Proof (I) Let O,G be the circumcenter and centroid of tetrahedron A1A2A3A4, respectively. Extend A1G to intersect the plane A2A3A4 at G1, then G1 is the centroid of △A2A3A4. Connect A2G1 to intersect A3A4 at point M, then M is the midpoint of A3A4, as shown in Figure 22-9.
From the proof in Example 9, we know that
A1G12=91[3(A1A22+A1A32+A1A42)−(A2A32+A2A42+A3A42)].
Similarly, in the tetrahedron OA2A3A4, we have
OG12=91[3(OA22+OA32+OA42)−(A2A32+A2A42+A3A42)]=R2−91(A2A32+A2A42+A3A42).
Since G is the centroid of the tetrahedron, by property 3, we know GG1A1G=13. Thus, in △A1OG1, applying Stewart's theorem to point G, we have
OG2=161[4(3OG12+OA12)−3A1G12]=161[16R2−(A1A22+A1A32+A1A42)−(A2A32+A2A42+A3A42)].
Since OG2⩾0, it follows that ∑1⩽k<j⩽4AkAj⩽16R2.
(II) Clearly, A1G1⩾h, then h12⩽91[3(A1A22+A1A32+A1A42)−(A2A32+A2A42+A3A42)]. Similarly, for h2,h3,h4, we have similar inequalities.
Adding these four inequalities, we get ∑1⩽k<j⩽4AkAj⩾49∑i=14hi2.
(III) By V=31Sihi(i=1,2,3,4), we have ∑i=14hi1=3V1∑i=14Si.
Also, by V=31∑i=14Si⋅r, we have ∑i=14hi1=r1.
By ∑i=14hi1⩾4(h1h2h3h41)41, we have h1h2h3h4⩾(4r)4.
Thus, h12+h22+h32+h42⩾4(h1h2h3h4)21⩾4[(4r)4]21=(8r)2.
(IV) Let the circumradius of tetrahedron A′1A′2A′3A′4 be R′, then R′2⩾161∑1⩽k<j⩽4A′kA′j.
The circumradius of tetrahedron A′1A′2A′3A′4 is exactly the inradius of tetrahedron A1A2A3A4, so R′=r. Thus, R2⩾161∑1⩽k<j⩽4AkAj⩾649∑i=14hi2⩾9r2=9R′2⩾169∑1⩽k<j⩽4A′kA′j.
Therefore, ∑1⩽k<j⩽4AkAj⩾91⩽k<j⩽4A′kA′j.