Multivariable calculus

Limit of xyz/(x^2+y^2+z^2) at the Origin

The limit of xyz / (x^2 + y^2 + z^2) as (x, y, z) approaches the origin is 0. Since 2|xy| is at most x^2 + y^2, the quotient is bounded in size by |z|/2, which goes to 0 under any approach. In spherical coordinates the function is rho times a bounded trigonometric factor.

f(x,y,z)=xyzx2+y2+z2f(x,y,z) = \frac{xyz}{x^2 + y^2 + z^2}

The limit exists

lim(x,y,z)(0,0,0)xyzx2+y2+z2=0\lim_{(x,y,z) \to (0,0,0)} \frac{xyz}{x^2 + y^2 + z^2} = 0
PathLimit along it
along the x axis0
along the diagonal x = y = z0
along y = x in the plane z = 00

Paths in space

Start with an axis, the main diagonal x=y=zx = y = z, and a line lying in the plane z=0z = 0.

f(t,0,0)=0t2=0f(t,0,0) = \frac{0}{t^2} = 0
f(t,t,t)=t33t2=t30f(t,t,t) = \frac{t^3}{3t^2} = \frac{t}{3} \to 0
f(t,t,0)=02t2=0f(t,t,0) = \frac{0}{2t^2} = 0

In three variables the path problem gets worse, not better. Approaches can run along lines, curves, or spirals, and the set of directions is a whole sphere rather than a circle. Testing a few and generalising is even less defensible than it was in the plane.

A one line bound

Start from (xy)20\left(|x| - |y|\right)^2 \ge 0, which rearranges to 2xyx2+y22|xy| \le x^2 + y^2. Feed that into the numerator.

f=xyzx2+y2+z2z(x2+y2)2(x2+y2+z2)z2\left|f\right| = \frac{|x||y||z|}{x^2+y^2+z^2} \le \frac{|z|\left(x^2+y^2\right)}{2\left(x^2+y^2+z^2\right)} \le \frac{|z|}{2}

The last step uses x2+y2x2+y2+z2x^2 + y^2 \le x^2+y^2+z^2, so that fraction is at most 11. Every approach to the origin forces z0z \to 0, so the bound collapses and ff is squeezed to 00.

0fz200 \le \left|f\right| \le \frac{|z|}{2} \to 0

Spherical coordinates give the same conclusion and show why it happens. With x=ρsinφcosθx = \rho\sin\varphi\cos\theta, y=ρsinφsinθy = \rho\sin\varphi\sin\theta, z=ρcosφz = \rho\cos\varphi, the numerator is degree three in ρ\rho and the denominator degree two.

f=ρsin2φcosφcosθsinθ,fρf = \rho\sin^2\varphi\cos\varphi\cos\theta\sin\theta, \qquad \left|f\right| \le \rho

The mistake: checking the coordinate axes only

Along each coordinate axis two of the three variables are zero, so the numerator xyzxyz is identically zero and the function is flat at 00. Three effortless zeros feel conclusive and carry no information at all: the path killed the numerator before the denominator ever had a chance to matter.

A useful habit follows from that. When a path makes the numerator vanish identically, that path tells you nothing about the limit. Pick approaches where the numerator survives, such as x=y=z=tx = y = z = t above, and then close the argument with a bound that holds on a whole ball.

The same trap appears in two variables with functions like xyx2+y2\frac{xy}{x^2+y^2}, which is 00 along both axes and 12\frac{1}{2} along y=xy = x, and has no limit. Axis testing would have reported a limit of 00 with total confidence.

Frequently asked questions

Does this limit exist in every dimension?

The same argument works for x1x2xnx12++xn2\frac{x_1 x_2 \cdots x_n}{x_1^2 + \cdots + x_n^2} whenever n3n \ge 3, because the numerator has degree nn and the denominator degree 22. At n=2n = 2 the quotient is xyx2+y2\frac{xy}{x^2+y^2}, which has no limit: 00 along the axes, 12\frac{1}{2} along y=xy = x.

Is the bound |z|/2 better than the spherical one?

They are the same fact in different clothes. The bound z2\frac{|z|}{2} needs no change of coordinates and is quicker to write down. The spherical version makes the reason visible: three factors of the radius upstairs against two downstairs, leaving one factor of ρ\rho to drive everything to zero.