric
pressure sensors followed by their relative high drifts due to material ageing can
be partially solved by design of the diaphragm (thickness and Young modulus of
organic substrate) as shown in [3,4] and by means of Wheatstone bridge-based
differential signal conditioning.
Therefore, much work should be devoted here, but such polymeric sensors are the
best candidate for low cost disposable pressure sensors for applications where the
accuracy is not critical.
3. Piezoresistive pressure sensors based on surface modified polymeric
diaphragm
Flexible electronics (flex circuits), where the traditional silicon integrated
circuits are placed by surface mounting technologies on a flexible plastic substrate
is rapidly advancing in many market applications, from portable video camera to
solar cell [11].
For these flex circuits, the metallic interconnections between different integrated
circuits are made by standard photolithographic techniques. In parallel, an all-
organic electronics is emerging where the semiconductor devices like organic
light emitting diodes (OLED) and organic field emission transistor (OFET) are
manufactured in the body of the organic semiconductor materials [12]. In this
context, the sensing domain is going to add new capabilities to the rigid and
flexible organic electronics. On this idea, recently, we have proposed a concept
for pressure sensors where a polymeric pressure diaphragm is surface modified in
order to obtain selectively piezoresistive effect [7].
The novelty of our approach comes from the technology of the piezoresistor
realization, where we have applied ion implantation technique for the local
realization of the piezoresistivity and electrical conduction enhancement, as it will
be described below. In Fig. 3, we show a cross section view of an-all plastic
piezoresistive pressure sensor, before packaging, where the substrate and the
diaphragm come from the same starting rigid plastic material, which could be a
polyimide, like Kapton from Dupont or liquid crystal polymer. Alternatively,
polystyrene-co-acrylonitrile (SAN) 80/20 could be used. Such organic polymers
are dielectric materials, which have a very high electrical resistivity, and therefore
they cannot be used as they are for reaching the intended function.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Polymeric Pressure Sensors:
A Conceptual View 75
Fig. 3. Schematic of a pressure sensor with
diaphragm obtained by etching back side of the
substrate and piezoresistor fabricated by ion
implantation.
However, it was already shown in the literature that the ion implantation can
dramatically increase the electrical conduction of these plastic materials [13].
Therefore, we have proposed to perform a high dose implantation of the nitrogen
species, in well-defined regions of the diaphragm for the generation of the
piezoresistive regions. This process was followed by a phosphorus ion
implantation in the same region for further enhancement of the electrical
conductivity of the piezoresistive regions.
For generating such electrical and piezoelectrical properties selectively, in the
organic diaphragm, standard IC photo-lithographical processes can be used. This is
possible due to the chemical resistance of these rigid plastic materials to the
solvents and the other solutions used to remove the photoresist, at the end of the ion
implantation processes. Subsequently, the electrical contacts to the piezoresistors
are done by IC technology processes like electron-gun physical vapor deposition,
where, for example, a thin film combination like chromium/gold can be used.
Chromium is assuring the adherence of the gold layer to the plastic substrate and
ion-implanted regions of the piezoresistors.
For the realization of the diaphragm in the starting plastic material, the back-side
etching of the substrate is performed by plasma techniques, like reactive ion
etching (RIE). The realization of such pressure diaphragms from Fig. 3. is
possible by metal masking of the entire front side, and of selective regions of back
side regions of the substrate, which should survive after deep etching and
subsequently form what is called the rim of the pressure sensor. The thickness of
the pressure diaphragm is determined by the pressure range needed to be
measured and the important requirement that the deflection of the pressure
diaphragm under external pressure to be in the elasticity domain of that material.
Such design conditions are preserving the linearity of the sensor response and also
minimizing the hysteresis and long-term drifts of the sensors.
Fig. 4. Pressure with plastic diaphragm bonded
by adhesive to glass.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
76 Cornel Cobianu, Bogdan Serban
In other applications, it may be useful to use a glass substrate and a plastic diaphragm
which can be bonded together for defining the pressure sensor. Such an approach is
also possible within the above concept, where initially a plastic foil (similar to the
silicon wafer from the point of view of batch processing) of the thickness required by
the pressure sensing application is processed, as presented above, but, in this case
there is no need for the back side etching, as the entire thickness will play the role of
the diaphragm. In this case, which is shown in Fig. 4., micromachining of the glass
substrate is needed for the realization of the rim of the sensor, which is also allowing
the access of the fluid pressure (air, liquid) to the pressure diaphragm. Glass MEMS
is a well-established batch technology for sensor applications, and also for sensor
packaging, in general.
In this case, the key process is glass drilling at the wafer level for the sensor rim
realization, and this can be done by either wet etching in HF based solutions, or
by RIE or even laser drilling. The signal conditioning techniques are standard
consisting in Wheatstone bridge, which will be described in more details, in the
next section.
4. Low cost all-plastic piezoresistive pressure sensors
In different industrial domains, there is a strong demand for disposable, low cost
pressure sensors. In such cases, the silicon technology may still be expensive,
considering the cost of clean room processes, and of the monocrystalline silicon
substrate itself.
Fig. 5a. Top view of the low cost pressure sensor
made by plasting molding and direct printing of organic piezoresistive layer.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Polymeric Pressure Sensors:
A Conceptual View 77
Fig. 5b. Cross section view through 2A-2A axis.
For such applications, where even the photolithographical processing should be
avoided, we have developed a concept for a simple technology based on plastic
injection molding of the sensor rim and pressure diaphragm in conjunction with
mask-less, direct printing methods for the deposition of the metal, and novel
organic piezoresistive layers [8].
In Fig. 5. a), we show a top view of the pressure sensor based on four
piezoresistors, R
1
-R
4
, all of them being located in well-defined positions of the
molded plastic diaphragm, while in Fig. 5. b) we show a cross section view
through the central region of the pressure diaphragm.
These piezoresistors are electrically interconnected in a Wheatstone bridge, as
shown in Fig. 6. In the absence of an external pressure, all four piezoresistors
have the same value of the resistance, and the output voltage of the Wheatstone
bridge is equal to zero.
In order to obtain this zero voltage, two potentiometers are also connected to the
Wheatstone bridge, as in Fig. 5. a) and Fig. 6.
In the presence of an external pressure, the piezoresistors R
1
and R
3
, which are
positioned in the central region of the diaphragm, are exposed to tensile stress,
while R
2
and R
4
, which are located at the periphery of the diaphragm are exposed
to the compressive stress.
The variation of the piezoresistance as a function of pressure can be written as
follows:
R
1
= R
o
(1 + x)
R
2
= R
o
(1 x)
R
3
= R
o
(1 + x)
R
4
= R
o
(1 x)
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
78 Cornel Cobianu, Bogdan Serban
where, R
o
is the resistance value of the piezoresistor at the reference pressure, while
x = G*, where G is the gauge factor of piezoresistance, and is relative deformation
of the length of the piezoresistor, =/ [10].
Fig. 6. Wheatstone bridge with four piezoresistors for maximum pressure sensitivity.
As mentioned above, the novel aspects of the pressure sensor realization come
from the new organic piezoresistive materials, the maskless method for film
resistor deposition, and as well as the cheap technology proposed for the plastic
diaphragm realization.
For the realization of the piezoresistive films, we are considering more chemical
synthesis routes, which are aiming an increased electrical conductivity and
piezoresistivity of the organic layer. In the case of an all-organic piezoresistive
layer, for example, we suggest starting with polyaniline and doping it with large
molecules like p-sulfonato-calix[n]arenes (n = 4, 6, 8), p-sulfonated-
calix[n]arenes (n = 4, 6, 8), tosylates, carboxylic acids of calix[n]arenes
(n = 4, 6, 8), sulfonated crown ethers, sulfonated cyclodextrines, carboxylic acid
nanotubes, or carboxylic acid of fullerenes.
All these compounds can be dissolved in water and other solvents. They can
generate -stacking interactions with polyanilines and thus contributing to the
increase of electrical conductivity of the polymeric film.
In Fig. 7. we show the chemical formulae of the p-sulfonated-calix[n]arene for
n = 4, 6, 8, while in Fig. 8., we show the reaction for doping the aniline by
p-sulfonated calix[4]arene. The synthesis of a soluble conducting polymer can be
as follows. One can start with aniline substituted with an o-methoxy group and an
o-ethoxy group in equimolar amounts which can be polymerized by combining
with hydrogen peroxide, in aqueous solution. The dopant can be p-sulfonated
calix[n]arenes (n = 4, 6, 8), tosylates, carboxylic acids of calix[n]arenes (n =
4, 6, 8), or sulfonated crown ethers [14].
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Polymeric Pressure Sensors:
A Conceptual View 79
Fig. 7. Chemical formula of p-sulfonated calix[n]arene, where n = 4, 6, 8.
Fig. 8. Synthesis of doped polyaniline; HA stands for p-sulfonated calix [4]arene.
as described above. If the electrical and piezoresistive behavior of the all-organic
polymer film should be further increased, then, metal nanoparticles may be added
to the above polymeric sol, creating thus a heterogeneous inorganic-organic
mixture in the liquid state, which will be further used for the direct printing
method to be described below. Thus, an efficient increase of the gauge factor of
the piezoresistors is expected by the addition of metal nanoparticles to the liquid
phase of the initial pure organic solution.
Fig. 9. Schematics of the direct printing method for the maskless preparation of solid films.
6
CH
2
SO
3
H
OH
CH
2
SO
3
H
OH
4
CH
2
SO
3
H
OH
8
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
80 Cornel Cobianu, Bogdan Serban
In Fig. 9., we show a schematic of direct printing apparatus. More data about this
additive deposition method can be found elsewhere [15]. Specific to this direct
printing method is the formation of the liquid precursor of the future solid film, by
any chemical syntheses routes, which is then deposited in the right location of the
substrate by a moving nozzle, while its travel above the substrate is computer-
controlled. If we go into more details, as can be seen in Fig. 8., a gas flow is used
to carry the liquid phase of the atomized deposition material to a nozzle, which
is printing the liquid state on the substrate for the realization the pattern of the
future solid film. After printing, the gelly layers are dried and thermally
consolidated at the temperatures allowed by the organic materials, so that their
chemical properties to be preserved. The advantage of this deposition method is
that the solid film is printed from the very beginning in the right position and
pattern, and there is no need for additional photolithographic and etching process
for the layer delineation. Under such conditions, there is no loss of material and
many technological steps are eliminated.
By this additive technology we can deposit not only the polymeric films as
described above, but also the metallic films used to interconnect electrically the
piezoresistors from Figs. 5 and 6. For these metallic conductors, one can use
organic conductors, or silver based pastes, which have a low resistivity and do not
introduce parasitic resistances to the Wheatstone bridge.
Finally, a novel aspect of our concept is the monolithic realization of the sensor
rim and pressure diaphragm in a single process, from the same material by means
of plastic injection molding. This is possible by the progress in the field of this
molding process, where pressure diaphragms as thin as 75 m are possible to be
obtained. As plastic materials one can mention here, polycarbonates, polyesters
such as PET or nylon, or PVC.
As mentioned from very beginning, one of the most important drivers for such all-
plastic piezoresistive pressure sensors, based on injection molding and direct
printing of the organic conductive and piezoresistive solid films is the high
potential for a low cost fabrication process and associated materials. Such
concepts show the potential of the organic semiconductors to open the way
towards a new family of applications like flexible, intelligent microsystems,
where both the electronic circuit signal processing and the sensing devices to be
performed on the organic substrate.
5. Conclusions
In this paper we have reviewed the key materials and processes for the realization
of the polymeric thick films piezoresistive, and we introduced our concepts for the
preparation of novel organic piezoresistive thin films to be used in the next
generation of piezoresistive pressure sensor on rigid and flexible substrates.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Polymeric Pressure Sensors:
A Conceptual View 81
The polymeric pressure sensor domain was developed in the last two decades
starting from the useful results obtained in the years 1980s on piezoresistive
ruthenate thick films. Such polymeric thick films consisted of carbon loaded
polyimide films have shown a gauge factor of about 10, and they have been used
for pressure sensor operating in the range of about 0-10
6
Pa on either rigid or
flexible substrate.
The relatively high temperature coefficient of resistance (+/- 500 ppm/
o
C) and
long-term drift of these organic polymeric piezoresistive films (0.5-2)% used for
either planar and sandwich device configuration are partially solved by
Wheatstone bridge-based signal conditioning.
More work should be further devoted to the repeatability of fabrication
technology so that the organic piezoresistors connected in the Wheatstone bridge
to have similar temperature coefficient of resistance and gauge factor, and thus
minimize the long- term drift of sensor output.
We have introduced two novel approaches for the preparation of the thin film
polymeric piezoresistive pressure sensors. First concept consisted in the selective
surface modification of the organic substrate by ion implantation of nitrogen
species for inducing local piezorestivity, followed by phosphorus or boron
implantion for enhancing the electrical conductivity or the organic piezoresistors.
The novelty of the second concept consisted in the original chemical synthesis of
piezoresistive organic material, deposition method by direct printing, and
monolithic fabrication of plastic diaphragm by injection molding.
The new organic synthesis consist in doping of polyaniline as free base(
emeraldine) with large organic molecules (p-sulfonated calix[n]arenes p-
sulfonated calix[n]arenes, tosylates, carboxylic acids of calix[n]arenes, sulfonated
crown ethers, sulfonated cyclodextrines, carboxylic acid nanotubes, or carboxylic
acid of fullerenes. The bulk counterions of dopants improve the conductivity of
polyanilines
Acknowledgement
The authors would like to expresses their thanks to Honeywell International for
their support to work on this topic and write this paper.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
82 Cornel Cobianu, Bogdan Serban
R E F E R E N C E S
[1] Ch.Smith, Piezoresistive effect in germanium and silicon, Physical Review, 1954, vol. 94,
pp. 42-49.
[2] Commercial devices have been available for some time. See, for example, the National
Semiconductor Catalog, Transducers; Pressure and Temperature for August 1974.
[3] G. Harsanyi, Polymer Thick Films Technology: a Possibility to Obtain Very Low Cost
Pressure Sensors?, Sensors and Actuators A, 25-25 (1991), pp. 853-857.
[4] C. Csazar, G.r Harsanyi and R. P. Agarwal, A very low cost pressure sensor with extremely
high sensitivity, Sensors and Actuators A, vol. 41-42 (1994), pp. 417-420.
[5] K.I. Arshak, A.K. Ray, C.A. Hogarth, D.G.Collins, F. Ansari, An Analysis of polymeric thick
films as pressure sensors, Sensors and Actuators, A 4 vol. 49, (1995) pp. 41-45.
[6] N.J. Hendreson, N.M. White, T.V. Papakostas, P.H. Hartel, Low-Cost planar PTF Sensors
for Identity Verification of Smartcard Holders , Invited paper 2002 IEEE, Flexible Sensors in
Smart Applications session.
[7] C. Cobianu, M. Gologanu, I. Pavelescu, B. Serban, Micro-machined Pressure Sensor with
Polymer Diaphragm, US Patent 7401525 B2, Date of Patent July 22, 2008.
[8] C. Cobianu, S. R. Shiffer, B. Serban, A.D. Bradley, M. N. Mihaila, Pressure Sensor,
US Patent 7 318351 B2, Date of Patent, January 15, 2008.
[9] C. Canali, D. Malavasi, B. Moretn, M. Prudenziati, A. Taroni, Strain Sensitivity in Thick Film
Resistors, IEEE Transactions on Components, Hybrids, and Manufacturing Technology, vol.
CHMT-3, No. 3, September 1980, pp. 421-423.
[10] R. P. Arreny and J. G. Webster, Sensors and Signal Conditioning, Second Edition, John
Wiley and Sons, 2001, ISBN 0-471-33232-1.
[11] D. Shavit: The developments of LEDs and SMD Electronics on Transparent Conductive
Polyester Film, Vacuum International, 1/2007, S. 35 ff.
[12] H. Sirringhaus, C.W. Sele, T. von Werne, C. Ramsdale, 2007 Manufacturing of organic
Transistor Circuits by Solution-based Printing in G. Hadziioannou, G.G. Malliaras,
Semiconducting Polymers: Chemistry, Physics and Engineering, vol. 2 (2
nd
Edition),Wiley-
VCH. pp. 667694. ISBN 978352731271913]
[13] R. E. Giedd, M.G. Moss, M.M. Craig, and D.E. Robertson, Temperature sensitive ion-
implanted polymer films," Nuclear Instruments and Methods in Physics Research [Netherlands],
vol. B59/60, 1991, pp. 1253-1256.
[14] B. Serban, M. Bercu, S. Voicu, M. Mihail, Gh. Nechifor, C. Cobianu Calixarene doped
polyaniline for applications in sensing, International Semiconductor Conference, CAS 2006,
Proceedings, pp. 257-260.
[15] M. Hedges, Aerosol Jet Technology for Printed and Organic Electronic Devices,
Proceedings of LOPE Conference, 2009.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Annals of the Academy of Romanian Scientists
Series on Science and Technology of Information
ISSN 2066-8562 Volume 5, Number 1/2012 83
AN ANALYSIS REGARDING THE DECREASING OF THE
IMAGE QUALITY WITH THE OPTICAL MISALIGNMENT
CATALIN SPULBER
1
, OCTAVIA BORCAN
2
Rezumat. Performantele de informaie vizuala achiziionate de o camera termala sunt
determinate de doi parametri distinci, NETD (Noise equivalent temperature difference)
si MTF (Modulation Transfer Function); valoarea fiecruia dintre aceti parametri este
dependent de caracteristicile si de metodologia de msurare a componentelor de baz
ale unei camere termale: obiectivul i matricea de detecie. Autorii analizeaz unele
probleme legate de msurarea NETD i a MTF n cazul n care variaz distana focal a
obiectivului i apare o dezaliniere optic la montajul camerei termale. Experimentele
realizate demonstreaz c o distan focal mai mare asigur un MTF mai bun, iar
evaluarea NETD cu instrumentar electronic este mai adecvat.
Abstract. The performances of visual information acquired with a thermal camera are
determined by two distinct parameters, NETD (Noise equivalent temperature difference)
and MTF (Modulation Transfer Function) values of each of these parameters is
dependent on the characteristics and the methodology for measuring the basic
components of thermal camera: the lens and the starring detector. The authors analyze
some problems related to measurement NETD and MTF where the focal lens is variable
and optical misalignment occurs when mounting thermal camera. Experiments
demonstrate that a longer focal distance provides a better MTF and NETD evaluation
with electronic instruments is most appropriate.
Keywords: Thermal camera, NETD, MTF, optical misalignment, lens focal
1. Introduction
It is known that, in the field of actual research of vision using thermal cameras,
the most frequent question refers to the performance of the distance observation,
as is shown in the paper [1, 2].
This performance is described by some main parameters, as Noise Equivalent
Temperature Difference (NETD), Modulation Transfer Function (MTF), and
Minimum Resolvable Temperature Difference (MRTD) [3].A thermal camera is
commonly composed of an optical assembly, a FPA detector module (starring
detector), and signal processing electronics.
The imaging process can be described as a functional flow diagram (fig. 1.) from
scene (input information) to observer (output information).
1
Pro Optica Service&Components, e-mail: catalin.spulber@yahoo.com.,
Academy of Romanian
Scientists, www.aosromania.ro.
2
Pro Optica Service&Components, Romania, e-mail: borcan_octavia@yahoo.com.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
84 Catalin Spulber, Octavia Borcan
Fig. 1. The flow information between thermal camera components.
The lens gathers the scene radiance onto the detector array (FPA). The detector
module converts the radiation (photons) into an electrical signal, which enters a signal
processing unit. As it already was mentioned by authors in a previous article [2], the
environment (the atmosphere with aerosols or thermal perturbations) and the main
components of camera can deteriorates the final image.
An object is visible in an image because it has a different brightness than its
surroundings (target contrast). The input to any thermal camera system is photon
number N from the scene. The contrast of the object (i.e., the signal) must overcome
the image noise. Noise increases with the signal level results when the image has
been represented by a small number of individual particles. The signal-noise ratio
(SNR) is defined as the contrast divided by the standard deviation of the noise [4].
The mathematics governing these variations is called counting statistics or Poisson
statistics. That is, if there are N particles in each pixel on display, the mean is equal to
N and the standard deviation is equal to N . This makes the signal-to-noise ratio
equal to N . Detection is a noisy process. The noise V is composed of shot noise,
spatial noise, and excess noise. There are additional noise sources, including readout
amplifier noise and digital quantization noise. This is modelled as follows [4]:
noise read
V V V + = (1)
The noise equivalent temperature difference (NETD) is a widely used performance
parameter that characterizes the sensitivity of thermal imaging sensors.
The Modulation Transfer Function (MTF) is a quantitative measure of thermal
camera capability to transfer contrasts from the object plan in the image plan
depending on the aims of the lens and detector details. MTF is used to approximate
the position of best focus.
2. The problematic
Two problems are import ants if there is misalignment between the optical lens
and FPA:
a) Although NETD has been used for many years, there has always been some
confusion and misunderstanding about how to measure it. Differences in opinion
on how this measurement should be made can cause substantial variations in
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
An Analysis Regarding the Decreasing of the Image Quality
with the Optical Misalignment
reported values of NETD measurements [5]; for example, subjective digital noise
introduced by the display. The NETD will then calculated from the experimental
data as follows [6,7]:
N / S
T
NETD
A
= (2)
where
background et arg t
T T T = A , and the SNR is the signal-to-noise ratio of the
thermal camera.
b) The projection of the scene on the FPA detector is not perfect; since the
optical elements create blur [8]. For a quantitative analysis it used a test thermal
pattern as in figure 2., and the relation [3]:
eye display FPA optics
MTF MTF MTF MTF MTF
= (3)
Fig. 2. Thermal pattern used in the thermal camera performance analysis.
The MTF value varies with the lens focal length and their aperture or F number
(F#). The calculus of MTF values is based on the followings relations:
|
|
.
|
\
| t
=
|
|
.
|
\
| t
|
|
.
|
\
| t
=
cutoff
cutoff
cutoff
FPA
f
f
sinc
f
f
f
f
sin
MTF
(4)
where the spatial frequency f [cycles/mm], the cut-off frequency c
ut-off
[cycles/mm],
the horizontal and vertical size of detector (dH and dV, respectively) is exprimed as:
]
mrad
lp
[
dasV
1
f ];
mrad
lp
[
dasH
1
f
V cutoff H cutoff
= =
;
] mrad [
f
dV
dasV ]; mrad [
f
dH
dasH
ob ob
= =
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
86 Catalin Spulber, Octavia Borcan
|
.
|
\
|
t =
H
CCH
H H display
f
fl
d
FF sinc ) f ( MTF
(5) where FF
H
and FF
V
fill factors for
detector on the horizontal, respectively vertical direction.
MTF
optics
and MTF
FPA
have the configuration in figure 3.
Fig. 3. MTF Diagram for the MTF realized with the Maviis 1.5 software (JCD Publishing).
where
4 ) (
) (
4 1
4 1
+
=
V V
V V
MTF
t
(6)
On the other hand, it is known that [3, 9]:
|
(
t
o
|
|
.
|
\
|
A
=
2 / 1
frame
2 / 1
2 / 1
d
T
2 / 1
R
) t t ( MTF
f
) f (
NETD
3
MRTD (7)
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
An Analysis Regarding the Decreasing of the Image Quality
with the Optical Misalignment
or [7]:
f
f
f
K NETD
MRTD
MTF
= (8)
where: Af
R
-electronic frequency bandwidth, o- the field of view of the FPA
starring, f-the spatial frequency of the target being observed, t
d
[s] response time
of the detector, t- the integration time of the observer eye, t
frame
- the frame time.
So, it can write:
) f ( MRTD
NETD
const ) f ( MTF
1
o
= ;
o
=
2
const
NETD ;
f f
1
ob
= o (9)
For comparison of two MRTD as the same spatial frequency, in which one of the
thermal camera has a certain optical misalignment, it can by written as:
mis
mis
) f ( MRTD
) f ( MRTD
NETD
NETD
) f ( MTF
) f ( MTF
mis
= (10)
3. Experiments and results
For the experimental determination a thermal camera with the following technical
characteristics was used: the working spectral range 812 m, detection matrix
with micro-bolometers and resolution of 640 x 480 detection items, the size of the
detection item being 0.17 m, the focal distance for the objective being of 45 and
135 mm and f number (F#) between f/1.1 (for 45 mm lens focal) and f/1.6 (for
135 mm lens focal). In laboratory conditions, in order to determine the MRTD
function thermal patterns with 4 cycles/target, the ambient temperature was
24.7
0
C, and the black body temperature was 29.7
0
C (fig.4-6).
Fig. 4. Example of image
acquired with thermal contrast
AT = 10 C
Fig. 5. Example of image
acquired with thermal contrast
AT=5 C
Fig. 6. Example of image
acquired with thermal contrast
AT=2 C
The procedure used to evaluate NETD was based on standard ASTM E 1543-00 [6].
The results are presented in figures 7-19.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
88 Catalin Spulber, Octavia Borcan
Fig. 7. Visual images signal and noise on the oscilloscope.
0 1 2 3 4 5 6 7 8 9
0
20
40
60
80
100
120
140
NETD
Noise N [mV]
Signal S [mV]
S
i
g
n
a
l
,
N
o
i
s
e
,
N
E
T
D
[
m
V
]
Measurement no.
Fig. 8. Variation of the NETD excluding subjective (digital) noise.
Fig. 9. Original Image for measurement MTF (aligned focal lens 135 mm) with DT 1500
tester thermal camera and related software from Inframet.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
An Analysis Regarding the Decreasing of the Image Quality
with the Optical Misalignment
Fig. 10. The diagram MTF vs. spatial frequency. The loss of image quality in different cases of a
lens misalignment with 45 mm focal length.
Fig. 11. The diagram contrast vs. off-axis distance.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
90 Catalin Spulber, Octavia Borcan
The loss of image contrast with off-axis distance of a 45 mm focal length.
Fig. 12. The diagram MTF vs. spatial frequency. Comparison between MTF diagrams for lenses
with 45 mm and 135 mm focal lengths, with well alignment done. One can see, for example, that
for a same resolution of 0,5 lp/mm, a 3.3 times improvement of contrast using a lens with a focal
length 3 times greater, can be obtained.
Fig. 13. MTF aligned focal lens 45 mm.
Fig. 14. MTF aligned focal lens 135 mm.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
An Analysis Regarding the Decreasing of the Image Quality
with the Optical Misalignment
Fig. 15. MTF with misaligned
2 mm focal lens 135 mm.
Fig. 16. MTF with misaligned
2 mm focal lens 45 mm.
Fig. 18. MTF with misaligned
5 mm focal lens 135 mm.
Fig. 19. MTF with misaligned
5 mm focal lens 45 mm.
4. Conclusions
A better MTF and NETD evaluation with electronic instruments is most
appropriate;
A significant decrease in MTF is obtained at low level variations of optical
misalignement;
4.3 The MTF decreases with decreasing lens focal.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
92 Catalin Spulber, Octavia Borcan
R E F E R E N C E S
[1] Owen, A.: "Surveillance cameras steal away the night, in Laser Focus World, 33, 111-115 (1997).
[2] Borcan, Octavia, Spulber, C: Experimental method for observation prediction based on the
decision matrix, through day/night equipments in NIR and LWIR spectral ranges, in Infrared Imaging
Systems: Design, Analysis, Modelling, and Testing XX, edited by Gerald C. Holst, Proceedings of
SPIE Vol. 7300 (SPIE, Bellingham, WA 2009) 730011.
[3] Holst, Gerald C. Infrared Imaging Systems: Design, Analysis, Modelling, and Testing IX.
[4]
Saar Bobrov* and Yoav Y. Schechner: Image-based prediction of imaging and vision
performance, in J. Opt. Soc. Am. A/Vol. 24, No. 7/July 2007.
[5] Ronald G. Driggers, Curtis M. Webb, Stanley J. Pruchnic, Carl E. Halford and Ellis E. Burroughs,
"Laboratory measurement of sampled infrared imaging system performance", in Opt. Eng. 38, 852
(1999).
[6] ***ASTM E 1543 00: Standard Test Method for Noise Equivalent Temperature Difference of
Thermal Imaging Systems.
[7]
Arnold Daniels: Field Guide to Infrared Systems Field Guide to Infrared Systems, Detectors, and
FPAs, Second Edition, SPIE Press Book, Bellingham, 2010, ISBN: 9780819480804.
[8] Steven W.Smith: The Scientist and Engineer's Guide to Digital Signal Processing, 1997, ISBN
978-0966017632.
[9] Alan Irwin, Robert L. Nicklin:Standard software for automated testing of infrared imagers,
IRWindows, in practical applications, in Proceedings of SPIE, SPIE 3377, 206 (1998).
[10] Paul A. Bell, Carl W. Hoover, Jr., Stanley J. Pruchnic, Jr., Standard NETD test procedure for
FLIR systems with video outputs, Proc. SPIE 1969, 194 (1993].
[11] Spulber, C, Borcan, Octavia: Some Aspects Regarding the Image Acquisition using Video
Systems under Low Vibrations, in Annals of the Academy of Romanian Scientists Series on Science
and Technology of Information 6, Vol. 2, 87-98, Number 1/2009, ISSN 20662742.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Annals of the Academy of Romanian Scientists
Series on Science and Technology of Information
ISSN 2066-8562 Volume 5, Number 1/2012 93
APPLICATIONS OF QUANTUM CRYPTOLOGY
FOR DATA TRANSMISSIONS
IMPLEMENTED IN A STUDENT LABORATORY
Bogdan-Adrian STEFANESCU
1
, Dan ANGHEL
1
, Octavian DANILA
1
,
Paul STERIAN
1
, Andreea Rodica STERIAN
1
Abstract. Quantum cryptography based on the BB84 protocol is discussed in the
following presentation, containing the concepts and the work that has been carried out in
the field, with some developments suitable for student research. Although it has not been
implemented on a commercial level, data transmissions based on quantum cryptology is a
good alternative for integration in optical fibers communications, with a wide range of
applications due to its securing capabilities. Evolution in photon-study related fields,
such as photon echo, contribute to the better understanding and further improvement of
the quantum key distribution protocol. An efficient way of encrypting the information is
by the use of a key. As it is well known, the encryption key uses very complex algorithms
that are very hard to break but the problem of key transmission between the transmitter
and receiver still remains. On a classical channel, the answer was given in the form of
RSA public keys that were sent between the transmitter and receiver several times, and
implied the use of randomizing algorithms by use of prime numbers. Quantum approach
of this problem can be solved through the following principle: If a quantum system that
resides in a defined state is observed, thus measured, the state of that system is
irreparably changed. This has a direct application in detecting whether an eavesdropper
has entered the quantum channel or not. A student-oriented experimental apparatus is
presented, together with a virtual simulation of the protocol that implements the
principles of quantum cryptography. Our optical channel can be improved using the
photon echo effect. Excitement of superradiant states by irradiating a probe with a
coherent optical impulse, with its duration and intensity conveniently chosen can be
shown with the photon echo. We demonstrated that the photon echo can improve the code
by adding either a controlled error on the channel or transforming the channel from a
binary channel to a ternary channel.
Keywords: Quantum cryptology, Data transmission, superradiant states, ternary channel
1. Introduction
The goal of this paper is to help students understand the application of quantum
physics in information security.
Why is information security so important? In present days, a lot of information is
exchanged via large networks, such as a LAN or the Internet [1-3, 11]. If sensitive
information is exchanged, a way of guarding the information from unwanted
eavesdroppers is needed [4-6]. A quite simple way of doing this is by encrypting
1
Academic Center for Optical Engineering and Photonics, Faculty of Applied Sciences, University
Politehnica Bucharest, Romania (sterian@physics.pub.ro).
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Bogdan-Adrian Stefanescu, Dan Anghel, Octavian Danila,
94 Paul Sterian, Andreea Rodica Sterian
the information. An efficient way of encrypting the information is by the use of a
key. The encryption key uses very complex algorithms that are very hard to break,
but there is another problem: How does the sender send the key to the receiver? In
early days, the key was transported to the receiver via physical medium such as
paper, punch card, floppy disk, EEPROM or CDROM. There was no insurance of
the interception of the key.
On a classical channel, used in the 1970-s, the answer was given in the form of
RSA public keys[5], that were sent between the transmitter and receiver several
times, and implied the use of randomizing algorithms by use of prime numbers.
Quantum approach of this problem can be solved through the following principle:
If a quantum system that resides in a defined state is observed, thus measured, the
state of that system is irreparably changed. This has a direct application in
detecting whether or not an eavesdropper has entered the quantum channel.
2. Quantum Cryptography
Quantum cryptography solves the key distribution problem by allowing the
exchange of a cryptographic key between two remote parties with absolute
security, guaranteed by the laws of physics. This key can then be used with
conventional cryptographic algorithms [5]. If one encodes the value of a digital bit
on a single quantum object, its interception will necessarily translate into a
perturbation, because the eavesdropper is forced to observe it. This perturbation
causes errors in the sequence of bits exchanged by the sender and recipient. By
checking for the presence of such errors, the two parties can verify whether their
key was intercepted or not. It is important to stress that since this verification
takes place after the exchange of bits, one finds out a posteriori whether the
communication was eavesdropped or not. That is why this technology is used to
exchange a key and not valuable information. Once the key is validated, it can be
used to encrypt data. In telecommunication networks, light is routinely used to
exchange information. For each bit of information, a pulse is emitted and sent
down an optical fiber to the receiver, where it is registered and transformed back
into an electronic signal. These pulses typically contain millions of photons [3,
14]. In quantum cryptography, one can follow the same approach, with the only
difference that the pulses contain only a single photon. In particular a photon
cannot be split into halves.[6].
3. The BB84 protocol
The first protocol for QC has been proposed in 1984 by Charles H. Bennett, from
IBM New-York, and Gilles Brassard, from the University of Montreal, hence the
name BB84 under which this protocol is recognized nowadays. They published
their work in a conference in India, totally unknown to physicists.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Applications of Quantum Cryptology for Data Transmissions
Implemented in a Student Laboratory 95
We shall explain the BB84 protocol using the language of spin 1/2 , any 2 level
system being equivalent to it. The protocol uses two interlocutors, Alice, as the
transmitter, and Bob, the receiver, as well as an eavesdropper, Eve. The photons
of use are divided into 4 quantum states that constitute 2 bases, think of the states
up | i, down | i, left | i and right | i. Conventionally, one attributes the
binary value 0 to states | i and | i and the value 1 to the other two states, and
calls the states qubits (for quantum bits). In the first step, Alice sends individual
spins to Bob in states chosen at random among the 4 basic states (the spin states |
i,| i, | i and | i are identified with the polarization states horizontal,
vertical, +45 degrees and -45 degrees, respectively). How she chooses at
random is a delicate problem in practice, but in principle she could use her free
will. The individual spins could be sent all at once, or one after the other (much
more practical); the only restriction being that Alice and Bob can establish a one-
to-one correspondence between the transmitted and the received spins[1]. Next,
Bob measures the incoming spins in one of the two bases, chosen at random
(using a random number generator independent from that of Alice). At this point,
whenever they used the same basis, they get perfectly correlated results. However,
whenever they used different basis, they get uncorrelated results. Hence, on
average, Bob obtains a string of bits with 25% errors, called the raw key. This
error rate is so large that standard error correction schemes would fail. But in this
protocol Alice and Bob know which bits are perfectly correlated (the ones for
which Alice and Bob used the same basis) and which ones are completely
uncorrelated (all the other ones). Hence, a straightforward error correction scheme
is possible: For each bit Bob announces publicly in which basis he measured the
corresponding qubit (but he does not tell the result he obtained). Alice then only
tells whether or not the state in which she encoded that qubit is compatible with
the basis announced by Bob. If the state is compatible, they keep the bit, if not
they disregard it. In this way about 50% of the bit string is discarded. This shorter
key obtained after bases reconciliation is called the sifted key. The fact that Alice
and Bob use a public channel at some stage of their protocol is very common in
crypto-protocols. This channel does not have to be confidential, but has to be
authentic. Hence, any adversary Eve can listen to it all the communication on the
public channel, but she cant modify it. In practice Alice and Bob may use the same
optical fiber to implement both the quantum and the classical channels. Note that
neither Alice nor Bob can decide which key results from the protocol. Indeed, it is
the conjunction of both of their random choices which produces the key [4-6].
Let us now consider the security of the above ideal protocol (ideal because so far
we did not take into account unavoidable noise due to technical imperfections).
Assume that some adversary Eve intercepts a qubit propagating from Alice to
Bob. This is very easy, but if Bob does not receive an expected qubit, he will
simply inform Alice to disregard it. Hence, in this way Eve only lowers the bit
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Bogdan-Adrian Stefanescu, Dan Anghel, Octavian Danila,
96 Paul Sterian, Andreea Rodica Sterian
rate (possibly down to zero), but she does not gain any useful information. For
real eavesdropping Eve must send a qubit to Bob. Ideally she would like to send
this qubit in its original state, keeping a copy for herself.
4. Simple software simulation of the BB84 protocol
There are many software implementations used to simulate the BB84 protocol.
Some are written in PHP code, Java ,C++, or native quantum simulation software.
For this paper we have chosen a software written in Visual C++ called QIT
designed by Fernando Lucas Rodriguez. This is a general public license (GPL)
software and it can be found on the Internet for student training and research. The
program has an interactive graphical user interface (GUI) that allows students to
watch a step by step execution of the protocol. Students will use the program to
find the different quantum keys.
The program will run from one of the 2 computers on the lab table witch must run
a Windows XP 2 operating system with .NET 2 installed. From the folder QIT13
the students must execute QIT IDE.exe [9]. After the execution the following
window will appear (Fig. 1.):
Fig. 1. Output window of QIT.exe
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Applications of Quantum Cryptology for Data Transmissions
Implemented in a Student Laboratory 97
Then from the Interactive menu ->Communications->InteractiveBB84
Cryptosystem is selected. With this selection the interface for simulating the
BB84 protocol will appear (Fig. 2.):
Fig.2. GUI of QIT.exe
Example: The secret data to be sent is: Information Security and Cryptology.
Press Start, then Finish. The resulting binary secret key is:
110101001101110011010000101010000000010011110111101101111001000010
011000110101011110101111010000010101001001010011100100110111100000
000010010001100111101110110101011010001010111010100110010101110010
010110000000001011000101101011100010111110001010000111000000111110
1100100111000100
A step by step execution of the protocol is available by pressing Start then
Execute Step [9].
The figure 1 and 2 represent the main steps in the execution of the protocol.
Behind the software the steps are as follow: the (4+delta)N qubits will be
transmitted. For each bit, Master station selects a basis randomly (horizontal-
vertical or diagonal), and also selects a random value (true or false). It sends the
resulting quantum state after encoding the true or false value into the selected basis.
It also stores the basis selected and the random value. The slave station measures
values received into a random basis (horizontal-vertical or diagonal). It stores the
basis used for measurement and the value measured. The master station sends the
basis used for values encoding. The values are not transmitted, they are the secret.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Bogdan-Adrian Stefanescu, Dan Anghel, Octavian Danila,
98 Paul Sterian, Andreea Rodica Sterian
The slave station stores the master's used basis for encoding. The master station
stores coincidences. Slave station informs Master about which basis there was
coincidence during qubit encoding and slave measurement. Values measured are
not transmitted, only if there was coincidence.
The master and slave generates an intermediary 'secret private' key with the values of
qubits of which where was coincidence on basis encoding and measurement and
checks that intermediary key has at least the double of bits needed for chippering data.
The master select random intermediary-key bits indexes for public comparisons,
and informs of those positions to the slave. The number of bits to compare will be
the size of intermediary key minus the number of bits needed for chippering data.
Slave stores the indexes of the positions that will be compared.
The master sends the values that tried to encode on quantum channel of the selected
indexes. Slave stores the values that Master sends. The master stores the values that
Slave sends. Slave sends the values that measured on quantum channel of the
selected indexes.
Master calculates noise level (it knows what it tried to send, and what Slave
received). If noise rate is too high (over 25%) with high probability there is an
Eavesdropper and data transmission is aborted. Slave calculates noise level (it
knows what Master tried to send and what it received). If noise rate is too high
(over 25%) with high probability there is an Eavesdropper and data transmission is
aborted. The data is sent using Vernam chippering with the BB84 key on a Classic
public channel. The Slave Station stores and decodes the ciphered message with the
BB84 secret, private & secure key (Fig. 2).
5. Implementing the BB84 protocol in a student oriented experimental
apparatus
5.1. Background
The experimental setup will follow the initial idea that was used in the Optical
Institute of Orsay, France. In that case, the photons were transmitted between two
windows of two separate buildings. For coding the qubits the experiment used the
four polarization states (horizontal, vertical, circular left and circular right).
Physically, they were created by applying a voltage on an electro-optical modulator.
The qubit sequence resulting from the coded polarization is generated by hardware
means, by using two Fibonacci configured linear registries. Each registry has an
output of
20
2 1 bits, and the 4 states of the protocol suffer 2-bit coding, each of the
bits being a part of a pseudo-random sequence. For minimizing diffraction effects, the
radius of the photon beam is extended to 2 cm, by using a two-lens system, before
being transmitted 30 m through open air. The photons are collected by Bob through
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Applications of Quantum Cryptology for Data Transmissions
Implemented in a Student Laboratory 99
the same system, at which 4 avalanche photo-diodes were added. Measurements of
polarization were made by selecting the states, as the photons are either transmitted or
reflected by a beam splitter at the incidence angle of 45. Horizontal and vertical
states were created by the beam splitter, while the circular left and right states were
separated by a special splitter that converts circular polarization into linear
polarization, and then discern them through the same splitting method.
5.2. The efficiency of the single photon source and the Poisson statistics
Primary characteristics of the single photon source quality made by Alice consist of
measuring the single and multi-photon emission probabilities, compared with weak
coherent pulses (WCP), with the same amount of photons per pulse. For a
transmission sequence of 0.2 s and pulse frequency of 5.3 MHz, a total of
4
8.8 10
photons are recorded by Alice. By correcting the efficiency of the photo-diodes
(APD) 0.6
APD
q = the global efficiency adds up to 2.8%. After passing through the
modulator, characterized by 0.9
EOM
T = and 0.94
optic
T = , the mean of the sent
photons per pulse was 0.0235 = . Reduction of the multi-photon emissions can be
set in Alice's part of apparatus. Photon statistics can be counted precisely by Bob's
measurements, thus resulting in the distribution of the photon numbers. Evaluations
were carried out on
6
40 10 pulses recorded by Bob [1]. For a time sample, detection
probabilities for a photon or photon pair are
3
1
7.6 10
d
P
= and
6
2
2.7 10
d
P
= .
Configuration of the photo-diodes shows that the detection probability
2d
P is 5/8 of
the reception probability of the same pair, while the probability that a photon pair
falls on the same photo-diode is of 3/8. The reduction factor is given by:
2
1
2
5
6.7
8 2
d
d
P
R
P
= = . This result is in accordance with the Poisson distribution, and
thus can be used in further calculus. Information leak to an intruder connected to the
quantum channel is
( ) m
S , which shows the probability that a photon leaves Alice's
system. For an equivalent WCP we have
( ) 4
1 (1 ) 2.7 10
m
WCP
S e
= + = , while for
the SPS
( ) 5
1
[1 (1 ) ] 4.1 10
6.7
m
SPS
S e
= + = [4].
5.3. Detection probabilities of Bob's system
Detection probability of Bob's apparatus over a time threshold is
3
exp
7.6 10 p
~ .
Assuming that the absorption of the beam is negligible through the 30 m
transmission, and taking into account that the mean 0.0235 = , an estimate of the
detection apparatus is 0.3
Bob
q ~ . While measuring photons, errors might appear,
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Bogdan-Adrian Stefanescu, Dan Anghel, Octavian Danila,
100 Paul Sterian, Andreea Rodica Sterian
because of unpolarized light. This contributes to compromising the security of the
transmission, and decrease the maximum transmission distance. By filtering and
spectral insulation, the optical environment can be protected. Measurements were
made at night time, the probability of recording an incorrect photon per time
threshold is
5 1
3.8 10
dark
p s
= [4].
6. Our experimental proposal setup
The experiment will take place on a lab table, so there is no need to have
complicated alignment system. A simple laser diode and the transmitter can be
used for alignment, by moving the two supporting legs.
Fig. 2. Experimental proposal.
The experiment consists of two main modules: Alice and Bob, in our case,
transmitting and receiving student, that communicate over open air, at the
designated distance for laboratory of about 1 m.
The classical channel is a basic TCP/IP connection (coaxial or UTP cable). The
experiments for optical fibers transmissions are also considered.
6.1. Transmitter Module
Fig. 3. The Alice module.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Applications of Quantum Cryptology for Data Transmissions
Implemented in a Student Laboratory 101
The module is designed to produce a stream of single polarized photons according
to the choice of basis and bit value. Because no source can produce single
photons, we use pulses that have the property of coherency. They are called weak
coherent pulses (WCP), of Poisson distribution and mean photon number
0.0235 = . To produce the pulses, we will use four laser diodes, oriented around
a conical mirror at the desired polarization angles. The polarization problem is
solved by the laser diodes, that have intrinsic polarization. After the beams are
reflected by the conical mirror, they pass a spatial filter, which consists of two
100 m at 0.9 cm apart. It serves a special purpose, that of making the pulse from
the four diodes indistinguishable from the others, in spatial terms. This measure
has to be taken because without the spatial filtering, the code can be broken quite
easily. In order to get as much light as possible through the spatial filter, there is a
lens with a focal length f = 2.75 mm between the conical mirror and the pinholes
of the spatial filter. Because of the very strong spatial filtering, the alignment of
the pinholes is crucial, otherwise the desired mean photon count will not be
achieved for all polarizations. [4]
6.2. Receiver Module
Fig. 4. Receiver module schematic.
The module is the heart of the receiver unit is connected directly to a receiver lens
and a spatial filter (SF), that are positioned so that the transmitted beam is focused
on the primary device of the module. The primary device is an interference filter
with a red color glass filter. This is important to allow for daylight operation,
because it rejects stray light, while permitting polarized light to pass. The
remaining optical devices divide the photon beams into bases H/V and +/-, the
construction being based on the idea by John Rarity and Paul Tapster. An incident
photon sees the 50/50 beam splitter (BS). If it is reflected it will see the polarizing
beam splitter(PBS) of the photon in the H/V basis, which in combination with the
two silicon avalanche photo diodes (APD) H and V. If APD V detects a photon it
is supposed to be in the V basis, whereas if APD H detects a photon, it is
supposed to be in the H basis.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Bogdan-Adrian Stefanescu, Dan Anghel, Octavian Danila,
102 Paul Sterian, Andreea Rodica Sterian
Any photon that is transmitted through the beam splitter passes through a half-wave
plate, set at an angle of 22.5, so that it rotates the linear polarization by 45 degrees.
Afterwards, a +45 degrees polarized photon is detected by APD 1 and converted into
the horizontal basis H, while a - 45 degrees polarized photon is detected by APD 3
and converted into the vertical basis V. Whenever a photon is measured in the wrong
basis, the measurement outcome is completely random. The APD-s have to be cooled
in order to reduce dark counts, at a temperature between - 25 and - 10 degrees. To
reach these temperatures, the photo-diodes are put into an aluminum block which is
cooled by a Peltier element glued to it from below [4]. The bit error rate (BER) [6] for
each channel was estimated from data taken during key exchange. It is given by the
expression:
wrong
total
N
BER
N
= , where N
wrong
is the number of bits in error and N
total
is the
number of bits received in total. This gives a measure of the likelihood of receiving a
zero when a one was sent from the transmitter. All but one of the BER values in table
1 are sufficiently low showing that optical imperfections from the equipment will
contribute little to the error in the sifted key. The system will operate in natural light
and artificial light in the lab, thus the background error rate must be considered. This
is a factor that could limit the entire experimental setup. We start from the signal
count for this rig [8]:
4
RMT
S
q
= , where R is the repetition rate, M is the average
number of photons per pulse, T is the lumped transmission and is the detection
efficiency of Bob's APDs. The product is divided by 4 because there are 4 detectors
(or 4 polarization states). The background rate is given by
b
P Bt = , where B is the
background the background count rate per APD and t is the time synchronization
gate. Half the counts induce errors and half of there are thrown away. The error
rate is:
b
base
P
E E
S
= + . E must be less the 0.07 therefore the maximum acceptable
background rate is:
75.5
MT
B
t
q
< . In considering the system presented here,
estimates can be made for the following values [8]:
1. M ~ 0.3, an accepted value for guaranteed security of low loss systems.
2. T ~ 1 since the source can be imaged on to the receiver and the system is
short range and thus atmospheric loss is negligible.
3. ~ 0.045 taking into account the quantum efficiency of the detectors and the
presence of the narrow band filter and polarizers.
4. t = 5 ns gate synchronization time.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Applications of Quantum Cryptology for Data Transmissions
Implemented in a Student Laboratory 103
Thus the maximum background count rate per detector can be given as roughly
B 36000 counts/s. In this system, the data is recorded first during the quantum
transmission and processed afterwards in a few seconds. The start of the
transmission is determined approximately by searching for a jump in the
frequency of time tags as Bob starts measuring before Alice begins her
transmission. Alice transmits sub-5 ns pulses every 200 ns, therefore a time
synchronization gate of 5 ns reduces the probability of registering a background
event within the gate by a factor of 40 [8]. The clock at the receiver is thus
synchronized with the clock at the transmitter by searching for time tags that sit at
separations of 200 ns and adjusting the time separation slightly every ~100 ms to
compensate for clock drift. The advantage of this set-up is that no timing
reference signal is needed. To determine the exact start time of the data, the
receiver reveals a random subset of his measured bit values and the basis he used
to the transmitter. The transmitter then finds the data start by performing a sparse
correlation against her stored data. This random subset can also be reused to
estimate the error rate.
Fig. 5. Full experimental rig.
It should be noted that our implementation of error correction requires that the two
parties both generate the same random factor graph. Once both of them know the
number of message bits they are error correcting over, and the measured error rate,
they seed a pseudo-random number generator from their OTP and use this to generate
an appropriate factor graph.
The eavesdropper, is assumed not to know which of the 2
256
, say, different factor
graphs the communicating parties are using [8].
For implementing this type of encryption we also can use fiber optics, as used in
Clavis [9, 10] devices for BB84 code implementation.
Fiber optics may provide a much longer distance for light propagation, thus
facilitating the wide area implementation for this type of secured data
transmissions.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Bogdan-Adrian Stefanescu, Dan Anghel, Octavian Danila,
104 Paul Sterian, Andreea Rodica Sterian
7. Improving BB84 by using photon echo
Our optical channel can be improved using the photon echo effect [3]. The code
encrypting can be made by introducing a supplementary key if we use the photon
echo. We have three states associated to the 1 and 0 qubits that are available for
encrypting. That correspond to the possibility of associating to the two input
impulses two or three output impulses, the third one corresponding to the photon
echo, generated after an algorithm or by our own will. Excitement of super radiant
states by irradiating a probe with an coherent optical impulse, with its duration
and intensity conveniently chosen can be shown with the photon echo [3]. In
principle, we consider the exciting of a ruby crystal, for example, with two
coherent, identical optical impulses A and B (emitted by a ruby laser), having a
duration of t (ns), the delay of B impulse when passing through the probe, will
be T t, showing in the figure 6 [12, 13].
Fig. 6. Principle of photon echo.
We can see at the exit, beside the two impulses A' and B', which correspond to the
emitted impulses A and B, a third impulse A'', symmetrical positioned to A' and
B', named photon echo [3]. For explaining the photon echo, we use the precession
equation, for a circular polarized radiation field, in a reference system which
rotates round the z axis with the frequency :
0
(
i
d
P P E k
dt
ee
= _ +
_
.
Due to the unequal width of the transition frequency for ruby atoms, and due to the
local field, variation, P and
0
will be affected by index k , which defines the k atom.
If before applying impulse A, the atoms were all in the fundamental state, the
vector
1
N
k
k
P P
=
=
=
+ =
n
i
i i
E C H p C H C C H
1
' ) ( '
, where
) ' ( C C H
and ( )
i
E C H ' are the uncertainty functions corresponding to the set { }
'
j i
E E =
Cartesian product of the statistical collectives C and C, and to the statistical
collective C, in conditions when the event
i
E appeared.
It was found [5] that the uncertainty function H(
1 2
, , ...
n
p p p ) fulfilling the above
indicated conditions is: H(
1 2
, , ...
n
p p p ) =
n
i
i b i
p p a
1
log , where a and b are
almost arbitrary constants, that satisfy the conditions: a > 0 and b > 1. One finds so
that the uncertainty function H(
1 2
, , ...
n
p p p ) represents the average (theoretical)
value of the so-called information entropy, defined by the relation:
i b i
p a S log = (1)
Similarly, for the continuous statistical collectives (described by the probability
density p(x)), the uncertainty function is given by the expression:
}
A = dx x x p x p a x p H
b
) ) ( ( log ) ( )) ( ( [where x is the (conveniently chosen)
quantum of the variable x] and the information entropy by the expression:
) ) ( ( log )) ( ( x x p a x p S
b
A = . (2)
We have to underline also that the expression (2) is absolutely similar to the
(previous) Planck-Boltzmann expression: = ln
. B therm
k S of the thermo-
dynamic entropy (k
B
is the Boltzmanns constant, while stands for the
probability density of micro-states localisation in the phases space).
2. Logical scheme of the humankind information accumulation
It is well known that the information processing and storage abilities of each
individual people brain are drastically limited. For this reason, the humankind
advance in its race for the complex systems knowledge and use imposes the
strong cooperation of the human beings by information transmission. Taking into
account that the information transmission is a resonance process (see fig. 1), it is
necessary to ensure: a) the obtained (got) information (see fig. 2) cleaning before
a new experiment (measurement, embryo development, Universe genesis, etc.),
b) a sufficiently broad and well-located information receiver bell, c) an implant
(inside the information receiver bell) of several connecting relays, achieving the
cross-fertilization between the information source(s) and its virtual applications,
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
The Mathematical theory of Communications versus the physical theory of Information.
Universe versus Multiverse 109
so that Toute la suite des hommes depuis le cours de tant de sicles est comme
un seul homme qui vit toujours et qui apprend continuellement (Blaise Pascal).
Fig. 1. Information transmission as a resonance process. Fig. 2. Magnetic memory
example of got information.
3. Main Conceptual Differences between Mathematics and Nature Sciences
3.1. Typical elementary object
While in Mathematics the typical elementary object (the problem unknown) is a
well-defined number or segment, in Nature Sciences this elementary object is a
parameter , described by a certain probability distribution P(p) of the individual
values p (see figure 3).
While the value of the unknown of a mathematical problem with a right
formulation is obtained exactly by means of the problem solution, the most
probable individual value (named also true value, or mathematical hope)
p
t
of the physical parameter p cannot be never exactly obtained!
For this reason, the definition of the real information amount has to be given by
means of the overlap area of the normalised to 1 probability distribution functions
corresponding to measurements and to the true parameter, respectively (fig. 4).
Fig. 3. Probability distribution Fig. 4. Definition of the true information amount
of a parameter p individual values. obtained by measurements.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
110 Dan Alexandru Iordache
3.2. Uniqueness parameters
While the number of the uniqueness parameters of a mathematical problem is
fixed [e.g.: 3 for an arbitrary triangle (the lengths of the 3 sides, or the lengths of 2
sides and the angle between them, etc.)], the number of the uniqueness
parameters of a physical system depends on the required accuracy for the
considered system description [e.g., the thermodynamic state of the air is
described by: (i) only 2 parameters (usually the temperature and the pressure) in a
first order approximation, (ii) by 3 physical parameters (adding e.g. the humidity)
in the frame of a better approximation, (iii) 4 physical parameters (adding also the
carbon dioxide content) in the frame of a still better approximation, etc.
3.3. Well-formulated problems
While in mathematics a well-formulated problem corresponds usually to a system
of compatible and non-redundant equations, the number of this system equations
being equal to the number of unknowns of the mathematical problem, in nature
sciences a well-formulated problem corresponds to a system of (slightly)
incompatible (and non-redundant) equations, and the number of equations has to
be considerably larger than that of unknowns. This fact is due to the fluctuations
of the individual values of the physical parameters and even to the presence of
some hysteretic behaviour (the individual values could depend on the system
previous history) of the physical systems.
3.4. Position of the incomplete induction method
While in mathematics the incomplete induction method represents only the first
step towards the inference (particularly, by the complete induction method) of a
theorem, in Physics this (incomplete induction) method represents an essential
method, because it allows the discovery of some truths which are not equivalent to
the information set used to formulate the respective hypothesis. The incomplete
induction method represents one of the most fertile methods used by the nature
sciences for the identification of some new plausible hypotheses and the
subsequent discovery of some new physical phenomena and laws.
4. On the bridge between the mathematical theory of communications
(information) and the physical theory of information
For a uniform distribution of the true value t
X
inside its corresponding confidence
interval:
} }
= = =
+
)
~
(
~
)
~
(
~
)
~
( 2 ) ( 1
n L n
n L n
x s z x
x s z x
n L X X X
x s z C dt C da t p ,
hence the corresponding expression of the uncertainty function is:
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
The Mathematical theory of Communications versus the physical theory of Information.
Universe versus Multiverse 111
}
A
= A =
x
x s z
a dt x t p t p a t p H
n L
b X X b X X
)
~
( 2
log ) ) ( ( log ) ( )) ( ( . (3)
It results that the apparent information obtained in frame of the n
th
physical
determination can be expressed by means of the square mean errors corresponding
to the sets of the obtained results obtained after the (n-1) determination and after
the n
th
determination as:
n
n
b n n n app
a H H I
o
o
1
1 .
log
= = .
One finds so that the usual I
app.n
> 0 values (corresponding to the convergence towards
the true value), it is possible to meet also values I
app.n
< 0, which could be due to:
a) rough errors (hence misinformation),
b) random gathering of the first individual values, the decision being
established by means of some statistical tests.
It results that the additional elements brought by the physical theory of
information refer mainly to: (i) experimental measurements, (ii) the corresponding
errors, (iii) the necessary statistical tests.
For this reason, the compatibility of a given theoretical relation y = f(x) with a certain
set of experimental individual values pairs x
s
, y
s
(s=1, 2, N) should be decided not
starting from the usual correlation coefficient which does take into account the
existing experimental errors, but from the error risks at the compatibility rejection of
each suspect pair x
s
, y
s
:
( )(
(
=
2
1 2
exp
r
p
q
s
s
, where (see also fig. 5):
(
+
(
=
) (
~
) (
~
2
) (
~
) (
~
2 2
y
y y
x
x x
r
y
y y
x
x x
p
s s s s
s
o o o o
. (4)
Fig. 5. Evaluation procedure of the error risk at the compatibility rejection
of a theoretical relation Y = f(X) relative to some local data.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
112 Dan Alexandru Iordache
5. Main features of the complex systems description
Because several completely different complex systems [computer arrays, robots,
networks, social sciences, biology (with some specific topics: colonies, swarms,
immunology, brain, genetics, proteomics), non-linear dynamics, economics,
mathematics, glasses, agents, cognition, etc.] have some common features centred
on their statistical behaviour and the corresponding phase transforms [6], [7], it
results that these complex systems have certain universality properties, which
due to their generality (see e.g.[7a]) - can be described only by some specific
numbers (the so-called similitude numbers, or criteria [8]).
If: | | | |
i
n
i
i
U
U P
o
[ =
=1
, where [P] is the physical dimension of a parameter P specific
to the studied state (or process), then 2 states (or processes) , are named
similar if the values of the parameters { } n i U
i
, 1 | = and P corresponding to these
states fulfil the relation [8]: ( )
i
U
n
i
i i
U U
P
P
o
[ =
=1
" '
"
'
. Some of the uniqueness
parameters could be similitude criteria, i.e. non-dimensional parameters: [s] = 1,
with equal values: s = s in all similar states or processes. While the first known
similitude criterion was introduced by Archimedes (287-231 b. Chr.):
2
3
v
A
=
gl
Ar , the first (existence) theorem of the similitude theory was stated by
Newton, all these theorems being presented in work [9].
Fig. 6. Plots of different pseudo- convergent Fig. 7. Gradual installation of instability in
simulations of elastic pulses propagation. simulations of elastic pulses propagation.
The accomplished study [1] of the typical study procedure of complex systems
pointed out that it involves the following main stages: a) identification of the
uniqueness parameters, b) identification of the characteristic similitude criteria, c)
obtainment of the set of irreducible criteria, d) translation of all relations of
scientific and/or technical interest in terms of similitude criteria, e) check of the
theoretical and experimental similitude models, f) test of compatibility of
theoretical relations and models relative to the existing experimental data.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
The Mathematical theory of Communications versus the physical theory of Information.
Universe versus Multiverse 113
Fig. 8. Distortions in the simulations of some random walk processes [10].
6. Main numerical phenomena intervening in the Data Processing
A detailed study of the main numerical phenomena: pseudo-convergence (fig. 6),
instability (fig. 7), distortions (fig. 8), intervening in the computer evaluations of
certain physical parameters and/or some simulations of different physical
phenomena was accomplished by work [10].
Taking into account the various errors types and numerical phenomena
intervening in the data processing, we consider as the most accurate data
processing procedure that presented in the frame of fig. 9.
Fig. 9. Basic Stages of the present Scientific Information Processing.
7. Interpretation of the physical information about Complex Systems
Unlike the classical (mathematical) information, the physical information (referring to
complex systems, especially) requires a very careful interpretation. First of all, it is
necessary to answer to the basic questions about the observed features:
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
114 Dan Alexandru Iordache
a) are they random or reproducible?
b) could they be connected to other results, obtained by different methods?
c) could they be explained by natural causes or it seems to intervene some
transcendent reasons?
We have to underline that the acceptance of a physical interpretation needs
multiple completely different experimental results, whose explanations converge
to this interpretation. E.g., the existence and parameters (charge, mass) of
electrons were established as a result of AT LEAST 5 completely different
experiments:
(i) the electrolysis (Faradays) laws leading to the elementary electrical charge
e = F/N,
(ii) the J.J. Thomsons experiment concerning the cathodic rays deviations in
an electrical field, which pointed out the existence of the electron,
(iii) the Millikans experiment which led to the electrical charge of the
electron,
(iv) the Lenards method of crossed (electrical and magnetic) fields, which
allowed the evaluation of the specific charge e/m of the electron,
(v) the Comptons effect which allowed the evaluation of the rest mass of the
electron.
Without redundant (completely different) experimental methods, the Physics is
often subject to major errors; some recent examples:
a) the so-called anomalons (1970-1980), erroneous interpretation supported
initially by several very good Physics reviews,
b) the so-called fusion nuclear reactions at low temperatures (Palladium
compounds, 1980-1990), again a mis-interpretation,
c) the trans-uranium 118 element, initially claimed by a research group of the
Berkeley University and vanished after 2-3 years.
For this reason, the interpretation of some new Physics experiments has to be
cautiously examined; some examples:
(i) the Palo Alto results concerning the magnetic monopoles and the exotic
particles, generally,
(ii) the very recent (2011) results of Pamelas orbital station, referring to the
anomalous strong fluxes of accelerated cosmic radiations around our planet,
which seem to indicate that the Earth has an absolutely singular location in our
(Milky Way) galaxy, etc.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
The Mathematical theory of Communications versus the physical theory of Information.
Universe versus Multiverse 115
8. Just Six Numbers seem be able to describe the Universe structure [12].
The Anthropic principle(s) [15]
In 1937, the British Physics Nobel prize laureate Paul A. M. Dirac had noted that the
number of baryons (basically protons plus neutrons) in the universe (~ 10
77
) is almost
equal to the inverse square of the gravitational coupling constant
|
|
.
|
\
|
~
=
39
2
10 906 . 5
c
m k
C
p
g
=
h
e
c
e
C
o o
e
W and Z
0
predicted by them were not still discovered (they were experimentally found by
C. Rubbia and S. van der Meer only 4 years later). For such reasons, the
confidence of physicists in the unified theories was so high that it was a deep
disappointment [17], [26] to find that these unified theories are not valid for the
Universe evolution descriptions.
Taking into account that:
a) while the masses of the intermediary vector bosons predicted by the unified
theory of weak and electromagnetic interactions are only 2 magnitude orders
larger than that of protons,
b) the parameters of the Big Bang processes are more than 25 magnitude orders
distant to somewhat usual ones, c) the Physics advance from Democrits
atomistic theory to the quantum atomic Physics (over 7 magnitude orders) require
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
120 Dan Alexandru Iordache
more than 20 centuries, we dont have to be exaggerate: even if now the Physics
progresses are much accelerated, its advance over more than 25 magnitude orders
(up to the Big Bang field) will require probably several decades (perhaps even
centuries)! We have to be patient to be accumulated in the following decades and
(probably) centuries sufficient experimental data to be able to formulate valid
theoretical cosmological models.
As it concerns the Bible, it seems that: a) inside the Bible structure are hidden
some important information, b) the modern Physics and the Bible predictions are
convergent. That is why the Bible deserves a thorough study, for its scientific
information, and not only for its outstanding ethics recommendations (cultivate
empathy, fight our selfishness, etc.). Despite the main goal of Bible is to improve
the human beings ethical behaviour, it involves also some (rather few) scientific
elements. However, one finds that the number of Bible sentences initially without
any scientific meaning that got in the last centuries such a connotation is
monotonically increasing, and even in an accelerated manner [e.g. (few examples)
from the: a) transcendent integers (John 21:11, Acts 27:37) to the: b) role of
Information in the building of complex systems (Genesis 1:3), c) appearance
order of light sources (Genesis 1:3 and 1:14), d) chromosomes (Genesis 2:22), d)
ADN defects and repairs, implicitly (Genesis 4 and 5), e) the special relativity
theory (2 Peter 3:8), f) main features of the Big Bang process (see above), etc.].
Either it happened that between a tremendous number of parallel Universes,
with different features [19a] our Universe be the unique (accidental) one [26]
with suitable conditions for the life existence and (on Earth) of the intelligent life
presence, or this Universe and Earth were created by supernatural design, it results
that the humankind has a huge responsibility to run optimally the only one
World experiment, whose main actors we are.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
The Mathematical theory of Communications versus the physical theory of Information.
Universe versus Multiverse 121
R E F E R E N C E S
1. a) DOBRESCU R., IORDACHE D., eds, Complexity Modelling (in Romanian), Politehnica
Press Printing House, Bucharest, 2007; b) DOBRESCU R., IORDACHE D., Complexity and
Information, Romanian Academy Printing House, 2010.
2. IORDACHE D., Main complexity features of the thermo-mechanical evolution of the universe,
chap. 5 in Research Trends in Mechanics, vol. 2, Romanian Academy Printing House, 2008.
3. HARTLEY R.V.L., Transmission of Information, Bell System Techn. J., July 1928.
4. a) SHANNON C., The Mathematical Theory of Communication, Bell Syst. Techn. J., 27, 379-
423, 623-56(1948); ibid., 30, 50(1951); b) SHANNON C. E., WEAVER W., The mathematical
theory of communications, Illinois Urbana Univ. Press, 1949.
5. a) KHINCHIN A.J., Mathematical Foundations of Information Theory, Dover, New York,
1957; b) KHINCHIN A.J., in Arbeiten zur Informationstheorie I, VEB Verlag, Berlin, 1961, pp. 7-
85; c) KOLMOGOROV A.N., ibid., pp. 91-116; d) GUIAU S., Information theory with
applications, McGraw Hill, New York, 1977.
6. a) ANDERSON P.W., Science, 177, 293(1972); b) ANDERSON P.W., Proc. Natl. Acad.
Science (USA), 92, 6653-6654(1995).
7. a) SOLOMON S., SHIR E., Europhysics News, 34(2) 54-57(2003); b) SOLOMON S., Annual
Reviews of Comp. Physics II, 243-294, D. Stauffer ed., World Scientific, 1995.
8. a) GUKHMAN A. A., Introduction to the Theory of Similarity, Academic Press, New York,
1965; b) BARENBLATT G. I., Dimensional Analysis, Gordon and Breach, New York, 1987; c)
BARENBLATT G. I., Scaling, Self-Similarity and Intermediate Asymptotics, Cambridge Texts in
Applied Mathematics, 1996.
9. BODEGOM E., IORDACHE D., Physics for Engineering Students, vol. 1, Classical Physics,
Politehnica Press, Bucharest, 2007
10. IORDACHE D., Contributions to the Study of Numerical Phenomena intervening in the
Computer Simulations of some Physical Processes, Credis Printing House, Bucharest, 2004.
11. DIRAC P.A.M., The Cosmological Constants, Nature 139, 323(1937).
12. REES M., Just Six Numbers, Basic Books, 2000.
13. DICKE R.H., Diracs Cosmology and Machs Principle, Nature 192, 440-441(1961).
14. a) ROOD R.T., TREFIL J.S., Are we alone? The possibility of Extraterrestrial Civilisations,
Scribners sons, New York, 1986; b) TIPLER F.J., The search for extraterrestrial life: recent
developments, Physics Today, 40, 92, December 1987.
15. WHEELER J.A., Foreword, in The Anthropic Cosmological Principle, Oxford, Clarendon
Press, UK, 1986.
16. EVERETT H., Many-World Interpretation of Quantum Mechanics, PhD Dissertation at
Princeton, 1956.
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
122 Dan Alexandru Iordache
17. SMOLIN L.: a) The Trouble with Physics, Houghton Mifflin Harcourt, 2006; b) Thread-Bar
theories, IEEE Spectrum, January 2007, www.Spectrum.ieee.org/geek-life/tools-toys/ threadbar-
theories.0
18. a) GUTH A. H., Inflationary Universe: A possible solution to the Horizon and Flatness
problems, Physical Review D, 23, pp. 347-356(1981); b) GUTH A. H., STEINHARDT P. J., The
Inflationary Universe, Scientific American, p. 116, May 1984; c) GUTH A. H., The inflationary
universe. The quest for a new of cosmic origins, Addison-Wesley, Reading, 1997.
19. a) LINDE A., The Self-Reproducing Inflationary Universe, Scientific American, pp. 48-55,
November 1994; b) LINDE A., Particle physics and Inflationary cosmology, Physics Today, 40(9)
61-68 (Sept. 1987), & treatise, Harwood Academic Publishers, 1990; c) LINDE A., Inflation and
Quantum cosmology, Academic Press, 1990.
20. FERMAT C.S., Diophantus Arithmetica containing (48) observations by P. de Fermat,
Toulouse, 1670.
21. A. WILES A., Modular elliptic curves and Fermats last theorem, Annals of Mathematics,
142, 443-551(1995).
22. SINGH S., Fermats Enigma: the Epic Quest to Solve the Worlds Greatest Mathematical
Problem, Walker Publishing Company, New York, 1997.
23. a) WITZTUM D., RIPS E., ROSENBERG Y., Equidistant letters sequences in Genesis,
Statistical Science, 9(3) 429-438(1994); b) M. Drosnin, The Bible code, World Media Inc., vol. 1
(1997), vol. 2 (2002).
24. HUBBLE E., A relation between the distance and radial velocity among extra-galactic
nebulae, Proc. National Academy of Sciences, 15, pp. 168-173(1929).
25. JOHNSON P., A History of Christianity, Athenaeum, New York, 1976, p. 413.
26. LIGHTMAN A.P., The accidental Universe: Sciences crisis of faith, Harpers Magazine,
December 2011; d) www.harpers.org/archive/2011/12/0083720
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Annals of the Academy of Romanian Scientists
Series on Science and Technology of Information
ISSN 2066-8562 Volume 5, Number 1/2012 123
CHAOS AND STABILIZATION
OF
SELF-REMISSION TUMOR SYSTEM
BY
SLIDING MODE
M.R. JAFARI
1
, M.R. ZARRABI
1,2
, S. EFFATI
1,3
Abstract. In this paper a pray-predator system that called self-remission tumor is
considered, and a new approach in order to stabilizing the unstable equilibrium
points of self-remission tumor system with sliding mode control is introduced.
The stability analysis of the biologically feasible equilibrium points is presented by
using the Lyapunov function.
A Lyapunov function is supposed for designing a sliding surface (SS).
Lyapunov function is constructed to establish the global asymptotic stability of the
uninfected and infected steady states by describing sliding surface (SS), after that
by considering the derivation of SS as zero, someone can achieve the equivalent
control that inbreed system stays on SS and tends to equilibrium point in infinite
horizon.
In addition, numerical examples are presented to illustrate the effectiveness of the
proposed method.
Keywords: Chaos, Tumor, Equivalent control, Sliding surface
1. Introduction
Cancer is one of the greatest killers in the world and the control of tumor growth
requires special attention [9].
The mathematical modeling of cancer self-remission and tumor has been
approached by a few numbers of researchers under using a variety of models over
the past decades [8, 9, 13, 18].
Many authors have discussed the problem of the chaotic behavior and stability
analysis of some biological models such as cancer and tumor model, genital
herpes epidemic, stochastic lattice gas prey-predator modes [7, 8, 13] and many
other models.
1
Department of Applied Mathematics, Ferdowsi University of Mashhad (see above picture),
Mashhad, Iran, (mreza.jafari26@gmail.com).
2
(mo.za870@gmail.com).
3
(s-effati@um.ac.ir).
124 M.R. Jafari, M.R. Zarrabi, S. Effati
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Chaos and Stabilization of self-Remission Tumor System by Sliding Mode 125
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
126 M.R. Jafari, M.R. Zarrabi, S. Effati
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Chaos and Stabilization of self-Remission Tumor System by Sliding Mode 127
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
128 M.R. Jafari, M.R. Zarrabi, S. Effati
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Chaos and Stabilization of self-Remission Tumor System by Sliding Mode 129
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
130 M.R. Jafari, M.R. Zarrabi, S. Effati
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Chaos and Stabilization of self-Remission Tumor System by Sliding Mode 131
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
132 M.R. Jafari, M.R. Zarrabi, S. Effati
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Chaos and Stabilization of self-Remission Tumor System by Sliding Mode 133
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
134 M.R. Jafari, M.R. Zarrabi, S. Effati
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected
Chaos and Stabilization of self-Remission Tumor System by Sliding Mode 135
Copyright Editura Academiei Oamenilor de tiin din Romnia, 2012
Watermark Protected