Two lemmas first. (A) The continuous image of a compact set is compact: if K is compact and f continuous, an open cover {Ui} of f(K) pulls back to an open cover {f−1(Ui)} of K, which has a finite subcover whose images cover f(K).
(B) A compact subset of a Hausdorff space is closed: if K⊆Y is compact and y∈/K, each x∈K has disjoint neighborhoods Ux∋x, Vx∋y (Hausdorff); finitely many Uxi cover K, so ⋂Vxi is a neighborhood of y missing K.
Netto's theorem: a continuous bijection from a compact space to a Hausdorff space is a homeomorphism. Indeed, if F⊆K is closed then F is compact — a closed subset of a compact set is compact (for K=[0,1] this is Heine–Borel) — so f(F) is compact by (A) and closed in Y by (B) — i.e. f is a closed map.
A closed bijection is a homeomorphism ((f−1)−1(F)=f(F) is closed, so f−1 is continuous).
If f:[0,1]→[0,1]2 were a continuous bijection, it would be a homeomorphism. Removing an interior point c∈(0,1) disconnects the interval: [0,1]∖{c}=[0,c)∪(c,1] is a separation into disjoint nonempty opens.
In contrast, [0,1]2 minus the corresponding point r:=f(c) is path-connected, shown here with explicit formulas.
Take two points p,q=r. If p=q, the constant path γ(t)≡p works. If p=q, look at the line L(t)=(1−t)p+tq (t∈[0,1]). If r∈/L([0,1]), then γ=L is the desired path.
If r∈L([0,1]), there is a unique t0∈(0,1) with r=L(t0) (since r=p,q). Set the midpoint m=2p+q, direction d=q−p, unit normal n=∣d∣(−d2,d1), and pick the detour point s=m+εn (ε=0, chosen explicitly below). The path is two segments joined: γ(t)=(1−2t)p+2ts (0≤t≤21), γ(t)=(2−2t)s+(2t−1)q (21≤t≤1).
Check that r is avoided. Points x on the line pq satisfy n⋅(x−m)=0 (since n⊥d), and r lies on it, so n⋅(r−m)=0. For a point (1−u)p+us=(1−2u)p+2uq+uεn on [p,s], n⋅(x−m)=uε; equaling r forces uε=0, i.e. u=0 — only the point p=r.
For a point (1−u)s+uq=21−up+21+uq+(1−u)εn on [s,q], n⋅(y−m)=(1−u)ε, so (1−u)ε=0 gives u=1 — only the point q=r. Hence γ never passes through r.
Explicit choice of ε (forcing s∈[0,1]2): m cannot be a corner (a corner m would give p=q=m, contradicting p=q). Put δ=min(m1,1−m1,m2,1−m2)≥0. If δ>0 (m interior), take ε=δ/2: ∣s−m∣=δ/2<δ, so s is interior.
If δ=0, then m lies on an edge but is not a corner. For instance with m=(m1,0), m1∈(0,1) (the other three edges are identical via symmetries like (x,y)↦(x,1−y)), put η=21min(m1,1−m1)>0 and take s=m+ηn if n2≥0, s=m−ηn if n2<0. Then s2=η∣n2∣∈[0,η]⊂[0,1] and ∣s1−m1∣≤η≤21min(m1,1−m1), so s1∈(0,1) — i.e. s∈[0,1]2. With s∈[0,1]2, convexity keeps both segments [p,s], [s,q] inside [0,1]2.
Thus any two points of [0,1]2∖{r} are joined by an explicit path avoiding r — path-connected, hence connected. A homeomorphism restricts to a homeomorphism of the punctured spaces — contradiction. So no continuous bijection exists, and the Peano surjection cannot be injective.