Our current recommendation is to use DI water that has been titrated to a neutral pH (using a bit of sodium hydroxide for example – see FAQ at https://www.ophiropt.com/laser--measurement/knowledge-center/faq/7805 The commercial additive Optishield Plus is also recommended for systems such as ours that have copper and aluminum in them. It has the additional benefit of including a biocide to prevent the buildup of bacteria and algae.
Historically, corrosion in our sensors manifests as pitting corrosion in the anodized aluminum. There are a number of factors that contribute to pitting corrosion including dissolved oxygen, chloride ions, more "noble"/cathodic metallic ions in solution, and excessive pH values (less than 4 or higher than 9). This is why we recommend neutral DI water, which maintains a healthy pH and removes chloride and metallic ions.
While it is commonly held that DI water is especially corrosive, we have not encountered many references with careful measurements of this. In one online study on the corrosion of copper in DI water (https://accelconf.web.cern.ch/accelconf/e88/PDF/EPAC1988_1067.PDF), the rate of corrosion is predominantly determined by the amount of dissolved oxygen and carbon dioxide. While DI water may be "hungry" to replace its lost ions, the absence of other metallic ions has little effect on the amount of copper entering solution (see "diverse ion effect"). In a closed system, only a small amount of copper and aluminum would potentially dissolve to replace what was removed by the deionization process. At that point, the water should be no "hungrier" than the original tap water, while still lacking the other ions that facilitate corrosion. Thus, it is not expected that DI water will accelerate corrosion, but rather just the opposite.
The copper ions that do dissolve are probably the main source of corrosion of the aluminum housing, since copper is more cathodic than aluminum. Copper ions will plate out, taking electrons from the aluminum and sending aluminum ions to replace them in solution. This exchange will probably occur at a crack in the oxide layer covering the aluminum. Since only a small fraction of the aluminum surface is thus exposed, the corrosion will be concentrated in small regions leading to pitting.
However, our experience suggests that this process is not very fast. The vast majority of our water-cooled sensors do not show problems with corrosion. If the inherent electron exchange between copper and aluminum was so pronounced, we would expect corrosion in all of these sensors. Furthermore, in studying this problem in house, we attempted to reproduce corrosion in a controlled manner, using both tap water and DI water and running continuously for over a month at elevated temperatures (to accelerate the chemical processes). No corrosion occurred.
Thus, it is likely that individual cases of corrosion are due to impurities in the particular water used that accelerate the corrosion process. Since tap water quality varies greatly and cannot be easily tied to specific cases of corrosion, another reason to use DI water is to have the system running under known starting conditions.
Lastly, as standing water left in a sensor evaporates, ions and impurities become concentrated which may accelerate corrosion locally. Thus, we recommend blowing remaining cooling water out of the sensor with compressed air or nitrogen after use.