Skip to content
LumenaryAcademy
Physical 70 min

Chemical kinetics

How fast reactions go and why — rate laws and half-lives, the Arrhenius equation fitted to data, mechanisms and the steady-state approximation checked against a numerical solution, and a first look at catalysts and enzymes.

Code tested · 6AI-draftedAwaiting review by a chemist

Your guides: Chloe · The Chemist, Margaret · The Analyst, Gideon · The Sceptic

MotivationH

Allowed is not the same as fast

Sugar and oxygen together are hugely unstable: the Gibbs energy of the reaction that turns them into carbon dioxide and water is about −2900 kJ per mole of glucose. Yet a bowl of sugar sits on the table for years. Diamonds are thermodynamically doomed to become graphite and never do. Petrol and air, paper and air, your own body and air: all are far from equilibrium and all are, on a human timescale, stable. The previous lessons said where a reaction can go. Kinetics asks how fast it gets there, and the answers are just as practical.

A refrigerator works because cooling food from 20 °C to 4 °C slows the reactions of spoilage several-fold. A medicine's shelf life is the time for a small percentage of it to decompose, and it is predicted by measuring the decomposition faster at higher temperatures. Ozone in the stratosphere is destroyed by a chain of fast reactions that a few chlorine atoms keep running. Industrial chemistry, from Haber–Bosch to plastics, depends on catalysts that make the wanted reaction fast and leave the unwanted ones slow, and life depends on enzymes, catalysts so good that they speed some reactions by factors above 101510^{15}.

This lesson builds the quantitative tools: rate laws and their integrated forms, the Arrhenius equation for temperature, and mechanisms. Throughout, it uses the mathematics of ordinary differential equations, solved both by hand and numerically.

IntuitionH

Collisions and barriers

For two molecules to react, they must meet. Double the concentration of one of them and, other things equal, they meet twice as often, so the reaction goes twice as fast. That is why rates depend on concentrations.

But most meetings achieve nothing. To react, old bonds must start to break before new ones form, and the half-broken arrangement in between, the transition state, is higher in energy than both reactants and products. Only collisions with enough energy, and the right orientation, get over this activation energy barrier. At room temperature only a tiny fraction of molecules have enough energy at any moment, and that fraction grows steeply as the temperature rises. That is why heat speeds reactions up so dramatically.

Analogy, not a formal construction

A ball in a valley, with a deeper valley beyond a ridge, will roll into the deeper valley eventually, but only if something kicks it over the ridge. Thermodynamics compares the depths of the two valleys; kinetics is about the height of the ridge and how hard the ball is being kicked. A catalyst digs a lower pass through the ridge.

A safe way to see kinetics in the kitchen, with an adult doing the cutting: cut an apple into slices, leave one on a plate, put one in the fridge, and brush one with lemon juice. The browning is catalysed by an enzyme in the apple (polyphenol oxidase) that uses oxygen from the air; the cold slows it down, and the acid and vitamin C in lemon juice hold it back. Compare the slices after an hour. Three slices, one reaction, three rates: temperature and catalysis are the two big levers of kinetics, and this lesson puts numbers on both.

DefinitionsH

Rates and rate laws

Definition— rate of reaction

For a reaction 0=∑iνiXi0 = \sum_i \nu_i X_i at constant volume, the rate is r=1νid[Xi]dtr = \dfrac{1}{\nu_i}\dfrac{d[X_i]}{dt}, the same whichever species is used. For N₂ + 3 H₂ → 2 NH₃, r=−d[N2]dt=−13d[H2]dt=12d[NH3]dtr = -\dfrac{d[\mathrm{N_2}]}{dt} = -\dfrac13\dfrac{d[\mathrm{H_2}]}{dt} = \dfrac12\dfrac{d[\mathrm{NH_3}]}{dt}. Its units are mol L⁻¹ s⁻¹.

Definition— rate law, order and rate constant

A rate law expresses the rate in terms of concentrations, often in the form r=k[A]m[B]nr = k[\mathrm{A}]^m[\mathrm{B}]^n. The exponents are the orders with respect to A and B, and m+nm + n is the overall order. The rate constant kk depends on temperature but not on concentration. Orders are found by experiment; they need not match the coefficients in the balanced equation, and they may be fractions or zero.

Definition— elementary reaction and mechanism

An elementary reaction happens in a single molecular event, such as one collision. Its molecularity is the number of molecules involved (unimolecular, bimolecular, rarely termolecular), and for an elementary step, and only then, the rate law follows from the equation: A + B → products has rate k[A][B]k[\mathrm{A}][\mathrm{B}]. A mechanism is a sequence of elementary steps that adds up to the overall reaction.

Definition— half-life

The half-life t1/2t_{1/2} is the time for the concentration of a reactant to fall to half its starting value.

The orders of a reaction are usually measured by the method of initial rates (change one starting concentration at a time and watch how the initial rate responds) or by following one experiment through time and testing which integrated rate law fits.

Example— Orders from initial rates

Nitrogen monoxide and hydrogen react at around 1100 K: 2 NO(g) + 2 H₂(g) → N₂(g) + 2 H₂O(g). Here are three illustrative runs, each measuring the rate in the first moments, before the concentrations have changed much:

Run[NO]0[\mathrm{NO}]_0 / mol L⁻¹[H2]0[\mathrm{H_2}]_0 / mol L⁻¹initial rate / mol L⁻¹ s⁻¹
10.00500.00202.0×10−52.0 \times 10^{-5}
20.01000.00208.0×10−58.0 \times 10^{-5}
30.00500.00404.0×10−54.0 \times 10^{-5}

From run 1 to run 2 only NO changes: doubling it multiplies the rate by 4 =22= 2^2, so the reaction is second order in NO. From run 1 to run 3 only H₂ changes: doubling it doubles the rate, so it is first order in H₂. (When the ratios are not neat, the order is log⁡(r2/r1)/log⁡(c2/c1)\log(r_2/r_1)/\log(c_2/c_1).) The rate law is r=k[NO]2[H2]r = k[\mathrm{NO}]^2[\mathrm{H_2}], third order overall, and any run gives k=r/([NO]2[H2])=2.0×10−5/(0.00502×0.0020)=400k = r/([\mathrm{NO}]^2[\mathrm{H_2}]) = 2.0 \times 10^{-5}/(0.0050^2 \times 0.0020) = 400 L² mol⁻² s⁻¹. Notice the units of kk, which change with the overall order, and notice that H₂ has coefficient 2 in the equation but order 1 in the rate law. The mechanism section below explains why.

Theorem— Integrated rate laws

For a reaction A → products with rate −d[A]/dt=k[A]n-d[\mathrm{A}]/dt = k[\mathrm{A}]^n and starting concentration [A]0[\mathrm{A}]_0:

OrderIntegrated lawStraight lineHalf-life
0[A]=[A]0−kt[\mathrm{A}] = [\mathrm{A}]_0 - kt[A][\mathrm{A}] against tt[A]0/2k[\mathrm{A}]_0/2k
1[A]=[A]0 e−kt[\mathrm{A}] = [\mathrm{A}]_0\,e^{-kt}ln⁡[A]\ln[\mathrm{A}] against ttln⁡2/k\ln 2/k
21/[A]=1/[A]0+kt1/[\mathrm{A}] = 1/[\mathrm{A}]_0 + kt1/[A]1/[\mathrm{A}] against tt1/(k[A]0)1/(k[\mathrm{A}]_0)

Proof

Each rate law is a separable differential equation. For n=1n = 1: d[A]/[A]=−k dtd[\mathrm{A}]/[\mathrm{A}] = -k\,dt; integrating from 0 to tt gives ln⁡([A]/[A]0)=−kt\ln([\mathrm{A}]/[\mathrm{A}]_0) = -kt. For n=2n = 2: d[A]/[A]2=−k dtd[\mathrm{A}]/[\mathrm{A}]^2 = -k\,dt integrates to 1/[A]−1/[A]0=kt1/[\mathrm{A}] - 1/[\mathrm{A}]_0 = kt. For n=0n = 0 the rate is constant. Each half-life follows by setting [A]=[A]0/2[\mathrm{A}] = [\mathrm{A}]_0/2.

∎

A second-order reaction is different: its half-life 1/(k[A]0)1/(k[\mathrm{A}]_0) depends on the starting concentration, so each successive half-life is twice as long as the one before, because the concentration it starts from has halved. That gives a quick test by eye, shown in the figure.

11/21/41/801234567time, in units of the first half-life[A] / [A]₀first order: equal half-livessecond order: 1, 2, 4, …
First-order (solid) and second-order (dashed) decay with the same first half-life. The first-order curve halves in every equal interval; the second-order curve takes one interval to fall to 1/2, two more to fall to 1/4 and four more to fall to 1/8, because its half-life is inversely proportional to the concentration it starts from.

Example— Coffee at bedtime

The body clears caffeine by roughly first-order kinetics, with a half-life that varies a lot between people but is about 5 hours in a typical healthy adult. Someone drinks a coffee with 80 mg of caffeine at 08:00. By 22:00, 14 hours later, the amount left is 80×2−14/5=11.580 \times 2^{-14/5} = 11.5 mg, about a seventh of the dose. To fall to 10 mg, one-eighth, takes exactly three half-lives, 15 hours: 23:00. In rate-constant form, k=ln⁡2/(5 h)=0.139k = \ln 2/(5\ \text{h}) = 0.139 h⁻¹, and the same answer comes from t=ln⁡(80/10)/kt = \ln(80/10)/k. The half-life is what pharmacologists quote, because it answers "how long until most of it is gone?" without knowing the dose.

ComputationC

Finding the order from data

Ludwig Wilhelmy made the first quantitative rate measurement in 1850, following the acid-catalysed breakdown of sucrose into glucose and fructose with a polarimeter, and found that the rate was proportional to the amount of sugar left. Here is an illustrative data set of the same kind: the concentration of a reactant, measured every so often. To find the order, test each integrated law for a straight line by least squares.

python
import numpy as np

# Illustrative measurements: concentration of a reactant (mol/L) against time (min).
t = np.array([0, 15, 30, 45, 60, 90, 120, 150, 180, 240], dtype=float)
c = np.array([0.3012, 0.2510, 0.2135, 0.1783, 0.1514, 0.1061, 0.0757, 0.0532, 0.0380, 0.0189])

def straight_line(x, y):
    """Least-squares fit y = a + b x; returns a, b and R^2."""
    b, a = np.polyfit(x, y, 1)
    residuals = y - (a + b * x)
    r2 = 1 - np.sum(residuals**2) / np.sum((y - y.mean())**2)
    return a, b, r2

# Each order predicts a different straight line:
#   zero order:   [A] = [A]0 - k t        first order: ln[A] = ln[A]0 - k t
#   second order: 1/[A] = 1/[A]0 + k t
tests = {"zero   ([A] vs t)": c, "first  (ln[A] vs t)": np.log(c), "second (1/[A] vs t)": 1 / c}
for name, y in tests.items():
    a, b, r2 = straight_line(t, y)
    print(f"{name:20s} slope = {b:+.4e}   R^2 = {r2:.5f}")

a, b, r2 = straight_line(t, np.log(c))
k = -b
print(f"\nfirst order: k = {k:.4e} per min, [A]0 = {np.exp(a):.4f} mol/L, half-life = {np.log(2) / k:.1f} min")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Only the first-order plot is straight (R2=0.99997R^2 = 0.99997, against 0.88 and 0.86 for the others), so the reaction is first order with k=1.15×10−2k = 1.15 \times 10^{-2} per minute and a half-life of 60.2 minutes. You can check the half-life by eye in the data: 0.30 mol/L at the start, 0.15 at 60 minutes, 0.076 at 120.

ComputationC

Temperature and the Arrhenius equation

In 1889 Svante Arrhenius, studying the same sucrose reaction, proposed that rate constants vary with temperature as

k=A e−Ea/RT,k = A\,e^{-E_\text{a}/RT},

where EaE_\text{a} is the activation energy and AA the pre-exponential factor. The exponential is the fraction of molecular encounters with enough energy to cross the barrier, from the Boltzmann distribution; AA collects how often encounters happen and how many have a suitable orientation. Taking logarithms gives a straight line, ln⁡k=ln⁡A−(Ea/R)(1/T)\ln k = \ln A - (E_\text{a}/R)(1/T), so an Arrhenius plot of ln⁡k\ln k against 1/T1/T has slope −Ea/R-E_\text{a}/R. The block fits illustrative rate constants at six temperatures by linear least squares, using NumPy's lstsq, and estimates the uncertainty of EaE_\text{a} from the scatter.

python
import numpy as np

R = 8.314462618                     # J/(mol K)
# Illustrative rate constants (per second) measured at six temperatures.
T = np.array([288.15, 298.15, 308.15, 318.15, 328.15, 338.15])
k = np.array([0.000794, 0.00208, 0.00584, 0.0149, 0.0323, 0.0749])

# Arrhenius: ln k = ln A - (Ea/R)(1/T).  A straight line in x = 1/T: solve by least squares.
X = np.column_stack([np.ones_like(T), 1 / T])          # design matrix: columns 1 and 1/T
coef, rss, rank, sv = np.linalg.lstsq(X, np.log(k), rcond=None)
lnA, slope = coef
Ea = -slope * R

# Standard error of the slope, from the scatter of the points about the line.
n, p = len(T), 2
s2 = rss[0] / (n - p)                                  # residual variance
cov = s2 * np.linalg.inv(X.T @ X)
Ea_err = np.sqrt(cov[1, 1]) * R
print(f"Ea = {Ea / 1000:.1f} ± {Ea_err / 1000:.1f} kJ/mol,   A = {np.exp(lnA):.2e} per s")

for Ti, ki in zip(T, k):
    fit = np.exp(lnA + slope / Ti)
    print(f"T = {Ti:.2f} K   k = {ki:.3e}   fit = {fit:.3e}   ({100 * (ki / fit - 1):+.1f} %)")

# What does the fit predict at body temperature, and how much faster per 10 K near 25 °C?
k_37 = np.exp(lnA + slope / 310.15)
print(f"\npredicted k at 37 °C: {k_37:.2e} per s, half-life {np.log(2) / k_37:.0f} s")
print(f"k(308.15)/k(298.15) = {np.exp(-Ea / R * (1 / 308.15 - 1 / 298.15)):.2f}")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

The fit gives Ea=73.9±0.8E_\text{a} = 73.9 \pm 0.8 kJ/mol, and every measured point lies within 5 % of the line. Near room temperature this reaction speeds up by a factor of 2.63 for 10 K. The old rule that "rates double for every 10 °C" holds only for activation energies near 53 kJ/mol at room temperature; a reaction with a larger barrier is more sensitive to temperature, which is why food keeps so much longer in a freezer.

ExplanationH

Mechanisms and the steady-state approximation

Most reactions happen in several elementary steps, and the rate law of the overall reaction comes from the mechanism. Two ideas make the algebra manageable.

The rate-determining step. If one step is much slower than the others, the overall rate is the rate of that step, as the speed of a queue is set by its slowest counter. For the gas reaction NO₂ + CO → NO + CO₂ below about 500 K, the measured rate law is k[NO2]2k[\mathrm{NO_2}]^2, independent of CO. The mechanism that explains it is 2 NO₂ → NO₃ + NO (slow), then NO₃ + CO → NO₂ + CO₂ (fast). CO enters only after the slow step, so it does not appear in the rate law.

Pre-equilibrium. If a fast, reversible step comes before the slow one, its equilibrium sets the concentration of the intermediate that the slow step uses. The NO + H₂ reaction of the example above is explained this way: 2 NO ⇌ N₂O₂ (fast, equilibrium constant K1K_1), then N₂O₂ + H₂ → N₂O + H₂O (slow, k2k_2), then N₂O + H₂ → N₂ + H₂O (fast). The rate is k2[N2O2][H2]=k2K1[NO]2[H2]k_2[\mathrm{N_2O_2}][\mathrm{H_2}] = k_2K_1[\mathrm{NO}]^2[\mathrm{H_2}]: second order in NO and first in H₂, as measured, because only one H₂ takes part in or before the slow step. A mechanism that fits the rate law is not thereby proved; it survives until an experiment, such as detecting the intermediate, rules it out.

The steady-state approximation. A reactive intermediate I that is made slowly and destroyed quickly never builds up: after a brief start, its concentration stays small and nearly constant, so we may set d[I]/dt≈0d[\mathrm{I}]/dt \approx 0. Max Bodenstein introduced the idea in 1913 for chain reactions. For the simplest case, A → I → P with rate constants k1k_1 and k2k_2, it gives [I]≈k1[A]/k2[\mathrm{I}] \approx k_1[\mathrm{A}]/k_2 and d[P]/dt≈k1[A]d[\mathrm{P}]/dt \approx k_1[\mathrm{A}]: the product appears at the rate of the first step.

How good is that? This mechanism has an exact solution, so we can check the approximation against it, and also against a numerical solution of the three rate equations by the classical fourth-order Runge–Kutta method (numerical methods), which is how real mechanisms with no exact solution are solved.

python
import numpy as np

def rk4(f, y0, t_end, dt):
    """Integrate dy/dt = f(y) from t = 0 with the classical fourth-order Runge-Kutta method."""
    y, t, out = np.array(y0, dtype=float), 0.0, {}
    for step in range(int(round(t_end / dt)) + 1):
        out[round(t, 6)] = y.copy()
        k1 = f(y); k2 = f(y + 0.5 * dt * k1); k3 = f(y + 0.5 * dt * k2); k4 = f(y + dt * k3)
        y = y + dt / 6 * (k1 + 2 * k2 + 2 * k3 + k4)
        t += dt
    return out

# Mechanism A -> I -> P (rate constants k1, k2 in 1/s), starting from [A]0 = 1 mol/L.
for k1, k2 in [(0.1, 2.0), (0.1, 0.2)]:
    rates = lambda y: np.array([-k1 * y[0], k1 * y[0] - k2 * y[1], k2 * y[1]])
    sol = rk4(rates, [1.0, 0.0, 0.0], 40.0, 0.01)
    print(f"k1 = {k1}, k2 = {k2}  (k2/k1 = {k2 / k1:g})")
    print("   t/s   [I] ODE   [I] steady state   [P] ODE   [P] exact   [P] steady state")
    for t in (1, 5, 10, 20, 40):
        A, I, P = sol[t]
        exact_P = 1 - np.exp(-k1 * t) - k1 / (k2 - k1) * (np.exp(-k1 * t) - np.exp(-k2 * t))
        ssa_I, ssa_P = k1 * A / k2, 1 - np.exp(-k1 * t)
        print(f"{t:6d} {I:9.4f} {ssa_I:14.4f} {P:13.4f} {exact_P:10.4f} {ssa_P:14.4f}")
    print()
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

The numerical solution matches the exact one to four decimal places throughout. When the intermediate reacts twenty times faster than it forms, the steady-state [I][\mathrm{I}] is within about 5 % of the truth after the first few seconds. The 5 % is no accident: once the fast start-up has died away, the exact solution gives [I]=k1[A]/(k2−k1)[\mathrm{I}] = k_1[\mathrm{A}]/(k_2 - k_1), so the approximation is low by the factor 1−k1/k2=0.951 - k_1/k_2 = 0.95, and it improves as the ratio grows. The product curve is close too: the remaining gap, about 0.02 mol/L at 10 s, is just the material sitting in I, which the approximation counts as product. When k2k_2 is only twice k1k_1, the intermediate builds up to almost a quarter of the material, and the approximation overestimates the product by a factor of 2.5 at 5 s. The steady state is an approximation for reactive intermediates, not a law.

Worked examplesH

Catalysts and enzymes: a preview

A catalyst speeds up a reaction by providing a different mechanism, with a lower activation energy, and is regenerated at the end. Because it lowers the barrier for the forward and reverse reactions equally, it changes how fast equilibrium is reached, never the equilibrium itself.

EaEa′ΔrHreactantsproductsintermediatereaction coordinateenergyuncatalysedcatalysed
Energy profiles for an uncatalysed reaction (solid) and a catalysed one (dashed). The catalysed route passes through an intermediate bound to the catalyst and has lower barriers, so it is faster; the reactants, products and reaction enthalpy ΔrH are unchanged, and so is the equilibrium constant.

Lowering EaE_\text{a} by even a little has a large effect, because it sits in an exponent: at 298 K, a barrier 10 kJ/mol lower makes a reaction e10 000/(8.314×298)≈57e^{10\,000/(8.314 \times 298)} \approx 57 times faster. Catalysts that are in the same phase as the reactants are homogeneous; solid catalysts that gases or liquids react on are heterogeneous, like the promoted iron of Haber–Bosch and the platinum, palladium and rhodium in a car's catalytic converter. They are the subject of Catalysis and green chemistry.

Enzymes are protein catalysts. In 1913 Leonor Michaelis and Maud Menten, a German biochemist and a Canadian physician-scientist working together in Berlin, described how an enzyme E binds its substrate S, converts it, and releases the product: E + S ⇌ ES → E + P. Applying a steady-state treatment to ES (in the form later given by George Briggs and J. B. S. Haldane) gives the Michaelis–Menten equation

v=Vmax[S]KM+[S],Vmax=kcat[E]0,KM=k−1+kcatk1.v = \frac{V_\text{max}[\mathrm{S}]}{K_\text{M} + [\mathrm{S}]}, \qquad V_\text{max} = k_\text{cat}[\mathrm{E}]_0, \quad K_\text{M} = \frac{k_{-1} + k_\text{cat}}{k_1}.

At low substrate concentration the rate is first order in S; at high concentration the enzyme is saturated and the rate levels off at VmaxV_\text{max}. The block integrates the full mechanism and the Michaelis–Menten rate law side by side, for an enzyme at 1 µmol/L acting on 100 µmol/L of substrate.

python
# Enzyme catalysis:  E + S <=> ES -> E + P   (concentrations in µmol/L, time in s)
k_on, k_off, k_cat = 1.0, 50.0, 10.0       # µM^-1 s^-1, s^-1, s^-1
E0, S0 = 1.0, 100.0
K_M = (k_off + k_cat) / k_on               # Michaelis constant, µM
V_max = k_cat * E0                         # µM/s

def full_model(S, ES):
    E = E0 - ES                            # enzyme is either free or bound
    bind = k_on * E * S - k_off * ES
    return -bind, bind - k_cat * ES, k_cat * ES       # dS/dt, dES/dt, dP/dt

def mm_rate(S):
    return V_max * S / (K_M + S)           # Michaelis-Menten (quasi-steady-state)

# Integrate both with RK4 (plain floats: fast enough for 40 000 steps).
dt, S, ES, P, S_mm = 0.001, S0, 0.0, 0.0, S0
print(f"K_M = {K_M:.0f} µM, V_max = {V_max:.0f} µM/s")
print("   t/s   [ES]    [P] full model   [P] Michaelis-Menten")
for step in range(40001):
    t = step * dt
    if step in (5, 20, 100, 2000, 10000, 20000, 40000):
        print(f"{t:6.3f} {ES:6.3f} {P:15.3f} {S0 - S_mm:20.3f}")
    a = full_model(S, ES)
    b = full_model(S + 0.5 * dt * a[0], ES + 0.5 * dt * a[1])
    c = full_model(S + 0.5 * dt * b[0], ES + 0.5 * dt * b[1])
    d = full_model(S + dt * c[0], ES + dt * c[1])
    S += dt / 6 * (a[0] + 2 * b[0] + 2 * c[0] + d[0])
    ES += dt / 6 * (a[1] + 2 * b[1] + 2 * c[1] + d[1])
    P += dt / 6 * (a[2] + 2 * b[2] + 2 * c[2] + d[2])
    m1 = mm_rate(S_mm); m2 = mm_rate(S_mm - 0.5 * dt * m1)
    m3 = mm_rate(S_mm - 0.5 * dt * m2); m4 = mm_rate(S_mm - dt * m3)
    S_mm -= dt / 6 * (m1 + 2 * m2 + 2 * m3 + m4)
print(f"initial rate predicted by Michaelis-Menten: {mm_rate(S0):.2f} µM/s")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Within a tenth of a second the enzyme–substrate complex reaches its steady level, 0.62 µmol/L (62 % of the enzyme is busy, as [S]/(KM+[S])=100/160[\mathrm{S}]/(K_\text{M} + [\mathrm{S}]) = 100/160 predicts), and from then on the simple rate law tracks the full model to within about 0.2 µmol/L of product. Biochemists measure KMK_\text{M} and kcatk_\text{cat} for an enzyme by fitting this equation to initial rates, much as we fitted Arrhenius above. Enzymes return in Biochemistry.

ExplanationH

Kinetics today

Watching fast reactions. Mixing two solutions takes milliseconds, so faster reactions need other tricks. Stopped-flow instruments mix and stop solutions in about a millisecond and follow the change by absorbance; flash photolysis, for which Ronald Norrish and George Porter shared the 1967 Nobel Prize (with Manfred Eigen, for relaxation methods), starts a reaction with a pulse of light. With femtosecond laser pulses, Ahmed Zewail's group watched bonds break in real time, work recognised by the 1999 Nobel Prize in Chemistry.

Theory. Henry Eyring's transition-state theory (1935) gives the rate constant as k=(kBT/h) e−ΔG‡/RTk = (k_\text{B}T/h)\,e^{-\Delta G^\ddagger/RT}, linking kinetics to the Gibbs energy of the transition state, which quantum chemistry can now compute (Computational chemistry). For many molecules, quantum tunnelling through the barrier, especially of hydrogen atoms, adds a correction.

Modelling. Combustion, atmospheric and industrial chemistry are described by mechanisms with hundreds or thousands of elementary steps. Evaluated rate constants are collected in databases such as the NIST Chemical Kinetics Database, and open-source tools such as Cantera integrate the resulting stiff systems of differential equations with implicit solvers. Mechanisms that large are no longer written entirely by hand: programs such as MIT's Reaction Mechanism Generator enumerate plausible elementary steps and estimate their rate constants, and as of 2026 machine-learned models trained on quantum-chemistry calculations are increasingly used to estimate barriers quickly, with the best estimates still checked against experiment. The same calculations underpin process safety: knowing how fast heat is released at each temperature tells engineers whether a reactor's cooling can keep up, as discussed in Thermochemistry.

Common misconceptionsH

Common misconceptions

Misconception

"The orders in the rate law are the coefficients of the balanced equation." Only for an elementary step. The overall equation NO₂ + CO → NO + CO₂ suggests first order in each, but the measured rate law is second order in NO₂ and zero order in CO. Orders are found by experiment and explained by mechanisms.

Misconception

"A catalyst is not involved in the reaction." It is deeply involved: it binds reactants, forms intermediates and changes the mechanism. It is regenerated at the end of each cycle, so it is not consumed, though in practice catalysts slowly lose activity and must be replaced.

Misconception

"Reactions with a large equilibrium constant are fast." KK depends on ΔrG∘\Delta_\text{r}G^\circ, the difference between reactants and products; the rate depends on the barrier between them. Diamond turning into graphite has K>1K > 1 and is immeasurably slow; some reactions with small KK reach their equilibrium in microseconds.

Misconception

"Reaction rates double for every 10 °C rise." That holds only for activation energies near 53 kJ/mol near room temperature. Our fitted reaction speeds up 2.6-fold per 10 K; reactions with low barriers barely respond to temperature, and enzyme-catalysed reactions slow down above an optimum temperature as the enzyme unfolds.

Exercises

Exercises

Take R=8.314R = 8.314 J/(mol K). Two exercises ask for Python fitting functions; NumPy's polyfit or linalg.lstsq will do the least squares.

1. Chemical kinetics

Warm-upHuman

A catalyst is added to a reaction mixture at constant temperature. Which of these does it change? Select all that apply.

Choose all that apply

2. Chemical kinetics

Warm-upHuman

For the reaction A + 2 B → C, three initial-rate runs give:

Run [A]₀ / mol L⁻¹ [B]₀ / mol L⁻¹ initial rate / mol L⁻¹ s⁻¹
1 0.10 0.10 1.5×10−31.5 \times 10^{-3}
2 0.20 0.10 6.0×10−36.0 \times 10^{-3}
3 0.10 0.30 1.5×10−31.5 \times 10^{-3}

What is the overall order of the reaction?

3. Chemical kinetics

CoreHuman

A reaction 2 A → products is second order in A. Starting from a certain concentration, half of A is gone after 12 minutes. How many minutes after the start does only one-eighth of the original A remain?

4. Chemical kinetics

CoreHuman

A rate constant is 1.0×10−31.0 \times 10^{-3} s⁻¹ at 300 K and 4.0×10−34.0 \times 10^{-3} s⁻¹ at 320 K. Assuming the Arrhenius equation holds with a constant activation energy, what is EaE_\text{a} in kJ/mol, to the nearest whole number? Take R=8.314R = 8.314 J/(mol K).

5. Chemical kinetics

StretchHuman

A proposed mechanism for the decomposition 2 O₃ → 3 O₂ is:

  1. O₃ ⇌ O₂ + O (fast, reversible, equilibrium constant K1K_1)
  2. O + O₃ → 2 O₂ (slow, rate constant k2k_2)

Which rate law does this mechanism predict?

Choose one

6. Chemical kinetics

CoreHuman

An enzyme obeys the Michaelis–Menten equation. At a substrate concentration equal to four times KMK_\text{M}, what percentage of VmaxV_\text{max} is the rate?

7. Chemical kinetics

CoreComputational

Write best_order(t, c), which takes times t and concentrations c of a reactant from one run and returns the order, 0, 1 or 2, whose integrated rate law fits best. For each order, fit a straight line by least squares to the right transform of c against t ([A][\mathrm{A}], ln⁡[A]\ln[\mathrm{A}] or 1/[A]1/[\mathrm{A}]) and compute R2=1−∑(residuals2)/∑(y−yˉ)2R^2 = 1 - \sum(\text{residuals}^2)/\sum(y - \bar y)^2; return the order with the largest R2R^2.

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

8. Chemical kinetics

StretchComputational

Write arrhenius_fit(T, k), which takes temperatures T (in K) and rate constants k and returns a pair (Ea_kJ, A): the activation energy in kJ/mol and the pre-exponential factor (in the units of k), from a least-squares straight line of ln⁡k\ln k against 1/T1/T. Use R=8.314462618R = 8.314462618 J/(mol K).

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Further reading and resources