Leak detection and localization are critical manufacturing quality-control processes. Many industrial and domestic machines use or convey pressurized gases or liquids. Unintended leaks from machine components may be detrimental to consumers, manufacturers, and the environment. This paper describes a leak detection technique based on photoacoustic sounds produced by the interaction of a carbon dioxide (CO2) laser tuned to 10.6 micrometers and a photoactive tracer gas, sulfur hexaflouride (SF6), emitted by calibrated leak sources. Acoustic signals generated by a high-speed scan of the laser beam through the cloud of tracer gas formed near the leak are recorded in a bandwidth from 3 to 52 kHz by multiple microphones. From the recorded signals, the presence or absence of a leak may be deduced by comparison with the background noise level at the signal frequencies, which occur at the harmonics of the scan rate. When a leak is present, its location is determined from a simple model of the acoustic environment and matched field processing (MFP). Current results show that a gas leak of 1 cm3 per day can be detected and localized to within ±3 mm in a few seconds using four microphones, placed 0.41 m from the leak location, and an incoherent average of the MFP ambiguity surfaces at eight signal frequencies. Comparisons of the Bartlett and minimum-variance-distortionless matched field processors are also presented. © 1999 Acoustical Society of America.