†Corresponding author. E-mail: weihua@semi.ac.cn
‡Corresponding author. E-mail: fhyang@semi.ac.cn
*Project supported by the National Natural Science Foundation of China (Grant Nos. 61376096, 61327813, and 61404126) and the National Basic Research Program of China (Grant No. 2010CB934104).
We investigate the conductivity characteristics in the surface accumulation layer of a junctionless nanowire transistor fabricated by the femtosecond laser lithography on a heavily n-doped silicon-on-insulator wafer. The conductivity of the accumulation region is totally suppressed when the gate voltage is more positive than the flatband voltage. The extracted low field electron mobility in the accumulation layer is estimated to be 1.25 cm2·V−1·s−1. A time-dependent drain current measured at 6 K predicts the existence of a complex trap state at the Si–SiO2 interface within the bandgap. The suppressed drain current and comparable low electron mobility of the accumulation layer can be well described by the large Coulomb scattering arising from the presence of a large density of interface charged traps. The effects of charge trapping and the scattering at interface states become the main reasons for mobility reduction for electrons in the accumulation region.
Recent years have witnessed the rapid downscaling of field effect transistors, causing many adverse issues such as short channel effect. In particular, it becomes exceedingly arduous to form abrupt p– n junctions between the source/drain and channel as the gate length is reduced down to 10 nm or even less. Recently, a unified heavily-doped nanowire resistor, called a “ junctionless nanowire transistor (JNT)” , has captured great attention as a promising candidate for the further continuation of Moore’ s law because of the excellent gate control ability and simplified fabrication process.[1– 4] The channel is fully depleted in subthreshold operation. Threshold voltage is reached when the central region of the nanowire becomes neutral. The depletion region decreases as the gate voltage further increases.[5] The flatband voltage is defined as the applied gate voltage which enforces flat bands at the insulator– semiconductor interface. The JNT has a bulk neutral conduction path less affected by the transverse electric field from the gate than by the surface accumulation layer separated by the flatband voltage (Vfb). The accumulation carriers are confined to a very thin layer along the silicon– gate oxide interface by the electric field originating from the gate electrode. Owing to the large surface-to-bulk aspect ratio, the trapping and detrapping of the surface defects at the Si– SiO2 interface become crucial in the nanowire transistor. It has been reported that Coulomb scattering introduced by charged interface traps can become a dominant mechanism of mobility degradation for metal– oxide– semiconductor field-effect transistors (MOSFETs) in low field regime.[6] In the accumulation region of JNTs, the charges trapped at the interface might work as the Coulomb scattering centers for carriers in the channel. The carrier mobility might be degraded by the trapped charges at the interface due to the reduced distance between the carrier in the accumulation layer and the Coulomb scattering center. Many of the research studies on JNTs have focused on the neutral channel region. However, relatively little work has been done to analyze the accumulation layer of the silicon nanowire channel. The present work aims at analyzing the electrical properties of JNT in the linear accumulation region under the condition of the silicon surface with the influence of interface traps.
The key to fabricating the JNTs is the formation of a ultranarrow channel to ensure the full depletion of carriers in the channel under zero gate voltage. Therefore, it is important to achieve the line size on the order of tens of nanometers. For decades, photolithography has been the predominant method for MOSFET fabrication. Since costs for lithographic equipment and mask sets keep rising with every technology generation, alternative cheaper patterning techniques are earnestly required. The femtosecond laser lithography has a host of advantages including no mask, three dimensions, high resolution and rapid prototyping, which make the lithography an attractive method of fabricating the nanostructures.[7, 8] In this approach, the focused femtosecond laser beam is used to directly write the desired image into the photoresist without lithographic masks. Furthermore, previous theoretical and experimental results have shown that femtosecond laser with two-photon or multiphoton absorption has a potential to exceed the diffraction limit due to the nonlinear effect.[9– 11] The development of this novel maskless lithography has a tremendous potential for opening a new nanomanufacturing market in the future. In the present work, we demonstrate the fabrication of n-type single-channel JNT on a silicon-on-insulator (SOI) substrate by two-photon absorption process. Two dry etching processes are implemented to transfer the desired image from the negative photoresist SCR500 onto the top silicon layer. Suppressed conductivity and degraded electron mobility are observed in the accumulation region, due to the generation of new important scattering mechanisms especially Coulomb scattering arising from the charged defects at the Si– SiO2 interface.
Figure 1 shows the schematic diagrams of the fabrication process of JNTs. The devices were fabricated on a (100)-oriented SOI substrate with a top Si thickness of 55 nm. After thermally growing 15-nm silicon dioxide, the SOI wafer was uniformly and heavily doped by phosphorus ion implantation with a dose of 5 × 1013 cm− 2, leading to a uniform concentration of 1 × 1019 cm− 3. It was followed by the deposition of 50-nm SiO2 on top of Si. The SiO2 layer was chosen to minimize the interference effect of femtosecond laser and eliminate the silicon surface damage directly induced by direct laser scanning.[12] Afterwards, two laser beams with optimized wavelengths and powers were used to pattern a nanowire connecting two pads.[13] The laser beams focused by an objective lens were scanned into the substrate covering negative photoresist by adjusting the position of the translational stage. The desired image was transferred from negative photoresist onto the top silicon layer by two-step inductively coupled plasma (ICP) etching processes as presented in Figs. 1(c)– 1(f). First, the deposited SiO2 layer was etched with photoresist as a hard mask, and the top silicon layer was etched with SiO2 as a mask. After rinsing in the HF solution, the Si nanowire was suspended by corroding the buried oxide layer. The scanning electron microscope (SEM) image presented in the insert of Fig. 2(a) shows that the width of the nanowire channel is 47 nm on average and the length of nanowire is about 500 nm. After removing the SiO2 mask by 5% HF, the sample was oxidized in dry oxygen at 900 ° C for 1 h, resulting in the formation of a Si core with a height of 28 nm (Tsi) and a width of 27 nm (Wsi). The thickness of SiO2 gate dielectric formed by thermal oxidation is about 22 nm. A 200-nm-thick poly-silicon was then deposited by low-pressure chemical vapor deposition on the gate oxide layer. The poly-silicon was doped by arsenic ion implantation at a dose of 2 × 1015 cm− 2, leading to a doping level of approximately 1 × 1020 cm− 3 after annealing at 1000 ° C for 10 s. The gate electrode pads were fabricated by evaporating a 20-nm-thick Ti film to form a Ni/Si ohmic contact, and then 300-nm-thick Al film is used for final metallization via conventional optical lithography.
The device is placed in a vacuum chamber which can be cooled down to ∼ 6 K with the help of Lakershore-340 temperature controller. The electrical characteristics of the transistor are investigated by Agilent B1500 semiconductor parameter analyzer. Figure 2(a) shows the drain current and voltage (IDS– VGS) characteristic curves of the JNT measured at room temperature. The threshold voltage Vth is approximately 0.62 V. As can be seen in the logarithmic scale of Fig. 2(a), IDS is almost independent of VGS beyond the flatband voltage Vfb whose value is determined in Fig. 2(b). It suggests that the electron mobility is relatively low in higher VGS. In addition, it should be noted that IDS is only tens of nA, which is totally suppressed.
Low field mobility represents the upper limit of carrier mobility for a given device. The Y-function
where W, L, μ acc, θ 0, and Cox represent the effective channel width, channel length, accumulation channel mobility, mobility attenuation factor and oxide capacitance per unit area, respectively. The first term on the right-hand side of Eq. (1) represents the accumulation current Iacc and the second term represents the bulk current in flatband conduction, which can be written as Ibulk = WSi/L × qμ bulkNdTSiVDS. Here μ bulk is the bulk channel mobility and Nd is the doping concentration. To obtain accurate and low field electron mobility of the accumulation layer, the bulk current Ibulk in flatband conduction should be subtracted from the total current IDS. Y-function of accumulation current can be expressed as
Figure 2(c) shows the Iacc separated by IDS– Ibulk at Vfb and Yacc of the accumulation channel at VDS = 20 mV. It should be noted that Yacc as a function of VGS shows only one single slope now. Finally, the low field mobility of the accumulation channel extracted from the square of Yacc slope is μ acc = 1.25 cm2· V− 1· s− 1. The effective channel width W equals 2(WSi + TSi), and the channel length approximately equals the nanowire length of 500 nm. The slope of S2 in Fig. 2(b) is a combined result of the bulk conduction and the accumulation conduction.
To further investigate the degradation mechanisms behind this low value of low field mobility, we measure the low temperature electrical properties of JNT. Figure 3(a) shows IDS– VGS curves with VDS values ranging from 2 mV to 10 mV at a temperature of 6 K. Obvious current jumps are observed. The amplitudes of those current jumps increase with increasing VDS. Typical time dependence of the drain current at VDS = 20 mV, measured in steps of 0.1 s for a total time interval of 100 s, is indicated in Fig. 3(b). The applied gate bias is 2.75 V.We could observe multiple current levels. It has been argued that this phenomenon could be due to the trapping and detrapping behavior of multiple traps, which will cause more dramatic conduction fluctuations of the drain current. The high current level corresponds to the empty trap state, while the low current level is related to the electrons captured by the traps. The observed four-level fluctuations show that two single-charge-state traps located within ∼ 2kBT of the Fermi level in the channel contribute to the current jumps.[18] The insert in Fig. 3(b) illustrates the band diagram for VGS = 2.75 V with two types of interface traps. Figure 4 manifests the histograms of the time domain random telegraph signals of the JNT at different VGS values of 2.4 V, 2.65 V, 2.75 V, and 2.9 V. The drain bias is constant at 20 mV. Complex current levels are observed as the gate voltage increases, strongly suggesting the existence of multiple types of traps. Four current levels observed at 2.4 V are ascribed to capture and emission of electrons from the other two different traps within ∼ 2kBT range just like the situation of 2.75 V, while two current levels at 2.65 V and 2.95 V are due to the other two single level traps.[19] The femtosecond laser gives rise to non-direct damage to the top silicon layer because of the deposition of 50-nm SiO2. In general, the traps at Si– SiO2 interface may be induced by etching procedure.
The charged traps can influence the electrical performance at large gate overdrives especially above flatband voltage, where we notice a current saturation with increasing VGS. This can be easily explained by the increase of Coulomb scattering of the charged traps at the increased gate voltage. As previously stated, the squeezing of the carriers toward the Si– SiO2 interfaces induces a reduced distance between the carriers of the accumulation layer and the charged traps as Coulomb scattering centers. The interface traps cause the increase of Coulomb scattering, which will eventually reduce the drive current. Furthermore, the influence of surface roughness scattering on electron mobility is very weak above Vfb because of the low perpendicular electric field of the accumulation layer in JNT compared with the scenario of inversion transistor. Actually, the extracted accumulation electron mobility in Ref. [17] is much higher without the influence of charged traps, inferring that charged trap scattering is the dominant scattering mechanism in this experiment. Hence, the charged traps have a detrimental influence on the performances of the device with a degenerated mobility in accumulation layer and the reduction of the gate-control ability.[20] Furthermore, one may consider that a large density of charged traps may form clusters of charges with high probability. In some cases, those charges of clusters work as a potential barrier, for electrons may form localized states, making electrons of the accumulation layer arduously transported through the channel. At room temperature, there will be additional charged traps activated by thermal energy around the Si– SiO2 surface. The flatband voltage is influenced by charges that are trapped in the insulator-semiconductor interface or within the gate dielectric. Thus, more positive gate voltage should be used to reach the flatband condition than the previous theoretical value under the condition of the same doping concentration.[21] The above-mentioned results suggest that the carrier transport is trap-limited in the accumulation layer.
In this work, the JNTs each with a physical channel crosssection area of 27 nm× 28 nm are fabricated by femtosecond laser lithograph. The low field mobility is found to be totally suppressed above the flatband voltage, due to the increased Coulomb scattering induced by charged traps at the Si– SiO2 interface. The Coulomb scattering introduced by charged traps becomes the dominated scattering mechanism of electron transport in accumulation region. Our result shows that charged interface traps play a significant role in degrading the mobility of JNT especially in the accumulation layer above flatband voltage. Those findings clearly highlight the drastic influence of the large density of charged traps on the mobility degradation for the accumulation layer of JNTs.
The authors thank professor Duan Xuan-Ming, Dr. Chen Shu, Mrs. Li Yan, and Mr. Wang De-Song for their technical support in device fabrication.
1 |
|
2 |
|
3 |
|
4 |
|
5 |
|
6 |
|
7 |
|
8 |
|
9 |
|
10 |
|
11 |
|
12 |
|
13 |
|
14 |
|
15 |
|
16 |
|
17 |
|
18 |
|
19 |
|
20 |
|
21 |
|