In this study, we report solution-processed amorphous zinc tin oxide transistors exhibiting high operational stability under positive gate bias stress, translated by a recoverable threshold voltage shift of about 20% of total applied stress voltage. Under vacuum condition, the threshold voltage shift saturates showing that the gate-bias stress is limited by trap exhaustion or balance between trap filling and emptying mechanism. In ambient atmosphere, the threshold voltage shift no longer saturates, stability is degraded and the recovering process is impeded. We suggest that the trapping time during the stress and detrapping time in recovering are affected by oxygen adsorption/desorption processes. The time constants extracted from stretched exponential fitting curves are ≈106 s and 105 s in vacuum and air, respectively.

Solution based amorphous oxide semiconductors (AOSs) provide an innovative and low-cost processing technology for thin-film transistors (TFTs).1–4 In order to exploit them into practical circuits, it is important to assure that their electrical performance is stable under the required working conditions. One of the sources of instability is related with the threshold voltage shift (Δ Vth) when the TFT is operated under constant gate voltage. The phenomena known as gate-bias-stress are common to all TFT technologies, including amorphous oxide TFTs.5–7 This instability is related with surface states of the dielectric which trap charge carriers and shield the externally applied gate-voltage. In the particular case of oxide semiconductors, oxygen adsorption/desorption processes are also enhanced by an externally applied electric field, contributing to increase instability.8 In spite of all efforts to mitigate this effect, the problem has not been yet properly solved. In fact, even for high quality thermal SiO2 typically used in test structures for performance evaluation of thin film semiconductors, it is not possible to assure an entirely perfect interface between dielectric and semiconductor. Previously, amorphous oxides using SiO2 as dielectric suffer from a pronounced Δ Vth, with time constants (τ) of 104-105 s.6,7,9 Here, we show evidences of intrinsically more stable solution-based zinc tin oxide (ZTO) TFTs from studying the time and temperature dependence of bias stress and recovering process. We propose that this enhanced stability is promoted by a defect passivation or neutralization of the SiO2 surface and consequently reducing the number of traps. We suggest that this passivation is brought out during the thin film fabrication process of the ZTO solution.

The TFTs are produced in a staggered bottom-gate, top-contact structure. Zinc nitrate and tin chloride-based precursors are used to prepare a ZTO solution which is then spin coated on top of a 100 nm thick thermal SiO2 on Si substrate. The film is annealed at 350 °C in air during 30 min. Next, Al source and drain electrodes (70 nm thick) are e-beam evaporated on top of the annealed ZTO film with typical channel width (W) and length (L) of 1400 μm and 100 μm, respectively (W/L = 14). Patterning is obtained via shadow masks. Figure 1 shows the typical output characteristic of the device. The schematic of device structure is shown in the inset of this figure. For the mobility μn, the value of the field-effect mobility at Vds = 1 V is chosen. The subthreshold swing (SS) is taken as the minimum value of d(log(Ids)/(Vgs))−1. Typical values for the as-fabricated devices are Vth = − 3.1 V, SS = 0.35 V/dec, and μn = 2.5 cm2/V s at room temperature.

During the positive gate-bias stress (PBS) experiment, all TFTs are biased with a constant gate voltage while source and drain are grounded. After a prefixed time, the bias stress is interrupted and the gate voltage (Vgs) is swept at Vds = 1 V to measure the Ids-Vgs characteristics of TFT. Stress measurements are performed in air and under vacuum (10−5 millibars, leaving the device at this pressure for 1 h prior starting the stress experiment) under dark condition. To overcome the problem of non-linearity of transfer characteristics, Vth is taken as the voltage corresponding to Ids = 1 nA. PBS displaces the transfer curves to the positive direction. The ΔVth occurs toward the applied stress voltage and is continuously shutting down the drain-source current. Figure 2(a) shows the shift of the transfer curves under constant Vgs = 5 V in vacuum. Interestingly, PBS in vacuum results into threshold voltage displacement during 8 h with ΔVth saturating after ≈1 V shift over this period. The time dependence of ΔVth can be well fitted by a stretched-exponential equation as below,

Δ V th = V 0 1 exp t τ β ,

where t is the stress time, V0 = |ΔVth| at the infinite time, β is the dispersion parameter which is related to the degree of barrier energy dispersion, and τ is the time constant. Table I shows all the fitting parameters values in different PBS conditions.

In order to get further insight into the physical mechanism of the gate-bias stress, the kinetics of the threshold voltage recovering process is also monitored while the device is kept unbiased under dark conditions. The stress and recovering times are compared in Fig. 2(b). The recovering time is fast <1 h, and Vth shift is linearly proportional to the logarithmic time, as observed in the inset of Fig. 2(b) supporting charge detrapping as the dominant mechanism. The fast device recovery is technologically relevant and to the best of our knowledge, has not been reported so far. The observed behavior can be interpreted in terms of a trap model. The fact that ΔVth reaches saturation shows that the number of states or defects available is limited. During stress time, the trap filling rate is dominated by the number of defects or available sites (Nt). ΔVth is then given as ΔVth = eNt/Cox, where Cox is the capacitance of the gate dielectric and e is the elementary charge. Throughout the stress measurement, more defect states are filled and consequently there are fewer states available; therefore at a long enough stress time, all defect states will be eventually filled leading to threshold voltage saturation. It is difficult to assess if the saturation of the threshold voltage is caused by trap exhaustion or alternatively by a competing trap filling/emptying mechanism. At long stress times, the backward (trap emptying) reaction eventually becomes faster than the forward reaction (trap filling). This stops further change in threshold voltage after a certain period of time. As a result, instantaneous threshold voltage shift becomes virtually zero.

Figure 3 shows the temperature dependence of the threshold voltage shift within the stress time. The time dependence of Δ Vth under stress is fitted by a stretched-exponential equation. The inset of Fig. 3 shows that τ is thermally activated by fitting the data to the following equation:

τ = τ 0 exp ( E τ / k T ) ,

where T is the absolute temperature, k is the Boltzmann constant, and Eτ is the activation energy. The obtained activation energy is 0.64 eV. This value has been already reported for TFTs based on other semiconductor technologies using SiO2 as a gate dielectric, supporting the view that this defect is not intrinsic to ZTO but instead related with the SiO2 surface.7,9

Under the presence of ambient atmosphere (295 K, 1 bar) during PBS process, the ZTO-TFT stability is severely degraded. However, the presence of oxygen does not affect the negative bias stress (NBS) instability. The fitting parameters β and τ are approximately equal for NBS measurements under vacuum and in the presence of atmosphere as shown in Table I.

Fig. 4(a) shows the typical electrical characteristics of the device under a gate-bias stress of 5 V in ambient atmosphere. The time dependence of ΔVth associated with fitting plots is shown in Fig. 4(b). The time constant (τ = 105 s) is now one order of magnitude lower than the one measured in vacuum (τ = 106 s). Higher value of τ indicates that at a particular temperature, the course of threshold voltage shift persists longer and the operational stability is enhanced. In this respect, τ can be used as a figure of merit to quantify operational stability. The value of β (β = 0.52) is higher in vacuum than in air (β = 0.36). This means that the distribution of energy barriers or trap sites becomes broader when oxygen adsorption occurs. Furthermore, the Vth does not saturate, suggesting that more defect states become available in the presence of atmosphere. Most relevant, the ability of the device to recover the original Vth is severely impeded, being the recovering time now more than 106 s. This time contrasts with the behavior in vacuum where a few minutes under unbiased conditions are enough to fully recover the original Vth. The recovering fitting shows the time constant of 5.5 × 104 s. This value is one order of magnitude smaller than trapping time during PBS in air. It supports that the recovering process is not only charge detrapping process but also the oxygen ion desorption from the channel surface contributes in the recovering mechanism. Nevertheless, the β is the same for stress/recovery process indicating the similar barrier energy dispersion of trap sites.

The fact that the threshold voltage can shift faster in air than in vacuum has been reported by a number of authors and tentatively explained by several mechanisms.10,11 The new finding reported here is that for solution based ZTO, the presence of atmospheric species impedes the fast recovering of the Vth observed under vacuum conditions. In order to explain this behavior, we propose that oxygen species adsorb to the ZTO back channel exposed surface. The effect of external electric field which can induce the chemisorption of oxygen on zinc oxide thin film has been already reported in 1968. A field which brings electrons to the surface would favor increased oxygen adsorption.8 Oxygen is a weak electron acceptor and removes free electrons from the bulk ZTO leading to overall reduction in the free carrier concentration. Therefore, it results in a depletion layer on the surface and increase of threshold voltage. During the recovering time when the device is kept unbiased, it is difficult to remove the oxygen molecules without any driving force. Therefore, the Vth hardly recovers back to its initial value. Two instability mechanisms simultaneously occur in TFT under gate bias stress experiment in air: (i) surface dielectric states trap electrons and shield the external gate bias, and (ii) adsorbed oxygen layer on the top exposed ZTO surface captures free carriers from ZTO. The two effects add to each other contributing to a further increase in threshold voltage shift in air.

In summary, we have shown that solution based ZTO-TFTs are highly stable in vacuum. The threshold voltage shift saturates in a time scale of 104 s. Furthermore, when the TFTs are kept unbiased, they exhibit a remarkable fast recovering behavior of the threshold voltage shift. We attributed this high operational stability to a passivation/neutralization of SiO2 surface defects occurring during the drying/annealing of the ZTO solution. Saturation of the threshold voltage shift may arise because the number of traps is finite, or alternatively because the internal field caused by the immobile trapped carriers will lead to a balance between trap filling and emptying. The presence of atmospheric species eliminates the ability of the stressed device to restore the original threshold voltage. We propose that a negatively charged oxygen layer is depleting the ZTO bulk from free carriers and impedes the Vth to be restored to original values. Encapsulation of the devices should prevent the interference of oxygen species and diminish this effect. The deposition of semiconductor layers under vacuum should in principle lead to a better control of impurity densities, interestingly the results reported here suggest that solution based amorphous oxides can intrinsically passivate the dielectric surface and lead to highly stable devices.

This work is funded by FEDER funds through the COMPETE 2020 Programme and National Funds through FCT–Portuguese Foundation for Science and Technology under the Project Nos. UID/CTM/50025/2013 and EXCL/CTM-NAN/0201/2012 and the European Communities 7th Framework Programme under grant agreement ICT-2013-10-611070 (i-FLEXIS project).

1.
R.
Branquinho
,
D.
Salgueiro
,
A.
Santa
,
A.
Kiazadeh
,
P.
Barquinha
,
L.
Pereira
,
R.
Martins
, and
E.
Fortunato
,
Semicond. Sci. Technol.
30
,
024007
(
2015
).
2.
L.
Jun seok
,
K.
Young-Jin
, and
C.
Woon-Seop
,
J. Korean Phys. Soc.
59
,
3055
(
2011
).
3.
S.-J.
Seo
,
C. G.
Choi
,
Y. H.
Hwang
, and
B.-S.
Bae
,
J. Phys. D: Appl. Phys.
42
,
035106
(
2009
).
4.
R.
Branquinho
,
D.
Salgueiro
,
L.
Santos
,
P.
Barquinha
,
L.
Pereira
,
R.
Martins
, and
E.
Fortunato
,
ACS Appl. Mater. Interfaces
6
,
19592
(
2014
).
5.
Y.
Jeong
,
C.
Bae
,
D.
Kim
,
K.
Song
,
K.
Woo
,
H.
Shin
,
G.
Cao
, and
J.
Moon
,
ACS Appl. Mater. Interfaces
2
,
611
(
2010
).
6.
J. S.
Park
,
W.-J.
Maeng
,
H.-S.
Kim
, and
J.-S.
Park
,
Thin Solid Films
520
,
1679
(
2012
).
7.
J. M.
Lee
,
I. T.
Cho
,
J. H.
Lee
, and
H. I.
Kwon
,
Appl. Phys. Lett.
93
,
093504
(
2008
).
8.
S. A.
Hoenig
and
J. R.
Lane
,
Surf. Sci.
11
,
163
(
1968
).
9.
M. E.
Lopes
,
H. L.
Gomes
,
M. C. R.
Medeiros
,
P.
Barquinha
,
L.
Pereira
,
E.
Fortunato
,
R.
Martins
, and
I.
Ferreira
,
Appl. Phys. Lett.
95
,
063502
(
2009
).
10.
D.
Kang
,
H.
Lim
,
C.
Kim
,
I.
Song
,
J.
Park
,
Y.
Park
, and
J.
Chung
,
Appl. Phys. Lett.
90
,
192101
(
2007
).
11.
J. S.
Park
,
J. K.
Jeong
,
H. J.
Chung
,
Y. G.
Mo
, and
H. D.
Kim
,
Appl. Phys. Lett.
92
,
072104
(
2008
).