1. Département de Physique and RQMP, Université de Sherbrooke, Sherbrooke, Québec J1K 2R1, Canada


2. Department of Physics and Astronomy, University of British Columbia, Vancouver, British Columbia V6T 1Z1, Canada


3. Canadian Institute for Advanced Research, Toronto, Ontario M5G 1Z8, Canada


4. Present address: Max Planck Institute for Chemical Physics of Solids, 01187 Dresden, Germany.

Correspondence to: Louis Taillefer^1,3 Correspondence and requests for materials should be addressed to L.T. (Email: louis.taillefer@physique.usherbrooke.ca).

The nature of the pseudogap phase is a central problem in the effort to understand the high-transition-temperature (high-T _c) copper oxide superconductors¹. A fundamental question is what symmetries are broken when the pseudogap phase sets in, which occurs when the temperature decreases below a value T*. There is evidence from measurements of both polarized neutron diffraction^{2, 3} and the polar Kerr effect⁴ that time-reversal symmetry is broken, but at temperatures that differ significantly from one another. Broken rotational symmetry was detected from both resistivity measurements⁵ and inelastic neutron scattering^{6, 7, 8} at low doping, and from scanning tunnelling spectroscopy^{9, 10} at low temperature, but showed no clear relation to T*. Here we report the observation of a large in-plane anisotropy of the Nernst effect in YBa₂Cu₃O_y that sets in precisely at T* throughout the doping phase diagram. We show that the CuO chains of the orthorhombic lattice are not responsible for this anisotropy, which is therefore an intrinsic property of the CuO₂ planes. We conclude that the pseudogap phase is an electronic state that strongly breaks four-fold rotational symmetry. This narrows the range of possible states considerably, pointing to stripe or nematic order^{11, 12}.