Limits and continuity in several variables
A point of the plane can be approached from infinitely many directions, along curves as well as lines. What does it mean for f(x, y) to approach a number there, and how do you prove that it does, or that it does not?
Where you are 8 in Partial derivatives
Why this exists
On the number line, x can approach a from two sides only, and a limit exists when the two one-sided limits agree (One sided limits and when a limit exists). A point (a, b) of the plane can be approached along every line through it, along parabolas and spirals, and along paths with no name at all. The traces of Functions of several variables and level sets show what f does along two of them.
Jasmine and Isaac run the sonar on a fishing trawler. The screen, centered on the sonar with x and y in meters east and north, shows the direction to each echo as a unit vector with east component \displaystyle u(x, y) = \dfrac{x}{\sqrt{x^2 + y^2}}. What should u show at the center? No value works: echoes arbitrarily close to the sonar have every east component from -1 to 1 (Example 4). Saying this exactly, and proving the limits that do exist, takes the \varepsilon-\delta definition of The epsilon delta definition of a limit with distance in the plane in place of \lvert x - a \rvert.
Definition
The distance from (x, y) to (a, b) is \displaystyle \sqrt{(x - a)^2 + (y - b)^2}, and the open disk of radius \delta > 0 about (a, b) is
A point (a, b) is a limit point of a set D \subseteq \mathbb{R}^2 if every disk B_\delta(a, b) contains a point of D other than (a, b). It need not lie in D: the origin is a limit point of the plane with the origin removed.
Limit. Let f : D \to \mathbb{R}, let (a, b) be a limit point of D, and let L be a real number. Then \lim_{(x, y) \to (a, b)} f(x, y) = L means:
> For every \varepsilon > 0 there is a \delta > 0 such that every (x, y) \in D with \displaystyle 0 < \sqrt{(x - a)^2 + (y - b)^2} < \delta satisfies \lvert f(x, y) - L \rvert < \varepsilon.
In plain words: every point of the domain close enough to (a, b), from whatever direction, other than (a, b) itself, has its value within \varepsilon of L. Only points of D are tested, and the value at (a, b), if there is one, plays no part. For f of n variables the definition is the same with the distance \lVert \mathbf{x} - \mathbf{a} \rVert of Dot product, length and orthogonality in n dimensions, and the disk becomes a ball.
Continuity. f : D \to \mathbb{R} is continuous at a point (a, b) of D if for every \varepsilon > 0 there is a \delta > 0 with \lvert f(x, y) - f(a, b) \rvert < \varepsilon for every (x, y) \in D within \delta of (a, b). When (a, b) is a limit point of D, this says \lim_{(x, y) \to (a, b)} f(x, y) = f(a, b), the condition of Continuity at a point. f is continuous on D if it is continuous at every point of D.
Path. A path to (a, b) in D is a function \mathbf{r}(t) = (x(t), y(t)), defined for 0 < t < \tau, with \mathbf{r}(t) \in D, \mathbf{r}(t) \ne (a, b), and x(t) \to a, y(t) \to b as t \to 0^+.
Intuition
In one variable the picture was a band of height 2\varepsilon around L and a strip around a. Here the strip becomes a disk: however thin the band, the limit asks for a disk around (a, b) whose points of D, the center excepted, all have values in the band. Lines, parabolas and spirals all enter the disk (Figure 1), and all of them count.
Two consequences shape every example. To prove a limit you need an estimate that holds at all points near (a, b) at once, usually a bound by the distance alone. To disprove one, two paths along which f tends to different numbers are enough, since the definition promises a single L along all of them. No list of paths you check proves that a limit exists, not even an infinite list such as every line through the point (Example 5).
Polar coordinates about (a, b), x = a + r\cos\theta and y = b + r\sin\theta (Polar coordinates), are made for this: r is exactly the distance to (a, b), and \theta is the direction of approach.
Results
Result 1 (uniqueness). A function has at most one limit at a limit point of its domain.
Result 2 (limit laws). If \lim_{(x, y) \to (a, b)} f(x, y) = L and \lim_{(x, y) \to (a, b)} g(x, y) = M, then f + g, f - g, cf and fg have limits L + M, L - M, cL and LM, and if M \ne 0, \displaystyle \frac{f}{g} has limit \displaystyle \frac{L}{M}.
Result 3 (continuous building blocks). The coordinate functions x and y are continuous on \mathbb{R}^2. So every polynomial in x and y is continuous on \mathbb{R}^2, and a rational function is continuous at every point where its denominator is not 0. If \lim_{(x, y) \to (a, b)} g(x, y) = c and h, a function of one variable defined at the values of g, is continuous at c, then \lim_{(x, y) \to (a, b)} h(g(x, y)) = h(c). In particular h \circ g is continuous wherever g is continuous and h is continuous at the value of g. All of this holds in \mathbb{R}^n.
Result 4 (a bound by the distance). Let r be the distance from (x, y) to (a, b). If \lvert f(x, y) - L \rvert \le g(r) for every (x, y) \in D with 0 < r < \rho, and g(r) \to 0 as r \to 0^+, then \lim_{(x, y) \to (a, b)} f(x, y) = L. In polar form: a bound \lvert f(a + r\cos\theta, b + r\sin\theta) - L \rvert \le g(r) that holds for every \theta.
Result 5 (paths). If \lim_{(x, y) \to (a, b)} f(x, y) = L and \mathbf{r} is a path to (a, b) in D, then \lim_{t \to 0^+} f(\mathbf{r}(t)) = L. So if f tends to different numbers along two paths, or has no limit along one, then f has no limit at (a, b).
Why it is true
Write r for the distance to (a, b). Since (p + q)^2 \ge p^2 + q^2 for p, q \ge 0, we have \max(\lvert x - a \rvert, \lvert y - b \rvert) \le r \le \lvert x - a \rvert + \lvert y - b \rvert.
Result 1. Suppose L \ne M are both limits, and let \displaystyle \varepsilon = \frac{\lvert L - M \rvert}{2}. Take \delta_1 for L and \delta_2 for M, and \delta = \min(\delta_1, \delta_2). Because (a, b) is a limit point of D, some \mathbf{p} \in D has 0 < r < \delta, and then \lvert L - M \rvert \le \lvert L - f(\mathbf{p}) \rvert + \lvert f(\mathbf{p}) - M \rvert < 2\varepsilon = \lvert L - M \rvert, a contradiction. Without a nearby point of D every number would pass the test, which is why the definition asks for a limit point.
Result 2. The sum law is Exercise 18. For the product, given \varepsilon, let \displaystyle \eta = \min\left(1, \frac{\varepsilon}{1 + \lvert L \rvert + \lvert M \rvert}\right) and take \delta so small that \lvert f - L \rvert < \eta and \lvert g - M \rvert < \eta on 0 < r < \delta. Since
and \eta^2 \le \eta, we get \lvert fg - LM \rvert < \eta(1 + \lvert M \rvert + \lvert L \rvert) \le \varepsilon. For the quotient it is enough to show \displaystyle \frac{1}{g} \to \frac{1}{M} and use the product law. Taking \displaystyle \varepsilon = \frac{\lvert M \rvert}{2} gives a disk where \displaystyle \lvert g \rvert > \frac{\lvert M \rvert}{2}, and there \displaystyle \left\lvert \frac{1}{g} - \frac{1}{M} \right\rvert = \frac{\lvert g - M \rvert}{\lvert g \rvert \lvert M \rvert} < \frac{2\lvert g - M \rvert}{M^2}, which is small once \lvert g - M \rvert is. The constant multiple and difference laws follow from these.
Result 3. For the coordinate function x, \lvert x - a \rvert \le r, so \delta = \varepsilon works; the same for y. Constants are continuous, and a polynomial is built from constants and coordinates by sums and products, so Result 2 makes it continuous; a quotient of polynomials is continuous where the denominator is not 0 by the quotient law. For the composition, let \varepsilon > 0. Continuity of h at c gives \eta > 0 with \lvert h(s) - h(c) \rvert < \varepsilon whenever \lvert s - c \rvert < \eta; the limit of g, with \eta in the role of \varepsilon, gives \delta with \lvert g(x, y) - c \rvert < \eta on 0 < r < \delta. There \lvert h(g(x, y)) - h(c) \rvert < \varepsilon.
Result 4. Given \varepsilon > 0, choose \delta \le \rho with g(r) < \varepsilon for 0 < r < \delta. Then \lvert f(x, y) - L \rvert \le g(r) < \varepsilon there.
Result 5. Let \varepsilon > 0 and take \delta from the limit of f. As x(t) \to a and y(t) \to b, there is \eta \le \tau with \displaystyle \lvert x(t) - a \rvert < \frac{\delta}{2} and \displaystyle \lvert y(t) - b \rvert < \frac{\delta}{2} for 0 < t < \eta. For those t, \mathbf{r}(t) \in D, its distance to (a, b) is positive because \mathbf{r}(t) \ne (a, b), and it is less than \displaystyle \frac{\delta}{2} + \frac{\delta}{2} = \delta. So \lvert f(\mathbf{r}(t)) - L \rvert < \varepsilon. If two paths gave limits M_1 \ne M_2, both would equal L, which is impossible.
Worked examples
Limits from continuity
Find \displaystyle \lim_{(x, y) \to (2, -1)} \sqrt{x^2 + 5y^2}\, e^{x + 2y} and \lim_{(x, y) \to (2, 2)} \dfrac{x^3 - y^3}{x - y}.
SolutionThe polynomial x^2 + 5y^2 is continuous with value 9 at (2, -1), and the square root is continuous at 9, so by Result 3 the first factor tends to 3. Likewise x + 2y \to 0 and e^{x + 2y} \to e^0 = 1. By the product law the limit is 3.
The second function has domain D = \{(x, y) : x \ne y\}. The point (2, 2) is not in D but is a limit point of it, through the points (2 + t, 2), t \ne 0. On D,
The definition tests only points of D, where f agrees with this polynomial, so the limit is 4 + 4 + 4 = 12.
A limit from the definition
Prove from the definition that \lim_{(x, y) \to (2, -1)} xy = -2.
SolutionLet \displaystyle r = \sqrt{(x - 2)^2 + (y + 1)^2}, so \lvert x - 2 \rvert \le r and \lvert y + 1 \rvert \le r. Write xy + 2 in the small quantities x - 2 and y + 1:
as expanding the right side shows. By the triangle inequality \lvert xy + 2 \rvert \le r^2 + r + 2r = r^2 + 3r. If r < 1 then r^2 < r, so \lvert xy + 2 \rvert < 4r.
Let \varepsilon > 0 and \displaystyle \delta = \min\left(1, \frac{\varepsilon}{4}\right). Every (x, y) with 0 < r < \delta has \displaystyle \lvert xy - (-2) \rvert < 4r < 4 \cdot \frac{\varepsilon}{4} = \varepsilon. The cap \delta \le 1 controls the quadratic term.
A bound in polar coordinates
Find \lim_{(x, y) \to (0, 0)} \dfrac{x^3 - 4y^3}{x^2 + y^2}, and decide whether f, given the value 0 at the origin, is continuous on \mathbb{R}^2.
SolutionOn the axes f(x, 0) = x and f(0, y) = -4y, both tending to 0, so by Result 5 the only possible limit is 0. To prove it, put x = r\cos\theta, y = r\sin\theta with r > 0:
Since \lvert \cos^3\theta - 4\sin^3\theta \rvert \le 1 + 4 = 5 for every \theta, \lvert f(x, y) \rvert \le 5r, and Result 4 with g(r) = 5r gives the limit 0; explicitly, \displaystyle \delta = \frac{\varepsilon}{5} works. The bound does not depend on \theta, and that is the whole point.
The extended function is continuous at the origin, where limit and value are both 0, and elsewhere by Result 3, so on all of \mathbb{R}^2.
The sonar has no direction at its center
Prove: if f is defined on the plane without the origin, f(tx, ty) = f(x, y) for every t > 0, and f takes at least two different values, then f has no limit at (0, 0). Apply this to the east component \displaystyle u(x, y) = \dfrac{x}{\sqrt{x^2 + y^2}} on Jasmine and Isaac’s screen.
SolutionChoose points \mathbf{p} and \mathbf{q}, not the origin, with f(\mathbf{p}) \ne f(\mathbf{q}). The ray \mathbf{r}(t) = t\mathbf{p}, 0 < t < 1, is a path to the origin in the domain: t\mathbf{p} \ne (0, 0) and t\mathbf{p} \to (0, 0). Along it f(t\mathbf{p}) = f(\mathbf{p}) for every t, so the limit along the path is f(\mathbf{p}). Along t\mathbf{q} it is f(\mathbf{q}). Two paths, two limits: by Result 5, f has no limit at the origin.
For t > 0, \displaystyle u(tx, ty) = \dfrac{tx}{\sqrt{t^2x^2 + t^2y^2}} = \dfrac{tx}{t\sqrt{x^2 + y^2}} = u(x, y), and u(1, 0) = 1 while u(0, 1) = 0. So u has no limit at the sonar, and no value at the center makes it continuous there. In polar coordinates u = \cos\theta, so for -1 < c < 1 the level curve u = c is the pair of rays at angles \pm\arccos c, each running into the center (Figure 2).
Every line agrees, and the limit still fails
Let f(x, y) = \dfrac{2x^2 y}{x^4 + 3y^2}. Show that f tends to 0 along every line through the origin, but has no limit at (0, 0).
SolutionBoth terms of the denominator are at least 0, and both vanish only when x = y = 0, so the domain is the plane without the origin. On the axes f(x, 0) = 0 and f(0, y) = 0. Along y = mx with m \ne 0, for x \ne 0,
Along the parabola y = kx^2, for x \ne 0,
a constant. For k = 1 it is \displaystyle \frac12, and along the x-axis it is 0, so by Result 5 there is no limit. The value \displaystyle \frac12 comes from two parabolas, since \displaystyle \frac{2k}{1 + 3k^2} = \frac12 means 3k^2 - 4k + 1 = 0, that is, k = 1 or \displaystyle k = \frac13 (Figure 3).
In polar coordinates f(r\cos\theta, r\sin\theta) \to 0 for each fixed \theta, which is the line test again; but the convergence is not uniform in \theta, so no bound g(r) as in Result 4 exists.
Three variables
Prove that F(x, y, z) = \dfrac{x^2 y + y^2 z + z^2 x}{x^2 + y^2 + z^2}, with F(0, 0, 0) = 0, is continuous on \mathbb{R}^3.
SolutionAway from the origin F is a rational function with nonzero denominator, continuous by Result 3. At the origin, let \displaystyle \rho = \sqrt{x^2 + y^2 + z^2}; each of \lvert x \rvert, \lvert y \rvert, \lvert z \rvert is at most \rho. So \lvert x^2 y \rvert \le x^2 \rho, \lvert y^2 z \rvert \le y^2 \rho and \lvert z^2 x \rvert \le z^2 \rho, and
Hence \lvert F \rvert \le \rho for \rho > 0, and Result 4 in \mathbb{R}^3 gives \lim F = 0 = F(0, 0, 0), with \delta = \varepsilon. Along the path x = y = z = t, \displaystyle F = \frac{3t^3}{3t^2} = t \to 0, which agrees, but only the bound proves it.
A function defined in two pieces
Let s(x, y) = \dfrac{\sin(xy)}{y} for y \ne 0 and s(x, 0) = x. Prove that s is continuous on \mathbb{R}^2, and find \lim_{(x, y) \to (3, 0)} s(x, y).
SolutionLet \sigma(w) = \dfrac{\sin w}{w} for w \ne 0 and \sigma(0) = 1. It is continuous at every w \ne 0 as a quotient, and at 0 because \displaystyle \lim_{w \to 0} \frac{\sin w}{w} = 1 (The squeeze theorem and the limit of sine x over x). We claim s(x, y) = x\,\sigma(xy) at every point. If y \ne 0 and x \ne 0, then xy \ne 0 and \displaystyle x\,\sigma(xy) = \frac{x\sin(xy)}{xy} = \frac{\sin(xy)}{y}. If y \ne 0 and x = 0, both sides are 0. If y = 0, x\,\sigma(0) = x = s(x, 0).
Now xy is a continuous polynomial, so \sigma(xy) is continuous by Result 3, and the product with x is continuous by Result 2. So s is continuous everywhere, including along the whole x-axis where the formula \displaystyle \frac{\sin(xy)}{y} means nothing. In particular \lim_{(x, y) \to (3, 0)} s(x, y) = s(3, 0) = 3.
Boundaries
No list of paths proves a limit. Example 5 and Exercise 10 pass along every line and still fail. Only an estimate that covers all nearby points at once proves that a limit exists.
A polar estimate must be free of \theta. Convergence for each fixed \theta is convergence along rays. Result 4 needs one g(r) for all directions.
Continuity in each variable separately is weaker than continuity (Exercise 16).
Limits are taken inside the domain, at limit points only. \displaystyle f(x, y) = \sqrt{-x^2 - y^2} has domain \{(0, 0)\}. The origin is not a limit point of it, so no limit is defined there, yet f is continuous at the origin, since every \delta works.
Where this goes next
The Partial derivatives are one-variable limits along the two lines through a point parallel to the axes, so, as Example 5 warns, they can exist at a point where f is not continuous. The definition of differentiability in Tangent planes and linear approximation is a limit in two variables, proved exactly as here, with a bound by the distance. A function continuous on a closed and bounded set has a largest and a smallest value, the fact behind Extrema and the second derivative test; it is proved in Real Analysis.
Exercises
Limits from continuity
- Find \lim_{(x, y) \to (1, -1)} \dfrac{x^2 + 3xy}{1 + y^2}.
- Find \lim_{(x, y, z) \to (2, 0, 1)} \left(z\,e^{xy} + \ln(x + z)\right).
- Find \displaystyle \lim_{(x, y) \to (4, 4)} \dfrac{x - y}{\sqrt{x} - \sqrt{y}}.
- Find \displaystyle \lim_{(x, y) \to (0, 0)} \dfrac{x^2 + y^2}{\sqrt{x^2 + y^2 + 9} - 3}.
Limits that exist
- Prove that \lim_{(x, y) \to (0, 0)} \dfrac{5x^4 - y^4}{x^2 + y^2} = 0, and give a \delta for each \varepsilon.
- Find \lim_{(x, y) \to (0, 0)} \dfrac{x^2 \sin y}{x^2 + 2y^2}.
- Find \lim_{(x, y) \to (0, 0)} (x^2 + y^2)\ln(x^2 + y^2).
- Find \lim_{(x, y) \to (1, 0)} \dfrac{(x - 1)^2 y}{(x - 1)^2 + y^2}.
Limits that do not exist
- Show that \dfrac{xy - y^2}{x^2 + y^2} has no limit at (0, 0).
- Show that \dfrac{x^3 y}{x^6 + 2y^2} tends to 0 along every line through the origin but has no limit there.
- Show that \dfrac{xy + yz + zx}{x^2 + y^2 + z^2} has no limit at (0, 0, 0).
- On the trawler’s screen, Isaac plots the north component \displaystyle v(x, y) = \dfrac{y}{\sqrt{x^2 + y^2}} and Jasmine plots w(x, y) = x\,v(x, y). Which of the two has a limit at the sonar?
Continuity
- Find the set on which f(x, y) = \dfrac{\ln(x - y)}{x^2 + y^2 - 4} is continuous.
- Let g(x, y) = \dfrac{e^{x - y} - 1}{x - y} for x \ne y and g(x, y) = c on the line y = x. Find every c that makes g continuous on \mathbb{R}^2.
- Let f(x, y) = \dfrac{x^3 + y^3}{x + y}. Find its limits at (1, -1) and at (0, 0), and say whether f extends to a function continuous on \mathbb{R}^2.
- Refute: a function of two variables that is continuous in each variable separately is continuous. Use g(x, y) = \dfrac{xy}{x^2 + 4y^2} with g(0, 0) = 0.
Proofs
- Prove from the definition that \lim_{(x, y) \to (2, 1)} \dfrac{1}{x + y} = \dfrac13.
- Prove the sum law: if f \to L and g \to M at (a, b), then f + g \to L + M.
- Prove that if f is continuous at (a, b) and f(a, b) > 0, then some disk B_\delta(a, b) has f > 0 at all its points in the domain of f.
- Let p, q > 0 and f(x, y) = \dfrac{\lvert x \rvert^p \lvert y \rvert^q}{x^2 + y^2}. Prove that the limit at (0, 0) is 0 when p + q > 2 and does not exist when p + q \le 2.
- Prove that \mathbf{x} \mapsto \lVert \mathbf{x} \rVert is continuous on \mathbb{R}^n.
Answers
- -1. A rational function with denominator 2 \ne 0 at the point: \displaystyle \frac{1 - 3}{2} = -1.
- 1 + \ln 3. The function is continuous at the point (Result 3), with value 1 \cdot e^0 + \ln 3.
- 4. On the domain (x, y \ge 0, x \ne y), \displaystyle x - y = (\sqrt{x} - \sqrt{y})(\sqrt{x} + \sqrt{y}), so the function is \displaystyle \sqrt{x} + \sqrt{y} \to 2 + 2.
- 6. With s = x^2 + y^2 > 0, multiplying by \displaystyle \sqrt{s + 9} + 3 gives \displaystyle \frac{s(\sqrt{s + 9} + 3)}{s} = \sqrt{s + 9} + 3 \to 3 + 3.
- \lvert 5x^4 - y^4 \rvert \le 5x^4 + y^4 \le 5(x^2 + y^2)^2, so \lvert f \rvert \le 5r^2. Result 4 gives 0; \displaystyle \delta = \sqrt{\varepsilon / 5} works.
- 0. Since \displaystyle \frac{x^2}{x^2 + 2y^2} \le 1 and \lvert \sin y \rvert \le \lvert y \rvert \le r, \lvert f \rvert \le r; \delta = \varepsilon.
- 0. Let h(s) = s\ln s for s > 0 and h(0) = 0. By l’Hôpital’s rule \displaystyle \lim_{s \to 0^+} \frac{\ln s}{1/s} = \lim_{s \to 0^+} (-s) = 0, so h is continuous at 0 on [0, \infty), and x^2 + y^2 \to 0: Result 3.
- 0. With s = x - 1, \displaystyle \frac{s^2}{s^2 + y^2} \le 1, so \lvert f \rvert \le \lvert y \rvert \le r, where r is the distance to (1, 0).
- Along the x-axis f = 0; along y = -x, \displaystyle f = \frac{-x^2 - x^2}{2x^2} = -1. Two paths, two limits.
- Along y = mx, m \ne 0: \displaystyle \frac{mx^4}{x^6 + 2m^2x^2} = \frac{mx^2}{x^4 + 2m^2} \to 0; on the axes f = 0. Along y = x^3: \displaystyle \frac{x^6}{x^6 + 2x^6} = \frac13. No limit.
- Along the x-axis f = 0; along x = y = z = t, \displaystyle f = \frac{3t^2}{3t^2} = 1. No limit.
- Only w, with limit 0. v(tx, ty) = v(x, y) for t > 0 with v(0, 1) = 1 and v(1, 0) = 0, so Example 4 applies. For w, \lvert y \rvert \le r gives \displaystyle \lvert w \rvert = \frac{\lvert x \rvert \lvert y \rvert}{r} \le \lvert x \rvert \le r.
- \{(x, y) : x > y,\ x^2 + y^2 \ne 4\}: \ln(x - y) is continuous where x - y > 0 (Result 3), and the quotient where the denominator is not 0, that is, off the circle of radius 2.
- c = 1 only. Let \displaystyle \varphi(w) = \frac{e^w - 1}{w}, \varphi(0) = 1; it is continuous since \displaystyle \lim_{w \to 0} \frac{e^w - 1}{w} = 1, the derivative of e^w at 0 (The derivative at a point). With c = 1, g = \varphi(x - y) everywhere, continuous by Result 3. If c \ne 1, along (a + t, a), \displaystyle g = \frac{e^t - 1}{t} \to 1 \ne c = g(a, a).
- 3 and 0; yes. On the domain x \ne -y, x^3 + y^3 = (x + y)(x^2 - xy + y^2), so f = x^2 - xy + y^2, which gives 1 + 1 + 1 = 3 at (1, -1) and 0 at (0, 0). The polynomial x^2 - xy + y^2 is the continuous extension.
- g(x, 0) = 0 and g(0, y) = 0 for all x and y, continuous functions of one variable. But \displaystyle g(t, t) = \frac{t^2}{5t^2} = \frac15 for t \ne 0, which does not tend to g(0, 0) = 0, so g is not continuous at the origin.
- Let r be the distance to (2, 1); then \lvert x + y - 3 \rvert \le \lvert x - 2 \rvert + \lvert y - 1 \rvert \le 2r. If r < 1, then x > 1 and y > 0, so x + y > 1, and \displaystyle \left\lvert \frac{1}{x + y} - \frac13 \right\rvert = \frac{\lvert x + y - 3 \rvert}{3(x + y)} < \frac{2r}{3} < r. So \delta = \min(1, \varepsilon) works.
- Let \varepsilon > 0. Take \delta_1 with \displaystyle \lvert f - L \rvert < \frac{\varepsilon}{2} and \delta_2 with \displaystyle \lvert g - M \rvert < \frac{\varepsilon}{2} at points of the common domain with 0 < r < \delta_i, and let \delta = \min(\delta_1, \delta_2). For 0 < r < \delta, \lvert (f + g) - (L + M) \rvert \le \lvert f - L \rvert + \lvert g - M \rvert < \varepsilon.
- Take \varepsilon = f(a, b) > 0 in the definition of continuity. There is \delta > 0 with \lvert f(x, y) - f(a, b) \rvert < f(a, b) at every point of the domain in B_\delta(a, b), and there f(x, y) > f(a, b) - f(a, b) = 0.
- \lvert x \rvert^p \lvert y \rvert^q \le r^p r^q, so f \le r^{p + q - 2}, which tends to 0 when p + q > 2 (Result 4). On the x-axis f = 0. On y = x, \displaystyle f = \frac{\lvert x \rvert^{p + q}}{2x^2} = \frac12 \lvert x \rvert^{p + q - 2}, which is \displaystyle \frac12 when p + q = 2 and tends to \infty when p + q < 2. Either way the two paths disagree, so by Result 5 there is no limit.
- By the triangle inequality, \lVert \mathbf{x} \rVert \le \lVert \mathbf{x} - \mathbf{a} \rVert + \lVert \mathbf{a} \rVert and \lVert \mathbf{a} \rVert \le \lVert \mathbf{a} - \mathbf{x} \rVert + \lVert \mathbf{x} \rVert, so \big\lvert \lVert \mathbf{x} \rVert - \lVert \mathbf{a} \rVert \big\rvert \le \lVert \mathbf{x} - \mathbf{a} \rVert. Given \varepsilon, \delta = \varepsilon works at every \mathbf{a}.