My recent work in mathematical analysis has moved from asking why the real line has no gaps to asking how “approaching a value” can be stated rigorously. My previous entry ended with density, bounds, and suprema. The next lectures introduced limits of sequences, where I had to replace an intuitive phrase such as “gets closer and closer” with a complete epsilon-N argument.
The recent line of study
The topics initially looked separate, but they now form a sequence. Density says that numbers can still be found inside every interval. The supremum principle expresses the completeness of the real numbers. Limits of sequences then ask whether the tail of a sequence eventually remains inside every prescribed error band.
I also worked through several smaller proofs: using contradiction to classify sums, differences, products, and quotients involving rational and irrational numbers; bounding |f(x)-f(y)| with the supremum and infimum of a function's range; and writing the quantifiers for both convergence and nonconvergence. The calculations were usually short. The difficult parts were keeping every condition, set notation, and intermediate inequality explicit.
Starting from the epsilon-N definition
A sequence aₙ converges to a if, for every ε>0, there is a positive integer N such that every n>N satisfies |aₙ-a|<ε.
My current interpretation is that someone else may choose ε as small as they want, while the proof must respond with a suitable N. After that point, every remaining term must stay inside the error band. Saying that the terms “get closer” is not enough, and finding only some close terms is not enough. The proof has to choose N from ε and then verify the choice.
Where the Bernoulli inequality project stalled
The project problem was to prove from the definition that qⁿ→0 whenever |q|<1.
My first approach used logarithms. The target is |q|ⁿ<ε. When 0<|q|<1, this can be rearranged into a lower bound on n, after which a larger positive integer N can be selected. This approach works, but q=0 must be handled separately, and the inequality reverses because the logarithm has a base between 0 and 1.
I then tried to apply Bernoulli's inequality directly to |q|ⁿ and obtained only a lower bound. That revealed the real issue: proving |q|ⁿ<ε requires an upper bound that tends to zero. A lower bound does not move the proof toward its target.
Why taking the reciprocal helps
For 0<|q|<1, set
h = 1/|q| - 1 > 0, so that 1/|q| = 1+h.
Bernoulli's inequality can now be applied to the reciprocal:
(1/|q|)ⁿ = (1+h)ⁿ ≥ 1+nh > nh.
All quantities are positive, so taking reciprocals reverses the inequalities:
|q|ⁿ ≤ 1/(1+nh) < 1/(nh).
It is then enough to make 1/(nh)<ε. Choosing
N = ⌊1/(hε)⌋ + 1
ensures that n≥N gives |qⁿ-0|=|q|ⁿ<ε. This completes the full path from an arbitrary ε to a verified choice of N.
The step 1/(1+nh)<1/(nh) is not essential. One could work directly with 1/(1+nh)<ε, but the resulting condition is less convenient. The estimate replaces the original expression with a slightly larger but simpler upper bound. If even that larger quantity is below ε, then |q|ⁿ must be below ε as well.
What I am taking from the project
The project ended with a complete handwritten proof and a PDF containing the full path from understanding the problem through checking the logarithmic approach, getting stuck with Bernoulli's inequality, switching to the reciprocal, and explaining the estimate. The useful result was not only the final proof, but also a checking routine: write the target first, decide whether an upper or lower bound is needed, inspect the object to which an inequality is applied, and substitute the chosen N back into the argument.
I can now follow and explain this proof, but I would not describe the method as mastered. I still need practice recognizing on my own that a quantity below 1 may be easier to study through its reciprocal, and choosing among several valid bounds the one that makes N easiest to construct.
Comments
0 comments