While it is possible to (at least partially) adapt the definition to handle negative numbers, I do not believe this is ever done.
(In passing, the are two more terms in Euclid VI that relate to the proportions like the above. The fourth proportional of the given numbers a, b, c is number x such that ab = c/x. The third proportional of two numbers a and b is number y such that a/b = b/y. Also, this same construction is used by Euclid II.14 to construct a square of the same area as a given rectangle.)
The above construction of the mean proportional is based on a Corollary from Euclid VI.8: If in a right-angled triangle a perpendicular is drawn from the right angle to the base, then the straight line so drawn is a mean proportional between the segments of the base. However this is not the only appearance of the mean proportional in right triangle. For example, in a right triangle with the altitude to the hypotenuse drawn we may observe three similar triangles: the given one, and the smaller ones cut off by the altitude. The corollary to VI.8 is derived from the similarity of two small triangles. If we pair the big triangle with any of the smaller ones, we'll find that a leg of a right triangle is the mean proportional between its projection on the hypotenuse and the hypotenuse itself.
In fact the geometric mean makes quite frequent appearances in a variety of geometric situations. I'll mention a few.
The length of the common tangent of two circles of diameters a and b that are tangent externally is the geometric mean of the radii:
The geometric mean appeared as a tangent to a circle in John Wallis' very first geometric interpretation of complex numbers:
In the framework of the See-Saw Lemma, if a semicircle is inscribed between two perpendiculars to its diameter and a transversal tangent to the semicircle cut segments of lengths a and b from the two lines, then the radius of the semicircle is the geometric mean of a and b.
In one of the simplest sangaku a square is inscribed into a right triangle. The process of inscribing a square continues for two more steps with the cut off triangles. Inscribe the incircles into three similar and similarly obtained triangles. Let their radii be a, p, b in the decreasing order, then p is the mean proportional of a and b.
Let AB be a chord in a circle and P a point on the circle. Draw perpendiculars PQ, PR, and PS from P to AB and the tangents to the circle at A and B. Then PQ² = PR·PS.
Let points C and D lie on a semicircle with diameter AB. Let E be the intersection of AC and BD and F the intersection of AD and BC. Let EF meet the semicircle in G and AB in H. Then GH² = EH·FH
.
[W. H. Besant, Conic Sections Treated Geometrically, George Bell & Sons, London, 1895, p. 28]. If from a point Q tangents QP, QP' be drawn to a parabola, the two triangles SPQ and SQP' (S the focus of the parabola), are similar, and SQ is a mean proportional between SP and SP'.
Produce PQ to meet the axis in T, and draw SY, SY' perpendicularly on the tangents. Then Y and Y' are points on the tangent at A.
| ∠SPQ | = ∠STY |
| | = ∠SYA |
| | = ∠SQP', |
since S, Y', Y, Q are points on a circle, and SYA, SQP' are in the same segment.
Also, since the tangents drawn from any point to a conic subtend equal angles at the focus, ∠PSQ = ∠QSP'; therefore the triangles PSQ, QSP' are similar, and
SP:SQ = SQ:SP'.
