Introduction to Physics for High School Students
Introduction
A car passes a speed camera and the reading says sixty miles an hour. Suppose the driver objects that she was on that road for seven minutes, so she never went sixty miles in any hour, and never could have. She has a point. The reading is a claim about one instant. An instant has no duration, and in no duration nothing travels any distance at all, so the quantity the camera reports is a distance of nothing divided by a time of nothing. Either the reading means nothing, or it means something that dividing a distance by a time cannot say.
Richard Feynman put that difficulty to the students of his introductory physics course and named what resolving it takes.
"This idea was invented by Newton and by Leibniz, independently, and is the beginning of a new branch of mathematics, called the differential calculus. Calculus was invented in order to describe motion, and its first application was to the problem of defining what is meant by going '60 miles an hour.'"
The calculus is the answer to the first question anyone asks about a moving body, and the question stays unanswerable without it. A course that means to discuss motion meets that fact in its opening minutes.
In the United States, most such courses do not. The American Institute of Physics counted about 1,543,000 students in a high school physics course of some kind in the 2018–19 school year, and about 79,000 of them in AP Physics C, the one course of the set built on the calculus. Five in a hundred. The other ninety-five are taught algebra-based or conceptual physics, in which the derivative is withheld and the results that need it are supplied ready-made.
This article is written for one of those ninety-five.
It assumes you can add and subtract whole numbers. Everything else it uses, it builds. Multiplication, division, fractions, negative numbers, the real numbers, functions, the limit, the derivative and the integral are all constructed here, in order, with nothing borrowed forward and nothing left standing on an appeal to a course you have taken or have yet to take. It is hard but possible, and what stands between a sixteen year old and the second law of motion is a timetable.
The split between algebra-based and calculus-based physics is a constraint on schools, on what can be staffed and sequenced across four years. Inside the subject the two are one thing. Newton made the calculus and the mechanics together, and neither was prerequisite to the other. An article has no timetable and can decline the split.
The plan is in four parts. Part I asks what a number is, what an equation asserts, and what a measurement means, then follows a single question to its end: you can add two arrows and you can scale one, so why has nobody told you how to multiply them. The answer is the geometric product, which the demand forces. It generates the algebra in which the rest of the article works. Part II builds the calculus: it defines the limit in full and the derivative through a slope function that makes the rules short to prove. Part III turns the machinery on motion: a constant force, a circle, a spring, a conserved quantity and an orbit, with every formula derived and its hypothesis left visible. Part IV tallies the formulas the derivative replaces and says what changes in three dimensions.
Each part opens with what the schoolroom treatment does at the same point and where it hides its working. Those passages quote their sources word for word, so nothing rests on my characterisation of a book you can open yourself.
Part I
1. Counting
Begin with what you can do already. You can name the whole numbers and so on without end, and you can add two of them or subtract one from another. That is the entire starting stock. Every other operation in this article is built from it, in the order it is needed.
Two words are wanted first, because Part I turns on them. A set here means a collection of things definite enough that for anything you can say whether it belongs. The whole numbers form a set, written . An operation on a set is a rule that takes two members of it and returns a single answer.
Two pieces of shorthand go with this. When a rule takes one member of a set and returns a member of a set , write , and write to name the answer it gives. Writing for the collection of ordered pairs, one drawn from each, an operation on is a rule . Section 11 returns to this and gives the two sets a role. Here the arrows abbreviate what was said in words.
The question to ask of any operation is whether the answer stays inside the set.
An operation on a set is closed when applying it to members of that set always produces a member of the same set.
Addition is closed on . Add two whole numbers and a whole number comes back every time. Subtraction fails.
There are whole numbers and for which is not a whole number.
Proof
Take and . A whole number with would satisfy . Adding to any whole number gives a result of at least , and is smaller than , so no such exists.
A failure of closure is an instruction. The set is too small for the operation, and the missing members are the answers to the questions that currently go unanswered. Following the instruction four times occupies the next section and produces every kind of number this article uses.
2. Ladder of numbers
Each rung of the ladder below is forced by an equation that has no solution on the rung beneath it. Nothing is invented for its own sake. At every stage the new numbers are exactly the answers that were missing.
Integers. The equation has no solution in , by the proposition above. Adjoin a solution to for every pair of whole numbers, call the result , and the enlarged set is the integers , on which subtraction is closed by construction. The negative numbers are the answers to subtractions that previously had none.
Multiplication. With in hand, define for a whole number as added to itself times, and extend it to negative by requiring that continue to hold. This is a definition, and it uses only the addition already available. Multiplication is closed on .
Rationals. The equation has no solution in , since multiplying any integer by gives an integer that is or at least in size. Adjoin a solution to for every pair of integers with , call it , and the result is the rationals . Division by anything other than zero is now closed. Division by zero is not adjoined, because for every , so the equation has no solution anywhere and no consistent answer can be supplied.
At this point every axiom of ordinary arithmetic is available. They are set down once here, so that later sections can say which of them apply.
A field is a set with two closed operations, addition and multiplication, such that for all members , , :
together with the distributive law , and with .
The rationals form a field. So do the real numbers of the next paragraph. Section 6 asks which of these axioms hold for arrows, and the answer drives the rest of Part I.
Reals. One rung remains, forced by a question about length. A square with sides of length has a diagonal whose length satisfies . No rational number does.
There is no with .
Proof
Suppose with and integers, , and the fraction in lowest terms, so that and share no factor greater than . From we get , so is even. An odd number squares to an odd number, since , so is even, say . Then , so , and the same argument makes even. Now and share the factor , contradicting the assumption that the fraction was in lowest terms.
So has holes, since a definite length is named by no member of it. The real numbers are what you get by filling them. Filling needs one more definition. This article characterises by the property that does the filling and does not build it piece by piece out of .
An upper bound for a set of numbers is a number with for every . A least upper bound for is an upper bound no larger than any other upper bound.
The real numbers are the field containing , ordered compatibly with its operations, in which every non-empty set that has an upper bound has a least upper bound. This last requirement is completeness.
Completeness is what lacks. The set of rationals whose square is below is bounded above, by for instance, and among the rationals it has no least upper bound, since any candidate can be beaten by one slightly closer to the diagonal's length. In that least upper bound exists, and it is the number written .
Completeness does more than fill one hole. It makes the limit of Section 12 exist, and through the limit it makes the derivative and the integral possible. Every result in Part III rests on it.
Everything numerical in the remainder of the article lives in .
3. Equals sign
Four different assertions get written with the same sign, and telling them apart is most of what algebra asks of anyone. They look identical on the page.
- Definition. The sign introduces a name for something already available. Writing creates the symbol and says what it abbreviates. Nothing is claimed, and nothing could turn out false. This article writes wherever a definition is meant.
- Identity. The sign asserts that two expressions agree for every value their symbols may take. is true whatever and are. An identity is a theorem, and it has a proof.
- Constraint. The sign asserts agreement for some values and denies it for others. is true when and false otherwise. Solving means finding which values satisfy it, and the equation poses a question.
- Law. The sign asserts something about the world that could have been otherwise. is a claim that measurements come out a particular way. Arithmetic does not prove it and experiment could have contradicted it.
The distinction matters because the permitted moves differ. Both sides of an identity may be replaced by anything they equal, anywhere the identity is stated to hold. A constraint may be manipulated only in ways that preserve its solution set, so multiplying both sides by an expression that might vanish can introduce solutions the original never had. A law may be rearranged as an identity once assumed, and its content lies entirely in the assumption.
A great deal of confusion in a physics class comes from meeting the fourth kind while trained only on the third. A student who has spent two years solving for reads as an instruction to find something. It is a statement that the world behaves in a certain way, from which things may then be found.
4. Measurement
A measurement is a number against a unit. The number alone says nothing: three is not a length, and three metres is. Writing a quantity means writing both, and the pair travels together through every calculation.
Units combine under multiplication and division exactly as numbers do. A distance divided by a time gives metres per second, and dividing that by a time again gives metres per second per second. Nothing here needs a new rule. What does need stating is the restriction that addition obeys.
Fix a set of base units, for mechanics the metre, the kilogram and the second. Every quantity built from them by multiplication and division has a list of three exponents, one for each base unit, and that list is its dimension. A length has , a mass , a speed , an acceleration , and a force, being mass times acceleration, has .
Multiplying two quantities adds their dimension lists entrywise, and dividing subtracts them. Two quantities may be added only when their dimension lists agree, and the sum then has that same list.
Proof
For multiplication, a quantity with exponents is a number times a product of base units raised to those powers, and multiplying two such expressions multiplies the powers of each base unit, which adds the exponents. Division inverts one factor, negating its exponents.
For addition, suppose has exponents and has , and consider changing the size of the first base unit by a factor . The number recording is then multiplied by and the number recording by . If is to name a quantity at all, the numbers recording it must rescale by a single common factor, which forces for every , and hence . The same argument applies to each base unit in turn.
A sum of unlike quantities is not a quantity because its numerical value would depend on which units you chose, and a physical statement cannot depend on that. Dimensional homogeneity is a consequence of the world not caring what units anyone uses.
The same shape recurs later. Quantities have a label. Multiplication combines the labels according to a rule. Addition refuses to mix quantities with different labels because the sum would fail to mean anything. Section 9 attaches a different label, called grade, to the elements of an algebra of arrows, and the same three sentences will be true of it. A reader who is comfortable with the fact that a length cannot be added to a mass has already met the only idea needed to accept that a number cannot be added to an area.
5. Positions in space
Physics is about things that are somewhere, so before any law can be written the space of positions has to be pinned down. The difficulty is that physical space, as anyone experiences it, has no special point in it. A stone falls the same way in Caracas and in Edinburgh, and no experiment picks out an origin. Any faithful model has to start without one.
An affine space modelled on a real vector space is a set of points, together with a rule assigning to each ordered pair of points a member of , called the displacement from to , subject to two demands. For any point and any there is exactly one point with , and for any three points the displacements chain,
That definition withholds a great deal. It gives differences of positions and nothing else. There is no origin and no way to add two points, because the sum of two places is not a place. This is faithful to the physics, in which separations and displacements are measurable and an absolute location never is.
Two small consequences follow immediately and get used without comment later.
For any points and , and .
Proof
Chaining [eq:affine-chain] with gives , and subtracting from both sides leaves . Chaining with then gives , which is the second claim.
Now choose a point and keep it.
An origin is a point , chosen once and kept fixed. Relative to it, every point acquires a unique position vector
and the assignment matches points with members of one for one. We write and from here on name points by their position vectors.
The identification [eq:position-vector] requires a choice, and the choice is arbitrary: a different origin gives every point a different position vector. It permits us to stop distinguishing a place from the arrow that reaches it, so that a moving body becomes a single arrow-valued function of time, and every derivative taken in Part III lands back in the same space where the positions live. Displacements are unaffected by the choice, since [eq:affine-chain] makes the displacement from to equal to whatever origin was picked. That is the only sense in which positions may be subtracted.
Everything now depends on , which has been named but not described.
6. Directed quantities
A displacement has a length and a direction, and the objects that do so are what this article calls arrows. Two operations on them need no justification, because both are forced by what displacements are. Arrows add, by laying one after the other, and [eq:affine-chain] is that addition. Arrows scale by real numbers, since half of a displacement is a definite displacement and so is its reverse.
A real vector space is a set with an addition and a scaling such that addition is commutative and associative with an identity and inverses, and for all and ,
Its members are vectors and the members of are scalars.
A set of vectors is linearly independent when the only way to combine them to is with every coefficient zero, and a basis is a linearly independent set whose combinations exhaust . Every basis of a given space has the same number of members, and that number is the dimension of the space. This word was used differently in Section 4, where the dimension of a quantity was its list of unit exponents. The two senses are unrelated and the context settles which is meant. This article works throughout in dimension two, the plane, with a chosen basis written , so that every vector is for exactly one pair of reals.
Now compare [eq:vector-axioms] with the field axioms [eq:field-axioms] of Section 2, and audit the arrows against the numbers.
Arrows add, and that addition is commutative, associative, has an identity, and has inverses. Every one of the additive field axioms holds.
Arrows scale by numbers coherently, as [eq:vector-axioms] says.
Of the multiplicative field axioms, not one is even stated, because there is no product of two arrows. There is no rule taking two arrows and returning an arrow. There is therefore no multiplicative identity among arrows, no reciprocal of an arrow, and no equation to solve.
That is the gap. School spends years establishing that numbers can be added, subtracted, multiplied and divided, then introduces a second kind of quantity, teaches two of the four operations, and moves on. A reader who noticed the omission and asked about it was probably told that arrows are not the sort of thing you multiply.
They are. The remaining sections of Part I construct the product, and it is forced. First consider the two things school offers in place of a product. Seeing why neither is a multiplication is most of the way to finding the one that is.
7. Two half-answers
The dot product of two vectors, in an orthonormal basis, is
and the length of a vector is , which exists by Section 2 because is a sum of squares. The dot product is symmetric, , it is linear in each slot separately, and .
The form usually met first, , needs the cosine, which Section 10 constructs. Nothing before then uses it.
The dot product measures how much of one arrow lies along another. Section 25 builds work and energy from it. As a candidate for multiplication it fails at the first test of Section 1. It is not closed: feed it two arrows and a number comes back. Multiplying twice is impossible, since asks for the dot product of a number with an arrow, and that is undefined. An operation you cannot apply twice is not an arithmetic.
The other offer is worse.
In three dimensions, the cross product is the vector perpendicular to both and , whose length is the area of the parallelogram they span, pointing along whichever of the two perpendicular directions a stated hand rule selects.
Three objections apply.
The first is that it does not exist here. This article works in the plane, and the definition asks for a direction perpendicular to two independent directions.
Let be linearly independent vectors in a two-dimensional space, and suppose satisfies and . Then .
Proof
Since and are two independent vectors in a space of dimension two, they form a basis, so for some reals . Then
using the linearity of [eq:dot-product] in its second slot. But , and the only vector of zero length is .
So in the plane the cross product returns nothing. Working in three dimensions does not remove the objection. Part III shows that a planet orbiting in a plane has a definite angular momentum and that a spinning wheel in a plane has a definite torque, and neither needs a third direction. The cross product is a substitute that three dimensions happen to allow.
The second objection is that even where it exists its answer does not behave like an arrow. The hand rule is a convention. Reflect the whole configuration in a mirror and every arrow reflects with it, while comes back pointing the wrong way, because the mirror image of a right hand is a left one. A quantity whose direction depends on a convention chosen by the person doing the calculation is a bookkeeping device.
The third objection is that neither offer is a multiplication at all, and school presents them as though between them they cover the ground. One returns a number and throws the plane away. The other returns an arrow and throws the magnitude of the plane into a direction that only exists in three dimensions. Each keeps half of what the pair of arrows knows.
David Hestenes spent a career on both physics teaching and the algebra this article is heading towards. He states the consequence bluntly.
"perhaps a third of the students seem incapable of reasoning with vectors as abstract elements of a linear space. Rather, they insist on conceiving a vector as a list of numbers or coordinates. I have come to regard this concept of vector as a kind of conceptual virus"
He is describing the graduate students he taught over some twenty-five years. The infection starts early, with an algebra so incomplete that coordinates are the only way to compute in it. The next section closes the gap by asking for one thing and taking whatever follows.
8. Geometric product
We ask for a product of arrows with the structure of multiplication and for one thing more. Everything else is then determined. Nothing is invented or chosen for convenience, and the object called a bivector arrives whether we want it or not.
A geometric product on a real vector space with a length is a way of multiplying that:
- is associative and distributes over addition, with real numbers commuting with everything, so that the usual rearrangements of an algebraic expression remain legal; and
- satisfies the contraction rule
The first demand asks only that the thing deserve the name multiplication. The second has the content and is the natural request to make. An arrow is not a number, so a product of arrows cannot be expected to be a number in general. There is exactly one number an arrow has on its own, its length, and the one product whose value is beyond argument is the product of an arrow with itself. Demand [eq:contraction] says that squaring an arrow gives the one number it already knows. That is the entire input.
Under the two demands, for any vectors and ,
so the part of the product that is unchanged by swapping the factors is the dot product and nothing else.
Proof
Apply [eq:contraction] to the vector :
Expand the left side using distributivity, which the first demand permits:
using [eq:contraction] twice. Expand the right side using [eq:dot-product]:
Cancel from both sides.
The dot product was never assumed. It was extracted from a demand that mentions only lengths. Half of the geometric product is already familiar.
The other half is equally forced. Every product splits into a part unchanged by swapping the factors and a part that changes sign, since
and this is an identity, true by inspection. The first bracket is the dot product. Name the second.
The wedge of two vectors is the part of their product that changes sign when the factors are swapped,
, and . The wedge vanishes exactly when the two vectors are parallel.
Proof
Antisymmetry is immediate from [eq:geometric-product], since swapping and negates the bracket. Setting gives , so it is zero.
For the last claim, take a basis and compute. With and similarly for , expanding [eq:geometric-product] and using leaves
That coefficient vanishes exactly when , which is the condition for the two vectors to be proportional.
The number in [eq:wedge-coefficient] is the signed area of the parallelogram the two arrows span, positive when the turn from to runs one way and negative when it runs the other. So gives the information the cross product was trying to deliver, an oriented area, without a third dimension or a hand.
Equation [eq:geometric-product] adds a number to an area. Section 4 anticipates the objection.
Section 4 established that a length cannot be added to a mass, because the sum would depend on a choice of units and so would fail to mean anything.
Nothing of the kind applies here. Writing does not claim that a number and an area are the same sort of thing, or that they can be merged into one. It keeps them side by side in a single expression, the way keeps two real numbers side by side, or the way a shopping list keeps apples and pears on separate lines while still being one list. The parts stay separable, and Section 9 gives the operation that separates them.
"You have probably noticed that the expanded multivector form (30) violates one of the basic math strictures that is drilled into our students, namely, that 'it is meaningless to add scalars to vectors,' not to mention bivectors and pseudoscalars. On the contrary, GA tells us that such addition is not only geometrically meaningful, it is essential to simplify and unify the language of physics"
9. Algebra of the plane
In the plane the algebra generated by the demands of Section 8 is small enough to write out completely.
For an orthonormal basis of the plane,
Proof
The squares are [eq:contraction] applied to unit vectors. For the anticommutation, [eq:polarisation] gives , since the basis is orthonormal.
Write , the oriented unit area of the plane.
Also anticommutes with every vector: .
Proof
Using associativity and then [eq:basis-relations] twice,
where the middle step swapped the adjacent for .
For the anticommutation it is enough to check the basis vectors, since every vector is a combination of them and the product distributes. , while . The same computation with the indices exchanged handles .
Equation [eq:i-squared] retires a supposed mystery. The square root of minus one is the oriented unit area of the plane you are standing in, and it squares to minus one by [eq:basis-relations], which says only that two perpendicular unit vectors anticommute.
Everything the algebra contains can now be listed.
The geometric algebra of the plane is the set of all sums
called multivectors. The four labels are grades: numbers have grade zero, vectors grade one, areas grade two. Writing for the grade- part of gives the operation that separates the pieces.
Compare this with the closing paragraph of Section 4. Quantities have a label. Multiplication combines the labels by a rule, since a vector times a vector gives grades zero and two, as [eq:geometric-product] says. Addition keeps different labels apart. The bookkeeping that a reader already accepts for metres and kilograms is the bookkeeping that runs this algebra, and is how you read one entry off the list.
The multiplication table follows from [eq:basis-relations] and [eq:i-squared] alone.
Full multiplication table
Every entry is one application of [eq:basis-relations]. As an example, .
One consequence explains where the complex numbers came from.
The multivectors of the form , with real, are closed under multiplication, and
Proof
Expand by distributivity and use [eq:i-squared] on the term. Since commutes with real numbers and with itself, no ordering question arises.
That is the multiplication rule for complex numbers with in the place of . The complex numbers are the even part of the algebra of the plane. Section 10 shows what they do.
10. Rotors and reflections
The usual notation is the exponential , which requires an infinite series and hence the limit of Section 12. This article borrows nothing forward, so everything below uses the algebra already in hand. Section 17 supplies the exponential once the limit exists to support it.
A rotor is a multivector with .
The condition says the pair lies on the unit circle, so the rotors form a circle, one for each angle of turning.
For a vector and a rotor , the product is again a vector, and .
Proof
With and , the multiplication table gives and , so
which has grade one only. Its squared length is
after expanding, cancelling the two cross terms, and using .
The product of two rotors is a rotor, and in terms of the circle parameters
Proof
The formula is [eq:even-product]. It remains to check the result is a rotor, which is the identity
obtained by expanding both squares and observing that the cross terms cancel.
Now attach the usual names. Write for the rotor whose circle parameters are the coordinates of the point reached by travelling a distance anticlockwise around the unit circle from , and call those coordinates and . That the arc length is a well-defined number is a fact about integration, supplied in Section 16. Until then labels a point on the circle and nothing below depends on more.
With those names, [eq:rotor-compose] reads
These are the angle-addition formulas, obtained as a statement about multiplying two elements of an algebra. They are the rule for composing turns, and composing turns is multiplying rotors.
Two further facts finish the section. The first is a dimensional accident.
For a rotor , write . Then for every vector ,
Proof
By [eq:i-squared] the unit area anticommutes with every vector, so and hence
Multiplying on the right by gives .
The two-sided form is the law in every dimension. In three dimensions the one-sided form is false, because the bivector of a rotor anticommutes only with vectors in its own plane and commutes with the direction perpendicular to it. In the plane the two coincide, so this article uses the shorter throughout. The general statement is [eq:sandwich]. Section 28 says what changes.
The second fact explains where turns come from.
For a unit vector , the map reflects in the line through . Reflecting in the line through and then in the line through turns every vector by the rotor .
Proof
Take without loss of generality, by choosing the basis. Then , so the direction along is fixed, and , so the perpendicular direction is reversed. Fixing one axis and reversing the other is exactly reflection in the first.
For the composition, applying the two reflections in turn sends to . Since is an even multivector of unit size, it is a rotor , and . The map is therefore , which by [eq:sandwich] is multiplication by .
Part II builds the calculus and Part III applies it with the algebra of Part I.
Part II
Part II builds the calculus, which the algebra-based course exists to avoid. First comes the case for avoiding it at full strength. It is better than its reputation and the article's answer depends on it.
Stewart Brekke taught physics and chemistry from 1975 to 2001, mostly in Chicago inner-city high schools where a majority of students were Black or Hispanic. He wrote to Physics Today in 2021 arguing for algebra-based physics, and he began from a fact that reframes the dispute.
"In most US public high schools, only the upper 35% of the student body is allowed or encouraged to enroll in algebra-based physics classes. Such restrictions often exclude many otherwise bright and capable youngsters—of all races and ethnicities—from that formative first step toward a future in physics."
The quarrel over whether the calculus belongs is conducted above a threshold that two thirds of students never reach. Before anyone is asked to choose between two kinds of physics course, most are offered neither.
Brekke's second fact concerns who teaches: about two thirds of US high schools, he writes, have no teacher who majored in physics or physics education. His source is the American Institute of Physics survey of the 2012–13 school year. That survey counted about 27,000 US high school physics teachers, of whom over 8,600, or 32 per cent, had a degree in physics or physics education, and 40 per cent taught a majority of their classes in other subjects (Tyler and White, What High School Physics Teachers Teach, AIP 2014). Brekke's figure counts schools and this one counts teachers. The teacher count is the one quoted here.
From those facts Brekke concludes that algebra-based physics should be opened to every student, with extra help in mathematics, and reports his method: drill with the standard formulas, about ten problems per formula, with the repetition through physics teaching the algebra.
The facts are real and the conclusion does not follow from them. That most physics teachers lack a physics degree is an argument about what teaching materials must contain, and it says nothing about how much mathematics a curriculum should contain. Brekke half concedes this, since his own remedy is a materials remedy. He chose thin materials and then treated the ceiling as fixed. Thin staffing argues for better materials. It never argues for less mathematics.
Subject knowledge is not dispensable. Section 20 quotes research whose entire method is diagnosing why a student's wrong answer is wrong, and telling a productive confusion from a plain error takes expertise no worksheet supplies. A teacher without it can hand a student this article. They cannot repair a student who has misread it.
The strongest version of the access argument is more recent and more careful. Suzanne White Brahmia and Geraldine Cochran put it this way.
"In the U.S., introductory calculus-based courses often acts as a gatekeeper to STEM degrees. But access to the prerequisite math is far from equal—students from low-income and racially marginalized communities are far less likely to have the chance to take calculus before college. These disparities are then compounded by rigid placement systems that reward procedural fluency in algebra and trigonometry, while overlooking the conceptual quantitative reasoning that physics truly demands. Labeling students as 'underprepared' ignores the structural causes of these gaps and places the burden of remediation on those least supported."
Every word of that is true. Their remedy is an extended, credit-bearing course that embeds quantitative reasoning and gives students more room, in their phrase to "expand access without compromising rigor". The strongest published argument for access does not conclude that the calculus should go. It concludes that the room should grow. An article has no timetable, and a reader may take a year over this one.
A word about where I am standing, since the argument is partly about students like the one I was. I came to the United States from Venezuela and went through its public schools. I sorted myself into the advanced track and stayed there. I took more than sixteen Advanced Placement courses across two Houston public high schools and did not find them hard. What I found was that difficulty, as school used the word, almost always meant procedural volume: longer computations, more of them, arranged to fill a timetable I was legally required to sit inside. Very little of it asked a question worth the time.
So I balk when the case for a thinner physics course is made on behalf of students like me. The proposal takes a subject that has been emptied of its reasoning and offers it to the students with the least access to anything better, then calls that access. Underrepresented students are owed the subject itself. Handing them a contrived version and calling it enriching is condescension in the language of help.
There is also a question of what the prerequisite is protecting. A study of 150 Calculus I final examinations from across US colleges and universities classified 3735 individual items and found that 85.21 per cent could be answered by recalling a fact or applying a memorised procedure, with the examinations seldom asking for explanation and rarely requiring a student to use the central ideas of the course at all. Whatever a calculus prerequisite certifies, it is largely technique.
Technique does not travel. The finding that ends the argument is nearly forty years old.
"Frequently students who have no trouble plotting points and computing slopes cannot apply what they have learned about graphs from their study of mathematics to physics."
Students who can compute a slope correctly cannot read a slope as a rate. The mathematics course delivered the procedure and the procedure did not transfer. Procedures learned without the ideas beneath them do not. Part II builds those ideas. It defines the limit in full, defines the derivative in a way that makes its rules provable in two lines each, constructs the integral and assumes nothing except Part I.
11. Function
A function consists of a set called the domain, a set called the codomain, and a rule assigning to each member of exactly one member of . The member assigned to is written .
Three parts of that definition matter. The domain is part of the function, so on the non-zero reals and the same formula on the positive reals are different functions. Each input gets exactly one output and so names something definite. Nothing requires the rule to be a formula. A function is an assignment, and formulas are one way of describing assignments.
The graph of a function is the set of pairs , drawn as a curve. The curve records the function without being the function, as a photograph of a bridge is not a bridge.
Section 20 examines what students do with graphs in physics, and the difficulty documented there is reading a picture back into the thing it pictures.
12. Limit
The camera in the introduction reported a speed at an instant, and dividing a distance by a time could not produce one. The way out is to ask what the ratio approaches as the interval shrinks, and to make "approaches" exact. This section gives the most demanding definition in the article.
The idea is a challenge and a response. Someone claims that approaches as approaches . To test the claim you name a tolerance: get within of for whatever small you like. The claim stands if for every tolerance you name, the claimant can name a nearness to that delivers it.
Let be defined near , though possibly not at itself. We say tends to as tends to , written , when
The clause excludes deliberately. The limit asks where the function is heading as the input closes in, and it must be able to ask that even where the function has no value at all. The speed camera creates exactly that situation.
.
Proof
Let be given. We must produce a . Since
so the requirement is the requirement . Take . Then whenever we have , as required.
Every proof from [eq:limit-def] has that shape. The tolerance arrives first and the nearness is built in response, with the response allowed to depend on the tolerance. Reversing the order would be a different and much weaker claim.
Two facts about limits are used constantly below.
If and , then .
Proof
Suppose and set . Applying [eq:limit-def] to each limit gives and ; let be the smaller. For any with , both and hold, so
using the triangle inequality. That says , which is impossible, so .
If and , then
Proof
Sum. Let . Choose for with tolerance and for with tolerance , and let be the smaller. For ,
Product. Write the difference so that one factor is controlled at a time:
First choose so that for , which forces . Let , and choose giving and giving . With the smallest of the three, the two terms above are each below , so their sum is below .
13. Continuity
A function is continuous at when is defined, exists, and the two agree:
It is continuous when it is continuous at every point of its domain.
All three clauses are needed. A function can fail to be continuous by having no value at the point, by having no limit there, or by having both and having them disagree. In the third case nothing appears wrong.
Let for every and . Then , since taking to be anything at all makes for . But , so is not continuous at .
Section 14 defines the derivative by building a function of exactly this shape and then insisting the value at the point be the right one. Repairing a hole by supplying the limit as the value is the mechanism.
Every polynomial is continuous at every real number.
Proof
The constant function has , taking any . The identity has , taking . A polynomial is built from these by finitely many sums and products, and [eq:limit-laws] says each such step preserves the limit, with the value at assembled the same way.
14. Derivative
Return to the speed camera. Over an interval from to the average speed is the distance divided by the time, and the quantity we want is what that ratio becomes when the interval closes up. Written out, the ratio is
This is a good number for every except , where it reads and means nothing.
Section 13 met a function with a hole and repaired it by supplying the limit as the value. The definition below performs the same repair by refusing to divide in the first place.
A function is differentiable at when there is a function , defined near and continuous at , with
The derivative of at is then . We call the slope function of at .
Away from the equation forces to be the ratio [eq:difference-quotient], so records the slope of the chord from to . At itself the equation says only and imposes nothing, and the entire content of the definition is the demand that be continuous there. Continuity is what ties the value at the point to the values around it.
One example shows the mechanism.
For factor:
so . That is a polynomial, hence continuous everywhere by Section 13, and .
No limit was taken and no appeared. The factor was cancelled before it could cause trouble. The cancellation is legitimate because it happens away from , and continuity extends the result to .
The same trick handles every power at once, with no binomial expansion.
For a whole number and , .
Proof
The identity
holds for every , as expanding the right side and cancelling the telescoping middle terms confirms. So is the bracket, a polynomial and therefore continuous, and evaluating at gives copies of .
One theorem relates the definition to the one every other book uses.
is differentiable at in the sense of [eq:caratheodory] exactly when
and then the two values of agree. In particular the derivative is unique.
Proof
Forward. Suppose [eq:caratheodory] holds with slope function . For we may divide, giving . Since is continuous at , [eq:continuity] says , and the left side is the limit in [eq:limit-derivative]. So the limit exists and equals .
Backward. Suppose the limit [eq:limit-derivative] exists and call it . Define
Then holds for by construction and for because both sides vanish. Also is continuous at , since is the hypothesis. So [eq:caratheodory] holds with this , and .
Uniqueness. Any two slope functions agree at every , being both equal to the ratio there, so their values at agree too by continuity and Theorem [eq:limit-unique].
In the next section every rule of differentiation becomes two lines of algebra with no limit manipulation.
Part III uses the following without comment.
If is differentiable at then it is continuous at .
Proof
From [eq:caratheodory], . As the factor tends to and the factor tends to , so by the product law [eq:limit-laws] the whole second term tends to and .
15. Rules of differentiation
Each rule below follows the same recipe: write the difference of the outputs, extract a factor of , and read off what multiplies it. The function that multiplies it is the slope function, and its value at is the derivative.
If and are differentiable at , so is , and
Proof
Let and be the slope functions of and at . Add and subtract one term:
The bracket is the slope function of . It is continuous at , being built from , and by products and sums, each continuous at by hypothesis and by Proposition [eq:diff-implies-cont]. Its value there is , which is [eq:product-rule].
In the usual treatment the chain rule is the hardest of the three, because dividing by is illegitimate whenever that difference vanishes, and the standard proofs work around it. Here nothing is divided and the difficulty does not arise.
If is differentiable at and is differentiable at , then is differentiable at and
Proof
Let be the slope function of at , and that of at , so that . Put :
The slope function of at is therefore . It is continuous at , since is continuous at , is continuous at , and a composition of functions continuous at the relevant points is continuous. Its value at is .
If is differentiable at and , then is differentiable at with
Proof
Since is continuous at and , it is non-zero near , so the reciprocal is defined there. With the slope function of ,
so the slope function of is , continuous at because the denominator does not vanish there. Its value is . The quotient rule follows by applying [eq:product-rule] to and collecting terms over .
The definition also gives a second result in Section 17: the same argument goes through unchanged for functions whose values are multivectors.
16. Integral
The derivative answers what a quantity is doing at an instant. The integral answers the reverse question: given the rate at every instant, what is the total. Section 20 needs it to turn an acceleration into a position, Section 25 needs it to turn a force into an energy, and Section 17 needs it to say what an angle is.
Let be bounded on . A partition is a finite list . On each piece write for the greatest lower bound of there and for the least upper bound, both existing by completeness, and form
Every lower sum is below every upper sum. When the least upper bound of the lower sums equals the greatest lower bound of the upper sums, is integrable and that common value is .
The picture is the familiar one of rectangles under a curve, with and squeezing the area from below and above. The definition does not say what area means. It says the two approximations agree, a statement about numbers alone.
Every continuous function on a closed bounded interval is integrable there.
Proof from completeness
The work is in showing that a continuous function on is uniformly continuous: for every there is one that works at every point at once.
Suppose not, and fix an for which no works. For an interval inside , say works for when for every in and every in with . Bisect . At least one half admits no working , since if and work for the two halves then the smaller works for the whole, every lying in one half or the other. Bisect that half, and continue. This produces nested intervals whose lengths halve each time. The left endpoints form a set bounded above, so by completeness it has a least upper bound , and lies in every one of the nested intervals.
Now is continuous at , so some gives whenever , and hence for any two such points. Take a nested interval shorter than . Every in it and every with lie within of . So works for that interval, against its choice.
With uniform continuity in hand, take , choose the corresponding , and take any partition with every piece shorter than . On each piece , so
Since was arbitrary, the upper and lower values are squeezed together and agree.
The fundamental theorem below needs one more theorem about derivatives.
Let be continuous on and differentiable on the interior. Then there is a point strictly between and with
In particular, a function whose derivative vanishes throughout an interval is constant there.
Proof in three steps from completeness
Attainment of a greatest value. The values are bounded above: otherwise, bisecting repeatedly and always keeping a half on which is unbounded produces nested intervals whose left endpoints have a least upper bound by completeness, and is then unbounded on every neighbourhood of , contradicting continuity at . So the values have a least upper bound . If never reached then would be continuous on , hence bounded by the argument just given, which would keep a fixed distance below and contradict being the least upper bound. The same argument applied to gives a least value.
Rolle's case. Suppose also . If the greatest and least values both occur at the endpoints then is constant and every interior point has zero derivative. Otherwise some interior is an extremum, say a maximum. Let be the slope function there, so with nearby. For the factor is positive, forcing ; for it is negative, forcing . Since is continuous at and is squeezed from both sides, , which is . A minimum is handled by the same argument with the inequalities reversed.
General case. Apply the previous step to
which is continuous, differentiable in the interior, and has . The interior point it supplies has , which rearranges to [eq:mvt].
Consequence. If throughout and for some pair, then [eq:mvt] on supplies a point where the derivative equals the non-zero chord slope, which is impossible.
Let be continuous on .
First part. The function is differentiable on with .
Second part. If is any function with on , then
Proof
First part. Fix in the interval. For , additivity of the integral over adjacent intervals gives . Since is continuous at , for any there is with for . For such , bounding the integrand between and gives
So the difference quotient tends to , and by Theorem [eq:caratheodory-equiv] that is exactly differentiability with .
Second part. Both and have derivative , so throughout, and Theorem [eq:mvt] makes constant. Hence
The length of the path traced by a differentiable between and is
the total of the speed. For the unit circle, the angle of a point is the length of the arc from to it, taken anticlockwise, and this is the number Section 10 named without yet being able to define. The total length of the unit circle is written and that defines .
That fills the gap recorded in Section 10. The rotor parameters and are now coordinates of a point located by a definite number, and Section 17 can differentiate them.
17. Differentiating multivectors
Part III studies bodies whose position is a vector varying with time, so the calculus of Part II must apply to multivectors as well as numbers. It applies unchanged, for the following reason.
Let assign a multivector to each real . It is differentiable at when there is a multivector-valued , continuous at , with
and then . Continuity of a multivector-valued function means continuity of each of its four components.
[eq:multivector-caratheodory] is [eq:caratheodory] with no change. The factor is a real number, and real numbers commute with everything in the algebra, so no question of ordering arises and the definition needs no repair. Differentiating componentwise gives the same answer, since each component equation is an instance of [eq:caratheodory].
The product rule changes in one respect.
If and are differentiable at , so is their geometric product, and
The two factors must be left in the order written.
Proof
Repeat the proof of [eq:product-rule], taking care never to commute anything:
where and are the slope functions and the scalar factor has been moved to the right, which it may be since scalars commute. The bracket is continuous at and its value there is [eq:multivector-product].
The order matters, and a familiar consequence fails: , which is only when commutes with its own derivative. Section 21 uses this.
What remains is the derivative of the rotor, and one limit does all the work.
Proof
Take . Then and are both positive. By Definition [eq:arclength] the angle is a length, and the task is to trace the arc and measure it with [eq:arc-length]. Trace it by
Since , every lies on the unit circle. It starts at , and substituting with gives and . The path therefore ends at , the point at angle . The quotient rule gives
Since , on the first component falls and the second rises: the path runs anticlockwise from to the point and never doubles back. Its length is therefore the angle,
The integrand is at most . It is at least the component of , since a vector is at least as long as any of its components, and by [eq:ftc] that component integrates to the change in the second coordinate. A larger integrand has a larger integral, as the lower and upper sums show. Hence . Dividing by and by ,
From also . Hence , the left side tends to and the middle term is squeezed to . The same is true for since both and change sign together.
For the second limit, multiply above and below by :
Write for the rotor at angle . Then is differentiable and
Proof
The composition law [eq:rotor-compose] says . So
and it remains to evaluate the last factor. By definition , so
by Lemma [eq:circle-limits]. Hence the difference quotient tends to , which equals because commutes with every even multivector. Theorem [eq:caratheodory-equiv] turns the limit into differentiability.
Because [eq:rotor-derivative] is the defining property of an exponential, that with , and because [eq:rotor-compose] is the law , we write
This is a name for an object already constructed, and no series is needed.
For a constant ,
Proof
The first is [eq:rotor-derivative] composed with through the chain rule [eq:chain-rule], whose proof in Section 15 used only the slope-function argument and so applies verbatim to multivector values. Differentiating a second time repeats it and gives a factor by [eq:i-squared].
Sections 21 and 22 rest on [eq:rotor-second]. Something turning at a steady rate has an acceleration equal to times its position, and circular motion, centripetal acceleration and the harmonic oscillator all follow from that with no further idea.
Part III
The order in which a subject is taught is often mistaken for the order in which it must be learned. Physics is presented as coming after the calculus, so it looks as though the calculus is a prerequisite. The history says otherwise.
Newton had the calculus before he wrote the Principia. He then wrote the Principia without it, in the geometrical style of the ancients, and the scholarship on why has run since the 1690s.
"This question comes very naturally to mind, since Newton discovered the calculus of fluxions before writing the Principia. It is just obvious to think that Newton had employed the calculus in order to mathematize his natural philosophy. However, very little trace of calculus techniques is to be found in the Principia, which are mostly written in 'geometric style'."
Guicciardini argues at length that the geometry of the Principia is more than calculus in disguise, and that the mathematicians who later translated it into calculus found difficulties. Only the ordering is claimed here. Newton did not complete a course in the calculus and then take up mechanics. He made both, and the sequence taught today is a decision about timetables.
Newton wrote in the late 1710s about readers who had by then changed, and said himself why the Principia looks as it does.
"To the mathematicians of the present century, however, versed almost wholly in algebra as they are, this synthetic style of writing is less pleasing, whether because it may seem too prolix and too akin to the method of the ancients, or because it is less revealing of the manner of discovery. And certainly I could have written analytically what I had found out analytically with less effort than it took me to compose it."
He found the results analytically and then rewrote them geometrically, with more effort, for an audience he judged unready. That is the algebra-based physics course, described three centuries early by the person whose name is on it.
The geometry available to him could not express the quantity his second law is about.
"the geometric mathematics Newton used in the Principia — and others were using before him — had no way of representing acceleration as a quantity in its own right." … "But the geometric mathematics used in the Principia offered no way of representing second derivatives."
Newton worked around it by using the curvature of the path, the circle that touches a curve, wherever a second derivative was wanted. Section 21 reaches for a second derivative directly and gets the same result in two lines.
One more constraint on this Part explains a choice made back in Section 6.
"GA also makes it possible to formulate and solve 2D physics problems in terms of vectors without introducing coordinates. Conventional vector algebra cannot do this, in part because the vector cross product is defined only in three dimensions. That is the main reason why coordinate methods dominate introductory physics. The available math tools are too weak to do otherwise. … For example, most of the mechanics problems in introductory physics are 2D problems."
Everything below happens in the plane, where the cross product does not exist and nothing is lost by its absence.
18. Kinematics
The motion of a body is a trajectory, a function assigning to each time a position vector in the plane, in the sense fixed by [eq:position-vector]. Its velocity and acceleration are
the derivatives of Section 17. The speed is .
Both derivatives are objects of the same kind as the position, since differentiating a vector-valued function of one real variable returns a vector and the identification of a point with an arrow permits this. Without it a velocity would live somewhere other than where positions live, and saying where would need the theory of manifolds.
The speed at an instant, which the introduction claimed was unavailable to arithmetic, is now . Nothing is divided by zero, because nothing is divided.
19. Second law
A body has a mass , a positive number, and a momentum
A force is a vector-valued quantity , and the law of motion is
with the second equality valid when the mass does not change.
Recall the four uses of the equals sign from Section 3. Equation [eq:newton-two] is the fourth kind. It asserts something about the world that could have been false, and its content is that the force determines the second derivative of the position and nothing else. It says nothing directly about where the body is or how fast it is going. Give it a force law, a starting position and a starting velocity, and everything else follows by integration.
That is the method of Part III, and the sections below apply it.
20. Constant force
Take the simplest case, a force that does not change. A dropped stone near the ground is the standard example, and [eq:newton-two] gives , a constant vector which we call .
If is constant, and the body starts at with velocity , then for all
and the speeds satisfy
Proof
The function has derivative , so it and differ by a constant by Theorem [eq:mvt] applied componentwise. At both equal , so the constant is zero. Integrating once more the same way gives the second formula, using [eq:ftc] and the power rule [eq:power-rule].
For [eq:suvat-energy], differentiate using the product rule, which holds for the dot product by the same slope-function argument:
so has zero derivative and is therefore constant. Evaluating at gives the stated form.
Those three formulas, together with the two that follow from eliminating between them, are the set a physics student is asked to memorise. They are a corollary here, obtained by integrating a constant twice, and every one of them has its hypothesis on its face: the acceleration is constant. A spring, a planet, a pendulum and a falling body meeting air all violate it, and for those the formulas say nothing.
School makes one further claim alongside them.
The average velocity over is the mean value of the velocity there,
which by [eq:ftc] is the displacement divided by the elapsed time.
holds whenever the velocity is a polynomial of degree at most one in , that is whenever the acceleration is constant. If the velocity has a further term , so that the acceleration changes at the constant rate , then over an interval of length the formula is wrong by exactly
Proof
Set and without loss. For ,
which agree. For the same two computations give and , whose difference is [eq:trapezoid-error].
[eq:trapezoid] is the trapezoid rule. It approximates the area under a graph by the area under the chord. It is exact for a straight graph and wrong for every other, and [eq:trapezoid-error] is the size of the error.
The most widely used free textbook in American classrooms, OpenStax College Physics, states the formula and marks its hypothesis correctly with "(constant a)" beside it. Its justification is the difficulty:
"The equation […] reflects the fact that, when acceleration is constant, v is just the simple average of the initial and final velocities."
Read as a statement it is true. Read as a reason it is empty, because "the simple average" is exactly what was to be shown. Proving it takes the definition [eq:average-velocity] and the fundamental theorem, and those are the two things an algebra-based course does not have. So the kinematic formulas, which the book derives correctly from that starting point, rest on a base that cannot be established within the course that teaches them.
The consequence shows up in what students can do afterwards. The research is nearly forty years old and has been replicated many times since.
McDermott, Rosenquist and van Zee studied several hundred university students in a laboratory-based preparatory physics course at the University of Washington over several years, found the same errors in introductory and high school classes, and reported (pp. 504–505) that students "frequently do not know whether to extract the desired information from the slope or the height of a graph", and that they "frequently do not realize that they should use the slope of an x vs t graph as the height of a v vs t graph".
Those are the two facts [eq:kinematics] and [eq:average-velocity] assert. A student who has met them as procedures has met them as separate tricks. A student who has met the derivative and the integral has met them as one statement seen twice.
21. Circular motion
A body going round a circle of radius at a steady rate has a position that is a fixed vector turned through an angle growing steadily with time. Section 10 says what turning is and Section 17 says how to differentiate it, so the trajectory can be written down at once:
with the constant rate of turning. Everything about the motion now follows by differentiating.
For the trajectory [eq:circular-trajectory],
The speed is constant at , and the acceleration has constant magnitude and points from the body towards the centre.
Proof
Differentiate [eq:circular-trajectory] using Corollary [eq:rotor-second], noting that is a constant vector and so passes through the derivative:
using from [eq:i-squared]. Multiplying a vector by and by a rotor preserves length, by Theorem [eq:rotor-turns], so and , both constant. Since with , the acceleration is the position reversed and rescaled, so it points at the centre. Finally .
Two differentiations of one line produce the result that directed inwards with no geometry and no approximation in the argument.
Set the standard derivation beside it. OpenStax College Physics obtains the same result from a picture of two similar triangles.
"Using the properties of two similar triangles, we obtain Δv/v = Δs/r." … "Acceleration is Δv/Δt, and so we first solve this expression for Δv: Δv = (v/r)Δs. Then we divide this by Δt, yielding Δv/Δt = (v/r) × (Δs/Δt). Finally, noting that Δv/Δt = a_c and that Δs/Δt = v … the magnitude of the centripetal acceleration is a_c = v²/r." … "(Because Δθ is very small, the arc length Δs is equal to the chord length Δr for small time differences.)"
Three of those steps are false as written, and each is false in the same way.
is not an equality. The ratio is an average acceleration over an interval, and is its value at an instant. The two agree only in a limit. is not an equality, for the same reason. An arc is never equal to its chord over an interval of positive length, however small. The two approach a common ratio, a statement about a limit.
The word limit does not appear in that passage. A limit has been taken three times, correctly, and each time recorded as an equals sign. The result is right because the errors vanish in a limit that is never mentioned. As written the reasoning is invalid, and a student who takes it at face value learns that mathematics is a procedure that emits correct answers for reasons one is not invited to examine.
Section 3 separated four uses of the equals sign. This passage adds a fifth, in which the sign means "becomes, once something we are not going to discuss has been done".
Newton faced the same obstacle and had no way round it. His geometry, as the opening of this Part quoted, "offered no way of representing second derivatives", so throughout the Principia he used the curvature of a path, the circle touching it, wherever a second derivative was wanted. That is a workaround for a limitation of the tools available in 1687.
An algebra-based course in 2026 reproduces the limitation on purpose.
22. Spring
A spring pulls back in proportion to how far it is stretched. This is Hooke's law: the force is for a positive constant , with the minus sign recording that the pull opposes the displacement. Feed that into [eq:newton-two]:
Compare [eq:oscillator] with [eq:centripetal]. They are the same equation, with . A mass on a spring and a body going round a circle are governed by one statement.
Set . Then
satisfies [eq:oscillator], each of its components separately satisfies the same equation, and the motion repeats after a time
Its component is and its component is .
Proof
That [eq:shm-solution] satisfies [eq:oscillator] is Theorem [eq:circular] with , since .
For the components, expand using [eq:euler] and the multiplication table: . Differentiating a sum differentiates each term, and and are constant, so each scalar coefficient satisfies the same scalar equation that the whole satisfies.
The period is the smallest positive with for all . By [eq:rotor-compose] that requires , so is a full turn of the circle, which is by Definition [eq:arclength]. Hence .
The figure traces the component, because a horizontal connector makes the projection visible. That component is the sine. It is the same motion started a quarter turn later and satisfies the same equation.
So the cosine is a theorem. It is one component of a rotor, and the oscillation is sinusoidal because turning at a steady rate solves [eq:oscillator].
The same result reaches a student in the algebra-based course like this.
"The displacement as a function of time t in any simple harmonic motion—that is, one in which the net restoring force can be described by Hooke's law, is given by x(t) = X cos(2πt/T)"
There is no derivation, because there cannot be one. Equation [eq:oscillator] relates a function to its own second derivative, and a course without derivatives has no way to write it down, let alone solve it. The answer is handed over, and the student takes on trust the main fact about oscillation, that it is circular motion seen from the side.
That fact is usually mentioned as an analogy. It is [eq:shm-solution].
23. Splitting the law
Section 8 built a product of vectors with two parts, a number and an area. Every use of it so far has been in one piece. This section takes it apart, because the two parts of one product are two conservation laws of mechanics.
Multiply the second law [eq:newton-two] by the position with the geometric product and split by [eq:geometric-product]:
Do the same with the velocity:
Two products, four pieces, and the two that matter are the bivector part of the first and the scalar part of the second. This section takes the bivector, and Section 25 takes the scalar.
The torque about the origin exerted by a force acting at position is the bivector
Torque is an oriented area of grade two and a definite object in the plane. In the treatment a student meets the torque has a magnitude and a direction supplied by holding up a hand.
There is no third direction in the plane, so no hand can be held up and no rule can select anything. A wheel turning in a plane experiences a torque, and Section 24 shows that a planet orbiting in a plane has angular momentum. Both quantities exist here and neither is a vector.
In three dimensions the right-hand rule converts an oriented plane into a normal direction. The conversion is available only because three dimensions supply exactly one direction perpendicular to a given plane. Section 28 gives the formula. The rule is dimensional good luck dressed up as a definition, as Section 7's second objection shows: reflect the whole apparatus in a mirror and the rule returns the wrong answer, because a mirror turns a right hand into a left one.
The bivector needs no hand and reflects correctly.
24. Angular momentum
The angular momentum of a body about the origin is the bivector
Proof
The wedge is built from the geometric product by [eq:wedge], so the product rule [eq:multivector-product] applies to it term by term and gives the same shape of answer:
The first term vanishes by Proposition [eq:wedge-props], since a wedge of a vector with itself is zero. Multiplying by the constant and using from [eq:newton-two] gives [eq:L-dot].
A force is central when it acts along the line joining the body to the origin, so that for some scalar-valued . Gravity and the spring of Section 22 are both central.
Under a central force is constant.
Proof
, using Proposition [eq:wedge-props] and the fact that a scalar factor passes through the wedge. So by [eq:L-dot], and a quantity with vanishing derivative is constant by Theorem [eq:mvt].
That is a conservation law, obtained in three lines from a differentiation and a term that vanishes because a wedge of parallel things is zero. It needs no appeal to symmetry, experiment or the inverse-square law. It holds for any central force.
The physical meaning of follows from the meaning of the wedge. Area has so far been given a meaning only for parallelograms. The area swept by a moving line needs a definition first.
Let a body move with polar form , where and are differentiable functions of time and is continuous. The area swept by the line from the origin to the body between times and is
Over a short time the line turns through about at radius about , and is the familiar area of a thin circular sector of that radius and angle. The definition adds these thin sectors and takes the limit through [eq:darboux]. For constant and a full turn it returns .
Let be the area swept by the line from the origin to the body. Then
Under a central force the right side is constant, so the line sweeps equal areas in equal times.
Proof
Write the position in the polar form , with and both functions of time. Differentiating with the product rule, Theorem [eq:rotor-derivative] and the chain rule,
the first term pointing along and the second perpendicular to it. Wedging with kills the first term, and the second contributes its full magnitude, so and hence .
For the area, the integrand in [eq:swept-area] is continuous and the first part of [eq:ftc] gives .
Kepler extracted that law from Tycho Brahe's observations of Mars and published it in Astronomia Nova (1609). Here it is a corollary of the conservation law above, and the derivation never mentions gravity or ellipses. The law applies to a planet, a mass whirled on a string and a charged particle attracted to a fixed charge.
The treatment a student receives states the law and says what it does not supply.
"Each planet moves so that an imaginary line drawn from the Sun to the planet sweeps out equal areas in equal times." … "The point is to demonstrate that the force of gravity is the cause for Kepler’s laws (although we will only derive the third one)."
So the second law is left where Kepler left it, as a regularity extracted from observation. Its cause is [eq:L-dot] with the right-hand side equal to zero. That is three lines from the second law of motion and unavailable to a course that cannot differentiate a product.
25. Work and energy
Section 23 split two products and set one piece aside. This section takes the scalar part of [eq:vF-split], the quantity , and finds the other conservation law of mechanics inside it.
The power delivered by a force to a body is , and the work done between times and is its total,
The kinetic energy is .
Equation [eq:work] is an ordinary integral of an ordinary function of , so Section 16 already covers it. The usual notation abbreviates the same thing.
Proof
Differentiate by the product rule, which holds for the dot product since it is built from the geometric product by [eq:geometric-product]:
using the symmetry of the dot product and then [eq:newton-two]. The second statement is the fundamental theorem [eq:ftc] applied to that derivative over .
So the work and energy theorem is the fundamental theorem of calculus, applied along a trajectory. A quantity was differentiated and the total of the derivative recovered the change in the quantity.
Some forces admit a second description.
A force is conservative with potential energy , a function of position, when along every trajectory
For a conservative force the total energy is constant along every trajectory.
Proof
by [eq:work-energy] and [eq:potential], and a quantity with vanishing derivative is constant by Theorem [eq:mvt].
Two cases cover everything in this article.
The spring force is conservative with . More generally a central force , with , is conservative with any satisfying .
Proof
From , differentiating gives , so . Then by the chain rule
which is [eq:potential]. The spring is the case constant, where integrates to by [eq:ftc].
The quantity is a formula students are given. Here it is the result of an integration, and the integration is the reason the one half is there. A course without the integral must supply the one half as a fact to be remembered, alongside the in the kinetic energy and the in [eq:suvat], as three separate things. They are the same thing three times: the integral of a linear function.
26. One product and two conservation laws
Part I asked for a product of arrows and found that one demand determined it. The product had two parts, a number and an oriented area, and Section 7 had shown that school offers each part separately while pretending between them they cover the ground. Keeping both gives the following.
Multiply the second law [eq:newton-two] by the velocity and by the position, and take one grade from each:
A force perpendicular to the motion conserves the first. A force along the position conserves the second.
Proof
The first line is Theorem [eq:work-energy] and the second is Theorem [eq:L-dot], each already proved. The only new observation is that the two left sides are the grade-zero and grade-two parts of [eq:vF-split] and [eq:xF-split].
For the final sentence: if then , so is constant. If is parallel to then by Proposition [eq:wedge-props], so is constant.
Equation [eq:summit] is one statement. There is a single product of vectors with a scalar part and a bivector part. Contracted against the velocity, its scalar part is the rate of change of energy. Contracted against the position, its bivector part is the rate of change of angular momentum. Energy and angular momentum are the two grades of one multiplication, and the conditions for conservation are the two ways a product can lose a grade: the dot vanishes when the factors are perpendicular and the wedge when they are parallel.
A circular orbit under gravity conserves both at once. The force is central, so and the angular momentum is conserved. The force is also perpendicular to the velocity, so and the speed is constant. Both facts are usually presented as separate observations about orbits. They are the two grades of and vanishing for the same reason: the force lies along one vector and across the other.
The dot product and the cross product cannot say this. Each keeps one grade and discards the other, so each states one conservation law and is blind to the second. The whole product states both as one statement read at two grades.
Part IV
27. Tally
Four results were examined in the form a student meets them.
| Result | In the algebra-based course | Here |
|---|---|---|
| The kinematic formulas | Derived correctly, from a base lemma about the "simple average" that the course has no means to prove | Section 20, by integrating a constant twice, with [eq:average-velocity] supplying the missing base |
| Centripetal acceleration | Similar triangles, with three limits written as equalities and the word limit absent | Section 21, two differentiations of [eq:circular-trajectory] |
| The oscillator's cosine | Stated with no derivation, since the governing equation cannot be written down | Section 22, one component of a rotor, with the period proved |
| Kepler's equal areas | Stated and not derived, the text deriving "only the third one" and that only for circular orbits | Section 24, three lines from [eq:L-dot], for any central force |
Now count what each route requires a student to remember.
The algebra-based route requires five kinematic formulas together with the condition restricting them to constant acceleration, a memorised , a memorised sinusoid, a memorised period, a right-hand rule, a separate statement of each conservation law, and Kepler's laws as three facts about planets. Each arrives on its own authority, and none can be checked from anything else on the list.
The route taken here requires one law of motion, [eq:newton-two]; one product of vectors, [eq:geometric-product]; and two operations of the calculus that are inverse to one another. Everything else in Part III was derived, and each derivation is short enough to redo from memory.
The second list is shorter than the first. That is the argument of this article: the material a student must remember is smaller when it is connected, and the algebra-based course is longer to learn because a list of unrelated facts is harder to remember than a short argument.
A second difference is harder to count. Every derivation above stated its hypotheses and used them. Where a result applied only under a condition, the condition appeared in the statement, and Section 20 could say exactly how wrong the average-velocity formula becomes when its hypothesis fails, by [eq:trapezoid-error]. A course that supplies results ready-made cannot do this. Its results have no conditions, and a student cannot know when one of them stops applying. Knowing when a formula fails is most of what it means to understand it.
The order was number, then equation, then measurement, then position, then the product of arrows, then function, limit, derivative, integral, then motion. At no point was anything used before it was built.
So a reader who has arrived here did not need a prerequisite. There was never one to have.
28. Three dimensions
Everything above happened in the plane. Most of it transfers without alteration. The parts that change are named below.
The construction of Section 8 never used the dimension. One demand, , gives [eq:polarisation] and hence the split [eq:geometric-product] in any number of dimensions. What changes is what the algebra contains.
In three dimensions there are three independent planes, spanned by , and , so bivectors form a three-dimensional family where Section 9 had a one-dimensional one. There is also a grade-three object, the oriented unit volume , and the algebra has eight dimensions in total.
Two facts about the plane were flagged as accidents when they were used, and both now fail.
The first is that anticommutes with vectors. That made rotation one-sided in Theorem [eq:sandwich]. In three dimensions a bivector anticommutes only with the vectors in its own plane and commutes with the vector perpendicular to it, so the one-sided formula is wrong. The unit volume commutes with everything. The two-sided sandwich is the law in every dimension, and the plane's shorter form is a coincidence of the unit area anticommuting.
The second is the origin of the cross product. Only in three dimensions do the bivectors and the vectors both form three-dimensional families, so only there can the two be matched up. The matching is multiplication by the unit volume:
The cross product is therefore the wedge converted into a vector by an operation available only because three happens to equal three. In the plane no vector is available for the conversion, by Proposition [eq:no-cross-in-plane], and in four dimensions the bivectors form a six-dimensional family too large to match with vectors. The right-hand rule is the record of a choice of sign in [eq:cross-dual], and its bad behaviour under reflection is the sign of changing when the orientation does.
Everything in Part III carries over to three dimensions. The second law is unchanged, angular momentum stays the bivector , torque stays , and Theorem [eq:summit] applies word for word. The three-dimensional treatment adds rigid bodies, which need the extra planes, and gyroscopes, which the plane cannot have because it offers only one axis of spin.
That development is the subject of Classical Mechanics from Zero in Two Languages. That article builds the same algebra over a general inner-product space, runs the matrix and geometric languages in parallel and reaches rigid-body dynamics and the Lagrangian and Hamiltonian formalisms. Linear Algebra I develops the linear algebra underneath as a subject in its own right.
This article is a first course in mathematics and a first course in physics, and it is one course because they were never two.