Shi Epoxidation

What the Shi epoxidation does

An epoxide is a three-membered ring holding two carbons and one oxygen — a strained little cap that later opens to install two new stereocenters in one shot. The trick is putting that oxygen on one face of the alkene and not the other, so the product is a single enantiomer instead of a racemic 50:50 mix. The Shi epoxidation solves that face-selection problem without any metal at all.

The reagent that actually delivers the oxygen is a dioxirane: a strained ring of one carbon and two oxygens, C(OO). You have almost certainly met its achiral cousin — dimethyldioxirane (DMDO), made from acetone — used to epoxidize alkenes on the bench. Shi's insight was that if the ketone you start from is chiral, the dioxirane it forms is chiral too, and it will hand its oxygen to only one face of an approaching alkene. His chiral ketone is derived in a couple of steps from D-fructose, the sugar in fruit and honey — cheap, abundant, and already loaded with the stereocenters that do the discriminating.

The headline features: it works on plain hydrocarbon alkenes with no directing group, it uses a cheap oxidant (Oxone), the chirality-defining catalyst is used substoichiometrically, and either enantiomer of the product is reachable by choosing D- or L-derived ketone.

The mechanism, arrow by arrow

There are two linked cycles: the ketone is turned into a dioxirane by Oxone, and the dioxirane then epoxidizes the alkene and reverts to the ketone. Trace the electrons:

1. Peroxymonosulfate attacks the ketone. Oxone contains the peroxymonosulfate anion, HSO₅⁻ (a peroxide, HO-O-SO₃⁻). Its terminal peroxide oxygen — a nucleophilic lone pair — adds to the electrophilic carbonyl carbon of the fructose ketone. The C=O π bond breaks and the former carbonyl oxygen picks up the negative charge; the attacking oxygen loses its proton (it is deprotonated at the operating pH), giving a tetrahedral alkoxide called the Criegee intermediate: R₂C(O⁻)(O-O-SO₃⁻).
2. Ring closure ejects sulfate. The alkoxide lone pair swings back down onto the near peroxide oxygen; that O-O σ bond electrons leave with sulfate (SO₄²⁻) as the leaving group. The result is the strained three-membered ring — the dioxirane. This is an intramolecular S_N2-like closure, favored at high pH where the alkoxide is deprotonated.
3. Concerted oxygen transfer to the alkene. The dioxirane presents one of its two oxygens to the alkene's π cloud. In a single concerted step, the alkene's π electrons attack that oxygen while the O-O bond of the dioxirane breaks; the two new C-O bonds form simultaneously and on the same face. Because it is concerted and suprafacial, the alkene's cis/trans geometry is preserved perfectly in the epoxide — a trans alkene gives a trans epoxide every time.
4. The ketone is regenerated. As the epoxide leaves with the transferred oxygen, the dioxirane's remaining C-O collapses back to a C=O double bond. The fructose ketone is reborn, unchanged, and rejoins step 1. That is why it is a catalyst: a little ketone shuttles many oxygens from a large pool of Oxone to the alkenes.

    Cycle A — form the chiral dioxirane
      ketone  +  HSO₅⁻   →   R₂C(O⁻)–O–O–SO₃⁻  +  H⁺   (Criegee intermediate)
      R₂C(O⁻)–O–O–SO₃⁻   →   R₂C<(OO) ring>  +  SO₄²⁻   (dioxirane + sulfate)

    Cycle B — transfer the oxygen
      dioxirane  +  C=C   →   epoxide  +  ketone      (concerted, spiro TS)

    Net:  C=C  +  Oxone  ──(cat. fructose ketone, K₂CO₃, pH 10.5)──→  chiral epoxide

The enantioselectivity is decided entirely in step 3, in the geometry of how the alkene meets the dioxirane — the subject of the next section.

Why "spiro": the transition state that picks the face

Two idealized geometries are possible for a dioxirane handing oxygen to an alkene: spiro and planar. In the spiro arrangement the plane of the alkene is perpendicular to the C-O-O plane of the dioxirane, so the two rings meet at a single point like the two loops of a spiro compound. In the planar arrangement the two π systems lie roughly coplanar.

Shi and others (with computational backing from Houk and Bartlett) showed the spiro transition state is strongly preferred, because it lets an oxygen lone pair align with the alkene's π* orbital — a stabilizing secondary orbital interaction that the planar geometry cannot achieve. Once you know the alkene must approach spiro and end-on, you can predict which face gets the oxygen: the alkene rotates so that its bulkier substituent points away from the fructose ketone's two bulky isopropylidene (acetonide) caps and slots into the one open quadrant. A trans substituent, or the extra group of a trisubstituted alkene, fits that open pocket cleanly; a cis substituent is forced to clash. That single steric requirement is why the method is superb on trans/trisubstituted alkenes and mediocre on cis/terminal ones.

Catalyst, oxidant, and real conditions

• The catalyst is the ketone from D-fructose: 1,2:4,5-di-O-isopropylidene-D-erythro-2,3-hexodiulo-2,6-pyranose, usually just called "the Shi ketone." D-fructose is protected as two acetonides, then the free primary alcohol is oxidized (PCC or similar) to the ketone. Two short, cheap steps from a bulk sugar. Loading is typically 20-30 mol% (early work needed stoichiometric ketone before the pH problem was solved).
• The terminal oxidant is Oxone, the triple salt 2KHSO₅·KHSO₄·K₂SO₄. The active species is peroxymonosulfate, KHSO₅. Cheap, water-soluble, and safe to weigh in air.
• Base / pH. The reaction is run near pH 10.5, held there with K₂CO₃ (added continuously or via a K₂CO₃/AcOH buffer, or a borate/phosphate buffer). This pH window suppresses both catalyst decomposition (self-Baeyer-Villiger) and epimerization while keeping the Criegee-forming step fast.
• Solvent and temperature. A homogeneous mix of acetonitrile / dimethoxyethane and aqueous buffer, run cold (0 °C is common) to slow catalyst decomposition and sharpen selectivity. A phase-transfer additive (Bu₄NHSO₄) helps shuttle the oxidant.
• Additive detail. EDTA is added to sequester trace metals that would otherwise catalyze unproductive Oxone decomposition.

    trans-stilbene  +  Oxone  ──(Shi ketone 30 mol%, K₂CO₃, Bu₄NHSO₄,──→  (R,R)-stilbene oxide
                                  CH₃CN/DME/H₂O buffer, 0 °C)              ~95% ee, ~85% yield

Scope, selectivity, and stereochemistry

The Shi ketone is a face-selective reagent that reads the alkene's substitution pattern. Because transfer is concerted and suprafacial, the alkene's geometry is always retained; the only question is which enantioface reacts.

• Trans-disubstituted alkenes — the sweet spot. trans-β-methylstyrene, trans-stilbene, and cinnamate esters give 90-98% ee.
• Trisubstituted alkenes — excellent, often >90% ee. This is a class Sharpless and many metal methods handle poorly.
• 2,2-disubstituted and certain conjugated dienes/enynes — good; Shi mapped out enol-ether and enyne substrates that give high ee and useful building blocks.
• Cis-disubstituted and terminal (mono-substituted) alkenes — the weakness of the original catalyst; ee often falls below 40%. Shi's answer was a second-generation glucose-derived oxazolidinone ketone (an amine-bearing analog) engineered for exactly these substrates, pushing many cis and terminal alkenes to 80-95% ee.

The absolute configuration is predictable from the spiro model. With the D-fructose ketone, drawing the alkene approaching end-on with its large group in the open pocket tells you which oxygen face is delivered. Want the mirror-image epoxide? Swap to the ent-catalyst from L-fructose. Both antipodes are practical, which is a real advantage over methods where only one chiral ligand is affordable.

Shi vs. Sharpless vs. Jacobsen-Katsuki

	Shi epoxidation	Sharpless epoxidation	Jacobsen-Katsuki
Chiral element	Fructose-derived ketone (organic)	Ti(OiPr)₄ + diethyl/diisopropyl tartrate	Mn(III)-salen complex
Metal?	None	Titanium	Manganese
Active oxidant	Chiral dioxirane (in situ)	Ti-peroxo (from t-BuOOH)	Mn(V)-oxo
Terminal oxidant	Oxone (KHSO₅)	t-BuOOH (TBHP)	NaOCl (bleach), or others
Directing group needed?	No — unfunctionalized alkenes	Yes — allylic alcohol required	No
Best substrates	trans- & trisubstituted	allylic alcohols (E or Z)	cis-disubstituted, conjugated
Typical ee	90-98% (matched substrate)	>90% for allylic alcohols	>90% for cis-alkenes
Both enantiomers?	Yes (D- or L-fructose ketone)	Yes ((+)- or (−)-tartrate)	Yes (either salen antipode)
Nobel Prize	—	Sharpless, 2001	—

The three are complementary, not competing. Sharpless owns allylic alcohols; Jacobsen-Katsuki owns cis-disubstituted and conjugated alkenes; Shi owns trans- and trisubstituted alkenes — and does it without a metal, which matters when metal contamination in a pharmaceutical is a problem.

Worked example: an epoxide en route to a natural product

Take a simple, textbook-clean case first: epoxidizing trans-β-methylstyrene.

    Ph–CH=CH–CH₃ (E)  +  Oxone
        │  Shi ketone (30 mol%), K₂CO₃, Bu₄NHSO₄, EDTA,
        │  CH₃CN / DME / aq. K₂CO₃-AcOH buffer (pH ≈ 10.5), 0 °C
        ▼
    (2R,3R)-2-methyl-3-phenyloxirane      ~94% ee, trans epoxide, high yield

• Why it works so well: a trans-disubstituted styrene — the phenyl and methyl point to opposite sides, so one of them slots into the open dioxirane pocket in the spiro TS with no clash.
• Stereochemistry: the concerted, suprafacial delivery keeps the two substituents trans in the epoxide, and the D-fructose ketone selects a single face → a single (2R,3R) enantiomer.

The famous application is the asymmetric total synthesis of complex polyol and polyene natural products where several epoxides must be set with defined stereochemistry — most famously Corey's synthesis of the C₂-symmetric oxasqualenoid glabrescol, where a single Shi tetra-epoxidation of a tetraene sets four epoxides that then cascade-cyclize into the fused polyether core. Because the catalyst reads local alkene geometry rather than a distant directing group, a substrate with several trans double bonds can be poly-epoxidized with predictable, matched stereochemistry — a control that is very hard to reproduce with a directing-group-dependent method. The Shi epoxidation is now a standard tool in the medicinal chemist's kit for setting epoxide stereocenters in candidate drugs, precisely because it leaves no metal behind.

Limitations and side reactions

• Cis and terminal alkenes (original ketone). Low ee, as discussed — use the second-generation glucose/oxazolidinone ketone for these.
• Catalyst self-destruction. The ketone can Baeyer-Villiger-oxidize itself (the same Criegee intermediate can migrate a C-C bond instead of closing to the dioxirane) and can epimerize. Both eat catalyst; both are why pH control and cold temperatures matter, and why loadings are 20-30 mol% rather than 1 mol%.
• Acid-sensitive products. Epoxides can ring-open under the aqueous conditions if the workup drifts acidic; buffering and prompt workup protect the strained ring.
• Oxidizable functional groups. Oxone/dioxirane will also oxidize sulfides to sulfoxides, amines to N-oxides, and can over-oxidize very electron-rich arenes; chemoselectivity must be considered.
• Oxone stoichiometry and gas. Excess oxidant is used, and CO₂ evolves from the K₂CO₃ buffer as acid builds — reactions are run open or vented, not sealed.

Who, and when

Yian Shi, then at Colorado State University, reported the fructose-derived ketone epoxidation in J. Am. Chem. Soc. in 1996 (Tu, Wang, Shi), with the key pH-control paper that made it truly catalytic following in 1997 (Wang, Tu, Frohn… Shi). The idea built on earlier demonstrations by Robert Curci in the 1980s that chiral ketones could transfer oxygen via dioxiranes with modest ee; Shi's leap was choosing a rigid, C₂-adjacent, sugar-based ketone whose stereocenters gave large, reliable selectivity, and then engineering the reaction pH so the sugar survived as a catalyst. The second-generation oxazolidinone (glucose-derived) ketone for cis and terminal alkenes came around 2000. The transition-state analysis that cemented the spiro model drew on computational work by K. N. Houk and mechanistic studies by Paul Bartlett, among others.

Green-chemistry and practical notes

• No metal to remove. The single biggest process advantage: no titanium or manganese to strip out of a drug substance to meet residual-metal limits.
• Renewable catalyst. The chiral information comes from a bulk sugar (fructose), not a designed ligand made from petrochemicals — a genuinely bio-derived catalyst.
• Benign byproducts. Oxone reduces to potassium sulfate; the only stoichiometric waste is water-soluble inorganic salt.
• Scale considerations. High catalyst loading and dilute cold buffered conditions are the trade-offs; process teams optimize buffer, phase-transfer catalyst, and oxidant addition rate to run it on multi-gram-to-kilogram scale.