Main

Cuprates are strongly anisotropic superconductors in which transport is made three-dimensional by Josephson tunnelling between the Cu–O planes. Tunnelling reduces the superfluid density in the direction perpendicular to the planes, and hence the frequency of the plasmon to below the average pair breaking gap. Weakly damped oscillations of the superfluid sustain transverse Josephson plasma waves (JPWs) that propagate along the planes.

Consider a complex superconducting order parameter in the ith Cu–O plane ψi(x, y, t) = |ψi(x, y, t)| expi(x, y, t), which depends on two in-plane spatial coordinates x and y and on time t. For a terahertz-frequency optical field polarized perpendicular to the planes, excitations above the superconducting gap are negligible and the modulus of the order parameter |ψ| 2 (number of Cooper pairs) is nearly constant in space and time. Hence, the electrodynamics is dominated by the order-parameter phase θi(x, y, t). Ignoring at first the spatial dependence of the phase, the local tunnelling strength can, from the Josephson equations10, be expressed in terms of an equivalent inductance, L, which depends on the local interlayer phase difference θi, i+1(t) = θi(t) − θi+1(t) as L(θi, i+1(t)) L0/cos(θi, i+1(t)) (i and i + 1 are the indices for two neighbouring layers). Here L0 = /2eIc is the inductance at equilibrium, the reduced Planck’s constant, 2e the Cooper pair charge and Ic the critical current. Denoting the capacitance of the Cu–O planes with a constant C, we express the Josephson plasma resonance (JPR) frequency as ωJP2 = 1/(L(θi, i+1(t))C) = ωJP02 cos[θi, i+1(t)], where ωJP02 = 1/L0C is the equilibrium value. Correspondingly, the oscillator strength f ωJP2 for the plasma oscillations11 is also a function of the interlayer phase, and scales as f = f0 cos[θi, i+1(t)]. The dependence of the oscillator strength f on the cosine of the superconducting phase corresponds to a third-order optical nonlinearity.

According to the second Josephson equation10, the interlayer phase difference θi, i+1(t) advances in time with the time integral of the interlayer voltage drop, as [θi, i+1(t)]/∂t = 2eV/. For an optical field made resonant with the Josephson plasma frequency E(t) = E0sin(ωJP0t), the interlayer phase oscillates as θi, i+1(t) = θ0 cos(ωJP0t), where E0 is the field amplitude and θ0 = (2ed/ωJP0)E0 (d 1 nm is the interlayer distance). This implies that the oscillator strength f(t) = f0 cos(θ0 cos(ωJP0t)) ≈ f0(1 − (θ02 + θ02 cos(2ωJP0t))/4) is modulated at a frequency 2ωJP0, whenever the field E0 is large enough to make the phase excursion θ0 sizeable.

Figure 1 provides a pictorial representation of this physics. We plot a vector that represents both the phase difference θi, i+1(t) (vector angle) and the oscillator strength f(t) (vector length).

Figure 1: Schematic representation of Josephson plasma waves.
figure 1

Schematic time-dependent representation of JPWs in the linear and nonlinear regime in the presence of a driving field E(t) = E0 sin(ωJP0t) polarized along the out-of-plane direction of a layered superconductor. Red arrows indicate the Josephson phase, while the corresponding oscillator strength f is represented by the black circle area. A JPW in the linear regime consists of small-amplitude modulations of θi, i+1 at constant oscillator strength f ωJP02. In the nonlinear regime, the Josephson phase oscillates at ωJP0, whereas f is modulated at 2ωJP0.

This picture shows how, for small driving fields, only θi, i+1(t) oscillates at the driving frequency ωJP0, whereas for larger fields these oscillations are accompanied by a 2ωJP0 modulation of the oscillator strength f(t).

Note also that the phenomena discussed above can be cast in terms of a Mathieu equation (see Supplementary Information 2). Thus, a 2ωJP0 modulation of the oscillator strength can serve as a pump for the parametric amplification of a second, weak plasma wave at frequency ωJP0. In this paper we demonstrate experimentally this effect in La1.905Ba0.095CuO4 (LBCO9.5), a cuprate superconductor with the equilibrium JPR at ωJP00.5 THz.

Terahertz pulses, generated with a photoconductive antenna12, were used as a weak probe of JPWs (a schematic drawing of the measurement geometry is reported in Supplementary Information 1). A typical terahertz-field trace13 reflected by the sample is shown in Fig. 2a. Two different measurements are displayed: one taken below (red line) and the other one above (black line) the superconducting transition temperature Tc = 32 K. In the superconducting state, long-lived oscillations with 2 ps period were observed on the trailing edge of the pulse, indicative of the JPR at ωJP00.5 THz. Figure 2b (solid red line) displays the corresponding reflectivity edge in frequency domain. The solid lines in Fig. 2c, d are the complex dielectric permittivity ɛ(ω) and the loss function L(ω) = −Im(1/ɛ(ω)) = ɛ2(ω)/(ɛ1(ω) + ɛ2(ω))2. L(ω) peaks at ωJP0, where the real part of the dielectric permittivity, ɛ1(ω), crosses zero.

Figure 2: Linear JPWs in LBCO9.5.
figure 2

aEprobe(τ) measured in the absence of a pump field both above and below Tc = 32 K. bd, Frequency-dependent, c-axis reflectivity at T = 5 K (b) and corresponding real and imaginary part of the complex permittivity (c) and energy loss function (d). Solid lines are based on the values extracted from the Eprobe(τ) trace of a. Dashed lines were calculated by numerically solving the sine-Gordon equation in the linear regime. e,f, Simulated phase θi, i+1(x, t) (e) and corresponding oscillator strength f (no change) (f) induced by a weak probe terahertz field. Horizontal dotted lines indicate the spatial coordinate x at which the line cuts are displayed (lower panels).

These optical properties could be reproduced well by solving the wave equation in the superconductor in one dimension8 (see Supplementary Information 3), which yields the space- and time-dependent order-parameter phase θi, i+1(x, t) (Fig. 2e) and the corresponding changes (negligible in linear response regime) of the oscillator strength f = f0 cos[θi, i+1(x, t)] (Fig. 2f). The reflectivity, complex permittivity, and loss function (dashed lines in Fig. 2b–d, respectively), calculated from these simulations by solving the electromagnetic field at the sample surface8, are in good agreement with the experimental data.

Amplification of a weak JPW like the one above (probe field) was achieved by mixing it with a second, intense pump field, which resonantly drove the Josephson phase to large amplitudes. Quasi-single cycle terahertz pulses, generated in LiNbO3 with the tilted pulse front method14 (yielding field strengths up to 100 kV cm−1), were used to excite these waves in the nonlinear regime. The spectral content of these pulses extended between 0.2 and 0.7 THz, centred at the JPR frequency (see Supplementary Information 4). Note that the pump field strength used in this experiment exceeds the expected threshold to access the nonlinear regime, defined by θ0 = (2eE0d)/(ωJP0) 1 and corresponding in this material to field amplitudes E0 = (ωJP0)/(2ed) 20 kV cm−1.

In Fig. 3, we report the time-delay-dependent, spectrally integrated pump–probe response of LBCO9.5. Changes in the reflected probe field were measured at one specific probe sampling time (τ = τ0) as a function of pump–probe time delay (t). For a system in which the optical properties are dominated by a single plasma resonance, the spectrally integrated response is proportional to the plasma oscillator strength f.

Figure 3: Nonlinear JPWs in LBCO9.5.
figure 3

a, Normalized spectrally integrated pump–probe response ΔEprobe/Eprobe(t, τ0 = 2 ps). Experimental data are shown along with calculations based on the sine-Gordon equation in the nonlinear regime (displayed with a vertical offset). Dashed lines indicate an average reduction that accompanies the oscillations (see model). The reduction was subtracted through Fourier filtering (>0.2 THz) to obtain the detail of the oscillations shown on the right. The signal buildup region affected by perturbed free induction decay is shaded in grey. b, Fourier transform of the extracted oscillations, showing a peak at 1 THz. c,d, Phase θi, i+1(x, t) (c) and corresponding oscillator strength f (d) induced by a strong terahertz pump field, as determined by numerically solving the sine-Gordon equation in the nonlinear regime. Horizontal dotted lines indicate the spatial coordinate x at which the line cuts are displayed (lower panels).

As shown in Fig. 3a, b, this integrated response exhibits a reduction of the signal and oscillations at a frequency 2ωJP0. Note that the oscillation frequency did not depend on the pump electric field strength E0, whereas the frequency reduced when the base temperature of the experiment was increased, consistent with the reduction of the equilibrium ωJP0 (see Supplementary Information 5 and 6). The effect completely disappeared at T > Tc.

Hence, the theoretically predicted 2ωJP0 modulation of the total oscillator strength f (see above) is reproduced well by the data in Fig. 3. This response could also be simulated using the space- and time-dependent sine-Gordon equation (see Fig. 3c, d), yielding good agreement between experiment and theory (see blue lines in Fig. 3a, b).

Note that here we only analyse pump–probe delays t 0 ps, because the response at the earliest times suffers from perturbed free induction decay15. This effect consists in the deformation of a coherent signal, which occurs when the pump strikes the sample during the oscillatory relaxation of the probe (for t 0 ps in our case).

Selected time-domain probe traces measured before and after excitation are displayed in Fig. 4. Crucially, at specific time delays the probe field is amplified (Fig. 4a), whereas at other delays it is suppressed (Fig. 4b) with respect to that measured at equilibrium.

Figure 4: Amplification and suppression of plasma oscillations.
figure 4

a,b, Eprobe(t, τ) traces measured by scanning the electro-optic sampling time τ at selected pump–probe delays t = 0 ps and t = 2 ps, respectively. Data are shown along with the same quantity measured at equilibrium (pump off). Plasma oscillations on the trailing edge of the pulses (τ 2 ps) are highlighted by thicker lines. Coloured shading indicates amplification (a) and suppression (b) of the JPW amplitude.

In Fig. 5 we report the time-delay-dependent and frequency-dependent loss function L(t, ω) = −Im(1/ɛ(t, ω)) = ɛ2(t, ω)/(ɛ1(t, ω) + ɛ2(t, ω))2, a quantity that peaks at the zero crossing of ɛ1(t, ω) and is always positive for a dissipative medium (that is, a medium with ɛ2(t, ω) > 0). The experimental data of Fig. 5a show that, after excitation, L(ω) acquires negative values around ωJP0 (red regions). This is indicative of a negative ɛ2(t, ω), and hence amplification. The effect is strong near zero pump–probe time delay, then disappears after 1 ps, and is observed again periodically with a repetition frequency of 2ωJP0. The same effect appears also in the simulations (Fig. 5b), yielding periodic amplification at a repetition frequency of 2ωJP0.

Figure 5: Time-delay-dependent and frequency-dependent loss function.
figure 5

a,b, Time-delay-dependent and frequency-dependent loss function L(t, ω) determined experimentally (a) and by numerically solving the sine-Gordon equation in nonlinear regime (b). Note that experimental and simulated L(ω) in the region between t = −4 ps and t = −2 ps have been multiplied by a factor of five to be better visualized with the other data.

In Supplementary Information 7 we report additional quantitative estimates of the degree of amplification. We include a negative absorption coefficient and a reflectivity larger than 1 at ω ωJP0. The extracted values are α = (−0.090 ± 0.003) μm−1 and R = (1.042 ± 0.008), respectively.

The data and simulations reported here demonstrate that terahertz JPWs can be parametrically amplified, exhibiting the expected sensitivity to the relative phase of strong and weak fields mixed in this process and the oscillatory dependence at twice the frequency of the drive.

Parametric amplification of terahertz light based on nonlinear optical techniques has already been shown in the past16. However, the physics demonstrated here extend beyond potential applications in photonics, directly leading to coherent parametric control of the superfluid in layered superconductors, and providing a means to manipulate the properties of the material or to probe them in new ways17.

Moreover, the ability to amplify a plasma wave could lead to terahertz single-phonon manipulation devices that may operate above 1 K temperatures. These would exploit concepts that to date have been developed only at microwave frequencies and in the millikelvin regime18,19,20,21. Finally, the parametric phenomena discussed here can also potentially be used to squeeze9,22,23 the superfluid phase, and may lead to control of fluctuating superconductivity24, perhaps even over a range of temperatures above Tc (refs 25,26).

Methods

Laser pulses with 100 fs duration and 5 mJ energy from a commercial Ti:Sa amplifier were split into three parts (92%, 7%, 1%). The most intense beam was used to generate strong-field terahertz pulses with energies up to 3 μJ via optical rectification in LiNbO3 with the tilted pulse front technique. These were collimated and then focused at normal incidence onto the sample (with polarization perpendicular to the Cu–O planes, that is, along the c axis) using a Teflon lens and a parabolic mirror, with focal lengths of 150 mm and 75 mm, respectively. The pump spot diameter at the sample position was 2.5 mm. The pump field strength was calibrated with electro-optic sampling in a 0.2-mm-thick GaP crystal, yielding a maximum value of 100 kV cm−1 (see Supplementary Information 4).

The 7% beam was used to generate the terahertz probe pulses with a photoconductive antenna. These had a dynamic bandwidth of 0.1–3 THz, corresponding to a time resolution of 250 fs. The c-axis optical properties of the superconductor (both at equilibrium and throughout the pump-induced dynamics) were probed in reflection geometry, with a probe incidence angle of 45° and a spot diameter at the sample position of 2 mm. The reflected probe pulses were electro-optically sampled in a 1-mm-thick ZnTe crystal, using the remaining 1% of the 800 nm beam. This measurement procedure returned the quantity Eprobe(t, τ), with t being the pump–probe delay and τ being the electro-optic sampling time coordinate.

The sample used in our experiment was a single crystal of La1.905Ba0.095CuO4 cut and polished along an ac oriented surface 3 × 3 mm in size. Its equilibrium optical response in the superconducting state was determined by measuring the complex-valued Eprobe(ω) (pump off) both at T < Tc and T > Tc, and by referencing it to the normal-state reflectivity measured in another crystal coming from the same batch of samples27.

The spectrally integrated pump–probe traces of Fig. 3 were measured by scanning the pump–probe delay t at a fixed sampling time τ = τ0. This was chosen to be on the trailing edge of the pulse, where the JPR oscillations are present. Note that the observed dynamics, and in particular the 2ωJP0 oscillations, did not depend significantly on the specific τ0 value at which the scan was performed.

The frequency and time-delay-dependent loss function of Fig. 5 (as well as all complex optical properties of the perturbed material) was determined by applying Fresnel equations11 to the pump-induced changes in the reflected electric field. These were normalized by independently recording Eprobe(t, τ) in the presence and absence of the terahertz pump field. Note that there was no need to take into account any pump–probe penetration depth mismatch in the calculation.

In the simulations, the Josephson phase evolution θi, i+1(x, t) was determined through the one-dimensional sine-Gordon equation6:

γ being a damping constant, c the speed of light, ɛr the dielectric permittivity, and ωJP0 the equilibrium JPR frequency. This equation was solved numerically, with the terahertz pump and probe fields overlapping at the vacuum/superconductor interface. For more details on this topic, we refer the reader to Supplementary Information 3.

Data availability.

The data supporting the plots within this paper and other findings of this study are available from the corresponding author on request.