Negative temperature is a thermodynamic condition in which a system’s entropy decreases as its internal energy increases, giving it a negative absolute temperature. It can occur in suitably isolated systems whose accessible energies have an upper bound, such as certain collections of magnetic spins. Negative temperatures are not colder than absolute zero: in the direction of spontaneous heat flow, they are hotter than every positive temperature. They are also distinct from ordinary negative Celsius or Fahrenheit readings, which usually correspond to positive absolute temperatures.
Thermodynamic definition
For a system in equilibrium, temperature is defined by
where is entropy, is internal energy, and volume and particle number are held fixed. Other external parameters must also be held fixed when relevant.
In most familiar systems, adding energy increases the number of accessible microscopic arrangements, so entropy increases and . In a bounded-energy system, however, entropy can reach a maximum at an intermediate energy and then decrease as energy approaches its upper limit. On this decreasing branch,
so .
This definition arises within statistical mechanics and depends on specifying the accessible states and the equilibrium description being used. Negative temperature is not a general property of any highly excited or externally driven system.
Inverse temperature and the ordering of hotness
It is often convenient to use the inverse-temperature parameter
where is the Boltzmann constant. Unlike , this parameter passes continuously through zero when the entropy reaches its maximum.
For a typical bounded system, increasing energy carries the system through the sequence
The two infinite-temperature limits meet at . The notation denotes the limiting negative-temperature state near the top of the energy spectrum; it does not mean a state colder than positive absolute zero.
Temperature therefore does not order thermal hotness by ordinary numerical comparison across its positive and negative branches. Inverse temperature provides the more useful ordering: energy tends to pass from smaller to larger .
Statistical interpretation
In a canonical equilibrium description, the probability of occupying a microscopic state of energy follows the Boltzmann distribution:
where is the partition function.
At positive temperature, , and individual higher-energy states have smaller probabilities than lower-energy states. At negative temperature, , so the weighting reverses: higher-energy states have larger probabilities. This reversal is called population inversion.
For energy levels with different degeneracies, the total population of a level also depends on how many microscopic states share its energy. Population inversion must therefore be interpreted with this multiplicity taken into account.
Why an upper energy bound matters
For negative , the factor grows with energy. If accessible energies extend without limit, the partition function ordinarily diverges and cannot define a normalized equilibrium distribution.
This excludes negative-temperature equilibrium for an ordinary gas with unrestricted kinetic energy:
Its momentum, and hence its kinetic energy, has no upper bound.
An upper energy bound makes negative temperature possible, but is not by itself sufficient. The accessible state space must also allow a normalizable distribution, and the system must be capable of reaching the relevant internal equilibrium. Degrees of freedom with unbounded energies must be absent, constrained, or sufficiently decoupled over the observation time.
A two-level example
Consider distinguishable, weakly interacting units, each with energy either or . If units occupy the upper level, the total energy is
The number of arrangements with that energy is
Using the Boltzmann entropy and a large- approximation gives
There are three regimes:
- Fewer than half excited: , so .
- Equal populations: , so , corresponding to infinite temperature.
- More than half excited: , so .
Entropy is greatest when the populations are equal. Beyond that point, further excitation reduces the number of possible arrangements. Near complete excitation, there are few ways to distribute the energy because nearly every unit must occupy the upper level.
The same population relation follows from
A ratio greater than one requires negative temperature.
Physical realization
Magnetic spin systems
A standard realization uses spins in a magnetic field. For a two-level spin degree of freedom, the magnetic interaction provides a lower and an upper energy level. A collection of such spins therefore has a bounded spin-energy spectrum.
Interactions between spins can redistribute energy and establish internal equilibrium. If this equilibration is much faster than energy exchange with the surrounding solid, the spin subsystem can possess a well-defined temperature different from the temperature of the material’s atomic motion.
Preparing an inverted population—for example, through a suitable field change or resonant pulse sequence—can produce a negative spin temperature. The solid as a whole need not have negative temperature: its vibrational degrees of freedom can remain at positive temperature.
This distinction between subsystem and whole-system temperature is central to experiments involving nuclear magnetic resonance.
Ultracold atoms in optical lattices
An optical lattice is a periodic potential produced by interfering laser beams. Within an isolated lattice band, atomic motion can have a bounded kinetic-energy spectrum rather than the unbounded spectrum of free particles.
In 2013, Simon Braun and collaborators reported negative absolute temperatures for motional degrees of freedom in an ultracold potassium atomic gas. The preparation used a lattice and controlled changes to interactions and confinement so that the relevant effective energy spectrum supported an inverted distribution.
Such experiments require attention to more than kinetic energy. Interaction and trapping energies must also be compatible with the bounded-energy description. Excitation into higher bands or coupling to uncontrolled degrees of freedom can eventually destroy the negative-temperature state.
Population inversion in lasers
A laser requires population inversion between relevant energy levels. Under suitable assumptions, the population ratio for a pair of levels can be assigned an effective negative temperature.
However, an operating laser is generally a driven, dissipative system with several levels and ongoing energy flows. An inverted transition does not establish that the entire laser medium is in negative-temperature thermodynamic equilibrium. A temperature inferred from one population ratio is not necessarily shared by other transitions or degrees of freedom.
Heat flow and the second law
Negative temperatures are consistent with the second law of thermodynamics.
Suppose an amount of energy passes from system A to system B. The corresponding total entropy change is
If and , the expression is positive. Energy can therefore flow spontaneously from the negative-temperature system to the positive-temperature system. This is the precise meaning of saying that negative temperatures are hotter than all positive temperatures.
A negative-temperature subsystem generally cannot remain indefinitely at that temperature when coupled to an ordinary positive-temperature environment. It transfers energy to the environment and loses its inversion.
Negative temperature also does not automatically imply negative heat capacity. For a canonical system with a temperature-independent energy spectrum,
This relation holds for either sign of temperature wherever the canonical distribution exists.
Heat engines
Formally substituting a negative hot-reservoir temperature and a positive cold-reservoir temperature into the reversible Carnot efficiency expression,
gives a value greater than one.
This does not violate energy conservation. In this unusual reversible cycle, the device absorbs energy from both reservoirs rather than taking energy only from the designated hot reservoir and rejecting part of it to the cold one. The ratio of work output to heat absorbed from the negative-temperature reservoir alone can exceed one because that denominator excludes the additional energy absorbed from the positive-temperature reservoir. Preparing and restoring an inverted reservoir also requires resources.
Historical development
The possibility of negative absolute temperature emerged from the statistical treatment of bounded-energy systems.
In 1951, Edward Purcell and Robert Pound reported negative spin temperatures in lithium fluoride using nuclear magnetic resonance techniques. Their experiment established that a nuclear-spin subsystem could display inverted populations described by negative absolute temperature.
In 1956, Norman Ramsey developed a systematic thermodynamic treatment of negative-temperature states, explaining their relation to entropy, equilibrium, and the second law.
Later work extended the subject beyond conventional spin systems. The 2013 ultracold-atom experiment was notable for realizing negative temperature in motional degrees of freedom within a carefully engineered lattice system.
Entropy definitions and theoretical debate
The interpretation of negative temperature is connected to a debate over entropy definitions for isolated systems, particularly those with bounded spectra.
The Boltzmann surface entropy counts states in a narrow energy interval:
If the number of states in that interval decreases at high energy, its derivative becomes negative and yields negative temperature.
The Gibbs volume entropy instead counts all states with energy at or below :
Because this cumulative count cannot decrease as increases, the temperature obtained from its derivative does not become negative.
The disagreement concerns which entropy definition supplies the appropriate thermodynamic description, and under what conditions. It is especially significant for finite isolated systems and for comparisons between different statistical ensembles.
The experimentally observed population inversions are not dependent on resolving this terminology. They can be measured directly and, where equilibrium assumptions are appropriate, described by canonical distributions with . The disputed issue is how such states should be incorporated into a general thermodynamic framework.
Scope and limitations
A meaningful negative-temperature description requires several distinctions:
- Absolute versus conventional scales: a negative Celsius reading is not evidence of negative absolute temperature.
- Equilibrium versus inversion alone: an arbitrarily prepared high-energy population need not have a thermodynamic temperature.
- Subsystem versus whole system: negative spin temperature can coexist with positive temperature in other degrees of freedom.
- Bounded effective dynamics versus unrestricted motion: a lattice-band description may support negative temperature even though access to higher bands would invalidate that description.
- Thermal state versus coherent state: a system prepared in a particular quantum superposition is not necessarily described by a thermal distribution.
Negative temperature is therefore a property of particular statistical states under identifiable physical constraints, not a way of cooling ordinary matter below absolute zero.