Hilbert Space: Where Quantum States Actually Live
Heisenberg and Schrödinger built two quantum theories that looked nothing alike but always agreed. Von Neumann's 1927 answer, the state vector in Hilbert space, explained why.

Ask where an electron is before anyone measures it and quantum mechanics declines to answer in the terms you used. It does not hand back a position in ordinary three-dimensional space. It hands back a direction: an arrow pointing through an abstract space of possible outcomes. That space is Hilbert space. If you try to map that arrow onto a location, the theory stops making sense.
This is the first rule of the formal theory, and it trips up most people coming from classical intuition, engineers included. In classical physics a state is a list of current properties. In quantum mechanics a state is a vector that encodes every way the object might turn out when you look. A quantum system is not described by where it is. It is described by how its possibilities are weighted.
What follows draws on Quanta Magazine's explainer on Hilbert space. It traces how the idea arose, what the vector means, and the two ways it moves.
The arrow does not point at a place
Lucien Hardy, a physicist at the Perimeter Institute for Theoretical Physics in Waterloo, Canada, puts the distinction plainly. The arrows of quantum mechanics are not aimed at locations. 'It's a much more abstract space than that,' he told Quanta. They are 'really pointing in a direction in a possibility space.'
This is the part worth absorbing before anything else. A vector in Hilbert space has components, but those components are not x, y and z in a room. Each axis stands for a possible outcome of an observation. The direction of the arrow describes how the object's possible futures are weighted against one another.
Quanta frames this as the first commandment of quantum mechanics: do not think about ordinary objects moving through ordinary space. The theory instead predicts, in detail, all the ways an object might turn out to be. Hilbert space is the arena where those predictions live.
Two theories that gave the same answers
The problem that Hilbert space solved was historical before it was conceptual. Quantum mechanics arrived in two incompatible-looking forms within two years.
- Matrix mechanics (1925). Werner Heisenberg's version used tables, and in later formulations long towers of numbers, to calculate the odds that an electron in an atom would jump to a higher or lower orbit.
- Wave mechanics (1926). Erwin Schrödinger's version used waves to track quantities such as the probability of finding a particle at a given location.
The pictures looked nothing alike, yet they produced identical predictions. That is the kind of result that should make anyone suspicious. When two implementations with no shared code always return the same output, you assume there is a common specification underneath that neither of them states. Physicists had two incarnations of one theory and no account of what the theory itself was.
Von Neumann writes the specification
David Hilbert, the mathematician who had spent much of his career trying to rebuild physics on crisp axioms, found the question compelling. He got the young John von Neumann working on it in the mid-1920s. In 1927, von Neumann, then 23, solved it in a trilogy of papers he wrote alone, building on insights from Paul Dirac.
The historian of mathematics Leo Corry separates who supplied what: 'The ideas specifically were von Neumann's, but the inspiration — why do you axiomatize and what for — this is something that he took from Hilbert.'
The papers defined the theory's central objects and the rules they obey. Their key result was that Heisenberg's towers of numbers and Schrödinger's waves are two notations for the same entity. Quanta's analogy is 0.5 and ½: different marks, one point on the number line. That entity is the quantum state. The early pioneers had not realised at first that their mathematics had left the real world behind. Von Neumann named the space it had moved into.
What a quantum state actually encodes
Everything has a state. A coin reads heads or tails. A grandfather clock's bob can sit at any of a continuum of positions as it swings. Most of physics consists of capturing a state and predicting how it changes.
Von Neumann's rules complicate this. Before observation, a quantum object has no fixed set of properties, such as a definite position. It has a superposition: a combination of possible properties that has no classical counterpart. For a particle, the superposition combines all the places it might end up being found.
A superposition is not the same thing as ignorance, and it is not vague. Quanta's example is a superposition in which there is a 99% chance of finding the particle to your left and a 1% chance of finding it to your right. You know how to place your bet. You do not know whether you won until you check. Von Neumann represented this weighted set of possibilities as a single vector, and the space of all such vectors for a given object is its Hilbert space.
A three-axis worked example
Quanta's illustration is a quantum traffic light with three possible outcomes: red, yellow, green. Its state vector lives in a three-dimensional Hilbert space, with one axis for each colour it could show.
Until someone looks, the light has no colour. It has a mixture of possible colours, so its arrow points into the interior of the space rather than along any single axis. The closer the arrow sits to the red axis, the more likely the light is to show red when observed. The same holds for yellow and green.
For readers who think in arrays, here is a rough sketch. The amplitudes are made up for illustration. Squaring a component to get a probability is the standard textbook rule.
import numpy as np
# axes: red, yellow, green
state = np.array([0.8, 0.36, 0.48])
state = state / np.linalg.norm(state) # states are unit vectors
probs = state ** 2 # weight on each outcome
print(dict(zip(['red', 'yellow', 'green'], probs.round(3))))
The main thing the sketch makes concrete is that the dimension of the space equals the number of distinguishable outcomes. Three colours gives three axes. According to von Neumann's first rule, the state of any object, from an electron to a galaxy, is captured by a vector like this one in a suitable Hilbert space. For most real objects that space is enormously larger.
Two ways the arrow moves
Von Neumann's remaining axioms say how a state changes. There are exactly two modes, and they behave very differently.
- Smooth evolution, between observations. As the world acts on the object, the arrow rotates continuously through Hilbert space. The traffic light's vector might drift toward the green axis, raising the odds of green, or swing toward red or yellow. This motion is smooth and predictable.
- Measurement, at observation. When you actually observe the system, the vector instantly and randomly snaps onto one axis: red, yellow or green. The more closely aligned it was with an axis, the more likely it is to land there rather than on the others.
The asymmetry is the point. The first mode is deterministic. Given the influences on the system, you can compute where the arrow goes. The second mode is not. Alignment sets the odds, but which axis the vector actually lands on is random. A model that gets the rotation right will still only ever give you a distribution over outcomes, never a single guaranteed result.
Why the abstraction paid off
Defining quantum mechanics as a Hilbert space did more than settle the dispute between two notations. According to Quanta, exploring the structure of Hilbert space led physicists to a deeper and more unified understanding of quantum physics.
The engineering analogy holds up. Once you have a single specification, the competing implementations become interchangeable views of it, and you can reason about the specification directly instead of translating between dialects. Heisenberg's numbers and Schrödinger's waves did not disappear. They became two coordinate choices for the same vector, each convenient for different problems.
Where the picture misleads, and what to watch for
The traffic-light model is useful precisely because it is small, and that same smallness is where it can mislead. Some cautions:
- Not every system has three tidy axes. The traffic light has a finite set of outcomes. Position, as in Schrödinger's waves, can take a continuum of values, and the clock bob in Quanta's own analogy hints at this. Intuition built on a three-dimensional picture does not carry over automatically to those cases.
- 'Any object, from an electron to a galaxy' is a statement of principle. The first rule says such a vector exists. It does not mean anyone can write it down or compute with it for a galaxy.
- Do not read the arrow spatially. As Hardy stresses, the axes are outcomes, not directions in a room. Most beginner confusion comes from forgetting this.
- This is an outline, not the full formalism. It does not cover how the smooth rotation is calculated, or how measurement is formalised beyond the snap-to-an-axis picture. Treat this article as an orientation to the framework rather than a working method.
The durable takeaway is von Neumann's first rule itself. A quantum object's state is a direction in a space of possibilities. It rotates smoothly until it is observed, and then it lands, at random, on one outcome.
- #Hilbert space
- #John von Neumann
- #matrix mechanics
- #quantum state
- #state vector
- #superposition



Comments