Lists, Vectors, and Normalization
A vector is just a list of numbers in a fixed order. A qubit's state is a two-number list of amplitudes. The list must obey one rule: the squares of its entries must add up to exactly 1. That rule is just "the chances must total 100%" in disguise.
What is a vector?
You already use vectors every week. A shopping list — 2 apples, 3 eggs, 1 loaf — is a list of numbers in a fixed order. The order matters: swap the 2 and the 3 and you come home with the wrong bag.
A GPS location is one too: (28.61, 77.21) is two numbers in a fixed order, latitude first, longitude second. Swap them and you are in the ocean.
That is the entire definition. A vector is a list of numbers in a fixed order. Nothing else. When math books draw arrows, they are drawing a picture of a vector; the vector itself is just the list. We will write lists in square brackets, like [0.6, 0.8].
How is a qubit's state a list of numbers?
Measure a qubit and you get 0 or 1. Before you measure, the qubit carries one amplitude for each possible reading. An amplitude is a number you square to get a chance. These are the "amounts that can cancel" from chapter 3. So a qubit's state is a two-entry vector: [amplitude for reading 0, amplitude for reading 1].
A fresh qubit is [1, 0]. All of the amount sits on 0 and none on 1, so every shot reads 0. After the fair-coin H gate from chapter 2, the state is [0.7071, 0.7071]. After the RY(π/3) turn from chapter 4, it is [0.8660, 0.5].
More qubits just means a longer list, with one entry per possible reading. Two qubits have four readings (00, 01, 10, 11), so the list has four entries. Three qubits need eight entries. In general, n qubits need a list with 2^n entries. That is the doubling explosion from chapter 1.
So 30 qubits already need a list about a billion entries long. That long list is part of what makes quantum computers interesting. It is also why ordinary computers that imitate them (called simulators) hit a wall.
What is the normalization rule?
Chapter 2 opened with one bedrock fact: the probabilities of all possible outcomes add up to 100%, that is, to 1. And squaring an amplitude gives that outcome's probability. Put those together and every legal state vector must pass one test:
Square every entry. The squares must add up to exactly 1.
This rule is called normalization. It is not a new law of nature — it is "probabilities total 100%" wearing vector clothes.
Worked example 1, digit by digit. Is [0.6, 0.8] a legal qubit state?
- 0.6 × 0.6 = 0.36
- 0.8 × 0.8 = 0.64
- 0.36 + 0.64 = 1.00 — legal. This qubit reads 0 on about 36% of shots and 1 on about 64%.
Check the fair coin [0.7071, 0.7071] too: 0.7071 × 0.7071 = 0.4999904…, which is exactly 0.5 once you use the full unrounded value 0.70710678… So the squares give 0.5 + 0.5 = 1. Legal.
One more thing squares are good at: ignoring signs. (−0.7071) × (−0.7071) is also +0.5, because a negative times a negative is positive — chapter 3's rule. The negative amplitudes that made cancellation possible are fully legal citizens of these lists.
How do you fix a list that fails the test?
The list [0.5, 0.5] looks innocent. Test it: 0.5 × 0.5 = 0.25, and 0.25 + 0.25 = 0.50. The squares total only 50%. Half the probability is missing, so this is not a legal state.
The repair is always the same: divide every entry by the list's length. The length is the square root — the un-squaring operation from chapter 1 — of the sum of the squares.
Worked example 2, digit by digit:
- Sum of squares: 0.25 + 0.25 = 0.50.
- Length: the square root of 0.50 is 0.7071, because 0.7071 × 0.7071 = 0.4999904… ≈ 0.50.
- Divide each entry by the length: 0.5 ÷ 0.7071 = 0.7071. (A charming habit of this particular number: dividing 0.5 by it hands it right back, precisely because its square is 0.5.)
- The repaired list is [0.7071, 0.7071]. Check: 0.5 + 0.5 = 1.00. Fixed.
This repair is called normalizing the vector. It is the same move as converting raw vote counts into percentages: you divide by the total so that everything accounts for exactly 100%.
How do independent qubits combine into one list?
Remember the two coins in chapter 2: independent events multiply, so heads-heads had probability 0.5 × 0.5 = 0.25, one cell of a 2-by-2 table. Amplitudes follow the same multiplication when qubits are independent.
Give each of two qubits the fair-coin treatment, so each is [0.7071, 0.7071]. The combined state lists all four readings — 00, 01, 10, 11 — and each entry is the product of the matching single-qubit amplitudes: 0.7071 × 0.7071 = 0.5, every time. The combined vector is [0.5, 0.5, 0.5, 0.5]. It is chapter 2's table again, flattened into a list.
Normalization check: each square is 0.5 × 0.5 = 0.25, and four of them make 0.25 + 0.25 + 0.25 + 0.25 = 1.00. Legal.
So the prediction writes itself: each of the four readings has probability exactly 0.25, which means about 250 of each in 1000 shots. Let's measure the list.
Run it: is [0.5, 0.5, 0.5, 0.5] real?
Where will you use this?
From now on, "the state" always means "the list". Normalization is the first check on any state you write down. You will use this in:
- Probability amplitudes, which tells the full story of the entries in these lists.
- Multiple qubits, where the 2^n-entry list becomes the main character. That chapter includes states that cannot be built by entry-by-entry multiplying.
- The next chapter, where repeated multiplying stops being a handy trick. It becomes the reason quantum hardware is hard to build.
- Amplitudes and phase, a quick summary you can come back to any time.
- The Lab, where you can look at the state list of any circuit you build. Watch the squares add up to 1.