betaFeedback

WorksPhysics and Astronomy

Comparison of Optical Response from DFT Random Phase Approximation and Low-Energy Effective Model: Strained Phosphorene

Abstract

The engineering of the optical response of materials is a paradigm that demands microscopic-level accuracy and reliable predictive theoretical tools. Here we compare and contrast the dispersive permittivity tensor, using both a low-energy effective model and density functional theory (DFT). As a representative material, phosphorene subject to strain is considered. Employing a low-energy model Hamiltonian with a Green's function current-current correlation function, we compute the dynamical optical conductivity and its associated permittivity tensor. For the DFT approach, first-principles calculations make use of the first-order random-phase approximation. Our results reveal that although the two models are generally in agreement within the low-strain and low-frequency regime, the intricate features associated with the fundamental physical properties of the system and optoelectronic device implementation such as band gap, Drude absorption response, vanishing real part, absorptivity, and sign of permittivity over the frequency range show significant discrepancies. Our results suggest that the random-phase approximation employed in widely used DFT packages should be revisited and improved to be able to predict these fundamental electronic characteristics of a given material with confidence. Furthermore, employing the permittivity results from both models, we uncover the pivotal role that phosphorene can play in optoelectronics devices to facilitate highly programable perfect absorption of electromagnetic waves by manipulating the chemical potential and exerting strain and illustrate how reliable predictions for the dielectric response of a given material are crucial to precise device design.

introduction

The dynamical finite-frequency optical conductivity and permittivity are the most pivotal quantities in designing optoelectronics devices.[1, 2] Various measurable optical properties such as the complex index of refraction, the reflectivity, and absorptivity are governed directly by the permittivity of the medium, which in turn is directly related to the optical conductivity of a time varying incident electromagnetic (EM) wave.[1, 3] The permittivity also connects the mutual influence of the medium and electric field of the incident EM wave, i.e., the light-matter interactions, and can reveal the precise nature of the medium.[1, 4] Moreover, there is an emerging need for optoelectronic chip architectures that require precise determination and manipulation of the permittivity and optical conductivity to benefit current fabrication techniques and advance technological applications.[2] Only then can fast, ultracompact low-power applications be efficiently realized.

First-principles calculations reside at the frontier of accurate simulations of various materials platforms.[5] For example, density functional theory (DFT) calculations have shown success in simulating the general band gap trend as a function of the number of layers in certain two-dimensional (2D) materials, such as black phosphorus, which constitutes a designed material (with desirable key characteristics such as epsilon-near-zero response [8, 9]), and observed in experiments. [10, 11, 12, 13, 14, 15, 16, 17] Many 2D materials consist of 2D layers of strongly bonded atoms attached to each other in the third dimension by weak forces. These weakly interacting 2D layers allow for designing novel materials with controllable electronics characteristics with low-cost operations, such as layer displacement. Nevertheless, it has proven that differing functionals and approximations in DFT calculations can modify the absolute band-gap of materials. This issue is more pronounced in 2D materials where both strong covalent bonds and weak van der Waals (vdW) forces are present. This is an important point in the context of DFT, which is unable to properly account for vdW forces without incorporating specific corrections. [5, 6, 7]

The weak interlayer vdW interactions provide a unique opportunity to peel off the layers and eventually create a one-atom-thick 2D sheet with drastically different electronics properties than the bulk material. Furthermore, performing mechanical operations such as the exertion of strain, on a single-layer material is much easier, as it responds more effectively to these operations compared to the bulk material. The most famous examples include graphene (a single layer of carbon atoms extracted from graphite)[18] and phosphorene (shown in Figure 1, a single layer of phosphorus atoms extracted from black phosphorus)[19]. Unlike graphene where carbon atoms reside in a single plane, phosphorene atoms reside within two planes with a finite separation distance, making a puckered structure [see Figs. 1(a) and 1(b)]. Compared to bulk black phosphorus, phosphorene acquires a fairly large band gap, ∼1.52\sim1.52 eV, very suitable for semiconductor and field-effect transistor technologies. [16] In the following, we specifically focus on phosphorene (with the possibility of incorporating strain) as its low energy Hamiltonian is available and provides an excellent semiconductor platform for strictly comparing and contrasting the results of DFT and those obtained by the low energy model.

Figure 1. An expanded unit cell of phosphorene. (a) and (b) exhibit the side views of the 2D material along the aa and bb principal axes, whereas (c) shows phosphorene from the top view along the cc-axis. Panel (d) displays the expanded crystal with a generic view. The 2D material is located at the middle of the unit cell with sufficiently large vacuum regions along the cc principal axis.

As the influence of vdW forces in a single layer of black phosphorus weakens, one may expect that the deficiencies in the various DFT simulations described earlier would consequently diminish. However, as we shall see below, DFT with a widely-used functional still underestimates the band gap of phosphorene. On the other hand, a low-energy effective model can incorporate a proper band gap, as it is calibrated through band structure calculations and experimental inputs when parametrizing a particular model. Furthermore, the low-energy effective model can provide precise and deep insights into the fundamental physical properties of the material, such as dominant transitions across the band gap, which are inaccessible in purely computational approaches like DFT.

In this paper, we compute each component of the permittivity tensor of phosphorene subject to in-plane strain. Two methods are used: One involves DFT combined with a random phase approximation (DFT-RPA),[20, 21, 22, 23, 24, 25, 26, 27] and the other uses a low-energy model Hamiltonian with the Green’s function current-current correlator. Our results reveal that the permittivity components calculated from DFT-RPA indicate an anisotropic band gap (direction dependent) with magnitude that is incompatible with the corresponding band structure obtained from DFT and the Perdew-Burke-Ernzerhof (PBE) functional.

The permittivity tensor computed by the low-energy model, however, is fully consistent with the associated band gap and clearly describes the underlying physical characteristics of phosphorene. The low energy model also allows for studying the influence of chemical potential or doping variations. We show that in addition to chemical potential variations, the application of strain provides an effective on/off switching mechanism for the Drude response. The underlaying mechanism is the on/off switching of the intraband transitions that can provide valuable information on the band structure of the system. It should be emphasized that although we have studied a specific 2D semiconductor, our conclusions are generalizable to other materials and point to the urgent need for revisiting DFT-RPA implementations used in many DFT packages. Finally, employing the permittivity data from the DFT and low-energy models, we demonstrate how their differing predictions can influence the precise design of an optoelectronics device. Nevertheless, our findings with both DFT-RPA and low-energy model reveal perfect absorption of electromagnetic waves in layered devices containing phosphorene, which is highly tunable by the chemical potential of phosphorene and the application of strain to the plane of phosphorene.

The paper is organized as follows. In Sec. , the formalisms used in both the DFT-RPA and low-energy models are summarized. In Sec. , the components of the permittivity tensor will be presented and the associated physics will be analyzed through band structure diagrams. It will be discussed how the inaccurate results of DFT-RPA are unable to provide correct information on the microscopic properties of the system. In Sec. , the Drude absorption response, and how it provides information on the band structure will be analyzed and discussed. In Sec. , the results of DFT and low-energy models will be contrasted in a practical device scenario, where the importance of accurate permittivity predictions are paramount to the proper design of a functional optical device. Finally, a summary and concluding remarks will be given in Sec. .

frameworks and formalisms

Below, in Secs. and, the basics of the two approaches, i.e., first-principles DFT-RPA and the low-energy effective model Hamiltonian used in the Green’s function current-current correlator are summarized.

First-principles density functional theory

The density functional calculations are based upon the charge density response to sufficiently weak external interactions, such as an electric field. In this case, the Kohn-Sham equations can be evaluated to obtain the dielectric response of the material. [28] In most DFT packages, such as GPAW, QUANTUM ESPRESSO, and VASP, a random phase approximation (RPA) [20, 21, 22] is implemented to evaluate the dielectric response, or permittivity tensor.[23, 29] Unfortunately, this approximation neglects the exchange-correlation contribution and can lead to unphysical modifications to the original band structure obtained through a specific functional and its associated exchange-correlation. [23] In the linear response regime, the dielectric matrix is given by

ϵG,G′(q,ω)=δG,G′−4π∣q+G∣χG,G′0(q,ω),(1)\epsilon_{G,G'}(\mathbf{q},\omega)=\delta_{G,G'}-\frac{4\pi}{|\mathbf{q}+\mathbf{G}|}\chi^{0}_{G,G'}(\mathbf{q},\omega), \tag*{(1)}

which is linked to the first-order density response χ0\chi^{0}, Bloch vector of the incident wave q\mathbf{q}, reciprocal lattice vectors G\mathbf{G}, and the conventional Kronecker delta δij\delta_{ij}. In the RPA regime, the dielectric function is obtained at the Γ\Gamma point so that

ϵ(q,ω)=1ϵ0,0−1(q,ω).(2)\epsilon(\mathbf{q},\omega)=\frac{1}{\epsilon^{-1}_{0,0}(\mathbf{q},\omega)}. \tag*{(2)}

In this work, first-principles DFT calculations of the dielectric response are performed using RPA as implemented in the GPAW DFT package. [29, 30, 31] The gradient-corrected functional by PBE is used for the exchange-correlation energy when calculating the electronic band structure and the dielectric response. To grid k space on the basis of the Monkhorst-Pack scheme, a sufficiently large value, i.e., 6.0 k-points per Å−1^{-1} is incorporated. The cut-off for the kinetic energy of the plane-waves is set to 800 eV, and 60 unoccupied electronic bands are included with a convergence on the first 50 bands. These high values ensure avoiding any artificial effects due to the application of strain in the subsequent calculations that follow. A small imaginary part is added to the frequency variable throughout the calculations, i.e., η=0.01\eta=0.01 eV, and the width of the Fermi-Dirac distribution is fixed at 0.01 eV.

We introduce the strain parameters siis_{ii} (for the i=x,y,zi=x,y,z directions), to describe the expansion and compression of the atom’s location and unit cell with respect to the relaxed unit cell in each direction, i.e., a=sxxa0a=s_{xx}a_{0}, b=syyb0b=s_{yy}b_{0}, and c=szzc0c=s_{zz}c_{0}. Here aa, bb, and cc are the three strained unit cell axis lengths, and the unstrained unit cell axis lengths are a0a_{0}, b0b_{0}, and c0c_{0}. Therefore, in this notation, sxx=syy=szz=0.9s_{xx}=s_{yy}=s_{zz}=0.9, 1.0, and 1.1 correspond to strains of −10%, 0%, and +10%, respectively. The strain-free expanded unit cell with differing view angles is shown in Figure 1. The phosphorene sheet is located in the aa-bb plane and a large vacuum region is included in the unit cell in the c\mathbf{c} direction. Since periodic boundary conditions in all directions are set in the numerical simulations, the vacuum spacing in the c\mathbf{c} direction ensures

zero overlap of the wave-functions in replicated sheets in the c\mathbf{c} direction. Additionally, as the system is non-magnetic, the permeability is isotropic and can be set to its vacuum value.

Low-energy effective model

To study the permittivity of phosphorene subject to an in-plane strain ϵii\epsilon_{ii} within the effective low-energy regime, we employ the model Hamiltonian presented in Refs. [32] and [33]:

H=∫dk(2π)2ψ^k†H(k)ψ^k=∫dk(2π)2ψ^k†{[u0+αisii+(ηj+βijsii)kj2]τ0+[δ0+μisii+(γj+νijsii)kj2]τx−χykyτy}ψ^k,(3)H=\int\frac{d\mathbf{k}}{(2\pi)^2}\hat{\psi}_{\mathbf{k}}^\dagger H(\mathbf{k})\hat{\psi}_{\mathbf{k}}= \int\frac{d\mathbf{k}}{(2\pi)^2}\hat{\psi}_{\mathbf{k}}^\dagger\left\{\left[u_{0}+\alpha_{i}s_{ii}+(\eta_{j}+\beta_{ij}s_{ii})k_{j}^{2}\right]\tau_{0}+ \left[\delta_{0}+\mu_{i}s_{ii}+(\gamma_{j}+\nu_{ij}s_{ii})k_{j}^{2}\right]\tau_{x}-\chi_{y}k_{y}\tau_{y}\right\}\hat{\psi}_{\mathbf{k}}, \tag*{(3)}

where the indices (i,j)(i,j) run over the coordinates x,yx,y. Here τi\tau_{i} are the Pauli matrices in pseudospin space (atomic sites), and k=(kx,ky)\mathbf{k}=(k_{x},k_{y}) is the momentum. The field operator associated with the Hamiltonian is given by ψ^†(k)=(ψA†,ψB†)\hat{\psi}^{\dagger}(\mathbf{k})=(\psi_{A}^{\dagger},\psi_{B}^{\dagger}), where the pseudospins are labeled by AA and BB. The parameters used for this model are summarized in Table I. This model has also been employed to study superconductivity and supercurrent in strained and magnetized phosphorene systems[34, 35] where it was found that strain can induce Majorana zero energy modes and drive ss-wave and pp-wave superconducting correlations to dd-wave and ff-wave correlations that might explain experimental observations in these contexts[36].

u0u_{0} (eV)δ0\delta_{0} (eV)αx\alpha_{x} (eV)αy\alpha_{y} (eV)μx\mu_{x} (eV)
-0.42+0.76+3.15-0.58+2.65
μy\mu_{y} (eV)ηx\eta_{x} (eV Å2^{2})ηy\eta_{y} (eV Å2^{2})γx\gamma_{x} (eV Å2^{2})γy\gamma_{y} (eV Å2^{2})
+2.16+0.58+1.01+3.93+3.83
βxx\beta_{xx} (eV Å2^{2})βyx\beta_{yx} (eV Å2^{2})βxy\beta_{xy} (eV Å2^{2})βyy\beta_{yy} (eV Å2^{2})
-3.48-0.57+0.80+2.39
νxx\nu_{xx} (eV Å2^{2})νyx\nu_{yx} (eV Å2^{2})νxy\nu_{xy} (eV Å2^{2})νyy\nu_{yy} (eV Å2^{2})χy\chi_{y} (eV)
-10.90-11.33-41.40-14.80+5.25

Table I. Band parameters of phosphorene subject to an in-plane strain [32, 33].

In the low-energy regime, the many-body dielectric response can be expressed by[37]

ϵab(ω)=δab−lim⁡∣q∣→0Πab(ω,q)−Πab(0,q)ϵ0ω2,(4)\epsilon_{ab}(\omega)=\delta_{ab}-\lim_{|\mathbf{q}|\to0}\frac{\Pi_{ab}(\omega,\mathbf{q})-\Pi_{ab}(0,\mathbf{q})}{\epsilon_{0}\omega^{2}}, \tag*{(4)}

in which δab\delta_{ab} is the Kronecker-delta and ϵ0\epsilon_{0} is the vacuum permittivity. The current-current correlation functions are given by:

Πab(ω,q)=e2T∑n∑sTr⁡∫d2p(2π)2JasGs(εn+ωk,p+q)JbsGs(εn,p)∣iωk→ω+iδ(5)\Pi_{ab}(\omega,\mathbf{q})=e^2T\sum_n\sum_s\operatorname{Tr}\int\frac{d^2p}{(2\pi)^2}J_a^sG_s(\varepsilon_n+\omega_k,\mathbf{p}+\mathbf{q})J_b^sG_s(\varepsilon_n,\mathbf{p})\bigg|_{i\omega_k\to\omega+i\delta} \tag*{(5)}

Here Ja,bsJ_{a,b}^s are the components of the current operators in the a,ba,b directions. The components of the Green’s function are labeled GsG_s, and ωk=2πTk\omega_k=2\pi T k and εn=πT(2n+1)\varepsilon_n=\pi T(2n+1) are the bosonic and fermionic Matsubara frequencies, respectively (k,nk,n are integers). Finally, the finite-frequency optical conductivity tensor can be obtained from,

σab(ω)=iωlim⁡∣q∣→0{Πab(ω,q)−Πab(0,q)}.(6)\sigma_{ab}(\omega)=\frac{i}{\omega}\lim_{|\mathbf{q}|\to0}\left\{\Pi_{ab}(\omega,\mathbf{q})-\Pi_{ab}(0,\mathbf{q})\right\}. \tag*{(6)}

Below, we employ separately these two frameworks discussed in Secs. and and compute the components of the permittivity tensor.

results and discussions

This section is divided into three subsections: In Sec. , the various aspects of permittivity and their underlying physical origins will be analyzed by visualizing the associated band structures. In Sec. , the Drude absorption of strained phosphorene will be discussed.[38, 39, 40, 41] In Sec. , the device implications will be presented.

Optical transitions and band gap equivalence

We begin with the DFT-RPA approach to calculate the anisotropic dielectric response and band structure of phosphorene. In Figure 2, the permittivity components εxx(ω)\varepsilon_{xx}(\omega) and εyy(ω)\varepsilon_{yy}(\omega) are shown as a function of frequency of the incident light. The associated band structure along the high-symmetry paths in k\mathbf{k}-space calculated by GPAW is also shown. Biaxial in-plane strains of representative strengths −10%-10\%, 0%0\%, and +10%+10\% are applied to the phosphorene in columns 2(a)-2(c), respectively. The blue and red curves in the top and middle rows of Figure 2 correspond to the real and imaginary parts of the permittivity components (as labeled). Both permittivity components ϵxx(ω)\epsilon_{xx}(\omega) and ϵyy(ω)\epsilon_{yy}(\omega) at zero strain [Figure 2(bFigure 1 and Figure 2(bFigure 2] are nonzero within the low-frequency regime, and approach zero at ω≈4.7\omega\approx4.7 eV. The imaginary parts of ϵxx(ω)\epsilon_{xx}(\omega) and ϵyy(ω)\epsilon_{yy}(\omega) exhibit zero loss at frequencies below ω≈1.5\omega\approx1.5 eV and ω≈0.63\omega\approx0.63 eV, respectively. As the onset of a nonzero imaginary permittivity generally points to photon energies that generate electron interband transitions in semiconductors and insulators, one may conclude that the corresponding band structure of the results shown in Figure 2(bFigure 1 and Figure 2(bFigure 2 should possesses bidirectional band gap on the orders of ω≈1.5\omega\approx1.5 eV and ω≈0.63\omega\approx0.63 eV. Note that the intraband transitions within the valence bands are not allowed due to the Pauli exclusion principle. Next, upon applying a compressive strain of −10%, Figure 2(aFigure 1 and Figure 2(aFigure 2 show that the real part of the permittivity now begins to diverge when ω→0\omega\to0. There are also multiple zero crossings over the given frequency range, and peaks at ω≈0.59\omega\approx0.59 eV [Figure 2(aFigure 1] and ω≈0.9\omega\approx0.9 eV [Figure 2(aFigure 2]. The different threshold frequencies for nonzero imaginary parts, i.e., ω≈0.45\omega\approx0.45 eV and ω≈0.55\omega\approx0.55 eV, suggest anisotropic interband transitions where the electronic transitions in the xx direction experience a larger gap than those occurring in the yy direction. Turning the strain type to tensile with the same magnitude, i.e., +10%, Figure 2(cFigure 1 and Figure 2(cFigure 2, show that the

Figure 2. Real (blue) and imaginary (red) parts of permittivity obtained from first-principles calculations in combination with the RPA (GPAW). The top and middle rows show εxx(ω)\varepsilon_{xx}(\omega) and εyy(ω)\varepsilon_{yy}(\omega), respectively, whereas the bottom row is the band structure of phosphorene along the different paths in k\mathbf{k} space. Column-wise, (a)-(c) correspond to biaxial strains of −10%-10\%, 0%0\%, and +10%+10\%, respectively.

permittivity components exhibit qualitatively similar behavior to those of zero strain shown in Figure 2(bFigure 1 and Figure 2(bFigure 2. As is also seen, the frequency thresholds where the imaginary parts vanish have increased to ω≈1.7\omega\approx1.7 eV and ω≈1.35\omega\approx1.35 eV, compared to the cases with strains of −10%-10\% and 0%0\%, suggesting that there is an increase in the energy gap for the interband transitions.

To confirm the correlation between the band gap transitions and key regions of the frequency dispersion of the permittivity, the band structure of phosphorene along high symmetry paths in kk space is plotted in Figure 2(aFigure 3, Figure 2(bFigure 3, and Figure 2(cFigure 3. The phosphorene layer is subject to the same biaxial strain column-wise. The energies are scaled so that the Fermi level resides at E=0E=0 (marked by the dashed line). As seen in Figure 2(bFigure 3, the unstrained system has a gap of ∼0.67\sim0.67 eV at the Γ\Gamma point. It is known that the band gap of phosphorene can be tuned by the number of layers, from ∼1.5\sim1.5 eV in a monolayer to ∼0.59\sim0.59 eV in a five-stack layer. [46, 47] Also, it was argued that the optical band gap of monolayer BP is around ∼1.5\sim1.5 eV, which is equivalent to a band gap of ∼2.3\sim2.3 eV down-shifted by ∼0.8\sim0.8 eV through the binding energy. [46, 47] Note that to improve band gap predictions, one can repeat the calculations with a hybrid functional, or make use of the GW approximation for the self-energy contribution. [44, 45, 46, 47].

Figure 3. The permittivity components and associated band structure when using the low-energy effective Hamiltonian. The two top rows show the ϵxx(ω)\epsilon_{xx}(\omega) and ϵyy(ω)\epsilon_{yy}(\omega) components, whereas the two bottom rows depict the band structure along the kxk_x and kyk_y directions. In columns (a), (b), and (c) biaxial in-plane strains are applied to the phosphorene system with strengths −10%, 0%, and +10%, respectively.

Although it is known that the bare PBE functional underestimates the band gap of phosphorene (∼1.52\sim1.52 eV) [8], the information extracted earlier from the permittivity components calculated through DFT-PRA are not consistent with this band gap either. By exerting +10%+10\% strain in Figure 2(cFigure 3, the band gap increases to ∼1.83\sim1.83 eV, consistent with the behavior of the permittivity seen in Figure 2(cFigure 1 and Figure 2(cFigure 2, although the intricate features that correlate with the interband transitions at low energies are not consistent with the band structure. With the application of −10%-10\% compressive strain, it is seen in Figure 2(aFigure 3 that the band gap closes and the conduction band at the Γ\Gamma point crosses the Fermi level. Therefore, the associated permittivity should show metallic characteristics at low energies. Indeed, the real part of permittivity in Figure 2(aFigure 1 and Figure 2(aFigure 2 acquires metallic properties with a Drude-type response, centered around ω=0\omega=0, due to the intraband transitions within the conduction band. Nevertheless, the imaginary part of permittivity in Figure 2(aFigure 1 and Figure 2(aFigure 2 does not overlap with the Drude peak, suggesting an anisotropic band gap, which is incompatible with the band structure.

We now discuss the finite-frequency optical conductivity and Drude response within the framework of the low-energy model. To simplify our notation in what follows, we rewrite the low-energy Hamiltonian model (3) by introducing new parameters a1,2,b1,2,c1,2a_{1,2},b_{1,2},c_{1,2}:

H(kx,ky)=(a1+b1kx2+b2ky2)τ0+(a2+c1kx2+c2ky2)τx−χykyτy.(7)H(k_x,k_y)=(a_1+b_1k_x^2+b_2k_y^2)\tau_0 +(a_2+c_1k_x^2+c_2k_y^2)\tau_x-\chi_y k_y\tau_y. \tag*{(7)}

In this notation, the components of Green’s function are given by

G11,22(kx,ky,iω)=12(1iω−f1+g1+1iω−f1−g1),(8)G_{11,22}(k_x,k_y,i\omega)= \frac{1}{2}\left(\frac{1}{i\omega-f_1+g_1}+\frac{1}{i\omega-f_1-g_1}\right), \tag*{(8)}
G12,21(kx,ky,iω)=f2±iχyky2g1(1iω−f1−g1−1iω−f1+g1),(9)G_{12,21}(k_x,k_y,i\omega)= \frac{f_2\pm i\chi_y k_y}{2g_1} \left(\frac{1}{i\omega-f_1-g_1}-\frac{1}{i\omega-f_1+g_1}\right), \tag*{(9)}

where the variables f1f_1, f2f_2, and g1g_1 are given by,

f1=a1+b1kx2+b2ky2,(10)f_1=a_1+b_1k_x^2+b_2k_y^2, \tag*{(10)}
f2=a2+c1kx2+c2ky2,(11)f_2=a_2+c_1k_x^2+c_2k_y^2, \tag*{(11)}
g1=f22+χy2ky2.(12)g_1=\sqrt{f_2^2+\chi_y^2k_y^2}. \tag*{(12)}

Substituting the Green’s function components (9) into Eq. (6), we obtain the real part of the optical conductivity tensor, expressed in terms of Dirac-delta functions:

σab(ω)=e24πℏω∬dΩ dkx dky F(Ω,ω,μ,T){hab[δ(Ω−f1+g1)δ(ω+Ω−f1−g1)+δ(Ω−f1−g1)δ(ω+Ω−f1+g1)]+gˉabδ(Ω−f1+g1)δ(ω+Ω−f1+g1)+gˉab+δ(Ω−f1−g1)δ(ω+Ω−f1−g1)}.(13)\sigma_{ab}(\omega)=\frac{e^2}{4\pi\hbar\omega}\iint d\Omega\,dk_x\,dk_y\,\mathcal{F}(\Omega,\omega,\mu,T) \left\{ h_{ab}\left[\delta(\Omega-f_1+g_1)\delta(\omega+\Omega-f_1-g_1) +\delta(\Omega-f_1-g_1)\delta(\omega+\Omega-f_1+g_1)\right] +\bar{g}_{ab}\delta(\Omega-f_1+g_1)\delta(\omega+\Omega-f_1+g_1) +\bar{g}^{+}_{ab}\delta(\Omega-f_1-g_1)\delta(\omega+\Omega-f_1-g_1) \right\}. \tag*{(13)}

The temperature dependence of the optical conductivity in the continuum regime is given by F(Ω,ω,μ,T)=f(Ω−μ,T)−f(Ω+ω−μ,T)\mathcal{F}(\Omega,\omega,\mu,T)=f(\Omega-\mu,T)-f(\Omega+\omega-\mu,T), in which μ\mu stands for the chemical potential and f(X,T)f(X,T) is the Fermi-Dirac distribution at temperature TT. Also, we have introduced the following new variables to further simplify the final expressions

hxx=4kx2(c12χy2ky2f22+χy2ky2),(14)h_{xx}=4k_x^2\left(\frac{c_1^2\chi_y^2k_y^2}{f_2^2+\chi_y^2k_y^2}\right), \tag*{(14)}
gxx±=4kx2(b12+c12f22f22+χy2ky2±2b1c1f2f22+χy2ky2),(15)g^{\pm}_{xx}=4k_x^2\left(b_1^2+\frac{c_1^2f_2^2}{f_2^2+\chi_y^2k_y^2}\pm\frac{2b_1c_1f_2}{\sqrt{f_2^2+\chi_y^2k_y^2}}\right), \tag*{(15)}
hxy=2c1kxkyχy22c2ky2−f2f22+χy2ky2,(16)h_{xy}=2c_1k_xk_y\chi_y^2\frac{2c_2k_y^2-f_2}{f_2^2+\chi_y^2k_y^2}, \tag*{(16)}
gxy±=2kxkyf22+χy2ky2(c1f2±b1f22+χy2ky2)(2c2f2±2b2f22+χy2ky2+χy2),(17)g^{\pm}_{xy}=\frac{2k_xk_y}{f_2^2+\chi_y^2k_y^2} \left(c_1f_2\pm b_1\sqrt{f_2^2+\chi_y^2k_y^2}\right) \left(2c_2f_2\pm2b_2\sqrt{f_2^2+\chi_y^2k_y^2}+\chi_y^2\right), \tag*{(17)}
hyy=(f2−2c2ky2)χy2f22+χy2ky2.(18)h_{yy}=\frac{(f_2-2c_2k_y^2)\chi_y^2}{f_2^2+\chi_y^2k_y^2}. \tag*{(18)}
gyy±=ky2(2b22+(2c2f2+χy2)2f22+χy2ky2±4b2(2c2f2+χy2)f22+χy2ky2)(19)g_{yy}^{\pm}=k_y^2\left(2b_2^2+\frac{(2c_2f_2+\chi_y^2)^2}{f_2^2+\chi_y^2k_y^2}\pm\frac{4b_2(2c_2f_2+\chi_y^2)}{\sqrt{f_2^2+\chi_y^2k_y^2}}\right) \tag*{(19)}

Here hyx=hxyh_{yx}=h_{xy}, gyx±=gxy±g_{yx}^{\pm}=g_{xy}^{\pm}, and the functions f1f_1 and f2f_2 are even functions of momenta kxk_x and kyk_y. Therefore, habh_{ab} and gab±g_{ab}^{\pm} determine the symmetry of the optical conductivity integrand Eq. (13) with respect to momenta. As is seen, the integrands of σxy(ω)\sigma_{xy}(\omega) and σyx(ω)\sigma_{yx}(\omega) are odd functions of momenta due to the aforementioned symmetry properties of hxyh_{xy}, hyxh_{yx}, gxy±g_{xy}^{\pm}, and gyx±g_{yx}^{\pm}. Hence, without performing any further calculations, we find that σxy(ω)=σyx(ω)=0\sigma_{xy}(\omega)=\sigma_{yx}(\omega)=0 in this system. On the other hand, hxxh_{xx}, gxx±g_{xx}^{\pm}, hyyh_{yy}, and gyy±g_{yy}^{\pm} are even functions of momenta, and determine the diagonal optical conductivity tensor components σxx(ω)\sigma_{xx}(\omega) and σyy(ω)\sigma_{yy}(\omega). The real part of the optical conductivity tensor, Eq. (6), is a complicated function of frequency and momenta that must be evaluated numerically. In what follows, we first compute the optical conductivity as a function of ω\omega and then obtain the components of permittivity through Eq. (4).

In Figure 3, a study comparable to Figure 2 is shown, except now we implement the method based on the low-energy effective Hamiltonian, Eq. (3). The parameters for the Hamiltonian are obtained through fitting the model Hamiltonian to the band structure obtained from first-principles around the Γ\Gamma point (summarized in Table I). The corrected band gap used in the low-energy Hamiltonian is on the order of 1.52 eV, although the magnitude of the band gap plays no role in our conclusions. Comparing Figure 2(bFigure 1 and Figure 2(bFigure 2 to Figure 3(bFigure 1 and Figure 3(bFigure 2, we see that the two approaches share similarities at low energies (ω≲3\omega\lesssim3 eV). For example, the generic behaviors are similar, namely, ϵxx(ω)\epsilon_{xx}(\omega) has a flat, smooth variation with ω\omega while ϵyy(ω)\epsilon_{yy}(\omega) has a clearly defined peak at low frequencies. Also, the strongly anisotropic nature of phosphorene is exhibited by the vastly different frequency-dependence of ϵxx(ω)\epsilon_{xx}(\omega) and ϵyy(ω)\epsilon_{yy}(\omega). Both DFT-RPA and the low-energy Hamiltonian model show some similar trends, i.e., the magnitudes follow ∣ϵxx(ω)∣≪∣ϵyy(ω)∣|\epsilon_{xx}(\omega)|\ll|\epsilon_{yy}(\omega)| within low energies. There are however significant quantitative differences between the two approaches. The origins of these disagreements between the two are two-fold: First, the threshold value for nonzero imaginary permittivity in both ϵxx(ω)\epsilon_{xx}(\omega) and ϵyy(ω)\epsilon_{yy}(\omega) obtained through the low-energy model are identical and equal to ω=1.52\omega=1.52 eV, unlike the different values obtained using DFT-RPA. Hence the low-energy effective model suggests that the same band gap exists in both directions. Second, we find from the low-energy model that the real part of ϵyy(ω)\epsilon_{yy}(\omega) vanishes at ω=2\omega=2 eV despite the fairly large nonzero imaginary part of ϵyy(ω)\epsilon_{yy}(\omega) at the same frequency. This feature is absent in the DFT-RPA results and can play a pivotal role in devising novel optoelectronics devices that are sensitive to loss.

Next, we incorporate strain, beginning with a +10% tensile strain [Figure 3(cFigure 1 and Figure 3(cFigure 2]. It is observed that now the frequency cutoff for a nonzero imaginary permittivity increases to ω=2.48\omega=2.48 eV, suggesting an enlarged band gap. Also, the permittivity is now nonzero over a larger interval of frequencies, indicating a flattening of the conduction and valence bands following application of this type of strain. Reversing the strain direction, Figure 3(aFigure 1 and Figure 3(aFigure 2 display the permittivity components subject to −10%-10\% compressive in-plane strain. As seen, both the real and imaginary parts possess Drude absorption peaks, when ω→0\omega\to0, indicating metallic behavior. Unlike the DFT-RPA results in Figure 2(aFigure 1 and Figure 2(aFigure 2, within the low-energy regime, the imaginary component of the permittivity has a diverging Drude response for low frequencies, and a secondary peak appears at ω=1.33\omega=1.33 eV and ω=1.30\omega=1.30 eV, for ϵxx(ω)\epsilon_{xx}(\omega) and ϵyy(ω)\epsilon_{yy}(\omega), respectively.

In order to fully understand these features, we have plotted the band structure associated with the low-energy effective model along both the kxk_x and kyk_y directions. For the strain-free case, the band structure in Figure 3(bFigure 3 and Figure 3(bFigure 4 illustrates that the bottom of valence band and the top of conduction band are separated by a gap of 1.52 eV for both directions. This clearly explains the identical threshold values for the nonzero

Figure 4. The normalized real part of the optical conductivity for a strain-free system obtained using the low-energy model. Panels (a) and (b) show the diagonal components of the optical conductivity tensor as a function of frequency. The chemical potential takes the values, μ=0,0.34,0.71,1.08,1.45\mu=0,0.34,0.71,1.08,1.45 eV.

imaginary permittivities shown in Figure 3(bFigure 1 and Figure 3(bFigure 2. The exertion of +10%+10\% tensile strain in Figure 3(cFigure 3 and Figure 3(cFigure 4, increases the band gap to 2.48 eV for both the kxk_x and kyk_y directions, and results in smaller band curvatures compared to unstrained phosphorene. This accounts for the nonzero imaginary permittivity for frequencies beyond the threshold ω=2.48\omega= 2.48 eV in Figure 3(cFigure 1 and Figure 3(cFigure 2. For a compressive strain of −10%-10\%, Figure 3(aFigure 3 and Figure 3(aFigure 4 show a closing of the gap, and the valence band now crosses the Fermi level. This crossing allows for intraband transitions, and thus the Drude peak for very low frequencies, ω→0\omega\to0, emerges. As seen, the band curvature now has further increased, resulting in a suppressed peak in the permittivity components [Figure 3(aFigure 1 and Figure 3(aFigure 2]. Also, the two transitions at the energies of ω=1.33\omega= 1.33 eV and ω=1.30\omega= 1.30 eV (marked in Figure 3(aFigure 3 and Figure 3(aFigure 4) follow from the anisotropic band curvatures in the kxk_x and kyk_y directions. Note that one is unable to make an immediate conclusion for identifying the locations of these peaks by looking at the band structure because ϵxx,yy(ω)\epsilon_{xx,yy}(\omega) are obtained by integrating over kxk_x and kyk_y. Unlike the zero strain and tensile strain cases, these differing transitions appear as peaks with differing locations in both permittivity components presented in Figure 3(aFigure 1 and Figure 3(aFigure 2.

We have performed the DFT-RPA calculations and obtained the permittivity components for a few materials and semiconductors with moderate band gaps. Our results reveal that the issues described here for the case of strained phosphorene also appear for other material platforms. This suggests that the adverse effects inherited from the RPA approach creates discrepancies that are generalizable to other systems.

Optical conductivity and Drude weight

For completeness, Figure 4 presents the absorptive components of the dynamical optical conductivity Re⁡{σxx(ω)}\operatorname{Re}\{\sigma_{xx}(\omega)\} and Re⁡{σyy(ω)}\operatorname{Re}\{\sigma_{yy}(\omega)\} as a function of frequency in the strain-free system. We normalize each component by σ0=e2/8ℏ\sigma_0 = e^2/8\hbar. To sample the different regions of the band structure, several representative values of the chemical potential are chosen. When the chemical potential is equal to zero or a value within the band gap, e.g., μ=0.34\mu= 0.34 eV (see Figure 3(bFigure 3 and Figure 3(bFigure 4), the optical conductivity is zero at low frequencies and then sharply rises at ω=1.52\omega= 1.52 eV (the band gap magnitude), corresponding to the absorption onset. In other words, the onset of nonzero optical absorption is controlled by the band gap edges. The associated transitions are schematically shown by arrows in Figure 4(a) and Figure 4(b). By increasing the chemical potential, the Drude absorption peak persists as ω→0\omega\to0, and the onset of nonzero optical absorption shifts to higher values of ω\omega. Note that when μ=0\mu= 0 and μ=0.34\mu= 0.34 eV, the optical conductivities do not exhibit low-frequency divergences, as those energies reside inside

the band gap. Meanwhile, the small σxx,yy(ω)\sigma_{xx,yy}(\omega) that is observed at low frequencies for μ=0.34\mu= 0.34 eV is due to a small imaginary term η=0.01\eta= 0.01 eV, added to the frequencies for numerical stability, and is physically equivalent to inelastic scattering. When the chemical potential crosses the valence band at a representative value, e.g., μ=0.71\mu= 0.71 eV, and larger values (see Figure 3(bFigure 3 and Figure 3(bFigure 4), the Drude response acquires more pronounced values as ω→0\omega\to0. Similar to the components of the permittivity tensor, the magnitudes of the components of the optical conductivity tensor obey ∣σxx(ω)∣<∣σyy(ω)∣|\sigma_{xx}(\omega)| < |\sigma_{yy}(\omega)|.

From observations of the strain-free case, it is straight-angled to conclude that the discrepancies observed in the DFT-RPA calculations are not present in the optical conductivity. forward to understand how strain affects the dynamical optical conductivity. In the presence of e.g., +10% tensile strain, the optical conductivity has the same structure as the strain-free case except now the band gap increases to 2.48 eV (see Figure 3(cFigure 3 and Figure 3(cFigure 4). Conversely, a −10% compressive strain closes the band gap and, therefore, the low-frequency Drude response appears (even when μ=0\mu= 0), with unequal dissipation threshold values, i.e., ω=1.33\omega= 1.33 eV and ω=1.30\omega= 1.30 eV at the first peak of the optical conductivity components σxx(ω)\sigma_{xx}(\omega) and σyy(ω)\sigma_{yy}(\omega), respectively. This anisotropy originates again, from the differing curvatures of the valence and conduction bands in different directions, causing the different interband transition gaps shown in Figure 3(aFigure 3 and Figure 3(aFigure 4.

The Drude weight of the anisotropic optical response can be obtained by integrating the Drude response part of the optical conductivity near ω→0\omega\to0: Dxx=lim⁡ω→0⟨σxx(ω)⟩ωD_{xx} = \lim_{\omega\to0}\langle\sigma_{xx}(\omega)\rangle_{\omega}, and Dyy=lim⁡ω→0⟨σyy(ω)⟩ωD_{yy} = \lim_{\omega\to0}\langle\sigma_{yy}(\omega)\rangle_{\omega}. Thus, Dxx,yyD_{xx,yy} gives a weight to the zero-frequency divergence of the optical conductivity and is closely associated with intraband transitions. In Figure 5, we illustrate the Drude weight for both components of the optical conductivity σxx(ω)\sigma_{xx}(\omega) and σyy(ω)\sigma_{yy}(\omega) as a function of chemical potential μ\mu, for three biaxial strain values of −10%, 0%, and +10%. The calculations shown in Figure 5(a) and Figure 5(b) reaffirm the anisotropic Drude response in this system. Nevertheless, the threshold chemical potential where the Drude weight becomes nonzero, is the same for both the xx and yy directions. Note that this threshold value for μ\mu, determines the distance between the Fermi level and the bottom of conduction band. When there is a compressive strain on phosphorene, there is a Drude response, even at μ=0\mu= 0. The remaining strain cases show that the distance between the Fermi level and the bottom of the conduction band is 0.32 eV and 1.04 eV in the presence of 0% and +10% strain, which are in excellent agreement with the band structure diagrams in Figure 3. To illustrate the nonuniformity of the Drude response, the Drude weight curves are shifted to the origin in Figure 5(c) and Figure 5(d). As seen, the steepest response belongs to DyyD_{yy} for the case of +10% strain, whereas DxxD_{xx} has a moderate response for the same strain value. This counterintuitive finding cannot be deduced by simple examination of the conduction bands in Figure 3. This is due to the fact that the band structure is nonuniform in the kxk_x-kyk_y plane (see e.g., the isoenergy contour plots presented in Ref. 36), and to obtain the Drude weight, the intraband transitions are integrated over the entire isoenergy curves in the kxk_x-kyk_y plane. Nevertheless, to confirm these findings, we have performed checks by summing up the contributions from the vertical intraband transitions through the following formula: [37, 42]

Figure 5. The Drude weight is shown as a function of chemical potential μ\mu. Three values of biaxial strain are considered: −10%-10\%, 0%0\%, and +10%+10\%. The panels (a) and (b) show the Drude weight for the conductivity components σxx(ω)\sigma_{xx}(\omega) and σyy(ω)\sigma_{yy}(\omega), respectively. For easier comparison, (c) and (d) show each curve from (a) and (b) shifted to begin at the origin.

lim⁡ω→0⟨σxx(yy)(ω)⟩∝∫∑kx(y)i∣dE(kx(y),ky(x))dkx(y)∣ dky(x).(20)\lim_{\omega\to0}\langle\sigma_{xx(yy)}(\omega)\rangle\propto\int\sum_{k^i_{x(y)}} \left|\frac{dE(k_{x(y)},k_{y(x)})}{dk_{x(y)}}\right|\,dk_{y(x)} . \tag*{(20)}

As seen, this formula accounts for the vertical intraband transitions through the slope of the conduction band E(kx,ky)E(k_x,k_y) at E(kx,ky)=μE(k_x,k_y) = \mu, where kx(y)ik^i_{x(y)} are the roots of E(kx,ky)−μ=0E(k_x,k_y) - \mu= 0. The total Drude weight is proportional to the integration of all these states over ky(x)k_{y(x)}. These calculations were found to have perfect agreement with those presented in Figure 5.

Implications for device design

We now demonstrate that having an accurate microscopic model for predicting the optical response of a material is crucial for the successful design of even a simple optics device, which in this case involves phosphorene. As seen in Figure 6, the basic design involves a phosphorene layer deposited on top of an insulator layer with thickness dd, and a perfect conductor, serving as a back plate. The incident EM wave, propagating through vacuum, impinges on the device from the phosphorene side with an angle of θ\theta measured from the normal to the phosphorene plane. In general, the permittivity tensor εˉˉ\bar{\bar{\varepsilon}} and permeability tensor μˉˉ\bar{\bar{\mu}} in the principal coordinates take the following biaxial forms for a given (uniform) region (n=0,1,2n = 0,1,2):

Schematic of the configuration involving a single-layer of phosphorene with an effective thickness t on top of a spacer layer and metal reflecting substrate

Figure 6. Schematic of the configuration involving a single-layer of phosphorene with an effective thickness tt on top of a spacer layer of thickness dd, and metal reflecting substrate. The phosphorene resides in the xx-yy plane. The phosphorene layer is exposed to an electromagnetic wave from the vacuum region, where the incident electric field is polarized along y^\hat{y}, and the magnetic field is polarized in the xx-zz plane. The incident wavevector k0\mathbf{k}_0 makes an angle θ\theta with the zz axis. The crystallography principal directions are set as follows: x≡ax \equiv a, y≡by \equiv b, and z≡cz \equiv c.

εˉˉn=εnxx^x^+εnyy^y^+εnzz^z^,(21)\bar{\bar{\varepsilon}}_n = \varepsilon_{nx}\hat{x}\hat{x} + \varepsilon_{ny}\hat{y}\hat{y} + \varepsilon_{nz}\hat{z}\hat{z}, \tag*{(21)}
μˉˉn=μnxx^x^+μnyy^y^+μnzz^z^,(22)\bar{\bar{\mu}}_n = \mu_{nx}\hat{x}\hat{x} + \mu_{ny}\hat{y}\hat{y} + \mu_{nz}\hat{z}\hat{z}, \tag*{(22)}

where nn denotes the vacuum region (n=0n = 0), the phosphorene layer (n=1n = 1), or the spacer layer (n=2n = 2).

We now investigate the absorption of EM waves from the layered configuration shown in Figure 6. The metallic back plate is taken to have perfect conductivity (PEC) for simplicity. The electric field of the incident wave is polarized along y^\hat{y}, and is incident from the vacuum region with wavevector k0\mathbf{k}_0 in the xx-zz plane: k0=x^k0x+z^k0z\mathbf{k}_0 = \hat{x}k_{0x} + \hat{z}k_{0z}, where k0x=k0sin⁡θk_{0x} = k_0 \sin\theta, and k0z=k0cos⁡θk_{0z} = k_0 \cos\theta. Since εˉˉn\bar{\bar{\varepsilon}}_n has no off-diagonal components, the transverse electric (TE) and transverse magnetic (TM) modes are decoupled. The absorption can be calculated from Maxwell’s equations. Assuming a harmonic time dependence exp⁡(−iωt)\exp(-i\omega t) for the EM field, we have,

∇×En=+iωμ0μˉˉn⋅Hn,∇×Hn=−iωε0εˉˉn⋅En.(23)\nabla\times\mathbf{E}_n = +i\omega\mu_0\bar{\bar{\mu}}_n\cdot\mathbf{H}_n, \\ \nabla\times\mathbf{H}_n = -i\omega\varepsilon_0\bar{\bar{\varepsilon}}_n\cdot\mathbf{E}_n. \tag*{(23)}

Combining Eqs. (23), we get,

∇×(μˉˉn−1⋅∇×E)=k02(εˉˉn⋅E),∇×(εˉˉn−1⋅∇×H)=k02(μˉˉn⋅H).(24)\nabla\times\left(\bar{\bar{\mu}}_n^{-1}\cdot\nabla\times\mathbf{E}\right) = k_0^2\left(\bar{\bar{\varepsilon}}_n\cdot\mathbf{E}\right), \\ \nabla\times\left(\bar{\bar{\varepsilon}}_n^{-1}\cdot\nabla\times\mathbf{H}\right) = k_0^2\left(\bar{\bar{\mu}}_n\cdot\mathbf{H}\right). \tag*{(24)}

We consider TE modes, corresponding to non-zero field components EnyE_{ny}, HnxH_{nx}, and HnzH_{nz}. The electric field EnyE_{ny} satisfies the following wave equation:

1μnz∂2Eny∂x2+1μnx∂2Eny∂z2+k02εnyEny=0,(25)\frac{1}{\mu_{nz}}\frac{\partial^2 E_{ny}}{\partial x^2}+\frac{1}{\mu_{nx}}\frac{\partial^2 E_{ny}}{\partial z^2}+k_0^2\varepsilon_{ny}E_{ny}=0, \tag*{(25)}

which admits separable solutions of the form ψ(z)exp⁡(ik0xx)\psi(z)\exp(ik_{0x}x). In what follows, we consider non-magnetic media, so that μnx=μnz=1\mu_{nx} = \mu_{nz} = 1. The parallel wave-vector k0xk_{0x} is determined by the incident wave, and is conserved across the interface. The form of ψ(z)\psi(z) then simply involves linear combinations of the exponential exp⁡(iknzz)\exp(ik_{nz}z) for a given region. Thus, the electric field in the vacuum region, E0E_0, is written in terms of incident and reflected waves: E0=(eik0zz+E0rye−ik0zz)eik0xxy^\mathbf{E}_0 = (e^{ik_{0z}z}+E_{0ry}e^{-ik_{0z}z})e^{ik_{0x}x}\hat{y}. From the electric field, we can use Eq. (23) to easily deduce the magnetic field components. Due to the presence of the perfect metal plate, in the spacer region, the general form of the electric field is written in terms of standing waves: E2y=E2sin⁡[k2(z−d)]eik0xxE_{2y}=E_2\sin[k_2(z-d)]e^{ik_{0x}x}, where from Eq. (25), the wave number k2k_2 is given by, k2=±ε2k02−k0x2k_2=\pm\sqrt{\varepsilon_2k_0^2-k_{0x}^2}. Note that the boundary condition that E2yE_{2y} vanishes at the ground plane (z=dz=d) is accounted for (see Figure 6). To construct the H\mathbf{H} fields we use Eqs. (23) to arrive at, Hn=(−∂zEny,∂xEny)/(iη0k0)\mathbf{H}_n=(-\partial_zE_{ny},\partial_xE_{ny})/(i\eta_0k_0), for n=0,2n=0,2, and where η0=μ0/ε0\eta_0=\sqrt{\mu_0/\varepsilon_0} is the impedance of free space. The presence of phosphorene enters in the boundary condition for the tangential component of the magnetic field by writing,

n^×(H0−H2)=J,(26)\hat{n}\times(\mathbf{H}_0-\mathbf{H}_2)=\mathbf{J}, \tag*{(26)}

where n^\hat{n} is the normal to the vacuum/phosphorene interface, and J\mathbf{J} is the current density. Thus, we have H2x(z=0)−H0x(z=0)=JyH_{2x}(z=0)-H_{0x}(z=0)=J_y, where Ohm’s law connects the surface current density JyJ_y in the phosphorene layer to the electric field in the usual way: J=σˉˉE\mathbf{J}=\bar{\bar{\sigma}}\mathbf{E}. The dielectric tensor in turn is defined through the surface conductivity tensor σˉˉ\bar{\bar{\sigma}} via,

εab(ω)=δab+iσab(ω)tε0ω,(27)\varepsilon_{ab}(\omega)=\delta_{ab}+\frac{i\sigma_{ab}(\omega)}{t\varepsilon_0\omega}, \tag*{(27)}

where tt is the effective thickness of the phosphorene layer, which we take to be ≈1 nm\approx1\,\mathrm{nm} (see Figure 6). One can also consider the phosphorene layer as a finite sized slab, like the spacer layer, and solve for the fields within the layer. This approach leads to equivalent results, but treating the phosphorene layer as a current sheet with infinitesimal thickness leads to simpler expressions. Upon matching the tangential electric fields at the vacuum/spacer interface, and using Eq. (26), it is straightforward to determine the unknown coefficients E0ryE_{0ry} and E2E_2. First, the reflection coefficient E0ryE_{0ry} is found to be,

E0ry=−1+2cos⁡θiκ2cot⁡(κ2d)+cos⁡θ+η0σyy,(28)E_{0ry}=-1+\frac{2\cos\theta}{i\kappa_2\cot(\kappa_2d)+\cos\theta+\eta_0\sigma_{yy}}, \tag*{(28)}

where we define κ2=k2/k0\kappa_2=k_2/k_0. The coefficient E2E_2 for the electric field in the spacer region is similarly found to be

E2=−2cos⁡θcsc⁡(κ2d)cos⁡θ+η0σyy+iκ2cot⁡(κ2d),(29)E_2=-\frac{2\cos\theta\csc(\kappa_2d)}{\cos\theta+\eta_0\sigma_{yy}+i\kappa_2\cot(\kappa_2d)}, \tag*{(29)}

where from Eq. (27), the dimensionless quantity η0σyy\eta_0\sigma_{yy} can be expressed in terms of the permittivity: η0σyy=i(1−εyy)k0t\eta_0\sigma_{yy}=i(1-\varepsilon_{yy})k_0t.

The fraction of energy that is absorbed by the system is determined by the absorptance (AA): A=1−RA=1-R, where RR is the reflectance index. Note that due to the metallic substrate, there is no transmission of EM fields into the region z>dz>d. In determining the absorptance of the phosphorene system, we consider the time-averaged Poynting vector in the direction perpendicular to the interfaces (the zz direction), S0z=ℜ{E0yH0x∗}/2S_{0z}=\Re\{E_{0y}H_{0x}^*\}/2. Upon inserting the electric and magnetic fields for the vacuum region, we find, A=1−∣E0ry∣2A=1-|E_{0ry}|^2, where A=Sz0/S0A=S_{z0}/S_0, and S0=k0z/(2ε0ω)S_0=k_0z/(2\varepsilon_0\omega) is the time-averaged Poynting vector for a plane wave traveling in the zz direction.

In the following, we consider representative material and geometric parameters, and demonstrate how the differing predictions from the low-energy model and DFT-RPA can considerably influence the absorption of EM energy in a phosphorene-based system. We illustrate in Figure 7 the absorptance as a function of incident angle θ\theta, and thickness of the insulator layer, dd. The absorptance is determined after the permittivity tensor is calculated from the DFT-RPA and low-energy models. Results are shown in Figure 7(a) and Figure 7(b) for the DFT-RPA approach, and in Figure 7(c) and Figure 7(d) for the low-energy model. The frequency of the incident EM wave is set to ω=0.06\omega=0.06 eV in Figure 7(a) and Figure 7(c), and ω=1.4\omega=1.4 eV in Figure 7(b) and Figure 7(d). In all cases, the chemical potential is set to zero and a compressive strain of −10%-10\% is considered. Considering first the DFT-RPA results in 7(a) and 7(b), it is evident that over most of the parameter space, the incident beam reflects completely off the structure (A≈0A \approx0). There is only a small region of the diagram in 7(b) where there is near perfect absorption for near-grazing incidences (θ≈90∘\theta\approx90^\circ). The low-energy model in Figure 7(c) however predicts an extremely strong absorption region within 2 μm≲d≲3 μm2\ \mu\mathrm{m} \lesssim d \lesssim3\ \mu\mathrm{m}. In contrast to the DFT approach, the low-energy model results in nearly perfect absorption within 45∘≲θ≲75∘45^\circ\lesssim\theta\lesssim75^\circ. Increasing the frequency to ω=1.4\omega= 1.4 eV in Figure 7(d), the results of DFT and the low-energy model become more similar, although the low-energy model still predicts stronger absorption over a broader range of spacer layer thicknesses and angles θ\theta.

Figure 7. The absorptance as a function of the angle of incident light θ\theta and the thickness of the insulator layer dd. The DFT data and low energy model are used in the top and bottom rows, respectively. The frequency of the incident light is set to ω=0.06\omega= 0.06 eV (left column) and 1.4 eV (right column). The biaxial strain is fixed at −10%-10\%, and for proper comparisons, the chemical potential is set to μ=0\mu= 0.

To further contrast the DFT-RPA method and low-energy model, Figure 8 displays the absorptance as a function of θ\theta (left column) and ω\omega (right column), for three values of the biaxial strain: −10%-10\%, 0%0\%, and +10%+10\%. Results in the left column have a set frequency ω=1.4\omega= 1.4 eV, and an insulator thickness of d=100d = 100 nm, whereas the right column has d=310d = 310 nm and θ=80∘\theta= 80^\circ. Figure 8(c) corresponds to a slice of Figure 7(b) and Figure 7(d), and more clearly shows how the DFT results cause a shifting of the near-perfect absorption peak towards θ=90∘\theta= 90^\circ. The discrepancies between the two models is seen to dramatically increase for the other strain values shown in 8(a) and 8(b). Examining the frequency response in Figure 8(d)-Figure 8(f), it is evident that the two models again lead to different absorption characteristics. For zero strain [Figure 8(d)] the absorptance profiles are similar but shifted in frequency. When a tensile strain of +10%+10\% is applied to phosphorene [Figure 8(e)], the DFT approach shows a small amount of absorption, but overall both models predict that the incident EM wave is mostly reflected over the given frequency window. When a compressive strain of −10%-10\% is applied [Figure 8(f)], there is again a shift similar to 8(d), but now the magnitudes of the peaks are much different, with the low-energy model exhibiting near perfect absorption at ω=1.4\omega= 1.4 eV. These discrepancies originate mainly from the different predictions for the permittivities, where e.g., the DFT and low-energy methods give different dissipa pfuna tion thresholds and significant amplitude variations (see Figure 2 and Figure 3).

Figure 8. The absorptance AA calculated by employing the results of DFT and low-energy models. Three biaxial strain values are considered (as labeled): −10%-10\%, 0%0\%, +10%+10\%. The left column shows AA as a function of the incident angle θ\theta, at fixed frequency ω=1.4\omega= 1.4 eV, and for an insulator thickness of d=100d = 100 nm. The right column has AA as a function of frequency of the incident EM wave at an incident angle of θ=80∘\theta= 80^\circ, and d=310d = 310 nm. In all cases, the chemical potential is zero; μ=0\mu= 0.

Regardless of the discrepancies and deviations discussed above (when using the permittivity data produced by the DFT-RPA and low-energy model), one can clearly observe the strong switching characteristics of the device considered (Figure 6) in absorbing the incident EM wave. As seen in Figure 7 and Figure 8, the device shown in Figure 6 can absorb nearly perfectly the incident EM wave within certain incident angles, strain, the thickness of the insulator layer, frequency, and chemical potential. For example, comparing the results obtained for different values of strain in Figure 8(a,Figure 8b,Figure 8c), we conclude that the application of low strains (less than 10%) into the plane of phosphorene can effectively control the absorptivity of this device, switching efficiently between vanishingly small absorption and nearly perfect absorption of an incident EM wave at certain incident angles. Although the overall behavior of the permittivity components in Figure 2 and Figure 3 at 0% and +10% seem to be the same, at the given frequency, i.e., ω=1.4\omega= 1.4 eV, these components possess substantially different imaginary and real parts, that together with the strong EM wave interference in the device of Figure 6, cause considerable differences in the absorptivity seen in Figure 7 and Figure 8. Both the interference phenomenon and Joule-heating effects are known to be

strongly dependent on the amount of loss in the material and are governed by the imaginary part of the relevant permittivity component. For example, in the left set of figures in Figure 7 (where the frequency is the same, and fixed at ω=0.06\omega= 0.06 eV), the imaginary part of εyy\varepsilon_{yy} is: 0.00906 (DFT) and 157.4 (low-energy model). These huge variations in the dissipation translate into the observed absorptivity differences.

conclusions

Due to the fundamental importance of light-matter interactions, we have investigated the permittivity of phosphorene, subject to in-plane strain, as a representative material platform, using two approaches: One approach employed density functional theory combined with the random phase approximation (DFT-RPA), and the other method involved a low-energy effective Hamiltonian model and Green’s function. The permittivity components for this strongly anisotropic material are fully explained by its associated band structures, transitions, and optical conductivities within the low-energy formalism. However, the results of DFT-RPA and the corresponding band structures calculated from the Perdew-Burke-Ernzerhof functional, showed considerable discrepancies. The DFT calculations were repeated using two different packages, and similar results were obtained. Although some broad, generic trends for the frequency dispersion of the permittivity components were in agreement for both approaches, several important differences stood out, including the onset of the imaginary part of the Drude response that revealed important fundamental physical characteristics of a material, such as the band gap, and interband and intraband transitions.

To illustrate the fundamental importance of accurate predictions of the permittivity response in designing new optoelectronics devices, we have compared the perfect absorption characteristics of a simple device, employing the permittivity data from the low-energy model and DFT-RPA. Our results suggest that the DFT-RPA method implemented in many types of DFT packages needs to be revisited, and improvements made so that the results are at least more consistent with the associated band structures. Accurate predictions for the permittivity and optical conductivity are of pivotal importance in determining the physical properties of materials and designing novel optoelectronics devices.

Interestingly, on the technological side of making use of phosphorene in optoelectronics devices, we find that strained phosphorene can serve as the switching element for the absorption of EM wave in EM wave absorbers. This switching effect can be effectively controlled by the application of relatively low mechanical strains (less than 10%) into the plane of phosphorene and/or manipulation of the chemical potential of phosphorene through a gate voltage.

Acknowledgments

The DFT calculations were performed using the resources provided by UNINETT Sigma2 - the National Infrastructure for High Performance Computing and Data Storage in Norway, NOTUR/Sigma2 project number: NN9497K. Part of the calculations were performed using HPC resources from the DOD High Performance Computing Modernization Program (HPCMP). K.H. is supported in part by the NAWCWD In Laboratory Independent Research (ILIR) program and a grant of HPC resources from the DOD HPCMP.

References

References

  1. [1]R. Gutzler, M. Garg, C. R. Ast, K. Kuhnke, K. Kern, Light-matter interaction at atomic scales, Nat. Rev. Phys. (2021). https://doi.org/10.1038/s42254-021-00306-5; M. Garg, etal, Multi-petahertz electronic metrology, Nature, 538, 359 (2016).
  2. [2]F. N. Xia, H. Wang, and Y. C. Jia, Rediscovering black phosphorus as an anisotropic layered material for optoelectronics and electronics, Nat. Commun. 5, 4458 (2014).
  3. [3]T. Ahmed, etal, Fully Light-Controlled Memory and Neuromorphic Computation in Layered Black Phosphorus (2021); T. Ahmed, etal, Multi-functional Optoelectronics via Harnessing Defects in Layered Black Phosphorus, Advanced Functional Materials, 29, 1 (2019).DOI
  4. [4]A. Rodin, M. Trushin, A. Carvalho, A. H. Castro Neto, Collective Excitations in 2D Materials, Nat. Rev. Phys. 2, 524 (2020).DOI
  5. [5]A. Carvalho, P. E. Trevisanutto, S. Taioli, A. H. Castro Neto, Computational methods for 2D materials modelling, arXiv:2101.06859.
  6. [6]M. Dion, H. Rydberg, E. Schroder, D. C. Langreth, B. I. Lundqvist, van der Waals Density Functional for General Geometries, Phys. Rev. Lett. 92, 246401 (2004).DOI
  7. [7]S. Grimme, Accurate description of van der Waals complexes by density functional theory including empirical corrections, J. Comput. Chem. 25, 1463 (2004).
  8. [8]M. Alidoust, K. Halterman, D. Pan, M. Willatzen, J. Akola, Strain-Engineered Widely-Tunable Perfect Absorption Angle in Black Phosphorus from First-Principles, Phys. Rev. B 102, 115307 (2020).DOI
  9. [9]S. Biswas, W. S. Whitney, M. Y. Grajower, K. Watanabe, T. Taniguchi, H. A. Bechtel, G. R. Rossman, H. A. Atwater, Tunable intraband optical conductivity and polarization-dependent epsilon-near-zero behavior in black phosphorus, Sci. Adv. 7, eabd4623 (2021).DOI
  10. [10]V. Tran, R. Soklaski, Y. Liang, and L. Yang, Layer-controlled band gap and anisotropic excitons in few-layer black phosphorus, Phys. Rev. B 89, 235319 (2014).
  11. [11]Y. Wei, F. Lu, T. Zhou, X. Luo, and Y. Zhao, Stacking sequences of black phosphorus allotropes and the corresponding few-layer phosphorenes, Physical Chemistry Chemical Physics 20, 10185 (2018).DOI
  12. [12]T. Fang, T. Liu, Z. Jiang, R. Yang, P. Servati, and G. Xia, Fabrication and the Interlayer Coupling Effect of Twisted Stacked Black Phosphorus for Optical Applications, ACS Appl. Nano Mater. 2, 3138 (2019).DOI
  13. [13]Z. Zhang, L. Li, J. Horng, N. Z. Wang, F. Yang, Y. Yu, Y. Zhang, G. Chen, K. Watanabe, T. Taniguchi, X. H. Chen, F. Wang, Y. Zhang, Strain-modulated bandgap and piezoresistive effect in black phosphorous field-effect transistors, Nano Lett. 17 6097-6103 (2017).DOI
  14. [14]S. Das, W. Zhang, M. Demarteau, A. Hoffmann, M. Dubey, A. Roelofs, Tunable transport gap in phosphorene, Nano Lett. 14 5733-5739 (2014).DOI
  15. [15]Z. Qin, G. Xie, H. Zhang, C. Zhao, P. Yuan, S. Wen, and L. Qian, Black phosphorous as saturable absorber for the Q-switched Er:ZBLAN fiber laser at 28 μm, Opt. Express 23, 24713 (2015).DOI
  16. [16]A. Carvalho, M. Wang, X. Zhu, A. Rodin, H. Su, and A. C. Neto, Phosphorene: from theory to applications, Nat. Rev. Materials 1, Article number: 16061 (2016).DOI
  17. [17]G. Zhang, S. Huang, F. Wang, Q. Xing, C. Song, C. Wang, Y. Lei, M. Huang, and H. Yan, The optical conductivity of few-layer black phosphorus by infrared spectroscopy, Nat. Comm.11, 1847 (2020).DOI
  18. [18]A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim, The electronic properties of graphene, Rev. Mod. Phys. 81, 109 (2009).DOI
  19. [19]H. Liu, A. T. Neal, Z. Zhu, Z. Luo, X. Xu, D. Tomanek, and P. D. Ye, Phosphorene: an unexplored 2D semiconductor with a high hole mobility, ACS NANO 8, 4033 (2014).DOI
  20. [20]D. Bohm and D. Pines, A collective description of electron interactions. I. Magnetic interactions Phys. Rev. 82, 625 (1951).
  21. [21]M. W. Jørgensen and S. P. A. Sauer, Benchmarking doubles-corrected random-phase approximation meth- ods for frequency dependent polarizabilities: Aromatic molecules calculated at the RPA, HRPA, RPA(D), HRPA(D), and SOPPA levels, J. Chem. Phys. 152, 234101 (2020).
  22. [22]E. van Loon, M. Rosner, M. I. Katsnelson, T. O. Wehling, Random Phase Approximation for gapped systems: role of vertex corrections and applicability of the constrained random phase approximation, arXiv:2103.04419.
  23. [23]M. Gajdos, K. Hummer, G. Kresse, J. Furthmüller and F. Bechstedt, Linear optical properties in the projected-augmented wave methodology, Phys. Rev. B 73, 045112 (2006).DOI
  24. [24]E. Sasioglu, C. Friedrich, and S. Blugel, Effective Coulomb interaction in transition metals from constrained random-phase approximation, Phys. Rev. B 83, 121101(R) (2011).DOI
  25. [25]H. Shinaoka, M. Troyer, and P. Werner, Accuracy of downfolding based on the constrained random-phase approximation, Phys. Rev. B 91, 245156 (2015).DOI
  26. [26]C. Honerkamp, H. Shinaoka, F. F. Assaad, and P. Werner Limitations of constrained random phase approximation downfolding, Phys. Rev. B 98, 235151 (2018).
  27. [27]X.-J. Han, P. Werner, and C. Honerkamp, Investigation of the effective interactions for the Emery model by the constrained random-phase approximation and constrained functional renormalization group, Phys. Rev. B 103, 125130 (2021).DOI
  28. [28]M. S. Hybertsen and S. G. Louie, Ab initio static dielectric matrices from the density-functional approach. I. Formulation and application to semiconductors and insulators, Phys. Rev. B 35, 5585 (1987).pubmed.ncbi.nlm.nih.gov/9940770
  29. [29]M. N. Gjer ding, M. Pandey and K. S. Thygesen, Band structure engineered layered metals for low-loss plasmonics, Nat. Commun. 8, 1 (2017).
  30. [30]J. J. Mortensen, L. B. Hansen, K. W. Jacobsen , Real-space grid implementation of the projector augmented wave method, Phys. Rev. B 71, 035109 (2005).DOI
  31. [31]J. Enkovaara, C. Rostgaard, J.J. Mortensen, J. Chen, M. Dulak, L. Ferrighi, J. Gavnholt, C. Glinsvad, V. Haikola, H.A. Hansen, et al., Electronic structure calculations with GPAW: a real-space implementation of the projector augmented-wave method, J. Phys. Condens. Matter 22, 253202 (2010).DOI
  32. [32]L. Voon, A. Lopez-Bezanilla, J. Wang, Y. Zhang, M. Willatzen, Effective Hamiltonians for phosphorene and silicene, New J. Phys. 17, 025004 (2015).DOI
  33. [33]L. Voon, J. Wang, Y. Zhang, and M. Willatzen, Band parameters of phosphorene, J. Phys.: Conf. Ser. 633, 012042 (2015).DOI
  34. [34]M. Alidoust, M. Willatzen, and A.-P. Jauho, Strain-engineered Majorana zero energy modes and φ0 Josephson state in black phosphorus, Phys. Rev. B 98, 085414 (2018).DOI
  35. [35]M. Alidoust, M. Willatzen, and A.-P. Jauho, Fraunhofer response and supercurrent spin switching in black phosphorus with strain and disorder, Phys. Rev. B 98, 184505 (2018).DOI
  36. [36]M. Alidoust, M. Willatzen, and A.-P. Jauho, Control of superconducting pairing symmetries in monolayer black phosphorus, Phys. Rev. B 99, 125417 (2019).DOI
  37. [37]G.D. Mahan, Many–Particle Physics (Plenum Press, New York) 1990.DOI
  38. [38]M. Tahir, P. Vasilopoulos, F. M. Peeters , Magneto-optical transport properties of monolayer phosphorene, Phys. Rev. B 92, 045420 (2015).DOI
  39. [39]J. Jang, S. Ahn, and H. Min, Optical conductivity of black phosphorus with a tunable electronic structure, 2D Mater. 6, 025029 (2019).DOI
  40. [40]C. H. Yang, J. Y. Zhang, G. X. Wang, and C. Zhang, Dependence of the optical conductivity on the uniaxial and biaxial strains in black phosphorene, Phys. Rev. B 97, 245408 (2018).DOI
  41. [41]D. Q. Khoa, B. D. Hoi, T. C. Phong, Transverse Zeeman magnetic field effects on the dynamical dielectric function of monolayer phosphorene: Beyond the continuum approximation, J. Mag. Mag. Mat. 491, 165637 (2019).DOI
  42. [42]C. Kittel, Introduction to Solid State Physics, California: John Wiley & Sons, Inc., 2004.
  43. [44]C. Adamo and V. Barone, Toward reliable density functional methods without adjustable parameters: The PBE0 model, J. Chem. Phys. 110, 6158 (1999).DOI
  44. [45]V. Tran, R. Soklaski, Y. Liang, and L. Yang, Layer-controlled band gap and anisotropic excitons in few-layer black phosphorus, Phys. Rev. B 89, 235319 (2014).
  45. [46]J. Kim, S. S. Baik, S. H. Ryu, Y. Sohn, S. Park, B.-G. Park, J. Denlinger, Y. Yi, H. J. Choi, K. S. Kim, Observation of tunable band gap and anisotropic Dirac semimetal state in black phosphorus, Science 349, 723 (2015).DOI
  46. [47]J. Qiao, X. Kong, Z.X. Hu, F. Yang, W. Ji, High-mobility transport anisotropy and linear dichroism in few-layer black phosphorus, Nat. Commun. 5, 4475 (2014).DOI

Paper details

vlnc:W_qfses3yv
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents