Hardware that actively forces two frequency combs into phase is a feedback system, and a feedback system on a moving vehicle has to hold lock against a disturbance that never stops. Li's own statement of the problem is that dual-comb spectroscopy buys simultaneous, high-precision multi-gas measurement at the cost of sensitivity to environmental noise [2]. Generating the combs in fiber lasers built to tolerate vibration, and then arranging for the two sources to stay synchronized without hardware to enforce it [3], removes the subsystem most likely to drop out first when the road surface changes.
The sampling geometry is the other design choice worth reading closely. The mid-infrared light passes through a compact open-path cell with an effective 25-metre path, working on air drawn from around the instrument [4]. That buys sensitivity without a long tube or a distant retroreflector, and it also fixes what the measurement is: the air the vehicle moves through, not a line of sight toward a target. Earlier field attempts leaned on precisely aligned optics, fixed telescopes and mirrors among them, which is exactly what pinned them in place [5].
The hour of driving yields roughly 3,600 one-second samples, about one per 13 metres of route [1], and a mean speed of 47 km/h [3], so the expressway stretches were a minority of it. At the top speed the system supports, a single sample averages the air over some 28 metres of road [2]. And the quoted 66 ppb precision [10] sits against a background that averaged 1.815 ppm [11], so one second's reading needs an enhancement of about 3.6 percent over ambient before it means anything [4].
The figure of merit, 3.4 x 10^6 Hz, is reported as comparable to typical laboratory mid-infrared dual-comb systems [9], and that is the honest version of the claim: the spectrometer did not degrade when it started moving. It is a number for comparing instruments, not a leak specification. The write-up does not give the emission rates of the simulated releases, their distance from the vehicle, or how many passes were needed to place them [12], and those are the figures a gas utility would want before trading a walking survey for a drive-by.
One more gap between the framing and the data: the technique's advertised edge is many species at once [2], while the road results report two, methane and water vapor [10]. What the run does establish is narrower and still useful. A laboratory-class interferometer survived an hour of ordinary traffic with its precision intact [7]. The remaining difficulty is outdoors, in the air itself, since sources are scattered and move with the wind [6]. That makes the alert Li sketches, a concentration with GPS coordinates sent to a maintenance crew [13], the start of a localization problem rather than its answer.
