Simulate
Check the claim with numbers
Lesson 2.1.1 found by substituting. Test it the way lesson PG4.1 tested sequences, with a tolerance for the output. Then look for a window of half-width around 3 that keeps every output within of 10.
The cell takes to be the smaller of 1 and , a choice the proof will explain. It tries 1,998 inputs inside the window, all but 3 itself, and reports the worst distance from 10.
Each worst distance is below its : 0.876653 for , and 0.008565 for . The last column is the widest window that would work. The proof's is a little smaller, and that is fine.
A function with a jump fails the test. Near 0, is on the left and 1 on the right (lesson 2.1.2).
However small the window, it holds inputs with outputs and 1. With no is within 0.5 of both, since they are 2 apart.
Conjecture
Three tolerances worked with the same rule for . A proof must show that the rule works for every , and for every input in the window, not just 1,998 of them.
The definition
The definition has the same shape as for sequences, with the same tolerance . The window (the Greek letter delta) takes the place of , because must be close to rather than far along.
Rule: Limit of a function
Only inputs in the domain of count. The condition leaves out itself: as in lesson 2.1.1, the value at plays no part.
In words: however narrow the band around , some window around sends every input in it, apart from , into the band. The window may depend on , but not on . It is chosen first, and then every in it must work.
Bound the extra factor
For the conjecture, . The window controls , but the factor needs a bound of its own.
So agree first that . Then means (lesson 0.3.3), so and .
Now , and when . Both demands hold for , the smaller of the two. The cap at 1 is a free choice: another cap works too, with its own bound in place of 7.
Format 1
Reorder the steps
Here is the ε–δ proof of the conjecture, with its lines shuffled. Put them in order with the arrows, or drag a line by its handle.
- Since , this is less than .
- Since , , so and .
- So every such has . As was arbitrary, the limit is 10.
- Then , because .
- Take any with .
- Let , and set , which is positive.
A bound that needs no cap
Lesson 2.1.1 handled near 4 by multiplying by . The same trick gives a for , and this time the bound holds on the whole domain, .
Try it
The bound is . Near 4 it is about twice the true distance: 0.025158 against 0.05 at .
With , the worst distances in the window are 0.05059 for and 0.005001 for . Each is below its .
Format 2
Fill the gap
The proof that , with its key steps left out. Type each missing piece in , or as a number, then check.
For , , so .
The bottom, , is never less than , because . So for every .
Let , and set . For every with , . So .
For example, with the proof takes .
How to type answers
- Powers with
^, brackets round a power that is a sum:r^k,r^(k+1). - Fractions with
/and brackets:(1-r)/(1+r). Multiply with*or side by side:2k+1. - Press Enter in any box to check them all.
Format 3
Find the error
A proof that . One line breaks the rules of the definition. Select it.
Summary card
- : for every there is a such that implies , for in the domain of .
- may depend on , never on : it is chosen before .
- An extra factor: cap , bound the factor, then shrink . For at 3, .
- A bound that holds on the whole domain needs no cap: gives .
- A jump defeats every : at 0, with .