Negative temperature
In physics and thermodynamics, negative temperature describes a counterintuitive state in certain isolated systems where the thermodynamic temperature falls below absolute zero on the Kelvin scale, arising when the entropy decreases as internal energy increases.[1] This phenomenon requires systems with a bounded energy spectrum, such as spin-1/2 particles in a magnetic field, where population inversion occurs—more particles occupy higher-energy states than lower-energy ones, inverting the typical Boltzmann distribution.[2][3] The concept stems from the fundamental thermodynamic relation T = \left( \frac{\partial E}{\partial S} \right)_{V,N}, where T is temperature, E is internal energy, S is entropy, V is volume, and N is particle number; negative temperatures emerge when \frac{\partial S}{\partial E} < 0, which happens after the system's entropy reaches a maximum at infinite positive temperature and begins to decline toward a finite maximum energy.[1][3] In such states, the system behaves as hotter than any positive temperature, including infinite temperature, because when placed in thermal contact with a positive-temperature system, heat flows from the negative-temperature system to the positive one, consistent with the second law of thermodynamics.[2][1] Negative temperatures were theoretically formalized in the mid-20th century and first experimentally realized in nuclear spin systems using radiofrequency techniques to achieve inversion without external pumping.[3] Modern experiments have extended this to larger scales, such as ultracold atomic gases of about 100,000 atoms confined in optical lattices formed by intersecting laser beams, where magnetic fields enable stable population inversion and negative temperatures.[2] These systems remain in internal thermal equilibrium but cannot be indefinitely maintained in contact with a heat bath, as they would relax to positive temperatures.[1] Key applications include precision measurements in nuclear magnetic resonance (NMR) calorimetry and enhancing the performance of masers and amplifiers, where the inverted populations amplify signals.[3] Despite their "hotter" nature, negative-temperature states are transient and limited to specific non-interacting or weakly interacting systems, underscoring their distinction from everyday thermal phenomena.[2]Fundamentals
Thermodynamic Definition
In thermodynamics, the temperature T of a system is defined by the relation T = \left( \frac{\partial U}{\partial S} \right)_{V,N}^{-1}, where U is the internal energy, S is the entropy, V is the volume, and N is the number of particles (or composition). This definition follows from the fundamental thermodynamic relation dS = \frac{1}{T} (dU + P dV - \mu dN), which integrates to the above form for processes at constant V and N, ensuring that temperature quantifies the rate of entropy change with energy in equilibrium. Negative temperatures arise when \left( \frac{\partial S}{\partial U} \right)_{V,N} < 0, meaning the entropy decreases as internal energy increases, inverting the usual monotonic increase of S with U. This condition is physically realizable only in systems with a bounded energy spectrum, where the possible energy values are limited from above (and typically from below), such as those with a finite number of accessible states. In such systems, the entropy S = k \ln \Omega, with k Boltzmann's constant and \Omega the number of microstates, reaches a maximum S_{\max} at an intermediate energy U^*, corresponding to the configuration of highest multiplicity. For U > U^*, further energy input reduces \Omega, causing S to decline toward zero at the upper energy bound, thereby yielding negative T.[4] The thermodynamic temperature scale accommodates negative values seamlessly, extending beyond positive infinity. As U approaches U^* from below, T \to +\infty; at U = U^*, the inverse temperature \beta = 1/(kT) = 0, corresponding to infinite temperature; and for U > U^*, T jumps to -\infty and increases toward zero (with \beta < 0) as U nears its maximum. This ordering implies that negative-temperature states are "hotter" than any positive-temperature state, as heat spontaneously flows from a negative-T system to a positive-T one until equilibrium.[4] In statistical mechanics, the inverse temperature admits a direct microcanonical formulation: \beta = \frac{1}{kT} = \frac{\partial \ln \Omega}{\partial U}, derived by considering the entropy's logarithmic dependence on the density of states and maximizing \Omega subject to fixed U. Negative T thus corresponds to negative \beta, occurring precisely when \frac{\partial \ln \Omega}{\partial U} < 0, beyond the energy of maximum \Omega. This statistical perspective unifies the thermodynamic definition with the underlying microstate counting, confirming the validity of negative temperatures within equilibrium statistical mechanics for bounded systems.[4]Relation to Entropy and Disorder
In statistical mechanics, entropy S is defined as S = k \ln \Omega, where k is Boltzmann's constant and \Omega is the number of accessible microstates, serving as a quantitative measure of a system's disorder or multiplicity. At negative temperatures, the system exhibits an inverted population distribution where high-energy states are more occupied than low-energy ones, leading to a reduction in \Omega despite the high total energy U; this counterintuitive decrease in disorder occurs because the phase space is bounded above, concentrating the system in fewer high-energy configurations.[4] Such states represent low-entropy, high-energy equilibria, which are transiently achievable through non-equilibrium processes like sudden energy injections.[2] The relationship between entropy and temperature is captured by the thermodynamic identity \frac{1}{T} = \left( \frac{\partial S}{\partial U} \right)_{V,N}, where negative T arises when \frac{\partial S}{\partial U} < 0, meaning entropy decreases as internal energy increases. Plotting S versus U yields a concave-down curve that rises from zero entropy at the ground state (U_{\min}), peaks at infinite temperature where all states are equally likely, and then declines toward a maximum energy (U_{\max}) on the high-U side; the negative temperature regime corresponds to this descending branch, where additional energy reduces the available microstates.[4] This peaked structure reflects the bounded density of states in suitable systems, such as spin ensembles, ensuring the curve's non-monotonic behavior.[4] Physically, negative temperature states are consistent with the second law of thermodynamics because they are not equilibrium states in isolation but evolve toward higher-entropy configurations by transferring heat to surrounding positive-temperature systems, thereby increasing overall entropy.[2] In contrast to positive temperatures, where increasing U broadens the energy distribution and heightens disorder (rising S), negative temperatures invert this: energy addition sharpens the distribution toward the upper bound, diminishing disorder and entropy as the system approaches a fully inverted, ordered state.[4] This inversion underscores why negative temperatures are "hotter" than any positive value, facilitating spontaneous heat flow from them to cooler reservoirs.Theoretical Foundations
Energy Distribution and Population Inversion
In thermal equilibrium, the distribution of particles across energy levels in a system is described by the Boltzmann distribution, where the probability p_i of occupying a state with energy E_i is proportional to e^{-E_i / kT}, with k being the Boltzmann constant and T the temperature. This canonical form arises from maximizing entropy subject to fixed energy and particle number constraints in statistical mechanics. For negative temperatures (T < 0), the exponent becomes positive because $1/T < 0, resulting in p_i \propto e^{|E_i| / |kT|}. Consequently, higher-energy states become more populated than lower-energy ones, inverting the usual thermal ordering observed at positive temperatures. This inverted distribution is a hallmark of negative temperature states and requires systems where energy is bounded from above to prevent divergence.[5] Population inversion refers to a configuration where the occupancy of an excited state exceeds that of a lower-energy state, such as more particles in a higher-energy level than in the ground state.[6] In the context of negative temperatures, this inversion emerges naturally from the equilibrium Boltzmann statistics, enabling processes like optical amplification where stimulated emission dominates absorption.[6] Consider a simple two-level system with energies E_\text{lower} and E_\text{upper} where E_\text{upper} > E_\text{lower}. The ratio of populations is given by \frac{n_\text{upper}}{n_\text{lower}} = e^{-(E_\text{upper} - E_\text{lower}) / kT} = e^{(E_\text{lower} - E_\text{upper}) / kT}. For T < 0, the exponent is positive, yielding n_\text{upper} / n_\text{lower} > 1, confirming population inversion. This concept extends to multi-level systems, where negative temperatures and population inversion are possible provided the energy spectrum is bounded from above, ensuring the partition function remains finite, and the system can be driven by external means such as optical or magnetic pumping to achieve the inverted state.[5] In such cases, the full Boltzmann distribution applies across all levels, with higher occupancies shifting toward the upper end of the spectrum under negative T.Heat Capacity and Non-Equilibrium Aspects
In negative temperature systems, the heat capacity C is defined thermodynamically as C = T \left( \frac{\partial S}{\partial T} \right)_V, where T is the temperature, S is the entropy, and the subscript V denotes constant volume. For such systems, where the entropy S reaches a maximum at infinite temperature and decreases with increasing energy in the negative temperature regime, the derivative \left( \frac{\partial S}{\partial T} \right)_V is negative while T < 0, resulting in a positive heat capacity overall.[7] This positive C ensures consistency with the second law of thermodynamics, as the addition of heat increases both the energy and the algebraic value of the temperature (making it less negative), thereby increasing the system's "hotness" on the extended temperature scale.[8] A representative example is the two-level system, such as non-interacting spins in a magnetic field, where the heat capacity takes the form C = N k \left( \frac{\epsilon}{k T} \right)^2 \sech^2 \left( \frac{\epsilon}{2 k T} \right), with N the number of particles, k Boltzmann's constant, and \epsilon the energy splitting between levels. This expression, derived from the canonical ensemble, yields positive values for both positive and negative T due to its dependence on $1/T, and peaks near |kT| \approx 0.417 \epsilon, reflecting the Schottky anomaly extended to the negative regime.[7] In this context, population inversion in the negative temperature state contributes to the entropy decrease with energy, but the macroscopic heat capacity remains positive.[8] The implications for heat flow are profound: systems at negative temperatures are hotter than any system at positive temperatures, as their inverse temperature \beta = 1/(kT) is negative and thus smaller than any positive \beta. Consequently, when a negative temperature system contacts a positive temperature one, heat spontaneously flows from the former to the latter, increasing the total entropy without external work.[8] This behavior upholds thermodynamic inequalities, such as the Clausius form, but requires careful consideration of the bounded energy spectrum inherent to negative temperature states.[7] Negative temperature states are inherently non-equilibrium with respect to typical environments, as they demand isolation from degrees of freedom with unbounded energy (e.g., phonons or photons) to prevent rapid relaxation to positive temperatures. These states are often transient, maintained transiently via sudden magnetization reversal or continuous pumping, and their longevity depends on weak coupling to the lattice or bath.[8] In extended non-equilibrium thermodynamics, such systems can form metastable steady states, for instance, in driven spin chains or discrete nonlinear Schrödinger models coupled to reservoirs, where local negative temperatures coexist with global positive ones, consistent with fluctuation-dissipation relations adapted for \beta < 0.[7]Historical Development
Early Theoretical Work
The concept of negative absolute temperature was first introduced theoretically by Lars Onsager in 1949, in the context of two-dimensional ideal fluid turbulence modeled by point vortices.[9] In his analysis of statistical hydrodynamics, Onsager applied Gibbsian equilibrium statistical mechanics to the point-vortex model, predicting that at energies exceeding a critical threshold, the system attains states of negative temperature.[9] This arises because the phase space for vortex angular momentum is effectively bounded from above, allowing the inverse temperature parameter β = 1/(k_B T)—where k_B is the Boltzmann constant—to take negative values when the energy distribution favors high-energy configurations to maximize entropy.[10] In such states, vortices of the same sign cluster into coherent, large-scale structures, explaining the observed inverse energy cascade in two-dimensional turbulence without violating thermodynamic principles.[9] Onsager's work laid the groundwork for understanding negative temperatures in systems with conserved angular momentum, highlighting how bounded phase spaces enable entropy to decrease with increasing energy beyond a maximum, thus permitting β < 0.[10] This theoretical insight resolved long-standing puzzles in hydrodynamic stability, such as the spontaneous formation of isolated vortices at high Reynolds numbers, by framing them as equilibrium phenomena rather than instabilities.[9] In 1956, Norman Ramsey extended and formalized the thermodynamics and statistical mechanics of negative temperatures, addressing key conceptual challenges and paradoxes.[11] Ramsey established that negative temperatures occur precisely in systems where the accessible phase space is bounded such that entropy S(E) reaches a maximum at some energy E_m, leading to negative β = 1/(k_B T) for E > E_m, as dS/dE = β becomes negative.[11] He clarified that such states are hotter than any positive temperature, with heat flowing from negative-T to positive-T systems, thereby preserving the second law.[11] Ramsey resolved apparent paradoxes, such as violations of the third law of thermodynamics, by showing that it extends to negative temperatures only if the heat capacity C approaches zero as T → 0 from either side, ensuring absolute zero remains unattainable.[11] For the zeroth law, he demonstrated that thermal equilibrium remains transitive, but systems at negative temperatures equilibrate preferentially with other high-"hotness" states, redefining temperature ordering to include the continuum from T = 0^+ through +∞ to -∞.[11] Regarding the Carnot cycle, Ramsey proved that the efficiency η = 1 - T_c/T_h holds as the maximum possible, irrespective of whether the hot (T_h) or cold (T_c) reservoir temperatures are positive or negative, thus upholding the first law without modification.[11] These extensions in statistical mechanics emphasized the generality of negative temperatures for any bounded-energy system, bridging Onsager's hydrodynamic insights to broader thermodynamic applicability.[11]Key Experimental Milestones
The first experimental demonstration of negative temperatures occurred in 1951, when Edward M. Purcell and Robert V. Pound observed population inversion in the nuclear spins of lithium-7 nuclei in a lithium fluoride crystal using nuclear magnetic resonance (NMR) techniques.[12] By rapidly reversing the magnetic field to invert the spin populations and measuring the subsequent relaxation, they confirmed that the higher-energy spin state had a greater population than the lower-energy state, corresponding to a negative absolute temperature of approximately -100 K. This experiment provided empirical validation of the thermodynamic concept, showing that the spin system absorbed energy from the field until inversion was achieved, after which it released energy more readily, behaving as if "hotter" than infinite temperature. In the 1960s, negative temperatures were extended to optical systems through the development of lasers, where population inversion is essential for stimulated emission. A pivotal experiment was conducted by Theodore H. Maiman in 1960, who achieved the first laser action in a ruby crystal by optically pumping chromium ions to create an inverted population, effectively realizing a negative temperature state in the electronic energy levels.[13] The ruby laser emitted a coherent pulse at 694 nm when the population of the excited metastable state exceeded that of the ground state, demonstrating how negative temperatures enable amplification of light and marking a practical application in photonics. This breakthrough, building on earlier NMR insights, spurred widespread laser research and confirmed negative temperatures in non-spin systems.[13] A significant advance came in 2013, when researchers at Ludwig Maximilian University of Munich and the Max Planck Institute of Quantum Optics created negative temperatures in the motional degrees of freedom of fermionic atoms confined in an optical lattice.[14] Using a Bose-Hubbard model with attractive interactions, the team prepared a gas of ^6Li atoms in a single band of the lattice potential, then suddenly reversed the interaction sign via a magnetic field sweep to induce population inversion in momentum space. Time-of-flight imaging revealed that high-momentum states were more occupied than low-momentum ones, yielding a negative temperature of about -0.3 times the Fermi temperature, with the system exhibiting negative pressure and rapid expansion dynamics. This experiment extended negative temperatures beyond internal degrees of freedom to kinetic motion, opening avenues for studying non-equilibrium many-body physics in ultracold gases.[14] In 2019, experimental studies of quantum vortices in two-dimensional Bose-Einstein condensates provided evidence of negative temperature analogs through vortex clustering. A team at the University of Queensland observed giant clusters of same-sign vortices in a ^87Rb condensate confined to an elliptical trap, where stirring induced turbulence leading to coherent structures at high vortex densities. These clusters formed via Onsager's predicted condensation mechanism, where the vortex gas reached a negative effective temperature, maximizing entropy by concentrating energy in large-scale structures rather than dissipating it. Absorption imaging after expansion showed up to 50 vortices coalescing into stable clusters, illustrating how negative temperatures govern inverse energy cascades in quantum fluids.[15] More recently, in 2024, synthetic negative temperatures have been explored in steady-state quantum thermodynamics using coupled positive-temperature baths weakly interacting with a qutrit system, enabling studies of heat engines and refrigerators at negative temperatures.[16]Examples and Applications
Two-Level Spin Systems
In idealized non-interacting two-level spin systems, such as a collection of N non-interacting spin-1/2 particles in an external magnetic field B, each spin has two possible energy states: -\epsilon/2 (aligned with the field) and +\epsilon/2 (aligned against the field), where \epsilon = g \mu_B B with g the Landé g-factor and \mu_B the Bohr magneton. The average energy U of the system in thermal equilibrium at temperature T is given by U = -\frac{N \epsilon}{2} \tanh\left(\frac{\epsilon}{2 k T}\right), where k is Boltzmann's constant; this expression arises from the Boltzmann distribution, with the lower-energy state more populated for positive T > 0. As T increases from 0 to \infty, U rises monotonically from -N \epsilon / 2 (all spins aligned with the field) to 0 (equal populations in both states). Negative temperatures in such systems are achieved through population inversion, where more spins occupy the higher-energy state than the lower-energy one, corresponding to energies U > 0. A standard method to realize this is by suddenly reversing the direction of the magnetic field while the spins remain fixed in their orientations due to their large magnetic moment and the rapidity of the reversal; this inverts the populations relative to the new field direction, instantaneously transforming the system from a state at T = +\infty (equal populations) to one effectively at T = -\infty (all spins now aligned against the new field, maximizing U). Subsequent evolution toward equilibrium at finite negative T (with U decreasing from +N \epsilon / 2 toward 0) occurs if the system is isolated from external influences, such as lattice vibrations that could cause relaxation to positive temperatures. For a system at negative temperature, the entropy S as a function of energy reflects the bounded spectrum: S increases with U for U < 0 (positive T), reaches a maximum at U = 0 (T = \pm \infty), and decreases for U > 0 (negative T), yielding T = \left( \frac{\partial S}{\partial U} \right)^{-1} < 0. Explicitly, S = N k \left[ \ln \left( 2 \cosh \frac{\epsilon}{2 k T} \right) - \frac{\epsilon}{2 k T} \tanh \frac{\epsilon}{2 k T} \right], which confirms the non-monotonic S(U) behavior enabling negative T. The magnetization M, proportional to the difference in populations, is M = N \mu \tanh \left( \frac{\mu B}{k T} \right), where \mu is the magnetic moment per spin (with \epsilon = 2 \mu B); at negative T, M inverts sign relative to B, indicating net alignment against the field. This equilibrium at negative T is thermodynamically stable only in isolated systems, as contact with a positive-temperature reservoir would drive relaxation to the ground state.Nuclear Spin Experiments
Nuclear spin experiments provide a practical realization of negative temperatures by manipulating the populations of spin states in solid-state systems, where the spins can be treated as weakly interacting two-level systems isolated from the lattice on short timescales. The seminal demonstration occurred in 1951, when Edward M. Purcell and Robert V. Pound achieved population inversion in the nuclear spins of lithium-7 (^7Li) in lithium fluoride (LiF) crystals. They accomplished this by rapidly reversing the direction of the external magnetic field—over a duration much shorter than the Larmor precession period (approximately 0.1 milliseconds at their field strength)—which preserved the total spin angular momentum but reversed the sign of the spin energy relative to the field, resulting in a negative spin temperature on the order of -300 K.[12] Subsequent experiments have employed radio-frequency (RF) pulses to induce population inversion via adiabatic fast passage, a technique that slowly sweeps the RF frequency through resonance while maintaining spin coherence. This method, first detailed by Anatole Abragam and William G. Proctor in 1958, allows precise control over the inversion in various paramagnetic solids, such as calcium fluoride (CaF_2), and extends the applicability to systems where field reversal is impractical. In these setups, the sample is placed in a strong static magnetic field (typically 0.1–1 T), and RF pulses with amplitudes on the order of the local field are applied perpendicular to the static field to flip the spin populations adiabatically. The effective negative temperature is measured through nuclear magnetic resonance (NMR) spectroscopy, where the population inversion manifests as a reversed or enhanced absorption signal compared to equilibrium at positive temperatures. The spin temperature T_s is determined from the ratio of excited to ground state populations, n_e / n_g = exp(-ΔE / k_B T_s), extracted from the amplitude of the free induction decay or steady-state magnetization; linewidths in the NMR spectrum, influenced by spin-spin relaxation, provide additional insight into the internal equilibrium but are secondary to signal intensity for thermometry. In the Purcell-Pound experiment, negative susceptibility was directly observed via the transient magnetization following inversion, confirming T_s < 0 for durations up to several minutes.[12] A key challenge in these experiments is the transient nature of the negative temperature state, limited by spin-lattice relaxation, which transfers energy from the spin system to the surrounding lattice phonons and restores equilibrium at positive temperature. In LiF at room temperature, this spin-lattice relaxation time T_1 is typically tens of seconds, allowing observation windows of similar length, while spin-spin relaxation T_2 (on the order of 0.1 seconds) ensures rapid internal thermalization among spins but broadens spectral lines if not minimized. Shorter T_1 values, down to milliseconds in doped or low-temperature samples, further constrain the duration, necessitating fast manipulation techniques and low-temperature environments to extend viable timescales.[12] These nuclear spin experiments have applications in precision thermometry, where the high sensitivity of spin polarization to temperature enables calibration of low-temperature scales below 1 K via adiabatic demagnetization extensions, and in fundamental tests of statistical mechanics, validating the consistency of negative temperatures with thermodynamic principles like the second law in isolated spin reservoirs. For instance, Norman F. Ramsey's 1956 analysis used such systems to explore entropy maximization and heat flow directions, confirming that negative-temperature spins appear "hotter" than any positive-temperature system despite lower entropy.Laser Population Inversion
In laser systems, population inversion is achieved through optical pumping, where an external light source excites atoms or ions from lower energy levels to higher ones, preferentially populating the upper lasing level over the lower one. This inversion between the two relevant lasing levels results in a non-equilibrium state describable by an effective negative temperature, as the population distribution follows a Boltzmann-like form but with more particles in the higher-energy state. The process typically involves a multi-level scheme, such as a three-level system, where rapid non-radiative relaxation from the pumped level to the upper lasing level occurs, while the lower lasing level remains sparsely populated due to its role as the ground or terminal state.[17] A classic example is the ruby laser, utilizing chromium ions (Cr³⁺) embedded in an aluminum oxide (Al₂O₃) host crystal. Optical pumping with intense flash lamps excites the Cr³⁺ ions from the ground state (⁴A₂) to broad absorption bands in the visible spectrum (around 400–550 nm), followed by quick relaxation to the metastable upper lasing level (²E). This creates a population inversion between the ²E level and the ground state for the lasing transition at 694 nm (R1 line), enabling stimulated emission. The resulting inverted population corresponds to a negative effective temperature for these optical levels, maintained transiently during the pump pulse due to the short lifetime of the inversion.[17][18] For lasing to occur, the population inversion must satisfy the threshold condition, where the gain from stimulated emission exceeds losses due to absorption, scattering, and output coupling. This requires a minimum inversion density such that the small-signal gain coefficient compensates for the round-trip loss in the cavity. The negative temperature state enhances this by favoring emission over absorption, but its inherent instability—arising from the fact that negative temperatures represent higher effective "hotness" than any positive temperature—drives rapid relaxation toward equilibrium unless continuously pumped, limiting continuous-wave operation in early systems like ruby to pulsed modes.[17] The gain coefficient g for the lasing transition is proportional to the population difference:g \propto (n_\upper - n_\lower)
where n_\upper and n_\lower are the population densities in the upper and lower levels (adjusted for degeneracies). At negative temperatures, n_\upper > n_\lower (accounting for statistical weights), yielding positive gain and enabling amplification of light. This relation stems from the Einstein coefficients for stimulated emission and absorption, underscoring how inversion transforms the medium into an optical amplifier.[17]