Image above: Spectrum of a high pressure sodium street lamp taken last night by Ebert. The spectrograph from Spectral Workbench is below the spectral image. X-axis in nm.
I was attracting some stares last night in the school parking lot as I stood under a street lamp, intent on the LCD of Ebert’s camera as its barrel pointed straight up toward the lamp. This was a sodium lamp, distinguished by its yellowish glow, and if Ebert could form a spectrum of the lamp some 20 feet above me, I expected to see the famous double emission lines of sodium. But the spectrum Ebert formed was exactly the opposite of that, and surprising in other ways as well.
High pressure sodium lamps produce a yellowish light because most of their emission is in two narrow peaks at 589 and 590 nm. But sodium atoms in the lamp also absorb light of these wavelengths, so almost no light of these wavelengths escapes the lamp. This results in the steep absorption valley in the spectrograph around 590 nm. On either side of that steep valley is a broad smear of orange and red wavelengths that are emitted by the lamp. This is a result of the high pressure in the lamp. There are enough atoms packed in there that they are colliding with one another a lot. When a collision happens as an atom is emitting a photon, there is increased uncertainty about the wavelength of that photon. So photons that would otherwise be 589 or 590 nm are emitted at nearby wavelengths causing the broad smear of color. This "collisional broadening" might also explain the broad smears near the mercury lines in the discharge tubes of the “Cigars” sign, but that’s just a guess.
Most of the emission lines in the sodium lamp spectrum are narrow blue and green lines, and most of these are from other gases in the lamp. Mercury and xenon are often included in these lamps, but I am not sure what is responsible for the lines in this spectrum. I don’t know how accurate the calibration of this spectrum is since I don’t know what the wavelengths should be. It was calibrated in the standard way at Spectral Workbench by comparing it with a spectrum of a CFL bulb taken when I returned home from the parking lot. The peaks are probably labelled within a few nm of where they should be. The spectrum can be seen at Spectral Workbench.