The problem of constructing magic squares with prime numbers only was
first discussed by myself in _The Weekly Dispatch_ for 22nd July and 5th
August 1900; but during the last three or four years it has received
great attention from American mathematicians. First, they have sought to
form these squares with the lowest possible constants. Thus, the first
nine prime numbers, 1 to 23 inclusive, sum to 99, which (being divisible
by 3) is theoretically a suitable series; yet it has been demonstrated
that the lowest possible constant is 111, and the required series as
follows: 1, 7, 13, 31, 37, 43, 61, 67, and 73. Similarly, in the case of
the fourth order, the lowest series of primes that are "theoretically
suitable" will not serve. But in every other order, up to the 12th
inclusive, magic squares have been constructed with the lowest series of
primes theoretically possible. And the 12th is the lowest order in which
a straight series of prime numbers, unbroken, from 1 upwards has been
made to work. In other words, the first 144 odd prime numbers have
actually been arranged in magic form. The following summary is taken
from _The Monist_ (Chicago) for October 1913:--

Order of Totals of Lowest Squares
Square. Series. Constants. made by--
(Henry E.
3rd 333 111 { Dudeney
( (1900).

(Ernest Bergholt
4th 408 102 { and C. D.
( Shuldham.

5th 1065 213 H. A. Sayles.