A.

ALTERNATIVE DEFINITION
Let 𝑆 ⊆ ℝ and 𝛼 ∈ 𝑆 with 𝛼 a limit point of 𝑆. Let 𝑓: 𝑆 → ℝ. We say that 𝑓 is 𝐷1 at 𝛼 if there
is a number 𝐴 such that for all 𝜖 > 0 there is 𝛿 > 0 such that if 𝑥 ∈ 𝑆 and |𝑥 − 𝛼| < 𝛿, then
|𝑓(𝑥) − 𝑓(𝛼) − 𝐴 ∙ (𝑥 − 𝛼)| ≤ 𝜖 ∙ |𝑥 − 𝛼|
It turns out that 𝐷1 is the same as differentiable, but do not assume that – just use the definition
of 𝐷1 as is.
Be able to prove the following.

(a) The function 𝑥 2 is 𝐷1 at 𝛼, for every 𝛼 ∈ ℝ. (Hint: let 𝐴 = 2 ∙ 𝛼.)

Let 𝜖 > 0, 𝛼 ∈ ℝ, and 𝐴 = 2 ∙ 𝛼

Let 𝛿 = 𝜖, then if 𝑥 ∈ 𝑆 and |𝑥 − 𝛼| < 𝛿, we have:
|𝑥 2 − 𝛼 2 − 2 ∙ 𝛼 ∙ (𝑥 − 𝛼)| = |𝑥 2 − 2 ∙ 𝛼 ∙ 𝑥 + 𝛼 2 | = |(𝑥 − 𝛼)2 | < 𝜖 ∙ |𝑥 − 𝛼|
Thus, 𝑥 2 is 𝐷1 at 𝛼, for every 𝛼 ∈ ℝ

(b) If 𝑓 is 𝐷1 at 𝛼, then the number 𝐴 involved is unique.

Suppose 𝑓 is 𝐷1 at 𝛼, and that there are 𝐴 and 𝐵 with 𝐴 ≠ 𝐵 that can fit in the inequality
Thus, for all 𝜖 > 0 there is 𝛿 > 0 such that if 𝑥 ∈ 𝑆 and |𝑥 − 𝛼| < 𝛿, then
|𝑓(𝑥) − 𝑓(𝛼) − 𝐴 ∙ (𝑥 − 𝛼)| ≤ 𝜖 ∙ |𝑥 − 𝛼| and |𝑓(𝑥) − 𝑓(𝛼) − 𝐵 ∙ (𝑥 − 𝛼)| ≤ 𝜖 ∙ |𝑥 − 𝛼|
We have:
|𝐴 − 𝐵| ∙ |𝑥 − 𝛼| = |(𝐴 − 𝐵) ∙ (𝑥 − 𝛼)| = |𝐴 ∙ (𝑥 − 𝛼) − 𝐵 ∙ (𝑥 − 𝛼)|
= |𝐴 ∙ (𝑥 − 𝛼) − 𝑓(𝑥) + 𝑓(𝛼) + 𝑓(𝑥) − 𝑓(𝛼) − 𝐵 ∙ (𝑥 − 𝛼)|
≤ |𝐴 ∙ (𝑥 − 𝛼) − 𝑓(𝑥) + 𝑓(𝛼)| + |𝑓(𝑥) − 𝑓(𝛼) − 𝐵 ∙ (𝑥 − 𝛼)|
≤ 𝜖 ∙ |𝑥 − 𝛼| + 𝜖 ∙ |𝑥 − 𝛼| = 2 ∙ 𝜖 ∙ |𝑥 − 𝛼|
Since 𝛼 ∈ 𝑆 and 𝛼 is a limit point of 𝑆, there’s 𝑥 ∈ (𝛼 − 𝛿, 𝛼 + 𝛿) with 𝑥 ≠ 𝛼, or |𝑥 − 𝛼| ≠ 0
Thus, |𝐴 − 𝐵| ≤ 2 ∙ 𝜖 arbitrary positive, i.e., 𝐴 = 𝐵

(c) If 𝑓: 𝑆 → 𝑅 and 𝑔: 𝑆 → 𝑅 and if both 𝑓 and 𝑔 are 𝐷1 at 𝛼 ∈ 𝑆, then 𝑓 + 𝑔 is 𝐷1 at 𝛼.

Let 𝜖 > 0
Since 𝑓 and 𝑔 are 𝐷1 at 𝛼, there is 𝐴 and 𝐵 and 𝛿 > 0 such that if 𝑥 ∈ 𝑆 and |𝑥 − 𝛼| < 𝛿, then
|𝑓(𝑥) − 𝑓(𝛼) − 𝐴 ∙ (𝑥 − 𝛼)| ≤ 𝜖 ∙ |𝑥 − 𝛼|
and |𝑔(𝑥) − 𝑔(𝛼) − 𝐵 ∙ (𝑥 − 𝛼)| ≤ 𝜖 ∙ |𝑥 − 𝛼|
For these 𝑥:

|(𝑓(𝑥) + 𝑔(𝑥)) − (𝑓(𝛼) + 𝑔(𝛼)) − (𝐴 + 𝐵) ∙ (𝑥 − 𝛼)|

 = |(𝑓(𝑥) − 𝑓(𝛼) − 𝐴 ∙ (𝑥 − 𝛼)) + (𝑔(𝑥) − 𝑔(𝛼) − 𝐵 ∙ (𝑥 − 𝛼))|

≤ |𝑓(𝑥) − 𝑓(𝛼) − 𝐴 ∙ (𝑥 − 𝛼)| + |𝑔(𝑥) − 𝑔(𝛼) − 𝐵 ∙ (𝑥 − 𝛼)|

≤ 𝜖 ∙ |𝑥 − 𝛼| + 𝜖 ∙ |𝑥 − 𝛼| = 2 ∙ 𝜖 ∙ |𝑥 − 𝛼|
Thus, 𝑓 + 𝑔 is 𝐷1 at 𝛼

B. PROPOSITION 4.6
Proposition 4.6. Let 𝑓: [𝑎, 𝑏] → ℝ be continuous. Then 𝑓(𝑥) is bounded.
Proof.
Suppose that 𝑓(𝑥) is not bounded
Then for each positive integer 𝑛, there is 𝑢[𝑛] ∈ [𝑎, 𝑏] such that |𝑓(𝑢[𝑛])| > 𝑛
By Bolzano-Weierstrass for sequences, since the sequence 𝑢[𝑛] is bounded, there is a
converging sequence 𝑢[𝑗𝑛 ] → 𝑐 as 𝑛 → ∞
Since 𝑢[𝑛] ∈ [𝑎, 𝑏] for all 𝑛, we have 𝑐 ∈ [𝑎, 𝑏]
By Archimedes, there is 𝑚 ∈ ℤ with 𝑚 > |𝑓(𝑐)|
Since 𝑓 is continuous at 𝑐 and 𝑚 > |𝑓(𝑐)|, there is 𝛿 > 0 such that if |𝑥 − 𝑐| < 𝛿 then
|𝑓(𝑥)| < 𝑚
But since 𝑢[𝑗𝑛 ] → 𝑐 as 𝑛 → ∞, there’s a positive integer 𝑁 such that if 𝑛 ≥ 𝑁, then
|𝑢[𝑗𝑛 ] − 𝑐| < 𝛿; thus, |𝑓(𝑢[𝑗𝑛 ])| < 𝑚
However, for 𝑛 ≥ 𝑁, by the construction of 𝑢[𝑛], we have |𝑓(𝑢[𝑗𝑛 ])| > 𝑗𝑛
Since the 𝑗𝑛 strictly increase, eventually 𝑗𝑛 > 𝑚 for some 𝑛 →←
Therefore, 𝑓(𝑥) is bounded
 C. PROOFS
Proposition 4.4. If 𝑆, 𝑇 ⊆ ℝ and 𝑓: 𝑆 → 𝑇 and 𝑔: 𝑇 → ℝ and if 𝛼 ∈ 𝑆 and 𝑓 is continuous at
𝛼 and 𝑔 is continuous at 𝑓(𝛼), then 𝑔(𝑓(𝑥)) is continuous at 𝛼.

Let 𝜖 > 0
Since 𝑔 is continuous at 𝑓(𝛼), there is 𝛿 > 0 such that if 𝑦 ∈ 𝑇 and |𝑦 − 𝑓(𝛼)| < 𝛿, then
|𝑔(𝑦) − 𝑔(𝑓(𝛼))| < 𝜖

Since 𝑓 is continuous at 𝛼, there is 𝜇 > 0 such that if 𝑥 ∈ 𝑆 and |𝑥 − 𝛼| < 𝜇, then

|𝑓(𝑥) − 𝑓(𝛼)| < 𝛿
Let 𝑥 ∈ 𝑆 and |𝑥 − 𝛼| < 𝜇
Then |𝑓(𝑥) − 𝑓(𝛼)| < 𝛿, and 𝑓(𝑥) ∈ 𝑓(𝑆) ⊆ 𝑇

Thus, |𝑔(𝑓(𝑥)) − 𝑔(𝑓(𝛼))| < 𝜖

Proposition 5.2. Let 𝑆 ⊆ ℝ and 𝛼 ∈ 𝑆 be a limit point of 𝑆. Let 𝑓: 𝑆 → ℝ. If 𝑓(𝑥) is

differentiable at 𝛼, then 𝑓(𝑥) is continuous at 𝛼.

Let 𝐹(𝑥) be a secant function for 𝑓(𝑥) at 𝛼

Since 𝐹 is continuous at 𝛼, there is 𝛿 > 0 such that if 𝑥 ∈ 𝑆 and |𝑥 − 𝛼| < 𝛿, then
|𝐹(𝑥) − 𝐹(𝛼)| < 1
For these 𝑥:
|𝐹(𝑥)| − |𝐹(𝛼)| ≤ |𝐹(𝑥) − 𝐹(𝛼)| < 1 ⇒ |𝐹(𝑥)| < |𝐹(𝛼)| + 1
Let 𝜖 > 0, and find 𝜇 > 0 less than 𝜖 and less than 𝛿
If 𝑥 ∈ 𝑆 and |𝑥 − 𝛼| < 𝜇, then also |𝑥 − 𝛼| < 𝛿, and we have:
|𝑓(𝑥) − 𝑓(𝛼)| = |𝐹(𝑥) ∙ (𝑥 − 𝛼)| ≤ (|𝐹(𝛼)| + 1) ∙ 𝜇 < (|𝐹(𝛼)| + 1) ∙ 𝜖
This number is arbitrarily small, and so 𝑓 is continuous at 𝛼

Chain Rule. If 𝑆, 𝑇 ⊆ ℝ and 𝑓: 𝑆 → 𝑇 and 𝑔: 𝑇 → ℝ and if 𝛼 ∈ 𝑆 and 𝑓 is differentiable at 𝛼

and 𝑔 is differentiable at 𝑓(𝛼), then 𝑔(𝑓(𝑥)) is differentiable at 𝛼, and
′
[𝑔(𝑓(𝑥))] | = 𝑔′ (𝑓(𝛼)) ∙ 𝑓 ′ (𝛼)
𝑥=𝛼

Let 𝐹 be a secant function for 𝑓 at 𝛼, and let 𝐺 be a secant function for 𝑔 at 𝑓(𝛼)

We claim that 𝐺(𝑓(𝑥)) ∙ 𝐹(𝑥) is a secant function for 𝑔(𝑓(𝑥)) at 𝛼

Indeed, for all 𝑥 ∈ 𝑆:

𝑔(𝑓(𝑥)) − 𝑔(𝑓(𝛼)) = 𝐺(𝑓(𝑥)) ∙ (𝑓(𝑥) − 𝑓(𝛼)) = 𝐺(𝑓(𝑥)) ∙ 𝐹(𝑥) ∙ (𝑥 − 𝛼)

Furthermore, 𝐹 is continuous at 𝛼, and 𝑓 is continuous at 𝛼 and 𝐺 is continuous at 𝑓(𝛼); thus,

𝐺(𝑓(𝑥)) ∙ 𝐹(𝑥) is continuous at 𝛼
This proves that 𝑔(𝑓(𝑥)) is differentiable at 𝛼

Then:
′
[𝑔(𝑓(𝑥))] | = 𝐺(𝑓(𝛼)) ∙ 𝐹(𝛼) = 𝑔′ (𝑓(𝛼)) ∙ 𝑓 ′ (𝛼)
𝑥=𝛼

Theorem on Interior Extremes. Let 𝐼 be an open interval and 𝑓: 𝐼 → ℝ. Suppose that 𝛼 ∈ 𝐼

and that 𝑓(𝛼) is the maximum of 𝑓(𝑥) on 𝐼. Suppose that 𝑓 ′ (𝛼) exists. Then 𝑓 ′ (𝛼) = 0.

Let 𝐹 be the secant function for 𝑓 at 𝛼

Suppose 𝑓 ′ (𝛼) < 0, so that 𝐹(𝛼) < 0
Since 𝐹 is continuous at 𝛼, get 𝛿 > 0 such that if 𝑥 ∈ 𝐼 and |𝑥 − 𝛼| < 𝛿, then 𝐹(𝑥) < 0
Since 𝐼 is an open interval, there is 𝛼1 ∈ 𝐼 such that 𝛼 − 𝛿 < 𝛼1 < 𝛼
Then:
𝑓(𝛼1 ) − 𝑓(𝛼) = 𝐹(𝛼1 ) ∙ (𝛼1 − 𝛼) > 0 ⇒ 𝑓(𝛼1 ) > 𝑓(𝛼) →←
Thus, 𝑓 ′ (𝛼) ≥ 0
Suppose 𝑓 ′ (𝛼) > 0, so that 𝐹(𝛼) > 0
Since 𝐹 is continuous at 𝛼, get 𝜇 > 0 such that if 𝑥 ∈ 𝐼 and |𝑥 − 𝛼| < 𝜇, then 𝐹(𝑥) > 0
Since 𝐼 is an open interval, there is 𝛼2 ∈ 𝐼 such that 𝛼 < 𝛼2 < 𝛼 + 𝜇
Then:
𝑓(𝛼2 ) − 𝑓(𝛼) = 𝐹(𝛼2 ) ∙ (𝛼2 − 𝛼) > 0 ⇒ 𝑓(𝛼2 ) > 𝑓(𝛼) →←
Therefore, 𝑓′(𝛼) = 0

Proposition 6.3. Let 𝑓: [𝑎, 𝑏] → ℝ be decreasing. Then 𝑓(𝑥) is integrable.

Since 𝑓 is decreasing, 𝑓(𝑎) and 𝑓(𝑏) are bounds on 𝑓(𝑥), so 𝑓(𝑥) is bounded
𝑏−𝑎
Let 𝑛 ∈ ℤ and 𝑛 ≥ 1, and 𝑃 a partition of [𝑎, 𝑏] into 𝑛 pieces with 𝐼 ∈ 𝑃 ⇒ |𝐼| =
𝑛

Let 𝐼 = [𝑐, 𝑑] ⊆ [𝑎, 𝑏]

Since 𝑓 is decreasing, sup(𝑓(𝐼)) = 𝑓(𝑐) and inf(𝑓(𝐼)) = 𝑓(𝑑)

We have: 𝑓(𝑎) ≥ 𝑓(𝑐) ≥ 𝑓(𝑑) ≥ 𝑓(𝑏) ⇒ var(𝑓, 𝐼) = 𝑓(𝑐) − 𝑓(𝑑) ≤ 𝑓(𝑎) − 𝑓(𝑏)
Therefore:
𝑛 𝑛
𝑢 𝑏−𝑎
∑ (𝑓, 𝑃) = ∑ var(𝑓, [𝑥𝑗−1 , 𝑥𝑗 ]) ∙ |[𝑥𝑗−1 , 𝑥𝑗 ]| = ∑[𝑓(𝑥𝑗−1 ) − 𝑓(𝑥𝑗 )] ∙
𝑙 𝑛
𝑗=1 𝑗=1

𝑏−𝑎
= (𝑓(𝑎) − 𝑓(𝑏)) ∙ → 0 as 𝑛 → ∞
𝑛
Thus, 𝑓(𝑥) is integrable
Proposition 6.14. Suppose that 𝑓(𝑥) is integrable on [𝑎, 𝑏]. Then there is a unique number 𝐼
such that
𝑢
∑ (𝑓, 𝑃) ≤ 𝐼 ≤ ∑ (𝑓, 𝑃)
𝑙

for all partitions 𝑃 on [𝑎, 𝑏].

Let 𝑃, 𝑄 be partitions of [𝑎, 𝑏], and 𝑅 a common refinement of 𝑃, 𝑄

By Proposition 6.11: ∑𝑙(𝑓, 𝑃) ≤ ∑𝑙(𝑓, 𝑅) ≤ ∑𝑢(𝑓, 𝑅) ≤ ∑𝑢(𝑓, 𝑄) for all 𝑃
Let 𝐼 = sup{∑𝑙(𝑓, 𝑃) | 𝑃 a partition of [𝑎, 𝑏]}
Then: 𝐼 ≤ ∑𝑢(𝑓, 𝑄) for all 𝑄
Therefore, ∑𝑙(𝑓, 𝑃) ≤ 𝐼 ≤ ∑𝑢(𝑓, 𝑃) for all 𝑃
Prove uniqueness: Let 𝐼, 𝐽 be in [∑𝑙(𝑓, 𝑃) , ∑𝑢(𝑓, 𝑃)] for all 𝑃
Let 𝜖 > 0, get a partition 𝑃 with ∑𝑢𝑙(𝑓, 𝑃) < 𝜖
We have: ∑𝑙(𝑓, 𝑃) ≤ 𝐼, 𝐽 ≤ ∑𝑢(𝑓, 𝑃)
Thus:
𝑢
|𝐼 − 𝐽| ≤ ∑ (𝑓, 𝑃) − ∑ (𝑓, 𝑃) = ∑ sup(𝑓, 𝐻) ∙ |𝐻| − ∑ inf(𝑓, 𝐻) ∙ |𝐻|
𝑙
𝐻∈𝑃 𝐻∈𝑃
𝑢
= ∑ var(𝑓, 𝐻) ∙ |𝐻| = ∑ (𝑓, 𝑃) < 𝜖
𝑙
𝐻∈𝑃

Therefore, 𝐼 = 𝐽