Ultranarrow heterojunctions of armchair-graphene nanoribbons as resonant-tunnelling devices

F. Sánchez-Ochoa*a, Jie Zhangb, Yueyao Dub, Zhiwei Huangb, G. Cantoc, Michael Springborgd and Gregorio H. Cocoletzie
aUniversidad Nacional Autónoma de México, Instituto de Física, Apartado Postal 20-364, Cd. de México 01000, Mexico. E-mail: fsanchez@fisica.unam.mx; franciscosno88@gmail.com
bDepartment of Environmental Science & Engineering, Huaqiao University, Xiamen 361021, People's Republic of China
cCentro de Investigación en Corrosión, Universidad Autónoma de Campeche, Av. Héroe de Nacozari 480, 24079 Campeche, Campeche, Mexico
dPhysical and Theoretical Chemistry, University of Saarland, 66123 Saarbrücken, Germany
eBenemérita Universidad Autónoma de Puebla, Instituto de Física, Apartado Postal J-48, Puebla 72570, Mexico

Received 6th August 2019 , Accepted 29th August 2019

First published on 13th September 2019

A systematic investigation is performed on the electronic transport properties of armchair-graphene nanoribbon (AGNR) heterojunctions using spin-polarized density functional theory calculations in combination with the non-equilibrium Green's function formalism. 9-AGNR and 5-AGNR structures are used to form a single-well configuration by sandwiching a 5-AGNR between two 9-AGNRs. At the same time, these 9-AGNRs are matched at the left and right to electrodes, 9 and 5 being the number of carbon dimers as width. This heterojunction mimics an electronic device with two potential barriers (9-AGNR) and one quantum well (5-AGNR) where quasi-bound states are confined. First, we study the ground state properties, and then we calculate the electron transport properties of this device as a function of the well width. We show the presence of electronic tunnelling resonances between the barriers by delocalized electron density inside the well's structure. This is corroborated by transmission curves, localized densities of states (LDOS), current-vs.-bias voltage results, and the trend of the resonances as a function of the well width. This work shows that carbon AGNRs may be used as resonant-tunnelling devices for applications in nanoelectronics.

1 Introduction

Molecular electronics is a promising research field in nanoelectronics as shown by many successful studies.1–6 In this field, a central issue is the miniaturization process of electronic devices, which is currently limited by the atomic scale. Some kinds of miniaturization make use of two-dimensional (2D)7 materials such as graphene8 resulting in one-dimensional (1D) strips or nanoribbons,9 and nanotubes or nanowires giving metallic chains also as 1D systems, all of them showing fascinating transport phenomena and potential applications as active components as well as interconnects in electronics.10–12 In particular, graphene nanoribbons (GNRs) are a new class of 1D materials that have promising applications in the next generation of electronic and optoelectronic devices.13–16

Theory and experiments show that the electronic and magnetic properties of GNRs can be tuned by varying their geometry, width and chemical composition of the edge termination17 inducing 1D quantum confinement, resulting in semiconductors, metals or spin-polarized semiconductors (half-metals).18–21 On the one hand, GNRs with zigzag edge-termination (ZGNRs) have spin-polarized edge states with applications in spintronics.21 However, GNRs with armchair termination, hereafter AGNRs, have an electronic bandgap which changes as a function of the ribbon width.15,18,19 The AGNRs can be grouped into three different subfamilies, that is, N = 3p, N = 3p + 1 and N = 3p + 2, where p is an integer and N is the number of carbon atom pairs across the GNR.19 N = 3p and N = 3p + 1 families have wide bandgaps that scale inversely with the ribbon width. In contrast, simple models, such as nearest-neighbour tight binding, predict the family N = 3p + 2 to be metallic with a zero bandgap.19

In the last decade, the design and study of more complex 1D nanostructures based on AGNRs, such as lateral heterojunctions, have attracted considerable attention.22–29 Computational and experimental efforts, mainly bottom-up chemical synthesis of GNRs from molecular precursors, have been made to explore quantum effects15,30–34 and demonstrate their usefulness.24–29 This is an interesting issue to address in order to develop new interconnection components with vanishing bandgaps and modern high-speed electronic devices as diodes (traditional p–n junctions)26 and transistors.35

Here, we use the width of AGNRs as a critical factor for bandgap engineering to design 1D lateral-heterojunction devices and then to study quantum resonant-tunnelling. The ground states were obtained using total energy density functional theory (DFT) calculations; these results were combined with the non-equilibrium Green's function (NEGF) formalism to calculate the curves of current-vs.-bias voltage characteristics. Systems with two potential barriers and one quantum well as finite superlattices were studied. This work shows that quasi-bound/resonance states are present in these heterojunctions. These electronic resonances can be tuned as a function of the geometry of devices, in particular through the quantum well width. Through this systematic study, we demonstrate that AGNR heterojunctions may be used as resonant-tunnelling devices in nanoelectronics.

The content of the paper is organized as follows: first, we present the models used and computational details in the Methodology section. Then, we discuss the ground state properties, the transmission curves and density of states (DOS) for the scattering region, and current-vs.-bias curves for the first six devices as a function of the well's width in the Results and discussion section. Finally, the conclusions are drawn in the last section.

2 Methodology section

2.1 Models

The device is sketched in Fig. 1. Two main parts can be identified, the source and drain extensions and the scattering region. Both source and drain are composed of right and left electrodes (shaded-purple areas), respectively, plus buffer atoms enclosed by dashed rectangles. As for electrodes, we take into account the first member, the narrowest AGNR, of the N = 3p + 2 family with p = 1. The leads are narrow bandgap semiconductors with five carbon atoms (5-AGNR). Meanwhile, the potential barriers are built with the third element of the N = 3p family with p = 3, hereafter 9-AGNR. The 9-AGNR structure should be a wider bandgap semiconductor,19 see Fig. 1. Edges (dangling bonds) are saturated with hydrogen atoms. The scattering region contains a segment of 5-AGNR (shaded-orange rectangle) sandwiched by two segments of 9-AGNR (shaded-red rectangle). The highlighted red segments contain two units of 9-AGNR, along the z-direction, acting as potential barriers, while one unit of 5-AGNR is placed in the middle, forming the quantum well. The device in Fig. 1 is the smallest of 16 devices with two potential barriers and one quantum well in this work. The remaining bigger 15 devices contain more units of 5-AGNR in the central region in order to address the role of the width of the quantum well in the transport properties. In addition, we discuss how the number of potential barriers with one 5-AGNR unit quantum well width affects the transmission properties. The xz is the plane of the heterojunction. Visualization of stick models and charge distributions was performed with the VESTA program.36
image file: c9cp04368c-f1.tif
Fig. 1 Stick model of the smallest heterojunction with the source, scattering, and drain regions. The source contact has buffer atoms enclosed by the dashed rectangle plus the left electrode highlighted by a shaded-purple rectangle with two and four 5-AGNR units, respectively. The drain contact is equivalent to the source. Two segments of 9-AGNR (shaded-red rectangles) sandwiching the central region with one 5-AGNR unit (shaded-orange rectangle). The electronic transport is along the z-direction. Carbon and hydrogen atoms are coloured black and white, respectively.

2.2 Computational details

Studies start with the structural relaxation of the electrodes as well as the device using spin-polarized DFT calculations with the SIESTA package.37,38 For the self-consistent process of valence electronic charge density, the states have been expanded in a linear combination of localized numerical atomic orbitals (LCAOs) using a minimal valence ζ (SZ)-basis set. Trouiller–Martins39 norm-conserving pseudopotentials were used to describe the interactions between core and valence electrons. The exchange–correlation energies have been treated with the Perdew–Zunger (PZ) functional in the local spin density approximation (LSDA).40 A 1 × 1 × 25 grid41 and an energy cutoff of 300 Ry for the grid integration, for the whole device (buffer atoms, electrodes, and scattering region), have been applied in the self-consistent calculations. The electron temperature was set equal to 300 K with a Fermi surface broadening. The atomic relaxation was achieved when the force reached the value 20 meV Å−1. The electronic relaxation was converged to 1.0 × 10−4. A vacuum gap of 15 Å in the lateral (x and y) directions has been used to prevent interactions between neighbouring AGNRs in adjacent supercells.

Once the ground state of the devices was determined with DFT, then the electronic transport properties of the optimized systems were calculated with the NEGF formalism as implemented in TranSIESTA.42 First, the calculation of the Hamiltonian for electrodes was done using a 1 × 1 × 1001 k-grid. Then employing the Hamiltonian of electrodes, the current (I) calculation was carried out through the scattering region sandwiched between two semi-infinite electrodes at a finite bias voltage (Vb) using the Landauer–Buttiker formula,43

image file: c9cp04368c-t1.tif(1)
where f is the Fermi distribution function, μL,R are the chemical potentials of the left and right electrodes with μL = EF + Vb/2 and μR = EFVb/2 up and down shifts relative to the EF level, respectively; and T(E,Vb) is the transmission coefficient at energy E and Vb = (μLμR)/e. The T(E,Vb) coefficient from the left to the right electrodes can be calculated by
T(e,Vb) = Tr[ΓL(E,Vb)G(E,Vb)ΓR(E,Vb)G(E,Vb)], (2)
where G(E,Vb) [G(E,Vb)] is the retarded (advanced) Green's function and ΓL,R are the coupling matrices. G0 is the quantized unit of electrical conductance, G0 = e2/h, defined by the elementary charge e and Planck constant h. Buffer atoms have been used to compute the I(Vb) curves, and the converged matrix density from the previous calculation is used to start a new calculation with a higher Vb to achieve a better self-consistent convergence. Finally, the electron transmission function and the current flow through the carbon heterojunctions at finite Vb have been calculated using the post-processing tool TBTrans.42

3 Results and discussion

We start with the discussion of the ground-state properties of the 5-AGNR and 9-AGNR systems, such as energetic stability, electronic bands, and transmission curves, since these structures are references to construct the heterojunctions studied subsequently. First, we find that the anti-ferromagnetic 5-AGNR and 9-AGNR systems are more stable than non-magnetic or ferromagnetic structures. As 5-AGNR, 9-AGNR, and the 16 devices studied here have different quantities of C and H atoms, the total energy is not a good parameter to determine their energetic stability. In this case, the relative formation energy (EF) formalism is used.44–46 EF is a function of the chemical potential of each element. The thermodynamic equilibrium is invoked to determine stability. In our methodology with T = 0 K, the next assumption G(μ1,μ2,…,μN) ≃ E(μ1,μ2,…,μN) can be used to determine the excess or deficit of C and H atoms in the heterojunctions. Thus, EF has the following form:
image file: c9cp04368c-t2.tif(3)
where ETotal is the total energy of the heterojunction under study, and Δni and μi are the deficit (or excess) and chemical potential of the i-th element, respectively, with μC-graphene and μH-molecule. The subscripts C-graphene and H-molecule refer to C atoms in the graphene monolayer and hydrogen molecule, respectively. Nt is the total number of atoms in the 5-AGNR, 9-AGNR or heterojunction. We take the energy of 9-AGNR as the reference energy (Eref). In Table 1, the calculated EF is presented for the first six heterojunctions as examples. As can be seen, the most stable system is 9-AGNR in agreement with the trend of energetic stability as a function of the width of AGNRs. On average, the HET-1 to HET-6 energies are 21.66 meV per atom. These EFs are slightly less stable than that of 9-AGNR, but they are very close (below) to that of the 5-AGNR system. This means that the present heterojunctions are energetically favorable to form as others synthesized22,23 as well as those 5-AGNR and 9-AGNR structures in experiments.25,27,30,47,48
Table 1 Formation energy, EF, in eV per atom and number of electronic resonances, NR, from 0 to 2 eV above the Fermi energy. C-atoms and H-atoms are the total number of atoms in the system. HET stands for heterojunction. The rest 10 heterojunctions have the same average EF
System C-atoms H-atoms EF NR
9-AGNR 22 3 0.0
HET-6 320 112 21.68 5
HET-5 310 108 21.65 4
HET-4 300 104 21.92 4
HET-3 290 100 21.65 3
HET-2 280 96 21.70 3
HET-1 270 92 21.35 2
5-AGNR 10 4 22.02

The electronic properties of 5-AGNR and 9-AGNR unit cells reveal bandgaps in Fig. S1 of the ESI, with a linear dispersion around the Fermi energy for the 5-AGNR system. Let us mention that the sizes of bandgaps in these materials are sensitive to the kind of theory employed.20 The 5-AGNR system can be considered as a metallic system due to the narrow bandgap, which is useful for leads in our device as well as a junction between two semiconducting structures forming two potential barriers. The calculated curve of transmission versus energy, only for spin-up states in Fig. S2 (ESI), shows a small bandgap of 0.2 (0.87) eV for an ideal and periodic device constructed with 5-AGNR (9-AGNR) in both leads and the scattering region. Additionally, a transmission of 1 is observed from ∼±0.1 eV to ∼±1.6 eV in 5-AGNR and from ∼±0.4 eV to ∼±1.0 eV in 9-AGNR, with zero transmission around the Fermi level in Fig. S2 (ESI).

Once we have determined the stability of 5- and 9-AGNR unit cells, and the band structure and transmission properties of ideal devices made of 5-AGNR and 9-AGNR units, we proceed with the calculation of the transmission versus energy for the first six heterojunctions in Table 1. The discussion will be focused on spin-up states, since the results for spin-down states are similar due to anti-ferromagnetic behaviour. In Fig. 2, we plot the square of transmission amplitude (on logarithmic scale) of six devices with two barriers, and one well, built with two 9-AGNR units. A sharp peak is manifested just at the zero eV (Fermi energy), and there are two broadened symmetric electronic resonances at ±0.65 eV corresponding to a single well system (one 5-AGNR unit), see Fig. 2(a). When the heterojunction has two units of 5-AGNR inside the well, the two resonances marked with black filled/empty circles in ±0.5 eV are closer to the Fermi energy. A pair of additional resonances located at ±1.4 eV, now red filled/empty circles, are generated as shown in Fig. 2(b). As the number of units inside the well goes from 3 to 6, the resonances shift towards the Fermi level where the sharp resonance at zero eV is always present. The heterostructure with six 5-AGNR units, see Fig. 2(f), shows five resonances in the valence and conduction bands by symmetry. This finding is clear evidence of quasi-bound states inside the quantum well formed by two potential barriers of 9-AGNR in Fig. 1. Thus, electronic resonances may be tuned by the well's width and, indeed, these resonances may split as a function of the number of potential wells and barriers in the heterojunctions.49 In Fig. S3 (ESI), we show the calculated transmission curves for two heterojunctions with three barriers (2 wells) and four barriers (3 wells) with one unit of 5-AGNR inside the well. The energy splittings of electronic resonances from two and three peaks are clear at ∼±0.6 eV, which therefore means that the heterostructures under study are finite superlattices of AGNRs.49

image file: c9cp04368c-f2.tif
Fig. 2 Square of transmission amplitude as a function of the electron energy for different sizes of the potential well. Only spin-up contributions are plotted by symmetry with spin-down states. From (a)–(f), the central region changes from 1 to 6 units of 5-AGNR. Filled and empty circles label electronic resonances in the valence and conduction regimes, respectively. Black, red, blue, and green denote the first, second, third, and fourth resonance groups near the Fermi level. The Fermi level is set to zero eV. The vertical axes are on the log-scale.

In order to get more insights into electronic transmission with the electronic properties such as the density of states (DOS) at zero eV (equilibrium conditions), we first calculate the DOS only for the scattering region in Fig. 1, using the Green's function formalism.42 Fig. 3 depicts the total DOS for the scattering region of six heterojunctions. The DOS calculated with Green's functions is more useful in this analysis, because quasi-bound/resonance states are differentiated from localized states by the size of peaks and by their position in energy in comparison with the total DOS calculation with the standard diagonalization method and Bloch states. With standard DFT calculations using the diagonalization method, the analysis of electronic states is quite complicated to determine the trend of quasi-bound/resonance and localized states. As shown in Fig. S4 (ESI), peaks at −0.49 and 0.51 eV do not change in position and size for heterojunctions with four, five and six 5-AGNR units forming the well. Others peaks at −1.33 and 1.41 eV change in size but their energetic positions remain the same. These peaks can be ascribed to localized states. On the other hand in Fig. 3, resonances marked by coloured filled/empty circles can be grouped into four sets and identified by small peaks approaching the Fermi energy when the width of the well tends to increase. In some cases, resonances and localized states are mixed, such as those in filled/empty-black and red circles in Fig. 3(b), and filled/empty-blue circles in Fig. 3(e). The mixing of the first resonances with localized states around the Fermi energy is due to the same width of the potential well and barriers in terms of 5-AGNR and 9-AGNR units. The relationship between resonances in transmission curves, in Fig. 2, and those small peaks in the total DOS curves for every heterojunction is in excellent agreement, showing the validity of the use of AGNR heterojunctions as finite superlattices. Also, resonances and localized states in the valence regime are symmetric, in energy and size, like those in the conduction regime around the Fermi level. This implies that these semiconductor-lateral heterojunctions have a type-I or straddling band alignment.50

image file: c9cp04368c-f3.tif
Fig. 3 Density of states for the scattering region calculated with the Green's function formalism. Only spin-up contributions are plotted by symmetry with spin-down states. The width of the quantum well increases from 1 to 6 units in the central region for (a)–(f). Filled and empty circles label electronic resonances in the valence and conduction regimes, respectively. Black, red, blue, and green denote the first, second, third, and fourth resonance groups near the Fermi level. The sharp and high peaks at ∓0.5 eV represent some localized states. The Fermi level is set to zero eV.

In what follows the analysis will focus on the heterojunction of two barriers and one well with six 5-AGNR units. With the knowledge of localized and quasi-bound/resonances states from DOS and transmission plots in Fig. 2 and 3, we then calculate the local DOS (LDOS), using the standard DFT formalism, in order to show the distribution of unoccupied states around specific energies under equilibrium conditions (zero-bias voltage). For this purpose, we employ the energy values of resonances and localized states presented in Fig. 3 for the calculation of conduction states within an energy window of 0.04 eV. These are the first four resonances at 0, +0.23, +0.76 and +1.22 eV; and the localized states at +0.52 and +1.43 eV. At zero eV, the distribution of unoccupied states is through the electrodes and buffer atoms, see the upper panel in Fig. S5 (ESI). In Fig. 4(a) and (c) the resonances at +0.23 and +0.76 eV show electronic states spreading inside the well with a symmetric distribution with quantum numbers n = 1 and n = 2 (n − 1 nodes), respectively. But the unoccupied states at +1.22 eV, see Fig. 4(d), are distributed inside the well with n = 3 (n − 1 nodes) and penetrate the barriers. This is because the heights of the potential barriers are finite, and the quantum confinement is weak owing to the narrow potentials. For each resonance, we can clearly note how the quantum confinement modifies the distribution of unoccupied states inside the 1D quantum well.

image file: c9cp04368c-f4.tif
Fig. 4 Calculated LDOS at selected energy values showing the distributions of unoccupied states in real space. (a) First quasi-bound, (b) first localized, (c) second and (d) third quasi-bound, and (e) second localized states above the Fermi level. Gray (white) sticks represent carbon (hydrogen) atoms. The red-dashed lines enclose (a) one, (c) two and (d) three lobes. The isosurface value is 1.27 × 10−5 e Å−3.

Red dashed rectangles show the quantized states with n = 1–3 of the sixth device. To demonstrate the recurrence of quantum confinement, we calculate the transmission and DOS of the scattering region for devices with seven to sixteen 5-AGNR units forming the quantum well. We plot the energy position of resonances as a function of the width of the quantum well, see Fig. S6 (ESI). Thus for the system with fourteen 5-AGNR units, we investigated the spatial distribution of unoccupied states in LDOS of the first four resonances and plotted them in the bottom panel of Fig. S5 (ESI). We only show the quantized states with n = 1–4, notwithstanding that this system has seven resonances within 0.1–1.6 eV. On the other hand, as depicted by Fig. 4(b) and (e), the localized states at +0.52 and +1.43 eV are highly confined on the 9-AGNR segments, respectively, and further are symmetric. In this context, both electronic quasi-bound and localized states in LDOS are due to C[2pz] orbitals, and the particularity of resonances in transmission studies is due to the edge states in 5-AGNR.9,18,22 Note that the states at the center of C-dimers are absent along the transport direction inside the quantum well. Furthermore, we calculated the electrostatic potential of this system along the transport direction with DFT. Fig. S7 (ESI) depicts the whole device and the planar (top panel) and averaged (bottom panel) electrostatic potentials. It should be noted that potential wells are formed inside 9-AGNRs sandwiched by 5-AGNR systems with an average of 0.96 eV and 0.27 eV, respectively, for the planar average and nanosmoothed one.51 The value of 0.27 eV is in agreement with the energy difference, of 0.33 eV, between the minima of conduction bands of 9-AGNR and 5-AGNR in Fig. S1 and S2 (ESI). The electrostatic potential was calculated for occupied states. A symmetric potential is generated for unoccupied states.

Electronic transport property calculations of the first six heterojunctions are now described. In particular, we have calculated the current-vs.-bias voltage characteristics from 0 to 2.5 V using a symmetrical bias (±Vb/2) applied on the leads. Fig. 5 shows the current results with a similar trend to that found in Fig. 2 and 3 corresponding to resonances. The first small peaks at low bias, marked with empty-black circles, approach the zero-bias voltages as the width of the quantum well increases from one to six 5-AGNR units, in Fig. 2(a)–(f). The bias values where the first maximum current occurs are approximately two times the values of energy in Fig. 2, omitting the universal constant, e, of electric charge.49 To enhance the effect of electronic resonances in electronic transport measurements, we have calculated the differential conductance (dI/dV), displayed in Fig. S8 (ESI), to point out the shift of resonances in terms of the quantum well width. Our findings are in excellent agreement with the behaviour of finite superlattices reported in the literature, with slabs of GaAs/AlGaAs heterostructures,49,52 or in Si/SiGe double-barrier diodes,53 to display a resonant tunnelling effect. Therefore, the heterojunctions made of AGNRs may be considered as finite superlattices with potential applications in nanoelectronics. Now we discuss an interesting non-linear effect in transport properties, i.e., the negative differential conductance (NDC), which is based on regions of negative current with positive voltage. 1D systems based on nanoribbons,54,55 metallic chains56,57 and nanotubes58,59 present the NDC effect due to the suppression or attenuation of conductive channels within a bias window. Comparing those 1D systems with the models considered here, we may mention that the resonant tunnelling is induced by the presence of quasi-bound states inside the quantum well. However, even when the well's width is three times larger than the barriers, for a heterojunction with six 5-AGNR units in Fig. 5(f), no decrease in electric current between the first (∼0.45 V) and second (∼1.5 V) resonant peaks is manifested. Note that Fig. S8(f) (ESI) shows a dI/dV < 0, which is characteristic of NDC, and after that the dI/dV is almost constant with a vanishing value, up to the second resonance where the dI/dV is positive exhibiting a peak (red-point). The absence of NDC in these graphene heterojunctions is due to a saturation effect. To explain the saturation effect, in Fig. 6, we plot the transmission and DOS (only for scattering region) as functions of the bias voltage (applied on two terminals) and energy for the sixth device. Only positive energies are plotted by symmetry. Also, as a reference in Fig. 6, we indicate the bias window, +Vb/2, with a purple (orange) dashed line for the transmission (DOS) curve. In Fig. 6(a), the bias +Vb/2 marked by the purple dashed line intersects electronic resonances at ∼0.25 and ∼0.75 eV in good agreement with the results in Fig. 2, but additional resonances are absent between 0.25 and 0.75 eV, and from 0 to ∼1.5 V.

image file: c9cp04368c-f5.tif
Fig. 5 Spin-up current versus bias voltage as a function of the potential width. The width of the quantum well increases from 1 to 6 units in the central region for (a)–(f). Empty circles label electronic resonances tending to low bias as the width of the quantum well increases. Black and red denote the first and second resonances near the zero bias, respectively.

image file: c9cp04368c-f6.tif
Fig. 6 (a) Square transmission on the log-scale to enhance the resonances, and (b) DOS curves versus energy for various bias values for the heterojunction with six 5-AGNR units in the central region. In this case, the DOS is calculated only for the scattering region. By symmetry, we only present the conduction regime, and also only spin-up contributions are plotted by symmetry with spin-down states. The dashed purple and orange lines denote the bias window, +Vb/2, as a reference line. The Fermi level is set to zero eV.

Note that the resonance at 0 eV splits into two peaks and follows the bias +Vb/2 in Fig. 6(a). On the other hand, looking at the calculated DOS for the scattering region in Fig. 6(b), the localized states at 0.5 and 1.4 eV also split into two branches as the bias voltage increases. In particular, the localized state at 0.5 eV mixes with the resonance at ∼0.25 eV starting from 0.5 V due to the perturbation of both resonance and localized states, and after that, the localized state follows the bias +Vb/2 window. Thus, the localized states saturate the electric current to a constant value, and dI/dV is zero around 1 V, see Fig. S8(f) (ESI). When the applied bias on both electrodes is 1.5 V, the electrons can tunnel once again through the structure. This result can be appreciated by the small peak in the current-vs.-voltage curve of Fig. 5(f). After that the current saturates once again to a constant value. Note that the quantum well width is quite small and as a result the electronic resonances face a larger separation, see Fig. S6 and S8 (ESI).

Finally, we study a graphene heterojunction with the quantum well formed by fourteen 5-AGNR units. With this structure, we have found the second and third resonances below the first localized state, see Fig. S6 (ESI). The length of this device is 15.68 nm formed by buffer atoms, leads, and the scattering region. Transmission and DOS (only for the scattering region) at zero bias are plotted in Fig. 7(a) and (b), respectively. In addition to the electronic resonance at 0 eV in Fig. 7(a), seven resonances can be identified in the range of 0.1–1.6 eV. Note that these resonances and the two well defined localized states are at 0.5 and 1.4 eV, as previously. The sixth resonance is mixed with the second localized state; meanwhile, the seventh resonance is mixed with other conduction states. The width ratio between the well and barrier equals 7 is enough to obtain two consecutive resonances at 0.1 and ∼0.36 eV which are below the energy of the first localized state at 0.5 eV, see Fig. 7(b) and Fig. S6 (ESI). The next step was to calculate the current-vs.-voltage properties. In Fig. 7(c), the electric current shows a linear (ohmic) behaviour for low bias until 0.22 V. The current has a non-linear feature when the bias is greater than 0.22 V. To be precise, the three peaks and valleys are manifestations due to NDC. The dI/dV shows three transitions from positive to negative values in Fig. S9 (ESI). These are fingerprints of NDC in the device.49,52

image file: c9cp04368c-f7.tif
Fig. 7 (a) Square transmission on the log-scale to enhance the resonances, and (b) total density of states curves versus energy under equilibrium conditions, i.e., zero volts, for the heterojunction with fourteen 5-AGNR units in the central region. By symmetry, we only present the conduction regime, and also only spin-up contributions are plotted by symmetry with spin-down states. Blue triangles (red stars) indicate resonances (localized states). (c) Spin-up current-vs.-voltage characteristics with three PVRs in different colours. (d) and (e) are the same plots (a) and (b), respectively, but for various bias voltages. The dashed purple and orange lines denote the bias window, +Vb/2, as a reference line. For plots in (a), (b), (d) and (e) the Fermi level is set to zero eV.

From Fig. 7(c), we can calculate the peak-to-valley current ratios (PVRs), which are crucial parameters for electronic applications of tunnel diodes. In this case, the PVR is calculated by the ratios of peak point – determined by peak voltage (PV) and peak current (PI), and valley point – determined by valley voltage (VV) and valley current (VI), leading to a dynamic resistance. In particular, the devices we are dealing with may be considered as resonant-tunnelling diodes because the quantum tunnelling is maximized when the applied bias is two times the energy of electrons in eV to tunnel. The conductance is negative for positive bias until a VV where the current increases again. As an example, in Fig. 7(c), the device with a quantum well formed by fourteen 5-AGNR units, between two potential barriers built with two 9-AGNRs, shows three PVRs. The PVR(1) is approximately 2, while the PVR(2) and PVR(3) approaches 1. It should be mentioned that because the PVR(1) ≃ 2, the heterojunctions studied here may not be suggested to form any logic device.35,60 However, changes in the geometrical configuration of potential barriers can induce higher PVRs such as wider barriers, or asymmetric barriers with different heights but with the same width.61 These topics will be addressed in future research. Furthermore, let us mention that similar 2D systems with different compositions could be used to achieve higher PVRs, such as TMDCs62 or allotropic forms of phosphorus.63,64

To end this section, we describe the NDC. To do this, we have calculated the transmission and DOS for the scattering region as depicted in Fig. 7(d) and (e), respectively. The baseline of bias, +Vb/2, is marked with dashed lines as a reference. In Fig. 7(d), seven resonances are indicated by vertically shaded segments in yellow, orange and red from 0 → 1.6 eV. The baseline of bias, +Vb/2, is superimposed onto the resonance at the Fermi level as in the previous device with six 5-AGNR units, see Fig. 6(a). It crosses four resonances. This is corroborated by the results in Fig. 7(c), where the fourth resonance is located near +1.5 V. On the other hand, as can be seen in Fig. 7(e), the dashed line of +Vb/2 intersects the first resonance in ∼0.1 eV giving a maximum in current at ∼0.22 V, see Fig. 7(c). After that, the dashed line does not intersect resonances or localized states until 0.4 eV where the second resonance and the first localized state are mixed. At this energy the current increases by 0.23 μA at 0.75 V. Between the first [PV(1) = 0.225 V] and second [PV(2) = 0.75 V] maxima of current, the current undergoes an attenuation up to VI(1) = 0.06 μA. The current attenuation is due to the absence of states to be occupied within 0.10–0.35 eV, as displayed in Fig. 7(e). A similar behaviour happens with the second and third resonances. The localized states do not mix themselves for all the bias. However, they mix with their neighbouring resonances. The aforementioned nonmixing behavior is owing to the steric separation, between potential barriers in this device with fourteen 5-AGNR units, which is greater than that in six 5-AGNR units. Therefore, the geometry and size of these AGNR heterojunctions are important for applications in nanoelectronics and optoelectronics devices.65–67 Additionally, considering defects could be an attractive complementary route to characterize these nanodevices.

4 Conclusions

In conclusion, we have investigated the electronic and transport properties of AGNR heterojunctions using density functional theory and the non-equilibrium Green's function formalism. We have shown that the present one-dimensional graphene systems behave as finite superlattices with potential applications in nanoelectronics as resonant-tunnelling devices. The geometry, edge-termination, and ribbon width are relevant parameters to design heterojunctions with different segments of AGNRs. Potential barriers can be tuned with different heights and quantum wells with tunable widths. Using the width of nanoribbons as a critical factor, we have proved that electronic resonances are present in these devices through the transmission and density of states in the scattering region, in addition to well defined localized states in the potential barriers. The combination of 5-AGNRs and 9-AGNRs has allowed getting barriers high enough to contain quasi-bound/resonances states not obtainable in nanodevices based only on 5-AGNR or 9-AGNR. Also, the transport properties of the combined AGNRs show resonant-tunnelling effects not present in pristine 5-AGNR or 9-AGNR nanodevices. The resonance energy can be tuned by varying the quantum well width. Finally, this effect enhances the negative differential conductance behaviour in the transport properties at low bias. Therefore, devices made of AGNRs can be applied as promising logical devices in nanoelectronics and optoelectronics with low electric power consumption when a large peak-to-valley current ratio is achieved.

Conflicts of interest

There are no conflicts to declare.


The authors gratefully acknowledge the partial financial support of projects Cuerpo Académico Física Computacional de la Materia Condensada (BUAP-CA-191), VIEP-BUAP 100071677-VIEP2019 and CONACYT project 223180. The numerical calculations were carried out using the computer facilities at the Instituto de Física (BUAP).


  1. A. Nitzan, Annu. Rev. Phys. Chem., 2001, 52, 681–750 CrossRef CAS.
  2. N. Agrat, A. L. Yeyati and J. M. Van Ruitenbeek, Phys. Rep., 2003, 377, 81–279 CrossRef.
  3. S. Datta, Quantum transport: Atom to transistor, Cambridge University Press, 2005 Search PubMed.
  4. S. Elke and C. J. Carlos, Molecular electronics: An introduction to theory and experiment, World Scientific, 2017, vol. 15 Search PubMed.
  5. L. Martini, Z. Chen, N. Mishra, G. B. Barin, P. Fantuzzi, P. Ruffieux, R. Fasel, X. Feng, A. Narita and C. Coletti, et al., Carbon, 2019, 146, 36–43 CrossRef CAS.
  6. M. Terrones, A. R. Botello-Méndez, J. Campos-Delgado, F. Lopez-Urias, Y. I. Vega-Cantú, F. J. Rodrguez-Macas, A. L. Elas, E. Munoz-Sandoval, A. G. Cano-Márquez and J.-C. Charlier, et al., Nano Today, 2010, 5, 351–372 CrossRef.
  7. G. Fiori, F. Bonaccorso, G. Iannaccone, T. Palacios, D. Neumaier, A. Seabaugh, S. K. Banerjee and L. Colombo, Nat. Nanotechnol., 2014, 9, 768 CrossRef CAS.
  8. A. K. Geim, Science, 2009, 324, 1530–1534 CrossRef CAS.
  9. A. C. Neto, F. Guinea, N. M. Peres, K. S. Novoselov and A. K. Geim, Rev. Mod. Phys., 2009, 81, 109 CrossRef.
  10. W. Y. Kim, Y. C. Choi and K. S. Kim, J. Mater. Chem., 2008, 18, 4510–4521 RSC.
  11. F. S. Ochoa, Z. Huang, X. Tang, G. H. Cocoletzi and M. Springborg, Phys. Chem. Chem. Phys., 2016, 18, 7442–7448 RSC.
  12. T. Terse-Thakoor, P. Ramnani, C. Villarreal, D. Yan, T.-T. Tran, T. Pham and A. Mulchandani, Biosens. Bioelectron., 2019, 126, 838–844 CrossRef CAS.
  13. A. C. Ferrari, F. Bonaccorso, V. Fal’Ko, K. S. Novoselov, S. Roche, P. Bøggild, S. Borini, F. H. Koppens, V. Palermo and N. Pugno, et al., Nanoscale, 2015, 7, 4598–4810 RSC.
  14. S. Dutta and S. K. Pati, J. Mater. Chem., 2010, 20, 8207–8223 RSC.
  15. P. Kim, M. Y. Han, A. F. Young, I. Meric and K. L. Shepard, 2009 IEEE International Electron Devices Meeting (IEDM), 2009, pp. 1-4.
  16. Z. Chen, Y.-M. Lin, M. J. Rooks and P. Avouris, Phys. E, 2007, 40, 228–232 CrossRef CAS.
  17. X. Zhang, J. Xin and F. Ding, Nanoscale, 2013, 5, 2556–2569 RSC.
  18. K. Nakada, M. Fujita, G. Dresselhaus and M. S. Dresselhaus, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 17954 CrossRef CAS.
  19. Y.-W. Son, M. L. Cohen and S. G. Louie, Phys. Rev. Lett., 2006, 97, 216803 CrossRef.
  20. M. Y. Han, B. Özyilmaz, Y. Zhang and P. Kim, Phys. Rev. Lett., 2007, 98, 206805 CrossRef.
  21. L. F. Huang, G. R. Zhang, X. H. Zheng, P. L. Gong, T. F. Cao and Z. Zeng, J. Phys.: Condens. Matter, 2013, 25, 055304 CrossRef.
  22. D. J. Rizzo, G. Veber, T. Cao, C. Bronner, T. Chen, F. Zhao, H. Rodriguez, S. G. Louie, M. F. Crommie and F. R. Fischer, Nature, 2018, 560, 204 CrossRef CAS.
  23. P. H. Jacobse, A. Kimouche, T. Gebraad, M. Ervasti, J. Thijssen, P. Liljeroth and I. Swart, Nat. Commun., 2017, 8, 119 CrossRef CAS.
  24. Z. Xiao, C. Ma, J. Huang, L. Liang, W. Lu, K. Hong, B. G. Sumpter, A.-P. Li and J. Bernholc, Adv. Theory Simul., 2019, 2, 1800172 CrossRef.
  25. A. Kimouche, M. M. Ervasti, R. Drost, S. Halonen, A. Harju, P. M. Joensuu, J. Sainio and P. Liljeroth, Nat. Commun., 2015, 6, 10177 CrossRef CAS.
  26. J. Cai, C. A. Pignedoli, L. Talirz, P. Ruffieux, H. Söde, L. Liang, V. Meunier, R. Berger, R. Li and X. Feng, et al., Nat. Nanotechnol., 2014, 9, 896 CrossRef CAS.
  27. S. Blankenburg, J. Cai, P. Ruffieux, R. Jaafar, D. Passerone, X. Feng, K. Müllen, R. Fasel and C. A. Pignedoli, ACS Nano, 2012, 6, 2020–2025 CrossRef CAS.
  28. D. Prezzi, D. Varsano, A. Ruini and E. Molinari, Phys. Rev. B: Condens. Matter Mater. Phys., 2011, 84, 041401 CrossRef.
  29. X.-F. Li, L.-L. Wang, K.-Q. Chen and Y. Luo, J. Phys. Chem. C, 2011, 115, 12616–12624 CrossRef CAS.
  30. Y.-C. Chen, T. Cao, C. Chen, Z. Pedramrazi, D. Haberer, D. G. De Oteyza, F. R. Fischer, S. G. Louie and M. F. Crommie, Nat. Nanotechnol., 2015, 10, 156 CrossRef CAS.
  31. A. F. Young and P. Kim, Nat. Phys., 2009, 5, 222 Search PubMed.
  32. H. Sevinçli, M. Topsakal and S. Ciraci, Phys. Rev. B: Condens. Matter Mater. Phys., 2008, 78, 245402 CrossRef.
  33. S. Li, C. K. Gan, Y.-W. Son, Y. P. Feng and S. Y. Quek, Appl. Phys. Lett., 2015, 106, 013302 CrossRef.
  34. G. Saha, A. K. Saha and A. H.-U. Rashid, 2015 IEEE 15th International Conference on Nanotechnology (IEEE-NANO), 2015, pp. 440-443.
  35. J. P. Llinas, A. Fairbrother, G. B. Barin, W. Shi, K. Lee, S. Wu, B. Y. Choi, R. Braganza, J. Lear and N. Kau, et al., Nat. Commun., 2017, 8, 633 CrossRef.
  36. K. Momma and F. Izumi, J. Appl. Crystallogr., 2011, 44, 1272–1276 CrossRef CAS.
  37. P. Ordejón, E. Artacho and J. M. Soler, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 53, R10441 CrossRef.
  38. J. M. Soler, E. Artacho, J. D. Gale, A. Garca, J. Junquera, P. Ordejón and D. Sánchez-Portal, J. Phys.: Condens. Matter, 2002, 14, 2745 CrossRef CAS.
  39. N. Troullier and J. L. Martins, Phys. Rev. B: Condens. Matter Mater. Phys., 1991, 43, 1993 CrossRef CAS.
  40. J. P. Perdew and A. Zunger, Phys. Rev. B: Condens. Matter Mater. Phys., 1981, 23, 5048 CrossRef CAS.
  41. H. J. Monkhorst and J. D. Pack, Phys. Rev. B: Solid State, 1976, 13, 5188 CrossRef.
  42. N. Papior, N. Lorente, T. Frederiksen, A. Garca and M. Brandbyge, Comput. Phys. Commun., 2017, 212, 8–24 CrossRef CAS.
  43. M. Büttiker, Y. Imry, R. Landauer and S. Pinhas, Phys. Rev. B: Condens. Matter Mater. Phys., 1985, 31, 6207 CrossRef.
  44. F. Sánchez-Ochoa, G. H. Cocoletzi and G. Canto, Microporous Mesoporous Mater., 2017, 249, 111–117 CrossRef.
  45. J. Moreno, J. Camacho-Garcia, R. Ponce-Pérez, F. Sánchez-Ochoa, M. T. R. de la Cruz and G. H. Cocoletzi, J. Cryst. Growth, 2019, 507, 370–378 CrossRef CAS.
  46. S. Gutierrez-Ojeda, R. Garcia-Diaz, F. Sanchez-Ochoa, J. Guerrero-Sanchez, L. M. de la Garza, J. Varalda, D. Mosca and G. H. Cocoletzi, Appl. Surf. Sci., 2018, 455, 1078–1085 CrossRef CAS.
  47. N. Merino-Dez, A. Garcia-Lekue, E. Carbonell-Sanroma, J. Li, M. Corso, L. Colazzo, F. Sedona, D. Sánchez-Portal, J. I. Pascual and D. G. de Oteyza, ACS Nano, 2017, 11, 11661–11668 CrossRef.
  48. M. Di Giovannantonio, O. Deniz, J. I. Urgel, R. Widmer, T. Dienel, S. Stolz, C. Sánchez-Sánchez, M. Muntwiler, T. Dumslaff and R. Berger, et al., ACS Nano, 2017, 12, 74–81 CrossRef.
  49. R. Tsu and L. Esaki, Appl. Phys. Lett., 1973, 22, 562–564 CrossRef CAS.
  50. A. Franciosi and C. G. Van de Walle, Surf. Sci. Rep., 1996, 25, 1–140 CrossRef CAS.
  51. J. Junquera, M. H. Cohen and K. M. Rabe, J. Phys.: Condens. Matter, 2007, 19, 213203 CrossRef.
  52. L. Chang, L. Esaki and R. Tsu, Appl. Phys. Lett., 1974, 24, 593–595 CrossRef CAS.
  53. K. Ismail, B. Meyerson and P. Wang, Appl. Phys. Lett., 1991, 59, 973–975 CrossRef CAS.
  54. X.-F. Li, K.-Y. Lian, Q. Qiu and Y. Luo, Nanoscale, 2015, 7, 4156–4162 RSC.
  55. C. Zhang, G. Xiang, M. Lan, Z. Tang, L. Deng and X. Zhang, RSC Adv., 2015, 5, 40358–40362 RSC.
  56. F. Sanchez-Ochoa, G. H. Cocoletzi, G. Canto and N. Takeuchi, RSC Adv., 2015, 5, 91288–91294 RSC.
  57. A. Kumar and P. Ahluwalia, J. Alloys Compd., 2014, 615, 194–203 CrossRef.
  58. F. Léonard and J. Tersoff, Phys. Rev. Lett., 2000, 85, 4767 CrossRef PubMed.
  59. K. H. Khoo, J. Neaton, Y. W. Son, M. L. Cohen and S. G. Louie, Nano Lett., 2008, 8, 2900–2905 CrossRef CAS.
  60. J. M. Marmolejo-Tejada and J. Velasco-Medina, Microelectron. J., 2016, 48, 18–38 CrossRef CAS.
  61. J. H. Davies, The physics of low-dimensional semiconductors: an introduction, Cambridge University Press, 1998 Search PubMed.
  62. H. Li, Z. Yin, Q. He, H. Li, X. Huang, G. Lu, D. W. H. Fam, A. I. Y. Tok, Q. Zhang and H. Zhang, Small, 2012, 8, 63–67 CrossRef CAS.
  63. J. Guerrero-Sánchez, M. Lopez-Fuentes, F. Sánchez-Ochoa, N. Takeuchi and G. H. Cocoletzi, RSC Adv., 2016, 6, 108621 RSC.
  64. M. Buscema, D. J. Groenendijk, S. I. Blanter, G. A. Steele, H. S. Van Der Zant and A. Castellanos-Gomez, Nano Lett., 2014, 14, 3347–3352 CrossRef CAS.
  65. B. Su, V. Goldman and J. Cunningham, Science, 1992, 255, 313–315 CrossRef CAS.
  66. N. Jin, S.-Y. Chung, R. Yu, R. M. Heyns, P. R. Berger and P. E. Thompson, IEEE Trans. Electron Devices, 2006, 53, 2243–2249 CAS.
  67. T. Kubis and P. Vogl, J. Comput. Electron., 2007, 6, 183–186 CrossRef CAS.


Electronic supplementary information (ESI) available. See DOI: 10.1039/c9cp04368c

This journal is © the Owner Societies 2019