Accepted Manuscript Organic physically unclonable function on flexible substrate operable at 2 V for IoT/ IoE security applications Kazunori Kuribara, Yohei Hori, Toshihiro Katashita, Kazuaki Kakita, Yasuhiro Tanaka, Manabu Yoshida PII:
S1566-1199(17)30420-2
DOI:
10.1016/j.orgel.2017.08.022
Reference:
ORGELE 4269
To appear in:
Organic Electronics
Received Date: 24 March 2017 Revised Date:
12 August 2017
Accepted Date: 18 August 2017
Please cite this article as: K. Kuribara, Y. Hori, T. Katashita, K. Kakita, Y. Tanaka, M. Yoshida, Organic physically unclonable function on flexible substrate operable at 2 V for IoT/IoE security applications, Organic Electronics (2017), doi: 10.1016/j.orgel.2017.08.022. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
Binarize
= finger print of device
Chip 2
Unique ID Bit error rate
SC
RI PT
Compare each ring oscillator for frequency
Chip 1
Same design
0.5 0.4 Low error rate 0.3 high reproducibility 0.2 0.1 0 1.7 1.8 1.9 2 2.1 2.2 2.3 Operation voltage [V]
ACCEPTED MANUSCRIPT
Organic physically unclonable function on flexible substrate operable at 2 V for IoT/IoE security applications Kazunori KURIBARA1*, Yohei HORI1, Toshihiro KATASHITA2, Kazuaki KAKITA2,
RI PT
Yasuhiro TANAKA2, and Manabu YOSHIDA1
1National Inst. of Advanced Industrial Science and Technology 1-1-1 Higashi, Tukuba, Ibaraki, 305-8565 Japan
2Ube Indus., 8-1 Minami-kaigan, Goi, Ichihara, Chiba, 290-0045 Japan
SC
*E-mail:
[email protected]
M AN U
Abstract We fabricated organic ring oscillators (ROs) on a flexible substrate and utilized them as the core circuit of a physically unclonable function (PUF). An RO-PUF is a security primitive that generates unique identification numbers (IDs) by extracting the frequency variation of the RO. We fabricated two RO-PUFs and evaluated their IDs in terms of stability and uniqueness at various operating voltages. The experimental results indicate that our RO-PUFs have a high degree of uniqueness and exhibit good stability relative to voltage fluctuations with a nominal operating voltage below 2 V.
1. Introduction the
Internet
intensively studied [2][3]. A PUF is a type of security device that utilizes the
of
internal variations that arise during the
Things/Everything (IoT/IoE) has become
fabrication process. For this reason, it is
a focus of constant attention. In the
fundamentally difficult to counterfeit,
IoT/IoE, various devices are assembled
imitate, or falsify PUF devices. Organic
into one or more networks. A large
circuits are advantageous for realizing
AC C
EP
Recently,
TE D
Keywords ‐ Organic transistor, Self-assembly, Low voltage, Flexible, Security, and Physically unclonable function
problem
in
the security of ambient electronics, such
determining how best to authorize or
as RFID tags, polymer banknotes and
identify devices that reside at the edges
wearable
of the network so they can communicate
flexibility
correctly. However, it is difficult to
implemented on polymer or thin film
arbitrarily create unique identification
substrates. In addition, some types of
numbers for large numbers of devices. To
organic/inorganic
address this problem from a security
dissolved into organic solvents to create
perspective,
inks with chemical modifications, which
functions
in
these
systems
physically (PUFs)
[1]
lies
unclonable have
been
electronics or
due
to
stretchability
materials
their when
can
be
are expected to reduce the production
ACCEPTED MANUSCRIPT
cost of printing technology in the future.
treatment for 30 min to form a 4-nm thick
However, to the best of our knowledge,
aluminum
organic devices, and PUFs in particular,
immersed the substrate into an isopropyl
have been rarely studied as security
alcohol solution of n-octadecylphosphonic
circuits.
acid for 2 h. After immersion, we rinsed
layer.
Then,
we
RI PT
oxide
In this study, we fabricated organic
the substrate with isopropyl alcohol and
PUFs on flexible substrates, evaluated
annealed it at 100 °C for 10 min [4]. The
their stability in terms of changes in the
combination
operating voltage, and estimated their
self-assembler functions as the gate
uniqueness
dielectrics. We then deposited organic
ID
generation
applications.
the
SC
for
of
AlOx
and
semiconductor DNTT (p-type) [5][6] and TU-1
(n-type)
[7][8]
onto
the
gate
M AN U
dielectrics layer by thermal evaporation.
2. Experimental 2.1. Fabrication process
Finally, Au was deposited onto the
We fabricated organic devices using
semiconductor to act as the source and
solution
drain electrodes, and the circuit traces.
processes. The transistor structure is
The nominal channel length was 10 µm,
shown in Fig. 1(a), and includes the top
and the channel widths were 250 µm (for
contact and bottom gate structure. First,
p-type) and 1,000 µm (n-type). Using
thermal
evaporation
and
these complementary transistor elements,
layer onto a 75-µm thick layer of
we fabricated 14 three-stage organic ring
polyimide based film to act as gate
oscillators (ROs) (see Fig. 1(b) and (c)).
TE D
we deposited a 25-nm thick aluminum
electrodes. Subsequently, this aluminum
2.2. Evaluation procedure for the organic RO
AC C
EP
layer was exposed to an oxygen plasma
Fig. 1 (a) Schematic structure of the 2-V operable complementary inverter with an organic thin film transistor. (b)(c) Optical image of the organic ring oscillator (b) and the ring oscillator array (c). (d) Schematic diagram of the number generation system.
ACCEPTED MANUSCRIPT
responses from the same PUF, and
PUF
inter-HD among the responses from
groups, and two organic RO PUFs were
different PUFs. The intra-HD represents
assembled that each consisted of 7 ROs. A
the stability of the PUF responses, and
schematic diagram of the RO PUF is
the ideal value is zero. The intra-HD is
shown in Fig. 1(d). The RO PUF receives
calculated
two inputs (= challenges) that target two
responses and the actual response matrix
ROs, then their oscillating frequencies
(see Fig. S2(a)). The expected response is
are compared to generate a 1-bit output
one in which each bit is determined by
(= response). To estimate the bit error
majority
rate, the frequency of an RO is obtained
responses. The inter-HD represents the
by averaging 1001 times of measurement
uniqueness of the PUF responses, and
with a 500-ms interval, which is limited
the ideal value is 0.5. The inter-HD is
by the measurement system. Therefore,
calculated
the measurement time becomes 500 s, in
matrices of two (or more) PUFs (see Fig.
total. After the 500 s, the measurement
S2(b)).
period elapses, 1001 frequency points are
2.3. Materials and equipment
by
RI PT
The 14 ROs were divided into two
XORing
over
expected
the
obtained
M AN U
SC
voting
the
by
XORing
the
response
A flexible polyimide film (UPILEX-75s,
× 1001 data points are obtained for each
Ube industries, Ltd.) was used for the
RO PUF.
substrate. For the organic semiconductor
TE D
obtained for each RO, and eventually 7
By comparing the frequency data of the
materials,
ROs at each measurement time, we
Dinaphtho[2,3-b:2′,3′-f]thieno[3,2-b]thiop
generate the responses of the RO PUF. In
hene was purchased from sigma-aldrich.
this
possible
An N-type soluble semiconductor named
challenges (= the combination among
TU-1 was provided from Ube industries,
seven ROs) becomes 21 = 7C2, and thus
Ltd.
we obtain a response matrix of size 21 ×
n-octadecylphosphonic acid is purchased
1001 for each RO PUF (see Fig. S1). The
from PCI synthesis.
number
of
EP
the
AC C
case,
columns of this matrix represent the
For
Self-assembled
measurement,
we
monolayer,
utilize
the
response to each challenge at each time,
DSO9104A digital oscilloscope (Keysight
and the rows of the matrix show the time
technologies)
evolution
software interface named BentchVeu. For
of
the
response
for
the
performance
characterized
its
data-logging
the voltage source of RO, the 2400 source
challenges. The
and
by
of the
the
PUF
is
intra-PUF
Hamming distance (intra-HD) among the
meter (Tektronix, Inc.) is also used. All measurements
were
carried
atmospheric air and in the dark.
out
in
ACCEPTED MANUSCRIPT
inverters connected in series. Therefore, the delay time, which is related to the
3. Results and discussion 3.1. Organic ring oscillator
inverse
frequency,
becomes
the
summation of the delays of all inverters.
waveform of our organic ring oscillator.
The frequency of the ring oscillator fosc
This three-stage RO works well with an
can be calculated as follows.
2(a)
shows
the
oscillation frequency of about 1 kHz. The p-type
transistors
and
that
n-type were
organic fabricated
simultaneously show mobilities of 0.6 and 0.06 cm2/Vs, respectively. The channel length L is 10 µm, and the channel
1 ∙ 2 2
1 ∙ 2 2
‐ ‐ ⋯ Eq. 1
M AN U
overlap Wc is also 10 µm. The cut-off
SC
single
RI PT
oscillation
Figure
In this equation, n is defined as the
to the transconductance gm, and inversely
number of invertor stages, µ is the field
proportional to the gate capacitance in a
effect mobility of the transistor, and Vdd
unit area C[9]. A ring oscillator consists of
and Vth are the operating and threshold
AC C
EP
TE D
frequency of the transistor is proportional
Fig. 2 (a) Typical wave form of the organic ring oscillator element. (d) Distribution of oscillation frequency in the seven organic ring oscillators. (c) Typical histogram at an oscillation frequency at 2.2 V. The inset blue histogram shows the calculated DC bias stress effect. (d) Histogram of the corrected frequency.
ACCEPTED MANUSCRIPT
voltages,
respectively.
The
frequency
dependence is also shown in Figs. 3 (a)
calculated using Eq. 1 for the fabricated
and (b) in which the operating voltage
ring oscillators is 1.2 kHz, which is
was adjusted from 1.8 to 2.2 V, which is a
consistent with the measured frequency.
10%
fluctuation
from
the
nominal
3.2. PUF characteristics
RI PT
operating voltage of 2 V. In Fig. 3(a), the intra-HD for both chips 1 and 2 increased up to 0.06 when the operating voltage
ROs with operating voltages of 2 V is
increased up to 2.2 V. However, the
shown in Fig. 2(b). The shape of each RO
average intra-HD remained below 0.1.
frequency histogram appears asymmetric,
On the other hand, the inter-HD was
which is likely due to the DC bias stress
hardly affected at all by the voltage
effect.
shape
fluctuation, and was consistently around
Gaussian
0.3. In this study, we calculate the
distribution if we remove the DC bias
inter-HD from a 1 bit length of data,
stress effect from the frequency (see Fig.
which means that the ideal inter-HD is
2 (c) and (d)).
0.5 if the frequency distribution is
is
approximates
because
that
of
the a
M AN U
This
SC
The frequency distribution from seven
In this study, the average intra-HD was
completely random. Compared with the
at an operating voltage of 2 V,
ideal value of 0.5, an inter-HD of 0.3 is
which is small in comparison to the
relatively high. In addition, the error rate
inter-HD of 0.3 at the same operating
estimated from the intra-HD of 4.8 × 10-5
AC C
EP
TE D
4.8 ×
10-5
Fig. 3 (a)(b) Operation voltage dependence on the intra-PUF Hamming distance (a) and inter-PUF Hamming distance (b). (c) Time evolution of the oscillation frequency among the 14 ring oscillators. voltage.
The
operating
voltage
is negligibly small for inter-HD, which
ACCEPTED MANUSCRIPT
oscillation
fosc
indicates that we can distinguish each
the
frequency
and
chip by the ID generated by the PUF.
operating time t is as shown below, noting that Vgs in Eq. 2 is changed to Vdd in Eq.
Our organic RO-PUF exhibited a low intra-HD over a period of 8 min, despite the large DC bias stress effect. The oscillation frequencies of the ROs as a
1. ' * ' ∙ "#$ % & ) + ‐ ‐ ⋯ 01. 3 (
RI PT
3.3. DC bias stress effect on the organic PUF
In this equation, f0 is the initial value of
shown in Fig. 3 (c). From this graph, it
the frequency (i.e., at t = 0).
SC
function of the measurement time are
We calculated the relaxation time and
the RO frequency to lower. However, this
stretch parameter for all ring oscillators
phenomenon
ROs,
using Eq. 3, and obtained the average of
simultaneously. Therefore, there are a
those parameters as a function of the
few cross points in the graph of the time
operating voltage (see Fig. 4). Both
evolution, which means that the order of
parameters were affected by the changed
frequencies among the 14 ROs seldom
operating voltage, as was the intra-HD.
changes during the measurement. It is
From Fig. 4, the carrier trap time τ was
for this reason that intra-HD is small
shortened by one order of magnitude,
enough below 2 V.
while the operating voltage increase from
on
all
TE D
occurs
M AN U
can be seen that DC bias stress causes
One of the reasons for the DC bias
1.8 to 2.2 V. The relaxation time of 103 s at
in either the semiconductor and insulator,
measurement time of 500 s. Therefore,
or at the interfaces. The threshold
the
voltage shift in this carrier trap model
affected by the DC bias stress at high
can be defined as follows [10].
operating voltages. It also means that our
AC C
EP
stress effect is carrier trapping at defects
' * ∆ !1 "#$ % & ) +, ‐ ‐ ⋯ 01. 2 ( - , /
2.2
V
was
intra-HD
is
comparable more
to
the
significantly
RO-PUF can show a lower bit error rate when
the
measurement
interval
is
shortened from 500 ms. From DC bias effect correction, it can be said that an intrinsic variation in organic
In this equation, τ is the carrier trapping
devices is large enough for thermal
relaxation time, and β is a stretch
fluctuation so that an organic PUF with
parameter that indicates the variation in
the variation can work well. We have
the trap state depth in this study. From
obtained
Eqs. 1 and 2, the relationship between
reproducibility of PUF ID generation in
a
low
error
rate
and
ACCEPTED MANUSCRIPT
short-term
measurement.
almost
organic
all
However,
materials
show
degradation by air or time. Although we utilize air stable materials [7][11] and air
[12],
these
materials
are
RI PT
materials that show thermal stability in also
degraded gradually. Therefore, in the next
step,
it
becomes
a
technical
challenge to investigate whether the circuit
can
keep
the
organic
PUF
characteristic in long-term use, i.e.,
a function of the operation voltage.
M AN U
several months or years.
Fig. 4 Fitting parameters in Eq. 3 as
SC
passivation layer or the compensation
Acknowledgement
4. Conclusion
In this study, we fabricated organic
This study was partially supported by
ring oscillators on flexible substrates for
the “Next-Generation Printed Electronics
use as the core circuits of security devices.
Material
The oscillators were then evaluated as
Technology Development” project of the
&
Process
Foundation
New Energy and Industrial Technology
number based on the variation of the
Development Organization (NEDO).
TE D
PUFs, which generate a unique binary oscillation. As a result, the error rate for generating unique numbers was 4.8 ×
EP
10-5 at 2 V. The intra-HD remained below 0.1 even when the operating voltage was
AC C
increased up to 2.2 V. On the other hand, the
inter-HD,
which
indicates
the
uniqueness of the chip, was 0.23–0.33. This is larger than the intra-HD value. Therefore, our ring oscillator PUF was found to be stable and unique. Moreover, an analysis of the DC bias stress characteristics suggests that the carrier trapping time plays a significant role in the stability of the fabricated RO-PUFs.
Reference [1] Ravikanth, Pappu Srinivasa. “Physical one-way functions.” Diss. Massachusetts Institute of Technology, (2001). [2] Yang, Kaiyuan, et al. "14.2 A physically unclonable function with BER< 10−8 for robust chip authentication using oscillator collapse in 40nm CMOS." 2015 IEEE International Solid-State Circuits Conference-(ISSCC) Digest of Technical Papers. IEEE, (2015). [3] Alvarez, Anastacia, Wenfeng Zhao, and Massimo Alioto. "14.3 15fJ/b static physically unclonable functions for secure chip identification with< 2% native bit instability and 140× Inter/Intra PUF hamming distance separation in 65nm." 2015 IEEE International Solid-State Circuits Conference-(ISSCC) Digest of Technical Papers. IEEE, (2015). [4] Klauk, Hagen, et al. "Ultralow-power organic complementary circuits." Nature 445.7129 (2007): 745-748. [5] Yamamoto, Tatsuya, and Kazuo Takimiya. "Facile synthesis of highly π-extended heteroarenes, dinaphtho [2, 3-b: 2', 3'-f] chalcogenopheno [3, 2-b] chalcogenophenes, and their application to
ACCEPTED MANUSCRIPT
SC
RI PT
[10] Chen, Te-Chih, et al. "Investigating the degradation behavior caused by charge trapping effect under DC and AC gate-bias stress for InGaZnO thin film transistor." Applied Physics Letters 99.2 (2011): 022104. [11] Zschieschang, Ute, et al. "Flexible Low Voltage Organic Transistors and Circuits Based on a High Mobility Organic Semiconductor with Good Air Stability." Advanced Materials 22.9 (2010): 982-985. [12] Kuribara, Kazunori, et al. "Organic transistors with high thermal stability for medical applications." Nature communications 3 (2012): 723.
M AN U
[9]
TE D
[8]
EP
[7]
AC C
[6]
field-effect transistors." Journal of the American Chemical Society 129.8 (2007): 2224-2225. Haas, Simon, et al. "High-performance dinaphtho- ‐ thieno-thiophene single crystal field-effect transistors." Applied Physics Letters 95.2 (2009). Mamada, Masashi, et al. "A unique solution-processable n-type semiconductor material design for high-performance organic field-effect transistors." Chemistry of Materials 27.1 (2014): 141-147. Mamada, Masashi, et al. "Crystal structure and modeled charge carrier mobility of benzobis (thiadiazole) derivatives." New Journal of Chemistry 40.2 (2016): 1403-1411. Lin, Y-M., et al. "100-GHz transistors from wafer-scale epitaxial graphene." Science 327.5966 (2010): 662-662.
ACCEPTED MANUSCRIPT
Highlights: Fabricated organic ring oscillators were used in physically unclonable functions
•
The fabricated devices generate IDs using their inherent frequency variation
•
The devices support a nominal operating voltage below 2 V
•
Results indicate our devices are unique and stable vs. voltage fluctuations.
AC C
EP
TE D
M AN U
SC
RI PT
•