This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Limit possible electric field in plasmon nanogap resonator

Andrey K. Sarychev [email protected] Institute for Theoretical and Applied Electrodynamics, Russian Academy of Sciences,125412, Moscow, Russia    Andrey V. Ivanov [email protected] Institute for Theoretical and Applied Electrodynamics, Russian Academy of Sciences,125412, Moscow, Russia    Grégory Barbillon [email protected] EPF-Ecole d’Ingénieurs, 92330 Sceaux, France
Abstract

We propose the theory of the plasmon excited in ultra-narrow plasmonic gap formed by a metal cylinder on a metal surface, that generates a huge resonant local electromagnetic field in optical frequencies. Resonance conditions are found. The maximum possible enhancement of the electric field in the nanogap is estimated as |Emax|/|E0||εm|3εd1(εm)2\left|E_{max}\right|/\left|E_{0}\right|\propto\left|\varepsilon_{m}\right|^{3}\varepsilon_{d}^{-1}\left(\Im\varepsilon_{m}\right)^{-2}. A simple ultranarrow-gaped resonator can be used for SERS sensing and infrared spectroscopy of adsorbed molecules. We assume that a plasmonic nanogap is simplest and most straightforward way to increase flexibility as well as sensitivity of the plasmon-enhanced spectroscopes.

keywords:
plasmon nanogap resonators, SERS

1 Introduction

Calculation of the electric field in the gap between two metallic cylinders is a classical problem of the classical electrodynamics. However, the metal permittivity is typically negative in optical spectral range, and plasmons are excited between the cylinders, which give a big enhancement of the local electric field in the crevice 1. Later, it was shown 2 that giant optical field fluctuations in random semicontinuous metallic films are due to the local field concentration in-between large metallic clusters. A precise separation of plasmonic nanostructures is of widespread interest, because it controls the build-up of intense and localized optical fields, as well as their spectral tuning 3, 4. Plasmonic enhancement can also be used to improve the efficiency of photodetectors 5. Similarly, the non-linear frequency conversion critically depends on the field enhancements, their spatial localization, and their spectral resonances whose all ultra-sensitivity depends on the plasmonic gaps 6. Fundamental processes like quantum tunneling can be observed optically, but only for gap separations in the subnanometer regime 7. Graphene could be used as the thinnest spacer between gold nanoparticles and a gold substrate 8. The atomic layer deposition of the Al2O3Al_{2}O_{3} film provides a subnanometer layer that has been employed to successfully fabricate high quality metallic nanogap arrays with sub-5 nm gap size 9. In the works 10, 11, 12, 13, high local electric fields were generated by ultra-narrow gap that formed by gold nanoparticle placed on gold mirror. Evolution of different gap modes with a gap size of 1-5 nm was considered. It was shown that the coupling of transverse and antenna modes is altered by varying the gap size and the nanoparticle shape from sphere to cube, resulting in strongly hybridized modes. Theoretical predictions and experimental studies have shown giant electromagnetic field fluctuations in case of almost touched plasmonic nanoparticles 15, 14, 16, 17, 18, 19, 20, 21. It was shown that plasmonic nanocavities confining the light to unprecedentedly small volumes, support multiple types of modes. Different nature of these modes leads to mode beating within the nanocavity and the Rabi oscillations, which alters the spatio-temporal dynamics of the hybrid system 13. By intermixing plasmonic excitation in nanoparticle arrays with excitons in a WS2WS_{2} monolayer inside a resonant metallic microcavity, the hierarchical system was fabricated with the collective microcavity-plasmon-exciton Rabi splitting exceeding 0.5eV0.5\,eV at room temperature. Gap-surface plasmon metasurfaces, which consist of a subwavelength thin dielectric spacer sandwiched between an optically thick film of metal and arrays of metallic subwavelength elements arranged in a strictly or quasiperiodic fashion, have a possibility to fully control the amplitude, phase, and polarization of the reflected light 22. Such systems are successfully used for surface-enhanced infrared absorption 23. Strong motivation for the investigation of the gap plasmons is the SERS effect, which relies on the plasmonic enhancement to enable identification of trace molecules captured within gaps. The SERS is extremely important for medical diagnostics, for instance, for cancer detection, imaging and therapy, drug delivery, quantitative control of biomarkers including glycated proteins and cardiovascular biomarkers 24, 25, 26, 27, 28. In recent years, there are many computer simulations as well as experimental works that are devoted to perfect coupled particles with ultra-narrow gaps 9, 29, 30, 31, 32, 33, 34, 36, 37, 38, 39, 40. In this paper, we propose the theory of the plasmons excited in a ultra-narrow gap formed by a metallic cylinder on a metallic surface for different metals. We find the resonance conditions, when it is possible to achieve possible limit electromagnetic field both inside and in the vicinity of the gap. This will increase the sensitivity of the SERS-probing as well as this of other existing surface enhanced spectroscopes.

2 Analytical theory

We consider two parallel metallic cylinders whose respective radii are aa and a1>aa_{1}>a that are separated by the distance d+d1d+d_{1} so that the spacing between the cylinder axes is a+d+a1+d1a+d+a_{1}+d_{1}. The xx-axis connects the cylinder centers so the centers have coordinates {x,y}={a+d,0}\{x,y\}=\{a+d,0\} and {a1d1,0}\{-a_{1}-d_{1},0\}, correspondingly, as shown in Fig.1. To find the optical electric field in the nanogap between the cylinders illuminated by the incident light (see Fig.1d), we use new coordinates uu and vv that convenient to introduce in the complex form z=Ltanhw/2z=L\tanh w/2, where z=x+iyz=x+iy, w=u+ivw=u+iv, and the characteristic scale L=d(2a+d)L=\sqrt{d(2a+d)}, where dd is the distance from the surface of the cylinder “aa" to the origin of the coordinates. Note aa, a1a_{1}, and LL fully define the system geometry, if we define d1=a12+L2a1<dd_{1}=\sqrt{a_{1}^{2}+L^{2}}-a_{1}<d. In the transition from zz to ww, the whole plane {x,y}\{x,y\} transforms to the strip <u<-\infty<u<\infty, π<v<π-\pi<v<\pi. The surfaces of the cylinders “aa" and “a1a_{1}" are transformed to the vertical lines ud=ln[(a+d)/(ad)]>0u_{d}=\ln[(a+d)/(a-d)]>0 and ud1=ln[(a1+d1)/(a1d1)]<0u_{d_{1}}=-\ln[(a_{1}+d_{1})/(a_{1}-d_{1})]<0 (Fig. 1).

Refer to caption
Figure 1: The design of cylindrical nanoparticle-on-mirror cavity. (a-b) Conformal map of two cylinders into uvu-v strip; (c) metallic cylinder above metal plate, electromagnetic wave is incident from the top; (d) electric field between Ag plate and cylinder with a radius of a=50nma=50\,nm, gap d=0.5nmd=0.5\,nm calculated at the excitation wavelength of λ=405nm\lambda=405\,nm.

All the space outer of the cylinders shrinks into the rectangular [ud1<u<ud,π<v<π][u_{d_{1}}<u<u_{d},\;-\pi<v<\pi]. The approach is similar to the map used in 16, 41. The solution of the Laplace equation in the rectangular can be presented as sum of the functions fRc(q)=equcosqv,fRs(q)=equsinqv,fLc(q)=equcosqv,fLs(q)=equsinqvf_{Rc}^{(q)}=e^{-qu}\cos qv,\;f_{Rs}^{(q)}=e^{-qu}\sin qv,\;f_{Lc}^{(q)}=e^{qu}\cos qv,\;f_{Ls}^{(q)}=e^{qu}\sin qv, q=1,2q=1,2\ldots :

φG(u,v)=φ0+q=1[A1(q)fRc(q)+A2(q)fRs(q)+A3(q)fLc(q)+A4(q)fLs(q)]\varphi^{G}(u,v)=\varphi_{0}+\\ \sum_{q=1}^{\infty}\left[A_{1}^{(q)}f_{Rc}^{(q)}+A_{2}^{(q)}f_{Rs}^{(q)}+A_{3}^{(q)}f_{Lc}^{(q)}+A_{4}^{(q)}f_{Ls}^{(q)}\right] (1)

where φ0\varphi_{0} is the potential of the external electric field. We assume that the electric field concentrates in the nanogap whose the size is much smaller than the wavelength of the incident light d+d1λd+d_{1}\ll\lambda. Electric potentials in the “aa" and “a1a_{1}" cylinders are given by the following expressions, respectively:

φR(u,v)=φ0+q=1[B1(q)fRc(q)+B2(q)fRs(q)],u>ud\varphi^{R}(u,v)=\varphi_{0}+\sum_{q=1}^{\infty}\left[B_{1}^{(q)}f_{Rc}^{(q)}+B_{2}^{(q)}f_{Rs}^{(q)}\right],\;\;u>u_{d} (2)
φL(u,v)=φ0+q=1[B3(q)fLc(q)+B4(q)fLs(q)],u<ud1\varphi^{L}(u,v)=\varphi_{0}+\sum_{q=1}^{\infty}\left[B_{3}^{(q)}f_{Lc}^{(q)}+B_{4}^{(q)}f_{Ls}^{(q)}\right],\;\;u<u_{d_{1}} (3)

The components of the electric field in {u,v}\{u,v\} space {Eu,Ev}=wφ={φ/u,\{E_{u},E_{v}\}=-\nabla_{w}\varphi=-\{\partial\varphi/\partial u, φ/v}\partial\varphi/\partial v\} are complex since the metal permittivity is complex. Electric field in {x,y}\{x,y\} space equals to {Ex,Ey}=J^{Eu,Ev}\{E_{x},E_{y}\}=\hat{J}\{E_{u},E_{v}\}, where the jacobian J^\hat{J} is obtained from the derivative wz=dw/dzw_{z}=dw/dz, namely, J11=J22=wzJ_{11}=J_{22}=\Re w_{z}, J21=J12=wzJ_{21}=-J_{12}=\Im w_{z}. The coefficients A(q)A^{(q)} and B(q)B^{(q)} in Eqs. (1), (2) and (3) are found from the boundary conditions for the electric fields ER=wφRE^{R}=-\nabla_{w}\varphi^{R}, EG=wφGE^{G}=-\nabla_{w}\varphi^{G}, and EL=wφLE^{L}=-\nabla_{w}\varphi^{L} at the surface of “aa" and “a1a_{1}" cylinders:

EvR=EvG,εmEuR=εdEuG,u=udE^{R}_{v}=E^{G}_{v},\;\varepsilon_{m}E^{R}_{u}=\varepsilon_{d}E^{G}_{u},\;u=u_{d} (4)
EvL=EvG,εm1EuL=εdEuG,u=ud1E^{L}_{v}=E^{G}_{v},\;\varepsilon_{m_{1}}E^{L}_{u}=\varepsilon_{d}E^{G}_{u},\;u=u_{d_{1}} (5)

where εm\varepsilon_{m}, εd\varepsilon_{d}, and εm1\varepsilon_{m_{1}} are the permittivities of the “aa" cylinder, the outer space, and the “a1a_{1}" cylinder, respectively. We expand the external field E0=E_{0}= φ0-\nabla\varphi_{0} ={E0x,E0y}=\{E_{0x},E_{0y}\} in a series of the functions fRc(q),fRs(q),f_{Rc}^{(q)},\,f_{Rs}^{(q)}, fLc(q),fLs(q)f_{Lc}^{(q)},\,f_{Ls}^{(q)}. For the simplest case of the constant external field, the expansion {E0u,E0v}=\left\{E_{0u},E_{0v}\right\}= q{E0u(q),E0v(q)}\sum_{q}\left\{E_{0u}^{(q)},E_{0v}^{(q)}\right\}, has the following form:

{E0u(q),E0v(q)}=2L(1)qq×{E0xfRc(q)E0yfRs(q),E0xfRs(q)+E0yfRc(q),u>0E0xfLc(q)+E0yfLs(q),E0yfLc(q)E0xfLs(q),u<0}\left\{E_{0u}^{(q)},E_{0v}^{(q)}\right\}=-2L(-1)^{q}q\times\\ \left\{\begin{matrix}E_{0x}f_{Rc}^{(q)}-E_{0y}f_{Rs}^{(q)},\;E_{0x}f_{Rs}^{(q)}+E_{0y}f_{Rc}^{(q)},\;u>0\\ E_{0x}f_{Lc}^{(q)}+E_{0y}f_{Ls}^{(q)},\;E_{0y}f_{Lc}^{(q)}-E_{0x}f_{Ls}^{(q)},\;u<0\end{matrix}\right\} (6)

where q=1,2,q=1,2,\ldots. We substitute the above equation in the boundary equation (4) for the surface of the “aa" cylinder and equate the coefficients at the same f(q)f^{(q)} functions obtaining the following equations for the coefficients {A(q),B(q)}\{A^{(q)},B^{(q)}\} in Eqs. (1), (2) and (3):

εd(A3(q)+A1(q)g2q)εmB1(q)=E0x(εmεd),εd(A4(q)A2(q)g2q)εmB2(q)=E0y(εmεd),A2(q)g2q+A4(q)B2(q)=A1(q)g2qA3(q)+B1(q)=0\varepsilon_{d}\left(A_{3}^{(q)}+A_{1}^{(q)}g^{2q}\right)-\varepsilon_{m}B_{1}^{(q)}=E_{0x}\left(\varepsilon_{m}-\varepsilon_{d}\right),\\ \varepsilon_{d}(A_{4}^{(q)}-A_{2}^{(q)}g^{2q})-\varepsilon_{m}B_{2}^{(q)}=E_{0y}\left(\varepsilon_{m}-\varepsilon_{d}\right),\\ A_{2}^{(q)}g^{2q}+A_{4}^{(q)}-B_{2}^{(q)}=A_{1}^{(q)}g^{2q}-A_{3}^{(q)}+B_{1}^{(q)}=0 (7)

where g=(L+d)/(Ld)g=(L+d)/(L-d). Matching electric field and displacement at the surface of the “a1a_{1}" cylinder, we obtain from Eq. (5) :

εd(A1(q)+A3(q)g12q)εm1B3(q)=E0x(εm1εd),εd(A2(q)A4(q)g12q)εm1B4(q)=E0y(εm1εd),A4(q)g12q+A2(q)B4(q)=A3(q)g12qA1(q)+B3(q)=0\varepsilon_{d}\left(A_{1}^{(q)}+A_{3}^{(q)}g_{1}^{2q}\right)-\varepsilon_{m_{1}}B_{3}^{(q)}=E_{0x}\left(\varepsilon_{m_{1}}-\varepsilon_{d}\right),\\ \varepsilon_{d}\left(A_{2}^{(q)}-A_{4}^{(q)}g_{1}^{2q}\right)-\varepsilon_{m_{1}}B_{4}^{(q)}=E_{0y}\left(\varepsilon_{m_{1}}-\varepsilon_{d}\right),\\ A_{4}^{(q)}g_{1}^{2q}+A_{2}^{(q)}-B_{4}^{(q)}=A_{3}^{(q)}g_{1}^{2q}-A_{1}^{(q)}+B_{3}^{(q)}=0 (8)

where g1=(L+d1)/(Ld1)g_{1}=(L+d_{1})/(L-d_{1}). Zeroing of the determinant of Eqs. (7) and (8) gives the condition for the “qq" resonance. The cylinders with the same permittivity εm1=εm\varepsilon_{m_{1}}=\varepsilon_{m} resonate when (gg1)2q(εd+εm)2(εmεd)2=0(gg_{1})^{2q}\left(\varepsilon_{d}+\varepsilon_{m}\right)^{2}-\left(\varepsilon_{m}-\varepsilon_{d}\right)^{2}=0. The first term dominates when the distance d+d1d+d_{1} between the cylinders increases and gg1gg_{1}\rightarrow\infty. Then, all the resonance frequencies ωr(q)\omega_{r}^{(q)} collapse to the single value ωr(1)\omega_{r}^{(1)} given by the well-known equation [εd(ωr(1))+εm(ωr(1))]=0\Re[\varepsilon_{d}(\omega_{r}^{(1)})+\varepsilon_{m}(\omega_{r}^{(1)})]=0 for the plasmon resonance in a metallic cylinder. On the other hand, when the gap size vanishes d+d10d+d_{1}\rightarrow 0, the factor gg11gg_{1}\rightarrow 1 from above and the resonance frequencies ωr(q)\omega_{r}^{(q)} spread out from ωr(1)\omega_{r}^{(1)} to the minimum resonance frequency estimated from the equation εm(ωr(m))2εdaa1L(a+a1)\varepsilon_{m}(\omega_{r}^{(m)})\simeq-\frac{2\varepsilon_{d}aa_{1}}{L(a+a_{1})}\rightarrow-\infty. The real part of the permittivity of the silver, gold, and many other metals is well described by the Drude formula in red and infrared spectral range εm(ω)\Re\varepsilon_{m}(\omega)\propto- (ωp/ω)2(\omega_{p}/\omega)^{2} 42 and the minimal resonance frequency ωr(m)\omega_{r}^{(m)}\propto ωp(d/a)1/4ωp\omega_{p}(d/a)^{1/4}\ll\omega_{p}.

Observation of the plasmon resonance between metal cylinders is a difficult experimental problem of positioning two nanocylinders parallel each other at a nanosize distance. Yet, the investigation of the plasmon resonance in the metallic nanocylinder placed at nanodistance above a flat metallic surface (i.e., a1=a_{1}=\infty) is in the center of nowadays experimental studies. To obtain the electric field in the gap between metallic cylinder and metallic surface, we are going to the limit a1a_{1}\rightarrow\infty, d1L2/2a10d_{1}\simeq L^{2}/2a_{1}\rightarrow 0. The electric field E0E_{0} is fixed at the surface of the large cylinder “a1a_{1}" far away from the cylinder “aa" and all the fields are expressed in terms of this field. After the limit a1a_{1}\rightarrow\infty is taken, the field E0E_{0} is the electric field at the metallic surface x=0,|y|ax=0,\;|y|\gg a (see Fig. 1c). Thus, we get the electric field in the center coordinates x=y=0x=y=0, i.e., u=v=0u=v=0. The electric field E(x,y)E(x,y) has the maximum at this characteristic point just below the cylinder (see Fig. 1 c,d), the jacobian JJ reduces to the scalar LL, and we obtain by solution of Eqs. (7) and (8) and substitution in Eq. (1):

E(0,0)={Ex0[1+8εm1Σεdεm],Ey0[18εdΣεdεm]},E(0,0)=\{{E}_{x0}\left[1\!+\frac{8\varepsilon_{m_{1}}\Sigma}{\varepsilon_{d}-\varepsilon_{m}}\right]\!,{E}_{y0}\left[1-\frac{8\varepsilon_{d}\Sigma}{\varepsilon_{d}-\varepsilon_{m}}\right]\}, (9)
Σ=q=1(1)qqh(b+1)qh,h=(εdεm)(εdεm1)(εd+εm)(εd+εm1),\Sigma=\sum^{\infty}_{q=1}\frac{(-1)^{q}q\,h}{(b+1)^{q}-h},h=\frac{\left(\varepsilon_{d}-\varepsilon_{m}\right)\left(\varepsilon_{d}-\varepsilon_{{m_{1}}}\right)}{\left(\varepsilon_{d}+\varepsilon_{m}\right)\left(\varepsilon_{d}+\varepsilon_{{m_{1}}}\right)}, (10)
b=(d+L)(2a+d+L)/a28d/a,b=(d+L)(2a+d+L)/a^{2}\simeq\sqrt{8d/a}, (11)

where εm\varepsilon_{m}, εm1\varepsilon_{m_{1}} and εd\varepsilon_{d} are the permittivities of the cylinder, metal plate, and outer space. aa and dd are the cylinder radius and the gap size, respectively. The last approximation in Eq. (11) holds for dad\ll a. The system resonates if a denominator in Eq. (10) almost vanishes, i.e., the dimensionless parameter h>1\Re h>1 and the loss factor κ=h/|h|2εd(εm|εm|2+εm1|εm1|2)1\kappa=\Im h/|h|\simeq 2\varepsilon_{d}\left(\Im\varepsilon_{m}\left|\varepsilon_{m}\right|^{-2}+\Im\varepsilon_{m1}{\left|\varepsilon_{m1}\right|^{-2}}\right)\ll 1 for |εm|,|εm1|εd\left|\varepsilon_{m}\right|,\left|\varepsilon_{m_{1}}\right|\gg\varepsilon_{d}.

In spite of simplicity of Eqs. (9) and (10), the gap field has rather rich behavior when the gap vanishes, i.e., b0b\rightarrow 0. Suppose that the parameter h<1\Re h<1, then the denominator n(q)=(b+1)qhn^{(q)}=(b+1)^{q}-h in the sum Σ\Sigma in Eq. (10) is not close to zero for all the values of qq, and therefore, it can be linearized as n(q)b1qh1n^{(q)}\simeq b_{1}q-h_{1}, where h1=h1h_{1}=h-1, h1<0\Re h_{1}<0, b1=log(1+b)bb_{1}=\log(1+b)\simeq b. Then, Σ\Sigma approximates as ΣS(h1,b1)=q(1)qq/(b1qh1)\Sigma\simeq S(h_{1},b_{1})=\sum_{q}^{\infty}\ (-1)^{q}q/(b_{1}q-h_{1}) and can be solved by Laplace transformation and Abel regularization. The function SS has simple integral presentation:

ΣS=h1(ψ1ψ2)+b12b12=01xh1b1(x+1)2dxb1\Sigma\simeq S=\frac{h_{1}\left(\psi_{1}-\psi_{2}\right)+b_{1}}{2b^{2}_{1}}=-\int_{0}^{1}\frac{x^{-\frac{h_{1}}{b_{1}}}}{(x+1)^{2}}\frac{dx}{b_{1}} (12)

where ψ1=ψ(b1h12b1)\psi_{1}=\psi\left(\frac{b_{1}-h_{1}}{2b_{1}}\right), ψ2=ψ(h12b1)\psi_{2}=\psi\left(-\frac{h_{1}}{2b_{1}}\right) and ψ=Γ/Γ\psi=\Gamma^{\prime}/\Gamma is the polygamma function. The sum approximates as Σ\Sigma\simeq 14(b1h1)-\frac{1}{4\left(b_{1}-h_{1}\right)} (b12b1h1+1)\left(\frac{b_{1}}{2b_{1}-h_{1}}+1\right) 14h1\rightarrow\frac{1}{4h_{1}} for the fixed h1h_{1} and |b1/h1|0\left|b_{1}/h_{1}\right|\rightarrow 0. That is the electric field tends towards a finite value when the gap between cylinder and plate vanishes. The integral (12) converges for h1/b1<1\Re h_{1}/b_{1}<1. In the opposite case, one of the denominators n(q)n^{(q)} in the sum Σ\Sigma in Eq. (10) vanishes for q=qc=logh/log(1+b)q=q_{c}=\log h/\log(1+b). We linearize the denominator at this point obtaining n(q)[dn(q)/dq](qqc)=qb2h2n^{(q)}\simeq[d\,n^{(q)}/d\,q](q-q_{c})=qb_{2}-h_{2}, where b2=hlog(1+b)b_{2}=h\log(1+b) and h2=hloghh_{2}=h\log h. The terms in Σ\Sigma with numbers close to q1=qcq_{1}=\Re q_{c} make maximum impact to the sum Σ\Sigma. The analytical continuation of the function S(b2,h2)S(b_{2},h_{2}) to the domain [h2/b2]<0\Re[h_{2}/b_{2}]<0 gives :

Σx2b2sin(x)S(h2,b2)\Sigma\simeq-\frac{x^{2}}{b_{2}\,\text{sin}(x)}-S\left(-h_{2},b_{2}\right) (13)

where x=πh2/b2=πloghlog(1+b)πlog|h|log(1+b)+iπκlog(1+b)x=\pi h_{2}/b_{2}=\frac{\pi\log h}{\log(1+b)}\simeq\frac{\pi\log|h|}{\log(1+b)}+i\frac{\pi\,\kappa}{\log(1+b)}. The first resonance term in Eq. (13) determines the singular behavior of the field when the loss factor κ1\kappa\ll 1. Suppose that the frequency is fixed, then the sum |Σ||\Sigma| oscillates as function of blog|h|b\ll\log\left|h\right| (see Fig. 2).

Refer to caption
Figure 2: Comparison of analytical (lines) and numerical (dots) electric field intensities |E/E0|2{|E/E_{0}|^{2}} in the gap between (a) Ag cylinder and Ag surface, (b) Au cylinder and Au surface, (c) Cu cylinder and Cu surface, (d) Si cylinder and Ag surface. Red, green and blue lines correspond to the excitation wavelengths of 785 nm/532 nm/405 nm, respectively. The excitation beam has an incidence angle of 4545^{\circ}, and a pp-polarization. The results are shown in logarithmic scale.

The oscillation period decreases with decreasing of the gap size. The local maxima |Σ|m|\Sigma|_{m} are achieved when x=mπ\Re x=m\pi. The value of these maxima is given by the following expression:

|Σ|mπm2|h2|sinh(πmκlog|h|)1|\Sigma|_{m}\simeq\frac{\pi m^{2}}{|h_{2}|}\sinh\left(\frac{\pi m\kappa}{\log|h|}\right)^{-1} (14)

This value linearly increases with the number mm: |Σ|mm/κ|\Sigma|_{m}\propto m/\kappa achieving absolute maximum for m=mmxlog|h|/πκm=m_{mx}\simeq\log|h|/\pi\kappa, namely, |Σ|mxlog|h|/(πκ2|h|)|\Sigma|_{mx}\simeq\log|h|/(\pi\kappa^{2}|h|). The oscillations collapse with further decrease the gap due to the losses in the system and |Σ|\left|\Sigma\right| exponentially drops down as function of bb. Substituting |Σ|mx|\Sigma|_{mx} in Eq. (9), we obtain the estimation of the limit of the electric field

|Emx|264|h|4log2|h|π2(h)4|E00|2\left|E_{mx}\right|^{2}\simeq\frac{64\left|h\right|^{4}\log^{2}|h|}{\pi^{2}(\Im h)^{4}}\left|E_{00}\right|^{2} (15)

where |E00|2=|E0xεm1|2+|E0yεd|2\left|E_{00}\right|^{2}=\left|E_{0x}\varepsilon_{m1}\right|^{2}+\left|E_{0y}\varepsilon_{d}\right|^{2}, ​ εm\varepsilon_{m}, εm1\varepsilon_{m_{1}} and εd\varepsilon_{d} are the permittivities of the metallic cylinder, metallic plate and surrounding space, respectively. In the red and infrared spectral ranges, the metal permittivity is large in absolute value (|εm|,|εm1|εd\left|\varepsilon_{m}\right|,\left|\varepsilon_{m_{1}}\right|\gg\varepsilon_{d}, εm/|εm|,εm1|εm1|1\Im\varepsilon_{m}/\left|\varepsilon_{m}\right|,\Im\varepsilon_{m_{1}}\left|\varepsilon_{m_{1}}\right|\ll 1) and Eq. (15) simplifies:

|Emx2||εm6||εm14|(|εm|+|εm1|)2εd2(|εm2|εm1+|εm12|εm)4|E00|2\left|E_{mx}^{2}\right|\propto\frac{\left|\varepsilon_{m}^{6}\right|\left|\varepsilon_{m_{1}}^{4}\right|\left(\left|\varepsilon_{m}\right|+\left|\varepsilon_{m_{1}}\right|\right)^{2}}{\varepsilon_{d}^{2}\left(\left|\varepsilon_{m}^{2}\right|\Im\varepsilon_{m_{1}}+\left|\varepsilon_{m_{1}}^{2}\right|\Im\varepsilon_{m}\right)^{4}}\left|E_{00}\right|^{2} (16)

The maximum field enhancement estimates as |Emx||εm|3εd1(εm)2|E0|\left|E_{mx}\right|\propto\left|\varepsilon_{m}\right|^{3}\varepsilon_{d}^{-1}\left(\Im\varepsilon_{m}\right)^{-2}\left|E_{0}\right| in the gap between cylinder and plate made of the same metal. For instance, the field could be as large as |Emx|>103|E0|\left|E_{mx}\right|>10^{3}\left|E_{0}\right| in the gap between the silver plate and a silver cylinder for the green light λ=532nm\lambda=532\,nm (Fig. 2a). It is an upper limit since the field could be restricted by radiation loss and the spatial dispersion of the electric charge in subnanometer gap 3.

3 Analytical theory vs. Simulations

In any real experiment, a layer of investigated molecule is placed in the gaps formed by metallic particles distributed over a metallic mirror. We use full-scale computer simulations in COMSOL environment to simulate the periodic array of metallic nanocylinders with the radius a=10nma=10\,nm, period of D=200nmD=200\,nm, and for “large" metallic nanocylinders, a=50nma=50\,nm and D=200nmD=200\,nm. The cylinders are placed over the metal surface and are excited by p-polarized light with an incidence angle of 4545^{\circ} similar to 10. The periodic boundary conditions and periodic ports are used to compute the reflectance as well as local electric field. Thus obtained, the enhancement of the gap electric field is in agreement with analytical theory as shown in Figs.2 a-c. The gap field could also achieve large values in the metal-semiconductor (dielectric) gap (Fig.2d).

Refer to caption
Figure 3: Analytical results in the gap between Ag cylinder and Ag surface as function of wavelength and d/ad/a ratio: (a) electric field intensity |E/E0|2{|E/E_{0}|^{2}} and (b) reflectance. Numerical results for “large" silver cylinder with a radius of a=50nma=50\,nm and a period of D=200nmD=200\,nm: (c) electric field intensity |E/E0|2{|E/E_{0}|^{2}} and (d) reflectance.

The field enhancement in the gap between a cylinder and the surface could revile itself in the anomalous reflectance (Fig. 3 b,d).

To compare the theory with computer simulations, we fix the electric field EsE_{s} at the point x=d+2a,y=0x=d+2a,y=0 on the surface of a cylinder and calculate the field E(x,y)E(x,y) in all the space by using our analytical theory. To simplify the consideration, the direct interaction between cylinders is neglected since the electric field is localized in the nanogaps with size dad\ll a. Then, we average the field over the layer 0<x<d+2a0<x<d+2a and introduce the effective permittivities εxx(e)=ε(x,y)Ex(x,y)/Ex(x,y)\varepsilon^{(e)}_{xx}=\left\langle\varepsilon(x,y)E_{x}(x,y)\right\rangle/\left\langle E_{x}(x,y)\right\rangle and εyy(e)=ε(x,y)Ey(x,y)/Ey(x,y)\varepsilon^{(e)}_{yy}=\left\langle\varepsilon(x,y)E_{y}(x,y)\right\rangle/\left\langle E_{y}(x,y)\right\rangle. Electromagnetic wave with an amplitude EinE_{in} and the pp-polarization is incident on the layer placed above the metal plate. Maxwell equations are solved in terms of the amplitude EinE_{in} in order to find electromagnetic field in the layer with thickness d+2ad+2a and permittivity ε^(e)\hat{\varepsilon}^{(e)}. In particular, we find the electric field E0={Ex0,Ey0}E_{0}=\{E_{x0},E_{y0}\} at the interface between the layer and the metal plate (see Fig. 1c). In the dilute case aDλa\ll D\ll\lambda, the field E0E_{0} equals to the field that were on the metal surface without cylinder, namely E0=Ein{εm1sin2θεdεm1εd2sin2θ+εm1cosθ,2cosθεdεm1εd2sin2θεdεm1εd2sin2θ+εm1cosθ}E_{0}=E_{in}\{\frac{\varepsilon_{m_{1}}\sin 2\theta}{\sqrt{\varepsilon_{d}\varepsilon_{m_{1}}-\varepsilon_{d}^{2}\sin^{2}\theta}+\varepsilon_{m_{1}}\cos\theta},\frac{2\cos\theta\sqrt{\varepsilon_{d}\varepsilon_{m_{1}}-\varepsilon_{d}^{2}\sin^{2}\theta}}{\sqrt{\varepsilon_{d}\varepsilon_{m_{1}}-\varepsilon_{d}^{2}\sin^{2}\theta}+\varepsilon_{m_{1}}\cos\theta}\}, where θ\theta is the angle of the incidence. This field is introduced in Eq. (9) in order to obtain the analytical esteem of the electric field in the nanogap. The theoretical results are in accordance with full-scale computer simulations for small cylinders (ka1ka\ll 1). The results of our quasistatic theory being extrapolated to ka1ka\sim 1 are still in a qualitative agreement with computer simulations of “large" nanocylinders (Fig.3). The agreement is better for the smaller gap size dd and higher plasmon resonance qq. The theory holds when the plasmon size Lqad/qL_{q}\sim\sqrt{ad}/q is smaller than the skin depth Lq1k|εm|L_{q}\ll\frac{1}{k\sqrt{\left|\varepsilon_{m}\right|}}. Figures 2 and 3 show that the largest field enhancement is achieved in red and infrared spectral bands, where optical loss in metal is relatively small. On the other hand, the silver-silicon system resonates for λ<550nm\lambda<550\,nm when the parameter h\Re h in Eq. (10) becomes larger than one.

In summary, we present a theory of the plasmons excited in the gap between a metal or dielectric nanocylinder and a metallic or dielectric plate. The resonance conditions for the gap plasmon are found as function of the metal permittivity and the gap size. When the cylinder-on-mirror cavity is excited by the incident light, the gap electric field increases and oscillates with decreasing the gap size achieving its limit value. Excitation of the gap plasmons results in the dips in the reflectance. The “qq-th" minimum of the reflectance corresponds to the excitation of the “qq-th" order plasmon. Fluctuations in the reflectance were probably observed, e.g., in experiments 3, 35. The quasistatic theory qualitatively describes the field enhancement up to ka1ka\leq 1, since the field concentrates in the nanogap whose size is much smaller than λ\lambda. Finally, this analytical theory can be used to design new SERS substrates and other optical sensors.

The work is supported by Russia Fund Basic Research grant 20-21-00080.

References

  • 1 F. J. Garcia-Vidal, and J. Pendry, Phys. Rev. Lett. 77, 1163 (1996).
  • 2 S. Grésillon et al., Phys. Rev. Lett. 82, 4520 (1999).
  • 3 C. Ciraci et al., Science 337, 1072 (2012).
  • 4 A. Moreau et al., Nature 492, 86 (2012).
  • 5 T. J. Echtermeyer et al., Nat. Commun. 2, 458 (2011).
  • 6 Y. Zhang et al., Nano Lett. 11, 5519 (2011).
  • 7 K. J. Savage et al., Nature 491, 574 (2012).
  • 8 J. Mertens et al., Nano Lett. 13, 5033 (2013).
  • 9 H. Cai et al., Opt. Express 24, 20808 (2016).
  • 10 S. Mubeen et al., Nano Lett. 12, 2088 (2012)
  • 11 R. Chikkaraddy et al., ACS Photonics 4, 4649 (2017).
  • 12 R. Chikkaraddy et al., Nano Lett. 18, 405 (2018).
  • 13 A. Demetriadou et al., ACS Photonics 4, 2410 (2017).
  • 14 K. Seal et al., Phys. Rev. Lett. 94, 226101 (2005).
  • 15 D. Genov, et al., Nano Lett. 4, 153 (2004).
  • 16 A. Ivanov et al., Appl. Phys. A 107, 17 (2012).
  • 17 L. L. Frumin, A. V. Nemykin, S. V. Perminov, and D. A. Shapiro, J. Opt. 30, 2048 (2013).
  • 18 Z.-q. Liu et al., Plasmonics 8, 1285 (2013).
  • 19 G.-q. Liu et al., Phys. Chem. Chem. Phys. 16, 4320 (2014).
  • 20 G.-q. Liu et al., Opt. Commun. 316, 111 (2014).
  • 21 I. L. Rasskazov, V. A. Markel, and S. V. Karpov, Opt. Spectrosc. 115, 666 (2013).
  • 22 F. Ding, Y. Yang, R. A. Deshpande, and S. I. Bozhevolnyi, Nanophotonics 7, 1129 (2018).
  • 23 L. Dong et al., Nano Lett. 17, 5768 (2017).
  • 24 J. Kneipp, ACS Nano 11, 1136 (2017).
  • 25 Y. Hu et al., ACS Nano 11, 5558 (2017).
  • 26 C. Andreou et al., ACS Nano 10, 5015 (2016).
  • 27 N. L. Nechaeva et al., Anal. Chim. Acta 1100, 2506 (2020).
  • 28 H. Chon et al., Chem. Commun. 50, 1058 (2014).
  • 29 Q. Fu et al., ACS Appl. Mater. Interfaces 7, 13322 (2015).
  • 30 T. Jiang et al., J. Am. Chem. Soc. 140, 15560 (2018).
  • 31 G. Liu et al., Sol. Energy Mater. Sol. Cells 186, 142 (2018).
  • 32 X. Lu et al., Chem. Mater. 30, 1989 (2018).
  • 33 C. Ma, Q. Gao, W. Hong, J. Fan, and J. Fang, Adv. Funct. Mater. 27, 1 (2016).
  • 34 J.-M. Nam, J.-W. Oh, H. Lee, and Y. D. Suh, Acc. Chem. Res. 49, 2746 (2016).
  • 35 K.N. Kanipe et al. J. Phys. Chem. C 121 14269 (2017)
  • 36 R. Pan et al., Nanoscale 10, 3171 (2018).
  • 37 Y. Shin, J. Song, D. Kim, and T. Kang, Adv. Mater. 27, 4344 (2015).
  • 38 D. O. Sigle et al., ACS Nano 9, 825 (2015).
  • 39 D. Yoo et al., Nano Lett. 18, 1930 (2018).
  • 40 J. Zhou et al., ACS Nano 10, 11066 (2016).
  • 41 M. Kraft et al., Phys. Rev. X 5, 031029 (2015).
  • 42 P. B. Johnson, and R. W. Christy, Phys. Rev. B 6, 4370 (1972).