Transport spectroscopy through dopant atom array in silicon junctionless nanowire transistors
Zhao Xiao-Song1, 2, Han Wei-Hua1, 2, †, Guo Yang-Yan1, 2, Dou Ya-Mei1, 2, Yang Fu-Hua1, 2, 3, ‡
Engineering Research Center for Semiconductor Integrated Technology, Beijing Engineering Center of Semiconductor Micro-Nano Integrated Technology, Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China
School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 100083, China
State Key Laboratory for Superlattices and Microstructures, Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China

 

† Corresponding author. E-mail: weihua@semi.ac.cn fhyang@semi.ac.cn

Project supported by the National Key R&D Program of China (Grant No. 2016YFA0200503) and the National Natural Science Foundation of China (Grant No. 61327813).

Abstract

We demonstrate electron transport spectroscopy through a dopant atom array in n-doped silicon junctionless nanowire transistors within a temperature range from 6 K to 250 K. Several current steps are observed at the initial stage of the transfer curves below 75 K, which result from the electron transport from Hubbard bands to one-dimensional conduction band. The current-off voltages in the transfer curves have a strikingly positive shift below 20 K and a negative shift above 20 K due to the electrostatic screening induced by the ionized dopant atoms. There exists the minimum electron mobility at a critical temperature of 20 K, resulting from the interplay between thermal activation and impurity scattering. Furthermore, electron transport behaviors change from hopping conductance to thermal activation conductance at the temperature of 30 K.

1. Introduction

During the past few years, much attention has been paid to silicon nanowire transistors which are considered key to future scaling beyond the 10 nm technology node due to their excellent electrostatic control and simple process.[14] Moreover, with the shrinking of the device scales, one-dimensional (1D) transport can be clearly observed in silicon nanowire transistors even at room temperature.[5,6] Meanwhile, the influence of dopant atoms is increasingly remarkable on the characteristics of devices in such a small dimension of sub-10 nm. Dopant distribution and fluctuations can cause a device-to-device fluctuation in threshold voltage[7] and on/off current.[8,9] Some work gained a remarkable insight into the transport of electrons in different dopant environments at low or room temperature, such as single dopant atom[10,11] or a few dopant-induced quantum dots.[1214] Furthermore, when the channel size approaches several tens of nanometers, the electron transport through a dopant-atom array can be observed.[12] The arrangement and position of the dopant atoms have a more significant influence on the transport property in nanowire devices with diameters less than 1 nm.[15,16] It is essential in understanding the influence of the ionized dopant on the electronic characteristic. Here, we explore the transport spectroscopy from Hubbard bands in a dopant atom array to 1D conduction band in n-doped silicon junctionless nanowire transistor (JNT) by varying the temperature from 6 K–250 K. In particular, we study the effect of dopant ionization at critical temperatures on the shifted current-off voltage, the minimum electron mobility, and the transition of electron transport mechanics.

2. Device fabrication

The device fabrication started from a (100)-oriented silicon-on-insulator (SOI) wafer with a top silicon thickness of 55 nm and buried oxide of 145 nm. After growing a 20-nm thermal oxidation layer, the SOI wafer was uniformly implanted by phosphorus ions with a doping fluence of 1×1013 cm−2 at an energy of 33 keV. A uniform phosphorus doping concentration of ND ≈ 2×1018 cm−3 could be achieved after annealing at 1000 °C in N2 for 15 s. Silicon nanowire was defined along the ⟨110⟩ direction by electron beam lithography (EBL) and inductively coupled plasma (ICP) etching. The sacrificial oxidation was introduced to reduce the etching damages. After being rinsed in the 5% HF, the nanowire was oxidized in dry oxygen at 900 °C for 1 h to form a 22-nm gate dielectric and to further reduce the cross-sectional area of the nanowire channel. Finally, a silicon nanowire channel with a width of 18 nm and height of 30 nm was achieved. After depositing a 200-nm-thick polysilicon layer to wrap the nanowires, EBL and ICP etching were used to obtain a polycrystalline silicon gate with a gate length of 280 nm. The top view scanning electron microscopy (SEM) image of our device is shown in Fig. 1(a). After depositing 200-nm SiO2 for passivation, 20-nm Ni and 300-nm Al were evaporated to form Ni/Si ohmic contact and final metallization with the help of conventional optical lithography. Figure 1(b) shows the schematic structure of our JNT. The single n-channel JNTs were measured in a vacuum chamber at temperatures ranging from 6 K–250 K with the Lakeshore-340 temperature controller and Agilent B1500 semiconductor parameter analyzer.

Fig. 1. (color online) (a) Top view SEM image of our device after gate formation. The cross section of channel is 18 nm × 30 nm and the gate length is 280 nm. (b) Three-dimensional schematic of JNT after Al pad formation.
3. Results and discussion

Temperature-dependent transfer characteristics in both linear and log scales are shown in Fig. 2(a) for the JNT at the bias Vds = 10 mV within a temperature range from 6 K–100 K. Current steps can be observed clearly below 50 K but gradually smear with the increase of the temperature, and eventually disappear beyond 75 K. The IdsVg curves show a current minimum at 20 K for a fixed drain voltage. Moreover, the curves increasing with the temperature rising have a positive shift below a temperature of 20 K and a negative shift above a temperature of 20 K. To show more clearly, the transconductance gm curves are presented in Fig. 2(b) as a function of gate voltage Vg at Vds = 10 mV for various temperatures. The gmVg curve has a maximum positive shift in gate voltage at 20 K, if we set the gate voltage at 6 K as a reference. Each current step in Fig. 2(a) corresponds to the population of an individual 1D sub-band.[1719] The experimental subband energy spacing can be estimated according to gate voltage spacing ΔVg from[12,20]

where m* is the effective mass of an electron and set to be 0.19 m0, ħ is the reduced Planck constant, e is the elementary charge, and Cox is the gate capacitance per unit area and estimated to be 1.56×10−7 F/cm2 for the gate-oxide layer thickness of 22 nm. The values of extracted gate voltage spacing ΔVg between adjacent transconductance peaks are 1 V, 0.85 V, and 0.91 V at 6 K, respectively, as shown in Fig. 2(b). Correspondingly, the values of energy spacing are estimated to be 6.17 meV, 5.25 meV, and 5.62 meV, which are less than the thermal energy of 6.5 meV at 75 K. Thus, the current steps disappear beyond 75 K where the thermal energy is higher than the subband spacing.

Fig. 2. (color online) (a) Drain current Ids and (b) transconductance gm (gm = ∂Ids/∂Vgs) versus gate voltage Vg of a single channel JNT at Vds = 10 mV at different temperatures. The curves in panel (b) are shifted for clarity.

In order to clarify the effect of temperature on the shift of the curve, the temperature-dependent tendencies of the current-off voltage Vgp and the gate voltage Vgm at the maximum gm peak are shown in Fig. 3(a). The current-off voltage Vgp is shifted positively from 6 K–20 K and negatively from 20 K–250 K. The gate voltage Vgm has a similar temperature-dependent tendency, confirming that the curves varying with temperature have shifted as a whole. To give a clear insight into the inner physics, the temperature dependencies of the effective mobility and electron concentration are drawn in Figs. 3(b) and 3(c), respectively. The effective mobility can be extracted from[21]μeff = gm max/Cox(Weff/L)Vds, where gm max is the maximum of transconductance peaks, L is the gate length, and Weff is the effective width and defined as twice the nanowire height plus the nanowire width. A prominent feature shown in Fig. 3(b) is the effective mobility minimum at 20 K. The mobility minimum at 20 K comes from the interplay between impurity scattering and thermal activation at low temperature.[22] As the ionized dopant atoms increase with the temperature rising from 6 K–20 K, more positive scattering centers are induced, resulting in the decrease of the mobility. At higher temperature, the ionized impurity scattering is gradually reduced by the shorter interaction time due to the thermal activation, causing the effective mobility to increase at temperature from 20 K–100 K. Figure 3(c) shows a quantitative result about the electron concentration. The electron concentration from 6 K–250 K can be expressed as[23]

where
Nc is the effective density of states in the conduction band, h is the Planck’s constant, and ED is set to be 2.6 meV according to the minimum effective mobility as discussed in Ref. [22]. The electron concentration is far less than the doping concentration but increases dramatically from 6 K–20 K and tends to be stable at a higher temperature. The increasing of the ionized dopant atoms can be estimated by the electron concentration according to the charge neutrality. The ionized dopant atoms increase dramatically from 6 K–20 K, which accords well with the tendency of the effective mobility. In the silicon nanowire channel, the positive ionized dopant atoms will induce negative charges at the surface of the nanowire[24] which certainly screen the gate voltage to some extent. Larger current-off voltage Vgp has to be used to neutralize the electrostatic screening as the ionized dopant atoms increase, resulting in a positive shift of Vgp as shown in Fig. 3(a). Above 20 K, the thermally activated electrons increase rapidly to make the current-off voltage Vgp decrease, which has been investigated in Refs. [25]–[27]. Therefore, the ionization of the dopant atoms enhances the electron scattering and the electrostatic screening, resulting in the suppression of the effective mobility and the positive shift of the current-off voltage in our JNT.

Fig. 3. (color online) (a) The current-off voltage Vgp and the gate voltage Vgm at the maximum gm peak versus temperature. The Vgp and Vgm are marked by dots and asterisks in Fig. 2(b). The current-off voltages are set to be the gate voltage at which the current is 10−12 A. (b) Effective mobility dependence on temperature. (c) Electron concentration dependence on temperature.

Figure 4 shows the temperature-dependent evolution of the electron transport spectroscopy from Hubbard bands in a dopant atom array to 1D conduction band. At the initial stage of the conductance below the temperature of 50 K, additional current oscillations before the first current step of 1D transport can be observed at Vds = 10 mV as shown in Fig. 4(a). The corresponding transconductance gm curve is drawn in Fig. 4(b) to clarify the temperature effect of the additional current oscillations. The fine gm peaks are sensitive to temperature and smear quickly with increasing temperature. At very low temperature, the ionized dopant atoms are isolated and form discrete quantum dots.[28] The three fine gm peaks at 6 K can be attributed to the transport through three discrete dopant-induced quantum dots.[29] The discrete ionized dopants in an array for the initial channel are coupling with each other to form the Hubbard band due to the repulsion of electrons with the same spin.[30] Two clear gm envelopes instead of three fine gm peaks can be observed in Fig. 4(b) from 10 K to 30 K, which validates the formation of the Hubbard band. The two transconductance envelopes are separated by an average gate-voltage spacing ΔVg = 0.464 V at the temperature range from 10 K to 30 K, corresponding to an energy spacing of 2.85 meV according to Eq. (1). The average energy spacing of 2.85 meV corresponds to the energy gap between two Hubbard bands, which accords well with the value given in Ref. [23] in a similar device. The thermal energy kBT of electrons at 30 K is about 2.6 meV, which is closer to the calculated energy gap of 2.85 meV. As a result, the merging of two Hubbard bands can be observed at the temperature of 50 K, in which the thermal energy is higher than the Hubbard band gap.

Fig. 4. (color online) (a) Conductance curves GVg (a) and (b) transconductance curves gmVg at Vds = 10 mV for different temperatures. The curves at each temperature are systematically shifted for the clarity of the fine current steps. The values of gate voltage spacing of two transconductance peaks are 0.464 V at 10 K, 15 K, 30 K and 0.4 V at 20 K, respectively.

To clarify the electron transport mechanisms at different temperatures, Arrhenius plots of conductance curve at three gm valleys and the corresponding maximum positions are drawn in Fig. 5(a). The conductances remain stable within the temperature range from 6 K–30 K. The activation energy can be estimated to be 0.14 meV, 0.09 meV, 0.06 meV, and 0.17 meV for the curves at Vg1, Vgm, Vg2, and Vg3, respectively, according to the slope of the curve in Fig. 5(a), in which the curve is described by

where G, Eact, and kB are the conductance, activation energy, and Boltzmann constant, respectively. We conclude that the electrons transport by variable range hopping in local states around the Fermi level.[31] However, the Arrhenius plot of conductance at the temperature of 30 K changes from relatively temperature-independent region to linearly dependent region. The lnG as a function of 1/T has a linear relation in the temperature range from 30 K–100 K. The electron behavior mainly relies on the hopping transport due to the thermal activation. With the temperature increasing, the conductance exponentially increases with the activation energy of 7.54 meV, 4.17 meV, 3.64 meV, and 2.93 meV at the gate voltage Vg1, Vgm, Vg2, and Vg3. In Fig. 5(b), the activation energy would decrease and tend to be a constant with the gate voltage increasing, for which the Fermi energy level is driven from the donor band to the conduction band by the gate voltages.

Fig. 5. (color online) (a) Arrhenius plots of conductance at various gate voltages, which correspond to maximum peaks at Vgm and three valleys Vg1, Vg2, Vg3 in the transconductance curves of Fig. 2(b). (b) Activation energies versus the gate voltage in different temperature regions.
4. Conclusion

In this research, we demonstrate the transport spectroscopy of dopant atom array in heavily n-doped silicon JNT from 6 K–100 K. The clear current steps for 1D subband transport are maintained below the temperature of 75 K. With the temperature increasing form 6 K–20 K, the ionization of the dopant atoms results in the suppression of the effective mobility and the positive shift of the current-off voltage. The transition of the electron transport from the Hubbard band to 1D conduction band is identified below the temperature of 50 K by the temperature-dependent evolution of gm peaks at the initial stage of the transfer characteristics. Arrhenius plots of conductance curves indicate the transition of electron transport mechanism at the temperature of 30 K from relatively temperature-independent region to linearly dependent region. Our study reveals that the ionized dopant atoms play an important role in the performance of silicon JNT. The in-depth understanding of dopant-induced QDs is highly desirable at giving better control for further scaling of transistors.

Reference
[1] Colinge J P Lee C W Afzalian A Akhavan N D Yan R Ferain I Razavi P O’Neill B Blake A White M Kelleher A M McCarthy B Murphy R 2010 Nat. Nanotechnol. 5 225
[2] Wang B Zhang H M Hu H Y Hu H Y Shi X W 2018 Chin. Phys. 27 067402
[3] Liu F Y Liu H Z Liu B W Guo Y F 2016 Chin. Phys. 25 047305
[4] Nazarov A N Rudenko T E 2015 Int. Semiconductor Conf. 27
[5] Yi K S Trivedi K Floresca H C Yuk H Hu W Kim M J 2011 Nano Lett. 11 5465
[6] Mirza M M Schupp F J Mol J A Maclaren D A Briggs G A D Paul D J 2017 Sci. Rep. 7 3004
[7] Shinada T Okamoto S Kobayashi T Ohdomari I 2005 Nature 437 1128
[8] Li Y Yu S M Hwang J R Yang F L 2008 IEEE Trans. Electron. Devices 55 1449
[9] Duan L Hu F Ding J W 2011 Acta Phys. Sin. 60 117201 in Chineses
[10] Tabe M Moraru D Ligowski M Anwar M Jablonski R Ono Y Mizuno T 2010 Phys. Rev. Lett. 105 016803
[11] Sellier H Lansbergen G P Caro J Rogge S Collaert N Ferain I Jurczak M Biesemans S 2006 Phys. Rev. Lett. 97 206805
[12] Han W H Wang H Yang X Ma L H Hong W T Lyu Q F Yang F H 2014 12th International Conference on Solid-State and Integrated Circuit Technology October 28–31 2014 Guilin, China 1 10.1109/ICSICT.2014.7021627
[13] Tyszka K Moraru D Samanta A Mizuno T Jabłoński R Tabe M 2015 Appl. Phys. 8 094202
[14] Samanta A Muruganathan M Hori M Ono Y Mizuta H Tabe M 2017 Appl. Phys. Lett. 110 093107
[15] Dong J C Li H Sun F W Zhang K Li Y F 2011 J. Phys. Chem. 115 13901
[16] Zhang X H Dong J C Wang Y Li L Li H 2013 J. Phys. Chem. 117 12958
[17] Guan X M Yu Z P 2005 Chin. Phys. Lett. 22 2651
[18] Wang H Han W H Ma L H Li X M Yang F H 2014 Chin. Phys. 23 088107
[19] Ma L H Han W H Wang H Yang X Yang F H 2015 IEEE Electron Device Lett. 36 941
[20] Morimoto K Hirai Y Yuki K Morita K 1996 Jpn. J. Appl. Phys. 35 853
[21] Colinge J P Xiong W Cleavelin C R Schulz T Schrüfer K Matthews K Patrunoolinge P 2006 IEEE Electron Device Lett. 27 120
[22] Li X M Han W H Wang H Ma L H Zhang Y Du Y D Yang F H 2013 Appl. Phys. Lett. 102 223507
[23] Ye L X 2012 Semiconductor Physics 1 2 Beijing Higher Education Press 118 121 in Chinese
[24] Diarra M Niquet Y M Delerue C Allan G 2007 Phys. Rev. 75 045301
[25] Joo M K Mouis M Jeon D Y Barraud S Park S J Kim G T Ghibaudo G 2014 Semicond. Sci. Technol. 29 045024
[26] de M Pavanello M A Trevisoli R D Doria R T Colinge J P 2011 IEEE Electron Device Lett. 32 1322
[27] Trevisoli R D Doria R T de M Pavanello M A 2011 Semicond. Sci. Technol. 26 105009
[28] Hong B H Jung Y C Rieh J S Hwang S W Cho K H Yeo K H Suk S D Yeoh Y Y Li M Kim D W Park D Oh K S Lee W S 2009 IEEE Trans. Nanotechnol. 8 713
[29] Ma L H Han W H Wang H Hong W T Lyu Q F Yang X Yang F H 2015 J. Appl. Phys. 117 034505
[30] Mott N F 1968 Rev. Mod. Phys. 40 677
[31] Ye L X 2009 Semiconductor Physics 2 2 Beijing Higher Education Press 540 542 in Chinese