The Arrhenius Equation Explained: Food Spoilage to Battery Ageing

The Arrhenius equation, k = A e^(−Ea/RT), says that a reaction's rate constant k grows exponentially as absolute temperature T rises, at a pace set by the activation energy Ea. A is the pre-exponential (frequency) factor and R is the molar gas constant, 8.314 462 618 J mol⁻¹ K⁻¹. That one line explains why food keeps longer in the fridge, why a hot car ages a phone battery, and why engineers bake chips at 125 °C to predict how they will fail at room temperature.

This post goes one level deeper than our tour of the laws of thermodynamics and pairs with the ideal gas law explained, which introduces the same gas constant R that sits in the exponent here. Where the thermodynamics post asks whether a change can happen, Arrhenius asks how fast it happens, and how much faster it goes when things warm up.

A note on numbers. The equation and the hydrogen iodide data come from OpenStax Chemistry 2e, the constants from NIST's CODATA tables, the reliability acceleration factors from the NIST/SEMATECH e-Handbook, storage conditions from ICH Q1A(R2) and ASTM F1980, the battery experiment from a peer-reviewed open access paper, and the food safety rules from the US Food and Drug Administration. Activation energies we chose ourselves are labelled as illustrative. Every number was recomputed with the Python script in the walkthrough.

Key takeaways

• The equation: k = A e^(−Ea/RT). k is the rate constant, A the pre-exponential factor (same units as k), Ea the activation energy in J/mol, R = 8.314 462 618 J mol⁻¹ K⁻¹ (exact since the 2019 SI revision), and T the absolute temperature in kelvin.
• The straight line: taking logs gives ln k = ln A − (Ea/R)(1/T). Plot ln k against 1/T and the slope is −Ea/R. That plot, the Arrhenius plot, is how activation energies are measured.
• The two-point form: ln(k₂/k₁) = (Ea/R)(1/T₁ − 1/T₂). It gives the speed-up between two temperatures, or Ea from two measured rates.
• History: Svante Arrhenius published the analysis in 1889 in Zeitschrift für Physikalische Chemie, volume 4, pages 226 to 248, in a paper on the acid-catalysed inversion of cane sugar. His 1903 Nobel Prize in Chemistry was for the electrolytic theory of dissociation, not for this equation.
• "Doubles every 10 degrees" is a rule of thumb, not a law: a reaction doubles between 25 °C and 35 °C only if Ea is about 53 kJ/mol. With Ea = 100 kJ/mol it speeds up 3.7 times; with 20 kJ/mol only 1.3 times. The same Ea gives a smaller ratio at higher temperatures.
• Accelerated ageing: NIST's handbook gives an acceleration factor of 133 between 25 °C and 125 °C for an activation energy of 0.5 eV, and 17,597 for 1.0 eV. The whole prediction rests on that one number, so it has to be measured for the failure mechanism in question.
• Food: temperature slows chemical quality loss in a fridge roughly as Arrhenius predicts, but bacterial growth does not follow it well (Ratkowsky and colleagues, 1982). Food safety comes from the FDA's rules: fridge at 40 °F (4 °C) or below, freezer at 0 °F (−18 °C), perishables refrigerated within 2 hours, or 1 hour above 90 °F. Never use an Arrhenius calculation to stretch those times.
• Breakdowns: Arrhenius behaviour fails when the mechanism changes with temperature, for example when proteins unfold, materials melt or a battery switches dominant ageing path.

What does the Arrhenius equation say?

OpenStax Chemistry 2e states it in the form most textbooks use:

k = A · e^(−Ea / RT)

It describes the rate constant, not the rate. For a simple reaction such as A + B → products, the rate is k[A][B]. Concentrations change as the reaction runs, but k depends only on temperature (and on things like the catalyst and solvent, which we hold fixed). So the Arrhenius equation tells you how the whole rate law scales when you change the temperature and keep everything else the same.

Read the equation from right to left. The exponent −Ea/RT is a ratio of two energies per mole: the energy barrier Ea that colliding molecules must clear, and RT, a measure of the typical thermal energy available at temperature T. At 298 K (25 °C), RT is about 2.48 kJ/mol. A barrier of 50 kJ/mol is about twenty times that, so e^(−Ea/RT) is tiny: 1.7 × 10⁻⁹. Only about two collisions in a billion carry enough energy along the right direction. Warm the system by 10 K to 308 K and that fraction rises to 3.3 × 10⁻⁹, nearly double. The exponential is where all the temperature sensitivity lives.

A sits in front. OpenStax calls it the frequency factor and says it "is related to the frequency of collisions and the orientation of the reacting molecules." If every energetic enough collision produced products, A would be the collision frequency. In practice many collisions have the right energy but the wrong geometry, so A also folds in an orientation, or steric, factor.

Each term, its meaning and its SI unit

The exponent must be dimensionless, so convert Ea from kJ/mol to J/mol before dividing by 8.314, and always use kelvin for T. Degrees Celsius in the exponent give answers that are wildly wrong rather than slightly off.

The gas constant, from CODATA

NIST's CODATA page lists the molar gas constant R as 8.314 462 618... J mol⁻¹ K⁻¹ with the note "exact". That is because the 2019 revision of the SI fixed the Boltzmann constant k_B at exactly 1.380 649 × 10⁻²³ J K⁻¹ and the Avogadro constant at exactly 6.022 140 76 × 10²³ mol⁻¹, and R is their product. OpenStax uses the rounded 8.314 J/mol/K, which is fine for every calculation in this post.

Per mole or per particle: kJ/mol versus eV

Chemists quote activation energies per mole and use R. Physicists and reliability engineers quote them per particle, usually in electronvolts, and use the Boltzmann constant instead:

k = A · e^(−Ea / k_B·T)     with k_B = 8.617 × 10⁻⁵ eV/K

The two are the same equation. One electronvolt per particle is 96.485 kJ/mol, so a semiconductor failure mechanism with Ea = 0.7 eV has the same temperature sensitivity as a chemical reaction with Ea = 67.5 kJ/mol. Keep that conversion handy, because worked example 3 switches between the two.

Three useful forms of the same law

Exponential form, for computing k at one temperature when A and Ea are known:

k = A · exp(−Ea / RT)

Linear form, for fitting data. Take natural logs of both sides:

ln k = ln A − (Ea / R) · (1 / T)

This is the equation of a straight line, y = c + mx, with y = ln k, x = 1/T, slope m = −Ea/R and intercept c = ln A. An Arrhenius plot is ln k on the vertical axis against 1/T on the horizontal axis. A straight line means one mechanism with one activation energy is in charge over that range. Curvature is a signal that something else is going on.

Two-point form, for ratios. Write the linear form at two temperatures and subtract. A cancels:

ln(k₂ / k₁) = (Ea / R) · (1/T₁ − 1/T₂)

This version is the workhorse. It answers "how much faster at T₂ than at T₁?" when you know Ea, and "what is Ea?" when you have measured k₁ and k₂. Because A cancels, you never need to know it for ratio questions, which is why reliability engineers can use the model without ever estimating A.

Who was Svante Arrhenius, and when did he propose the equation?

Svante August Arrhenius was a Swedish chemist. The Nobel Foundation's fact page records that he was born on 19 February 1859 in Vik, Sweden, and died on 2 October 1927 in Stockholm. It notes that he grew up in Uppsala, that his doctoral thesis was criticised at Uppsala University, and that he then went to Germany with help from Wilhelm Ostwald.

The rate equation that carries his name appeared in 1889. Crossref's record for the paper gives the title "Über die Reaktionsgeschwindigkeit bei der Inversion von Rohrzucker durch Säuren" (on the reaction rate of the inversion of cane sugar by acids), Zeitschrift für Physikalische Chemie, volume 4, pages 226 to 248, issued 1 July 1889. Sugar inversion is the acid-catalysed splitting of sucrose into glucose and fructose. The exponential temperature law that came out of that work is the one OpenStax and every other general chemistry text now teach, and the modern reading of it, that only the fraction of collisions energetic enough to clear a barrier reacts, is the collision theory picture in the next section.

He did not start from nothing. Jacobus Henricus van 't Hoff had worked on how temperature affects reaction rates before him. The Nobel Foundation's page for van 't Hoff, the first chemistry laureate in 1901, gives his prize motivation as "the discovery of the laws of chemical dynamics and osmotic pressure in solutions." Keith Laidler's 1984 article "The development of the Arrhenius equation" in the Journal of Chemical Education (volume 61, issue 6, page 494) traces how the competing temperature formulas of that period settled into the form we use.

Here is the part many summaries get wrong. Arrhenius won the 1903 Nobel Prize in Chemistry, alone, but the Nobel Foundation's motivation is "in recognition of the extraordinary services he has rendered to the advancement of chemistry by his electrolytic theory of dissociation." That is his theory that salts, acids and bases split into ions in solution. The rate equation was not cited.

Timeline

How does collision theory explain activation energy and the pre-exponential factor?

OpenStax Chemistry 2e, section 12.5, builds the explanation from three postulates of collision theory: the rate is proportional to the rate of collisions between reactants, the colliding species must be oriented in a way that lets the necessary bonds break and form, and the collision must carry enough energy to get over the barrier. Each postulate maps onto a piece of the Arrhenius equation.

The energy postulate becomes the exponential. In a gas or liquid at temperature T, molecular energies follow a distribution. Most molecules sit near the middle, and a thin tail reaches high energies. The fraction of collisions with energy above Ea is approximately e^(−Ea/RT). Raise T and the whole distribution shifts and flattens, and the tail above Ea grows much faster than the average energy does. That is why a 3 percent rise in absolute temperature (298 K to 308 K) can nearly double the rate: the average barely moves, but the population in the high energy tail roughly doubles.

The orientation postulate goes into A. OpenStax says A "reflects how well the reaction conditions favor properly oriented collisions between reactant molecules." Two molecules that hit with plenty of energy but with the wrong ends facing each other just bounce apart. Large, floppy molecules with a small reactive site have a low orientation factor and a smaller A.

The collision frequency also goes into A. Faster moving molecules collide more often, so the collision frequency itself rises slowly with temperature. That weak dependence is one reason a Tⁿ term appears in the modified Arrhenius form discussed later; over a modest range it is small next to the exponential, so A is treated as constant.

Activation energy is a barrier, not a release. The activation energy is the height of the energy hill between reactants and the highest point along the path, the activated complex or transition state. It is different from the overall energy change of the reaction, ΔH. A strongly exothermic reaction can still have a large Ea and be slow at room temperature; that is why paper does not burst into flame on a desk, even though burning it releases a great deal of energy. A catalyst works by providing a path with a lower barrier. It lowers Ea for the forward and reverse reactions alike and does not change ΔH or the position of equilibrium.

Worked example 1: how much faster from 25 °C to 35 °C?

Take a reaction with an activation energy of 50 kJ/mol. How much faster does it run at 35 °C than at 25 °C?

Step 1: convert to kelvin. T₁ = 25 + 273.15 = 298.15 K. T₂ = 35 + 273.15 = 308.15 K.

Step 2: compute the reciprocal difference.

1/T₁ − 1/T₂ = 1/298.15 − 1/308.15
            = 0.0033540 − 0.0032452
            = 1.0884 × 10⁻⁴ K⁻¹

Step 3: apply the two-point form.

ln(k₂/k₁) = (Ea/R) · (1/T₁ − 1/T₂)
          = (50,000 / 8.314) × 1.0884 × 10⁻⁴
          = 6,014 × 1.0884 × 10⁻⁴
          = 0.6546
k₂/k₁ = e^0.6546 = 1.92

The reaction runs 1.92 times faster: almost, but not quite, double.

Where "doubles every 10 degrees" comes from

Turn the question around: what activation energy makes a reaction exactly double between 25 °C and 35 °C? Set k₂/k₁ = 2:

Ea = R · ln 2 / (1/T₁ − 1/T₂)
   = 8.314 × 0.6931 / 1.0884 × 10⁻⁴
   = 52,900 J/mol ≈ 53 kJ/mol

Many reactions people meet at room temperature, in cooking, in biology and in the slow chemistry of materials, have activation energies in the range of tens of kilojoules per mole up to around 100 kJ/mol. So the factor for a 10 degree rise near room temperature is often somewhere between about 1.5 and 4, and "about two" is the easy number to remember. Chemists and biologists call this ratio Q10, the temperature coefficient over 10 degrees.

Where the rule of thumb fails

The rule fails in two distinct ways, and the table makes both visible.

It depends on Ea. Here is the 25 °C to 35 °C ratio for a range of activation energies:

It depends on where you start. Q10 is not a constant even for one reaction, because 1/T₁ − 1/T₂ for a fixed 10 kelvin step shrinks as the temperatures rise. Take the 52.9 kJ/mol reaction that exactly doubles between 25 °C and 35 °C and move the 10 degree window:

A constant Q10 and the Arrhenius equation are two different models that agree only over a narrow band. Over a wide range, an extrapolation that assumes Q10 = 2 everywhere will disagree with an Arrhenius extrapolation from the same starting point. That matters for accelerated ageing, where tests often run 30 degrees or more above the use temperature (see worked example 3).

Worked example 2: finding Ea from measured rate constants

Now go the other way. You have measured the rate constant at several temperatures and want the activation energy. OpenStax's Example 12.13 uses the decomposition of hydrogen iodide, 2HI(g) → H₂(g) + I₂(g), with these measured second order rate constants:

The rate constant rises by a factor of more than 100,000 over a 226 kelvin span. That is what a large activation energy looks like.

Method A: the two-point form

Use the first and last rows:

ln(k₂/k₁) = ln(3.95 × 10⁻² / 3.52 × 10⁻⁷) = ln(112,216) = 11.628
1/T₁ − 1/T₂ = 1/555 − 1/781 = 1.8018 × 10⁻³ − 1.2804 × 10⁻³ = 5.214 × 10⁻⁴ K⁻¹
Ea = R · ln(k₂/k₁) / (1/T₁ − 1/T₂)
   = 8.314 × 11.628 / 5.214 × 10⁻⁴
   = 185,000 J/mol ≈ 185 kJ/mol

OpenStax rounds the denominator to 0.52 × 10⁻³ and the slope to −2.2 × 10⁴ K, which gives its published answer of 180 kJ/mol. Carrying more digits gives 185 kJ/mol. Both are right to the precision the rounding allows; the difference is a reminder that subtracting two close numbers like 1/555 and 1/781 amplifies rounding error, so keep at least four significant figures in 1/T.

Method B: the Arrhenius plot with a least squares fit

With five points, a straight line fit uses all the data and gives an uncertainty estimate. OpenStax notes that "in practice, the equation of the line ... would be derived using a statistical process called regression." Fitting ln k against 1/T for the five rows:

slope       = −22,287 K
intercept   = 25.204
R²          = 0.9996
Ea = −slope × R = 22,287 × 8.314 = 185,300 J/mol ≈ 185 kJ/mol
A  = e^25.204 = 8.8 × 10¹⁰ L mol⁻¹ s⁻¹

The R² of 0.9996 says one straight line describes these five points very well, so a single activation energy is a fair summary over 555 K to 781 K. The fitted A of about 10¹¹ L mol⁻¹ s⁻¹ is in the range collision theory leads you to expect for a simple gas reaction.

What the fit lets you do, and what it does not

Inside the measured range, you can interpolate. The fitted line predicts k = 6.5 × 10⁻⁶ L mol⁻¹ s⁻¹ at 600 K, between the 575 K and 645 K measurements.

Far outside the range, be careful. The same line predicts k ≈ 3 × 10⁻²² L mol⁻¹ s⁻¹ at 298 K, which says hydrogen iodide gas is effectively stable at room temperature by this route. The direction is right, but the number is an extrapolation 257 kelvin beyond the coldest data point, across a region where a different mechanism (for example, light-driven or surface-catalysed decomposition) could dominate. Treat long extrapolations as hypotheses, not measurements. This is the same trap that awaits accelerated ageing tests.

Worked example 3: accelerated ageing of electronics and batteries

Nobody can wait ten years to find out whether a design lasts ten years. Accelerated ageing tests run the product hotter, so the chemistry that wears it out runs faster, then use a model, for temperature-driven wear-out usually Arrhenius, to translate test hours into field years.

The acceleration factor

The NIST/SEMATECH e-Handbook of Statistical Methods, section 8.1.5.1, describes the Arrhenius model as "one of the earliest and most successful acceleration models" for how time to fail varies with temperature. In the handbook's form, the time to failure is proportional to exp(ΔH/kT), where ΔH is the activation energy in electronvolts and k is Boltzmann's constant, 8.617 × 10⁻⁵ eV/K. The acceleration factor between a higher test temperature T₂ and a lower use temperature T₁ is:

AF = exp[ (ΔH / k) · (1/T₁ − 1/T₂) ]

It is the two-point form again, written for time to failure rather than rate. A rate that is AF times faster means a life that is AF times shorter.

NIST makes three points worth quoting closely. First, the constant in front "drops out when calculating acceleration factors", so only the activation energy matters. Second, ΔH "depends on the failure mechanism and the materials involved, and typically ranges from 0.3 or 0.4 up to 1.5, or even higher." Third, the model "has been used successfully for failure mechanisms that depend on chemical reactions, diffusion processes or migration processes," which covers many non-mechanical failures in electronics.

Numbers: chips baked at 125 °C

The NIST handbook gives its own example: between 25 °C and 125 °C, the acceleration factor is 133 if ΔH = 0.5 eV and 17,597 if ΔH = 1.0 eV. Recomputing:

T₁ = 298.15 K, T₂ = 398.15 K
1/T₁ − 1/T₂ = 8.424 × 10⁻⁴ K⁻¹
ΔH = 0.5 eV:  AF = exp(0.5 / 8.617×10⁻⁵ × 8.424×10⁻⁴) = exp(4.888) = 133
ΔH = 0.7 eV:  AF = exp(6.843) = 937
ΔH = 1.0 eV:  AF = exp(9.776) = 17,600

(NIST uses 273.16 for the Celsius offset; with 273.15 the 1.0 eV case comes out as 17,600 rather than 17,597. The difference does not matter.)

Read what those numbers mean for a 1,000 hour test at 125 °C. If the failure mechanism has ΔH = 0.5 eV, 1,000 hours at 125 °C stands in for about 133,000 hours, roughly 15 years, at 25 °C. If ΔH is really 0.7 eV, the same test stands in for about 937,000 hours. Guess 0.7 eV when the truth is 0.5 eV, and you overestimate field life by a factor of seven. The model is only as good as the activation energy, which is why NIST emphasises that ΔH "is all you need to know to calculate temperature acceleration" and that it depends on the specific mechanism.

Numbers: battery calendar ageing

Lithium-ion batteries age in two broad ways: through use (cycle ageing) and simply through time spent sitting at a given charge and temperature (calendar ageing). Calendar ageing is dominated by slow side reactions, which makes it a natural candidate for an Arrhenius treatment.

A peer-reviewed open access study by Peter Keil and colleagues, "Calendar Aging of Lithium-Ion Batteries" (Journal of The Electrochemical Society, volume 163, 2016, pages A1872 to A1880), stored three types of commercial 18650 cells with different cathode materials at 16 states of charge from 0 to 100 percent, in thermal chambers held at 25 °C, 40 °C and 50 °C. Capacity was checked periodically at 25 °C. The authors found that capacity fade was mainly caused by a shift in electrode balancing, driven by loss of cyclable lithium, and that low graphite anode potentials, which promote electrolyte reduction and growth of the solid electrolyte interphase (SEI) layer, were the main driver of fade during storage. They concluded that high storage states of charge, which correspond to low anode potential, "should be avoided" to maximise battery life.

To see how an engineer turns a hot storage test into a field estimate, take the study's three temperatures and assume, purely for illustration, three possible activation energies for the dominant fade reaction:

With an illustrative Ea of 50 kJ/mol, each month at 50 °C counts as 4.76 months at 25 °C, so ten years of storage at 25 °C would correspond to about 25 months at 50 °C. Notice how sensitive that is: at 30 kJ/mol the same ten years would need 47 months at 50 °C; at 70 kJ/mol only 13.5 months.

The limits, for batteries in particular

The Keil study is a good illustration of why a battery cannot be described by one Arrhenius line.

• State of charge matters as much as temperature. The study found that calendar ageing "does not increase steadily" with state of charge; instead there are plateaus spanning 20 to 30 percent of capacity where fade is similar, with step changes between them linked to the graphite anode's potential. An Arrhenius fit at one state of charge does not transfer to another.
• Mechanisms change. The authors observed signs of high voltage mechanisms only in a few cases, notably a substantially accelerated fade for one cell chemistry at 100 percent state of charge. If a hot test switches on a mechanism that is negligible at room temperature, the acceleration factor from the dominant room temperature mechanism no longer applies.
• Fade is not always linear in time. The capacity lost per month need not stay constant as a cell ages, so the "amount of ageing" is not simply a rate times a time. Life models apply the acceleration factor to an ageing law fitted to measured data, and the choice of law matters as much as the Ea.
• Hotter is not always better for the test. Push the temperature far enough and you reach behaviour that never occurs in service. Hot battery testing belongs in purpose-built thermal chambers under the test lab's own safety procedures. It is not something to try with a consumer device, and nothing in this post is advice to heat, charge or store a battery outside its maker's instructions.

How other regulated industries use the same idea

Accelerated ageing is common far beyond electronics, and regulators are careful about how far they trust it.

Medicines. The ICH Q1A(R2) guideline on stability testing of new drug substances and products defines accelerated testing as "studies designed to increase the rate of chemical degradation or physical change of a drug substance or drug product by using exaggerated storage conditions." Its general case pairs a long term condition of 25 °C ± 2 °C at 60 percent relative humidity (or 30 °C at 65 percent) with an accelerated condition of 40 °C ± 2 °C at 75 percent relative humidity for 6 months, sampled at a minimum of three time points such as 0, 3 and 6 months. Stress testing goes further, in 10 °C increments above the accelerated condition, such as 50 °C and 60 °C. The accelerated data support, but do not replace, long term data.

Sterile medical packaging. ASTM F1980 is a guide to accelerated ageing of sterile barrier systems and medical devices. Its scope says information from the guide "may be regarded as sufficient evidence for expiration date claims ... until data from real-time aging studies are available." Its summary says real time ageing studies "must be conducted in parallel" with accelerated ones, that conservative accelerated ageing factors must be used when little is known about the material, and that a shelf life based on a conservative estimate of the ageing factor, expressed as a Q10, is tentative until the real time results are complete. To see what a Q10 implies, take an illustrative Q10 of 2 applied over 30 degrees: the factor is 2³ = 8, so roughly 46 days at 55 °C would stand for one year at 25 °C. An Arrhenius model with the Ea that gives exactly Q10 = 2 at 25 °C to 35 °C would instead give a factor of 7.0 over the same 30 degrees, because Q10 drifts downward as temperature rises. The two methods are close, but not identical, and the standard's own caution, that real time data must follow, applies to both.

Worked example 4: why the fridge works, and why food safety is a different question

Food changes in two different ways, and the Arrhenius equation applies to them very differently.

Quality loss: chemistry that follows Arrhenius reasonably well

Much of what makes food stale, rancid, discoloured or flat tasting is ordinary chemistry: fats oxidising, vitamins breaking down, sugars and proteins browning, flavours reacting. These reactions slow down in the cold in roughly the way the Arrhenius equation predicts, which is why "best before" testing for shelf stable foods often uses warmer storage to speed up the work.

Suppose, for illustration, that the main quality loss reaction in a chilled product has Ea = 80 kJ/mol, and that the product keeps acceptable quality for 14 days at 4 °C. How long would it keep at 10 °C, or at 20 °C?

T₁ = 277.15 K (4 °C), T₂ = 283.15 K (10 °C), T₃ = 293.15 K (20 °C)

Ratio 4 °C → 10 °C:  exp[(80,000/8.314) × (1/277.15 − 1/283.15)] = 2.09
Ratio 4 °C → 20 °C:  exp[(80,000/8.314) × (1/277.15 − 1/293.15)] = 6.65

Quality life at 10 °C ≈ 14 / 2.09 ≈ 6.7 days
Quality life at 20 °C ≈ 14 / 6.65 ≈ 2.1 days

Swap the illustrative Ea and the answer moves a lot: at 50 kJ/mol the 4 °C to 20 °C ratio is 3.3, at 100 kJ/mol it is 10.7. A fridge that runs a few degrees warm costs real quality days, and a product left out loses quality several times faster.

Safety: bacteria do not follow the Arrhenius line

The tempting next step is to say "bacteria grow faster when warm too, so use the same equation to decide how long food is safe." Do not. A peer-reviewed study by D. A. Ratkowsky, J. Olley, T. A. McMeekin and A. Ball in the Journal of Bacteriology (volume 149, 1982, pages 1 to 5) found that the Arrhenius law "does not adequately describe the effect of temperature on bacterial growth." When microbiologists plotted the logarithm of growth rate against reciprocal absolute temperature, they got curves rather than straight lines. The authors proposed instead that the square root of the growth rate is linear in temperature, √r = b(T − T₀), where T₀ is a notional minimum growth temperature specific to the organism. They found this held for a wide range of bacteria and also for yeasts, moulds and nucleotide breakdown. A follow-up paper in 1983 (Journal of Bacteriology, volume 154, pages 1222 to 1226) extended the model to the full temperature range of growth, adding minimum and maximum temperature bounds, and reported that it applied to all 30 bacterial strains for which the authors could obtain data.

That result has a practical edge. The Arrhenius curve flattens toward zero rate only as temperature approaches absolute zero; the square root model reaches zero growth at a real temperature, T₀, and different organisms have different T₀ values, some of which grow slowly even at fridge temperatures. A pathogen's growth does not shrink smoothly the way a quality reaction does, and food safety limits are set from microbiological evidence, not from a chemistry formula.

So food safety comes from official rules, not from this post's arithmetic. The US Food and Drug Administration's safe food handling guidance says to use an appliance thermometer to make sure the refrigerator is consistently at 40 °F (4 °C) or below and the freezer at 0 °F (−18 °C) or below; to refrigerate or freeze meat, poultry, eggs, seafood and other perishables within 2 hours of cooking or purchasing, or within 1 hour if the outside temperature is above 90 °F (32 °C); and never to thaw food at room temperature on the counter. Those are the numbers to use. An Arrhenius calculation can explain why cold storage slows quality loss; it can never justify leaving perishable food out longer than those limits.

Where do real rate data come from?

The textbook examples use clean numbers. Real kinetics depend on measured rate constants, collected from thousands of papers and evaluated for quality.

The largest public compilation for gas phase reactions is the NIST Chemical Kinetics Database, Standard Reference Database 17. Its home page describes it as "a compilation of kinetics data on gas-phase reactions" and, as of this writing, lists Version 7.1 (web version), release 1.6.8, with a 2026 data version. You can search by reactants and products, or by bibliographic reference, and NIST runs targeted evaluations in specific areas. NIST also accepts requests to abstract kinetics data from journal articles that the database has missed.

What you find in an entry is typically a rate expression fitted by the original authors, the temperature range over which it was measured, the experimental or theoretical method, and the reference. The database presents rate expressions in a three-parameter Arrhenius form with a reference temperature, A·(T/T_ref)ⁿ·exp(−Ea/RT); its unit preferences page notes that setting the reference temperature to 1.0 displays the plain Tⁿ form. The same page lets you choose volume units (cm³ or L), per molecule or per mole, and energy units of K, J, kJ, cal or kcal. Three habits make that data safe to use:

1. Check the temperature range. A rate expression measured from 300 K to 500 K says nothing reliable about 1,500 K. Use it inside its range or treat the extrapolation as a hypothesis.
2. Check the units. Gas phase kinetics is often reported per molecule in cm³ (for a second order reaction, cm³ molecule⁻¹ s⁻¹) rather than per mole in litres. Converting between them involves Avogadro's number and a factor of 1,000. Energies may be given as Ea/R in kelvin rather than in kJ/mol; multiply by R to convert.
3. Compare several measurements. For important reactions there are often many entries that disagree. Evaluated recommendations from review panels exist precisely because individual measurements scatter.

For reactions in solution, in solids, in food or in devices, there is no single database; data come from the literature, standard test methods and your own measurements.

When does Arrhenius behaviour break down?

A straight Arrhenius plot says one mechanism with one barrier controls the rate across the range. The plot curves, or the model fails outright, when that stops being true. Common cases:

• Biology. Enzymes speed up with temperature until the protein starts to unfold, then activity collapses. Whole organisms show the curved behaviour Ratkowsky and colleagues documented. A single Ea cannot describe growth from fridge to body temperature and beyond.
• Phase changes and glass transitions. When a fat melts, a polymer passes its glass transition, or water freezes, molecular mobility changes abruptly. A frozen food can still degrade in pockets of unfrozen, concentrated liquid, at rates a simple extrapolation from warm temperatures would not predict.
• Mechanism switching. A reaction that goes by one path at low temperature can switch to a different path with a different Ea at high temperature. The Arrhenius plot then shows two lines with a bend between them. This is the main risk in accelerated ageing.
• Diffusion control. Very fast reactions in solution can be limited by how quickly reactants find each other rather than by the chemical barrier. The apparent Ea then reflects the temperature dependence of the solvent's viscosity.
• Quantum tunnelling. For reactions that move a hydrogen atom, especially at low temperatures, particles can pass through the barrier rather than over it. Rates then fall off much more slowly with cooling than Arrhenius predicts.
• Wide temperature ranges. Even for simple gas reactions, A has a mild temperature dependence that matters over thousands of kelvin, as in combustion. That is what the modified Arrhenius form is for.
• Apparent negative activation energies. Some reactions go slower when heated, usually because they proceed through a weakly bound intermediate that falls apart more easily at higher temperatures. The fitted Ea is below zero, which is a sign that the elementary steps need separate treatment.

Comparison: Arrhenius, Eyring and the modified forms

The Arrhenius equation is empirical: it was found to fit data before anyone could derive it. Later theories refined it. Here is how the common temperature models compare.

What Eyring adds

Henry Eyring's 1935 paper in the Journal of Chemical Physics, "The Activated Complex in Chemical Reactions," gave the rate constant a statistical mechanics foundation. Transition state theory pictures the reactants in a quasi-equilibrium with an activated complex at the top of the barrier, which then crosses to products at a universal frequency k_B·T/h. Using the exact CODATA values of the Boltzmann constant (1.380 649 × 10⁻²³ J/K) and the Planck constant (6.626 070 15 × 10⁻³⁴ J s), that frequency is 6.2 × 10¹² s⁻¹ at 298.15 K.

The practical gain is that Eyring splits the barrier into an enthalpy part, ΔH‡, and an entropy part, ΔS‡. The entropy of activation says how ordered the transition state is compared with the reactants. A large negative ΔS‡ suggests two molecules locking together in a tight, specific geometry; a positive one suggests a bond loosening and the complex gaining freedom. That is information the Arrhenius A hides. For a simple reaction in solution, the two are linked by Ea = ΔH‡ + RT, so over modest temperature ranges an Arrhenius fit and an Eyring fit describe the same data equally well.

Walkthrough: reproduce every number in Python

Every number in the worked examples came from a short script. Here is how to rebuild it in a clean, pinned environment, so it gives the same answer next year.

Step 1: create an isolated environment

mkdir arrhenius && cd arrhenius
python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
pip install "numpy==2.1.3"
pip freeze > requirements.txt

Pinning numpy to an exact version and saving requirements.txt means a colleague, or a CI job, gets the same library you tested with.

Step 2: write the calculations

Save this as arrhenius.py:

import math
import numpy as np

R = 8.314462618          # J mol^-1 K^-1, CODATA (exact)
KB_EV = 8.617333262e-5   # eV K^-1, Boltzmann constant in eV
K = lambda c: c + 273.15 # Celsius to kelvin

def ratio(ea_j_mol, t1_k, t2_k):
    """k(T2)/k(T1) from the two-point Arrhenius form."""
    return math.exp(ea_j_mol / R * (1 / t1_k - 1 / t2_k))

def accel_factor_ev(ea_ev, t_use_k, t_test_k):
    """NIST-style acceleration factor with Ea in eV."""
    return math.exp(ea_ev / KB_EV * (1 / t_use_k - 1 / t_test_k))

# Worked example 1: 25 to 35 C
print("Ea 50 kJ/mol, 25->35 C:", round(ratio(50e3, K(25), K(35)), 2))
ea_double = R * math.log(2) / (1 / K(25) - 1 / K(35))
print("Ea for exact doubling:", round(ea_double / 1e3, 1), "kJ/mol")

# Worked example 2: OpenStax HI data
T = np.array([555, 575, 645, 700, 781], dtype=float)
k = np.array([3.52e-7, 1.22e-6, 8.59e-5, 1.16e-3, 3.95e-2])
slope, intercept = np.polyfit(1 / T, np.log(k), 1)
print("Fitted Ea:", round(-slope * R / 1e3, 1), "kJ/mol")
print("Fitted A :", f"{math.exp(intercept):.2e}", "L/mol/s")
two_point = R * math.log(k[-1] / k[0]) / (1 / T[0] - 1 / T[-1])
print("Two-point Ea:", round(two_point / 1e3, 1), "kJ/mol")

# Worked example 3: NIST handbook acceleration factors
for ea in (0.5, 0.7, 1.0):
    print(f"AF {ea} eV, 25->125 C:", round(accel_factor_ev(ea, K(25), K(125))))

# Worked example 4: illustrative food quality reaction
for c in (10, 20):
    r = ratio(80e3, K(4), K(c))
    print(f"4->{c} C ratio {r:.2f}; 14-day quality life becomes {14 / r:.1f} days")

Step 3: run it and compare

python arrhenius.py

You should see:

Ea 50 kJ/mol, 25->35 C: 1.92
Ea for exact doubling: 52.9 kJ/mol
Fitted Ea: 185.3 kJ/mol
Fitted A : 8.83e+10 L/mol/s
Two-point Ea: 185.4 kJ/mol
AF 0.5 eV, 25->125 C: 133
AF 0.7 eV, 25->125 C: 937
AF 1.0 eV, 25->125 C: 17600
4->10 C ratio 2.09; 14-day quality life becomes 6.7 days
4->20 C ratio 6.65; 14-day quality life becomes 2.1 days

If any number differs in the first two significant figures, check first that you used kelvin and joules, which are the two errors that change the answer by orders of magnitude.

Step 4: add a sanity check for curvature

Before trusting one Ea, look at the residuals of the straight line fit. Systematic curvature, with residuals positive at both ends and negative in the middle or the reverse, means more than one mechanism is involved.

pred = intercept + slope / T
residuals = np.log(k) - pred
for t, r in zip(T, residuals):
    print(f"{t:.0f} K residual {r:+.3f}")

For the hydrogen iodide data the residuals run +0.094, −0.060, −0.012, −0.124 and +0.102: no larger than about 0.12 in ln k, or roughly 12 percent in k. The ends sit slightly above the line and the middle slightly below, but with only five points and that much scatter it is not evidence of a second mechanism. With more data over a wider range, that is exactly the pattern to watch for.

Step 5: keep the script's dependencies healthy

A calculation script is software, and software ages too. If the script grows into a lab tool, a shelf life calculator or a battery test dashboard, it will pull in more libraries. Keep requirements.txt pinned, rebuild the environment from it rather than by hand, and scan the project for known vulnerable components before you share it:

safeguard scan --dir .

Results are written as JSON under ./output, so they can be checked into a review or picked up by CI.

Common mistakes

• Using Celsius in the exponent. T must be in kelvin. Using 25 instead of 298.15 changes the answer by many orders of magnitude.
• Mixing kJ and J. Ea in kJ/mol divided by R in J/(mol·K) gives an exponent 1,000 times too small. Convert first.
• Treating "doubles every 10 degrees" as a law. It holds only for Ea near 53 kJ/mol around room temperature, and Q10 drifts with temperature even for one reaction.
• Confusing activation energy with reaction enthalpy. Ea is the height of the barrier; ΔH is the net energy change. A very exothermic reaction can still be slow.
• Extrapolating far beyond the data. An Arrhenius line fitted from 555 K to 781 K says little that is reliable about 298 K. Every accelerated ageing prediction is an extrapolation and needs real time confirmation.
• Guessing the activation energy in an accelerated test. As NIST's examples show, 0.5 eV versus 1.0 eV changes the acceleration factor from 133 to 17,597. Measure it for the mechanism, ideally at three or more temperatures.
• Assuming one mechanism across a phase change. Melting, freezing and glass transitions change mobility and can change the controlling reaction.
• Ignoring other stresses. Humidity, state of charge, voltage, light and mechanical load all change ageing. ICH Q1A(R2) pairs every temperature with a relative humidity for a reason, and the Keil battery study shows state of charge can matter as much as temperature.
• Applying chemistry models to food safety. Bacterial growth does not follow Arrhenius well. Use the FDA's time and temperature rules for safety decisions.

FAQ

What is the Arrhenius equation in simple terms?

It says that reactions speed up exponentially as temperature rises, and that the size of the speed-up is set by the activation energy, the energy barrier molecules must clear. The formula is k = A e^(−Ea/RT), with T in kelvin and R = 8.314 J mol⁻¹ K⁻¹.

What are the units of the Arrhenius equation?

Ea is in J/mol (usually quoted in kJ/mol), R is in J mol⁻¹ K⁻¹, T is in K, so Ea/RT has no units. A has the same units as k, which depend on reaction order: s⁻¹ for first order, L mol⁻¹ s⁻¹ for second order. In reliability work Ea is in eV and the Boltzmann constant in eV/K replaces R.

How do you calculate activation energy from two temperatures?

Use ln(k₂/k₁) = (Ea/R)(1/T₁ − 1/T₂) and solve for Ea: Ea = R·ln(k₂/k₁)/(1/T₁ − 1/T₂). Keep temperatures in kelvin and carry at least four significant figures in 1/T, because you are subtracting two close numbers.

Why do reactions double every 10 degrees?

They do not always. A doubling between 25 °C and 35 °C corresponds to an activation energy of about 53 kJ/mol. Many everyday reactions have activation energies in the tens of kJ/mol, so "about double" is a fair rough guide near room temperature, but the true factor can be 1.3 or 7 depending on Ea, and it shrinks at higher temperatures.

What does the Arrhenius plot show?

It plots ln k against 1/T. For a reaction with one mechanism the points fall on a straight line with slope −Ea/R and intercept ln A. Curvature means the mechanism or the controlling step changes over the range.

What is the difference between the Arrhenius and Eyring equations?

Arrhenius is empirical, with two fitted parameters, A and Ea. Eyring comes from transition state theory and expresses the rate constant through the enthalpy and entropy of activation, with a universal prefactor k_B·T/h. Over modest ranges both fit data equally well; Eyring's entropy term gives extra insight into mechanism.

How are accelerated ageing tests calculated?

The test temperature and use temperature are plugged into an acceleration factor, AF = exp[(Ea/k_B)(1/T_use − 1/T_test)] with Ea in eV, or an equivalent Q10 formula. Test time multiplied by AF estimates equivalent field time. The estimate is only as good as the activation energy and the assumption that the same mechanism operates at both temperatures, which is why standards treat it as provisional until real time data arrive.

Does the Arrhenius equation apply to food spoilage?

Partly. Chemical quality loss, such as fat oxidation or vitamin loss, often follows Arrhenius behaviour reasonably well. Bacterial growth does not, according to Ratkowsky and colleagues' 1982 study, which proposed a square root model instead. For safety, follow FDA guidance: fridge at 40 °F (4 °C) or below, perishables refrigerated within 2 hours, or 1 hour above 90 °F.

Why do batteries age faster when hot?

Calendar ageing in lithium-ion cells is driven largely by slow side reactions, such as electrolyte reduction and growth of the solid electrolyte interphase at the graphite anode, which consume cyclable lithium. Like most chemical reactions, these speed up with temperature. A peer-reviewed study of cells stored at 25, 40 and 50 °C also found that high states of charge accelerate this fade, so temperature and charge level both matter.

The software around the chemistry: an honest bridge

The Arrhenius equation is chemistry, and Safeguard does not do chemistry. Its models are single purpose security models for software. They do not calculate rate constants, predict shelf life, estimate battery life or analyse test data, and nothing in this post was computed by them. The connection is narrower. Stability study systems, battery test rigs, cold chain monitoring dashboards and laboratory data pipelines that apply these equations are software, and much of that code, the instrument drivers, the data services, the web dashboards and analysis scripts like the one in the walkthrough, is built from open source dependencies. That is where Safeguard helps: scanning repositories, container images and SBOM files for known vulnerable open source components, showing which projects carry a given version, and, when a fixed version exists, opening an upgrade pull request for review.

Every rebuild of a lab data service or test dashboard pulls its dependencies again, and the controllers and loggers inside thermal chambers, battery cyclers and cold chain sensors ship open source components of their own. The fit is indirect for the kinetics itself, but for the code your team builds and ships around these systems the same dependency hygiene applies, and the section below shows how to close the gap whether or not upstream has a fix yet.

Fix it without waiting for upstream

Whatever the vulnerability, most of the risk sits in two gaps. The first is the time between a fix being published and that fix actually running in your production builds. The second, and worse, is the time before any fix exists at all, while a maintainer works out a patch and a release schedule. Waiting through both gaps is a habit, not a requirement. Safeguard gives teams three ways to close them.

1. Autonomous remediation: pull requests, not tickets

When a fixed version exists, Safeguard's Griffin models open a tested, policy gated pull request that moves the affected dependency to the safe version. You can run it on one repository or in bulk across every project that carries the vulnerable version. Remediation covers eight categories, including dependencies, licenses, security posture and findings. Reachability analysis ranks the vulnerable code paths your application actually calls first, so the first pull requests you merge are the ones that matter most. Autonomous remediation ships switched off, with every category set to Manual, so your team decides how much runs on its own and what still needs a human approval.

2. Gold Open Source: the safe version in seconds, free

gold.safeguard.sh is free and needs no login. Search any CVE or package to see affected and fixed versions, CVSS, EPSS, CISA KEV status and a recommended safe version, or use the "Am I affected?" checker. The same data is available through a free read API, README status badges and RSS threat feeds. The safeguard-sh/gold-scan-action GitHub Action can fail a build on a known exploited vulnerability, on any known vulnerability, or never, so a version you have just removed cannot quietly come back in next week's merge.

3. Gold Registry: start from components that are already fixed

registry.safeguard.sh provides hardened, drop in replacements for upstream packages and container images. Every Gold artifact is held continuously to zero exploitable critical or high CVEs, zero malware and a completed license check, with SLSA L3 or higher provenance, a Sigstore signature and a public evidence page you can hand to an auditor. An artifact loses its Gold badge if it regresses. Teams can point their own package directory at the registry so hardened equivalents resolve without changing how developers install anything.

When upstream has no fix yet, or will not ship one, Safeguard first works to get the fix accepted upstream. If the issue is severe or upstream will not move, Safeguard builds a fork, fixes the vulnerability there and publishes a hardened Gold build, so you are not stuck on a maintainer's release schedule. You can also request a hardened, compatibility tested build of a specific package. This fork and patch flow is in early access today, so ask which packages it covers before you plan around it.

One honest limit: when the vulnerable code lives in a vendor appliance, a hosted service or an operating system runtime you do not build yourself, the fix still has to come from that vendor or distribution. The steps above apply to the open source dependencies and container images you build and ship, which is where most teams carry the largest and least tracked share of their exposure.

See how Safeguard would help

None of these demos involve chemistry or kinetics calculations; they show the software side described in the bridge above.

The first demo is a real Safeguard CLI session: it installs the CLI, opens the interactive terminal and asks it to fix a lodash CVE in a repository. It reads the finding and advisory, asks before each command, edits package.json to lodash 4.17.21, runs npm install and npm test with approval, prints a summary and pushes nothing. The same flow applies to a stability study dashboard or a cold chain monitoring web app your team maintains: scan the repository, then let the assistant propose the dependency upgrade under your approval.

Click through it at your own pace, or open it full screen.

The second demo sets up the Safeguard runner on your own machine, approves it with its device code and then scans from the terminal. That suits lab and test engineering teams who keep analysis scripts, instrument integration code or battery test tooling on their own hardware and want the scan to run where the code lives.

Click through it at your own pace, or open it full screen.

The third demo connects Claude to the Safeguard MCP server with OAuth and lets you choose which tools Claude can use. For an engineer who already works with Claude while writing a data pipeline for accelerated ageing results, it brings dependency findings into the same conversation, limited to the tools you allow.

Click through it at your own pace, or open it full screen.

To try it on your own repositories, download Safeguard and follow the CLI documentation.

Quick check

Sources

• OpenStax, Chemistry 2e, 12.5 Collision Theory
• NIST CODATA, molar gas constant R
• NIST CODATA, Boltzmann constant
• NIST CODATA, Planck constant
• Nobel Foundation, Svante Arrhenius facts, Nobel Prize in Chemistry 1903
• Nobel Foundation, Jacobus H. van 't Hoff facts, Nobel Prize in Chemistry 1901
• Crossref record: Arrhenius, Über die Reaktionsgeschwindigkeit bei der Inversion von Rohrzucker durch Säuren, Z. Phys. Chem. 4 (1889) 226 to 248
• Crossref record: Eyring, The Activated Complex in Chemical Reactions, J. Chem. Phys. 3 (1935) 107 to 115
• Crossref record: Laidler, The development of the Arrhenius equation, J. Chem. Educ. 61 (1984) 494
• NIST Chemical Kinetics Database, Standard Reference Database 17
• NIST/SEMATECH e-Handbook of Statistical Methods, 8.1.5.1 Arrhenius
• ICH Q1A(R2), Stability Testing of New Drug Substances and Products (PDF)
• ASTM F1980, Standard Guide for Accelerated Aging of Sterile Barrier Systems and Medical Devices
• Keil et al., Calendar Aging of Lithium-Ion Batteries, J. Electrochem. Soc. 163 (2016) A1872 (open access)
• Ratkowsky et al., Relationship between temperature and growth rate of bacterial cultures, J. Bacteriol. 149 (1982) 1 to 5 (PMC)
• Ratkowsky et al., Model for bacterial culture growth rate throughout the entire biokinetic temperature range, J. Bacteriol. 154 (1983) 1222 to 1226 (PMC)
• NIST Chemical Kinetics Database, unit preferences and three-parameter Arrhenius display
• US Food and Drug Administration, Safe Food Handling

Last reviewed: 24 October 2026.