Idea. The level det=1 is an actual "edge" of the set, because scaling by a factor t<1 moves the determinant strictly down.
Let E={det≥1} and O={det>1}.
Step 1: E is closed, O is open. Preimages of [1,∞) and (1,∞) under the continuous map det.
Step 2: O⊂intE. O is open and O⊂E.
Step 3: intE⊂O. Let A∈E∖O, i.e. detA=1. For t∈(0,1), det(tA)=tndetA=tn<1, so tA∈/E; and ∥tA−A∥=(1−t)∥A∥→0 as t→1−. Thus every neighbourhood of A contains a point outside E, and A∈/intE. Hence intE=O.
Step 4: the boundary. E is closed, so E=E and
∂E=E∖intE={det≥1}∖{det>1}={det=1}=SLn(R).
Remark.
- The same proof shows ∂{det≥c}={det=c} for every c>0. For c=0 one needs the more delicate SVD argument ({det>0}={det≥0}, problem above), since scaling does not move det=0.
- For the set {det≥−1} and a point with detA=−1 the scaling must go the other way: det(tA)=−tn>−1 for t<1 stays inside, but for t>1 one gets −tn<−1, outside, so such A is a boundary point. Always check the sign.