Throughout, for x∈I and y∈I with y=x write
Qx(y)=y−xf(y)−f(x),so that Qx(y)=Qy(x);
and x↦cx, once cx exists, is a map I→R given by an infimum (Routine Facts Used Without Citation: Arithmetic, Maps, Predicativity and Indices (iv)), so the set in (v) is formed as written. Every x∈I lies strictly between two points of I: since I is open, there is a real r>0 with BdR(x,r)⊆I, and x−r/2 and x+r/2 lie in this ball by Open Ball in a Metric Space (ii), as their distance to x is r/2<r.
Step 1: increasing slopes. Let s,t,u∈I with s<t<u, and put λ=(u−t)/(u−s). Then 0<λ<1, 1−λ=(t−s)/(u−s) and λs+(1−λ)u=t, so convexity (Convex Functions on an Interval of ℝ (i), with s, u and the weight λ) gives f(t)≤λf(s)+(1−λ)f(u). Subtracting f(s), respectively f(u), from both sides and rearranging,
f(t)−f(s)≤u−st−s(f(u)−f(s)),f(u)−f(t)≥u−su−t(f(u)−f(s)).
Dividing the first inequality by t−s>0 and the second by u−t>0 gives the two inequalities of (i).
Step 2: the quotients at a point increase. Let x∈I and y,z∈I∖{x} with y<z. We show Qx(y)≤Qx(z). If x<y<z, this is the left inequality of (i) for (s,t,u)=(x,y,z). If y<z<x, it is the right inequality for (y,z,x), since Qx(y)=(f(x)−f(y))/(x−y) and Qx(z)=(f(x)−f(z))/(x−z). If y<x<z, it is the outer inequality for (y,x,z):
Qx(y)=x−yf(x)−f(y)≤z−xf(z)−f(x)=Qx(z).
Step 3: the infimum cx. Let x∈I, and choose s,t∈I with s<x<t, as above; so the set Rx={Qx(y):y∈I, y>x} contains Qx(t), and by Step 2 every s′∈I with s′<x gives the lower bound Qx(s′) of Rx. Hence Rx has an infimum cx∈R, by The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness (viii); as a lower bound of Rx it satisfies cx≤Qx(t′) for every t′∈I with t′>x, and as the greatest lower bound it satisfies Qx(s′)≤cx for every s′∈I with s′<x. Since Qx(s′)=(f(x)−f(s′))/(x−s′), this is (ii).
Step 4: supporting lines. Let x,y∈I. For y=x there is nothing to prove. If y>x, then cx≤Qx(y) by (ii), and multiplying by y−x>0 gives f(y)−f(x)≥cx(y−x). If y<x, then Qx(y)≤cx by (ii), and multiplying by y−x<0 gives f(y)−f(x)≥cx(y−x) again. This is (iii).
Step 5: continuity. We use the ε-δ form of continuity on I of The Real Line: Standing Notation and Background for Calculus on Intervals (vi). Let x∈I and let ε>0 be real. Choose s,t∈I with s<x<t, as above, and put M=∣Qx(s)∣+∣Qx(t)∣. For y∈I with s<y<t and y=x, Step 2 gives Qx(s)≤Qx(y)≤Qx(t), hence ∣Qx(y)∣≤M and
∣f(y)−f(x)∣=∣Qx(y)∣∣y−x∣≤M∣y−x∣,
which also holds for y=x. Put δ=min{x−s, t−x, ε/(M+1)}>0. Every y∈I with ∣y−x∣<δ satisfies s<y<t, so
∣f(y)−f(x)∣≤(M+1)∣y−x∣<(M+1)δ≤ε.
Hence f is continuous on I, which is (iv).
Step 6: the rational supporting lines. Let y∈I and let Sy={f(q)+cq(y−q):q∈I∩Q}. By (iii), applied at q, every element of Sy is at most f(y). Choose s∈I with s<y, as above; by The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness (x) there is q∈Q with s<q<y, and q∈I by Intervals of the Real Line (i). So Sy is nonempty and bounded above by f(y); it has a supremum by The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness (vii), and supSy≤f(y).
For the reverse inequality put r=(s+y)/2, so s<r<y and r∈I, and
L=r−sf(r)−f(s).
For q∈I with r<q<y, (ii) (with s<q) and then (i) (with s<r<q) give
cq≥q−sf(q)−f(s)≥L,hencef(q)+cq(y−q)≥f(q)+L(y−q)≥f(q)−∣L∣(y−q).
Now let ε>0 be real. By (iv) there is a real δ>0 such that ∣f(q)−f(y)∣<ε/2 for every q∈I with ∣q−y∣<δ. Put a=max{r, y−δ, y−ε/(2∣L∣+2)}, so a<y; by The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness (x) there is q∈Q with a<q<y, and q∈I since r<q<y. Then (∣L∣+1)(y−q)<ε/2, and
f(q)+cq(y−q)≥f(q)−∣L∣(y−q)>(f(y)−2ε)−2ε=f(y)−ε.
So supSy>f(y)−ε for every real ε>0, hence supSy≥f(y), and with the first part f(y)=supSy. This is (v).