Not much research has been done on this question in recent years, but some researchers are worried enough to research into wooden satellites.
The question on the environmental impact of deorbiting satellites burning up in the upper atmosphere was partly addressed in a 1994 report (warning: not peer reviewed) by the Environmental Management of the Space & Missile Systems Command in the United States. Their focus was to consider the impact of deorbiting space debris on ozone at the time, and their conclusion was that deorbiting space debris has very little impact on stratospheric ozone. They considered two types of impact using a combination of lab and model measurements:
Heterogeneous mechanisms, or small particles (in their case Al₂O₃) on which ozone depletion can occur (such as in polar stratospheric clouds). The report cites experiments by Marino Molina (MIT) from which they conclude it takes 10⁴ – 10⁵ years to destroy one percent of stratospheric ozone. Although small, it would seem that a major increase of re-entry mass should motivate revisiting this study; if their figures are accurate and numbers are linear, a factor 1000 increase would translate to 10–100 years to destroy 1% of stratospheric ozone, close enough to warrant at least some worry, although most likely still in the safe range (there is always a natural amount of ozone destruction and regeneration).
Homogeneous mechanisms: spacecraft paint and the Zeldovich mechanism produce nitric oxide, but the study estimates that the paint causes the destruction of one ozone molecule per billion days and the Zeldovich mechanism even less, so if their conclusions are correct, this mechanism is negligible.
There may be other impacts than ozone depletion, but ozone is probably the most sensitive substance that may suffer from the impact of orbital debris reentry for us to worry about in practice.
The impact is probably still minor, but to actually answer this question, you need to consider:
- Where does the satellite break up?
- What pollutants are released in the process, and how much of each?
- What is the lifetime of those pollutants?
- What is the ultimate fate of these pollutants after their lifetime?
A complete answer would take an in-depth study. We've made great progress in the recovery of the ozone hole and let's not be taken by surprise again; few if any people expected fridges to cause skin cancer, after all. To address the questions a bit:
Comparing with the total mass of the atmosphere is not useful. The total atmospheric mass is not relevant because satellites break up in the upper atmosphere, which is very thin compared to the rest. The stratosphere does not exchange much mass with the lower atmosphere, so substances can stay in the stratosphere very long (unless destroyed in chemical reactions or heavy enough to fall down due to gravity).
You don't need much mass to have a high impact. CFCs have concentrations in the parts per billion range, but with their lifetime of decades can and do break down large amounts of ozone. So we can't simply dismiss the problem based on atmospheric mass considerations alone.
The materials from which satellites are built are different from propellants involved in launch. Therefore, you can't simply dismiss satellites as far less massive than rocket propellants and therefore declare the impact of re-entry negligible compared to the the impact of launch. Much of a satellite material is metal, which will deposit rather quickly (see next question), but other materials could in theory have an impact (John et al. considered paint and concluded its impact was negligible). Some satellites contain unusual materials: for example, Kosmos 1402 was a Soviet spy satellite containing a nuclear reactor and thus fuel. Leifer et al (1987) have shown that more than a year after Cosmos-1402 deorbited a 53±20% excess of 235U was measured at an elevation of 36 km. I don't know what happened since then. Fortunately Starlink will not contain nuclear reactors.
Lifetime, often defined using the half-life or the time it takes for the concentration to halve, is crucial to determine impact. A piece of metal that falls down has no impact on the atmosphere, but exclusively anthropogenic CFCs lingering around for decades may, even in relatively small concentrations. I don't know if anyone has estimated the lifetime of the 235U from Kosmos, but Murphy et al. (2018) may have detected evidence of it in the upper troposphere (they found one particle and could not tell the source). Due to its low concentration this is rather of academic interest than something to really worry about.
There are only two ways a substance can leave the upper atmosphere: by physically leaving (to the troposphere) or by destruction (chemical reaction). If it reaches the troposphere it will have a very low concentration compared to pollutants originating from the surface, and if it reacts we go back to question 2. Larger particles may fall down quickly, but molecules can stay around for a while. The John et al. found that mostly the small particles were enhancing ozone depletion.
In conclusion: the impact is probably small, but the potential for ozone depletion if re-entry flux is increased by several orders of magnitude is probably sufficient to warrant a dedicated research project to quantify this again. And any detectable impact is an impact, which is worth monitoring even if the impact is well within safe limits. Someone go write a research grant proposal ;)
Relevant papers I found:
- Leifer et al., Detection of Uranium from Cosmos-1402 in the Stratosphere, Science, 23 Oct 2017, doi: 10.1126/science.238.4826.512
- Murphy et al., An aerosol particle containing enriched uranium encountered in the remote upper troposphere, Journal of Environmental Radioactivity, Volume 184-185, April 2018, Pages 95-100.
- John et al., The Impact of Deorbiting Space Debris on Stratospheric Ozone, 1994 (see online PDF).