Chapter VI · Derivatives and integrals
The notation dy/dx
Why the derivative is written dy/dx, what that symbol does and does not mean, and the rules of differentiation restated in Leibniz’s notation.
Why this matters
Most books, and most people, write the derivative of as . The notation is extraordinarily convenient: the chain rule looks like cancelling fractions, and the inverse-function rule looks like turning a fraction upside down. It is also misleading, because is not a fraction. In §116 Hardy explains where the notation comes from, warns against reading too much into it, and restates the rules of §114 in it. This lesson makes sure you can use the notation freely without being misled by it.
The idea
When changes from to , the dependent variable changes from to . The two changes are called increments and written
The chord slope is then , a genuine quotient of two numbers. The derivative is its limit as , and Leibniz’s notation writes this limit as , with Latin in place of Greek , as a reminder of where it came from.
Try this. Drag down towards and watch the three numbers in the read-out. Then switch on the zoom, which keeps the little triangle the same size on screen, and do it again.
- δx (increment of x)
- δy (increment of y)
- chord, slope δy/δx
- tangent, slope dy/dx
What to notice
and both shrink to , while settles on the derivative ( for at ). With the zoom on, the triangle keeps its shape and the curve straightens onto the tangent: what survives in the limit is the shape of the triangle, its slope, not its size.
So is the limit of a ratio. It is not a ratio of limits: and , and “” means nothing. Hardy insists that at this stage and have no meaning on their own; the symbol must be read as a whole.
Precise statements
If , its derivative is written in any of the ways
Here , or , denotes the operation of forming and letting . The symbol means “the result of applying to ”. It is not a quotient.
If , are differentiable functions of and is a constant:
- (1) , and (2) ;
- (3) ;
- (4) , and (5) ;
- (6) if is a function of and a function of , then ;
- (7) if and are inverse functions of each other, then (where ).
Rules (6) and (7) are where the notation earns its keep. It suggests the right formula: “cancel the ” in (6), “invert the fraction” in (7). But the suggestion is not a proof. The proof of (6) is the careful one in §114, which had to deal with for small , exactly the case in which “” is meaningless.
Proofs
There is nothing new to prove: each line above is the corresponding rule of §114 in new clothes. It is worth doing the translation once, slowly, for the chain rule, because that is where readers are most often misled.
Name the functions behind the letters, write the rule in Hardy’s functional notation, and check that the two statements say the same thing at the same points.
Name the functions
“ is a function of ” means ; “ is a function of ” means . Then , as a function of , is .
Translate each symbol
means . means . And means , evaluated at . This last point is the one the notation hides: is computed as a function of and then evaluated at the current value of .
Compare
The statement therefore reads , which is exactly §114 (6), with the same hypotheses: differentiable at , differentiable at .
Use rule (7) and keep track of where each derivative is evaluated.
The functions
Let be strictly monotonic and continuous, with inverse .
The rule
By §114 (7), if then , where and correspond, that is, .
Translate
is and is , so at corresponding points. When the rule is silent, and has a vertical tangent as a function of .
Worked examples and exercises
Let and . Find , and say at which point must be evaluated.
Hint
Solution
. The factor is evaluated at , the value of that corresponds to the current . Leaving the answer as is correct but incomplete: the answer should be a function of .
If , write the formula for in Leibniz notation, and deduce that if then .
Solution
From the product of factors, (where no vanishes). With every , the equal terms add up to .
A curve is given by , . Show that, where , the slope of the curve is , and deduce the equation of the tangent at the point with parameter . Apply it to the circle , .
Hint
Solution
Suppose keeps a fixed sign, say positive, on an interval around , so that is strictly increasing there (this is Cor. 1 of §122). Then is a function of with by (7), and by (6)
The tangent at is therefore , or, clearing the denominator, . This last form also makes sense when , where it gives a vertical tangent. For the circle, and . The tangent is , that is, , the familiar tangent to a circle.
For at , compute and as , and the limit of . Explain in one sentence why cannot be defined as .
Solution
. As , and , while . The quotient of the limits would be , which is not a number. The derivative is the limit of the quotient, taken before either increment reaches .
Check yourself
In the notation , the symbol stands for .
False. That limit is for every continuous function, so “” would always be . In Hardy’s treatment here has no meaning of its own; the symbol is read as a whole. (In Chapter VII Hardy gives and a separate meaning, as differentials, chosen so that their quotient is the derivative; but they are still not limits of increments.)
Let . What is when ?
At , and . So . Directly: gives .
Let and . What is at ?
. At , , so this is . Forgetting to evaluate at (rather than at ) gives the wrong answer .
For , which of these does not denote the derivative of with respect to ?
is the chord slope for a particular increment : a genuine quotient, but only an approximation to the derivative. The other three are different names for the limit.
Hardy’s language
The reader must however be careful to remember that dy/dx does not mean ‘a certain number dy divided by another number dx’.
Summary
- and are increments; is a chord slope; is its limit.
- is a single symbol, the result of the operation applied to . It is not a quotient, and is not , which is .
- The notation makes rules (6) and (7) look like algebra with fractions: and . These are good mnemonics but not proofs.
- Always evaluate at the value of corresponding to the current .