Operation mechanisms and electrical characteristics of tunneling field-effect transistors (TFETs) employing a hetero tunneling junction by utilizing an n-type oxide-semiconductor (OS) and a p-type group-IV-semiconductor are comprehensibly analyzed. Gate-normal band-to-band tunneling (BTBT) has high potential for the superior TFET performance such as high on-state current and small sub-threshold swing (S.S.). Additionally, a hetero tunneling junction with type-II energy band alignment is promising to exponentially increase tunneling probability with keeping small off-state current. Therefore, in this study, we investigate the impact of key material and device parameters such as energy band alignment of source/channel regions and thickness of the OS channel layer or gate insulator based on technology computer aided design (TCAD) simulation. The gate-controlled uniform band bending along the source-drain direction realizes uniform BTBT in the entire region of the hetero tunneling junction. Also, the reduction of the tunneling barrier height, which is continuously controlled by the conduction band minimum of the OS-channel and the valence band maximum of the IV-source, is effective to increases on-state current and decrease S.S. value. On the other hand, the thicknesses of OS channel layer and gate insulator have strong influences on tunneling probability and threshold voltage. Therefore, the sub-threshold characteristics of TFETs are sensitive to non-uniformities in the tunneling junction such as channel thickness fluctuation and surface potential fluctuation at the metal-oxide-semiconductor (MOS) interfaces. These numerical analyses of the device operation are essentially important to understand the effects of key device parameters on the TFET performance and to realize the superior electrical performance.

A bilayer tunneling field effect transistor (TFET) with gate-normal band-to-band tunneling (BTBT) is one of ideal device structures to realize extremely-small subthreshold swing (S.S.) because overlap of the density of states (DOS) between the source and the channel, inducing the BTBT current, can be directly modulated over the entire region of the tunneling junction by the gate voltage (Vg).1–5 Additionally, the large on-state current (Ion) is expected in the bilayer TFETs thanks to the larger tunneling area and the shorter tunneling distance which is controlled by the thickness of the upper channel layer. On the other hand, type-II energy band alignment, which has the higher valence band maximum in the source region and the lower conduction band minimum in the channel region for n-channel TFETs, is promising to enhance the BTBT probability because of reduction in the effective barrier height (Eb-eff) and the resulting shorter tunneling distance.5–10 As a result, a combination of these two approaches, bilayer TFETs composed of source and channel materials with type-II energy band alignment, is one of the most effective structures for satisfying both high performance and extremely-steep S.S. In reality, however, there are several difficult challenges for fabricating these devices because of the structural complexity.3,4 Also, in drawing out the tunneling current from the tunneling junction to the drain contact, non-ideal leakage currents near the drain edge induced by the concentration of the electrical field tend to significantly deteriorate the off currents (Ioff) and the S.S. properties.11,12

In order to overcome these obstacles, we have recently proposed a new bilayer TFET structure by utilizing a hetero tunneling junction composed of an n-type oxide-semiconductor (n-OS) and a p-type group-IV-semiconductor (p-IV).13Figure 1(a) illustrates the concept of this bilayer TFET structure and its energy band diagram normal to the gate/high-k/n-OS/p-IV tunneling junction under (b) off-state and (c) on-state. When positive voltage (Vg) is applied to the gate electrode formed on the n-OS side, DOS of the valence band in the p-IV source region is overlapped with that of conduction band in the n-OS channel region overlaps, leading to the generation of BTBT at the tunneling junction. This proposed tunneling junction has several attractive features as follows; (I) The proposed material combination allows us to provide the type-II energy band structure with the small effective barrier height (Eb-eff = Ec-OSEv-IV, where Ec-OS is the conduction band minimum of n-OS and Ev-IV is the valence band maximum of p-IV) under the semiconductors with relatively large bandgap (Eg), which keep small Ioff. Furthermore, this Eb-eff value is tunable by changing the material combination and these compositions, as shown in Fig. 1(d).14–16 (II) OS materials having the conduction band minimum at the Γ point allows direct tunneling, which is in contrast with typical group-IV semiconductors which do not have the conduction band minimum at the Γ point.17 This tunneling path is effective for realizing high tunneling probability.18,19 (III) The nature of n-type carrier conduction in OS layers can realize the junction-less channel structure, where additional n-type doping is not needed in the gate-underlap region. This junction-less structures as well as the large bandgap of OS can effectively suppress any non-ideal leakage current near the drain edge. (IV) The simple device structure and the fabrication processes by utilizing well-developed deposition techniques of OS layers can lead to high compatibility to the CMOS platform with a large-size Si substrate. As a result, we can expect to realize high performance bilayer TFETs with high Ion and small S.S. on the Si CMOS platform.

FIG. 1.

(a) Device structure image of an n-OS/p-IV bilayer TFET and its energy band diagrams in (b) off state and (c) on state, respectively. (d) Material candidates of p-IV source and n-OS channel.

FIG. 1.

(a) Device structure image of an n-OS/p-IV bilayer TFET and its energy band diagrams in (b) off state and (c) on state, respectively. (d) Material candidates of p-IV source and n-OS channel.

Close modal

Therefore, in this study, we investigate the potential of the proposed bilayer TFET with type-II hetero tunneling junction as a high-performance TFET based on technology computer aided design (TCAD) simulation.20 Here, an n-ZnO/p-Ge bilayer TFET is mainly examined because this structure can provide smaller Eb-eff by utilizing a realistic material combination. In particular, ZnO or ZnO-based oxide semiconductors have been well-studied. The high electron mobilities have already been reported everywhere21–25 including our previous work,26 even in the poly-crystalline layers. Also, we examine influences of principal device parameters such as Eb-eff by using a SiGe source and other OS materials, and channel layer thickness on the electrical performance of the proposed bilayer TFET for clear guideline of the device design. Then, we investigate the influences of the non-uniformities of the tunneling junction such as the channel thickness and surface potential at the high-k/OS interface, which could be other key factors for the actual device fabrications with drawing out the attractive potential of the proposed TFET.

First, the ideal performance of the proposed n-OS/p-IV bilayer TFET is investigated by using TCAD simulation based on Synopsys Sentaurus simulator with non-local BTBT model.17Figures 2(a) and 2(b) show the device structure used in the simulation and the simulated 2-dimensional (2D) distribution of the BTBT rate near the n-OS/p-IV hetero tunneling junction, respectively. The red color in Fig. 2(b) indicates the amount of the BTBT rate in logarithmic scale. Here, as the material parameters of n-OS, those of bulk ZnO such as the relative permittivity of 8.8,27–29 the electron affinity of 4.2 eV13,30 and the electron/hole effective masses of 0.28m0/0.59 m0,27,28,31 where m0 is mass of a free electron, were used. The impurity concentrations in both n-type channel and p-type source regions were taken to be 3×1018 cm–3. It is clearly seen that vertical tunneling from near the surface of the p-IV source to the top surface of the n-OS channel, that is high-k/n-OS interface, is uniformly generated by applying positive Vg. Figure 2(c), 2(d) and 2(e) show the energy band diagrams of n-Ge/p-Ge homo, n-ZnO/p-Ge hetero and n-ZnO/p-Si hetero tunneling junctions, respectively, in off- and on-states. Here, we assume the applied Vg values in the off- and the on-state as –5 mV and +0.5 V, respectively, from Voff, defined as Vg at Id of 0.1 pA/μm in simulation. It can be seen that band bending in ZnO, Ge, and Si are different because of the difference of the permittivity. For example, in the n-ZnO/p-Ge hetero junction, the amount of band bending in the n-ZnO channel is much larger than that in the p-Ge source, owing to the lower permittivity of ZnO than Ge. In the n-ZnO/p-Si hetero junction, the difference in the amount of band bending between the n-ZnO channel and the p-Si source is smaller because of the smaller difference in the permittivity between ZnO and Si, though that of n-ZnO is still larger than that of p-Si. Additionally, the energy band of the n-ZnO/p-Ge hetero junction is almost flat under its off-state condition thanks to the small Eb-eff. On the other hand, the energy band of the n-ZnO/p-Si hetero junction is already bent even in its off-state because the n-ZnO/p-Si junction forms a long depletion region in the p-Si source due to the large Eb-eff between ZnO and Si. The large permittivity difference and the small Eb-eff in the n-ZnO/p-Ge is effective to reduce the tunneling distance and resulting increase in the BTBT probability.

FIG. 2.

(a) 3D image of device structure for TCAD simulation, (b) 2D distribution of BTBT rate from p-IV source to n-OS channel surface, and energy band diagrams of (c) n-Ge/p-Ge homo, (d) n-ZnO/p-Ge hetero, and (e) n-ZnO/p-Si hetero tunneling junctions, respectively. Impurity concentrations in both source and channel regions are 3×1018 cm−3, Ec-OS is 4.2 eV, and relative permittivities of ZnO, Ge, and Si are 8.8, 16.2, and 11.2, respectively.

FIG. 2.

(a) 3D image of device structure for TCAD simulation, (b) 2D distribution of BTBT rate from p-IV source to n-OS channel surface, and energy band diagrams of (c) n-Ge/p-Ge homo, (d) n-ZnO/p-Ge hetero, and (e) n-ZnO/p-Si hetero tunneling junctions, respectively. Impurity concentrations in both source and channel regions are 3×1018 cm−3, Ec-OS is 4.2 eV, and relative permittivities of ZnO, Ge, and Si are 8.8, 16.2, and 11.2, respectively.

Close modal

Figure 3(a) shows Id-Vg characteristics of these three TFETs with the lateral junction length (Lg) of 100 nm and equivalent oxide thickness (EOT) of 1 nm. It is found that the n-ZnO/p-Ge hetero-junction can enhance Ion by more than 2 orders than the n-Ge/p-Ge homo-junction thanks to the smaller Eb-eff and the shorter tunneling distance, as explained in the energy band diagram in Fig. 2. On the other hand, Ion of n-ZnO/p-Si TFET is not sufficiently high, although it works as a bilayer TFET as predicted from Fig. 2. It is found that the on/off switching of the bilayer TFETs is extremely steep and that the minimum S.S. value is as small as 1 mV/dec. Also, the remarkably large Ion/Ioff current ratio more than 8 orders is obtained. It is also observed, on the other hand, that the change in the Id-Vg characteristics with applied Vd from 50 to 350 mV is largely different between the n-ZnO/p-Ge and n-ZnO/p-Ge devices. Figures 3(b) and 3(c) show Id-Vd characteristics of the n-ZnO/p-Ge and the n-ZnO/p-Si TFET, respectively, and it is clearly observed that Id-Vd characteristics of these TFETs are different each other. The typical linear and saturation regions are observed in the Ge TFET for small Vg, although Id tends not to saturate with increasing Vd particularly for larger Vg. Furthermore, the Id value as high as 100 μA/μm can be expected. In the Si device, on the other hand, the Id-Vd curves show the Schottky-like non-ohmic characteristics.

FIG. 3.

(a) Id-Vg characteristics of n-ZnO/p-Ge and n-ZnO/p-Si bilayer TFETs, and Id-Vd characteristics of (b) n-ZnO/p-Ge and (c) n-ZnO/p-Si bilayer TFETs, respectively. Impurity concentrations in both source and channel regions are 3×1018 cm−3, Ec-OS is 4.2 eV. OS channel thickness and EOT are set to be 5 and 1 nm, respectively.

FIG. 3.

(a) Id-Vg characteristics of n-ZnO/p-Ge and n-ZnO/p-Si bilayer TFETs, and Id-Vd characteristics of (b) n-ZnO/p-Ge and (c) n-ZnO/p-Si bilayer TFETs, respectively. Impurity concentrations in both source and channel regions are 3×1018 cm−3, Ec-OS is 4.2 eV. OS channel thickness and EOT are set to be 5 and 1 nm, respectively.

Close modal

Here, the Schottky-like non-ohmic Id-Vd characteristic is often observed for various types of TFETs and some possible origins for this characteristic have been still discussed.32 Furthermore, in the bilayer TFET structure, uniformity of the BTBT probability over the channel under the gate electrode might be degraded with non-uniform band bending along the source-drain direction after applying Vd. Therefore, the profiles of Ec-OS along the source-drain direction is simulated, in order to examine the uniformity of the energy band bending with Vg and Vd, which directly changes the BTBT rate. Figures 4(a) and 4(b) show the profiles of Ec-OS at the most ZnO channel surface (∼0.2 nm away from the high-k/ZnO interface) for the ZnO/Ge and ZnO/Si TFETs, respectively. Also, Figs. 4(c) and 4(d) show the 2D distributions of BTBT rate with low Vd of 50 mV and high Vd of 300 mV, respectively. Here, Vg swing is defined as Vg - Voff. It is clearly observed that Ec-OS is almost uniform along the source-to-drain direction in any bias conditions, except for in the gate edge regions. Additionally, it can be clearly confirmed that uniformity of BTBT at the tunneling junction along the source-to-drain direction is not degraded even with high Vd of 300 mV. On the other hand, Ec-OS at the ZnO channel surface is largely modulated by Vd, which increases the potential drop at the high-k/ZnO interface and the resulting BTBT rate. This is because the electrical conduction in the ZnO channel is much higher than that through the tunneling junction, which can be an origin of the continuous increase in Id with Vd, shown in Figs. 3(b) and 3(c).

FIG. 4.

Profiles of Ec-OS at the most surface of ZnO channel in (a) ZnO/Ge and (b) ZnO/Si TFETs along source-to-drain direction. 2D distributions of electron band-to-band tunneling rate with (c) low Vd of 50 mV and (d) high Vd of 300 mV, respectively. Vg swing is defined as VgVoff, which is Vg at Id of 0.1 pA/μm.

FIG. 4.

Profiles of Ec-OS at the most surface of ZnO channel in (a) ZnO/Ge and (b) ZnO/Si TFETs along source-to-drain direction. 2D distributions of electron band-to-band tunneling rate with (c) low Vd of 50 mV and (d) high Vd of 300 mV, respectively. Vg swing is defined as VgVoff, which is Vg at Id of 0.1 pA/μm.

Close modal

Also, the change in Ec-OS and the increase in the BTBT rate with increasing Vd are more significant for the ZnO/Si TFET than for the ZnO/Ge TFET. The remarkably large moderation of the energy band of the ZnO/Si TFET caused with Vd (not with Vg) can lead to the non-ohmic Id-Vd characteristic. Here, when we compare the Id-Vd characteristics between ZnO/Ge and ZnO/Si TFETs under same Eb-eff given by the intentional adjustment of the Ec-OS position, the Schottky-like Id-Vd characteristic has been still observed in the ZnO/Si TFET (not shown). This result indicates that the difference in the Id-Vd characteristics between ZnO/Ge and ZnO/Si TFETs is attributable to that in the material parameters such as permittivity between p-Si and p-Ge, which can modulate the voltages across the ZnO channel and the source region under a give Vg value.

The tunability of the band lineup of the n-OS and p-IV hetero tunneling junction is one of the advantages of the present TFET structures. Here, the Ec-OS position can be changed by the nature of the n-OS materials and the Ev-IV position can be continuingly optimized by the Ge fraction in the SiGe source. Ion is calculated as a function of the Ge fraction in the SiGe source for various Ec-OS values in order to understand the impact of Eb-eff on Ion. The results are shown in Fig. 5(a). Here, Ion is defined as the Id value at Vg = Voff + 0.3 V, where Ioff is fixed to be 1 pA/μm. The material parameters of ZnO are still used for those of the OS channel except Ec-OS, which is intentionally changed from 4.0 to 4.8 eV. The Ion values exponentially increase with decreasing Ec-OS and increasing the Ge fraction. In particular, it is found that a slight change of Ev-IV by using Si0.7Ge0.3, commonly used in advanced Si technologies, significantly enhances Ion in comparison with Si. Also, Ion tends to saturate in high Ge content and large Ec-OS regions. This saturation is attributable to the parasitic resistance of the gate-underlapped OS region between the tunneling junction and the drain electrode. Figures 5(b) and 5(c) show minimum S.S. near Id of 1 pA/μm and average S.S. over the Vg swing of 0.3 V, respectively, as a function of the Ge fraction for various Ec-OS values. Reduction of Eb-eff also improves the sub-threshold properties. The minimum and the average S.S. values of ∼1 mV/dec. and ∼40 mV/dec., respectively, can be obtained for small Eb-eff.

FIG. 5.

Impacts of Eb-eff on performance of n-ZnO/p-SiGe bilayer TFETs; (a) Ion at Vg swing of 0.3 V, (b) minimum S.S. at Id of 1 pA/μm, and (c) average S.S. in Vg swing of 0.3 V as a function of Ge fraction in SiGe source for various Ec-OS cases.

FIG. 5.

Impacts of Eb-eff on performance of n-ZnO/p-SiGe bilayer TFETs; (a) Ion at Vg swing of 0.3 V, (b) minimum S.S. at Id of 1 pA/μm, and (c) average S.S. in Vg swing of 0.3 V as a function of Ge fraction in SiGe source for various Ec-OS cases.

Close modal

The choice of the thickness of the n-OS channel layer (tOS) is also an important device design issue, because it directly changes the tunneling distance and the resulting tunneling probability. Figure 6(a) and 6(b) show the change in the Id-Vg characteristics of the n-ZnO/p-Ge bilayer TFET with changing tOS and the energy band profiles in the tunneling junction at a certain Vg of 0.6 V, respectively. Here, Ge is selected as the source material in this calculation to examine the influences of tOS on high-performance devices with small Eb-eff. EOT and the channel length are taken to be 4 nm and 50 nm, respectively. Also, the carrier concentrations in the n-ZnO channel and the p-Ge source are assumed to be 3×1018 and 5×1018 cm−3, respectively. The shift of off-state voltage (Voff), defined as Vg at Id of 0.1 pA/μm here, and the change in Ion are clearly observed with changing tOS. Figures 7(a) and 7(b) summarize Voff and Ion at Vg swing of 0.3 V, respectively, as a function of tOS for various EOT values. The Voff values shift toward a negative Vg side with increasing tOS and this shift is more significant in thicker EOT. The threshold voltage for starting to generate BTBT in the TFET is determined by the energy position of Ec-OS at the ZnO channel surface, independent of tOS. This fact means that the amount of the total potential drop across the Ge source and the ZnO channel under the threshold condition does not change with tOS, although the electrical field across the ZnO/Ge tunneling junction decreases with increasing tOS. At the same time, the electrical field through the gate insulator is also reduced with increasing tOS. As a result, the Voff value decreases. On the other hand, Ion exponentially increases with thinner tOS because of the shorter tunneling distance, while the EOT dependence of Ion is quite small. It is also found that S.S. is not degraded obviously by thicker or thinner channels. Here, the similar tOS dependence has also been confirmed in n-OS/p-Si TFET (not shown).

FIG. 6.

Change in (a) Id-Vg characteristics and (b) energy band diagrams of n-ZnO/p-Ge TFET with ZnO channel thickness. Impurity concentrations in ZnO channel and Ge source are 3×1018 and 5×1018 cm−3, respectively.

FIG. 6.

Change in (a) Id-Vg characteristics and (b) energy band diagrams of n-ZnO/p-Ge TFET with ZnO channel thickness. Impurity concentrations in ZnO channel and Ge source are 3×1018 and 5×1018 cm−3, respectively.

Close modal
FIG. 7.

Impact of OS channel thickness on (a) Voff and (b) Ion for various EOT cases.

FIG. 7.

Impact of OS channel thickness on (a) Voff and (b) Ion for various EOT cases.

Close modal

As a result, it has been found from the TCAD simulation results that the bilayer TFETs composed of the n-OS/p-IV hetero-junction with the type-II energy band alignment have high potential for steep on-off switching with small Vg swing. In particular, the band profile at the hetero tunneling junction along the source-to-drain direction can be uniformly controlled by Vg, which is a promising feature of the bilayer TFETs to realize small S.S. Also, the bilayer TFET with Eb-eff typically lower than 0.4 V, which can be obtained by the Ge fraction higher than 70% and/or Ec-OS larger than 4.3 eV, are effective to achieve high Ion of ∼100 μA/μm, the Ion/Ioff ratio higher than 8 orders and the small average S.S. of ∼40 mV/dec. under the supply voltage less than 0.3 V.

We have recently succeeded the experimental demonstration of the operation of the present bilayer TFETs fabricated by using an n-ZnO channel and a p-Si or a p-Ge source.13 However, the experimentally achieved S.S. value is much higher than the values predicted by TCAD simulation. Here, we have also observed the non-uniform channel thickness from cross sectional transmission electron microscope (TEM) images of the TiN/Al2O3/ZnO/Si tunneling junction after the crystallization of the ZnO layer.13,26 Thus, this thickness fluctuation might be one of possible origins of the poor subthreshold properties of the experimental results. On the other hand, the stretch out of the Id-Vg characteristics could also be caused by charge trapping/de-trapping related to interface states at the Al2O3/ZnO interface. Additionally, it is known that interface states created at oxide/semiconductor interfaces can cause the fluctuation of the potential at the semiconductor surfaces.33 If the channel thickness and the surface potential are spatially fluctuated across a tunneling junction area, the energy band diagram must have the non-uniformity over the tunneling junction area, affecting the sub-threshold properties. Therefore, we investigate the influences of the channel thickness fluctuation (CTF), interface states and the surface potential fluctuation (SPF) at the high-k/OS interface on the degradation of the sub-threshold characteristics of the bilayer TFETs.

As found in Fig. 6, the threshold voltage of the n-OS/p-IV bilayer TFETs largely shifts with changing tOS, while the sub-threshold characteristics are not degraded. However, if tOS is fluctuated over a tunneling junction region of one device, the total Id-Vg curves becomes less steep because of the summation of the multiple Id-Vg curves with different threshold voltage. Such a degradation of the sub-threshold slope is schematically illustrated in Fig. 8(a). Actually, it has been observed from cross-sectional TEM images of the fabricated device that the thickness of the ZnO channel layer is not uniform.13Figure 8(b) shows the height histogram of the ZnO surface reproduced from an atomic force microscope (AFM) image of a ZnO/Si hetero-structure before the gate oxide deposition. Since the ZnO/Si interface is much smoother than the ZnO layer surface, this roughness directly corresponds to the fluctuation of the thickness of the ZnO channel layer. The main portion of the histogram can be well fitted by the Gauss distribution with a standard deviation of (σt) of 0.58 nm, while the distribution has an additional tail in the larger height (thickness) side.

FIG. 8.

(a) Schematic illustration of degradation of TFET Id-Vg characteristic caused with CTF. (b) Experimentally observed height histogram of ZnO channel surface reproduced from AFM image of ZnO/Si structure before the gate oxide deposition.

FIG. 8.

(a) Schematic illustration of degradation of TFET Id-Vg characteristic caused with CTF. (b) Experimentally observed height histogram of ZnO channel surface reproduced from AFM image of ZnO/Si structure before the gate oxide deposition.

Close modal

Figure 9 shows the calculated Id-Vg characteristics of a ZnO/Ge TFET with the OS channel thickness fluctuation. The Id value with the OS thickness fluctuation (Id-TF) is calculated as follows,

IdTFVg=idtOS,VgPtOSdtOS
(1)

where, id(tOS, Vg) is the drain current of a TFET with uniform tOS, and P(tOS) is the probability of each tOS. Here, P(tOS) is assumed to be the height histogram obtained from the AFM result of Fig. 8(d). It is clearly found that the sub-threshold characteristics of the bilayer TFET is degraded by CTF. It is also observed from the comparison between the two CTF distributions, the fitted Gaussian distribution and the AFM histogram, that the tail of the fluctuation toward the thicker tOS side leads to the further degradation of sub-threshold characteristic. As a result, Ion obtained by a certain Vg swing from Voff is significantly reduced due to the negative shift of Voff. Therefore, these results emphasize the necessity of improvement of the tOS uniformity for the better performance with maximizing the merits of the bilayer TFETs. The detailed S.S. and Ion values degraded by CTF will be discussed later.

FIG. 9.

Degradation of TFET Id-Vg characteristic with CTF. Two cases of thickness distributions, Gaussian fitting and AFM histogram as shown in Fig. 8(b), are assumed.

FIG. 9.

Degradation of TFET Id-Vg characteristic with CTF. Two cases of thickness distributions, Gaussian fitting and AFM histogram as shown in Fig. 8(b), are assumed.

Close modal

Generally, interface states at gate-oxide/semiconductor interfaces stretch out the Id-Vg characteristics because of charge trapping and/or de-trapping associated with the surface potential change. Additionally, in bilayer tunneling FETs with potential fluctuation at the OS-channel surface, the energy band diagram in the tunneling junction along the gate metal direction is no longer uniform over a tunneling area, suggesting that the spatial non-uniformity of the tunneling probability might not be ignored, as shown in Fig. 10(a). Therefore, influences of interface states and the surface potential fluctuation due to trapped charges are examined independently.

FIG. 10.

(a) Energy band diagrams of ZnO/Ge TFET with SPF at the ZnO channel surface. (b) Relationship between Ψs-OS and Vg of n-ZnO/p-Ge TFET.

FIG. 10.

(a) Energy band diagrams of ZnO/Ge TFET with SPF at the ZnO channel surface. (b) Relationship between Ψs-OS and Vg of n-ZnO/p-Ge TFET.

Close modal

The degradation of the Id-Vg characteristic of the ZnO/Ge bilayer TFET due to charge trapping/de-trapping attributable to interface states is evaluated, first. Figure 10(b) shows the surface potential (Ψs-OS), which is the position of Ec-OS at the surface of the OS channel measured from the Fermi level, and Id as a function of Vg corresponding to the on/off switching. Then, the Id-Vg characteristics are calculated with assuming a uniform energy distribution of an interface state density (Dit) over the bandgap of ZnO and a Vg shift by the density of the trapped charges of ΔQit = qDitΔΨs-OS, where q is the elementary charge. It is found in Fig. 10(b) that the change in Ψs-OS is extremely small in the sub-threshold region of this bilayer TFET. Figure 11(a) shows the calculated Id-Vg characteristics including the influence of interface states. The sub-threshold characteristics are still extremely steep even with a large amount of Dit. This is because ΔΨs-OS in the sub-threshold region is very small and the amount of the trapped charges itself can be ignored. On the other hand, since the extremely high Dit of 1013 to 1014 cm-2eV-1 prevents the deep band bending of the surface potential in the on-state, Ion decreases with increasing Dit.

FIG. 11.

Degradation of TFET Id-Vg characteristic attributable to interface states; (a) influence of charge trapping into interface states on stretch out of Id-Vg characteristics and (b) influence of surface potential fluctuation (SPF) due to trapped charge at the interface.

FIG. 11.

Degradation of TFET Id-Vg characteristic attributable to interface states; (a) influence of charge trapping into interface states on stretch out of Id-Vg characteristics and (b) influence of surface potential fluctuation (SPF) due to trapped charge at the interface.

Close modal

Subsequently, the influence of SPF on the TFET performance is examined. We have already obtained Ψs-OS and Id as a function of Vg without taking the surface potential fluctuation into account, as shown in Fig. 10(b). In this study, by assuming the gauss distribution for the fluctuation of Ψs-OS, the Id values with SPF (Id-PF) at each Vg can be numerically calculated as follows.

IdPF(Vg)=idψsOSPψsOSdψsOS
(2)
PψsOS=12πσs2expψsOSψsOS22σs2
(3)

Here, σs is the standard deviation of the surface potential. The values of σs are calculated on a basis of the Brews model given by the following equations,33,34

σs2=q2Ncc4πεs+εox2qkT2ln1+CλCox+Cd2
(4)
Cλ=εs+εoxλ
(5)

where εs and εox are the permittivity of a semiconductor and a gate insulator, respectively, Cox and Cd are capacitances of a gate insulator and a semiconductor, respectively, Ncc is the charged center density, and λ is a parameter related to the surface potential fluctuation. Here, λ of 7.5 nm was used.33 A capacitance associated with interface traps is not included in this equation. It should be noted here that positive voltage is applied to the gate metal to turn-on the proposed n-OS/p-IV TFETs, meaning that the OS channel surface is accumulated. Hence, in accumulation condition, σs is represented by using accumulation layer capacitance (Cacc) with quantum effect as follows.34 

σs2=q2Ncc4πλ2qkT21Cacc2=164π3qkT2h2m*q2Ncc

where, h is Planck’s constant and m* is the electron effective mass in OS channel (=0.28m0). Figure 11(b) shows the influence of SPF on the Id-Vg characteristics of the n-ZnO/p-Ge TFET with σskT/q) of 5, 10, or 20 meV. Similar to the influences of CTF, Voff shifts toward the negative Vg side. This shift is caused by BTBT in the region with deeper Ec-OS, which has more negative Voff. As a result, the total Id-Vg characteristic becomes less steep than that without SPF. The σskT/q) values corresponding to Ncc of 1×1012, 1×1013 and 1×1014 cm–2 are estimated to be 3.2, 10, and 32 meV, respectively, under the accumulation condition of the OS channel surface. As a result, the influence of SPF on the performance of the present TFETs is expected not to be significant. More detailed results on the bilayer TFET performance with various σs such as Ion and average S.S. are described in the next section with an emphasis on the impact of EOT scaling.

Based on these simulation results, we can understand that influence of charge trapping/de-trapping in interface states at the gate-oxide/OS-channel interface is negligibly small, because the change in the surface potential of the OS channel in the sub-threshold region is extremely small. On the other hand, this fact means that a small variation of Ψs-OS by any mechanisms clearly shifts the threshold voltage. Therefore, if trapped charges at interface states can cause the spatial fluctuation of Ψs-OS, Voff can be shifted toward the negative Vg side and S.S. can be increased.

In the section II, we have shown that there is no clear impact of EOT on the sub-threshold properties of the TFET without any fluctuations.13 On the other hand, if the non-uniformities in the TFET are not negligible as discussed above, EOT can change the magnitude of the CTF/SPF influences. Figure 12(a) shows the degradation of the Id-Vg characteristics of the ZnO/Ge bilayer TFETs caused by CTF for EOT of 1 or 4 nm. EOT scaling does not increase Ion drastically but it significantly suppresses the negative Voff shift caused by CTF, resulting in the better sub-threshold property. Figures 12(b) and (c) benchmark the impact of EOT scaling on average S.S. and Ion, respectively, of bilayer TFETs with CTF under a Vg swing of 0.1 or 0.3 V from Voff. In thicker EOT, the average S.S. increases and Ion decreases largely because of the CTF. Meanwhile, the average S.S. under small EOT hardly increases even under the same degree of CTF. Also, small EOT increases Ion under a certain Vg swing thanks to the suppression of the negative Voff shift.

FIG. 12.

Impact of gate insulator thickness on TFET Id-Vg characteristics degraded by CTF. (a) Id-Vg characteristics with CTF, (b) average S.S. as a function of EOT, and (c) Ion as a function of EOT at a Vg swing of 0.1 or 0.3 V.

FIG. 12.

Impact of gate insulator thickness on TFET Id-Vg characteristics degraded by CTF. (a) Id-Vg characteristics with CTF, (b) average S.S. as a function of EOT, and (c) Ion as a function of EOT at a Vg swing of 0.1 or 0.3 V.

Close modal

Similarly, reduction of EOT can mitigate influences of SPF. Figure 13(a) shows the Id-Vg characteristics of the bilayer TFET degraded by SPF for EOT of 1 or 4 nm. Figure 13(b) and (c) also plot the average S.S. and Ion, respectively, of the bilayer TFET under a Vg swing of 0.1 or 0.3 V, as a function of EOT. It is clearly found that both the smaller average S.S. and the higher Ion are achievable by thinning EOT even if the performance of the TFET is strongly affected by SPF.

FIG. 13.

Impact of gate insulator thickness on TFET Id-Vg characteristics degraded by SPF. (a) Id-Vg characteristics with CTF, (b) average S.S. as a function of EOT, and (c) Ion as a function of EOT at a Vg swing of 0.3 V.

FIG. 13.

Impact of gate insulator thickness on TFET Id-Vg characteristics degraded by SPF. (a) Id-Vg characteristics with CTF, (b) average S.S. as a function of EOT, and (c) Ion as a function of EOT at a Vg swing of 0.3 V.

Close modal

As a result, it can be concluded that reduction of EOT as well as the improvement of the tOS uniformity and the high-k/OS interfacial quality are important directions to improve the bilayer TFET performance. The influences of the non-uniformities of the OS thickness and the MOS interface charges are particularly important for the bilayer TFETs because the current is composed of the gate-normal BTBT current over the entire region of the tunneling junction, while a typical planar TFET structure utilizes BTBT only near the surface of the source/channel junction just below the gate electrode. Therefore, the improvements of the uniformities to suppress the performance degradation and EOT scaling to mitigate these influences are both the important guidelines for any kinds of bi-layer TFETs utilizing gate-normal BTBT.

In this study, we have proposed a new concept on a hetero tunneling junction with the type-II energy band alignment realized by an n-OS channel and a p-IV source and have addressed the effectiveness of a bilayer TFET employing this hetero tunneling junction with gate-normal BTBT. TCAD simulations have revealed the unique and remarkably high potential of the proposed bilayer TFET such as high Ion, larger Ion/Ioff ratio, near-zero minimum S.S. and small average S.S. average. It has also been clarified that reduction of Eb-eff by the OS material selection and the Ge fraction in the SiGe source effectively increases Ion with decreasing S.S. and that tOS is another key parameter to increase Ion. On the other hand, it has been found that non-uniformities associated with the tunneling junction thickness and the gate stack structure significantly degrade the sub-threshold properties of bilayer TFETs. The threshold Vg to generate BTBT, which varies with the channel thickness and the Ec-OS position at the OS channel surface, is negatively shifted in the Id-Vg characteristics under the existence of CTF or SPF, which increases the S.S. value and decreases Ion at certain Vg swing from Voff. Thus, the improvement of the uniformities related to the tunneling junction is critical to the high performance of the bi-layer TFETs. It has been found, in addition, that thinning of EOT is effective to mitigate these influences. Therefore, not only the improvements of the uniformities but also thickness of gate insulator are important guidelines for bilayer TFETs with gate-normal BTBT.

This work was supported by JST (Japan Science and Technology Agency) CREST Grant Number JPMJCR1332, Japan.

1.
L.
Lattanzio
,
N.
Dagtekin
,
L.
De Michielis
, and
A. M.
Ionescu
,
IEEE Trans. Electron Dev.
59
,
2932
(
2012
).
2.
S.
Agarwal
,
J. T.
Teherani
,
J. L.
Hoyt
,
D. A.
Antoniadis
, and
E.
Yablonovitch
,
IEEE Trans. Electron Dev.
61
,
1599
(
2014
).
3.
S. W.
Kim
,
J. H.
Kim
,
T.-J. K.
Liu
,
W. Y.
Choi
, and
B.-G.
Park
,
IEEE Trans. Electron Dev.
63
,
1774
(
2015
).
4.
L.
Lattanzio
,
L.
De Michielis
, and
A. M.
Ionescu
,
Solid-State Electron.
74
,
85
(
2012
).
5.
D.
Sarkar
,
X.
Xie
,
W.
Liu
,
W.
Cao
,
J.
Kang
,
Y.
Gong
,
S.
Kraemer
,
P. M.
Ajayan
, and
K.
Banerjee
,
Nature
526
,
91
(
2015
).
6.
M.
Kim
,
Y.-H.
Kim
,
M.
Yokoyama
,
R.
Nakane
,
S.-H.
Kim
,
M.
Takenaka
, and
S.
Takagi
,
Thin Solid Films
557
,
298
(
2014
).
7.
Y.
Zhu
,
M. K.
Hudait
,
D. K.
Mohata
,
B.
Rajamohanan
,
S.
Datta
,
D.
Lubyshev
,
J. M.
Fastenau
, and
A. K.
Liu
,
J. Vac. Sci. Tech. B
31
,
041203
(
2013
).
8.
G.
Dewey
,
B.
Chu-Kung
,
J.
Boardman
,
J. M.
Fastenau
,
J.
Kavalieros
,
R.
Kotlyar
,
W. K.
Liu
,
D.
Lubyshev
,
M.
Metz
,
N.
Mukherjee
,
P.
Oakey
,
R.
Pillarisetty
,
M.
Radosavljevic
,
H. W.
Then
, and
R.
Chau
,
Tech. Dig. IEEE Int. Electron Device Meeting
, p.
785
(
2011
).
9.
Q. T.
Zhao
,
W. J.
Yu
,
B.
Zhang
,
M.
Schmidt
,
S.
Richter
,
D.
Buca
,
J.-M.
Hartmann
,
R.
Luptak
,
A.
Fox
,
K. K.
Bourdelle
, and
S.
Mantl
,
Solid-State Electron.
74
,
97
(
2012
).
10.
T.
Roy
,
M.
Tosun
,
M.
Hettick
,
G. H.
Ahn
,
C.
Hu
, and
A.
Javey
,
Appl. Phys. Lett.
108
,
083111
(
2016
).
11.
T.
Krishnamohan
,
D.
Kim
,
S.
Raghunathan
, and
K.
Saraswat
,
Tech. Dig. IEEE Int. Electron Device Meeting
, p.
947
(
2008
).
12.
T.
Yu
,
J. T.
Teherani
,
D. A.
Antoniadis
, and
J. L.
Hoyt
,
IEEE Electron Dev. Lett.
34
,
1503
(
2013
).
13.
K.
Kato
,
H.
Matsui
,
H.
Tabata
,
M.
Takenaka
, and
S.
Takagi
,
Tech. Dig. IEEE Int. Electron Device Meeting
, p.
377
(
2017
).
14.
C. G.
Van de Walle
and
J.
Neugebauer
,
Nature
423
,
626
(
2003
).
15.
J.
Robertson
and
S. J.
Clark
,
Phys. Rev. B
83
,
075205
(
2011
).
16.
W.
Wu
,
C.
Jiang
, and
V. A. L.
Roy
,
Nanoscale
7
,
38
(
2015
).
17.
Y.
Mi
,
H.
Odaka
, and
S.
Iwata
,
Jpn. J. Appl. Phys.
38
,
3453
(
1999
).
18.
K.-H.
Kao
,
A. S.
Verhulst
,
W. G.
Vandenberghe
,
B.
Sorée
,
G.
Groeseneken
, and
K.
De Meyer
,
IEEE Trans. Electron Dev.
59
,
292
(
2012
).
19.
E. O.
Kane
,
J. Appl. Phys.
32
,
83
(
1961
).
20.
TCAD Sentaurus Device Manual Version H-2013.03 (
Synopsys
,
2013
).
21.
J.
Nishii
,
F. M.
Hossain
,
S.
Takagi
,
T.
Aita
,
K.
Saikusa
,
Y.
Ohmaki
,
I.
Ohkubo
,
S.
Kishimoto
,
A.
Ohtomo
, and
T.
Fukumura
,
Jpn. J. Appl. Phys.
42
,
L347
(
2003
).
22.
K.
Ellmer
and
R.
Mientus
,
Thin Solid Films
516
,
4620
(
2008
).
23.
T.
Hirao
,
M.
Furuta
,
H.
Furuta
,
T.
Matsuda
,
T.
Hiramatsu
,
H.
Hokari
,
M.
Yoshida
,
H.
Ishii
, and
M.
Kakegawa
,
J. Sci. Inf. Disp.
15
,
17
(
2007
).
24.
Y. V.
Li
,
J. I.
Ramirez
,
K. G.
Sun
, and
T. N.
Jackson
,
IEEE Elec. Dev. Lett.
34
,
891
(
2013
).
25.
H.
Bong
,
W. H.
Lee
,
D. Y.
Lee
,
B. J.
Kim
,
J. H.
Cho
, and
K.
Cho
,
Appl. Phys. Lett.
96
,
192115
(
2010
).
26.
K.
Kato
,
H.
Matsui
,
H.
Tabata
,
M.
Takenaka
, and
S.
Takagi
,
Appl. Phys. Lett.
112
,
162105
(
2018
).
27.
K.
Ellmer
and
R.
Mientus
,
Thin Solid Films
516
,
4620
(
2008
).
28.
K.
Ellmer
,
J. Phys. D: Appl. Phys.
34
,
3097
(
2001
).
29.
A.
Osinsky
and
S.
Karpov
,
Zinc Oxide Bulk, Thin Films and Nanostructures –Processing, Properties and Applications–
, Edited by
C.
Jagadish
and
S. J.
Pearton
, Chap. 15,
Elsevier
,
Amsterdam, The Netherlands
,
2006
.
30.
Y.-J.
Kim
,
J.
Cho
,
Y. J.
Hong
,
J.-M.
Jeon
,
M.
Kim
,
C.
Liu
, and
G.-C.
Yi
,
Nanotechnology
21
,
055303
(
2010
).
31.
S. J.
Pearton
,
D. P.
Norton
,
K.
Ip
,
Y. W.
Heo
, and
T.
Steiner
,
Progress in Mater. Sci.
50
,
293
(
2005
).
32.
L.
De Michielis
,
L.
Lattanzio
, and
A. M.
Ionescu
,
IEEE Electron Dev. Lett.
33
,
1523
(
2012
).
33.
J. R.
Brews
,
J. Appl. Phys.
43
,
2306
(
1972
).
34.
E. H.
Nicollian
and
J. R.
Brews
,
MOS Physics and Technology
, Chap. 5,
Wiley
,
New York, NY, USA
,
2003
.