Real Analysis Qualifying Exam Solutions

What You Actually Need to Know Before Opening a Practice Problem

Most people approaching their real analysis qualifying exam will spend weeks drilling measure theory problems without ever realizing that the exam is testing something slightly different than what they studied. The qualifying exam doesn't want you to recite theorems. It wants to see whether you can construct a clean proof under pressure when you've never seen the exact problem before. That distinction matters more than you probably think right now. I remember sitting in a practice session where the problem asked me to construct a measurable function whose derivative exists almost everywhere but is not Lebesgue integrable. The standard Cantor function came to mind immediately, but the derivative of the Cantor function is zero almost everywhere, which means it's trivially integrable. That was the trap. The actual solution required modifying the construction — using a variant where the derivative blows up on the complement of the Cantor set in a controlled way, producing something like 1/x on certain intervals removed during the construction. I spent about forty minutes going down the wrong path before I stopped and re-read the question carefully. That minutes could have been fifteen if I had recognized the pattern earlier.

Where to Find Real Analysis Qualifying Exam Solutions

The best collection of practice problems and worked solutions lives scattered across university mathematics department websites. Berkeley, UCLA, University of Chicago, and UT Austin all archive their past qualifying exams with full solutions. MIT OpenCourseWare has some materials, though the coverage is uneven. The Math Stack Exchange community also maintains a running thread collection tagged with qualifying exam questions, but those solutions vary wildly in quality. Some are correct and elegant. Others contain subtle gaps that look fine at first glance and will cost you points on your actual exam if you try to copy that reasoning. I typically start with the UChicago archive because their exams emphasize proof writing over computation, which matches most program expectations. The Berkeley exams are useful for seeing how measure-theoretic arguments are structured at a higher level. I cross-reference solutions between schools because different graders approach the same problem differently, and seeing two valid proof strategies for one question usually gives you more flexibility than seeing the same approach repeated four times. There are commercially available solution manuals out there, but I wouldn't recommend relying on them. The ones I've seen are often transcriptions from students who passed once, meaning the solutions reflect what that person could produce under exam conditions, not what a clean proof actually looks like. You'll pick up bad habits faster from a commercial manual than from a raw exam PDF with no solutions at all.

The Three Topics That Actually Show Up

Real analysis qualifying exams concentrate on three areas with heavy overlap: measure theory and Lebesgue integration, metric and topological spaces, and basic functional analysis including Banach and Hilbert space theory. Don't be fooled into thinking topology questions will be easy because they're usually paired with measure theory in ways that require both tools simultaneously. One counter-intuitive thing about these exams that most students don't figure out until after they take one: compactness arguments are almost always the intended path, even when the problem is stated purely in terms of integration. I saw this repeatedly. A question about uniform integrability will often resolve cleanly with a compactness argument on the underlying measure space rather than a direct epsilon-delta estimate. If you're pushing through with brute force estimation and your solution is running eight paragraphs long, you've probably missed the structural insight the question is looking for. A proper approach using compactness or subsequence extraction will typically come in under four lines. Another thing that catches people off guard: the relationship between pointwise convergence and convergence in measure. Students memorize that L^p convergence implies convergence in measure, which implies a.e. convergence along a subsequence. But the exam loves to ask whether convergence in measure plus a domination condition implies L^1 convergence, and the answer requires citing the generalized dominated convergence theorem rather than the standard version most textbooks present first. The standard version assumes almost everywhere convergence of the dominating sequence. The generalized version relaxes that to convergence in measure. Getting this distinction wrong during an exam costs you the entire problem, and it's a very common mistake because introductory courses rarely emphasize the generalized form.

How to Actually Use Past Exams

Open a past exam. Set a timer for three hours. Work it under actual conditions — no notes, no breaks, no checking solutions mid-problem. This is non-negotiable. Reading a solution and understanding it is not the same as producing one. The exam tests production under time pressure, and if you only study by reading solutions you will significantly underestimate how much time each problem actually takes to write cleanly. After completing the timed run, compare your solutions against the posted answers. Pay attention to three things: where your proof has unnecessary case splits, where you used a heavier theorem than required, and where you left gaps that a strict grader would mark down. Most students skip this analysis step and just check whether they got the right answer. That's insufficient. The path matters more than the result on a qualifying exam. I found that working through five full past exams this way — roughly three weeks of study time — was more effective than doing twenty standalone problems from a textbook. The exam format itself is a skill, and practicing the format directly builds it faster than any other method I've tried.

What These Exams Won't Test (And What That Means)

Real analysis qualifying exams rarely test complex analysis beyond what's needed for basic function space arguments. You won't see contour integration or residue calculus unless your program happens to combine the two exams. Similarly, probability theory concepts appear only when they intersect with measure theory directly. Don't waste time studying martingales or stochastic processes unless your specific program's past exams demonstrate that it does.The main limitation of relying on past exam solutions is that programs occasionally recycle problem structures with different numerical constants or modified hypotheses. A solution you memorize from three years ago might not apply directly to this year's version if the question writer changed the space from L^2 to L^p for p not equal to 2. The fix is to understand the proof technique behind the solution, not the solution itself. When you can reconstruct a proof from its key lemmas rather than recalling it verbatim, recycling questions stop being a problem. Another limitation worth noting: solution quality on free online sources is inconsistent. A solution posted by a graduate student who barely passed won't teach you anything about elegant proof structure. I always verify solutions from unofficial sources against two or three other references before trusting them. The verification step adds time but prevents you from internalizing incorrect reasoning, which is harder to unlearn than not knowing the material at all.