Why Does Iron Rust? 7 Steps to Understand the Electrochemical Corrosion Mechanism

1. Why Corrosion Is an Electrochemical Process, Not a Chemical One

Before we can build the mechanism, we need to be precise about one distinction. Students often arrive at this topic thinking of rusting as a chemical reaction between iron and oxygen — something like iron simply "burning slowly" in air. That picture is not wrong in its final accounting, but it is wrong in its route, and the route is what determines how fast corrosion happens and how we can stop it.

Let me define the two terms carefully.

Chemical corrosion (also called dry corrosion or direct chemical attack) occurs when a metal reacts directly with a gas or a non-conducting liquid, with no liquid electrolyte present. Electrons pass from the metal atom to the oxidising species at the same spot, across a distance of a few ångströms. The high-temperature oxidation of steel in a furnace, where iron reacts directly with O₂ to build a scale of FeO, Fe₃O₄ and Fe₂O₃, is the classic case.

Electrochemical corrosion (also called wet corrosion) occurs when the metal is in contact with a liquid electrolyte — and here is the essential difference — the oxidation and the reduction happen at physically separated locations on the surface, connected by two conducting paths running in parallel. This spatial separation is not a detail. It is the entire mechanism.

Once you accept the separation, several everyday observations that seem mysterious become obvious. Rust deposits form near a pit but not inside it. A steel pipe buried in soil corrodes fastest exactly where it passes from clay into sand. A scratch through galvanised coating does not rust, while a scratch through tin plating does. Every one of these follows directly from where the anodic and cathodic regions sit relative to one another.

2. The Four Components Every Corrosion Cell Must Have

A region of corroding metal is, quite literally, a short-circuited electrochemical cell — a battery that has been wired to itself and is discharging its own structure. For any such cell to operate, four things must be present simultaneously. Learn these four, and you have a diagnostic checklist you can apply to any corrosion problem you will ever meet.

Notice the symmetry of the last two. Charge flows in a complete loop: as electrons through the metal in one direction, and as ions through the electrolyte in the other. Break the loop anywhere and the corrosion current falls to zero. Almost every practical corrosion control method in industry is, at heart, an attack on one of these four components.

3. How a Uniform Iron Surface Develops Anodic and Cathodic Regions

This is the section that answers our opening question, so let us take it slowly and build it in stages. The puzzle is this: if a piece of iron is chemically identical from one end to the other, what decides which patch becomes the anode and which becomes the cathode?

The answer has two halves. First, no real metal surface is ever truly uniform — the uniformity is an idealisation we use in introductory diagrams and nowhere else. Second, even if a surface somehow started perfectly uniform, the environment above it is not uniform either, and the corrosion process itself immediately generates non-uniformity. Let us take these in turn.

3.1 Every Real Metal Surface Is Heterogeneous

Zoom in on a piece of ordinary mild steel at the micrometre scale and you find a landscape, not a plane. Here is what makes different points on that landscape electrochemically different from one another.

Grain boundaries. A polycrystalline metal is an assembly of crystallites — grains — each with its own lattice orientation. Where two grains meet, atoms sit in a disordered, strained arrangement. These boundary atoms have higher energy than atoms sitting comfortably in the ordered interior, so they are easier to remove. Grain boundaries therefore tend to be anodic relative to grain interiors. If you have worked through our tutorials on unit cells and lattice parameters, you already have the mental picture: the boundary is where the periodic order breaks down.

Second-phase particles and inclusions. Commercial steel is not pure iron. It contains cementite (Fe₃C), manganese sulphide inclusions (MnS), oxide particles, and other phases. Each phase has its own electrochemical character. Cementite, being an electronic conductor with a more noble potential than ferrite, acts as an efficient cathode, while the surrounding ferrite grains act as anodes. This is precisely why pearlitic steel corrodes faster than pure iron: it comes pre-supplied with millions of microscopic cathodes.

Cold work and residual stress. Bend a wire, punch a hole, or thread a bolt, and you introduce dislocations and residual stress into that region. The stored elastic and plastic strain energy raises the free energy of those atoms, making them thermodynamically easier to oxidise. Deformed regions become anodic to undeformed regions. This is why the bent portion of a nail and the threaded portion of a bolt corrode preferentially.

Surface films and their defects. Iron in air always carries a thin native oxide film. That film is never perfect — it has pores, scratches, and thin spots. Where the film is intact, the metal is shielded; where it is broken, bare metal is exposed. The bare patch becomes a small anode facing a very large film-covered cathode, and that area ratio is dangerous, as we will see in Section 11.

Temperature and surface roughness variations. A warmer region of a metal surface generally has a higher exchange current density and slightly shifted equilibrium potential relative to a cooler region, creating a thermogalvanic cell. Rough, high-surface-area regions expose more atoms per unit projected area.

3.2 The Environment Adds Its Own Heterogeneity: Differential Aeration

Suppose we could somehow produce a perfectly homogeneous iron surface. Would corrosion stop? It would not, because the electrolyte above it supplies its own asymmetry. The most important such effect — and the one that explains more real-world corrosion damage than any other single mechanism — is the differential aeration cell, first analysed systematically by U. R. Evans in the 1920s.

Consider the classic demonstration: a single drop of salt water resting on a clean iron plate. Oxygen from the air dissolves into the drop and diffuses to the metal. But diffusion distance is not the same everywhere. At the thin edge of the drop, oxygen has only a short path; at the thick centre, the path is much longer. The result is a steep oxygen gradient across a drop only a few millimetres wide.

The oxygen-rich rim supports rapid oxygen reduction, so it becomes cathodic. The oxygen-starved centre cannot sustain that reduction, so iron dissolution takes over there and it becomes anodic. Within an hour you can see the outcome: a ring of brown rust deposit, with a clean, actively pitting patch of bare metal at the centre.

Let us quantify the driving force, because a number makes the argument concrete. Using the Nernst equation for the oxygen reduction half-reaction at pH 7 and 25 °C, with the rim exposed to atmospheric oxygen (p_O2 = 0.21 atm) and the centre depleted to roughly 10⁻⁴ atm:

3.3 Concentration Cells: The Process Creates Its Own Gradients

There is a third source of asymmetry, and it is self-generating. Once iron begins dissolving at some point, Fe²⁺ ions accumulate locally. The Nernst equation tells us that a region with a higher concentration of its own metal ions sits at a more positive (more noble) potential. Working the numbers for iron at 25 °C:

Note carefully what this means. The ion-depleted region becomes the anode, so wherever fresh electrolyte flushes ions away, dissolution is encouraged. Combined with differential aeration, this produces the autocatalytic behaviour of pitting that we will study in a later lesson.

3.4 Putting It Together: The Sequence in Seven Steps

We now have everything we need to write the mechanism as an ordered sequence. Read these as a narrative of what happens in the first fraction of a second after a clean iron surface meets water, and then continuously thereafter.

4. The Anodic Reaction: Metal Dissolution

We now examine each half of the mechanism in detail, starting with the half where damage occurs. For iron in near-neutral or acidic aqueous environments, the anodic reaction is:

Read that equation as a physical event, not just a symbol string. An iron atom sitting in the body-centred cubic lattice at the surface is held by metallic bonding to its neighbours. A water molecule approaches. The atom breaks free from the lattice, leaves two of its valence electrons behind in the metal's conduction band, and moves into solution as a hydrated Fe²⁺ ion surrounded by a shell of water molecules. A tiny crater is left behind.

4.1 Why Iron Dissolves at All: The Thermodynamic Driving Force

A reaction proceeds spontaneously when it lowers the Gibbs free energy of the system. For an electrochemical cell, the connection between cell potential and free energy is one of the most useful relations in the whole subject:

For iron in aerated neutral water, pairing the anodic reaction with oxygen reduction:

This is worth stating plainly to students, because it reframes the entire engineering problem: we can almost never make a structural metal thermodynamically stable in a wet environment. What we can do is make the reaction slow. Every coating, inhibitor and cathodic protection system is a kinetic intervention, not a thermodynamic one.

4.2 What Happens to the Fe²⁺ Ion Afterwards

Once released, the ferrous ion has several possible fates depending on pH and oxygen availability. It may remain dissolved as hydrated Fe²⁺ in acidic or oxygen-poor conditions; it may react with hydroxide to precipitate Fe(OH)₂; or it may be oxidised further to Fe³⁺. We will follow this chain carefully in Section 9, because a great deal of student confusion about "where rust comes from" is resolved there.

One point deserves emphasis now. In some environments iron can oxidise directly to the ferric state, and under alkaline or strongly oxidising conditions it may form a protective passive film. That passivating behaviour is what makes stainless steel possible, and it is the subject of its own lesson in this series.

5. The Cathodic Reactions: Where the Electrons Go

The anodic reaction cannot run in isolation. Electrons released by dissolving iron must be consumed at the same rate, or the metal would accumulate negative charge and the reaction would stop almost immediately. The cathodic reaction is therefore not a bystander — in most practical situations it is the step that actually controls the corrosion rate.

Which reduction reaction takes place depends on what is available in the electrolyte. Here are the four you need to know, in order of practical importance.

5.1 Oxygen Reduction in Neutral or Alkaline Solution

Two features of this reaction are worth dwelling on. First, it produces hydroxide ions, so the electrolyte immediately adjacent to the cathode becomes locally alkaline. Second, it consumes dissolved oxygen, which must be continuously resupplied by diffusion through the water film. Because oxygen is only sparingly soluble in water, this resupply is slow, and in quiet water it becomes the bottleneck for the entire corrosion process. We call this diffusion-limited or concentration-polarised behaviour.

5.2 Hydrogen Evolution in Acid Solution

When you drop a piece of iron into dilute hydrochloric acid and see it fizz, you are watching this cathodic reaction on the cathodic patches while iron dissolves on the anodic patches. The driving force is E°_cell = 0.000 − (−0.440) = 0.440 V.

Hydrogen evolution carries a second hazard beyond metal loss. Some of the atomic hydrogen produced at the surface, instead of combining into H₂ gas, diffuses into the metal lattice, where it can cause loss of ductility and delayed cracking. This phenomenon — hydrogen embrittlement — is the subject of Lesson 3 in this series.

5.3 Oxygen Reduction in Acid Solution

5.4 Reduction of Dissolved Metal Ions

5.5 Reference Table of Standard Potentials

The following table collects the standard electrode potentials used throughout this lesson, all referred to the Standard Hydrogen Electrode at 25 °C. Values are drawn from standard electrochemical data compilations. For engineering work in a specific electrolyte, a measured galvanic series (ASTM G82) is a better guide than this thermodynamic table — Section 11.4 explains why.

6. Electron Movement Through the Metal

We have two reaction sites. Now we need to understand the two connections between them, and they work by completely different physics. Let us take the metallic path first.

6.1 Why Metals Conduct Electrons So Well

In a metallic solid, the outer valence electrons are not bound to individual atoms. They are delocalised across the entire crystal, occupying a partially filled conduction band. The conventional picture — a rigid array of positive ion cores immersed in a mobile "sea" of electrons — is a simplification of band theory, but it captures the essential point: charge carriers in a metal are free to move throughout the whole solid.

Iron has a conductivity of roughly 1×10⁷ S m⁻¹. To appreciate what that means for corrosion, compare it with the electrolyte:

Iron is about two million times more conductive than seawater, and about two billion times more conductive than distilled water. This enormous asymmetry has a direct and important consequence for the mechanism.

6.2 What Actually Moves, and How Fast

Students are sometimes surprised to learn how slowly individual electrons move. For a typical corrosion current density of 1 µA cm⁻², the electron drift velocity in the metal is of the order of 10⁻¹² m s⁻¹ — slower than a growing fingernail by an enormous factor. Yet the corrosion current is established essentially instantaneously.

The resolution is the same as for any electrical circuit: the signal propagates at nearly the speed of light through the electromagnetic field, while individual carriers merely drift. There are so many free electrons in a metal — of order 10²⁸ m⁻³ — that an imperceptible average drift of the whole population carries the required current.

6.3 The Direction of Electron Flow (and a Warning About Sign Conventions)

Electrons flow from anode to cathode through the metal. Conventional current, being defined as the flow of positive charge, therefore flows in the opposite sense inside the metal: from cathode to anode. In the electrolyte, conventional current flows from anode to cathode, carried by positive ions moving that way and negative ions moving the other way.

6.4 What Happens If You Cut the Metallic Path

Break the electronic connection and the cell stops, no matter how favourable the thermodynamics. This is exactly the principle behind an insulating flange or dielectric union, used routinely where a steel pipe meets a copper fitting: an insulating gasket interrupts the electron path, and the galvanic cell that would otherwise consume the steel simply cannot operate.

7. Ion Movement Through the Electrolyte

The return half of the circuit works on entirely different physics, and this is where a great deal of practical corrosion behaviour is decided.

7.1 Why Ions Must Move at All

Think about what would happen if ions could not move. At the anode, Fe²⁺ ions accumulate, building positive charge in the adjacent solution. At the cathode, OH⁻ ions accumulate, building negative charge. Within microseconds these charge build-ups create an electric field opposing further reaction, and the corrosion current collapses to zero.

Ion migration is therefore not an afterthought — it is a hard requirement for the reaction to continue. Charge neutrality must be maintained in the electrolyte, and ions are the only means available.

7.2 The Three Transport Mechanisms

Ions in solution move by three distinct mechanisms, all of which operate simultaneously in a corrosion cell.

Migration is motion driven by the electric field between anodic and cathodic regions. Cations (Fe²⁺, Na⁺, H⁺) drift toward the cathode; anions (OH⁻, Cl⁻, SO₄²⁻) drift toward the anode. This is the mechanism that directly carries the ionic current.

Diffusion is motion driven by concentration gradients, described by Fick's laws. Oxygen diffusing from the air–water interface to the cathode, and Fe²⁺ diffusing away from the anode, are both diffusion processes. In quiescent solutions diffusion is usually the slowest and therefore rate-controlling step.

Convection is bulk fluid motion — stirring, pumping, natural buoyancy currents, wave action. Convection short-circuits diffusion by physically carrying fresh electrolyte to the surface, which is why flowing seawater is so much more corrosive than still seawater.

7.3 Why Electrolyte Conductivity Controls Corrosion Range

The conductivity of the electrolyte does two things. It sets the ohmic resistance of the return path, and through that it sets how far apart the anode and cathode can usefully be.

In highly conductive electrolytes such as seawater or salted road water, the ohmic drop is small, so an anode can drive a cathode several metres away. This produces macrocell corrosion — large separated anodic and cathodic areas — and often severe, localised attack. It is also why a small anode area can be so dangerous: the entire current from a huge cathode concentrates into a tiny spot.

In poorly conductive electrolytes such as clean rainwater, the ohmic drop is large and anode–cathode pairs must sit close together. The result is many small, closely spaced micro-cells, which appear macroscopically as slow, fairly uniform general corrosion.

7.4 What Happens If You Cut the Ionic Path

Just as with the electronic path, breaking the ionic path stops the cell. Keeping steel genuinely dry is the simplest and most effective corrosion control there is. Paints, greases and polymer coatings work primarily by placing a high-resistance barrier in the ionic path, not by chemically blocking oxygen — which is why a coating with a single pinhole can fail so dramatically. The pinhole becomes a tiny anode connected to a large cathode, and the concentrated current drills straight through the steel.

8. Closing the Loop: Mixed Potential, E_corr and i_corr

We have assembled all four components. Now let us see how they settle into a steady state, because this is the framework that lets us turn the qualitative mechanism into numbers.

8.1 The Charge Balance Requirement

An isolated, freely corroding piece of metal is electrically neutral overall and is not connected to any external circuit. Charge cannot accumulate anywhere. Therefore, over the whole surface:

Notice that the balance is between total currents, not current densities. The consequence is the area-ratio effect, and it explains a great deal of practical damage. Rearranged:

8.2 The Corrosion Potential and the Evans Diagram

These ideas were formalised as mixed potential theory by Wagner and Traud in 1938, and are conventionally displayed on a plot of potential against the logarithm of current density — the Evans diagram, named after U. R. Evans. Reading one is a skill worth acquiring now.

Read Figure 5 as follows. Each half-reaction starts from its own equilibrium potential at its own exchange current density, and departs from it as net current is drawn — the anodic branch rising, the cathodic branch falling. There is only one potential at which the two currents are equal, and that is where the branches cross. The metal has no choice but to sit there.

9. Where Rust Actually Forms — and Why It Is a Secondary Product

We arrive at a point where careful teaching matters, because the standard school-level account is misleading. Rust is not the anodic product, and it does not form at the anode. It forms in the solution, in the region between anode and cathode, by a sequence of ordinary chemical reactions that occur after the electrochemistry is finished.

Here is the chain, in order.

9.1 Why Rust Does Not Protect the Metal Underneath

Some metals form oxide films that shield the surface. Aluminium and chromium are the standard examples, and their protection is the reason stainless steel exists. Iron does not, and the reason is largely geometric.

The Pilling–Bedworth ratio (PBR) compares the volume of oxide produced with the volume of metal consumed. A ratio slightly above 1 gives a compact, adherent, protective film. A ratio well below 1 gives a porous, cracked film. A ratio well above 1 gives a film in compression that buckles and spalls off.

The situation for hydrated rust is worse still. Taking Fe(OH)₃ at a typical density of about 3.4 g cm⁻³, one mole of iron occupies 7.09 cm³ as metal but about 31 cm³ as hydrated rust — roughly a fourfold volume expansion. The product cannot possibly sit compactly on the surface it came from. It flakes, exposing fresh metal, and the process repeats indefinitely. This is why steel corrodes until it is gone, while aluminium stops after a few nanometres.

10. Worked Numerical Examples

The mechanism becomes far more solid once you can put numbers to it. All three examples below use Faraday's laws of electrolysis, which connect charge passed to mass transformed — the bridge between the electrochemistry and the engineering.

Example 1 — From current density to penetration rate

A steel plate corrodes freely in aerated water at a corrosion current density of 1 µA cm⁻². Find the annual mass loss per square centimetre and the penetration depth per year. Density of iron = 7.874 g cm⁻³.

Pause on how modest that sounds — and then notice the trap. If this current is uniform over the whole surface, a 6 mm plate survives more than five centuries. But if the same total current is concentrated into a pit occupying one thousandth of the area, the local penetration rate becomes 11.6 mm yr⁻¹ and the plate is perforated in under seven months. Localised corrosion is dangerous not because the current is large, but because the area is small. This is the area-ratio effect of Section 8.1 expressed in the units an engineer cares about.

Example 2 — From measured current to mass loss

A galvanic couple between a steel bolt and a copper plate in seawater is measured to pass a steady current of 5 mA. How much iron is lost in 24 hours?

Example 3 — Predicting the driving force from tabulated potentials

Compare the thermodynamic driving force for iron corroding in (a) deaerated acid at pH 0, and (b) aerated acid at pH 0. Comment on the practical significance.

11. Engineering Applications of the Mechanism

Everything we have built now pays off, because each control method is simply an attack on one of the four components from Section 2. Organise your thinking this way and you will never have to memorise a list.

11.1 Remove the Electrolyte

Dehumidification of storage warehouses, vapour-phase inhibitors in packaging, drainage design that prevents water pooling in structural sections, and simply avoiding crevices where moisture can sit. The cheapest corrosion control there is, when the application permits it.

11.2 Interrupt the Ionic Path — Coatings and Linings

Paints, epoxies, polymer linings and greases work chiefly by inserting a very high resistance into the ionic return path. Their weakness follows directly from Section 8.1: a single pinhole creates a minute anode facing an enormous coated cathode, and the concentrated current perforates the metal rapidly. Coating integrity matters more than coating thickness.

11.3 Interrupt the Electronic Path — Insulating Joints

Insulating flanges, dielectric unions and non-conductive gaskets are placed wherever dissimilar metals must be joined. The thermodynamic driving force remains, but with the electron path broken, no current can flow.

11.4 Suppress the Cathodic Reaction

Deaeration of boiler feedwater; oxygen scavengers such as sodium sulphite and hydrazine; and cathodic inhibitors that deposit a film specifically on cathodic sites. On the Evans diagram of Figure 5, all of these push the cathodic branch to lower current and slide the intersection to smaller i_corr.

11.5 Suppress the Anodic Reaction

Anodic inhibitors such as chromates, nitrites and phosphates promote or repair a passive film on anodic sites. A warning that GATE examiners are fond of: an underdosed anodic inhibitor is worse than none at all, because it blocks most of the anodic area while leaving a few small sites active, converting broad general corrosion into deep, dangerous pitting.

11.6 Make the Whole Structure a Cathode — Cathodic Protection

This is the most elegant application of the mechanism, because it does not fight corrosion — it relocates it.

Sacrificial anode (galvanic) protection. Attach a metal more active than iron — zinc (−0.763 V), aluminium (−1.662 V) or magnesium (−2.372 V). The attached metal becomes the anode of the new, larger cell and dissolves, while the steel is polarised into the cathodic region and stops dissolving. Galvanised steel works exactly this way, which is why a scratch through zinc coating does not rust: the surrounding zinc protects the exposed steel. Contrast this with tin plating on steel, where tin is more noble than iron; a scratch there produces a small steel anode facing a large tin cathode, and the steel corrodes rapidly.

Impressed current cathodic protection. An external DC power supply forces electrons into the structure, holding it below E_eq(Fe). Used on pipelines, ship hulls, jetties and large storage tanks where the current demand is too high for sacrificial anodes.

11.7 Change the Material or the Design

Alloying to promote passivity — adding at least about 11% chromium to steel to form a self-healing Cr₂O₃ film — is a materials solution. Equally important is design: avoiding crevices, eliminating dead legs where stagnant water sits, providing drainage, avoiding contact between widely separated metals, and never pairing a small anode with a large cathode.

12. Six Misconceptions Worth Correcting

In my experience teaching this material, the same six errors appear year after year. Reading them explicitly, side by side with the correct statement, tends to fix them permanently.

13. Summary and Key Takeaways

Let us close the loop on the question we opened with. Why does a nail rust in patches rather than uniformly? Because a nail is not a uniform object in a uniform environment. Its grain boundaries, its cold-worked bend, its inclusions and its broken oxide film all sit at slightly different electrochemical potentials; the water film above it delivers oxygen unevenly; and because the whole nail is a superb electronic conductor, these differences cannot coexist as separate potentials. The surface is forced to a single compromise potential, and in reaching it, it partitions itself into anodes that dissolve and cathodes that survive.

The full mechanism, compressed into one sentence: local heterogeneity establishes anodic and cathodic regions; iron dissolves at the anode releasing electrons; those electrons cross the metal to the cathode where oxygen or hydrogen ions consume them; ions migrate through the electrolyte to close the circuit; and rust precipitates in the solution in between.

14. Practice Questions (GATE Pattern)

The following eight questions are original items written in the GATE XE-C style to test the concepts of this lesson. Correct answers are marked, with reasoning given below each.

15. Frequently Asked Questions

16. References

1. M. G. Fontana, Corrosion Engineering, 3rd ed. New York, NY, USA: McGraw-Hill, 1986.
2. D. A. Jones, Principles and Prevention of Corrosion, 2nd ed. Upper Saddle River, NJ, USA: Prentice Hall, 1996.
3. U. R. Evans, The Corrosion and Oxidation of Metals: Scientific Principles and Practical Applications. London, U.K.: Edward Arnold, 1960.
4. C. Wagner and W. Traud, "On the interpretation of corrosion processes through the superposition of electrochemical partial processes and on the potential of mixed electrodes," Zeitschrift für Elektrochemie, vol. 44, no. 7, pp. 391–402, 1938. doi:10.1002/bbpc.19380440702. (English translation: Corrosion, vol. 62, no. 10, pp. 843–855, 2006. doi:10.5006/1.3279894.)
5. M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, 2nd English ed. Houston, TX, USA: NACE International, 1974.
6. A. J. Bard and L. R. Faulkner, Electrochemical Methods: Fundamentals and Applications, 2nd ed. New York, NY, USA: Wiley, 2001.
7. N. B. Pilling and R. E. Bedworth, "The oxidation of metals at high temperatures," Journal of the Institute of Metals, vol. 29, pp. 529–582, 1923.
8. P. R. Roberge, Handbook of Corrosion Engineering, 2nd ed. New York, NY, USA: McGraw-Hill, 2012.
9. ASTM G102-89(2015)e1, Standard Practice for Calculation of Corrosion Rates and Related Information from Electrochemical Measurements. West Conshohocken, PA, USA: ASTM International. store.astm.org
10. ASTM G82-98(2021)e1, Standard Guide for Development and Use of a Galvanic Series for Predicting Galvanic Corrosion Performance. West Conshohocken, PA, USA: ASTM International. store.astm.org
11. C. Kittel, Introduction to Solid State Physics, 8th ed. Hoboken, NJ, USA: Wiley, 2005.
12. P. Vanysek, "Electrochemical series," in CRC Handbook of Chemistry and Physics, 97th ed., W. M. Haynes, Ed. Boca Raton, FL, USA: CRC Press, 2016.