SHORTEST DISTANCE BETWEEN TWO POINTS ON A SPHERE
It is known that the shortest distance between point A and point B on the surface of
a sphere of radius R is part of a great circle lying in a plane intersecting the sphere
surface and containing the points A and B and the point C at the sphere center. Let
us use the calculus of variations and spherical coordinates to define this great circle
and show how to calculate the geodesic distance between points A and B on the
surface. One starts with the definition of length between points A and B along the
great circle. Mathematically one has-
θ =θ B dϕ 2
L=R ∫ dθ 1 + sin 2 θ ( )
θ =θ A dθ
where one is using spherical coordinates defined as-
x = R sin θ cosϕ , y = R sin θ sin ϕ , z = R cosθ
With θ being the polar angle and φ the azimuthal angle. In terms of latitude and
longitude on the earth one has LAT=(π/2-θ) and LONG=φ , when measured relative
to Greenwich. Now according to the Euler-Lagrange equation one has that the
integrand is an extremum(here a minimum) when-
d ∂ dϕ 2 c 21
=0 so that ( ) =
dθ ∂ϕ ' dθ sin 2 θ [sin θ 2 − c 2 ]
where c is an adjustable constant. The geodesic distance between point A and B is
thus-
θB sin θ
L=R ∫ dθ
2 2
θA sin θ − c
On integrating one finds-
R 2i (δ − γ ) [2a − (1 − c 2 )] [2b − (1 + c 2 ]
L = arctan with δ = and γ =
2 γδ − 4 2 2
a − a (1 + c ) + c 2
b 2 − b(1 + c 2 ) + c 2
and a=sin(θa)2, b=sin(θb)2. The constant c2 is determined by solving the first order
equation for dφ/dθ which brings in the values of φa and φb.
To test out this result consider point A at θ=π/4 and φ=0 and point B at to θ=π/2 and
φ=π/2. Here we have a=sin(π/4)2 =1/2 and b=sin(π/2)2 =1 so that γ=∞ and
δ=2c2/sqrt(2c2-1). This in turn implies that-
R 2i (−∞) R i 2c 2 − 1
L = arctan = arctan[ ]
2 (−∞)2c 2 / 2c 2 − 1 2 2
c
and will have a minimum positive value of L =πR/2 when c2=0.5, a value consistent
with the latitudes of the end points of the geodesic. That this result must be correct
is seen by visualizing a globe and noting that the distance from A to B will be just
equal to one fourth of the global circumference. In general to find c2 one needs to
first solve the dφ/dθ equation and then plug into the above solution for L. This can
be a rather lengthy procedure and instead one usually uses an alternate route based
upon spherical geometry( see our earlier discussion on spherical geometry). The law
of cosines for a spherical triangle having corners at A and B plus a third corner at
the pole P of the globe leads to the simple formula-
L=Rcos-1{cos[LAT(A)-LAT(B)]-[cos(LAT(B)cos(LAT(A)][1-cos(LONG(B)-
LONG(A))]}
where LAT and LONG refer to the latitude and longitude of the end points A and B
, respectively.
Thus for the same end points considered above we have LAT(A)=π/4, LONG(A)=0,
LAT(B)=0, and LONG(B)=π/2. This leads (with much less effort) to the same result-
L=Rcos-1 {cos[π/4-0]-[cos(0)cos(π/4)][1- cos(π/2-0)]}
=Rcos-1 {1/sqrt(2)-1/sqrt(2)}=πR/2
The length of the geodesic going from the north pole to the south pole is calculated
using LAT(A)=π/2, LAT(B)=- π/2 with LONG(A)=LONG(B)=0 say. This yields at
once that –
L= Rcos-1 { cos(π/2+ π/2)-[cos(-π/2)cos(π/2)][1-cos(0-0)]=πR
We conclude our discussion by looking at Erastosthenes measurement of the earth’s
circumference C=2πR. He noted that at the summer solstice the sun in Aswan at
noon was directly overhead while the sun at his home in Alexandria, Egypt was at
7deg from the zenith at noon of the same day He also knew the distance between
Alexandria and Syene(Aswan)was 500 miles directly south. Although he had no
knowledge of extremum principles and the formulas of spherical trigonometry, he
nevertheless came up with the first (and surprisingly accurate) measure of the
earth’s circumference. In terms of the above discussion one has at Alexandria LAT=
31degN and LONG(A)=30degE and at Aswan LAT=24degN and LONG=30degE.
Also L=500miles. Plugging these numbers into the above equation for L one has
approximately-
cos(500/R)= {cos(7deg)-[cos(24deg)cos(31deg)][1-cos(0)]
or
500/R=cos-1{cos(7/57.29)}=0.12218 or R=4092 miles
A not a bad result compared to the accurate value of R=3960 miles, especially since
it was carried out some2249 years ago. Columbus would have done well to study
Eratosthenes results so that he would have had a better estimate of the distance to
East Indies when sailing west from Spain.
June 2009
