The Wittig Reaction
A phosphorus ylide walks up to a ketone and, in one smooth choreography of bond-making and bond-breaking, swaps the carbonyl oxygen for a new carbon–carbon double bond. That single transformation — a C=O turned cleanly into a C=C — solved a problem that had frustrated chemists for a century: how do you build an alkene at a specific position, with a specific geometry, without the tars and rearrangements that plague acid-catalyzed dehydrations and eliminations? Georg Wittig's answer, developed in the early 1950s, earned him a share of the 1979 Nobel Prize and handed synthetic chemistry one of its most reliable and most-used bond constructions. The reaction matters because carbonyls are everywhere and alkenes are valuable real estate: they are the entry points to epoxides, diols, alkanes, polymers, and — critically — to further carbonyls once ozonolysis or dihydroxylation cuts them back open. The Wittig reaction lets a chemist choose exactly where a double bond will sit, because the new C=C forms precisely at the old C=O, no ambiguity, no competing regiochemistry the way E1/E2 eliminations suffer from. This "traceless" retrosynthetic logic — disconnect any alkene straight back to a carbonyl plus an ylide — is why the Wittig reaction appears in essentially every total synthesis textbook and in industrial-scale vitamin A and vitamin D manufacture. The mechanistic story is a beautiful piece of orbital reasoning built directly on Nucleophilic Addition to the Carbonyl: The Core Mechanism: a stabilized carbanion (the ylide) adds to the electrophilic carbonyl carbon, but instead of stopping at an alkoxide, the phosphorus and oxygen — sitting right next to each other — snap together into a four-membered ring, the oxaphosphetane. That ring then falls apart in a retro-[2+2]-like step, expelling triphenylphosphine oxide and releasing the alkene. Every stereochemical outcome of the reaction — the famous E/Z selectivity switch between "reactive" (non-stabilized) and "stabilized" ylides — is a direct consequence of which diastereomer of oxaphosphetane forms first and how reversibly it forms. This article works through ylide formation and its electronic structure (building on ylide stabilization), the full betaine/oxaphosphetane mechanism with curved arrows at every step, the orbital and steric arguments that control E versus Z alkene geometry, and three fully worked examples spanning a classroom exercise, an applied synthesis (a fragrance-industry alkene), and an exam-level stereochemical prediction problem.
- The Wittig reaction converts a ketone or aldehyde (C=O) plus a phosphorus ylide (Ph₃P=CR₂) into an alkene (C=C) plus triphenylphosphine oxide (Ph₃P=O).
- Ylides are generated by deprotonating a phosphonium salt with a strong base; the carbanion is stabilized by negative hyperconjugation into σ*(P–C) and by electrostatic P⁺–C⁻ attraction, not by classical P=C π-bonding.
- The mechanism proceeds through a betaine or, more directly, a four-membered oxaphosphetane, formed by [2+2]-like addition of the ylide carbon and phosphorus across the carbonyl.
- The oxaphosphetane collapses by a retro-[2+2] cycloreversion, driven by the enormous strength of the P=O bond, expelling Ph₃P=O and generating the alkene.
- Non-stabilized ylides give predominantly cis (Z) alkenes via kinetically controlled, irreversible formation of the cis-oxaphosphetane; stabilized ylides give predominantly trans (E) alkenes because oxaphosphetane formation is reversible and equilibrates to the more stable trans isomer.
- The Schlosser modification and the Horner–Wadsworth–Emmons variant extend stereochemical control to give E-selective alkenes even from otherwise Z-selective ylide types.
- Key idea: the Wittig reaction is a [2+2]/retro-[2+2] cycloaddition sequence in which the driving force is the formation of an extremely strong P=O bond, and the E/Z outcome is decided entirely by whether oxaphosphetane formation is reversible.
An ylide adds to a carbonyl to form a four-membered oxaphosphetane ring; this ring collapses via retro-[2+2] cycloreversion, trading a C=O and a P–C(ylide) bond for a new C=C and an exceptionally strong P=O bond — and it is the reversibility of the ring-forming step (governed by ylide stabilization) that dictates whether the alkene comes out cis or trans.
Where this lives today
Industrial vitamin A and vitamin D₃ manufacturing routes use Wittig-type olefinations at multi-tonne scale to install specific polyene double bonds with defined geometry — the retinal/retinol side chain, with its precise all-trans polyene pattern essential for visual-cycle biological activity, is built up using stereocontrolled phosphonium or phosphonate olefinations at key disconnections. BASF's industrial vitamin A synthesis is one of the classic examples of Wittig chemistry running at true bulk-chemical scale rather than just in academic milligram quantities. In pheromone and fragrance chemistry, Z-selective Wittig reactions with non-stabilized ylides remain the standard method for making the long-chain Z-alkenes found in insect sex pheromones (used in mating-disruption pest control) and in numerous fragrance-industry unsaturated alcohols and esters, precisely because ordinary elimination or partial-hydrogenation methods cannot deliver comparable geometric purity. In pharmaceutical process chemistry, Wittig and Horner–Wadsworth–Emmons olefinations are routinely used to install α,β-unsaturated ester or ketone side chains (Michael acceptors, statin side chains, and prostaglandin analog syntheses) — the E-selectivity available from stabilized ylides is exactly what is needed for these conjugated systems, since biological receptors are typically geometry-selective for the E-configured enone or enoate. In total synthesis more broadly, the Wittig disconnection (alkene ⇒ carbonyl + ylide) is one of the first disconnections taught in retrosynthetic analysis courses, precisely because it is so mechanistically transparent and stereochemically predictable — a direct payoff of understanding the oxaphosphetane mechanism rather than merely memorizing "Wittig makes alkenes."
:::quiz[Which statement about the oxaphosphetane retro-[2+2] step is correct?]
- It scrambles stereochemistry via a long-lived betaine intermediate in all cases
- It proceeds with retention of the ring's relative stereochemistry, translating oxaphosphetane diastereomer directly into alkene geometry ✓
- It requires a separate acid catalyst to proceed
- It re-forms the starting carbonyl compound as a side product
Ask the tutor: How does the Schlosser modification force an E-alkene even when starting from a normally Z-selective non-stabilized ylide?
The physical picture
Start with the ylide itself. A phosphonium ylide is written R₃P=CR′₂, but that double-bond notation is a convenient fiction. Phosphorus is in the third row, and while early theories invoked P=C π-bonding using phosphorus 3d orbitals, modern computational and structural studies show the true picture is closer to a phosphonium–carbanion ion pair with substantial negative hyperconjugation: the carbanion lone pair on Cα donates into the antibonding σ orbitals of the three P–C(phenyl) bonds. This is exactly the electronic stabilization logic developed for <span class="wiki-link-missing">ylide stabilization</span> in general — a formally charged carbanion adjacent to an electropositive, polarizable heteroatom is stabilized by delocalization of its lone pair away from carbon and into empty or antibonding orbitals on the neighbor, without requiring a "real" π-bond. The carbanion carbon is nonetheless highly nucleophilic — pyramidalized enough to retain substantial p-character and a real lone pair, but flattened somewhat by the hyperconjugative donation. This lone pair is what attacks the carbonyl carbon of the aldehyde or ketone: a standard nucleophile-into-π(C=O) addition, precisely the step you already know from carbonyl addition chemistry, except that here the incoming carbon carries a positively charged phosphonium group riding along with it. Once the new C–C bond forms and the oxygen becomes a negatively charged alkoxide, that oxygen is held in close proximity to the positively charged phosphorus — geometrically forced into a four-membered ring. Phosphorus, being both large, polarizable, and oxophilic (P=O bonds are around 130–140 kcal/mol, among the strongest single "π-type" bonds known for a main-group element), reacts with that alkoxide oxygen to close the ring: this is the oxaphosphetane. The ring then opens the other way — a retro-[2+2] — breaking the C–O and P–C bonds and forming the new C=C π bond plus the P=O π bond. Because P=O is so much stronger than the C=O bond it replaces (P=O bond energy substantially exceeds a typical carbonyl C=O in the transition-state bookkeeping, once you also count the new strong Si-like affinity of phosphorus for oxygen), the whole sequence is strongly exergonic and essentially irreversible once past the oxaphosphetane.
:::svg[The oxaphosphetane: formation and retro-[2+2] collapse]
The core ideas
The overall stoichiometry is simple:
where: R, R′ are the substituents on the carbonyl carbon and the ylide carbon respectively; the phosphorus byproduct is triphenylphosphine oxide, removed as an easily separated, highly polar solid in workup. The thermodynamic driving force can be expressed as a bond-energy bookkeeping relationship:
where: is the carbonyl bond broken (≈ 175–180 kcal/mol for a typical ketone C=O, counting both σ and π), is the ylide P–C bond broken, is the new alkene π-bond formed (≈ 65 kcal/mol π-component on top of the retained σ), and is the newly formed phosphoryl bond (≈ 130–140 kcal/mol). Because so greatly exceeds any reasonable value for , the reaction is strongly exothermic regardless of substituents — this is why the Wittig reaction works reliably even with fairly unreactive, hindered ketones. The ylide's basicity/nucleophilicity, which sets the reaction's rate and the reversibility of the first C–C bond-forming step, correlates with the stabilization energy of the anion — the same stabilization logic from ylide stabilization:
where: a non-stabilized ylide (R′ = alkyl, e.g. CH₃, giving Ph₃P=CH₂) has a highly basic, reactive, essentially localized carbanion (p of the conjugate phosphonium salt ~22–25 in DMSO); a stabilized ylide (R′ = ester, ketone, or other π-acceptor, e.g. Ph₃P=CHCO₂Et) delocalizes the negative charge into an adjacent carbonyl, lowering basicity/nucleophilicity substantially (p ~ 6–9) and making every step of the mechanism—including oxaphosphetane formation—reversible. The E/Z outcome is captured qualitatively by a kinetic-vs-thermodynamic control statement:
where: the cis-oxaphosphetane, with its incipient eclipsing 1,2-substituent interactions, is the kinetically favored transition state for non-stabilized ylides — but because it collapses irreversibly before it can equilibrate, the kinetic product (Z-alkene) dominates. For stabilized ylides, betaine/oxaphosphetane formation is reversible enough that the thermodynamically more stable trans-oxaphosphetane (fewer 1,2-eclipsing interactions) becomes the major pathway, funneling to the E-alkene.
Goal: show, with curved arrows, how an alkyl halide plus triphenylphosphine become a stable phosphonium salt, and then how base deprotonation generates the reactive ylide.
Step 1 — SN2 alkylation of phosphorus. Triphenylphosphine's lone pair on phosphorus is a good, soft, highly polarizable nucleophile (isolobal comparison to a tertiary amine, but P is bigger and more nucleophilic in the soft sense). It attacks the electrophilic carbon of a primary alkyl halide, R–CH₂–X, in a backside SN2 displacement. The curved arrow runs from the phosphorus lone pair into the C–X σ* orbital, while the C–X bonding electrons flow fully onto the leaving halide. This inverts configuration at carbon (irrelevant here if R–CH₂–X is primary/unsubstituted at that carbon) and generates a phosphonium halide salt, Ph₃P⁺–CH₂R X⁻.
Step 2 — Deprotonation adjacent to P⁺. The C–H bonds on the carbon directly attached to the positively charged phosphorus are unusually acidic (p roughly 22–25 for a simple alkyl case in DMSO, dramatically lower than a normal alkane C–H, because the resulting carbanion is stabilized by the adjacent cationic, polarizable phosphorus via negative hyperconjugation into σ*(P–Cipso) of the phenyl rings, plus simple electrostatic ion-pairing). A strong, non-nucleophilic or moderately hindered base — classically n-BuLi, but also NaH, KOtBu, or NaNH₂ depending on the acidity needed — removes one of these protons. The curved arrow starts at the base's lone pair, goes to the C–H bond; the C–H bonding electrons flow onto that carbon, generating the carbanion (ylide) center and depositing a proton onto the base (e.g., forming butane from n-BuLi).
Step 3 — Resulting ylide. The product, Ph₃P⁺–CH(R)⁻ (drawn conventionally as Ph₃P=CHR), is isolable in many cases (especially stabilized ylides, which can even be purified by recrystallization) or generated and used in situ for reactive, non-stabilized ylides such as Ph₃P=CH₂ (methylenetriphenylphosphorane), which is unstable to air and moisture and is typically used immediately after generation.
Check — asymmetric worked numbers. Take ethyltriphenylphosphonium bromide, Ph₃P⁺–CH₂CH₃ Br⁻ (from PPh₃ + bromoethane). Deprotonation with n-BuLi (1.05 equiv, THF, –78 °C then warmed) removes a proton from the CH₂ group (not the terminal CH₃, since that carbon is not adjacent to P⁺ and its C–H bonds are ordinary alkyl C–H, p ~ 50), giving the ethylidene ylide Ph₃P=CHCH₃. This is a non-stabilized, alkyl-substituted ylide — it will show Z-selectivity in the alkene-forming step below.
:::derivation[Mechanism 2 — Full Wittig olefination: betaine/oxaphosphetane and retro-[2+2] collapse] Goal: trace every arrow from ylide + carbonyl to alkene + Ph₃P=O, and show explicitly why the cis- vs trans-oxaphosphetane choice controls E/Z alkene geometry.
Step 1 — Nucleophilic addition of the ylide carbon to the carbonyl carbon. This step is mechanistically identical to any nucleophile-into-carbonyl addition. The ylide carbanion lone pair attacks the carbonyl carbon along the Bürgi–Dunitz trajectory (~107° from the C=O axis), donating into the empty π*(C=O) orbital. The curved arrow: lone pair on Cα(ylide) → new σ bond to carbonyl carbon; simultaneously, the π electrons of C=O shift fully onto oxygen, generating an alkoxide. The immediate product is a zwitterion called the betaine: a phosphonium cation and an alkoxide anion joined by the new C–C bond, but not yet cyclized. (For very hindered or specific substrates the betaine can be a discrete, observable intermediate; for most simple cases the oxaphosphetane forms so fast that the betaine is better regarded as a transition-state feature rather than a true minimum.)
Step 2 — Ring closure to the oxaphosphetane. The alkoxide oxygen's lone pair attacks the adjacent, electrophilic, positively charged phosphorus center intramolecularly. Curved arrow: O lone pair → new σ bond to P. This closes a four-membered ring — the 1,2-oxaphosphetane — with phosphorus now pentacoordinate (a trigonal-bipyramidal-like or simply tetrahedral-plus picture depending on the exact bonding description) bearing three phenyl groups, the ring oxygen, and the ring carbon. This ring-closure step is the stereochemistry-determining step: it can form as either the cis (both new substituents R and R′ on the same face of the four-membered ring, syn) or trans diastereomer, and — crucially — for non-stabilized ylides this ring closure is effectively irreversible (or nearly so) and strongly kinetically preferred to give the cis oxaphosphetane, via a puckered, roughly planar transition state that minimizes torsional strain along the forming ring, even though the fully-formed cis ring itself is less stable (1,2-eclipsing substituents) than the trans ring.
Step 3 — Retro-[2+2] cycloreversion. The oxaphosphetane collapses in a concerted, four-centered retro-cycloaddition. Three curved arrows in a cyclic array: the C–P σ bond breaks, its electrons flowing to form the new C=C π bond between the two ring carbons; the C–O σ bond breaks, its electrons remaining on oxygen as it becomes doubly bonded to phosphorus; and the P–O σ bond's electron pair — actually, more precisely, the ring oxygen's lone pair that had formed the P–O bond is retained on oxygen while a new π bond forms between P and O using the breaking P–C(ring) bonding pair reorganized into P–O π overlap. The net electron count is a concerted four-electron (retro-[2+2]) shift: two σ bonds (C–P and C–O) break, and two π bonds (C=C and P=O) form. This step proceeds with retention of the relative stereochemistry set in the oxaphosphetane — a cis-oxaphosphetane collapses stereospecifically to the Z-alkene, and a trans-oxaphosphetane collapses stereospecifically to the E-alkene. No bond rotation is possible once the rigid four-membered ring has formed, so whichever diastereomer of oxaphosphetane you make, that is the alkene geometry you get.
Step 4 — Why non-stabilized ylides are Z-selective and stabilized ylides are E-selective. For a non-stabilized ylide (alkyl substituent on Cα), the initial nucleophilic addition and the subsequent ring closure are fast and essentially irreversible: the transition state leading to the cis-oxaphosphetane is lower in energy than the one leading to trans, primarily because the cis transition state can adopt a more nearly planar, "puckered-envelope" four-center geometry with the bulky substituents oriented pseudo-equatorially relative to the developing ring, whereas the trans-forming transition state suffers greater torsional strain along the C–C bond as it forms. Because collapse to alkene (Step 3) is fast and irreversible relative to any re-opening of the oxaphosphetane, the kinetic cis product is trapped and delivered directly to the Z-alkene — this is kinetic control. For a stabilized ylide (Cα bears an ester, ketone, nitrile, or other conjugating group), the initial C–C bond formation is reversible: the betaine/oxaphosphetane can revert back to separate ylide plus carbonyl compound (retro-addition), because the stabilized carbanion is a much better leaving group in reverse and the forward addition itself is less exothermic. Given enough time to equilibrate, the system funnels through the lower-energy trans-oxaphosphetane (less steric strain in the ring, substituents anti rather than eclipsing), which then collapses irreversibly to the E-alkene — this is thermodynamic control exercised at the level of a pre-equilibrium between oxaphosphetane diastereomers, not at the level of the alkene product itself (alkene formation, once it happens, is not reversible under standard Wittig conditions).
Check — asymmetric worked example. Take the ethylidene ylide Ph₃P=CHCH₃ (non-stabilized, from Mechanism 1) reacting with benzaldehyde, PhCHO. Nucleophilic addition places Cα (bearing CH₃) next to the carbonyl carbon (bearing Ph and H); ring closure preferentially forms the cis-oxaphosphetane, in which the CH₃ and Ph groups sit on the same face; retro-[2+2] collapse gives predominantly (Z)-β-methylstyrene, i.e., (Z)-1-phenylpropene, with the methyl and phenyl groups cis across the new double bond — typically obtained in roughly 80:20 to 95:5 Z:E ratio depending on exact conditions (salt-free vs. salt-containing, solvent, temperature).
Worked examples
Example 1 — Foundational: methylenation of cyclohexanone. React cyclohexanone with methylenetriphenylphosphorane, Ph₃P=CH₂ (generated from methyltriphenylphosphonium bromide + n-BuLi, THF, 0 °C). Step through the mechanism explicitly. The ylide carbanion (CH₂⁻, non-stabilized, no substituents to complicate E/Z since one carbon of the eventual alkene bears two identical H's) attacks the carbonyl carbon of cyclohexanone. Curved arrow: Cα lone pair → carbonyl carbon; C=O π → O, forming the alkoxide/betaine. Ring closure: alkoxide O lone pair → P, forming the oxaphosphetane spiro-fused to the cyclohexane ring. Retro-[2+2]: C–P σ → new C=C π; C–O σ stays on O; P–C(ring... wait, here it's simpler) bond reorganizes to P=O. The product is methylenecyclohexane, (C₆H₁₀=CH₂), plus Ph₃P=O. There is no E/Z question here because the exocyclic methylene carbon carries two identical hydrogens — a useful reminder that stereochemical analysis only applies when the new alkene carbon(s) bear two different substituents. This reaction is the standard method for converting a ketone directly into a terminal, exocyclic methylene alkene, something very difficult to do cleanly by any elimination method (which would require installing a leaving group first, and would risk regiochemical scrambling in an unsymmetrical ketone).
Example 2 — Applied: industrial-style synthesis of a fragrance alkene (Z-selective, kinetic Wittig). Many fragrance and pheromone molecules require a Z-alkene at a specific, otherwise hard-to-control position — the Wittig reaction's Z-selectivity with non-stabilized ylides is exploited industrially for exactly this reason. Consider the synthesis of (Z)-9-tricosene-type insect pheromone alkenes (a real industrial application, used for insect mating-disruption formulations): a long-chain aldehyde, say octanal, CH₃(CH₂)₆CHO, is reacted with a non-stabilized ylide derived from a long-chain alkyl phosphonium salt, e.g. Ph₃P=CH(CH₂)₁₂CH₃ (from the corresponding tetradecyl bromide). Mechanistically nothing changes from the general scheme: nucleophilic addition of the ylide carbon to the aldehyde carbonyl carbon, ring closure to the cis-oxaphosphetane (kinetically preferred, since both substituents — the heptyl chain from the aldehyde and the tridecyl chain from the ylide — are simple alkyl groups with no stabilizing conjugation), and irreversible retro-[2+2] collapse locking in the Z geometry. The result is a long-chain (Z)-alkene with the double bond positioned exactly where the aldehyde carbonyl and the ylide carbon met — critical, because pheromone bioactivity is exquisitely sensitive to both double-bond position and E/Z geometry; even a small percentage of the E-isomer can drastically reduce or abolish activity. Non-stabilized-ylide Wittig chemistry, run under carefully controlled "salt-free" conditions (using bases like NaHMDS instead of lithium bases, since lithium salts can erode Z-selectivity by templating oxaphosphetane equilibration), routinely delivers >95:5 Z:E selectivity for such substrates.
Example 3 — Exam/competition level: predicting E/Z with a stabilized ylide and rationalizing a mixed-selectivity substrate. Consider the reaction of an aliphatic aldehyde, 2-methylpropanal ((CH₃)₂CHCHO, isobutyraldehyde), with the stabilized ylide Ph₃P=CHCO₂Et (from ethyl bromoacetate + PPh₃, then base — this ylide is stabilized because the ester carbonyl conjugates with and delocalizes the Cα carbanion; its conjugate acid, the phosphonium salt, has p around 6, meaning even a mild base like K₂CO₃ deprotonates it, unlike the non-stabilized case requiring n-BuLi). Predict the major product and its stereochemistry, with full justification. Reasoning. Because Ph₃P=CHCO₂Et is a stabilized ylide, invoke the reversibility argument from Mechanism 2, Step 4: nucleophilic addition to the aldehyde carbonyl and the subsequent oxaphosphetane ring closure are reversible, allowing the system to equilibrate toward the lower-energy trans-oxaphosphetane diastereomer before irreversible retro-[2+2] collapse. The trans ring places the isopropyl group (from the aldehyde) and the CO₂Et group (from the ylide) anti to each other around the four-membered ring, minimizing 1,2-strain — note that the isopropyl group is itself moderately bulky, reinforcing the thermodynamic preference for trans even more strongly than a less hindered aldehyde would. Prediction: the major product is ethyl (E)-4-methylpent-2-enoate, (E)-(CH₃)₂CHCH=CHCO�2Et, expected in a ratio on the order of 85:15 to 95:5 E:Z (stabilized ylides with α,β-unsaturated ester ylides routinely give E:Z ratios in this range or higher, especially with branched, sterically demanding aldehydes like isobutyraldehyde which further disfavor the crowded cis oxaphosphetane transition state). Limiting-case sanity check built into the answer: if instead the aldehyde substituent were replaced by something with negligible steric demand (a hypothetical extreme, not formaldehyde specifically since formaldehyde has no stereocenter issue at all, but conceptually a very small substituent), the thermodynamic preference for trans over cis oxaphosphetane would shrink (less steric differentiation), and E-selectivity would be expected to decrease toward closer-to-50:50 — consistent with the general trend that stabilized-ylide E-selectivity is steric-difference-dependent, not an absolute mechanistic guarantee. This is exactly the check an exam grader looks for: recognizing that E-selectivity from stabilized ylides is a thermodynamic preference of degree, tunable by sterics, rather than an all-or-nothing rule.
Limiting cases, scope & common traps
| Situation | Expected behavior | Mechanistic reason |
|---|---|---|
| Non-stabilized ylide (alkyl Cα) | Z-alkene major (often >90:10) | Kinetic, irreversible ring closure to cis-oxaphosphetane |
| Stabilized ylide (ester/ketone/nitrile on Cα) | E-alkene major | Reversible ring closure; equilibration to lower-energy trans-oxaphosphetane |
| Semi-stabilized ylide (allyl, benzyl on Cα) | Mixed E/Z, often modest selectivity, solvent/base-dependent | Intermediate reversibility; both kinetic and thermodynamic pathways compete |
| Lithium salts present (from n-BuLi/LiBr) | Erodes Z-selectivity for non-stabilized ylides | Li⁺ can template betaine equilibration/epimerization before ring closure |
| "Salt-free" ylide (K, Na bases, or ylide pre-isolated) | Improves Z-selectivity | Removes Li⁺-mediated equilibration pathway |
| Base-sensitive substrates (enolizable ketones, epimerizable stereocenters) | Risk of substrate epimerization/aldol side reactions under strongly basic ylide-generation conditions | Ylide-forming bases (n-BuLi, NaH) are strong enough to deprotonate other acidic C–H's in the substrate |
| Ester, amide, nitrile carbonyls as electrophile | Generally unreactive with ordinary ylides | Poorer electrophilicity / poor overlap and unfavorable thermodynamics without a leaving-group-assisted addition–elimination pathway (contrast Horner–Wadsworth–Emmons, which uses stabilized phosphonate carbanions but still targets aldehydes/ketones, not esters) |
| Need for E-alkene from a non-stabilized ylide precursor | Use Schlosser modification (excess base, "instant ylide," or betaine equilibration at low temperature with a second equivalent of strong base to epimerize the betaine before oxaphosphetane closure) | Forces thermodynamic equilibration even for normally Z-selective ylides |
The chemist behind it
Georg Wittig (1897–1987) was a German organic chemist whose career, remarkably, only reached its most celebrated discovery in his mid-fifties. Trained at the University of Marburg, Wittig worked for decades on carbanion chemistry, ate ether chemistry, and the reactivity of organolithium and organophosphorus reagents — patiently building the physical organic groundwork that would let him recognize, in 1953, that a phosphonium ylide reacting with a carbonyl compound could deliver an alkene with a cleanly predictable double-bond position. Working with his student Georg Geissler, Wittig published the reaction that now bears his name, and spent the subsequent decades (mostly at Heidelberg, where he held a chair from 1956) refining its scope, mechanism, and stereochemical behavior — including recognizing, well ahead of full physical-organic proof, that stabilized and non-stabilized ylides behaved in mechanistically distinct ways. He shared the 1979 Nobel Prize in Chemistry with Herbert C. Brown (for hydroboration), a fitting pairing: both prizes recognized methods that gave synthetic chemists precise, predictable control over molecular architecture where previous methods had offered only rough, mixture-generating alternatives.
Origins · 1 minGeorg Wittig and how this was discovered
Wittig's path to his namesake reaction ran through a long apprenticeship in carbanion and ylide chemistry — he had studied ammonium and sulfonium ylides for years before recognizing phosphonium ylides could be tamed into a general, mild carbonyl olefination. His insistence on rigorous mechanistic reasoning over empirical trial-and-error is precisely the tradition this article follows: every curved arrow in the oxaphosphetane mechanism traces back to structural and kinetic studies he and later workers (notably Bestmann and Vedejs) carried out to distinguish the betaine pathway from a concerted, direct-cycloaddition pathway.
Why can't you just use a Grignard reagent to make an alkene from a ketone instead of a Wittig ylide?
A Grignard reagent adds to a ketone to give a tertiary alkoxide, which upon workup gives a tertiary alcohol, not an alkene — there is no good intramolecular pathway analogous to oxaphosphetane collapse, because carbon and magnesium don't form a favorable four-membered ring with a strong enough thermodynamic payoff comparable to P=O formation. You could dehydrate that alcohol afterward with acid, but that reintroduces the classic elimination problems: regiochemical mixtures (Zaitsev vs. Hofmann products), possible carbocation rearrangements, and loss of stereocontrol. The Wittig reaction's genius is combining the addition and elimination into one intramolecular retro-cycloaddition step so the new alkene forms with a predictable, traceable relationship to the original carbonyl geometry.
Does the Wittig reaction work on esters, amides, or carboxylic acids?
Not under normal conditions. Ylides are reasonably good nucleophiles toward the electrophilic, relatively unhindered carbonyl carbon of aldehydes and ketones, but esters, amides, and acids are far less electrophilic (their carbonyl carbon is stabilized by lone-pair donation from the adjacent heteroatom) and, more importantly, even if addition occurred, there is no good four-membered oxaphosphetane pathway available because the tetrahedral intermediate would need to expel an alkoxide leaving group first (an addition–elimination sequence), which ordinary ylides are not set up to do efficiently. Specialized phosphorus ylides and related reagents exist for related transformations, but the classical Wittig reaction is a ketone/aldehyde-specific method.
What is the difference between the Wittig reaction and the Horner–Wadsworth–Emmons (HWE) reaction?
HWE uses a phosphonate ester carbanion, (RO)₂P(=O)–CHR′⁻, generated from a phosphonate ester rather than a phosphonium salt. Mechanistically it proceeds through an analogous betaine/oxaphosphetane-like intermediate, but the phosphonate byproduct, a water-soluble dialkyl phosphate, is much easier to remove than triphenylphosphine oxide, and HWE carbanions are inherently stabilized (by the phosphonate itself), giving reliably strong E-selectivity. HWE is generally preferred industrially for large-scale E-selective olefinations of aldehydes with ester- or ketone-stabilized carbanions, precisely because of this easier purification and highly reliable E-selectivity.
Why does adding lithium salts reduce Z-selectivity in the "classic" Wittig reaction with non-stabilized ylides?
When lithium counterions are present (as they typically are when n-BuLi is used to generate the ylide), Li⁺ can coordinate to the developing alkoxide oxygen in the betaine intermediate, and this coordination opens a pathway for the betaine to reversibly re-open and re-close, allowing some equilibration toward the more stable trans-oxaphosphetane before final collapse. This erodes the normally high Z-selectivity. Using "salt-free" conditions — generating the ylide with a base whose conjugate cation does not template this equilibration (such as sodium bis(trimethylsilyl)amide, NaHMDS, or potassium tert-butoxide with careful control) or physically removing lithium halide byproducts before adding the aldehyde — restores high Z-selectivity by keeping the initial kinetic ring-closure pathway dominant.
Can the Wittig reaction distinguish between two different carbonyl groups in the same molecule (chemoselectivity)?
Yes, in a useful and predictable way: aldehydes are markedly more electrophilic than ketones (less steric hindrance around the carbonyl carbon, and less electron donation from only one alkyl/aryl substituent versus two), so non-stabilized ylides react preferentially, often exclusively, with an aldehyde carbonyl in the presence of an unprotected ketone in the same molecule, provided stoichiometry is controlled (using one equivalent of ylide). This chemoselectivity is heavily exploited in multistep total synthesis to install one alkene at an aldehyde site while leaving a ketone untouched for a later, separate transformation.
Is the oxaphosphetane a real, observable intermediate, or just a mechanistic convenience?
It is real and has been directly observed. Low-temperature ³¹P NMR studies (pioneered by Vedejs and others) have detected oxaphosphetanes as genuine intermediates for a range of substrates, showing characteristic upfield ³¹P shifts distinct from both the starting ylide and the phosphine oxide product, and in some cases oxaphosphetanes have even been crystallographically characterized. This experimental evidence is what shifted mechanistic consensus away from earlier proposals of a long-lived, freely-rotating betaine as the sole or dominant intermediate for typical substrates, toward the current understanding that oxaphosphetane formation (whether via a betaine-like or more concerted pathway) is the stereochemistry-determining step.
Why is triphenylphosphine oxide such an effective driving force, in energetic terms?
The P=O bond in phosphine oxides is unusually strong — commonly cited around 130–140 kcal/mol — substantially stronger than a typical carbonyl C=O bond (roughly 175–180 kcal/mol total bond energy, but recall the reaction is trading one C=O and one P–C bond for one C=C and one P=O bond, not comparing C=O to P=O directly). The key comparison in the bond-energy bookkeeping is that forming P=O from P–C is far more favorable than forming C=C from C=O would be on its own, because phosphorus's larger, more polarizable orbitals and greater electropositivity give unusually favorable orbital overlap and electrostatic stabilization with oxygen. This large thermodynamic sink is what makes the overall four-centered retro-cycloreversion essentially irreversible under standard conditions.