One of the most surprising predictions of modern quantum theory is that the vacuum of space is not empty. In fact, quantum theory predicts that it teems with virtual particles flitting in and out of existence. Although initially a curiosity, it was quickly realized that these vacuum fluctuations had measurable consequences—for instance, producing the Lamb shift1 of atomic spectra and modifying the magnetic moment of the electron2. This type of renormalization due to vacuum fluctuations is now central to our understanding of nature. However, these effects provide indirect evidence for the existence of vacuum fluctuations. From early on, it was discussed whether it might be possible to more directly observe the virtual particles that compose the quantum vacuum. Forty years ago, it was suggested3 that a mirror undergoing relativistic motion could convert virtual photons into directly observable real photons. The phenomenon, later termed the dynamical Casimir effect4, 5, has not been demonstrated previously. Here we observe the dynamical Casimir effect in a superconducting circuit consisting of a coplanar transmission line with a tunable electrical length. The rate of change of the electrical length can be made very fast (a substantial fraction of the speed of light) by modulating the inductance of a superconducting quantum interference device at high frequencies (>10 gigahertz). In addition to observing the creation of real photons, we detect two-mode squeezing in the emitted radiation, which is a signature of the quantum character of the generation process.
Figures at a glance
Figure 1: Experimental overview. 
a, Optical micrograph of sample 2. Light parts are Al, which fills most of the image, while the dark parts are the Si substrate, visible where the Al has been removed to define the transmission lines. The output line is labelled CPW and the drive line enters from the top. Both lines converge near the SQUID (boxed). b, A scanning-electron micrograph of the SQUID. The SQUID has a vertical dimension of 13 μm and a normal state resistance of 218 Ω (170 Ω) implying LJ(0) = 0.23 nH (0.18 nH) for sample 2 (sample 1). A basic electrical characterization of the SQUID is presented in Supplementary Fig. 1. c, A simplified schematic of the measurement set-up. The SQUID is indicated by the box with two crosses, suggestive of the SQUID loop interrupted by Josephson junctions. A small external coil is also used to apply a d.c. flux bias through a lowpass filter (LP). The driving line has 36 dB of cold attenuation, along with an 8.4–12 GHz bandpass filter (BP). The filter ensures that no thermal radiation couples to the transmission line in the frequency region were we expect DCE radiation. (For sample 1, the last 6 dB of attenuation were at base temperature.) The outgoing field of the CPW is coupled through two circulators to a cryogenic low-noise amplifier (LNA) with a system noise temperature of TN ≈ 6 K. At room temperature, the signal is further amplified before being captured by two vector microwave digitizers. The dashed boxes delineate portions of the set-up at different temperatures, T, which are labelled.
Figure 2: Photons generated by the dynamical Casimir effect. 
Here we show the output flux of the transmission line while driving sample 1 at fd = 10.30 GHz. a, b, Broadband photon generation. We plot the dimensionless photon flux density, nout (photons s−1 Hz−1), which is the measured power spectral density normalized to the photon energy,
ω, as a function of pump power and detuning, δω/2π. Panel a shows negative detunings (axis reversed), while b shows positive detunings. The symmetry of the spectrum is apparent. Positive and negative detunings are recorded simultaneously. The plots are stitched together from several separate scans, between which we have changed image rejection filters at the input of the analysers. c, d, The photon flux density for positive and negative detunings averaged over frequency (at fixed power) for two different symmetric bands, showing the symmetry of the spectrum. Error bars, s.d. e, A section through a at δω/2π = −764 MHz, along with a fit to the full theory of ref. 17.
Figure 3: Two-mode squeezing of the DCE field. 
