Take 1-phenylpropan-1-ol: a benzene ring bonded to a three-carbon chain that carries an OH. Suppose you have to make it from benzene and pieces of three carbons or fewer. The OH could be put in place several ways, but the bond that joins the ring to the chain can come from only a few reactions. Choosing that reaction is most of the problem.
The same is true of synthesis in general. Changing one functional group into another, without touching the carbon skeleton, is called functional group interconversion. There are many of them, and there is usually more than one way to reach a given group. Building the skeleton is different. A first-year course has only about ten reactions that make a carbon–carbon bond, so choosing among them is where a synthesis problem is decided.
In the language of Thinking backwards, each reaction on the list is a disconnection you can make. A disconnection is a bond in the target that you imagine cutting, to find the two pieces that could have formed it. This page collects the ten reactions, shows how counting carbons narrows the choice, and marks the ones that close rings.
The list
Learn the reactions as a list. When you are stuck on a target, ask which of the ten could have made each C–C bond in it. That question usually finds the route.
| Reaction | Joins | Gives |
|---|---|---|
| Grignard / organolithium + carbonyl | R⁻ + C=O | Alcohol |
| Grignard + CO₂ | R⁻ + CO₂ | Carboxylic acid (one carbon longer) |
| Acetylide + alkyl halide (SN2) | RC≡C⁻ + R'X | Alkyne (internal unless R = H) |
| Cyanide + alkyl halide (SN2) | ⁻CN + RX | Nitrile (one carbon longer) |
| Aldol | enolate + C=O | β-hydroxy carbonyl |
| Claisen | enolate + ester | β-keto ester |
| Michael (conjugate) addition | enolate + enone | 1,5-dicarbonyl |
| Diels–Alder | diene + dienophile | Cyclohexene (two bonds at once) |
| Friedel–Crafts alkylation / acylation | arene + R⁺ or RCO⁺ | Substituted arene |
| Wittig | ylide + C=O | Alkene |
The Joins column uses short names. R⁻ stands for the carbon of a Grignard reagent (RMgBr) or an organolithium (RLi). That carbon reacts as if it were a carbanion. An acetylide is the anion of a terminal alkyne. An enolate is the anion formed when a base removes a hydrogen from the carbon next to a C=O. An enone is a ketone whose C=O is conjugated with a C=C. An arene is an aromatic ring such as benzene, and an ylide is the Wittig reagent, such as Ph₃P=CH₂.
The figure draws one example of each reaction as its product, with the new bond in color.
Several of the ten are one idea. In the aldol, the Claisen and the Michael, an enolate attacks something electrophilic. What it attacks decides which of the three you get.
The colored bonds in the figure also sort the ten by where the new bond lands. Four put the new carbon onto a carbonyl carbon, or onto the carbon that was one: the Grignard on a ketone, the Grignard on CO₂, Friedel–Crafts acylation and the Wittig. Two run from an α carbon to a carbonyl carbon: the aldol and the Claisen. The Michael runs from an α carbon to the β carbon of an enone. The last three, the Diels–Alder and the acetylide and cyanide alkylations, form a bond with no carbonyl carbon at either end. The nitrile carbon becomes a carbonyl carbon only if you hydrolyze it afterwards.
A few close relatives use the same bond-making step on a different partner. An enolate attacks an alkyl halide in the malonic and acetoacetic ester syntheses. A Grignard opens an epoxide at its less hindered carbon (Grignard reagents). A lithium dialkylcuprate, R₂CuLi, adds to the β carbon of an enone, just as a Michael donor does (Cuprates and conjugate addition). When a bond in a target fits none of the ten, check these next.
One reaction is kept off the list because it does a job none of the ten can do. Palladium cross-coupling joins two sp2 carbons: aryl to aryl, or aryl to alkenyl. An aryl carbon is a carbon of a benzene ring, and an alkenyl carbon is a carbon of a C=C. A biaryl bond, one that joins two benzene rings directly, has no disconnection anywhere in the table. When a target has one, cross-coupling is the reaction to reach for (Palladium cross-coupling).
Four limits that rule out routes
Each of these limits rules out a route that looks fine on paper.
- The acetylide and cyanide reactions are SN2, so they work well only on methyl and primary halides. A secondary halide gives a lot of elimination. Acetylide is the worse case; cyanide is a weaker base and tolerates a secondary halide better. A tertiary halide gives the alkene instead of any substitution product.
- A Grignard or organolithium cannot be stopped at the ketone when it reacts with an ester or an acid chloride. It adds twice, and the product is a tertiary alcohol carrying two identical R groups.
- A crossed aldol or crossed Claisen, between two different carbonyl partners, is clean only in two cases. Either one partner has no α hydrogen, so it cannot form an enolate, or one partner is turned completely into its enolate first with a strong, hindered base such as LDA.
- Friedel–Crafts alkylation has two problems. Its carbocation can rearrange, and its product is more reactive than benzene, so it reacts again (polyalkylation). Acylation has neither problem, which is why it is almost always the one to plan with.
Count the carbons first
Target: 1-phenylpropan-1-ol, from benzene and any pieces of three carbons or fewer.
Count. The target has nine carbons: six in the ring and three in the chain. Benzene supplies the six, and one three-carbon piece supplies the rest. Two pieces need exactly one C–C bond, and it is bond a, the one that joins the ring to the chain.
Find the reactions that make bond a. The chain carbon at bond a is the carbinol carbon, the carbon that carries the OH. A C–C bond to a carbinol carbon is the retron for a Grignard; a retron is the structural pattern that signals a disconnection. So PhMgBr and propanal can make bond a. Benzene brominates to bromobenzene (Br₂, FeBr₃), which gives PhMgBr with magnesium.
Check the other reaction that joins a ring to a chain. Friedel–Crafts acylation with propanoyl chloride (AlCl₃) makes bond a too, at the ketone stage. NaBH₄ then reduces the ketone to the alcohol. That route is one step shorter, and it is also a complete answer.
Rule out bond b. Ethylmagnesium bromide and benzaldehyde would make bond b. But benzaldehyde already has a carbon on its ring, so that route needs a second C–C bond. Benzaldehyde can be made from benzene in four steps (Br₂ and FeBr₃, then Mg, then HCHO, then PCC), and one of those steps is that second C–C bond. The count shows the route is a detour. A Friedel–Crafts acylation cannot shortcut it, because formyl chloride is too unstable to use.
The count did the real work. It showed that only one bond had to form: the one joining the ring to the chain.
In general, count the carbons in the target and in the permitted starting materials before you plan anything. The difference tells you how many C–C bonds you must form and roughly what size the pieces are.
- If the target has the same number of carbons as the starting material, compare the two skeletons. Butan-1-ol and 2-methylpropan-2-ol both have four carbons, but in butan-1-ol they form an unbranched chain, and in 2-methylpropan-2-ol three of them branch from a central carbon (see the figure below). When the carbons are joined in the same pattern, no C–C bond has to form, and the whole problem is functional group interconversion (Functional group interconversion). When they are joined differently, at least one C–C bond has to change, even though no carbon is added.
- If the target has one more carbon, one C–C bond forms. When the new carbon carries the functional group, the reaction is a one-carbon extension, and there are four of them, set out below. When the new carbon is a plain CH₃ group with no functional group of its own, a methyl reagent adds it, as the carbonyl section shows.
- If the target has several more carbons, count the pieces. Each piece joined on needs one new bond, so three pieces in a row need two bonds. Closing a ring needs one bond more. Look for disconnections that split the target into pieces of the sizes you were given.
The one-carbon extensions are worth knowing as a set. "Make this molecule one carbon longer" is a very common exam instruction, and each route ends somewhere different. The figure starts each route from 1-bromobutane.
Target: pentanoic acid, from butan-1-ol and any one-carbon reagent.
Count. The target has five carbons and butan-1-ol has four. One carbon is missing, so one C–C bond forms, and it is a one-carbon extension. The Wittig makes an alkene, so three of the four are worth comparing, and each leaves the new carbon in a different functional group.
Choose the extension by what the new carbon must become. Here it has to become the carboxyl carbon. The cyanide route gets there after hydrolysis, and the CO₂ route gets there directly, in the same number of steps. The formaldehyde route stops at a primary alcohol and needs one more step, an oxidation.
Forward, via cyanide.
1. PBr₃ → 1-bromobutane.
2. NaCN in DMSO → pentanenitrile. The halide is primary, so the SN2 works.
3. H₃O⁺, heat → pentanoic acid.
Forward, via CO₂.
1. PBr₃ → 1-bromobutane.
2. Mg, dry ether → butylmagnesium bromide.
3. CO₂, then H₃O⁺ → pentanoic acid.
Both routes have the same length and the same first step, and either is a complete answer. Oxidizing butan-1-ol would not be an answer. Jones reagent turns it into butanoic acid, which still has four carbons, because an oxidation never adds a carbon. Counting first stops you from writing that route down.
Where the new bond can go relative to a carbonyl
Take pentan-3-one, a five-carbon chain with its C=O on the middle carbon. The carbon next to the C=O carbon is the α carbon, and the one after that is the β carbon. A methyl group can be joined at three different carbons. The first two start from pentan-3-one itself. The third, at the β carbon, starts from the enone pent-1-en-3-one, which has a C=C next to the C=O. The figure shows all three.
- A nucleophile that attacks the carbonyl carbon builds a bond at that carbon. Grignard reagents, organolithiums, cyanide and hydride all react here.
- An enolate is nucleophilic at the α carbon, so it builds a bond next to the carbonyl. The ketone itself becomes the nucleophile, and the partner is the electrophile, such as an alkyl halide, an aldehyde or ketone (the aldol), or an ester (the Claisen).
- Conjugate addition to an enone builds a bond at the β carbon, two carbons away from the C=O carbon. Only an enone offers this site. Its C=O pulls electron density out of the C=C, and that leaves the β carbon electron-poor. A cuprate or a Michael donor adds here.
So a carbonyl group gives you three places to attach a new carbon (the third only when it is part of an enone), and the reagent chooses among them. That is a large part of why carbonyl compounds dominate synthesis.
Reactions that build rings
Most of the list joins two pieces into a chain. Two kinds of reaction also close rings, which is otherwise hard to do. One is the Diels–Alder. The other is an aldol or a Claisen with both partners in one molecule.
The Diels–Alder joins a diene, two C=C bonds separated by one single bond, to a dienophile, a C=C that usually carries a C=O. It forms two C–C bonds in one step and closes a cyclohexene (The Diels–Alder reaction).
It also sets stereochemistry in a way you can predict. Dimethyl maleate has its two esters cis on the C=C, and it gives a ring with the two esters cis. Dimethyl fumarate has them trans, and it gives them trans. The dienophile keeps its cis or trans relationship because both new bonds form on the same face of each partner.
When the dienophile carries a C=O, the major product is usually the endo adduct. The endo adduct forms with the dienophile's C=O group lying under the diene. In the bicyclic adduct from cyclopentadiene and maleic anhydride, that puts the anhydride ring on the side away from the CH₂ bridge.
So when a target has a cyclohexene ring with defined stereocenters on it, one Diels–Alder disconnection often settles both the skeleton and the stereochemistry. That makes it the most valuable single disconnection on the list.
The second kind is an aldol or a Claisen with both partners in the same molecule. Take heptane-2,6-dione, a seven-carbon chain with C=O groups on C2 and C6. With base, the end carbon C1 becomes an enolate and attacks the C6 carbonyl carbon. That closes a six-membered ring. The first product is a β-hydroxy ketone, and with heat it loses water to give 3-methylcyclohex-2-en-1-one. This is an intramolecular aldol: an aldol inside one molecule. The Robinson annulation ends with this same step.
To predict the ring size, number the chain from one C=O carbon (1) to the other. In heptane-2,6-dione the second C=O carbon is atom 5, so the compound is a 1,5-dicarbonyl. The attacking α carbon lies outside the stretch from atom 1 to atom 5, so the ring holds the α carbon plus atoms 1 to 5: six atoms. A 1,4-diketone such as hexane-2,5-dione closes a five-membered ring the same way, and gives 3-methylcyclopent-2-en-1-one.
The attacking α carbon can also lie inside that stretch. Hexanedial, a 1,6-dialdehyde, closes when its C2 enolate attacks C6. That ring holds atoms 2 to 6, five atoms, and C1 stays outside it as a CHO group; loss of water gives cyclopent-1-ene-1-carbaldehyde (Aldol reactions). One rule covers every case: the ring holds each atom from the attacking carbon to the attacked C=O carbon. Five- and six-membered rings are the ones that form easily.
The Dieckmann condensation is the Claisen run inside one molecule (Claisen reactions). Take diethyl hexanedioate, whose two ester carbons are atoms 1 and 6. Ethoxide removes a hydrogen from C2, and that enolate attacks the ester carbon C6, closing a ring of atoms 2 to 6: five atoms. Ethoxide then leaves, and the C=O re-forms. After acid workup, the product is ethyl 2-oxocyclopentane-1-carboxylate, a cyclic β-keto ester.
In a diester the attacking carbon always lies inside the stretch between the ester carbons, because only OEt lies beyond each ester carbon. So a 1,6-diester closes a five-membered ring, and a 1,7-diester such as diethyl heptanedioate closes a six-membered ring.
In a target, tell the two apart by the product each one leaves. The intramolecular aldol, after loss of water, leaves a ring C=C conjugated with a C=O. The C=O is either in the ring (a cyclic enone) or just outside it (the CHO from hexanedial). The Dieckmann leaves a ring C=O with an ester on the next ring carbon: a cyclic β-keto ester. Its 1,3-dicarbonyl relationship appears in the product, not in the starting diester. Each of those product patterns is the retron for its reaction.
What carries forward
This list is the skeleton of every synthesis question you will meet. Functional group interconversions decorate it, protecting groups keep it from interfering with itself, and route planning puts the steps in the right order. The C–C bonds are the decisions, and everything else is bookkeeping around them.