Half-wave voltage (Vπ)

Apply a voltage to an electro-optic phase shifter and the optical phase at its output moves. The half-wave voltage $V_\pi$ is the voltage that moves it by exactly $\pi$: half an optical wave. In a Mach-Zehnder modulator the transfer function is

so $V_\pi$ is the voltage separation between a transmission maximum and the adjacent minimum, and the pattern repeats every $2V_\pi$. The bias phase $\phi_b$ sets the operating point; digital links usually hold the modulator at quadrature, halfway down the slope.

Why it is the headline number. The drive electronics must swing of order $V_\pi$ to switch the modulator fully, and RF driver power scales as the square of that swing. A modulator with half the $V_\pi$ needs a quarter of the drive power, which at datacenter volumes is the difference between needing a driver amplifier and running directly off CMOS. "CMOS-compatible" in modulator papers means $V_\pi$ around one volt.

VπL: the length trade. In a traveling-wave device the phase shift accumulates along the electrode, so $V_\pi$ falls as the modulator gets longer. The platform figure of merit is therefore the product $V_\pi L$ (V·cm), which is fixed by the material and cross-section:

A length of centimeters buys a low $V_\pi$ but costs optical loss and demands velocity matching between the RF and optical waves, so a companion figure $\alpha \cdot V_\pi L$ (V·dB) is often quoted: how much loss you pay per volt saved.

Fine print worth knowing. $V_\pi$ is wavelength-dependent (the phase shift is $2\pi \Delta n L/\lambda$) and frequency-dependent: the DC value measured with a slow sweep is smaller than the value at 50 GHz, where electrode loss and velocity mismatch dilute the interaction, which is why datasheets distinguish DC $V_\pi$ from RF $V_\pi(f)$. In silicon, the plasma-dispersion phase shift is not linear in voltage, so the quoted $V_\pi$ depends on the bias around which it was measured. In lithium niobate, DC bias drift slowly walks $\phi_b$, which is a measurement trap as much as an operational one. And push-pull drive, where the two arms move in opposite directions, halves the effective $V_\pi$ and cancels chirp; x-cut thin-film lithium niobate gets this from electrode geometry alone.

Measuring it. The clean method is to sweep the bias slowly, record the full transfer curve, and fit the cosine rather than eyeballing peak-to-null distance; the fit also returns the extinction ratio and the bias point. The procedure, including the Bessel-null method for phase modulators and the swept-frequency $V_\pi(f)$ measurement, is in Measuring modulator Vπ and EO bandwidth, and the Vπ Fitter will fit a pasted sweep directly.

References: Wooten et al., IEEE J. Sel. Top. Quantum Electron. 6, 69 (2000); Wang et al., Nature 562, 101 (2018); Chrostowski & Hochberg, Silicon Photonics Design (2015).