4.1.20 · D2Calculus I — Limits & Derivatives

Visual walkthrough — Derivatives of ln x and logₐ(x)

1,956 words9 min readBack to topic

Before we start, one word we lean on: a slope is "how steep a curve is" — how much the height changes when you step a tiny bit to the right. A slope of means "step right by , climb by ." A slope of means "step right by , climb by only a half." That is the whole game: we want the slope of the curve at every point.


Step 1 — What is , really? (a picture, not a formula)

WHAT. is defined as the inverse of the function . Saying "" is exactly the same statement as "." They are two ways to write one fact.

WHY. We are not allowed to use any slope formula yet — we don't have one. So we must anchor to something we do trust: , whose slope law we take from Derivative of e^x and a^x (it is its own slope). Every result below is squeezed out of that single relation .

PICTURE. Look at the figure. The purple curve is . The coral curve is . They are mirror images across the dashed grey line — that is what "inverse" looks like: swap the horizontal and vertical axes and one curve becomes the other.

Figure — Derivatives of ln x and logₐ(x)

Step 2 — Reading slope off the mirror

WHAT. A mirror across swaps the roles of "horizontal step" and "vertical step". So it swaps rise and run. And slope is — so a mirror flips the slope upside down (takes its reciprocal).

WHY. This is the geometric heart of the whole page. We know the slope of ; we want the slope of its mirror image . If mirroring flips slope, we can get one from the other with no algebra at all.

PICTURE. At the point where has a little right-triangle "step" of run , rise (slope ), the mirrored point on has the same triangle with the legs swapped: run , rise — slope . The red triangle and its reflected teal copy show this leg-swap.

Figure — Derivatives of ln x and logₐ(x)

We now turn this picture into an equation. That is Step 3.


Step 3 — Differentiate both sides of

WHAT. Take the relation and ask "how does each side change as nudges right?" We apply the same operation (="rate of change with respect to ") to both sides.

WHY. An equation stays true if you do the same thing to both sides. The right side changes at rate (step right by a bit, grows by exactly that bit). The left side changes too — but itself is silently sliding as moves. This is the implicit idea.

Here because 's rate of change with respect to itself is .

PICTURE. The figure shows the two curves stacked as a chain: . A tiny push at the bottom travels up through and out through . We must track how big the push is at each stage — that is what the Chain Rule does next.

Figure — Derivatives of ln x and logₐ(x)

Step 4 — The chain rule unpacks the left side

WHAT. The left side is a function of , and is a function of . The chain rule says: total rate = (rate of per unit ) (rate of per unit ).

WHY. We only know one slope law for the exponential: grows at rate per unit of . But is the variable being pushed, and only moves a fraction as fast. Multiply the two fractions of speed together — that is the chain rule. Without it we'd be measuring speed in the wrong units.

PICTURE. Two gears meshed: turning the -gear turns the -gear at ratio , which turns the -gear at ratio . The overall gearing multiplies. The equation from Step 3 now reads:

Figure — Derivatives of ln x and logₐ(x)

Step 5 — Solve, and put back in

WHAT. We have . Divide both sides by :

Now recall the relation from Step 1: . Substitute:

WHY. We want the answer expressed in terms of where we are, — not in terms of the height . The relation is the bridge that swaps for the plain number . This is the exact same reciprocal from the mirror in Step 2: is "one over the slope of ", now written using .

PICTURE. The tangent line to drawn at three spots — , , — with its slope printed. Slopes are , , : exactly . Watch the tangent flatten as grows.

Figure — Derivatives of ln x and logₐ(x)

Step 6 — The same answer from first principles (a sanity check)

WHAT. Slope is officially defined as a limit: shrink a step and watch the average slope settle.

Using the log law from Logarithm Laws, and the substitution :

WHY. The pure geometric mirror argument (Steps 1–5) is beautiful but leans on 's slope. This second route leans instead on the standard limit — a completely independent foundation. When two independent roads reach , we trust it.

PICTURE. The chord (a straight line joining two nearby points on ) tilts and settles onto the tangent as shrinks. Its printed slope creeps toward .

Figure — Derivatives of ln x and logₐ(x)

Step 7 — Any base, and the empty basement of

WHAT. For a general base (with ), change base: . Since is just a fixed number, we divide the slope by it:

WHY. We refuse to memorise a second rule. Every is a rescaled copy of : dividing the whole curve's height by the constant divides every slope by too. Setting gives , so the factor vanishes and we're back to — the "basement is empty."

PICTURE. Three logs — , , — are the same shape squashed vertically by . At their slopes are , , : a bigger base means a flatter, gentler curve.

Figure — Derivatives of ln x and logₐ(x)

Step 8 — The degenerate & edge cases (never leave a gap)

WHAT & WHY. Three boundaries the smooth story quietly relies on:

  • (left edge of the domain). Slope : the curve dives down to with a vertical tangent. It never reaches ; the -axis is a wall (an asymptote). simply does not exist for — matching never producing a non-positive .
  • . : the one point where the log-hill is as steep as a ramp. Here , so the curve crosses the axis.
  • Negative via . For we can still take . Its slope is still (now a negative number for ) — the formula extends cleanly to every , but only through the absolute value.

PICTURE. The full curve with the wall marked, the crossing at flagged, and the mirrored branch for drawn faint.

Figure — Derivatives of ln x and logₐ(x)

The one-picture summary

Figure — Derivatives of ln x and logₐ(x)

Everything at once: and mirrored across ; a matched pair of step-triangles with legs swapped (slope becomes ); the tangent to at a general labelled with slope ; and printed as the bridge.

Recall Feynman retelling — the whole walk in plain words

A logarithm is just a reflection of the exponential in a mirror tilted at . A mirror swaps left–right steps with up–down steps, and slope is up-over-across — so a mirror turns every slope upside-down. The exponential's slope at a point of height is itself (it's the self-copying function). Flip that upside-down and you get — the slope of . To be safe I did it a second, totally different way: I took tiny chords, used the log law to fold two logs into one, and rode the standard limit to the finish — same . Other bases are just the same hill squashed shorter by the fixed number , so every slope shrinks to ; base leaves the basement empty because . And the fine print: the hill only exists to the right of zero, at it's a perfect ramp, and it dives to a vertical wall as you approach .

Recall Rapid self-check

Slope of at ? ::: . Slope of at ? ::: (a crossing). Why does a mirror across invert slopes? ::: It swaps rise and run, and slope is rise/run. Slope of at ? ::: . Where does have a vertical tangent? ::: As , where .


Connections

  • Derivative of e^x and a^x — the self-copying slope we mirror.
  • Implicit Differentiation — Steps 3–5.
  • Chain Rule — the meshed gears of Step 4.
  • Logarithm Laws — folds the chord in Step 6.
  • Standard Limit (1+t)^{1/t} → e — the independent foundation of Step 6.
  • Logarithmic Differentiation — where this slope law is put to work.
Mirror across does what to a slope?
Inverts it (reciprocal), by swapping rise and run.
Slope of from the mirror argument
since .
Slope of and why
; the curve is squashed by the constant .
Behaviour of slope as
, a vertical tangent (asymptote wall).