Wenjing
Jie
ab and
Jianhua
Hao
*b
aCollege of Chemistry and Materials Science, Sichuan Normal University, Chengdu, China
bDepartment of Applied Physics, The Hong Kong Polytechnic University, Hung Hom, Hong Kong. E-mail: jh.hao@polyu.edu.hk
First published on 22nd November 2017
Graphene-based field effect transistors (FETs) were fabricated by employing ferroelectric Pb(Mg1/3Nb2/3)O3-PbTiO3 (PMN-PT) as a gate insulator. The co-existing effects of ferroelectric gating and interface charge trapping on the transport properties of graphene were investigated with respect to the FET structure. The sheet resistance (Rs) of graphene shows a slight decay under a small applied voltage, which is much less than the coercive voltage of the ferroelectric PMN-PT, suggesting non-negligible charge trapping effects. Moreover, when the applied voltage is increased up to a value larger than the coercive voltage, Rs exhibits three states: an initial rapid change, followed by a slow nearly exponential evolution, and finally a saturated state either during the applied voltage is retained or after it is released. In particular, a high-resistance state is finally reached due to the ferroelectric gating, implying that ferroelectric effects dominate this process. The underlying physical mechanism was fully investigated to effectively address the observed evolution of time-dependent Rs. Such a finding provides us an opportunity to understand the co-existing effects of ferroelectric gating and charge trapping and tune the transport properties of graphene through the interface effects.
Then, the transport characteristics of the graphene were systematically studied in the GFeFETs by PMN-PT gating. The sheet resistance (Rs) of graphene is calculated by measuring the drain current (Ids) when a source–drain voltage (Vds) of 0.5 V is biased. Fig. 2 shows the Rs of graphene as a function of the gate voltage (Vg). It is observed that the range of applied Vg is gradually increased from 50 to 100 to 150 V as shown in Fig. 2a, b and c, respectively. From Fig. 2a, it is clear that the Rs decreases with backward sweeping Vg, while it increases with forward sweeping, suggesting that the CVD-grown graphene is p-type in air at room-temperature. Moreover, a slight hysteresis loop is observed in the resistance curves. Similar p-type characteristics with an on/off ratio of about 3 can be observed for graphene layers on both PMN-PT and SiO2 when the range of Vg increases up to 100 V as shown in Fig. 2b and d, respectively. A relatively large hysteresis loop is observed in graphene on PMN-PT compared to that on SiO2. It should be pointed out that the charge concentration and resistivity of graphene primarily depend on the dielectric gate materials. To be more specific, the doping concentration is proportional to the dielectric constant of the gating insulator, while varying in inverse ratio to its thickness under the same Vg. Although the room-temperature dielectric constant of PMN-PT (∼5000) is over 3 orders-of-magnitude higher than that of SiO2 (∼3.9), the thickness of PMN-PT (0.5 mm) is over 1000 times higher than that of SiO2 (300 nm). Therefore, it is understandable that the E doping effects on graphene by PMN-PT or SiO2 gating are comparable under identical Vg. Accordingly, Rs of graphene on the PMN-PT is slightly smaller than that on SiO2. For the Vg range up to 150 V, graphene on SiO2 still shows similar transport behaviors. In our experiments, the transport properties of graphene were measured at room-temperature in air ambient. The fabricated FETs exposed to the laboratory environment were always found to be initially p-type due to the adsorbed molecules in air ambient.32 The Dirac point, located at about 130 V, can be observed in graphene/SiO2 FETs when the gate voltage is increased to 150 V as shown in the inset of Fig. 2d. However, for graphene on PMN-PT, the backward and forward scans of Rs show the typical ferroelectric butterfly-shape hysteresis loop when the Vg range is up to 150 V as shown in Fig. 2c. The ferroelectric polarization field in PMN-PT can contribute to the adjacent graphene layer with electrons or holes depending on the applied Vg or polarization direction. As a consequence, the hysteresis in ferroelectric polarization can offset the Dirac point and induce the butterfly-shape hysteresis in Rs of graphene with backward and forward sweeping Vg. Although the fabricated GFeFETs exhibit asymmetry in the hysteresis, reproducible and consistent behaviors are achieved as illustrated in Fig. 2c. This can be caused by some additional doping of water and oxygen from the air as aforementioned. Moreover, in the transport measurements, the Vg was swept from the positive maximum to the negative minimum, and then followed by forward sweeping with a time gap between the backward and forward sweeping. There could be a slight difference in the resistance for backward and forward sweeping due to the nonvolatile ferroelectric polarization; also, the charge trapped at the interface could induce the slight difference in the carrier density as well as the resistance of graphene when Vg is swept in a different direction. Thus, the transport curves are not closed (Fig. 2).
To further understand the correlation between the transport properties of graphene and ferroelectric polarization of PMN-PT, we measured the hysteresis loop of the polarization (P) versus E from the PMN-PT single crystal with graphene serving as top electrode and Au acting as bottom electrode as shown in Fig. 3a. The P of PMN-PT begins changing at about 75 V (E = 1.5 kV cm−1) and saturates at about 150 V with forward sweeping, while the P remains almost constant until −75 V and saturates at −150 V for backward sweeping with the coercive voltage (Vc) of about ±100 V (coercive field, Ec = 2 kV cm−1). To better understand the measured Rs (Fig. 2), we combined P and Rs dependences on Vg for different Vg scale as shown in Fig. 2a, b and c. In graphene/PMN-PT heterostructure, the ferroelectric P of PMN-PT gradually changes depending on Vg with a slight hysteresis loop. The ferroelectric gating modulates the transport properties of graphene as a conventional dielectric of SiO2 within the Vg regions, where the spontaneous polarization does not reverse, i.e., Vg is less than Vc. The ferroelectric polarization could not be reversed in the Vg range of 50 and 100 V. Herein, the resistance of graphene shows a gradual change when Vg is smaller than the coercive voltage as shown in Fig. 2a and b. As a consequence, the hysteresis loops observed for both devices by PMN-PT or SiO2 gating are primarily a result of the charge trapping at the interface.21,33 Moreover, when Vg exceeds Vc, the polarization can be switched. The simultaneous rise in resistance with the reversal of ferroelectric polarization is expected at Vc (about 100 V), inducing the hysteresis loop in the resistance behaviors when sweeping Vg in forward and backward directions. It is observed that Rs in graphene shows a consistent hysteresis loop with the polarization loop when Vg is beyond Vc. Furthermore, the Hall effect measurement has been performed for the graphene sample on the PMN-PT substrate. The carrier type and density of graphene on PMN-PT substrates under different gate voltage can be obtained as shown in Fig. 3b. The gate voltages in the range of −200 to 200 V with step of 50 V are applied during the measurements. The black circles represent holes, while the blue circles indicate electrons. The red circles stand for the Dirac points. The green arrows stand for the ramping direction of the gate voltage. The CVD-grown monolayer graphene samples exposed to laboratory environment are initially p-type. The excess hole concentration of graphene on PMN-PT under zero gate is measured to be about 5 × 1012 cm−2, which is slightly larger than that of graphene on SiO2 (about 2 × 1012 cm−2). The carrier type is changed into electron when the gate voltage is in excess of 100 V. Similar to the ferroelectric hysteresis in the resistance curve (Fig. 2c), there is a hysteresis in the carrier density of graphene when swithching the gate voltage forward and backward.
Furthermore, it should be noted that the hysteresis direction with the Vg range of 150 V is reversed compared to those with Vg ranges of 50 and 100 V as well as graphene on the SiO2 substrate. Since the interface charge trapping effect dominates the ferroelectric polarization effect, anti-hysteresis can be observed in the transport properties of GFeFETs.14,26,29 Similar transport properties caused by the charge trapping effect have been previously observed even in graphene on conventional dielectric materials, which do not exhibit a spontaneous polarization.21,33 Efforts have been made to eliminate the effect of charge traps on the transport properties of GFeFETs, such as improving the crystalline qualities of the ferroelectric materials or heating the ferroelectric substrates,26 avoiding exposure of GFeFETs to air during the transfer and measurement process,14 or measuring the transport properties after removing the gate voltage.29 In this study, the reverse of anti-hysteresis was realized by the pre-poling the PMN-PT crystal with Vg of −250 V for 10 min prior to the resistance measurements, resulting in the downward polarization state, i.e., the electric dipole moments in the PMN-PT pointing away from the graphene sheet. Then, the anti-hysteresis behaviors can be reversed when the sweeping Vg was larger than the coercive voltage of PMN-PT. This can be caused by the sharp transition of ferroelectric switching of the PMN-PT crystal. The reversed hysteresis suggests that the ferroelectric polarization can dominate the transport properties of graphene under high gate voltage. Moreover, the extrinsic charge trapping effects also can modify the transport properties of graphene. The evidence is that the Dirac points for forward and backward sweeping shift closer to each other when the sweeping rate is changed from 1 to 10 V s−1. Fig. 3b presents the normalized Rs as a function of Vg with different sweeping rates from 1 to 10 V s−1. For fast ramping, the charge trapping effect should be less significant, hence gating should be more effective. The sweeping rate dependent Dirac point shifts are caused from the time-dependent charge trapping or charge distribution effects at the interface. Therefore, the transport behaviors of graphene gating by ferroelectric PMN-PT are tuned by both the underlying ferroelectric polarization and the interface charge trapping. Clearly, the ferroelectric doping is dominated in the transport behaviors of graphene when the Vg range is in excess of the Vc of PMN-PT. Moreover, it should be mentioned that the Dirac points show little variation at the sweeping speeds below 1 V s−1 (not shown here). Actually, not only the Dirac points remain constant but also the transport curves remain unchanged when the sweeping speeds are below 1 V s−1.
To further explore the co-existing effects of ferroelectric polarization and charge trapping, time-dependent Rs was measured under Vg, which is smaller than Vc. In general, the Rs of graphene should remain invariable under a constant E, while graphene on PMN-PT substrate exhibits time-dependent behaviors as shown in Fig. 4a. The pulse signals with the peak-to-peak voltage of 10 V, i.e., from −5 V to +5 V, are used to serve as the Vg. For comparison, we also measured the resistance response of graphene in graphene/SiO2 heterostructures. Both graphene layers are zero biased for more than 1 h until their resistances reach a stable state. Then, gate pulses were applied and the resistances were measured accordingly. Both resistances were modulated at the same frequency as that of the driving signals (10 mHz), and the positive Vg gives rise to the rapid increase of resistance corresponding to the decrease of carrier concentration, while the negative Vg leads to the dramatic decrease in resistance. The observed characteristics are in accordance with the p-type transport properties of graphene. Moreover, there is a slight and negligible change in the resistance during the retention of both positive and negative gate voltages for graphene/SiO2 system caused by the time-dependent interface charge trapping effect. However, the resistance for graphene on PMN-PT shows a gradual evolution under the small Vg, which is much less than Vc. To understand the unique behaviours for graphene on PMN-PT, the surface morphology of the PMN-PT substrate and the as-fabricated graphene were characterized using an atomic force microscope (AFM) (Fig. 4b). It is known that single-crystal PMN-PT possesses a rhombohedral structure and its spontaneous polarization points to the (111)-direction in the pseudo-cubic unit cell.34,35 Thus, some structural kinks inevitably exist at the domain walls, inducing some contrast patterns on the surface of (100)-oriented PMN-PT as shown in the inset of Fig. 4b. Such kinks are also observed in the AFM image of graphene on the top surface of the PMN-PT substrate due to the ultra-thin graphene layer. It was reported that the uneven surface morphology could enhance the charge trapping effects at the interface between the graphene layer and the underlying ferroelectric materials.26 Typically, the charge traps can be divided into two types. One is the surface defects related “intrinsic” traps and the other is the adsorbates related “external” traps. Uneven surface morphology would not only induce more surface defects for the underlying ferroelectric materials but also provide more locations for adsorbates. Thus, we believe that the surface morphology can affect the charge traps at the interface of graphene/PMN-PT. At the rising edge of the gate pulse, Rs increases instantly due to the hole density decrease in graphene caused by the electron accumulation at the interface. Rs then decreases gradually when Vg is maintained at +5 V. In this process, the accumulated electrons at the interface are trapped by the trapping states. As a consequence, the graphene layer gets less gate potential from the underlying ferroelectric gate of PMN-PT. At the falling edge of the gate pulse, Rs decreases rapidly due to the increase in carrier density of holes caused by the hole accumulation at the interface. Rs increases quite slowly during the period of the negative voltage due to the de-trapping process of the trapped electrons. Thus, the carrier density of holes decreases and the resistance increases accordingly.19 By considering that magnitude of the applied gate pulses is much smaller than the coercive voltage, the spontaneous polarization of PMN-PT cannot be switched during the entire process. Herein, the time-dependent behaviors of graphene/PMN-PT are dominated by the charge trapping and de-trapping at the interface when the Vg is much smaller than Vc.
Moreover, it should be mentioned that single-crystal PMN-PT is also a piezoelectric material possessing a large converse piezoelectric response.36,37 Thus, the Vg not only provides E but also induces in-plane strain, and such strain could be potentially delivered to the above graphene layer.31,38,39 Thus, there could be two possible co-existing tuning mechanisms, i.e., field effect and strain effect with regards to employing PMN-PT as a gate dielectric. The strain provided by the PMN-PT can be calculated according to the change in lattice constant under external E. Actually, the strain is much smaller, typically less than 0.1% for Vg of 500 V.31 However, at least 0.5% strain is required to access the strain dependent resistance.40,41 In addition, typically, a large strain in excess of 20% is needed to induce a band gap in graphene.42,43 Moreover, ohmic contact can be formed between graphene and the electrodes and can be retained even when Vg is up to ±200 V. Accordingly, the effect of piezoelectric strain on the contact resistance can be ignored in our experiments. Therefore, the small piezoelectric strain is negligible in our sample and the strain effects on our GFeFETs can be ruled out, while the ferroelectric field effects and the charge trapping effects are dominant in the transport properties.
By considering that both the reversal of ferroelectric polarization and interface charge trapping can contribute to the observed transport properties in GFeFETs, a large Vg capable of switching the spontaneous polarization in PMN-PT is used in the measurement of time evolution of Rs as illustrated in Fig. 5a. Rs is recorded as a function of time during the retention of a Vg of 200 V, which is sufficiently large to ensure complete switching of the polarization. The time-dependent Rs exhibits three distinct steps when Vg is applied. The initial rapid sharp rise of Rs is followed by a slow exponential increase, and finally, Rs reaches saturation and remains unchanged in the high resistance state after a period of about 100 s. The observed high-resistance state at 200 V can be understood from the initial low-resistance state of p-type properties. Graphene is actually p-doped at zero gated by adsorbates of molecules in air. The ferroelectric spontaneous polarization can further enhance the p-doping properties and contribute to the initial low resistance state. The applied Vg of 200 V is large enough to align the polarization in PMN-PT in the upward direction. By considering the Dirac point located at about 100 V for graphene on PMN-PT, the gate voltage of 200 V can generate electron doping in graphene and the generated electrons can act as current carriers. Moreover, the generated electrons normally “screen” the charge carriers resulting in a weaker “trapping” effect. Due to the time-dependent charge trapping and de-trapping at the interface, a steady exponential increase in resistance is actually observed. When Rs reaches a steady high resistance state, the Vg voltage is removed, and Rs is recorded as a function of time as shown in Fig. 5b. The Rs decreases gradually and reaches a saturation regime after 5000 s. The inset of Fig. 5b shows the log scale data of R(t) − Rh, where R(t) is the time-dependent Rs and Rh stands for Rs at the finally saturated high-resistance state. By considering the unique non-volatility in ferroelectric PMN-PT, the ferroelectric polarization should be retained even when the external field is removed.44 The PMN-PT would finally reach the state with remnant polarization, which can contribute to the high resistance in graphene regardless of the slightly shifted dipoles in PMN-PT. The log scale curve fits well with the red dash line, presenting a linear relationship with time as shown in the inset of Fig. 5b. This presents the exponential decay Rs over time. The initial non-linear or non-exponential change can result from a gradual decrease in the ferroelectric polarization of PMN-PT, while the exponential decay suggests the time-dependent charge injection and trapping at the interface between graphene and PMN-PT. Herein, we can conclude that the Rs evolution exhibits three states, namely an initial rapid change, followed by a slow nearly exponential evolution, and finally a saturated state no matter when the large Vg is retained or after it is released. During the entire process, the transport properties of graphene are tuned by multiple factors, including ferroelectric polarization, non-volatility and charge trapping. However, graphene initially reaches a high-resistance state due to the ferroelectric doping and finally reaches a steady high-resistance state due to the ferroelectric non-volatility from initial retention and subsequent release of Vg. Therefore, the ferroelectric effects are dominated in this process when Vg is larger than Vc.
This journal is © The Royal Society of Chemistry 2018 |