As HPS lamps age, their characteristics change. The lamp voltage rises and the discharge tube gets darker and it eventually cycles at the end of its useful life. This is all fine and dandy, but I have questions about why exactly this happens:

---

1) Sodium Loss:
According to Wikipedia (reliable, I know), the aluminum oxide discharge tube and the sodium metal reacts over time to become sodium oxide and aluminum metal. This obviously makes sense because sodium is probably a lot more reactive than aluminum. But, since the discharge tube is saturated with sodium, isn't there an excess of sodium? Does the arc tube really eat all of that sodium up to the point that it becomes unsaturated?

2) Metallic Aluminum:
Since the alumina arc tube decomposes into metallic aluminum, what happens to it? Does it ever vaporize and go into the arc stream? I know that it has to at least melt because the arc tube does get yellow-hot.

3) Voltage Rise:
Why does this happen? Is it sodium loss? Is it electrode wear? Does anyone know?

---

Thanks so much!

One major contributor for the last one is rather simple: As the arctube becomes darker, it absorbs more of the otherwise radiated power and runs hotter. Higher temperature then means higher vapor pressure and that means higher arc voltage drop.
Yes, there are design features to suppress this effec (detaching the amalgam reservoir from the arctube around the arc,...),but that only reduces the effect, it won't eliminate it completely.
In fact the effect is to some extend needed: Aging lamp has lower efficacy, so a mechanism operating it at higher power compensates for it, yielding more stable lumen output over lifetime. And that means the system does not be designed with so much lumen output margin as others, so has lower power consumption (because of less lamps needed) when the lamps are fresh, yet keeps the illumination even as the lamps age (because the elevated run temperature mean higher voltage so somewhat higher power for the lamp). My guess will be, the lamps would be really engineered to match the efficacy loss by the elevated power on the gear the particular lamps are designed for (CWA in the "120V" market, series choke in the "230V" market) over the rated lifetime.

The sodium loss is an issue only for unsaturated vapor lamps, the normal saturated vapor types have so much excess of it, the sodium loss won't ever be a problem. Of course, I do not mean cases of failed seals or arctube, when everything leaks out.

So most of the question comes back to why the tube darkens over time. There one reason wouod be the same as with any discharge using electrodes: Electrode sputtering and evaporation coating the electrode material onto surrounding structures, so onto the arctube wall, turning it dark. In a HPS is no cleaning mechanism like halogen cycle (in MH's) or so, so the only thing helping somehow reducing the effect to lumen depression is the arctube being thin, so the sputtering/evaporation concentrates the darkening mainly just around the electrodes, but over time the whole arctube accumulates some tungsten coat too.

@Medved
Thanks so much, that does make sense. I never thought about how darkening walls would increase pressure. Why does this not happen to MV lamps? I know they can get very dark at the end of their lives, but I have never heard of any voltage increase associated with that.

It is linked to the fact the (standard) HPS are saturated vapor lamps, so only part of the fill is vaporized, the vapor pressure (or better to say density) is strongly linked to the temperature of the liquid fill amalgam reservoir - the higher the temperature, the more of the fill evaporates, so the more of the fill material is within the arctube, hence the higher density and pressure.

MV are of unsaturated vapor lamps, so always (at operating temperatures) the complete fillis evaporated,so the fill density is constant, regardless what the temperature is doing. So the pressure then just follows gas laws,so proportional to just direct absolute temperature, so very small pressure change with large temperature change (compare to the radiated heat power by the arctube is proportional to T^4, so way steeper, keeping the temperature rather constant).

Same saturated vapor effects are affecting MH, there the problem is even worse because there are multiple components involved and each of them behaves differently, so the fillcomposition, so relative strengths of radiation from each component are changing with the arctube temperature, consequently the radiated color is shifting.

@Medved
Wow, I feel like I am really approaching enlightenment about lighting knowledge. I wonder why they don't make MV or MH lamps saturated, but whatever. That clears that up.

it should be noted that Sodium loss does play a big roll in the aging of HPS lamps, and is the primary life-limiting factor of a HPS lamp


remember that in a HPS lamp its a sodium-mercury amalgam,  and there is a golden ratio so to speak of sodium to mercury, and as sodium is lost during the life time of the lamp, more and more mercury is vaporised, which causes the arc voltage to go up, until eventually the arc voltage becomes too high for the ballast to sustain and the arc extinguishes, (until it cools down the lamp restrikes, and so starts the cycle/cycling)


the effects that Medved mention do play a roll, but they are secondary to the main mechanism of sodium loss throughout life


this sodium loss is also why HPS lamps towards/at end of life often turn reddish in colour as the higher mercury/sodium ratio shifts the colour of the lamp

One major contributor for the last one is rather simple: As the arctube becomes darker, it absorbs more of the otherwise radiated power and runs hotter. Higher temperature then means higher vapor pressure and that means higher arc voltage drop.
Yes, there are design features to suppress this effec (detaching the amalgam reservoir from the arctube around the arc,...),but that only reduces the effect, it won't eliminate it completely.
In fact the effect is to some extend needed: Aging lamp has lower efficacy, so a mechanism operating it at higher power compensates for it, yielding more stable lumen output over lifetime. And that means the system does not be designed with so much lumen output margin as others, so has lower power consumption (because of less lamps needed) when the lamps are fresh, yet keeps the illumination even as the lamps age (because the elevated run temperature mean higher voltage so somewhat higher power for the lamp). My guess will be, the lamps would be really engineered to match the efficacy loss by the elevated power on the gear the particular lamps are designed for (CWA in the "120V" market, series choke in the "230V" market) over the rated lifetime.

The sodium loss is an issue only for unsaturated vapor lamps, the normal saturated vapor types have so much excess of it, the sodium loss won't ever be a problem. Of course, I do not mean cases of failed seals or arctube, when everything leaks out.

So most of the question comes back to why the tube darkens over time. There one reason wouod be the same as with any discharge using electrodes: Electrode sputtering and evaporation coating the electrode material onto surrounding structures, so onto the arctube wall, turning it dark. In a HPS is no cleaning mechanism like halogen cycle (in MH's) or so, so the only thing helping somehow reducing the effect to lumen depression is the arctube being thin, so the sputtering/evaporation concentrates the darkening mainly just around the electrodes, but over time the whole arctube accumulates some tungsten coat too.

Wow, I'm starting to like discharge tube bulbs more. This information cleared up some questions in my mind. 

The above comments are correct, but it should also be noted that the sodium does not actually react with the aluminium oxide arc tube to any great extent.  Wikipedia is completely wrong in this respect.  The sodium does react slightly with some dopants in the crystalline structure of the alumina, notably the magnesium oxide and calcium oxide components which are present at the grain boundaries.  This gradually makes the arc tube wall more porous during life.  Sodium atoms are then able to diffuse along the grain boundaries and leak into the outer bulb, where they react with the glass bulb and cause additional darkening which reduces luminous flux and traps more heat inside the lamp, further exacerbating the voltage rise caused by arc tube blackening.

There is also extensive reaction between sodium and the glassy frit-seals at the arc tube ends, and with the barium tungstate emitter coating of the electrodes.  In that case the sodium does not leave the arc tube but becomes chemically bound with those components.

As Dez mentioned, the primary ageing mechanism of standard HPS lamps is the changing amalgam ratio.  HPS lamps reach peak efficiency and luminous flux when the delta-lambda distance between the two spectral peaks either side of the sodium resonance radiation is about 120 Angstroms.  But as the sodium-mercury ratio increases and the arc tube temperature rises due to blackening, the red-wing of the sodium spectrum is broadened and the d-line is also broadened.  More red and infrared radiation is produced by the plasma, so the luminous flux must decrease.

Note that these failure mechanisms are entirely different in the higher performance Unsaturated Vapour HPS lamps.  These are actually characterised by a falling lamp voltage during life, since there is no excess of sodium and the whole dose is vaporised.  Therefore no longer so dependent on the cold spot temperature and its changes during life as a result of blackening.  USV lamps are only feasible when the sodium-sinks are reduced or eliminated.  So the alumina arc tubes tend to be doped with zirconium or erbium oxides in addition to magnesium oxide, and with reduced calcium oxide impurity, which reduces sodium reactions at the grain boundaries and hence the rate of sodium loss.  Of far more importance though is the change of electrode emitter material, usually one of the biggest sodium sinks.  The traditional barium tungstate emitter is changed to something less reactive - in the case of the Sylvania lamps that was BSY2, barium strontium yttrate combined with a special sintering procedure.  Some companies also used more resistant frit sealing glasses.  Incidentally just before the USV lamps fail and the last of their sodium is consumed, there is a sudden rapid rise in voltage back up to something close to the original level.  This characteristic ensures that there is no end-of-life cycling.  Lamps turn completely blue when the sodium is gone, and continue to burn as a pure mercury discharge.  But then of course the plasma temperature increases, and since the ceramic arc tube is not chemically stable enough to withstand a pure mercury arc (sodium or metal halides are required to protect it), the ceramic eventually disintegrates and leads to complete lamp failure. 

If you want to learn more about this I can highly reccomend the book of one of my former colleagues, Sjef de Groot at Philips Eindhoven, who wrote "The High Pressure Sodium Lamp".  There was a copy going cheaply on Ebay for a long time that seems not to have sold but now I cannot find it.  That book is rather old and does not cover the newer developments like the USV, deluxe/white, retrofit, high xenon pressure, mercury-free lamps etc.  The subject was brought just about fully up to date in a comprehensive IEEE paper written in 1993 by my old boss at Sylvania, Rudy Geens, and our American colleague Elliot Wyner.  See https://digital-library.theiet.org/doi/abs/10.1049/ip-a-3.1993.0070  I can send a copy if you do not manage to find it via the usual scientific literature sources.

@James
Wow, as always, that was a lot of good information. Thank you so much.

@James Do you have more information about USV lamps?