Why only square arrays get this number, and what it settles about a linear system

0:00Two equations in two unknowns. Each one is a straight line, and there are only three things two straight lines can do. They cross once. They never meet. Or they are the same line twice over, and meet everywhere. Here is the claim this whole topic is built on. You can tell which of those three you are in before you solve anything -- and you can tell it from the four coefficients alone.

0:27Not from the right-hand sides. Not from a graph. From four numbers, one multiplication, another multiplication, and a subtraction. That is a strong claim, so it deserves to be tested rather than announced. Write the two equations with lettered coefficients and pull the four of them out into a square array. a and b in the first row, c and d in the second. Now multiply along one diagonal: a times d.

1:00Multiply along the other: b times c. Subtract the second from the first. a d minus b c. That is the number. One number, out of four, and it never once looks at the right-hand sides. The claim is that this number is not zero exactly when the two lines cross at a single point. Notice how little it is doing. No solving, no rearranging, no graph. Two products and a subtraction.

1:31Take a pair with actual numbers in it. Two x plus five y equals one, and three x plus two y equals seven. The four coefficients are two and five above three and two. Two times two is four. Five times three is fifteen. Four minus fifteen is minus eleven. Not zero. So the claim says: one crossing point, exactly one. Solve it the long way and you get x equals three, y equals minus one. One point, as promised.

2:07Now a second pair. One x plus three y equals five, and two x plus six y equals eight. One times six is six. Three times two is six. Six minus six is zero. And these two lines never meet. The left side of the second equation is exactly twice the first, but the right side is not. Same slope, different height. Parallel, and no solution anywhere. You may have met this test before in a different costume.

2:41Compare the ratio of the two x-coefficients with the ratio of the two y-coefficients. If those ratios differ, the lines cross. That is the same test. Clear the denominators and the ratios differing becomes exactly a d minus b c not being zero. Two forms, one statement, and students routinely meet them years apart without ever being told they are one thing. But the two forms are not equally usable, and it is worth seeing why.

3:16Of the two hundred and fifty-six arrays built from minus one, zero, one and two, the ratio form can be written down for a hundred and forty-four. For the other hundred and twelve it cannot, because a denominator is zero. On the hundred and forty-four where both forms can be asked, they agree every single time. Zero disagreements. So the cross-difference form is not a different test. It is the same test with the division taken out, which is why it always answers.

3:54Now, that claim about crossing points is true. Which makes it dangerous. If I compute a d minus b c and then tell you the answer is right, I have shown you nothing, because I decided both halves. So let us do something harder. Take six different rules that turn four entries into one number, and ask all six the same question. Multiply across and subtract, which is the real one. Multiply across and ADD. Add the two entries on the diagonal.

4:29Subtract the other way round. Multiply all four entries together. And take the first row's total minus the second's. Every one of those is a perfectly definite recipe. Four numbers in, one number out. Now four thousand and ninety-six pairs of equations: all two hundred and fifty-six coefficient arrays, against all sixteen right-hand sides. Each one gets classified the slow, honest way -- by row reduction, comparing ranks, with no deciding number anywhere on that path.

5:08Three thousand and forty cross at exactly one point. Eight hundred and twenty-seven have no solution. Two hundred and twenty-nine have infinitely many. And now we ask each of the six rules: does your number being non-zero match that verdict? Multiply across and subtract: four thousand and ninety-six agreements. Zero disagreements. It is right every time. Multiply across and ADD: three thousand five hundred and sixty-eight right, five hundred and twenty-eight wrong.

5:46And the errors go both ways -- two hundred and seventy-two times it promises a single crossing point that is not there, two hundred and fifty-six times it denies one that is. Add the two diagonal entries: two thousand six hundred and twenty-four right, one thousand four hundred and seventy-two wrong. Multiply all four entries: one thousand eight hundred and eight right, two thousand two hundred and eighty-eight wrong. That one fails in a particular way. Two thousand and sixteen of its errors are denials -- any array with a single zero in it makes its product vanish, whatever the lines are doing.

6:32The first row's total minus the second's: three thousand and forty right, one thousand and fifty-six wrong. Four of those five are wrong somewhere, and one is right every time. The sixth rule's numbers we have not put on the board yet, and that is deliberate. Because the rule we skipped passes too, and the reason is worth a detour. Look at the fourth rule: subtract the other way round. b c minus a d, instead of a d minus b c.

7:08It also gets four thousand and ninety-six agreements and zero disagreements. Of course it does. It is the negative of the real rule, and a number is zero exactly when its negative is. So on the question the topic asks, that rival is perfect. On the number itself, it is wrong almost always. Across those two hundred and fifty-six arrays, the two rules hand back the same value only sixty-six times -- exactly the arrays where the value is zero, and nowhere else.

7:45Take the array holding two and two above four and zero. The real rule gives minus eight. The rival gives plus eight. Same verdict, opposite number. Which tells you something the crossing question alone cannot: this number is going to be used for more than a yes or a no. Everything later in the subject -- areas, inverses, whole solution formulas -- uses the value and its sign, not just whether it vanished.

8:18So we should name the thing. The number is called the determinant of the array. And the framing that matters is this: it is a rule that takes in a square array and hands back one number. Every square array of a given size gets a number attached to it. The determinant of A is a value, not an action. That distinction earns its keep almost immediately, because there turn out to be several quite different ways to compute the same value.

8:50If the determinant is a value, six recipes agreeing is unremarkable. If it were an action, six recipes would be six different things. There is a notational point too, and it is not decoration. The array itself is written in square brackets. Its determinant is written in vertical bars. Brackets: an array of numbers. Bars: one number. Same entries on the page, two completely different objects. A drawing that puts brackets round something while the voice says determinant undoes that distinction in a single shot.

9:32The rule accepts a square array and turns away everything else. That is usually stated and almost never defended, so let us defend it. Offer the rule the sixty-four arrays with two rows and three columns. It accepts none of them. Offer it the sixty-four with three rows and two columns. None of those either. Offer it the two hundred and fifty-six square ones and it accepts all two hundred and fifty-six.

10:03And it does not stop at size two. Give it a three by three and it answers: for the array holding one, two, three above four, five, six above seven, eight, ten, the number is minus three. So squareness is the condition, not smallness. Why squareness? Go back to where the number came from. Two equations in two unknowns. As many equations as unknowns, so the coefficients came out square.

10:35A three-by-two array of coefficients would be three equations in two unknowns. That is not a harder version of the same question. It is a different question. It is worth seeing exactly how different, because this is the step people skip. Take three equations in two unknowns: x equals one, y equals one, and x plus y equals three. Pull out any two of those three and you get a square array. All three of those two-row arrays have a non-zero determinant.

11:12Every pair of the three lines crosses at exactly one point. And the three equations together have no solution at all. Change one right-hand side -- x plus y equals two instead of three -- and suddenly there is exactly one solution, with not a single coefficient touched. So for a non-square system the coefficients genuinely do not settle it. The right-hand sides get a vote. That is why the rule refuses. It is not being fussy. It has nothing to say about that question.

11:51Now the single most expensive misreading in this whole subject, and it is caused by punctuation. The determinant is written between two vertical bars. Those are the same two marks used for the absolute value of a number. They are not the same thing, and the giveaway is that a determinant can be negative. That array from earlier -- two and two above four and zero -- has determinant minus eight. An absolute value is never minus eight.

12:24And it is not a rare accident. Of the two hundred and fifty-six arrays in this population, ninety-five have a negative determinant. Ninety-five values that no absolute value could ever produce. The two readings even collide on the smallest possible case. A one by one array holding a single entry has that entry as its determinant. Take the five entries minus two, minus one, zero, one and two. On two of them -- the two negative ones -- the determinant and the absolute value disagree.

13:03So the bars mean the determinant of. They are not a modulus. Read them as one and you will get the sign wrong for the rest of the subject. One more thing to be careful about, and it is the opposite mistake. The number being non-zero settles the question completely: exactly one solution. The number being zero settles much less than people assume. Of the four thousand and ninety-six pairs, one thousand and fifty-six have a vanishing number.

13:39Of those, eight hundred and twenty-seven have no solution at all, and two hundred and twenty-nine have infinitely many. And exactly zero of them have a single solution -- so the rule never lies. But it does not choose. A vanishing number rules out one outcome and rules in two, and it cannot tell you which. Worse: the coefficients alone genuinely cannot tell you. Sixty-six of those arrays appear in BOTH counts -- with one set of right-hand sides they give no solution, with another they give infinitely many.

14:18Take our parallel pair and change the eight on the right to a ten. The determinant is untouched, still zero, and now the two lines are the same line and there are infinitely many solutions. So: not zero means one point. Zero means not one point, and you have more work to do. Everything so far has been counted over a particular collection of arrays with small whole-number entries. That is a deliberate choice and it has a cost.

14:52A count over a finite collection is evidence. It is not a proof about all arrays. So run the rule once on four distinct symbols and read the answer as an expression rather than a number. a d minus b c. Each of the four letters appears exactly once, to the first power. The rival that adds gives a d plus b c -- the same two products, and a genuinely different expression.

15:22That is the identity, and it holds whatever the entries are. Which raises the question of what the entries are allowed to be. In general: any numbers at all, real or complex. In practice, at this stage, real entries and arrays no bigger than three by three. Say the general case once, then stay where the work is. And notice what the symbols also show. Put the same letter in two corners and it comes back squared -- so the rule is not merely shuffling entries around, it is genuinely multiplying them.

16:02So take stock of what this one number is and is not. It is a rule from square arrays to numbers. It refuses everything that is not square, and it refuses it on grounds of shape, before any entry is read. Its value is a signed number, not a size, and the sign will matter later. It is not zero exactly when a pair of equations has one solution -- verified four thousand and ninety-six times out of four thousand and ninety-six, against five rival rules that each fail somewhere.

16:40And when it is zero it tells you honestly that it does not know which of two things happened. That is a remarkable amount of information for two multiplications and a subtraction. What it does not give you is the solution itself. Getting the actual numbers out takes a good deal more machinery, and that machinery is what comes next. The number arrives first, the method arrives later, and knowing which is which is most of what makes this subject navigable.