Reason: First published version. Continuity is obtained by gluing over the two open sets {y : x-y in Omega} and {||y|| > delta}, which cover R^n exactly because the closed delta-ball about x lies in Omega; support, integrability and the bound then follow from the toolkit for continuous compactly supported functions.
Step 1: U1 and U2 are open and cover Rn. Let y0∈U1. Since x−y0∈Ω and Ω is open there is a real r>0 with the open ballB(x−y0,r) contained in Ω. If d(y,y0)<r then, since (x−y)−(x−y0)=y0−y, we get d(x−y,x−y0)=∥y0−y∥=d(y,y0)<r, so x−y∈Ω and y∈U1. Hence U1 is open.
Let y0∈U2 and put r=∥y0∥−δ>0. If d(y,y0)<r then claim 6 of Elementary Properties of the Euclidean Norm on Rn, applied to the points y and y0−y whose sum is y0, gives ∥y0∥≤∥y∥+∥y0−y∥<∥y∥+r, so ∥y∥>∥y0∥−r=δ and y∈U2. Hence U2 is open.
Step 2: hx is continuous. On U1 we have hx(y)=f(x−y)ρ(y) by definition. The map y↦f(x−y) is continuous on U1: given y0∈U1 and a real ε>0, continuity of f at x−y0 relative to Ω supplies a real θ>0 such that every z∈Ω with d(z,x−y0)<θ satisfies ∣f(z)−f(x−y0)∣<ε; and for y∈U1 with d(y,y0)<θ the point z=x−y lies in Ω and satisfies d(x−y,x−y0)=d(y,y0)<θ, as computed in step 1. The map ρ, being continuous on Rn, is continuous on U1 relative to U1, since a θ that works for points of Rn works in particular for points of U1. Hence hx is continuous on U1 by Continuity of Sums and Products of Real-Valued Functions on a Metric Space.
On U2 we have hx(y)=0: either x−y∈/Ω, and then hx(y)=0 by definition, or x−y∈Ω, and then hx(y)=f(x−y)ρ(y)=0 because ∥y∥>δ forces ρ(y)=0. A constant function is continuous on U2.
Now let y0∈Rn and let ε>0 be real. By step 1 the point y0 lies in Ui for some i∈{1,2}, and Ui is open, so there is a real r>0 with B(y0,r)⊆Ui. Continuity of hx on Ui at y0 relative to Ui supplies a real θ>0 such that every y∈Ui with d(y,y0)<θ satisfies ∣hx(y)−hx(y0)∣<ε. Let θ′ be the smaller of r and θ; every y∈Rn with d(y,y0)<θ′ lies in Ui and satisfies ∣hx(y)−hx(y0)∣<ε. Hence hx is continuous at y0 relative to Rn, and as y0 was arbitrary, hx is continuous on Rn.
Step 4: the bound. Let M≥0 be real with ∣f(z)∣≤M for every z∈Bˉ(x,δ), and let y∈Rn. If ∥y∥>δ then hx(y)=0 and ρ(y)=0, so ∣hx(y)∣=0=M∣ρ(y)∣. If ∥y∥≤δ then, as in step 1, x−y∈Bˉ(x,δ)⊆Ω, so hx(y)=f(x−y)ρ(y) and ∣hx(y)∣=∣f(x−y)∣∣ρ(y)∣≤M∣ρ(y)∣. Thus ∣hx∣≤M∣ρ∣ pointwise.
Both ∣hx∣ and M∣ρ∣ are integrable with respect to λn: the first because hx is integrable, the second by claim 2 of Linearity and Monotonicity of the Lebesgue Integral applied to the integrable function ∣ρ∣. By claim 2 of that theorem,