A multivariable limit compares all sufficiently close domain points with one target value. A sequence tests the same requirement by choosing those points one at a time. The sequential criterion makes this connection precise, including the reverse direction: if one radius cannot control all nearby inputs, a sequence of failures can be constructed.
This section revisits sequences and multivariable limits, emphasizing the quantifiers in the relative-domain sequential criterion.
Two worked applications follow this section: coordinate tails and one radius and domain and admissible sequences.
One tail must control all coordinates
Definition 5.1.1
(Convergence of a sequence in )
A sequence tends to if, for every , there is an integer such that
The condition concerns every term after that index, not merely infinitely many good terms. For example, a sequence with two subsequences approaching distinct points does not converge.
Proposition 5.1.2
(Coordinate convergence)
A sequence converges to if and only if each coordinate sequence converges to .
Proof
(Coordinate errors against the norm)
Coordinate convergence is equivalent to norm convergence. For each coordinate,
Conversely, if every coordinate converges, choose a tail on which each coordinate error is smaller than . Take the largest of the finitely many required indices. Then
The same tail controls every coordinate. Taking a different index after seeing which coordinate or term is being tested would not establish this statement.
The domain belongs in the definition
Definition 5.1.3
(Accumulation point and relative-domain limit)
Let . Assume is an accumulation point of : every ball centered at contains a point of different from . The relative-domain limit is
The center is excluded. The function need not be defined at that center, and changing its value there cannot change the punctured limit. Continuity at a point in the domain adds the requirement that the limit agree with the actual function value.
The accumulation hypothesis prevents a vacuous test. At an isolated domain point a small enough punctured ball has no admissible inputs. The displayed implication would then hold for every target, so we do not use it to assign a unique punctured limit there. Continuity at an isolated domain point is still meaningful and automatic: a sufficiently small relative neighborhood contains only the point itself.
The input radius can depend on the output tolerance and the fixed point. It must work for every nearby admissible input. A radius chosen separately for each direction does not provide that common control.
The exact negation produces a bad sequence
Failure of the proposed limit means
There is one fixed positive output error. A bad input exists in every punctured radius, but that input may depend on the radius.
For each positive integer , take radius and choose one such bad point . Then
The inputs tend to , while the outputs stay a fixed distance from the proposed target. The sequence need not have a simple formula; its existence follows from the failed quantified statement.
Sequential criterion and uniqueness
Theorem 5.1.4
(Sequential criterion)
Under the accumulation hypothesis, the relative-domain limit equals if and only if every sequence with satisfies .
Proof
(Both directions of the sequential criterion)
For the forward implication, fix an output tolerance and choose the limit's radius. The input sequence eventually lies inside that punctured radius, so its outputs eventually satisfy the tolerance.
For the reverse implication, suppose the sequential condition holds but the function limit fails. The section on the exact negation constructs an admissible input sequence whose outputs stay at least from . This contradicts the sequential condition and proves the limit.
Proof X-ray
Where the quantifiers sit in the criterion
The forward direction chains two thresholds: the limit supplies a radius for the tolerance, and convergence of the inputs supplies an index after which the inputs lie inside that radius. The reverse direction is a contrapositive whose key point is the order of quantifiers in the negation: is fixed first and does not depend on the radius, so choosing one bad point for each radius yields a single sequence with a uniform output gap. Had depended on , the outputs could still converge. The accumulation hypothesis is not used in these two directions; it is what makes the criterion non-vacuous, since at an isolated point there are no punctured sequences at all.
The universal word “every” is essential. One good sequence does not prove a function limit. Two admissible sequences with different output limits disprove existence of any common finite limit.
Theorem 5.1.5
(Uniqueness of the limit)
If is an accumulation point of , a limit is unique.
Proof
(Two limits cannot differ)
Limits are unique. If two distinct vectors were limits, use tolerance for both and take the smaller of their radii. The accumulation hypothesis supplies an admissible point in this common punctured ball. Then
a contradiction. The hypothesis is used to obtain a point, not merely to manipulate inequalities.
Counterexample mode
Without an accumulation point every vector is a limit
Let , and . The ball of radius about contains no point of other than , so is isolated. The condition “ implies ” is then vacuous, and it holds for and for alike, with . Two different values both satisfy the definition, so uniqueness fails, and the sequential condition is satisfied by both as well because no punctured sequence exists. The accumulation hypothesis is what prevents this.
Turn a norm estimate into an actual radius
Proposition 5.1.6
(Power estimate gives a radius)
If for and , with , then , and is a valid radius for each .
Proof
(From a power estimate to a radius)
Suppose constants , and satisfy
Then works. Every admissible input within it has output error strictly below . This proves existence directly and, through the criterion, controls all admissible sequences at once.
For nonlinear expressions it is often useful first to restrict the distance below one so a varying coefficient is bounded, then impose the desired tolerance. The radius must satisfy both requirements. The worked example at a nonzero point supplies a concrete calculation.
A finite limit also forces local boundedness: choose output tolerance one to obtain sufficiently close to the point. An unbounded output sequence approaching the point therefore excludes any finite limit.
Worked examples
Worked example 5.1.7
(A radius from a power estimate)
Let for . Show at the origin and give .
Since , . So the estimate holds with , and any , and works: if then .
Worked example 5.1.8
(A radius at a nonzero point)
Show that as .
Put , , , so . Then . Using , we get for : . Hence works. The cap is needed so that and .
Worked example 5.1.9
(Bounded output but no limit)
Let on at .
The function is bounded by . The sequences and lie in for and tend to . But while . The outputs have different limits, so by the criterion no limit exists. Local boundedness is necessary for a finite limit but is not sufficient.
Common mistakes
Common mistake
(Treating one good sequence as a proof)
Checking that shows only that the limit, if it exists, can only be . The function limit needs every admissible sequence, or an estimate that controls all inputs near the point at once.
Common mistake
(Choosing the radius separately in each direction)
A radius that works along each line through but shrinks with the direction does not give one for all nearby inputs. The definition requires a single for a given .
Checklist
- Check that the point is an accumulation point of the actual domain.
- Exclude the center in a punctured limit and retain the domain restriction.
- To prove convergence, supply one estimate or a statement about every admissible sequence.
- To disprove a target, keep one fixed positive output gap while the input radius tends to zero.
- Do not test along deleted lines, or call a direction with no approaching domain points a successful limit test.
- Distinguish a punctured limit, continuity at the center and boundedness near the center.
Quick checks
Quick check
(Does converge in , and to what?)
Check each coordinate.
Solution · Answer
Both coordinates tend to because . By coordinate convergence .
Quick check
(Write the negation of as a quantified statement.)
Swap each quantifier and negate the implication.
Solution · Answer
There exists such that for every there is with and .
Quick check
(For , which guarantees near the origin?)
Use .
Solution · Answer
: if then .