Can a Single Particle Have a Temperature?

The one-particle average of (3/2)k_B T holds for any N. What N changes is fluctuation, and how much noise you accept determines how many particles you need.

Binary Buddies · 5 min read · 8 views

Textbooks often say that temperature is the average kinetic energy of the particles in a system. Take that literally and one particle has an average of one, so its kinetic energy should be its temperature. A question on Physics Stack Exchange tests this. Isolate a single atom in a perfect vacuum. It has a velocity and a kinetic energy. Does it have a temperature?

The question then asks the harder thing. If temperature belongs to an ensemble, where does the change happen? Two particles, ten, a hundred, a thousand? Is there a fundamental lower limit on system size, or only a practical one?

The two highest-voted answers agree on the core point. The statistical mechanics is cleaner than the textbook slogan suggests, and the answer to 'how many particles' is a number you choose, not one nature gives you.

Temperature is defined by entropy, not kinetic energy

The slogan has the logic backwards. As the 12-vote answer puts it, in statistical mechanics temperature is not defined from the average kinetic energy. It is defined by the thermodynamic relation 1/T = ∂S/∂E, which says how a system's entropy changes as its energy changes. The link to kinetic energy is something the theory derives. It is not the definition.

This matters for the single-atom question. A temperature defined through entropy is a property of a probability distribution over states. It is not a property of one trajectory. A lone atom moving at one fixed velocity has a kinetic energy. Whether it has a temperature depends on whether you are describing it with an equilibrium ensemble at all.

The single-particle result holds for any N

Here the slogan is more right than you might expect. The same answer points out that in classical statistical mechanics, the average kinetic energy of a single particle is proportional to the thermodynamic temperature. It can be proved with the virial theorem. It holds in any equilibrium ensemble, and it does not depend on the number of particles.

In the canonical ensemble, with a classical Hamiltonian split into kinetic energy K = Σ p_i²/2m and a potential V, the averages are:

  • Total kinetic energy: ⟨K⟩ = (3N/2) k_B T
  • One particle: ⟨K_1⟩ = (3/2) k_B T

So for one particle, averaged over the ensemble, the formula still holds exactly. The catch is the word 'averaged'. The relation tells you nothing about what a single measurement of that particle will show.

Fluctuations are where N enters

N does not change the averages. It changes how far individual values stray from them. In the canonical ensemble the variance of the total kinetic energy is ⟨(K − ⟨K⟩)²⟩ = (3/2) N (k_B T)². Divide its square root by ⟨K⟩ and the relative fluctuation scales as 1/√N. The answer drops the constant. Carried through, it is √(2/(3N)).

The one-particle kinetic energy behaves differently. Its fluctuation does not depend on N, and it is the same order as its average value. A single atom's kinetic energy is not a good estimate of T, however large the system around it. The total is a good estimate, and it gets better as N grows. Plugging values into that formula:

NRelative fluctuation of total K, √(2/(3N))
1≈ 82%
2≈ 58%
10≈ 26%
100≈ 8.2%
1000≈ 2.6%

These figures come from the formula above, not from any experiment. Notice what the table does not contain: a step where temperature 'switches on'. The noise shrinks smoothly. If you compute a temperature-like number from a handful of particles in your own code, this scaling tells you how noisy that number will be.

The thermodynamic limit, and why there is no magic number

The accepted answer names the idea the question is looking for: the thermodynamic limit. This is the framework that connects microscopic particle properties to macroscopic quantities such as temperature. The intuition is that the system must be large enough that fluctuations are negligible. Then the Law of Large Numbers holds as a good approximation.

That answer also says there is no fundamental, universal minimum particle count. 'Negligible' and 'good approximation' are judgements made for a given context. If you want to be rigorous, the steps are:

  1. Fix an upper bound on the fluctuations you will accept.
  2. Know the statistical distribution of the quantity you are averaging. For temperature, that is kinetic energy.
  3. Derive the smallest system size that keeps fluctuations within your bound.

So the lower limit is real, but you set it. Ask for 26% and ten particles are enough. Ask for under 3% and you need about a thousand.

Where this framing does not fit

Every result above is classical. The derivations use a classical Hamiltonian and the virial theorem. The thread does not address the quantum case, so do not carry these formulas over to systems where quantum effects dominate.

The fluctuation numbers come from the canonical ensemble specifically. The thread does not work through other ensembles, so treat the 1/√N scaling as a canonical-ensemble result. The single-particle average holding for any N is stated for any equilibrium ensemble. That leaves the question's original setup open. An atom moving alone in a vacuum with one fixed velocity is not obviously drawn from any equilibrium ensemble, and the averages only describe it if it is.

Finally, avoid both easy answers. 'One particle has no temperature' is too strong, since the ensemble average still gives (3/2) k_B T. 'One particle's energy is its temperature' is wrong, since a single sample fluctuates as much as its mean. The accurate statement is narrower: temperature is a parameter of a distribution, and N decides how well any one sample reveals it.

This article draws on If temperature is related to the average kinetic energy of particles, what does temperature actually mean for a single particle? and answers to it (1, 2), on the Stack Exchange network, licensed under CC BY-SA 4.0.

  • #canonical ensemble
  • #kinetic energy fluctuations
  • #statistical-mechanics
  • #temperature
  • #thermodynamic limit
  • #thermodynamics

Comments

Add a comment
Sign in with Google to join the conversation