How this calculation works
The explorer samples six straight or curved paths from both parameter directions. Different sampled trends can expose a path-dependent limit, while agreement among a finite list cannot establish a general limit.
A multivariable limit must agree along every valid approach pathA worked example
For x*y/(x^2+y^2) at (0,0), the horizontal path gives zero and the diagonal y=x gives one-half, so the actual path limits disagree.
Good to know
Bounded real expressions and floating-point arithmetic. Finite path and surface samples do not prove arbitrary limits or differentiability.
A limit must survive every path
A two-input point can be approached along horizontal and vertical lines, diagonals, parabolas and infinitely many other curves. A limit exists only when the function approaches the same value along all valid domain paths toward that point.
To disprove a limit, establish two actual path limits that differ. For xy/(x²+y²), substituting y=0 gives zero for nonzero x, while substituting y=x gives one-half. This algebraic argument establishes nonexistence; the numerical table illustrates it.
Straight lines can miss a curved-path failure
Consider x²y/(x⁴+y²) near the origin. Many straight-line approaches tend to zero, but the parabola y=x² gives one-half. Checking only coordinate axes would therefore suggest the wrong conclusion.
The explorer includes y−b=(x−a)² and its exchanged version alongside four straight paths, with positive and negative parameter values. These extra examples improve inspection but still represent only a finite collection. Other curves can expose behavior the tool misses.
Separate the center from the approaching values
The assigned value at the target can be undefined even when a limit exists. Conversely, a formula that returns a finite center value is not automatically continuous there. Keep substitution, domain conditions and nearby behavior as separate questions.
For ordinary compositions, a conservative continuity-rule assessment can support direct substitution. For an indeterminate quotient, examine cancellations or a suitable inequality. A polar-coordinate argument can establish a limit when its bound tends to zero uniformly in angle; testing a few angles does not establish that uniform condition.
Use the samples as evidence with a scale
Each path is sampled at seven shrinking offsets scaled by the initial probe width. Very small offsets can magnify floating-point cancellation, while very large ones may not show local behavior. Compare the table with an algebraic derivation before trusting an apparent pattern.
The partial derivative and gradient calculator examines coordinate slopes at a point, which is a different question from continuity along every path. Its derivative expressions and this explorer’s rotatable surface are learning aids; neither turns a finite numerical experiment into an epsilon-delta proof.
A few questions, answered
Does agreement mean the limit exists?
No. Untested paths or oscillations can behave differently.
Why include parabolas?
Some expressions agree on straight lines but differ along a curved path.
Does undefined at the center mean no limit?
No. Limits concern nearby domain values, not only the assigned center value.