My BS detector went off when I saw that graph, so I had to actually read the paper... and now I understand the graph and it's not at all what it seems to be.
See that giant circle for silver? That doesn't mean "a massive amount of silver", or even "a massive amount of environmental impact from silver". These circles are sized proportionally to how much more of a resource is needed than in the current generation mix, if all power came from that source. Since almost no silver is used in the current generation mix, anything that actually uses silver - even if a rather small amount - will dramatically inflate its circle.
In all of the study's graphs, the amount of silver used is so small that the bars don't even register one pixel tall, so you can't really estimate how much they're talking about. The only graph where you see any height at all, Fig 8, is the graph relative to total world production - but even in their highest scenario it's less than 0.5% of world production. And the study itself notes, "Silver in PV cells might be replaced by other metals". Silver has only a 6% better conductivity than copper, it's not a big difference. And if you're willing to use slightly thicker or more frequent connects, you can use cheap and hyper-abundant aluminum. Either way, the amount of metal involved is practically irrelevant, the interconnects we're talking about here are practically microscopic.