3.2.1 Electrical reliability analysis under gate bias stress
As shown in Fig.3-7, the result shows that devices with thickness10nm have the lowest variation whether on mobility, threshold voltage, or sub-threshold swing. For the discussion of the effect on different oxygen flow rate, we uses the a-IZTO TFTs with thickness10nm to have reliability analysis in vacuum in order to isolate from ambiance influence.
Fig.3-11(a),(b) shows the threshold voltage (Vth) shift of a-IZTO TFT devices under pos-itive gate bias stress (PGBS) and negative gate bias stress (NGBS) for the time duration of 2000 seconds, respectively. And Fig.3-12(a),(b) shows subthreshold swing (S.S.) shift of a-IZTO TFT devices under positive gate bias stress (PGBS) and negative gate bias stress (NGBS) for the time duration of 2000 seconds. The Vth shift of PGBS apparently increase with oxygen flow rate increases except the oxygen flow rate 0.1 sccm is 1.5V, a little bit smaller than 0 sccm about 1.6V.
The a-IZTO TFT devices with different oxygen flow rate don’t have significant change
20
in subthreshold swing (S.S.) shift under positive gate bias stress (PGBS) and negative gate bias stress (NGBS) as shown in Fig.3-12. The phenomenon means the shift of Vth is not ac-companied by the sub-threshold slope degradation, which indicates that the Vth shift in a-IZTO is attributed to the trapping of electrons in the interface or bulk dielectric layers with negligible creation of additional interface traps[19,20]. The trapped electrons partially screen the applied electric field so that the effective applied gate voltage is smaller when trapped electrons exist at the interface or bulk dielectric layers. When negative gate bias stress is ap-plied to the gate electrode, the transfer curve hardly moves from the initial one, which is due to the depletion of carriers in the active layer. To better understand the mechanism in the process of the charge trapping and detrapping through bias stress, we found that our results show that the time dependence of the Vth is in agreement with the stretched exponential equation, which has been developed to model Vth by the charge trapping mechanism in a-Si TFTs. [21]
The stretched-exponential equation for the Vth is defined as :
| | = | | { [ ( ) ]} (3-2)
Where
= G_S ress _ini ial (3-3)
and
τ= τ (kTEτ) (3-4)
21
is effective stress voltage
_ini ial is the initial threshold voltage τ is the characteristic trapping time of carriers β is the stretched exponential exponent
Eτ is the average effective energy barrier that electrons in the a-IZTO TFT channel need to overcome before they can enter the insulator. And τ is a significant parameter which can be
inferred as the time duration of electron needed to be trapped or detrapped from the defect, which is mean, when the τ is increase, the carriers need more time to be trapped by defects as
we can say the defects or trap states decrease, and vice versa. Fig.3-13 shows that the
loga-rithmic time dependence of Vth can be well fitted with the stretched-exponential equation, and the Table3-4 are the τ and correlative parameters. From Table3-4, device with oxygen flow rate 0.1 sccm has the largest τ, and it decease when oxygen flow rate increases. This re-sult is consistent with the previous experiment in trap state extraction, the τ of device with no
oxygen incorporation is smaller than of 0.1 sccm. It confirms that defects increase when oxy-gen flow rate increases, but the proper quantity of oxyoxy-gen can repair defects and optimize the characteristic.
22
3.2.2 The photosensitivity analysis
As shown in Fig.3-14(a), the 30nm film with different oxygen flow rate transparent through the visible spectrum. All the film transmittance is great than 80% from 400nm to 700nm. And from Tauc model as mentioned in eq.(2-7), Fig.3-14 (b) shows the relationships between (αhν)2 and the photon energy (hν), the optical band-gaps, Eg, can be obtained by ex-trapolating from the linear region onto the axis of photon energy throught the fitting curve.
And the Table3-5 shows the optical band-gaps for 30nm a-IZTO thin film with different oxy-gen flow rate extracted from Fig.3-14 (b). From the table, we observed that thin films do not have obvious difference approximately 3.7eV with different oxygen flow rate.
In order to study the instability under visible light illumination, for device different oxy-gen flow rate, the devices in 10nm a-IZTO expose in different wavelength illumination were investigated. Fig.3-15(a)~(e) show the corresponding transfer curves of devices with different
oxygen flow rate, respectively, which are subjected to various colored visible light illumina-tion (380nm~700nm) without applying bias. From Fig.3-15(a)~(e), we can find that the
de-vices with different oxygen flow rate are insensitivity to the visible light.
For the further investigation of instability of reality, we use the light source with full-band spectrum similar to flat panel back light illumination. Fig.3-16 shows the back light spectrum with intensity. Fig.3-17(a) shows the the threshold voltage (Vth) shift under light il-lumination for 2000 seconds without applying bias, and Fig.3-17(b) is recovery mesurement
23
at 1 hour, 3 hour, and 6 hour after the light illumination. From the figure, threshold voltage (Vth) of all devices shift negatively for small amount about 1.3V, and will recover for a period of time. This result shows, the light generated electron hole pairs, causing more conductive on active layer. As in Fig.3-16, the wavelengh in blue light has the highest intensity in this full band spectrum, even so, the threshold voltage (Vth) shows a mall amount of shift on a-IZTO TFT. It is clearly again to observe that the devices with different oxygen flow rate are insensi-tivity to the visible light.
Moreover, the negative bias illumination stress (NBIS) induced instability of the electrical properties was also investigated under the back light illumination. In the stress conduction, the gate voltage was fixed at the VG of -25V and the S/D electrodes were grounded. In Fig.3-18(a), voltage shift dosen’t have obvious tendency in different oxygen flow rate. The voltage shifts for five oxygen flow rates are about -15V which is much bigger than negative gate bias stress without illumination. The light induced electron hole pairs from deep level state, and free hole carriers were attracted to the interface by negative bias, as a result, elec-trons in channel result in the negative voltage shift. As we know from literature, electron hole pairs generated from illumination are come from deep level state [26], and this also means that oxygen incorporation doesn’t have significant influence in deep level trap state since the voltage shifts are about -15V and with no obvious tendency. Again with stretch exponential fitting shown in Fig.3-18(b) and Table 3-6, τ has no obvious tendency as well.
24
3.2.3 Summary
Through stretched exponential equation analysis under positive gate bias stress, the fit-ting result shows that defects increase when oxygen flow rate increases. And shifts of thresh-old voltage do not have obvious change in negative bias stress. The devices show insensitivity to the visible light, only under the full band spectrum with the highest intensity in blue light, the threshold voltage shifts negatively for about 1.3V. In NBIS, voltage shift dosen’t have obvious tendency in different oxygen flow rate, that means, oxygen incorporation doesn’t have significant influence in deep level trap state.