Metamaterials and Negative-Index Media

Every result here is squeezed out of one equation the electromagnetic-optics head already owns: the plane-wave form of Maxwell's equations, $\mathbf{k}\times\mathbf{E}=\omega\mu\mu_0\mathbf{H}$ and $\mathbf{k}\times\mathbf{H}=-\omega\varepsilon\varepsilon_0\mathbf{E}$ . Read those relations for a transparent dielectric and the wave triad $(\mathbf{E},\mathbf{H},\mathbf{k})$ is right-handed, the phase index is $n=+\sqrt{\varepsilon_r\mu_r}$ , and Snell's law bends rays the familiar way. But nothing in Maxwell's equations demands that $\varepsilon$ and $\mu$ be positive. Feed them both negative and every sign in the same two lines flips: the triad becomes left-handed, the wavevector runs backward against the flow of energy, and the index that multiplies $\sin\theta$ turns negative. Natural materials rarely oblige — a negative $\varepsilon$ is common (any metal below its plasma frequency) but a negative $\mu$ at optical frequencies essentially never occurs in nature. A metamaterial manufactures what nature withholds: an artificial medium built from a periodic array of subwavelength inclusions — thin metal wires, split-ring resonators, dielectric spheres — with a lattice period $a\ll\lambda$ so small that the wave sees only smoothed, effective parameters $\varepsilon(\omega)$ and $\mu(\omega)$ the designer dials to order. Wires furnish a diluted electron gas with $\varepsilon<0$ ; split rings furnish nanoscale LC loops with $\mu<0$ . Overlap the two negative bands and the composite refracts light the wrong way. Veselago worked out the consequences in 1968 as a paper exercise; Pendry supplied the wire and ring recipes in the late 1990s; Smith and Shelby built and measured the first negative-index prism in 2000–2001. Out of it came the perfect lens, cloaking, and transformation optics — a whole engineering language for telling light where to go.

Key takeaways

A metamaterial is a composite whose subwavelength structure ( $a\ll\lambda$ ) produces effective $\varepsilon(\omega)$ and $\mu(\omega)$ unavailable in natural materials, including simultaneously negative values.
When both $\varepsilon_r<0$ and $\mu_r<0$ the phase index is $n=-\sqrt{\varepsilon_r\mu_r}<0$ : wavefronts run backward against the energy flow, and Snell's law refracts rays to the same side of the normal.
Thin-wire arrays give a Drude $\varepsilon_r(\omega)=1-\omega_p^2/\omega^2$ with a plasma frequency pulled from ultraviolet down to a few GHz; split-ring resonators give a Lorentzian $\mu_r(\omega)$ that dips negative just above their LC resonance.
Pendry's perfect lens: an $n=-1$ slab amplifies a source's evanescent spectrum, in principle beating the diffraction limit — capped in practice by loss.
Negative index is forbidden to be constant: the dispersive energy density stays positive only because $\varepsilon_r,\mu_r$ vary steeply near a resonance, so left-handed behaviour is intrinsically narrowband and lossy.
Key idea: simultaneously negative $\varepsilon$ and $\mu$ make the wavevector antiparallel to the energy flow — the medium is left-handed, the phase index is $n=-\sqrt{\varepsilon_r\mu_r}$ , and light refracts to the same side of the normal.

The core insight

Simultaneously negative $\varepsilon$ and $\mu$ make the wavevector antiparallel to energy flow: the medium is left-handed, the phase index is negative ( $n=-\sqrt{\varepsilon_r\mu_r}$ ), and Snell's law refracts light to the same side of the normal.

Where this lives today

The metamaterial idea long ago outgrew the bulky microwave prism. At radio and microwave frequencies it is now standard engineering: metamaterial resonators shrink antennas below the half-wavelength limit, leaky-wave antennas scan a beam from backward to forward angles by exploiting the reversed phase progression, and metasurfaces — single-layer arrays of subwavelength phase-shifting elements — steer 5G and satellite beams with no moving parts. The optical frontier has shifted from bulk negative index, where metallic loss is punishing, to the flat metalens: a dielectric metasurface of nanopillars that imprints an arbitrary phase profile and focuses light in a film a fraction of a wavelength thick, now shipping in commercial cameras. And on silicon photonic chips, subwavelength gratings emulate a designer effective index for low-loss fibre-to-chip coupling and label-free refractive-index sensing — a direct descendant of the homogenization argument that started with a wire and a ring.

The physical picture

Start with the two building blocks as circuits. A thin metal wire is a diluted electron gas: confine the conduction electrons to a wire of radius $r\sim1\ \mu\mathrm{m}$ on a lattice of period $a\sim5\ \mathrm{mm}$ and two things happen — the cell-averaged electron density collapses by the filling fraction $\pi r^2/a^2$ , and the magnetic field wrapped around each current-carrying wire adds a huge kinetic inertia, so the effective electron mass balloons. Both drive the plasma frequency down from the ultraviolet of bulk metal to a few gigahertz; below it the array reflects like a metal with negative permittivity, exactly as $\varepsilon_r=1-\omega_p^2/\omega^2$ dictates. A split-ring resonator is the magnetic partner: a gapped metal loop is a nanoscale LC circuit, the ring supplying inductance $L$ and the gap capacitance $C$ . An oscillating magnetic field threads the ring and drives a circulating current, and just above the resonance $\omega_0=1/\sqrt{LC}$ the induced moment overshoots and points against the drive — a negative permeability over a narrow band.

Figure

Effective ε_r and μ_r for a wire-plus-ring metamaterial versus frequency: the Drude wires give ε_r below zero everywhere under the plasma frequency f_p ≈ 8.2 GHz while the split rings give μ_r below zero only between the resonance f₀ = 4 GHz and the magnetic-plasma edge f_mp ≈ 5.7 GHz — the shaded overlap from 4.0 to 5.7 GHz is the negative-index band where n = −√(ε_r μ_r)

In an ordinary medium the crests march the way energy travels; in a left-handed medium picture them marching backward, into the source, while power still streams outward — a treadmill wave. That single reversal explains the whole zoo of left-handed effects at once: Snell refraction to the wrong side, an inverted Doppler shift, a reversed Cherenkov cone. When such a wave crosses a flat interface the tangential wavevector is preserved but causality forces energy away from the boundary, so the refracted ray tips to the same side of the normal as the incoming one — and a flat slab, with no curved surface, brings a diverging bundle back to a focus.

Figure

Negative refraction at a vacuum-to-metamaterial interface: a ray incident at θ_i = 30° in vacuum (n₁ = 1.0) refracts into n₂ = −1.3 not downward as ordinary Snell would send it dashed but upward to the same side of the normal at θ_t = −22.6°; inside the medium the energy flow S runs along the ray while the wavevector k points antiparallel, the signature of a backward wave

The perfect lens is the same reversal pushed to its limit. The fine detail of a source lives in evanescent components whose transverse spatial frequency $k_x$ exceeds $\omega/c$ ; these have imaginary longitudinal wavevector and decay exponentially in vacuum — precisely the diffraction limit. A slab with $\varepsilon=\mu=-1$ does not merely transmit them but resonantly amplifies them by the exact factor they lost in the gap, restoring subwavelength information at the image. And because $n(\omega)$ must sweep from positive to negative across a resonance, negative index is inseparable from strong dispersion — nature's bookkeeping forbids a static free lunch.

The core equations

The entire subject hangs on the sign of one square root. For an isotropic medium the phase index is the geometric mean of the relative permittivity and permeability, but the sign of the root is not free — it is fixed by requiring a passive (lossy) medium to absorb rather than amplify.

n(\omega)=\pm\sqrt{\varepsilon_r(\omega)\,\mu_r(\omega)},\qquad n<0\ \text{iff}\ \varepsilon_r<0\ \text{and}\ \mu_r<0 \qquad (1)

where: $n$ — phase refractive index; $\varepsilon_r$ — relative permittivity; $\mu_r$ — relative permeability; $\omega$ — angular frequency. Physical reading: the magnitude of $n$ follows the product $\varepsilon_r\mu_r$ as in ordinary optics, but the sign is fixed separately by causality — negative only in the doubly-negative quadrant. The reversal is written into the plane-wave form of Maxwell's equations, where Faraday's and Ampère's laws tie $\mathbf{k}$ , $\mathbf{E}$ , and $\mathbf{H}$ together with $\mu$ and $\varepsilon$ in front carrying their signs.

\mathbf{k}\times\mathbf{E}=\omega\mu\mu_0\mathbf{H},\qquad \mathbf{k}\times\mathbf{H}=-\omega\varepsilon\varepsilon_0\mathbf{E} \qquad (2)

where: $\mathbf{k}$ — wavevector (phase advance); $\mathbf{E},\mathbf{H}$ — field amplitudes; $\varepsilon_0,\mu_0$ — vacuum permittivity and permeability. Physical reading: with $\varepsilon,\mu>0$ the triad $(\mathbf{E},\mathbf{H},\mathbf{k})$ is right-handed; flip both signs and $\mathbf{k}$ reverses relative to the fields — the set turns left-handed. The consequence is a sign on the energy flow. The Poynting vector points along $\mathbf{E}\times\mathbf{H}$ , which the fields alone fix, so when $\mathbf{k}$ reverses the two point opposite ways.

\mathbf{S}=\tfrac12\,\mathrm{Re}(\mathbf{E}\times\mathbf{H}^*),\qquad \mathbf{k}\cdot\mathbf{S}<0\ \ \text{in a left-handed medium} \qquad (3)

where: $\mathbf{S}$ — time-averaged energy flux; $\mathbf{k}\cdot\mathbf{S}$ — projection of energy flow onto the phase-propagation direction. Physical reading: energy and wavefronts travel in opposite directions — the defining property of a backward wave, and the root cause of every reversed effect. The electric half of the recipe comes from thin metal wires, which behave as a dilute plasma. The effective permittivity is Drude-like, with a plasma frequency set entirely by geometry.

\varepsilon_r(\omega)=1-\frac{\omega_p^2}{\omega^2},\qquad \omega_p^2=\frac{2\pi c^2}{a^2\,\ln(a/r)} \qquad (4)

where: $\omega_p$ — effective plasma frequency of the wire lattice; $a$ — lattice period; $r$ — wire radius; $c$ — vacuum light speed. Physical reading: thinner wires and a wider lattice push $\omega_p$ down, so $\varepsilon_r<0$ is engineered across a chosen microwave band instead of inherited in the ultraviolet. The magnetic half comes from split-ring resonators, whose effective permeability is a Lorentzian resonance. Above the ring's LC resonance the response overshoots and $\mathrm{Re}\,\mu_r$ dips below zero up to a magnetic-plasma edge.

\mu_r(\omega)=1-\frac{F\,\omega^2}{\omega^2-\omega_0^2+i\gamma\omega},\qquad \omega_{mp}=\frac{\omega_0}{\sqrt{1-F}} \qquad (5)

where: $F$ — geometric filling factor of the ring array ( $0<F<1$ ); $\omega_0=1/\sqrt{LC}$ — LC resonance frequency; $\gamma$ — resistive damping; $\omega_{mp}$ — upper edge of the negative- $\mu$ band. Physical reading: $\mathrm{Re}\,\mu_r<0$ only in the narrow window $\omega_0<\omega<\omega_{mp}$ , so the wire and ring bands must be co-designed to overlap. A negative index cannot be frequency-independent, and the reason is energy. The correct dispersive energy density is not $\tfrac12(\varepsilon_0\varepsilon_r|E|^2+\mu_0\mu_r|H|^2)$ — which would go negative — but the Brillouin form with frequency derivatives, whose positivity is a real constraint.

u=\tfrac12\!\left[\frac{\mathrm{d}(\omega\varepsilon_r)}{\mathrm{d}\omega}\,\varepsilon_0|E|^2+\frac{\mathrm{d}(\omega\mu_r)}{\mathrm{d}\omega}\,\mu_0|H|^2\right]>0 \qquad (6)

where: $u$ — time-averaged electromagnetic energy density; $\mathrm{d}(\omega\varepsilon_r)/\mathrm{d}\omega$ and $\mathrm{d}(\omega\mu_r)/\mathrm{d}\omega$ — dispersive corrections that must stay positive. Physical reading: $u>0$ demands steep dispersion, achievable only near a resonance, so negative index is intrinsically narrowband — a constant $n<0$ would violate causality. Snell's law is unchanged in form but the negative index tilts the refracted ray to the opposite side. The tangential wavevector is conserved, and the sign of $n_2$ carries straight through.

n_1\sin\theta_i=n_2\sin\theta_t,\qquad n_2<0\Rightarrow\theta_t<0 \qquad (7)

where: $\theta_i,\theta_t$ — incidence and refraction angles measured from the normal; $n_1,n_2$ — indices of the two media. Physical reading: a negative $n_2$ forces a negative $\theta_t$ , i.e. the refracted ray lands on the same side of the normal as the incident ray — the observable hallmark of a left-handed medium. Finally the perfect lens. For evanescent components ( $k_x>\omega/c$ ) the longitudinal wavevector is imaginary, $k_z=i|k_z|$ , and an $\varepsilon=\mu=-1$ slab of thickness $d$ has a transmission that grows rather than decays.

t(k_x)\to e^{+|k_z|\,d}\quad\text{for}\ k_x>\omega/c,\ \ \varepsilon=\mu=-1 \qquad (8)

where: $t(k_x)$ — slab transmission for transverse spatial frequency $k_x$ ; $|k_z|=\sqrt{k_x^2-(\omega/c)^2}$ ; $d$ — slab thickness. Physical reading: the slab amplifies each evanescent component by exactly the factor it lost crossing the vacuum gap, restoring subwavelength detail — until real loss caps the amplification.

Derivation AWhy the negative root of the index is mandatory

The goal is to prove that the sign in $n=\pm\sqrt{\varepsilon_r\mu_r}$ is not a matter of taste: when both parameters are negative the physically correct branch is the negative one. We establish it two independent ways — by tracking the complex phase of the square root for a lossless left-handed medium, and by adding infinitesimal loss and demanding that the medium absorb rather than amplify — and both must reduce to the ordinary positive root when either parameter is positive.

Step 1. Write each parameter in polar form for a lossless doubly-negative medium: $\varepsilon_r=|\varepsilon_r|e^{i\pi}$ and $\mu_r=|\mu_r|e^{i\pi}$ , since a negative real number has phase $\pi$ . Then $n=\sqrt{|\varepsilon_r|e^{i\pi}\,|\mu_r|e^{i\pi}}=\sqrt{|\varepsilon_r\mu_r|}\;e^{i\pi}$ . Why: the two factors each contribute phase $\pi$ , their product carries $2\pi$ , and the square root halves that to $\pi$ — not $0$ — which is the whole crux of the sign.

Step 2. Evaluate the phase factor: $e^{i\pi}=-1$ , so $n=-\sqrt{|\varepsilon_r\mu_r|}$ . The negative branch is selected purely by the branch cut of the square root, with no extra physics assumed beyond the medium being doubly negative. Why: had only one parameter been negative the total phase would have been $\pi$ , the root phase $\pi/2$ , and $n$ imaginary — the evanescent, non-propagating case, not a negative real index.

Step 3. Confirm by passivity: give the parameters a small positive imaginary part, $\varepsilon_r=-|\varepsilon_r|+i\varepsilon''$ and $\mu_r=-|\mu_r|+i\mu''$ with $\varepsilon'',\mu''\to0^+$ . Demanding $\mathrm{Im}(n)>0$ so a forward wave decays rather than grows picks the same negative-real branch as the loss vanishes. Why: a passive medium must attenuate, so the loss-selection rule is unambiguous and agrees with the phase-tracking result — the negative root is forced, not chosen.

Go deeper · Derivation B · 1 minBackward-wave energy flow from the Poynting vector

Derivation BBackward-wave energy flow from the Poynting vector

The goal is to show directly from Maxwell's plane-wave relations that energy flow and phase progression point opposite ways whenever $\mu<0$ (and, symmetrically, whenever $\varepsilon<0$ ). Rather than assert left-handedness, we compute the time-averaged Poynting vector for a transverse plane wave, read off its direction relative to $\mathbf{k}$ , and check that the ordinary result reappears for a normal medium — turning the loose treadmill picture into a signed statement.

Step 1. From Faraday's law, equation (2), solve for the magnetic field of a plane wave: $\mathbf{H}=(\mathbf{k}\times\mathbf{E})/(\omega\mu\mu_0)$ . The magnetic amplitude inherits an explicit $1/\mu$ , which is where the sign of $\mu$ will enter. Why: isolating $\mathbf{H}$ in terms of $\mathbf{E}$ and $\mathbf{k}$ is what lets the constitutive sign propagate into the energy flux.

Step 2. Form $\mathbf{S}=\tfrac12\,\mathrm{Re}(\mathbf{E}\times\mathbf{H}^*)$ and substitute, using transversality $\mathbf{k}\cdot\mathbf{E}=0$ to expand the double cross product: $\mathbf{S}=\dfrac{|E|^2}{2\omega\mu\mu_0}\,\mathbf{k}$ . So $\mathbf{S}$ is parallel to $\mathbf{k}/\mu$ , with the permeability in the denominator. Why: the identity collapses everything to a scalar prefactor times $\mathbf{k}$ , making the sign question purely about the sign of $\mu$ .

Step 3. Read the sign. For $\mu>0$ the prefactor is positive and $\mathbf{S}\parallel\mathbf{k}$ — the familiar forward wave. For $\mu<0$ it flips and $\mathbf{S}$ is antiparallel to $\mathbf{k}$ : energy flows one way while crests advance the other. The identical conclusion follows from the $\varepsilon$ -form via Ampère's law. Why: recovering $\mathbf{S}\parallel\mathbf{k}$ for ordinary media confirms the algebra, and the sign flip is the microscopic origin of negative refraction and the reversed Doppler and Cherenkov effects.

Go deeper · Derivation C · 1 minPositivity of energy density forces strong dispersion

Derivation CPositivity of energy density forces strong dispersion

The goal is to prove that a negative index cannot be frequency-flat. A naive energy density built from the instantaneous constitutive relations would be negative in a doubly-negative medium — impossible for a passive material storing real energy. The resolution is that in a dispersive medium the stored energy includes the resonant motion of the bound charges (Brillouin's formula), and requiring it positive forces the constitutive functions to rise steeply with frequency — the steepness that ties negative index to a resonance and to loss.

Step 1. Write the naive density $u_{\text{naive}}=\tfrac12(\varepsilon_0\varepsilon_r|E|^2+\mu_0\mu_r|H|^2)$ . With $\varepsilon_r,\mu_r<0$ this is manifestly negative, which cannot be the energy stored in a passive medium. Why: exposing the contradiction pins down that the naive expression is simply the wrong bookkeeping once the medium responds resonantly.

Step 2. Replace it with the Brillouin dispersive form, equation (6), in which $\varepsilon_r$ and $\mu_r$ are promoted to $\mathrm{d}(\omega\varepsilon_r)/\mathrm{d}\omega$ and $\mathrm{d}(\omega\mu_r)/\mathrm{d}\omega$ . These derivatives account for energy temporarily parked in the oscillating charges and currents, not just in the fields. Why: only the frequency-derivative form conserves energy for a medium whose polarization lags the drive, so it is the physically correct density.

Step 3. Impose $u>0$ . Take the Drude wire result $\varepsilon_r=1-\omega_p^2/\omega^2$ : then $\mathrm{d}(\omega\varepsilon_r)/\mathrm{d}\omega=1+\omega_p^2/\omega^2>0$ always, even where $\varepsilon_r$ itself is negative — positivity is rescued entirely by the dispersion term, and only because $\varepsilon_r$ varies steeply. Why: this is the quantitative statement that a constant negative index is forbidden — steep dispersion near a resonance is mandatory, so left-handed media are intrinsically narrowband, consistent with Kramers–Kronig causality.

Limiting cases

The four sign combinations of $(\varepsilon_r,\mu_r)$ classify every isotropic medium and locate the left-handed quadrant among its neighbours.

Case	Result	Meaning
Right-handed ( $\varepsilon_r>0,\mu_r>0$ )	$n=+\sqrt{\varepsilon_r\mu_r}>0$	Ordinary transparent dielectrics; forward waves, normal Snell refraction
Electric plasma ( $\varepsilon_r<0,\mu_r>0$ )	$n^2<0\Rightarrow n$ imaginary	Wave evanescent, medium reflective — metals below $\omega_p$ , or a bare wire array
Magnetic plasma ( $\varepsilon_r>0,\mu_r<0$ )	$n^2<0\Rightarrow n$ imaginary	Also evanescent; the narrow band just above split-ring resonance, rings alone
Left-handed ( $\varepsilon_r<0,\mu_r<0$ )	$n=-\sqrt{\varepsilon_r\mu_r}<0$	Propagating backward wave, negative refraction; needs overlapping wire and ring bands
Impedance-matched ( $\varepsilon_r=\mu_r=-1$ )	$n=-1$ , $Z=Z_0$	Perfect-lens point: zero reflection, evanescent amplification $t=e^{+\|k_z\|d}$

Common trapA frequency-independent negative index is possible if you just build it carefully

It is not — and the sign of the square root is not yours to choose either. Two hard constraints bite. First, causality (Kramers–Kronig) forbids a resonant response flat in frequency, and the dispersive energy density (6) goes negative for any constant $\varepsilon_r,\mu_r<0$ , so every real negative-index material is narrowband and lossy near its operating point. Second, when both parameters are negative you must take $n=-\sqrt{\varepsilon_r\mu_r}$ , fixed by requiring $\mathrm{Im}(n)>0$ for a lossy medium (equivalently, that the Poynting vector point away from the source). Picking the positive root by habit gives a wave that grows without limit — the classic sign error. Negative index is a resonance phenomenon, not a material constant.

Worked examples

1 · Tuning a wire array to a microwave plasma frequency (confidence builder)

A negative-permittivity medium is made of thin metal wires of radius $r=1\ \mu\mathrm{m}$ arranged on a square lattice of period $a=5\ \mathrm{mm}$ . Using Pendry's thin-wire result $\omega_p^2=2\pi c^2/[a^2\ln(a/r)]$ , find the effective plasma frequency and compare it with bulk copper.

Step 1. Evaluate the geometric logarithm: $a/r=5\times10^{-3}/1\times10^{-6}=5000$ , so $\ln(a/r)=8.517$ . The denominator is $a^2\ln(a/r)=(5\times10^{-3})^2(8.517)=2.129\times10^{-4}\ \mathrm{m}^2$ .

Step 2. The numerator is $2\pi c^2=2\pi(2.998\times10^8)^2=5.648\times10^{17}\ \mathrm{m^2\,s^{-2}}$ . Hence $\omega_p^2=5.648\times10^{17}/2.129\times10^{-4}=2.653\times10^{21}\ \mathrm{s^{-2}}$ and $\omega_p=5.15\times10^{10}\ \mathrm{rad\,s^{-1}}$ .

Step 3. Convert to ordinary frequency: $f_p=\omega_p/2\pi=\mathbf{8.2\ GHz}$ . Bulk copper has $\omega_p\approx1.6\times10^{16}\ \mathrm{rad\,s^{-1}}$ , i.e. $f_p\approx2500\ \mathrm{THz}$ in the ultraviolet. The wire lattice has pulled the plasma frequency down by a factor of about $3\times10^5$ , so $\varepsilon_r<0$ now lives across the entire microwave band instead of the ultraviolet. The wire array behaves as a metal with $f_p = 8.2$ GHz, roughly $3\times10^5$ times lower than bulk copper.

2 · The angle of negative refraction

A ray in air ( $n_1=1.0$ ) strikes the flat face of a left-handed metamaterial with effective index $n_2=-1.3$ at an incidence angle $\theta_i=30°$ . Find the refraction angle, and contrast it with an ordinary glass of $n_2=+1.3$ .

Step 1. Apply Snell's law, equation (7): $\sin\theta_t=(n_1/n_2)\sin\theta_i=(1.0/-1.3)\sin30°=-0.5/1.3=-0.3846$ .

Step 2. Invert: $\theta_t=\arcsin(-0.3846)=\mathbf{-22.6°}$ . The magnitude, $22.6°$ , is identical to what an ordinary $n_2=+1.3$ glass would give — but the sign is negative.

Step 3. Read the geometry. Ordinary glass sends the refracted ray across the normal into the lower quadrant; here the negative sign keeps it on the same side as the incident ray, folded back across the normal. Trace a second ray and the refracted rays converge: a flat $n_2=-1$ slab of thickness $d$ brings a source at distance $\ell<d$ to an internal focus, then to an external focus $d-\ell$ beyond — total object-to-image separation $2d$ , a lens with no curved surface. $\theta_t = -22.6°$ : same magnitude as ordinary refraction, opposite side of the normal.

3 · How much a perfect lens amplifies the near field

A source has subwavelength structure at transverse spatial period $\Lambda=\lambda_0/5$ , so its dominant evanescent component has $k_x=2\pi/\Lambda=5\,(\omega/c)$ . It is imaged by an $\varepsilon=\mu=-1$ slab of thickness $d=\lambda_0/10$ . By what factor does the slab amplify this component, and what did the same distance of vacuum do to it?

Step 1. Find the decay constant. With $k_0\equiv\omega/c$ and $k_x=5k_0$ , the longitudinal wavevector is imaginary: $|k_z|=\sqrt{k_x^2-k_0^2}=k_0\sqrt{25-1}=\sqrt{24}\,k_0=4.90\,k_0$ .

Step 2. Form the exponent for the slab thickness $d=\lambda_0/10$ . Since $k_0=2\pi/\lambda_0$ , we get $|k_z|\,d=4.90\,(2\pi/\lambda_0)(\lambda_0/10)=4.90\times0.628=3.08$ .

Step 3. The slab transmission, equation (8), is $t=e^{+|k_z|d}=e^{3.08}=\mathbf{21.7}$ . Over the same thickness of empty space that component would have decayed by $e^{-3.08}=0.046$ — a factor of about $22$ down. The slab restores exactly what the vacuum gap removed, the mechanism of subwavelength imaging. A modest loss (finite $\mathrm{Im}\,\varepsilon$ ) detunes the resonance and caps the gain, which is why real superlenses reach a few times the diffraction limit, not infinity. The slab amplifies the evanescent component about $22\times$ , precisely undoing the vacuum decay.

Physics of the portrait

Negative-index optics is unusual in physics: a fully worked-out theory sat idle for three decades because no material could realize it, and one person supplied the missing designs that woke it up.

Origins · 1 minJohn Pendry and how this was discovered

John Pendry

1943– · proposed the structured-media route to negative ε and μ and the perfect lens

Sir John Brian Pendry (born 1943), a condensed-matter theorist at Imperial College London, turned negative-index media from a curiosity into an engineering discipline. Victor Veselago had shown in 1968 that a material with simultaneously negative $\varepsilon$ and $\mu$ would refract light backward, invert the Doppler shift, and reverse radiation pressure — but no such material existed, and the paper drew a handful of citations a year for thirty years. Pendry supplied the recipes. In 1996 he showed that a lattice of thin metal wires behaves as a dilute plasma with a plasma frequency dragged down to gigahertz, delivering $\varepsilon<0$ on demand; in 1999 that split-ring resonators, tiny nonmagnetic LC loops, produce $\mu<0$ just above their resonance. Overlap the two and you have a negative index — which David Smith's group built in 2000 and Shelby measured in a microwave prism in 2001. Then in 2000 Pendry pushed further with "Negative Refraction Makes a Perfect Lens," arguing that an $n=-1$ slab restores the evanescent spectrum and beats the diffraction limit — a claim that provoked years of productive controversy before near-field superlenses confirmed the core physics. In 2006, with Smith and Schurig, he formulated transformation optics and demonstrated the first microwave cloak. The through-line is one idea: structure, not chemistry, sets the electromagnetic response.

History

1968

— Veselago analyzes left-handed media with simultaneously negative ε and μ and predicts backward waves and reversed refraction

1996

— Pendry shows thin-wire arrays act as a low-frequency electric plasma, giving effective ε below zero

1999

— Pendry introduces the split-ring resonator, giving effective μ below zero just above its LC resonance

2000

— Smith and co-workers build the first negative-index metamaterial; Pendry predicts the perfect lens

2001

— Shelby, Smith and Schultz verify negative refraction in a microwave wedge experiment

2006

— Transformation optics is formulated and the first microwave invisibility cloak is demonstrated

Going further

Texts. Saleh and Teich, Fundamentals of Photonics, 3rd ed., §5.7 and the metamaterials chapter · Hecht, Optics, 5th ed., §4 (refraction, dispersion, and the negative-index note) · Jackson, Classical Electrodynamics, 3rd ed., Ch. 7 (dispersion, causality, energy density) · Landau and Lifshitz, Electrodynamics of Continuous Media, §80 (the Brillouin dispersive energy density) · Ramakrishna and Grzegorczyk, Physics and Applications of Negative Refractive Index Materials.

Provenance. The sign-of-index result and the reversed Snell law follow Veselago (1968); a verifier should recompute the branch of $\sqrt{\varepsilon_r\mu_r}$ for both-negative parameters and confirm $n<0$ , and check $\mathrm{Im}(n)>0$ under infinitesimal loss. The Drude thin-wire permittivity with $\omega_p^2=2\pi c^2/[a^2\ln(a/r)]$ and the split-ring Lorentzian permeability are Pendry, Holden, Robbins and Stewart, IEEE Trans. Microwave Theory Tech. 47, 2075 (1999); the worked wire-array value $f_p=8.2$ GHz for $a=5$ mm, $r=1\ \mu$ m should be reproduced. The dispersive energy density is the Brillouin form (Landau–Lifshitz); the derivative check $\mathrm{d}(\omega\varepsilon_r)/\mathrm{d}\omega=1+\omega_p^2/\omega^2>0$ is exact. The evanescent transmission $t=e^{+|k_z|d}$ and the perfect lens are Pendry, Phys. Rev. Lett. 85, 3966 (2000); the amplification factor $e^{3.08}\approx21.7$ for the stated geometry should be recomputed. Experimental negative refraction: Smith et al., Phys. Rev. Lett. 84, 4184 (2000) and Shelby, Smith and Schultz, Science 292, 77 (2001). No numerical constants beyond standard textbook forms are asserted.

Prerequisites. Waveguide Modes and the Wave Equation · Total Internal Reflection and Optical Guiding · Fiber Attenuation and Loss Mechanisms. Leads to. Integrated Photonic Waveguides and Circuits · Microring Resonators and Optical Cavities on Chip · Dispersion Management and Compensation.

Ask the tutor: Want to see how the wire radius and lattice spacing trade off to place the plasma frequency exactly where the split-ring band sits — or how a single fishnet meta-atom delivers both negative responses at once?

Is a negative refractive index the same as an evanescent or imaginary index?

No — these are different quadrants of the sign map. An imaginary index means $n^2<0$ , which happens when exactly one of $\varepsilon_r$ or $\mu_r$ is negative: the wave does not propagate, it decays evanescently, and the medium reflects — a metal below its plasma frequency ( $\varepsilon<0$ , $\mu>0$ ) is the everyday example. A genuinely negative real index needs both $\varepsilon_r<0$ and $\mu_r<0$ together, so that $n^2=\varepsilon_r\mu_r>0$ and the wave propagates — but as a backward wave with $n=-\sqrt{\varepsilon_r\mu_r}$ . So negative index is a propagating phenomenon, distinct from the non-propagating single-negative cases, which is why both a wire array and a ring array are required at once.

Does negative index mean signals travel backward or faster than light?

Neither. The wavevector points backward, so the phase velocity is antiparallel to the energy flow — the crests appear to move toward the source. But a signal rides the group velocity and the Poynting energy flow, both of which move forward, away from the source, at a speed at or below $c$ . Nothing overtakes light and no information runs backward. The reversal is a statement about wavefronts, not about the causal front of a pulse — the same distinction that lets an ordinary medium have a phase index below one near a resonance without violating relativity.

Why do you need both wires and split rings — can one structure do both?

Because $\varepsilon$ and $\mu$ are independent responses to the electric and magnetic parts of the wave, and the two classic structures each address one. Thin wires respond to the electric field, carrying a current that mimics a dilute plasma and delivering $\varepsilon_r<0$ below the engineered plasma frequency. Split rings respond to the magnetic field threading the loop, resonating like an LC circuit and delivering $\mu_r<0$ just above resonance. A negative index needs both negative in the same window, so the bands must be co-designed to overlap. Modern designs do sometimes merge both functions into one meta-atom — fishnet structures and paired nanorods carry electric and magnetic resonances together — but the requirement of two overlapping negative responses is unchanged.

How can a flat slab focus light with no curved surface?

Through negative refraction. An ordinary lens uses curvature to bend each ray by a different amount and bring a bundle to a focus; in a Veselago flat lens the bending happens at a flat face, because a negative index sends every refracted ray to the opposite side of the normal. A diverging bundle entering an $n=-1$ slab converges to an internal focus, diverges, then refracts again at the exit face to a second, external focus. A slab of thickness $d$ images a source a distance $\ell$ in front of it to a point $d-\ell$ beyond, so it works only for sources closer than $d$ and gives no magnification — a near-field relay, not a camera lens.

What ultimately limits the resolution of a perfect lens?

Loss and imperfection, not the principle. The ideal argument assumes $\varepsilon=\mu=-1$ exactly and a lossless slab, giving unbounded amplification of the evanescent spectrum and formally unlimited resolution. Any real material has a finite imaginary part in $\varepsilon$ and $\mu$ , which detunes the surface-plasmon resonance doing the amplifying and caps the growth; finite thickness, meta-atom granularity (the effective-medium picture fails as $k_x$ approaches $2\pi/a$ ), and deviations of $n$ from $-1$ degrade it further. In practice near-field superlenses — including silver-film versions using only $\varepsilon<0$ for TM polarization — reach a few times below the diffraction limit, a real gain but far short of the ideal.

What is transformation optics, and how is it related to negative index?

Transformation optics is the design principle that grew out of the same effective-medium thinking. Maxwell's equations keep their form under a coordinate transformation if you reinterpret the stretching of space as a spatially varying $\varepsilon$ and $\mu$ tensor. So to route light along any desired path, imagine deforming space to carry it there, read off the material recipe the deformation demands, and build it as a graded metamaterial. Steering light around a hidden region gives a cloak; other transformations give flat lenses, beam benders, and field concentrators. Negative index is the historical entry point: it proved $\varepsilon$ and $\mu$ could be engineered almost arbitrarily — including negative and near-zero values — which is exactly the design freedom transformation optics assumes.