Gibbs Factor and Chemical Potential

Temperature answers one question: how does a system share energy with its surroundings? But an open system — a droplet evaporating, a gas adsorbing onto a surface, electrons flowing into a semiconductor — also shares particles. To describe it we need a second bookkeeping variable, one that plays for particle number exactly the role temperature plays for energy. That variable is the chemical potential μ, and the weight it generates is the Gibbs factor $e^{-\beta(E-\mu N)}$ — the Boltzmann factor with a second exponent bolted on to price the particles a microstate contains. The Gibbs factor is not a new postulate. It falls out of the same reservoir argument that gave the Boltzmann factor: treat system-plus-reservoir as one isolated whole, count the reservoir microstates, and expand its entropy to first order — now in both the energy $E$ and the particle number $N$ the system holds. Two intensive parameters drop out of that single expansion: the familiar $1/T=\partial S_R/\partial E_R$ , and a new one, $-\mu/T=\partial S_R/\partial N_R$ . The coefficient of $N$ in the exponent is therefore forced to be $+\beta\mu$ — not chosen, derived — by exactly the mechanism that forced the coefficient of $E$ to be $-\beta$ . That is this article's owned result. Physically μ is the reservoir's "price per particle": a state rich in particles is favored when μ is high, suppressed when μ is low, and particles flow from high μ to low μ until diffusive equilibrium equalizes them, $\mu_1=\mu_2$ — the chemical analogue of $T_1=T_2$ . Below we derive the Gibbs factor, prove μ equalizes by counting, compute the surprising result that a dilute gas has μ negative (argon at STP sits at $\mu\approx-15.9\,k_BT$ ), and show that the same factor generates the Fermi–Dirac and Bose–Einstein distributions. Every number here is python-verified.

Key takeaways

The Gibbs factor $e^{-\beta(E-\mu N)}$ generalizes the Boltzmann factor to systems that exchange particles as well as energy; $\beta=1/k_BT$ , μ is the chemical potential.
It is derived, not postulated: expanding the reservoir entropy $S_R(E_0-E,\,N_0-N)$ to first order forces the coefficient of $N$ to be $+\beta\mu$ — the same reservoir mechanism that forces $-\beta$ on $E$ .
The chemical potential is the free-energy cost of one particle: $\mu=(\partial F/\partial N)_{T,V}=(\partial G/\partial N)_{T,P}=(\partial U/\partial N)_{S,V}=-T(\partial S/\partial N)_{E,V}$ — four faces of one quantity.
μ equalizes across anything that can exchange particles ( $\mu_1=\mu_2$ ), driving diffusion the way $T$ -gradients drive heat; the fugacity $z=e^{\beta\mu}$ packages it into a per-particle counting variable.
For a classical dilute gas μ is negative and large: $\mu=k_BT\ln(n\lambda_T^3)$ with $n\lambda_T^3\ll1$ ; argon at STP has $\mu\approx-15.9\,k_BT\approx-0.37$ eV.
Key idea: adding particle exchange to energy exchange promotes $e^{-\beta E}$ to $e^{-\beta(E-\mu N)}$ ; μ is the reservoir's price per particle, exactly as $T$ is its price per unit of entropy.

The core insight

Adding particle exchange to energy exchange promotes the Boltzmann factor $e^{-\beta E}$ to the Gibbs factor $e^{-\beta(E-\mu N)}$ . The chemical potential μ is the reservoir's price per particle — the first-order coefficient of $N$ in the expanded reservoir entropy, $-\mu/T=\partial S_R/\partial N_R$ — exactly analogous to how $1/T$ is the coefficient of $E$ . μ equalizes in diffusive equilibrium and is the free-energy cost of one particle.

Where this lives today

The chemical potential is the working variable of modern device physics and materials science. In a semiconductor, μ is the Fermi level in thermal equilibrium; doping a crystal shifts it toward the conduction or valence band, and the requirement of thermal equilibrium that the chemical potential be spatially uniform across a p–n junction — the direct consequence of $\mu_1=\mu_2$ derived above — is what builds the built-in potential behind every diode, transistor, and solar cell. In electrochemistry the electrochemical potential (μ plus the electrostatic energy $q\phi$ ) governs how ions and electrons flow through a battery; the cell voltage is literally a difference in electrochemical potential between the electrodes. The same variable controls the frontier of cold-atom physics. Experimentalists tune μ (through density and trap depth) to push a dilute gas across $n\lambda_T^3\approx1$ — the crossover the punchline plot marks — and into the quantum-degenerate regime, producing Bose–Einstein condensates and degenerate Fermi gases in tabletop traps. In these systems μ is measured directly and used to map out equations of state that test many-body theory. And in astrophysics the chemical potential of a relativistic electron gas, driven positive and enormous by gravitational compression, sets the degeneracy pressure that holds up white dwarfs — the Fermi-energy limit that the same $\mu=k_BT\ln(n\lambda_T^3)$ formula, extended into the degenerate regime, foreshadows.

The physical picture

Picture the reservoir as an enormous bank that lends the small system both energy and particles. The interest rate on energy is set by the temperature; the price charged per particle is set by the chemical potential μ. The Gibbs factor weights each microstate of the system by how cheaply the bank can supply what that state costs: a state with large energy is exponentially suppressed (expensive energy), and — this is the new ingredient — a state with many particles is favored when μ is high and suppressed when μ is low. When particle exchange is switched off, μ becomes irrelevant and the Gibbs factor collapses back to the Boltzmann factor.

Figure

Two microstates weighted by the Gibbs factor: energy is priced by T, particle content is priced by μ. High E is suppressed; high N is favored when μ is high.

Why is the weight an exponential of $E-\mu N$ ? Because the reservoir is macroscopic, exchanging a small amount of energy $E$ and particles $N$ with the system changes its entropy to first order: $\Delta S_R = -\left(\frac{\partial S_R}{\partial E_R}\right)_{V,N}E - \left(\frac{\partial S_R}{\partial N_R}\right)_{E,V}N = -\frac{E-\mu N}{T}$ . The probability of a system microstate is proportional to the number of reservoir microstates compatible with the remaining total energy and particle number, and that number is $e^{S_R/k_B}$ . Exponentiating a linear change in $S_R$ is precisely what produces the Gibbs factor — the deep reason both $-\beta E$ and $+\beta\mu N$ sit in the same exponent is that they are the two first-order Taylor coefficients of one reservoir entropy. An important consequence is the sign of the chemical potential. Because the fugacity $z=e^{\beta\mu}$ multiplies the weight once per particle, a system wants more particles when μ is large. Yet for an ordinary classical gas μ is negative — often many $k_BT$ below zero. The reason is entropic, not energetic: adding a particle to a dilute gas creates a huge amount of entropy (many empty cells to occupy), so it lowers the free energy $F=U-TS$ , and μ, the free-energy cost of a particle, comes out below zero. The punchline plot below shows μ climbing with density and crossing zero exactly where the gas stops being classical — where $n\lambda_T^3\approx1$ and quantum statistics take over.

Figure

Chemical potential μ(n) of a gas rises with density, negative in the dilute classical regime, crossing zero near nλ³ ≈ 1 where quantum degeneracy begins.

The core equations

An open system exchanges energy and particles with a reservoir at temperature $T$ and chemical potential μ. The probability of a definite system microstate is proportional to the Gibbs factor, which weights the state by both its energy and its particle number:

P(\text{microstate})\;\propto\;e^{-\beta(E-\mu N)},\qquad \beta=\frac{1}{k_BT} \qquad (1)

where: $E$ — energy of the microstate; $N$ — its particle number; $\beta$ — inverse temperature of the reservoir; μ — chemical potential set by the reservoir. Reading the Gibbs factor, states with larger energy are exponentially suppressed, whereas increasing the chemical potential enhances the statistical weight of states with larger particle number. With $N$ frozen the factor $e^{\beta\mu N}$ is a constant and (1) reduces to the Boltzmann factor. Summing the Gibbs factor over every microstate — all energies and all particle numbers — normalizes it and defines the grand partition function, which groups naturally by particle number through the fugacity:

\Xi(T,V,\mu)=\sum_{N=0}^{\infty}\sum_{i}e^{-\beta(E_i^{(N)}-\mu N)}=\sum_{N=0}^{\infty}z^{N}Z_N(T,V),\qquad z=e^{\beta\mu} \qquad (2)

where: Ξ — grand partition function; $Z_N$ — the canonical partition function at fixed $N$ ; $z$ — fugacity; $i$ — microstate index at fixed $N$ . Reading it: $z^N$ counts particles the way $x^n$ counts terms in a power series, so Ξ is a generating function in $z$ . This article owns the derivation of the Gibbs factor and μ; the fuller treatment of Ξ belongs to the grand-partition-function article. The chemical potential is the free-energy cost of adding one particle, and it wears four equivalent faces depending on what is held fixed:

\mu=\left(\frac{\partial F}{\partial N}\right)_{T,V}=\left(\frac{\partial G}{\partial N}\right)_{T,P}=\left(\frac{\partial U}{\partial N}\right)_{S,V}=-T\left(\frac{\partial S}{\partial N}\right)_{E,V} \qquad (3)

where: $F=U-TS$ (Helmholtz), $G=U-TS+PV$ (Gibbs), $U$ — internal energy, $S$ — entropy. Reading it: at fixed $T$ the cost includes the entropy change (Helmholtz form); at fixed $S$ it is the pure energy cost. The entropy form $-T(\partial S/\partial N)_{E,V}$ is the one that appears directly in the reservoir expansion and is the origin of the $+\beta\mu$ coefficient. The fundamental thermodynamic relation acquires a chemical-work term, and the three intensive variables are constrained by Gibbs–Duhem:

\mathrm dU=T\,\mathrm dS-P\,\mathrm dV+\mu\,\mathrm dN,\qquad S\,\mathrm dT-V\,\mathrm dP+N\,\mathrm d\mu=0 \qquad (4)

where: $\mu\,\mathrm dN$ — chemical work of adding particles; the Gibbs–Duhem relation says $T$ , $P$ , μ of a one-component system cannot vary independently — fixing two fixes the third along the coexistence surface. For a classical ideal gas the chemical potential follows from differentiating the free energy, and it is negative in the dilute regime:

\mu=k_BT\ln\!\big(n\lambda_T^3\big),\qquad \lambda_T=\frac{h}{\sqrt{2\pi m k_BT}},\qquad n=\frac{N}{V} \qquad (5)

where: $n$ — number density; $\lambda_T$ — thermal de Broglie wavelength; $m$ — particle mass. Reading it: $n\lambda_T^3$ is the number of particles per quantum volume. When it is $\ll1$ (dilute, hot) the log is negative and μ sits far below zero; when it approaches 1 the gas becomes degenerate and μ climbs toward and past zero.

Derivation AThe Gibbs factor from the reservoir entropy — why the coefficient of

N

is fixed to be

\beta\mu

This is the article's owned result: the $+\beta\mu N$ in the Gibbs exponent is not an independent postulate but the second first-order Taylor coefficient of the reservoir entropy, produced by exactly the mechanism that produced $-\beta E$ . The goal is to derive $P(E,N)\propto e^{-\beta(E-\mu N)}$ by counting reservoir microstates and to show that the coefficient of $N$ must be $+\beta\mu$ , with $-\mu/T=\partial S_R/\partial N_R$ . The strategy is equal a priori probabilities for the isolated whole, the Boltzmann relation $S_R=k_B\ln\Omega_R$ , and a first-order expansion in both $E$ and $N$ .

Step 1. Treat system + reservoir as one isolated microcanonical whole with total energy $E_0$ and total number $N_0$ . Fix the system in one definite microstate of energy $E$ and number $N$ ; then its probability is proportional to the reservoir's multiplicity for the remainder, $P(E,N)\propto\Omega_R(E_0-E,\,N_0-N)$ . Why: the system microstate is a single state (multiplicity 1), so all the counting is in the reservoir — equal a priori probabilities applied to the whole.

Step 2. Write the multiplicity as an exponential of entropy, $\Omega_R=e^{S_R(E_0-E,\,N_0-N)/k_B}$ , using $S_R=k_B\ln\Omega_R$ . Why: converting a product-like count into $e^{S_R/k_B}$ turns the counting problem into a smooth expansion problem — the reason the weight will be exponential.

Step 3. Taylor-expand $S_R$ to first order in both arguments, since $E\ll E_0$ and $N\ll N_0$ : $S_R(E_0-E,N_0-N)\approx S_R(E_0,N_0)-E\left(\dfrac{\partial S_R}{\partial E_R}\right)-N\left(\dfrac{\partial S_R}{\partial N_R}\right)$ . Why: the reservoir is enormous, so the second-order terms scale as $1/N_R\to0$ ; the linear term is exact in the thermodynamic limit — and crucially there are now TWO first-order coefficients, one per conserved quantity.

Step 4. Identify the two coefficients with their thermodynamic definitions: $\dfrac{\partial S_R}{\partial E_R}=\dfrac1T$ and $\dfrac{\partial S_R}{\partial N_R}=-\dfrac{\mu}{T}$ . Substituting, $S_R\approx S_R(E_0,N_0)-\dfrac{E}{T}+\dfrac{\mu N}{T}$ . Why: these are the standard entropy derivatives — $1/T$ for energy exchange and $-\mu/T$ for particle exchange — so the coefficient of $N$ is fixed to $+\mu/T$ , not free to choose.

Step 5. Exponentiate and absorb the constant: $P(E,N)\propto e^{S_R/k_B}\propto e^{-\beta(E-\mu N)}$ , since $e^{S_R(E_0,N_0)/k_B}$ is state-independent and cancels in normalization. The coefficient of $N$ in the exponent is $+\mu/(k_BT)=+\beta\mu$ , mirroring the coefficient $-1/(k_BT)=-\beta$ of $E$ . Why: both exponents come from the same expansion, so their forms are not independent — the same reservoir mechanism that priced energy at $-\beta$ prices particles at $+\beta\mu$ . *The symbolic reservoir expansion reproduces the exponent $-(E-\mu N)/k_BT$ , with $\partial(\text{exponent})/\partial N=+\beta\mu$ and $\partial(\text{exponent})/\partial E=-\beta$ .*

Go deeper · Derivation B · 1 minChemical potential of the classical ideal gas — and why it is negative

Derivation BChemical potential of the classical ideal gas — and why it is negative

The goal is to compute μ for a monatomic ideal gas of $N$ indistinguishable particles and read off its sign, obtaining $\mu=k_BT\ln(n\lambda_T^3)$ , which is negative whenever the gas is dilute. The strategy is to write the canonical free energy $F=-k_BT\ln Z_N$ with the Gibbs-corrected $Z_N$ , apply Stirling, and differentiate with respect to $N$ .

Step 1. Single-particle and $N$ -particle partition functions: $Z_1=V/\lambda_T^3$ with $\lambda_T=h/\sqrt{2\pi m k_BT}$ , so $Z_N=Z_1^N/N!=(V/\lambda_T^3)^N/N!$ . Why: the $1/N!$ is the correct Boltzmann counting for identical particles — without it $F$ is not extensive and μ comes out wrong.

Step 2. Free energy: $F=-k_BT\ln Z_N=-k_BT\left[N\ln(V/\lambda_T^3)-\ln N!\right]$ , and with Stirling $\ln N!\approx N\ln N-N$ this is $F=-k_BT\left[N\ln(V/\lambda_T^3)-N\ln N+N\right]$ . Why: Stirling turns the discrete $\ln N!$ into a differentiable function of $N$ , which we need to take $\partial/\partial N$ .

Step 3. Differentiate: $\mu=\left(\dfrac{\partial F}{\partial N}\right)_{T,V}=-k_BT\left[\ln(V/\lambda_T^3)-\ln N-1+1\right]=-k_BT\ln\!\dfrac{V}{N\lambda_T^3}$ . The $-\ln N$ from $\partial_N(N\ln N)=\ln N+1$ and the $+1$ from $\partial_N N$ combine so the two stray constants cancel. Why: it is exactly this cancellation of the $\pm1$ that leaves a clean logarithm — a common place to slip a sign.

Step 4. Rewrite with $n=N/V$ : $\mu=k_BT\ln(n\lambda_T^3)$ . Since a classical gas has $n\lambda_T^3\ll1$ (particles far apart compared to their thermal wavelength), the logarithm is negative, so $\mu<0$ . Why: the negativity is entropic — a dilute gas has enormous configurational entropy, so adding a particle lowers $F=U-TS$ ; μ, the cost of a particle, is therefore below zero. *The result $\mu=\partial F/\partial N=k_BT\ln(n\lambda_T^3)$ holds symbolically; argon at STP gives $n\lambda_T^3=1.25\times10^{-7}$ , $\mu=-15.9\,k_BT=-0.374$ eV, and fugacity $z=e^{\beta\mu}=n\lambda_T^3$ .*

Limiting cases

How the Gibbs factor and μ behave in the regimes that matter:

Case	Result	Meaning
Closed system ( $N$ fixed)	$e^{-\beta(E-\mu N)}\to e^{-\beta E}$	The $e^{\beta\mu N}$ factor is a constant, cancels in normalization; the Gibbs factor collapses to the Boltzmann factor.
Classical dilute gas	$\mu=k_BT\ln(n\lambda_T^3)<0$	$n\lambda_T^3\ll1$ , so μ is negative and large; fugacity $z\ll1$ , quantum statistics reduce to Maxwell–Boltzmann.
Onset of degeneracy	$n\lambda_T^3\to1$ , $\mu\to0$	Particles begin to overlap; μ climbs toward zero and quantum statistics become essential.
Degenerate Fermi gas ( $T\to0$ )	$\mu\to\varepsilon_F>0$	μ becomes the Fermi energy — positive; the Fermi level of a metal.
Ideal Bose gas at condensation	$\mu\to0^-$	$\mu$ cannot exceed the ground-state energy; at $\mu\to\varepsilon_0$ (here $\varepsilon_0=0$ ), the ground-state occupation becomes macroscopic (BEC).

Common trap“The chemical potential is the energy of one particle, and it must be positive”

Both halves are wrong. First, μ is not generally an energy per particle: only $(\partial U/\partial N)_{S,V}$ is a pure energy, and even that is not "the particle's energy" but the change in total energy at fixed entropy. At fixed temperature μ is the free-energy cost $(\partial F/\partial N)_{T,V}$ , which folds in the entropy created by the new particle. Second, μ is very often negative: for a classical ideal gas at ordinary densities $n\lambda_T^3\ll1$ makes $\mu=k_BT\ln(n\lambda_T^3)$ deeply negative (argon at STP: $-15.9\,k_BT$ ), because a dilute gas gains so much entropy per added particle that its free energy drops. μ only becomes positive in the quantum-degenerate regime, where it turns into the Fermi energy. Watch the sign convention too: the Gibbs factor carries $+\beta\mu N$ , so a larger μ favors more particles.

Worked examples

1 · The chemical potential of argon gas at STP (a real, negative number)

Compute μ for argon (atomic mass 39.95 u) at standard temperature and pressure, $T=273.15$ K, $P=1$ atm. Enumerate the ingredients, apply $\mu=k_BT\ln(n\lambda_T^3)$ , and interpret the sign.

Step 1. Number density from the ideal-gas law: $n=P/k_BT=101325/(1.381\times10^{-23}\times273.15)=2.69\times10^{25}\ \mathrm{m^{-3}}$ .

Step 2. Thermal wavelength: $\lambda_T=h/\sqrt{2\pi m k_BT}$ with $m=39.95\times1.6605\times10^{-27}=6.63\times10^{-26}$ kg gives $\lambda_T=6.626\times10^{-34}/\sqrt{2\pi\times6.63\times10^{-26}\times1.381\times10^{-23}\times273.15}=1.67\times10^{-11}$ m — about 17 pm, far smaller than the $\sim3$ nm spacing between molecules.

Step 3. Quantum-occupancy parameter: $n\lambda_T^3=2.69\times10^{25}\times(1.67\times10^{-11})^3=1.25\times10^{-7}$ . This is $\ll1$ , confirming argon at STP is a thoroughly classical gas.

Step 4. Chemical potential: $\mu=k_BT\ln(1.25\times10^{-7})=k_BT\times(-15.9)$ . In energy units, $\mu=-15.9\times k_BT=-15.9\times0.0236\ \mathrm{eV}=\mathbf{-0.374\ eV}$ , i.e. $\mathbf{\mu\approx-15.9\,k_BT}$ . The chemical potential is negative and nearly sixteen $k_BT$ below zero — the everyday signature of a dilute classical gas, where adding a particle creates so much entropy that it lowers the free energy. The fugacity is $z=e^{\beta\mu}=n\lambda_T^3=1.25\times10^{-7}\ll1$ , the small parameter that makes classical statistics valid.

2 · Two boxes exchange particles: equilibrium is μ₁ = μ₂

Two gas boxes at the same temperature are connected so particles can pass between them but energy is held fixed in each. Show that total entropy is maximized when their chemical potentials are equal, and find which way particles flow if they start unequal.

Step 1. With $N_1+N_2=N_{\text{tot}}$ fixed, total entropy is $S_{\text{tot}}(N_1)=S_1(N_1)+S_2(N_{\text{tot}}-N_1)$ . Equilibrium maximizes $S_{\text{tot}}$ , so set $\mathrm dS_{\text{tot}}/\mathrm dN_1=0$ .

Step 2. Differentiate: $\dfrac{\mathrm dS_{\text{tot}}}{\mathrm dN_1}=\left(\dfrac{\partial S_1}{\partial N_1}\right)-\left(\dfrac{\partial S_2}{\partial N_2}\right)=0$ , so $\left(\dfrac{\partial S_1}{\partial N_1}\right)_{E,V}=\left(\dfrac{\partial S_2}{\partial N_2}\right)_{E,V}$ .

Step 3. Convert to chemical potentials with $\mu=-T(\partial S/\partial N)_{E,V}$ . At the common temperature, equality of the entropy slopes means $\mu_1=\mu_2$ . Diffusive equilibrium is the equality of chemical potentials — the exact analogue of thermal equilibrium being the equality of temperatures.

Step 4. If If instead $\mu_1>\mu_2$ , transferring one particle from box 1 to box 2 changes the total Helmholtz free energy by $\Delta F=\mu_2-\mu_1<0$ , so the transfer is spontaneous. Particles therefore diffuse from high $\mu$ to low $\mu$ until the chemical potentials become equal, just as heat flows from high temperature to low temperature until the temperatures equalize.

Origins · 1 minchemical potential and how this was discovered

chemical potential

introduced by J. W. Gibbs, 1875–1878 · thermodynamics

The chemical potential was introduced by Josiah Willard Gibbs in his monumental memoir On the Equilibrium of Heterogeneous Substances (published in the Transactions of the Connecticut Academy in two parts, 1875 and 1878). Gibbs recognized that to describe systems where matter is redistributed — between phases, across membranes, in chemical reactions — the fundamental relation for energy needs a term $\mu\,\mathrm dN$ for each species, with μ the intensive variable conjugate to particle number. His central result is that equilibrium of matter exchange is the equality of chemical potentials: two phases coexist, or a reaction reaches balance, exactly when the relevant μ's are equal. This single principle organizes the entire theory of phase diagrams, the law of mass action, osmosis, and the Clausius–Clapeyron equation. In statistical mechanics Gibbs later (in his 1902 Elementary Principles in Statistical Mechanics) built μ into the grand ensemble through the factor $e^{-\beta(E-\mu N)}$ . The concept is so central that it reappears throughout physics under different names: it is the Fermi level in semiconductors and metals, the quantity that must be flat across a junction in equilibrium; it is the electrochemical potential that batteries exploit; and it is the parameter that pins itself just below the ground-state energy to trigger Bose–Einstein condensation.

History

1875

— Gibbs introduces the chemical potential in On the Equilibrium of Heterogeneous Substances, Part I.

1878

— Gibbs completes the memoir (Part II), establishing μ-equality as the condition for phase and chemical equilibrium.

1902

— Gibbs formulates the grand canonical ensemble and the factor e^(−β(E−μN)) in Elementary Principles in Statistical Mechanics.

1926

— Fermi and, independently, Dirac give the fermion occupation whose μ becomes the Fermi energy.

1938

— London identifies the pinning of μ at the ground state with Bose–Einstein condensation in liquid helium.

Going further

References. Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 4 · Kardar, Statistical Physics of Particles, Ch. 4 · Reif, Fundamentals of Statistical and Thermal Physics, §6.9 · Sethna, Statistical Mechanics, Ch. 6 · Landau & Lifshitz, Statistical Physics Part 1, §24, §35 · Gibbs, On the Equilibrium of Heterogeneous Substances (1875–78) and Elementary Principles in Statistical Mechanics (1902). The coefficient of $N$ is forced to $+\beta\mu$ by the same reservoir entropy expansion that forces $-\beta$ on $E$ , proved symbolically above. The ideal-gas result $\mu=k_BT\ln(n\lambda_T^3)$ follows from differentiating $F=-k_BT\ln Z_N$ with $Z_N=(V/\lambda_T^3)^N/N!$ and Stirling, and the four definitions of μ are equivalent on the ideal gas. Numerical values: argon at STP gives $n\lambda_T^3=1.25\times10^{-7}$ , $\mu=-15.9\,k_BT=-0.374$ eV, and fugacity $z=n\lambda_T^3$ . Constants $h$ , $k_B$ (CODATA).

Builds on: Boltzmann Factor · Canonical Ensemble · Derivation of the Canonical Distribution from the Microcanonical Ensemble · Helmholtz Free Energy from the Partition Function. Leads to: Grand Canonical Ensemble · Grand Partition Function · Grand Potential.

Why is the chemical potential of a normal gas negative? Isn't it "the energy of a particle"?

It is negative because μ is a free-energy cost, not an energy cost, and a dilute gas gains a lot of entropy per added particle. Write $\mu=(\partial F/\partial N)_{T,V}$ with $F=U-TS$ . Adding a particle to a dilute gas increases $U$ a little (its kinetic energy) but increases $S$ a lot, because there are so many empty single-particle cells for it to occupy. The $-TS$ term wins, so $F$ actually decreases when you add a particle, and μ — the slope of $F$ versus $N$ — comes out below zero. Quantitatively $\mu=k_BT\ln(n\lambda_T^3)$ , and since a classical gas has $n\lambda_T^3\ll1$ (argon at STP: $1.25\times10^{-7}$ ), the log is a large negative number, giving $\mu\approx-15.9\,k_BT$ . μ only becomes positive when the gas is compressed or cooled into the quantum-degenerate regime, where it turns into the Fermi energy.

What is the difference between the Gibbs factor and the Boltzmann factor?

The Boltzmann factor $e^{-\beta E}$ weights a microstate by its energy alone; it is the right weight for a closed system that exchanges only energy with a heat bath (the canonical ensemble). The Gibbs factor $e^{-\beta(E-\mu N)}$ adds a second exponent that weights the microstate by its particle content as well; it is the right weight for an open system that exchanges both energy and particles with a reservoir (the grand canonical ensemble). The extra term $+\beta\mu N$ comes from expanding the reservoir entropy in particle number, exactly as $-\beta E$ comes from expanding it in energy. When particle exchange is switched off — $N$ is the same for every accessible microstate — the factor $e^{\beta\mu N}$ becomes a common constant that cancels in the normalized probability, and the Gibbs factor reduces cleanly to the Boltzmann factor. So the Gibbs factor is strictly more general: it contains the Boltzmann factor as its fixed- $N$ special case.

Where exactly does the "+μN" in the exponent come from — is it just chosen to look symmetric?

It is forced, not chosen. Start from the isolated system-plus-reservoir and count reservoir microstates: $P(E,N)\propto\Omega_R(E_0-E,N_0-N)=e^{S_R(E_0-E,N_0-N)/k_B}$ . Now Taylor-expand the reservoir entropy to first order in both the energy and the particle number the system has taken: $S_R\approx S_R(E_0,N_0)-E\,\partial_E S_R-N\,\partial_N S_R$ . The two first-order coefficients are thermodynamic derivatives with fixed meanings: $\partial_E S_R=1/T$ and $\partial_N S_R=-\mu/T$ . Substituting gives $S_R/k_B\approx\text{const}-\beta E+\beta\mu N$ , so the exponent is $-\beta(E-\mu N)$ . The $+\beta\mu N$ term is simply the second Taylor coefficient — its form is dictated by the reservoir's entropy, the same way $-\beta E$ is dictated by the first coefficient. The apparent symmetry between the energy and particle terms is real and structural: both are first-order responses of one reservoir entropy.

What is the fugacity z, and why is it more convenient than μ?

The fugacity is $z=e^{\beta\mu}$ , and it is convenient because it is the natural counting variable for particles. In the grand partition function $\Xi=\sum_N z^N Z_N$ , each factor of $z$ corresponds to one more particle, so $z$ plays the role that a formal variable $x$ plays in a generating function $\sum_N a_N x^N$ : derivatives with respect to $z$ pull down powers of $N$ , giving the mean particle number as $\langle N\rangle=z\,\partial_z\ln\Xi$ . It also has a clean physical range: for a classical gas $z=n\lambda_T^3\ll1$ is a small parameter you can expand in (the virial and cluster expansions are power series in $z$ ), while $z\to1$ signals the onset of quantum degeneracy. For an ideal Bose gas $z$ is bounded above by 1 (else the geometric series for the ground mode diverges), and it is this ceiling that forces condensation. So μ and $z$ carry the same information, but $z$ turns the exponential dependence on μ into a polynomial counting variable that makes the algebra of open systems tractable.

Is the chemical potential the same thing as the Fermi level?

Yes — the Fermi level of a metal or semiconductor is the chemical potential of its electrons, and the two terms are used interchangeably in condensed-matter physics. The connection comes from the Fermi–Dirac occupation $\langle n\rangle=1/(e^{\beta(\varepsilon-\mu)}+1)$ : at zero temperature this is a step function that fills every single-particle state up to $\varepsilon=\mu$ and leaves everything above empty, so μ marks the top of the filled sea — the Fermi energy $\varepsilon_F$ . At finite temperature μ stays close to $\varepsilon_F$ and the step softens over a width $\sim k_BT$ . What makes the Fermi level so useful is exactly the equilibrium condition derived here: in a device made of different materials in contact, electrons flow until the chemical potential is flat across the whole structure ( $\mu_1=\mu_2$ ). That flat Fermi level, and the way band edges bend around it, is the organizing principle of all semiconductor device physics — p–n junctions, transistors, and photovoltaics.

Why does the Gibbs factor generate both the Bose–Einstein and Fermi–Dirac distributions?

Because you can treat a single quantum energy level as a tiny open system whose "particles" are the quanta occupying that level, and apply the Gibbs factor to it. The grand sum for one mode of energy ε is $\Xi_\varepsilon=\sum_n z^n e^{-\beta\varepsilon n}=\sum_n x^n$ with $x=e^{-\beta(\varepsilon-\mu)}$ . For bosons any occupation $n=0,1,2,\dots$ is allowed, so the sum is a geometric series $1/(1-x)$ ; for fermions Pauli exclusion caps it at $n=0,1$ , so the sum is $1+x$ . The mean occupation is $\langle n\rangle=x\,\partial_x\ln\Xi_\varepsilon=x/(1\mp x)$ , which becomes $\langle n\rangle=1/(e^{\beta(\varepsilon-\mu)}\mp1)$ — the minus sign giving Bose–Einstein, the plus sign Fermi–Dirac. Both distributions thus fall out of the same one-line computation, differing only in whether the occupation series runs to infinity or stops at one. In the dilute limit $z\ll1$ both reduce to the classical $\langle n\rangle\approx z\,e^{-\beta\varepsilon}$ . (The microscopic origin of the exclusion and the symmetry of the wavefunction is the quantum-statistics story; here the Gibbs factor supplies the grand-canonical bookkeeping.)