I have been asked to submit a list of my N most important papers for a certain nomination. In this case, N is five. In the past, for various reasons, N was often three, including for my tenure package years ago.
It’s always hard to chose; what does it even mean — “most important papers”?
There are highly cited papers that have an experimentalist as the lead senior author, so even though my student/postdoc and I did a lot of theoretical work and wrote large sections of those papers, I don’t consider them my papers, but the experimental group’s papers.
There are highly cited papers that have me as the lead senior author, but many of those I honestly think are boring. People read them and cite them because we did what needed to be done at the time when it needed to be done, we did it well, and we wrote it up clearly and compellingly. But I never thought that they were particularly technically exciting, at least to me. Maybe part of what makes them well cited is that they are appreciated even by the people who are not as enamored of theory and math as I am.
Then there are the papers that are not yet highly cited as they were only published 1-2 years ago, but I think will eventually get there, because the work is cool and their citation rate (number of citations per unit time) is pretty high.
Then there are those that often aren’t highly cited, but that I think were really exciting to work on or that really broke new ground. They may or may not get citation traction, because some of them are really complicated, but I think they are important, and sometimes they do start to pick up citations after a few years delay.
Why am I even hung up on citations? Well, they are a metric for how much we influenced the field. Still, I don’t really want to list the highly cited papers that I don’t think are technically beautiful.
I have to remark that, post tenure, I definitely have a much higher proportion of exciting, technically challenging work to the obvious-next-step-that-we-can-do-faster-and-better-than-others papers than what I did on the tenure track. And that’s a major perk of tenure — having the security to work on the harder, higher-risk stuff.
What say you, blogosphere? How would you pick your 3–5 most important papers?
xykademiqz 
Curious to see what other people have figured out for this. I’ve had to do this for some faculty apps at the junior level, too, and don’t have quite as much citation data to guide me.
I’ve ended up with this set:
1) Clever, but technically quite simple paper that sets up some of my future research
2) Graduate paper that I think hits my intersection of “papers I like” and “reasonably useful to a lot of people in the field,” even though it’s definitely not the most cited or fanciest journal.
3) Paper in fancy-pants journal that got there because it was the first, and shows off some more intense computation, but doesn’t actually have the most compelling scientific results.
I’d stick to mostly papers that feature your group. I think others will also think that papers with experimentalists as lead authors are experimental papers, regardless of how significant the theory contribution was. It is an unfortunate truth that most fields are biased towards experimentalists, and consider them the drivers, regardless of how the situation actually is.
For me, I use a highly cited paper in a fancy journal that is getting on in age, but remains a really nice publication that is a foundation for my current work. I then use 2-3 newer papers with good citation rates that I think are exciting. If I need a fifth, I use another paper that is foundational for my current work with good citations, but that I find boring now.
In your situation, I’d pick these 5:
1 highly cited paper with you as senior lead author (even if you think it is boring) that is very clear and in a well-known journal.
1 paper you think is ground-breaking and exciting, even if not highly cited
2-3 newer papers with high citation rates that you really like
and maybe 1 highly cited paper with an experimental group (especially if it is in a “name” journal), since I think people expect theory people to have that sort of paper.
The key is, you need to select “your most important papers” using the criteria of the people you believe will be reading and evaluating your selections. It isn’t about the absolute value of your papers at all, or about the future worth of your recent papers.
Thus, you should only select papers from your group, and only those published since you have been publishing independently. In addition, you should select older papers where the test of time has shown they are important to others in the field (e.g. Those papers with larger numbers of citations).
In most cases, it is important that you are selecting a set, that makes sense as a set for the purpose involved; it is not just “calculate an importantness metric, and then select the papers with highest value for importantness”. In grant applications, relevance to the work proposed is a factor, so I may well submit different sets of best papers, for grants being submitted at the same time! But there is also a need for spread: sometimes I try to ensure the set includes a paper with a great idea, and a paper that shows really careful technical mastery; a paper that shows I can work with big-name collaborators, and a paper where I worked alone; a paper that shows I was early in an agenda that had lots of influence, and a paper showing my latest direction. Of course, one paper can cover multiple of these aspects.