Sei sulla pagina 1di 8

Solid-State Electronics 49 (2005) 1009–1016

www.elsevier.com/locate/sse

Accurate modeling and parameter extraction method


for organic TFTs
a,*
M. Estrada , A. Cerdeira a, J. Puigdollers b, L. Reséndiz a, J. Pallares c,
L.F. Marsal c, C. Voz b, B. Iñiguez c
a
Sección de Electrónica del Estado Sólido (SEES), Departamento de Ingenierı́a Eléctrica, CINVESTAV, Av. IPN No. 2508,
Apto. Postal 14-740, 07300 DF, Mexico
b
Departament dÉnginyeria Electronica, Universitat Politecnica de Catalunya, C/Jordi Girona, Modul C4, Barcelona 08034, Spain
c
Departament dÉnginyeria Electronica Eléctrica i Automatica, Universitat Rovira i Virgili, Avda. Paisos Catalans 26, 43007 Tarragona, Spain

Received 18 November 2004; received in revised form 14 February 2005; accepted 15 February 2005
Available online 31 March 2005

The review of this paper was arranged by Prof. A. Zaslavsky

Abstract

In this paper we demonstrate the applicability of the unified model and parameter extraction method (UMEM), previously devel-
oped by us, to organic thin film transistors, OTFTs.
The UMEM, which has been previously used with a-Si:H, polysilicon and nanocrystalline TFTs, provides a much rigorous and
accurate determination of main electrical parameters of organic TFTs than previous methods. Device parameters are extracted in a
simple and direct way from the experimental measurements, with no need of assigning predetermined values to any other model
parameter or using optimization methods. The method can be applied to both experimental and simulated characteristics of organic
TFTs, having different geometries and mobility. It provides a very good agreement between transfer, transconductance and output
characteristics calculated using parameter values obtained with our extraction procedure and experimental curves. Differences in
mobility behavior, as well as other device features that can be analyzed using UMEM are discussed.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Organic TFT; Parameter extraction; TFT modeling

1. Introduction MOS crystalline devices. However, OTFTs present sev-


eral differences with respect to crystalline (MOSFETs),
Simulation of devices requires both, an accurate amorphous, (a-Si:H TFTs) or polycrystalline Si (polySi
model to represent the behavior of the device and an TFTs) devices. According to measured transfer and out-
extraction method strictly related to this model to deter- put characteristics of organic TFTs, it has been observed
mine with precision the device parameters required by it that mobility usually increases with gate voltage. It has
and thus by simulation. Modeling of organic TFTs also been reported that mobility increases with VDS
(OTFTs) has been made, up to now, analytically or for a given value of VGS, in a more pronounced way
numerically, mostly using the same expressions as for than the mobility of a-Si:H devices with similar geome-
try, [1–3]. This increase of mobility with VDS has been
*
Corresponding author. Tel.: +52 55 50613786; fax: +52 55
related to hopping [4]; to a Frenkel–Poole type effect
50613978. associated to traps in the material, [5] as well as; to
E-mail address: mestrada@cinvestav.mx (M. Estrada). the lowering of the emission barrier in traps located at

0038-1101/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.sse.2005.02.004
1010 M. Estrada et al. / Solid-State Electronics 49 (2005) 1009–1016

the boundaries between crystals, since most materials Finally we present a complement to the UMEM in
used to fabricate OTFTs are polycrystalline, [6,7]. order to consider, if necessary, the non-ohmic contact
OTFTs can present deformation at the origin of the at drain and source, sometimes present in OTFTs. For
output characteristics, due to non-ohmic contacts at this last treatment, analytical expressions are also pro-
drain and source [8–10]. They can also have leakage cur- vided for modeling this effect and for the extraction of
rent across the gate dielectric [11] and/or polarization ef- the required parameters.
fects [8], which can be significant and have to be taken
into account.
In numerical simulations, the longitudinal field 2. Main features of unified model and parameter
dependence of the mobility has been introduced by extraction method (UMEM)
external models, [5], evaluated by Monte Carlo [1], or
simply not considered, [12]. 2.1. Summary of steps to extract model parameters
Some authors have used simulation in SPICE using
the MOSFET model, to which some features as series The mobility dependence with gate voltage is de-
and non-linear resistance at drain and source are some- scribed as [15]:
times added, [8,2]. The use of the model 15 for a-Si:H  c
ðV GS  V T Þ a
devices in AIMSpice was reported in [13], with an addi- lFET ¼ l0 ¼ lFET0  ðV GS  V T Þca ð1Þ
tional treatment of the series and non-linear OTFTs V aa
resistance. The non-linear OTFT resistance was also where lFET0 is the a value of mobility for low perpendi-
treated in [14]. Although this last approach has some cular and longitudinal electric field, VT the threshold
advantages regarding previous, the method used voltage and ca and Vaa are fitting parameters. l0 is usu-
for parameter extraction is numerical and among its ally taken as the band mobility for the material of the
disadvantages are a relatively big number of fitting para- TFT under analysis. Since this parameter is usually esti-
meters, which can be determined by optimization mated, the fitting parameter Vaa is used to adjust lFET0
methods when using AIMSpice, or through graphical to the experimental value of the low field mobility for
or derivative methods which are troublesome and impre- the device being modeled. Parameter ca is related to
cise due to experimental errors. For example, with AIM- the conduction mechanism and can describe both an in-
Spice extractor, it is difficult to obtain the same group of crease or decrease in mobility with VGS. In the first case,
fitting parameters for all modeled transistor curves, ca > 0 and in the second ca < 0. The first behavior is typ-
which also lack physical meaning. In addition, some of ical of amorphous and nanocrystalline devices and is re-
the extraction procedures, require to previously assume lated to trap conduction mechanism, while a decrease in
values to some parameters in order to extract the rest of mobility with gate voltage appears in polycrystalline
them, that is, some parameters cannot be extracted inde- TFTs when surface scattering starts to be important.
pendently one from another [15]. In organic TFTs, as will be shown later, both behaviors
In the last years, we have been applying our unified were observed. In this way, extracted fitting parameters
model and parameter extraction method (UMEM) first can be used nevertheless, to analyze physical mecha-
to a-Si:H, [15,16], to polycrystalline, [17] and recently nisms that take place in the devices.
to nanocrystalline silicon transistors [18]. Among the Drain current in the linear and saturation regions is
advantages of this method are that all above threshold modeled as:
parameters are extracted from two transfer characteris-
W lFET  ðV GS  V T Þ
tics, one in the linear and the other in the saturation re- I DS ¼  C diel  
gion and from the output characteristic of the device L 1 þ R WL  C diel lFET  ðV GS  V T Þ
under study, using a single mathematical processing V DS ð1 þ k  V DS Þ
h h im i1=m þ I 0 ð2Þ
involving and integral method that additionally reduces
experimental noise. The method can be used to compare 1 þ VVDSsat
DS

devices with different geometries and fabrication condi- where W is the channel width, L is the channel length,
tions under the same parameter extraction conditions. Cdiel is the gate capacitance, R is source plus drain resis-
This method uses analytical expressions for both model- tance, I0 is the leakage current and m and k are fitting
ing and parameter extraction, so calculations can be parameters related to the sharpness of the knee region
made using any program for mathematical calculations. and to the channel length modulation respectively.
In addition it was also implemented in AIMSpice [19]. Parameter k describes the variation of conductance with
In this paper we show that UMEM provides an excel- VDS in the saturation region.
lent tool for modeling OTFTs with different geometries, The saturation voltage is defined through the satura-
dielectric materials and fabrication processes. We apply tion modulation parameter aS:
it to analyze and compare the mobility behavior of four
different types of pentacene TFTs. V DSsat ¼ aS ðV GS  V T Þ ð3Þ
M. Estrada et al. / Solid-State Electronics 49 (2005) 1009–1016 1011

For the extraction procedure in the above threshold the maximum gate voltage value you want to model.
regime, we will use the same integral function H(VGS) Select a value of VDS in the saturation region, not too
defined in [15]: far from the knee of the output selected curve for which
R V GS the drain current IDS1 is known from measurements. This
I DS ðxÞdx
H ðV GS Þ ¼ 0 ð4Þ drain voltage value will be referred as VDS1 and should be
I DS ðV GS Þ equal to:
After substituting (1) in (2) and calculating H(VGS) from V DS1 ¼ aS  ðV GS1  V T Þ ð9Þ
(4), the following expression is obtained for above
threshold regime: This value of IDS1 corresponds to IDSsat(VDSsat) in
1 expression (8).
H a ðV GS Þ ¼ ðV GS  V T Þ ð5Þ Step no. 5: Parameter k is then extracted from:
2 þ ca

8 h h im i1=m 9
>
< ðI DS2 Þ2  ½1 þ R  K  lFET ðV GS1 Þ  ðV GS1  V T Þ  1 þ asðVV DS2V Þ >
=
ðV DS2 Þ GS1 T 1
k¼  ; ð10Þ
>
: K  lFET ðV GS1 Þ  ðV GS1  V T Þ >
; V DS2

The generalized parameter extraction procedure for dif- where VDS2 is selected not too far from the maximum
ferent devices and operation regimes can be found in drain voltage measured in the characteristic to be mod-
[15–18]. The steps for the determination of the model eled. IDS2 is the measured drain current for the selected
parameters VT, ca, aS Vaa, m and k are summarized drain voltage VDS2.
below.
If the slope of the IDS–VGS curve in the linear region
Step no. 1: VT is obtained from the intercept, and ca is observed to decrease due to the presence of a series
from the slope in expression (5). resistance, the value of this resistance can be determined
Step no. 2: From 1 the experimental data, calculate as indicated in [15].
ðI DS Þ1þca vs. (VGS  VT) and its slope Sl. OTFT can show a non-linear contact at drain and
The value of Vaa is extracted as: source. To model also this region, the following simple
 1=ca and precise procedure can be used.
KV DS
V aa ¼ ; ð6Þ
Sl1þca 2.2. Modeling of OTFTs output characteristics when
non-ohmic contacts are present
where K ¼ WL  l0  C diel . In this form, the three parame-
ters that determine the effective change of the field effect If a non-ohmic contact is present at drain and source,
mobility in the above threshold regime are extracted. which is frequently observed in OTFTs, the external bias
Step no. 3: Using the saturation current characteristic applied, Vext, will fall part on the non-ohmic contact,
for VDS P VGS  VT, calculate the slope (diode), Vdiode and part on the transistor, VDS, including
1=ð2þcaÞ
Ss in the linear region of I DSsat vs. its series resistance, that is:
(VGS  VT). Parameter aS is extracted as:
pffiffiffi V DSext ¼ V DS þ V diode
Ss2þca V caaa 2  
aS ¼ ð7Þ I DS kT I DS
K ¼ þn  log
GðV GS ; V DS Þ q I do
Step no. 4: Parameter m, is calculated evaluating (2) at  
VDSsat for a value of gate voltage near the I DS kT I DS
¼ þn  log ; ð11Þ
maximum measured, neglecting R and k: GðV GS ; V DSext Þ  n q I do
 where
K  lFET ðV GS Þ
m ¼ log 2= log
½1 þ K  lFET ðV GS Þ  ðV GS  V T Þ
 GðV GS ; V DSext Þ
aS ðV GS1  V T Þ
 ð8Þ KlFET ðV GS Þ  ðV GS  V T Þ  ð1 þ k  V DSext Þ
I DSsat ðV DSsat Þ ¼ h  m i1=m
ð1 þ R  lFET ðV GS Þ  ðV GS  V T ÞÞ  1 þ VV DSext
DSsat
To determine IDSsat(VDSsat) select a measured output
characteristic for VGS = VGS1, where VGS1 is a value near ð12Þ
1012 M. Estrada et al. / Solid-State Electronics 49 (2005) 1009–1016

and n is a fitting parameter to account for the real volt- The effect of n is to provide the necessary displace-
age across the transistor when the diode resistance is ment and adjustment of the slope of the modeled curve
significant. to the measured one, in the region of deformation.
To model output characteristics presenting this effect, After determining n, calculate k from (11) and (12),
after following steps 1–4 previously described, the diode solving for a value of drain voltage VDSext = VDS2 (and
parameters must be determined. It must be indicated its corresponding IDS = IDS2) not too far from the maxi-
that for steps 1–4, the transfer curve selected must cor- mum drain voltage measured in the selected output
respond to a VDSext in the linear region, that at the same curve for VGS2.
time lies outside of the region affected by the deforma- Finally, IDS vs. VDS is modeled substituting the
tion of the non-linear contact. extracted parameters in (11).

2.2.1. Extraction of diode parameters


After parameters VT, ca, aS, Vaa and m are deter- 3. Experimental part and results
mined, select an output characteristic with VGS = VGS2
where the non-ohmic contact effect is clearly seen. Plot Four different pentacene TFTs, T1, T2, T3 and T4
the curve log(IDS) vs. VDSext and determine the slope, were studied. All of them are bottom gate pentacene
B, and the intercept, A, of this curve near the origin. TFTs on glass substrate. Transistor T1 uses polymethyl
Using the normal procedure for a diode: methacrylate (PMMA) as gate dielectric and were pre-
pared as indicated in [20]. The PMMA dielectric thick-
log e
n¼ ð13Þ ness is 700 nm thick with a dielectric constant of 3.9.
B  kT
q Pentacene thickness was 160 nm. These devices have
L = 120 and W = 600 lm.
and T2 type transistors were fabricated by a process sim-
I do ¼ 10A : ð14Þ ilar as indicated in [21]. The gate dielectric is a 100 nm
polyvinylphenol (PVP) layer. Pentacene layer thickness
was of 30 nm, W = 500 lm and L = 50 lm.
2.2.2. Extraction of fitting parameter n Transistors T3 type correspond to devices previously
The effect of the presence of this diode is to reduce the reported in [21], with 120 nm polyvinylphenol (PVP)
drain current at small values of bias, when the diode layer as gate dielectric, 30 nm of pentacene layer,
resistance is high. As the external bias is increased be- W = 500 lm and L = 50 lm.
yond the knee of the I–V characteristic of the diode, Transistor T4 corresponds to devices previously re-
the drain current increases and starts to be determined ported in [22]. The gate dielectric was a 400 nm SiO2
mainly by the transistor. This means that in the output layer. The thickness of the pentacene layer was 50 nm,
characteristic, near or after the maximum slope in this W = 220 lm and L = 20 lm. Since we did not have mea-
region of deformation, most of the applied voltage starts sured characteristics of T4 type devices, we used simu-
to fall across the transistor. For this reason, parameter n lated characteristics in ATLAS1 to compare with
is determined as: output characteristics modeled using UMEM. Parame-
    ters used for simulation were taken from [5], where it
I DS ðV GS2 ; V ext2 Þ I DS ðV GS2 ; V ext1 Þ
 was shown that they provide an excellent agreement
G1ðV GS2 Þ G1ðV GS2 Þ with measured curves for these devices.
n¼ ð15Þ
½V DS2  V DS1  Trap distribution was taken as:
 
where IDS(VGS2, Vext2) and IDS(VGS2, Vext1) are the mea- ðEc  EÞ
gatail ðEÞ ¼ gatail0 exp  ð17Þ
sured current for the selected output characteristic at Eatail
two external bias Vext2 and Vext1 limiting an approxi-  
mately linear region of the curve IDS vs. VDSext,, where ðE  EV Þ
gdtail ðEÞ ¼ gdtail0 exp  ð18Þ
the maximum slope of this curve is included. VDS2, Edtail
(VDS1) are determined substituting Vext2, (Vext1) and its
where gatail0 is the concentration of acceptor tail states at
corresponding IDS in (8).
conduction band Ec equal to1 2.5 · 1018 cm3; gdtail0 in
G1(VGS2) is the conductance in the linear region eval-
the concentration of donor tail states at valence band
uated for VGS = VGS2:
EV, equal to 1 · 1018 cm3 and Eatail = 0.129 eV and
KlFET ðV GS2 Þ  ðV GS2  V T Þ Edtail = 0.5 eV are the activation energy for acceptor
G1ðV GS2 Þ ¼ : ð16Þ and donor tail states respectively.
ð1 þ R  lFET ðV GS2 Þ  ðV GS2  V T ÞÞ

R and lFET are extracted and calculated from extracted


parameters, respectively. 1
ATLAS is property of Silvaco International.
M. Estrada et al. / Solid-State Electronics 49 (2005) 1009–1016 1013

Table 1
Technological parameters of pentacene transistors
Transistor Pentacene layer [nm] Dielectric gate W [lm] L [lm] Reference
Type [nm]
T1 160 PMMA 700 600 120 [20]
T2 30 PVP 100 500 50
T3 30 PVP 120 500 50 [21]
T4 50 SiO2 400 220 20 [5,22]

The energy gap was taken as 2.8 eV and permitivity


of ei = 3. The extended states concentration at valence
and conduction bands was taken as 2.88 · 1021 cm3.
Pentacene is considered to be p-type with hole concen-
tration of 5.75 · 1017 cm3.
According to [5], externally it was pffiffiffiadded
ffi a field
dependence in the form of a  exp½b  F , where F is
the longitudinal electric field, a = 0.5 and b = 3 · 103
(V/cm)1/2, to account for the presence of a Frenkel–
Poole effect.
Table 1 shows the summary of the fabrication param-
eters for the four types of transistors.

3.1. Modeling of current–voltage characteristics


Fig. 2. Measured and modeled output characteristics of T2 type
Figs. 1–4 show measured and modeled output charac- transistor.
teristics of the four types of transistors described. As can
be seen, all devices characteristics analyzed in this work
can be very well modeled using UMEM, without the
need of any additional term to consider Frenkel–Poole
effect.
In some OTFTs, the dependence of the drain current
with drain voltage in the saturation region of the output
characteristic is more pronounced that the observed in
a-Si:H devices, however it can be still well modeled
through parameter k. This is also in agreement with
reported by [13].

Fig. 3. Measured and modeled output characteristics of T3 type


transistor.

Table 2 shows the summary of extracted parameters


for one transistor of each type of devices.
As we mentioned, one of the advantages of using
UMEM is the possibility of monitoring the behavior
of mobility, as well as to obtain distinguishing features
of each device. For example, from the extracted para-
meters indicated in Table 2, we can see that transistors
T1, T2 and T3 type are normally off, while T4 is nor-
Fig. 1. Measured and modeled output characteristics of T1 type mally on. Series resistance reduces from T1 to T4. In this
transistor. later case it is so small that cannot be determined.
1014 M. Estrada et al. / Solid-State Electronics 49 (2005) 1009–1016

Fig. 4. Measured and modeled output characteristics of T4 type


transistor.

Fig. 5. Measured and modeled transfer characteristics of T1 type


transistor in the linear and in the saturation regions.
Parameter as, which is related to the behavior of the out-
put characteristic at the onset of saturation, is higher for
transistors T2 and T3 type, which is also observed in the transfer curve, trying to cover the widest possible region
output characteristics in Figs. 2 and 3, where the linear where function H(VGS) is linear with VGS.
region is larger relative to the measured range of VDS. Another feature worth to remark is the comparison
Also parameter m, related to the knee of the curve at on- of the VT values obtained by our method and by the
set of saturation is approximately half the observed for standard
p ffiffiffiffiffiffiffi method for crystalline devices pffiffiffiffiffiffiffi basedð1þc
in the
Þ
transistors T2, T3 and T4 type. I DS vs. VGS. In Fig. 6 we plotted I DS and I DS a vs.
Fig. 5 shows agreement between modeled transfer VGS curves for one T1 type transistor. If VT is obtained
pffiffiffiffiffiffiffi
curves for T1 type transistor for two values of VDS, for this device using a I DS vs. VGS curve, as is fre-
one in the linear region (VDS = 1 V) and the other in quently done by authors, the value obtained is around
saturation, (VDS = 40 V). As can be seen, the agree- 18 V. This will not correspond to what is observed
ment is excellent, using the same group of fitting param- from measured output characteristics, where transistor
eters for both curves. This is really difficult to obtain is clearly off for VGS around 4 V. The cause of this is
with other extraction procedures, while using UMEM understood
pffiffiffiffiffiffiffi from Fig. 6, where it is clear that the curve
is easy, provided you correctly select the voltage range I DS vs. VGS does not present any region where you
in the linear and saturation region where the integral can assure that there is a linear behavior. However, with
ð1þc Þ
function H is evaluated. Table 2 also indicates the volt- the representation I DS a vs. VGS, there is a wide linear
age range used for each case. In general, best results are region, in which VT can be determined accurately. Say-
obtained when the voltage range in both regions is se- ing it in other words, OTFTs do not show the square
lected near the maximum slope of the corresponding root behavior of drain current with drain voltage typical

Table 2
Extracted parameters
T1 T2 T3 T4
Linear region voltage range (35–39) (8–11) (10–20) (60–100)
Saturation region voltage range (30–39) (8–11) (15–20) (80–100)
Output characteristic VGS [V] 40 14 20 40
VT [V] 4.1 2.6 3.9 +12.3
ca 1.9 0.6 0.15 0.072
Vaa [V] 1.7 · 103 106 1.4 · 103 1.7 · 104
lFET0 [cm2/V s] 7.4 · 107 0.058 0.52 0.7
R [kX] 1.1 · 104 200 14 0
as 0.39 1.7 1.4 0.935
m 1.27 2.8 2.97 2.8
k [1/V] 3.5 · 103 1.1 · 102 9.6 · 105 2 · 104
n – 8.9 – –
I0 [A] – 1.9 · 108 – –
n – 1.37 – –
M. Estrada et al. / Solid-State Electronics 49 (2005) 1009–1016 1015

0.30 0.08

T1 type transistor Transistors T2, T3 and T4 type


1/(1+γ a)

I IDSI
0.25 VDS = 40 V
5
γα = 3.07 0.06 T3
[µA]

1/2

a
0.20

γ
[V]
1/(1+γa )

0.15 0.04

µFET/µFETo
I IDS I

[µA]
0.10
0.02 T2

1/2
0.05
T4

0.00 0.00
-10 0 10 20 30 40 0
0 5 10 15
-VGS [V]
(a) IVGS-VTI [V]

Fig. 6. Comparison of the drain current dependence with gate voltage -4


7x10
if the present mobility model, (1), or the crystalline-like model is used.
6x10
-4 Transistor T1 type

-4

a
5x10

γ
of crystalline devices. In fact, other amorphous devices,

[V]
as for example a-Si:H and nanocrystalline Si TFTs, do 4x10
-4

µFET/µFETo
not show this dependence either. -4
3x10
OTFTs can present significant VT shifts at room tem-
-4
perature when a positive or negative gate voltage is ap- 2x10

plied. This situation must be considered when modeling -4


1x10
the current voltage characteristics. Since UMEM re-
quires measurements of transfer curves in linear and sat- 0
0 5 10 15 20 25 30 35
uration regions as well as the output characteristic, in (b) IVGS-VTI [V]
order to extract the device parameters, measurements
must be done so that gate and drain voltage variations Fig. 7. Relative behavior of mobility normalized to lFET0 as function
start from the same initial value in the three measured of the gate overvoltage. (a) For transistors T2, T3, and T4. (b) For
transistor T1.
curves, usually in the sense of increasing its absolute
value.
UMEM can also be used to determine VT shifts as dinal field as the one included in [5]. According to results
was done in [23]. presented, it was perfectly possible to model the OTFTs
UMEM also allows to extract subthreshold parame- behavior of all OTFTs analyzed, with the common
ters in a-Si:H, polycrystalline and Si nanocrystalline expression (2), through parameter k.
TFTs, [16–18]. In the case of OTFTs, surface and gate For the other three types of transistors, mobility in-
leakage currents can be significant, so subthreshold re- creases with VGS, resulting in positive values of ca. This
gion is usually determined by them and not by traps. mobility dependence with VGS of transistors T2 and T3
For this reason, it has not been possible up to now to type is similar to the observed for a-Si:H TFTs. Mobility
deal with this region. for T2 type transistors depends much less with VGS than
for T3 type transistors, which can be associated to a
3.2. Mobility behavior higher value of resistance, an order more. Transistor
T1 has much higher series resistance and lower effective
Regarding the mobility model, parameter ca is related mobility compared to T2 and T3. The effective mobility
to the dependence of mobility with VGS, while increases more rapidly with VGS, but remains at lower
lFET0 ¼ Vlc0a fits the modeled value of mobility to its mea- absolute values, see Fig. 7b. This is supposed to be ob-
aa
sured value at low gate voltages slightly higher than VT. tained because the pentacene layer in T1 was deposited
Fig. 7a shows the relative behavior of mobility normal- at lower temperature and directly on top of the dielec-
ized to lFET0 for transistors T2, T3 and T4 type. As can tric, which is expected to produce smaller grains.
be seen transistor T4 shows a mobility decreasing with With the above analysis of the four different types of
VGS, resulting in a negative value for parameter ca. As OTFTs, we have illustrated the applicability of UMEM
already indicated, the series resistance is very small for to modeling OTFTs, as well as its advantages in inter-
T4 type devices and VT is positive and quite large, which preting and comparing differences in their behavior
is not a typical behavior for these OTFTs. However, the through the extracted parameters, which in this case
model can represent well their output characteristics are not only fitting parameters. This advantage is not
without adding any other dependence with the longitu- trivial compared with other methods and is especially
1016 M. Estrada et al. / Solid-State Electronics 49 (2005) 1009–1016

important for modeling and analyzing OTFTs, where the tics of pentacene-based thin film transistors. Thin Solid Films
electrical characteristics of devices fabricated under dif- 2004;467:215.
[4] Arkhipov VI, Heremans P, Emelianova EV, Adriaenseens GJ,
ferent conditions and materials can be quite significant. Bässler H. Charge carrier mobility in doped disordered organic
semiconductors. J Non-Cryst Solids 2004;338–340:603.
[5] Silvaco International. Two dimensional ATLAS device simulation
4. Conclusion of pentacene organic thin film transistors. Simulation Standard
2003;12:1.
[6] Levinson J, Shepherd FR, Scanlon PJ, Westwood WD, Este G,
We presented the application of our unified model Rider M. Conductivity behavior in polycrystalline semiconductor
and parameter extraction method, (UMEM), to organic thin film transistors. J Appl Phys 1982;53:1193.
TFTs. The procedure for the extraction of the model [7] Verlaak S, Arkhipov V, Heremans P. Modeling of transport in
parameters was complemented to consider the presence polycrystalline organic semiconductor films. Appl Phys Lett
of non-ohmic contacts that frequently appear in these 2003;82:745.
[8] Brederlow R, Briole S, Klauk H, Halik M, Zschieschang U,
devices. It was demonstrated that UMEM provides a Schmid G, et al. Evaluation of the performance potential of
more precise modeling with simpler parameter extrac- organics TFT circuits. In: Proc. IEEE International Solid State
tion method for a wide variety of organic TFTs, fabri- Circuits Conference (ISSCC 2003), paper 21.6.
cated under different conditions and having different [9] Necliudov PV, Shur M. Contact resistance extraction in pentacene
materials and geometries. thin film transistors. Solid-State Electron 2003;47:259.
[10] Klauk H, Schmid G, Radlik W, Weber W, Zhou L, Sheraw CD,
In the analysis of four different types of OTFTs, we et al. Contact Resistance in organic thin film transistors. Solid-
also demonstrated its advantages in interpreting and State Electron 2003;47:297.
comparing differences in the behavior of the devices [11] Horowitz G, Delannoy P. An analytical model for organics based
through the extracted parameters. This advantage is thin film transistors. J Appl Phys 1991;70:469.
not trivial compared with other methods and is specially [12] Alam MA, Dodabalapur A, Pinto MR. A two dimensional
simulation of organic transistors. IEEE Trans Electron Dev
useful in characterizing OTFTs, where differences in the 1997;44:1332.
behavior of devices fabricated under different conditions [13] Necliudov PV, Shur M. Modeling of organic thin film transistors
and materials can be quite significant. of different designs. J Appl Phys 2000;88:6894.
According to our results, it is perfectly possible to [14] Street RA, Salleo A. Contact effects in polymer transistors. Appl
model the OTFTs current–voltage characteristics with Phys Lett 2002;81:2887.
[15] Cerdeira A, Estrada M, Garcia R, Ortiz Conde A, Garcia FJ.
the common expression (2), through parameter k. New procedure for the extraction of basic a-Si:H TFT model
parameters in the linear and saturation regions. Solid-State
Electron 2001;45:1077.
Acknowledgements [16] Reséndiz L, Estrada M, Cerdeira A. New procedure for the
extraction of a-Si:H TFTs model parameters in the subthreshold
region. Solid-State Electron 2003;47:1351.
We want to thank J. Deen, from McMaster Univer- [17] Estrada M, Cerdeira A, Ortiz Conde A, Garcia FJ, Iñiguez B.
sity and M. Halik and H. Klauk from Infineon Technol- Extraction method for polycrystalline TFT above and below
ogies for providing OTFTs measurements, as well as for threshold parameters. Solid-State Electron 2002;46:2295.
valuable discussions. [18] Cerdeira A, Estrada M, Iñiguez B, Pallares J, Marsal LF.
This work was supported by CONACYT project Modeling and parameter extraction procedure for nanocrystal-
lines TFTs. Solid-State Electron 2004;48:103.
39708 in Mexico, by CICYT of the Spanish Government [19] Reséndiz L, Iñiguez B, Estrada M, Cerdeira A. Modification of
under program TIC2002-04263 and TIC2002-04184, as amorphous level 15 AIM-Spice model to include new subthresh-
well as by the Generalitat de Catalunya in Spain under old model. In: Proc. 24th Conf. on Microelec. (Miel 2004), 2004.
program PICS2003-21. vol. 1. p. 593.
[20] Puigdollers J, Voz C, Orpella A, Quidant R, Martin I, Vetter M,
et al. Pentacene thin-film transistors with polymeric gate dielec-
tric. Org Electron 2004;5:67.
References [21] Klauk H, Halik M, Zschieschang U, Eder F, Schmid G, Dehm C.
Pentacene organic transistors and ring oscillators on glass and
[1] Bolognesi A, Di Carlo A, Lugli P. Influence of carrier mobility on flexible polymeric substrates. Appl Phys Lett 2003;82:
and contact barrier height on the electrical characteristics of 4175.
organic transistors. Appl Phys Lett 2002;81:4646–8. [22] Lin YY, Gundlach DJ, Nelson SF, Jackson TN. Pentacene-based
[2] Torsi L, Dodabalapur A, Katz HE. An analytical model for short organic thin-film transistors. IEEE Trans Electron Dev 1997;44:
channel organic thin film transistors. J Appl Phys 1995;78: 1325.
1088–93. [23] Gomes HL, Stallinga P, Leeuw DM, Muck T, Geurts J,
[3] Wang WY, Cheng HL, Wang YK, Hu TH, Ho JC, Lee CC, et al. Molenkamp W, et al. Bias induced threshold voltage shifts in
Influence of measuring environment on the electrical characteris- thin film organic transistors. Appl Phys Lett 2004;843184.