(Disclaimer: I work for a solar company, but the following is my personal opinion.)
The simple, straightforward is yes – by far. No mass-produced anything is pollution free, but solar emits much, much less lifetime pollution than fossil fuels no matter how you look at it. Solar emits less pollution, isn’t enough by itself, but is a so-far essential part of a cleaner energy grid. Some regions are lucky enough to have geothermal or wind resources, both of which can produce more and cleaner energy than solar, but the rest of us aren’t so lucky and for us solar remains one of the cleanest energy sources. Note that “cleanest” does not mean “pollution-free”, it means “much less evil than fossil fuels, less habitat-raping than large-scale hydropower, less geography-dependent than wind or microhydro, and much less controversial/less long-term waste than nuclear”.
lazybratsche’s linked lifecycle meta-analysis covers a lot of studies, aggregated and summarized by NREL (a renewables research lab owned by the US Department of Energy). These sort of studies, also known as “cradle to grave analyses”, attempt to quantify accumulated pollution over a product’s entire lifecycle by normalizing different pollutants into carbon dioxide-equivalent units (in terms of its potential impact on climate).
That normalization is done because there isn’t one magic number that can encompass all of a product’s potential impacts on the environment and human safety: something can have minimal impact on the climate but be incredibly toxic (yet only produces minute quantities of the toxic waste), or kills a lot of frogs and nothing else, or completely decimates a local waterway but then operates cleanly for the remaining 25 years where the final product is installed, etc. For that kind of data, you need to look at the individual studies and sometimes they will also cover toxicology from pollutants that don’t have direct impacts on climate.
That NeoGeo article CarnalK linked to is a great start, and the Solar Scorecard mentioned in the article is also a good read. Sadly, as a still-young industry with many players outside of the reach of Western environmental legislation and watchdogs, “self-policing” and self-reporting is pretty much the norm… and the history of industry has taught us that rarely works. But as far as anyone can tell so far, even panels produced by the dirtiest factories in China are still much cleaner, in terms of pollution per kWh of energy generated over its lifetime, than fossil fuels.
In terms of energy return on energy invested (EROEI), over its lifetime, a solar panel will generate about 7x the amount of energy it took to be produced. This is already higher than, say, corn ethanol. Panel efficiency (measured as % of incoming solar energy converted into usable electricity) is always increasing, and in the lab the efficiencies are already more than 2x what they are in the consumer sphere, but we are already at less than $1/kW for the panels themselves. This means that the price of the panels themselves are already low, and only one portion of the overall cost of an installation – the rest is due to inverters, racking/mounting, (optional) batteries, and (optional) professional installation. So further improving panel efficiency will only have limited impacts on the end price of a complete system; for that to go down much more, economies of scale in inverter production, installation methodology, etc. will have to be improved.
(Not a sales pitch, but sales-related) The 30% federal tax break, coupled with any potential local incentives, is the only thing on the immediate horizon that could alter prices that drastically for most people. For the portion of solar users who need or want batteries, Tesla’s mass-production of their Powerwall units could have big impacts on the market – but most people don’t really need batteries with their solar. So if she wants to wait for the roof to be redone, just do it before the tax credit expires at the end of 2016 (unless it gets extended).
Call it 662 gallons of gasoline then. Still sounds like a good trade.