1592 IEEE ELECTRON DEVICE LETTERS, VOL. 41, NO. 10, OCTOBER 2020 Ultrahigh Doping of Graphene Using Flame-Deposited MoO3 Sam Vaziri , Member, IEEE, Victoria Chen, Lili Cai, Yue Jiang , Michelle E. Chen, Ryan W. Grady, Xiaolin Zheng, and Eric Pop , Senior Member, IEEE Abstract— The expected high performance of graphene- result, applying conventional semiconductor doping methods based electronics is often hindered by lack of adequate (e.g. substituting C atoms with dopants) is very challenging doping, which causes low carrier density and large sheet for graphene and other 2D materials. resistance. Many reported graphene doping schemes also suffer from instability or incompatibility with existing semi- A more promising route to control the doping of graphene conductor processing. Here we report ultrahigh and stable and other 2D materials is to introduce charge carriers from p-type doping up to ∼7 × 1013 cm−2 (∼2 × 1021 cm−3) of the outside, either with a field-effect (electrostatically or with monolayer graphene grown by chemical vapor deposition. external charge dipoles) or by external dopants (similar to This is achieved by direct polycrystalline MoO3 growth on modulation doping in conventional semiconductors) [10]–[13]. graphene using a rapid flame synthesis technique. With this approach, the metal-graphene contact resistance for For example, charge transfer from external metal-oxide layers holes is reduced to ∼200 · μm. We also demonstrate is one of the most promising mechanisms, being compati- that flame-deposited MoO3 provides over 5× higher doping ble with semiconductor device processing while minimally of graphene, as well as superior thermal and long-term degrading the electronic properties of graphene [14], [15]. stability, compared to electron-beam deposited MoO3. In this doping scheme, a metal-oxide with a different electron Index Terms— Graphene, doping, contact resistance. affinity is brought into contact with graphene, transferring charge carriers across the interface and forming interface I. INTRODUCTION dipoles. For example, p-type doping of graphene can be RAPHENE is a two-dimensional (2D) material with achieved by capping with molybdenum oxide (MoO where unique physics and intrinsic properties [1]. This has x G x ∼ 3) with large electron affinity >6 eV [10]. MoO also resulted in many reports on proof-of-concept electronic and 3 has a sufficiently large band gap (E ∼ 3.7 eV) to be optoelectronic devices with functionalities that cannot be G compatible with transparent electronic applications. However, achieved by conventional bulk materials [2]–[4]. However, due to varying material quality and measurement conditions, many such devices still suffer from issues such as low perfor- reported results have not been consistent [10], [15]–[17]. Thus, mance, poor reliability, device-to-device variation, and lack of the effectiveness of graphene doping remains challenging, reproducibility [5]–[7]. This partly occurs because atomically ∼ while little is known about long-term stability [13]. thin graphene ( 0.335 nm monolayer) is very sensitive to In this letter, we apply MoO using a rapid flame vapor perturbations introduced during or after its integration with 3 deposition technique [18] to effectively induce ultrahigh and other materials [8]. This high sensitivity also makes it chal- stable p-type doping of graphene monolayers grown by chem- lenging to achieve effective, reliable, and uniform doping of ical vapor deposition (CVD). We use electrical measurements, graphene without sacrificing its electronic quality [9]. As a Raman spectroscopy, and modeling to show that a thin layer of ∼5 nm flame-deposited MoO3 on monolayer graphene can Manuscript received July 24, 2020; revised August 11, 2020; accepted − August 18, 2020. Date of publication August 21, 2020; date of current induce a hole density of ∼7×1013 cm 2 and a metal-graphene version September 25, 2020. This work was supported by Bosch through contact resistance of ∼200 · μm. We compare the effective- the Stanford SystemX Alliance. The work of Sam Vaziri was supported by the Knut and Alice Wallenberg Foundation Fellowship. The work ness of this method to the conventional approach of electron- of Ryan W. Grady was supported by the NSF Graduate Research beam (e-beam) evaporated MoO3. The results are averaged Fellowship Program (GRFP) under Grant DGE-1656518. The review of over tens of test structures and show long-term stability. this letter was arranged by Editor D. Shahrjerdi. (Corresponding author: Eric Pop.) Sam Vaziri, Victoria Chen, and Ryan W. Grady are with the Department II. GRAPHENE FIELD-EFFECT TEST STRUCTURES of Electrical Engineering, Stanford University, Stanford, CA 94305 USA. Monolayer graphene was grown on Cu foil using CVD, and Lili Cai, Yue Jiang, and Xiaolin Zheng are with the Department of + Mechanical Engineering, Stanford University, Stanford, CA 94305 USA. subsequently transferred [19] onto 90 nm SiO2 on p Si sub- Michelle E. Chen is with the Department of Materials Science and strates which also serve as back-gates. After defining graphene Engineering, Stanford University, Stanford, CA 94305 USA. channels by photolithography and oxygen plasma, 50 nm-thick Eric Pop is with the Department of Electrical Engineering, Stanford University, Stanford, CA 94305 USA, and also with the Department Pd contacts were formed using e-beam lithography, e-beam of Materials Science and Engineering, Stanford University, Stanford, deposition, and lift-off. Similar steps were repeated to form CA 94305 USA (e-mail: [email protected]). Ti(3 nm)/Au(70 nm) electrical contact pads. Figure 1(a) shows Color versions of one or more of the figures in this letter are available online at http://ieeexplore.ieee.org. a representative transfer length method (TLM) test device Digital Object Identifier 10.1109/LED.2020.3018485 with 8 channel lengths, 500 nm ≤ L ≤ 22 μm. Then, the 0741-3106 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See https://www.ieee.org/publications/rights/index.html for more information. Authorized licensed use limited to: Stanford University. Downloaded on September 27,2020 at 01:12:37 UTC from IEEE Xplore. Restrictions apply. VAZIRI et al.: ULTRAHIGH DOPING OF GRAPHENE USING FLAME-DEPOSITED MoO3 1593 Fig. 1. Transfer characteristics at room temperature. (a) Optical image of a TLM test structure with channel lengths of 0.5 µm, 1 µm, 2 µm, 3 µm, 4 µm, 6 µm, 12 µm and 22 µm. (b) Measured current (ID) vs. gate Fig. 2. Contact and sheet resistance from TLM. (a) Pd-graphene voltage (VGS) of uncapped graphene channels in the TLM. (c) Measured (Gr) contact resistance RC vs. VGS with Gr channel uncapped (blue) and ID-VGS of a graphene (Gr) channel with L = 4 µm: uncapped (black line), capped with flame-grown MoO3 (red). Inset shows TLM plot and fitting capped with evaporated (magenta line) and flame deposited (red line) (dashed lines). (b) Extracted graphene sheet resistance for uncapped MoO3. The blue line shows the same device as in red, after 8 months. and capped graphene channels before and after annealing. Solid lines The schematic shows the crystalline MoO3 interface with graphene. show the average resistance and the shaded regions put bounds on the maximum and minimum values over all the measured TLMs. Black dashes show the average sheet resistance of the flame-deposited MoO3 after annealing. samples were annealed at 250◦C in vacuum (5 × 10−5 Torr). Figure 1(b) shows the transfer characteristics of the back-gated for 30 min. The effect of annealing is discussed in the next graphene field effect transistors (GFETs) making up the TLM section. For the same channel length of 4 μm, the GFET in vacuum. These display very symmetric ambipolar behavior capped with the evaporated MoO [Fig. 1(c), magenta curve] ) 3 with Dirac voltage (V0 near zero and minimal hysteresis, has lower drain current than GFETs capped with the flame- indicating clean, nearly intrinsic graphene channels. deposited MoO3. In previous work [18] we showed that atomically thin MoO3 can be grown on layered materials such as WSe2 and graphene, III. SHEET AND CONTACT RESISTANCE using flame synthesis by van der Waals epitaxy. Unlike e-beam After process development, we performed six MoO3 flame evaporation, the flame-deposited MoO3 on layered materials depositions and two MoO3 evaporation runs. Then we charac- shows a high degree of crystallinity. Here, we use this method terized on average twelve TLM structures [20] for each depo- to deposit thin layers (∼5to10nm)ofMoO3 on the graphene sition run, with Fig. 2 showing extracted contact resistance RC channels to induce strong p-type doping. and sheet resistance Rsh for the uncapped and MoO3-capped Figure 1(c) shows measurements of an uncapped GFET in samples. The inset shows the TLM plot of resistance vs. length vacuum (black line) which exhibits a maximum hole mobility for uncapped devices. In Fig. 2(a), the uncapped devices (blue ∼2700 cm2V−1s−1 and an on/off current ratio of ∼7.5 with line) show relatively symmetric Pd-graphene contacts for elec- a hysteresis |V0|≈0.5 V (not shown). After the flame trons and holes, suggesting Fermi level pinning near the Dirac deposition of MoO3, the transfer characteristics of the same point. When approaching the Dirac voltage, RC increases as device show a significant positive shift of the Dirac voltage the graphene density of states decreases. For holes in doped (beyond the VGS range we can apply safely without damag- graphene (red line), we note RC ∼ 200 · μm at room ing the back-gate oxide), indicative of strong p-type doping temperature, which improves at least by a factor of 2.5 the [Fig. 1(c), red line]. lowest RC achieved in the uncapped devices. This contact Importantly, re-measuring the same device after 8 months resistance is comparable to the lowest reported values for Pd [Fig. 1(c), blue line] shows no significant variation or degra- with exfoliated monolayer graphene [21].