The Complete Set of Symmetry Operations

Every modern electronic-structure calculation on a molecule begins, invisibly, with exactly this operation census. Quantum chemistry packages such as Gaussian, ORCA, and Psi4 run a numerical symmetry-detection routine on input Cartesian coordinates that tests candidate axes, planes, and inversion centers for each of the five operation types (within a numerical tolerance for bond-length and angle imprecision), assigns the point group, and then uses that group to symmetrize the basis set — cutting computational cost dramatically because integrals that must vanish by symmetry (forced to zero by an odd-under-some-operation character) are never even evaluated. Getting the operation inventory wrong at this first step propagates directly into wrong predicted orbital degeneracies and wrong vibrational mode counts. In vibrational spectroscopy, the mutual exclusion rule for centrosymmetric molecules — a vibration active in the infrared cannot also be Raman-active, and vice versa — is a direct corollary of whether $i$ is present in the operation inventory (a fact developed fully in group-theoretic treatments of Introduction to Point Groups but which starts from correctly identifying $i$'s presence or absence, exactly the census performed here). Chemists routinely use this rule diagnostically: observing that no bands coincide between the measured IR and Raman spectra of a compound is taken as strong structural evidence for a centrosymmetric (inversion-containing) geometry, without needing a crystal structure. In stereochemistry and drug design, the presence or absence of improper operations in a candidate molecule's symmetry inventory determines whether it is chiral — directly relevant to pharmaceuticals where one enantiomer is therapeutic and the mirror-image enantiomer is inactive or harmful (thalidomide being the textbook cautionary tale). Chiral catalyst design in asymmetric synthesis explicitly engineers ligand environments around a metal center to eliminate every improper symmetry operation, guaranteeing the resulting complex is chiral and can therefore discriminate between enantiomeric transition states. Materials scientists use the same census to predict whether a crystal can be piezoelectric or exhibit second-harmonic generation — both properties require the absence of an inversion center, so identifying $i$ (or its absence) in a unit cell's point group is the very first screening step in searching for new nonlinear-optical materials.

Ask the tutor: For a molecule like allene ($\text{H}_2\text{C=C=CH}_2$), how do you locate its $S_4$ axis when the two CH₂ planes are perpendicular to each other, and why does it lack a simple $C_4$?

A symmetry operation is a spatial transformation — a rotation, reflection, inversion, or improper rotation — that carries a molecule into a configuration physically indistinguishable from the original. "Indistinguishable" is stronger than "similar": every atom must land exactly on top of an atom of the same element, with all bond lengths and angles reproduced exactly. The geometric object that anchors the motion — the axis, plane, or point about which it occurs — is the symmetry element. It is essential to keep operation and element conceptually separate even though chemists reuse the same symbol for both: $C_2$ can mean "the twofold axis" (element, a line) or "rotate by $180°$" (operation, a motion). Context always disambiguates.

Identity, $E$. The trivial operation: do nothing. Every molecule, however irregular, possesses $E$, because "leave it alone" trivially reproduces the original configuration. Its inclusion looks pedantic until you need Symmetry Operations as a Mathematical Group to have a group identity element — then it is indispensable, not decorative.

Proper rotation, $C_n$. Rotate the molecule by $360°/n$ about an axis; if this produces an indistinguishable configuration, that axis is an $n$-fold rotation axis and the operation is $C_n$. A single axis generates a whole family: $C_n^1, C_n^2, \dots, C_n^{n}$, where $C_n^n = E$ by definition (a full turn). If several axes of different order pass through the molecule, the one of highest order is called the principal axis and conventionally defines the $z$-direction. Water has one $C_2$ axis (through O, bisecting H–O–H); benzene has one $C_6$ axis perpendicular to the ring plus six $C_2$ axes in that plane.

Reflection, $\sigma$. Reflect every point through a plane; if the image is indistinguishable from the original, that plane is a mirror plane. Three sub-labels track the plane's relationship to the principal axis: $\sigma_h$ (horizontal, perpendicular to the principal axis), $\sigma_v$ (vertical, containing the principal axis), and $\sigma_d$ (dihedral, a vertical plane that additionally bisects the angle between two $C_2$ axes perpendicular to the principal axis). Applying the same $\sigma$ twice returns the original: $\sigma^2 = E$.

Inversion, $i$. Project every atom through a single central point and out to the same distance on the opposite side: $(x,y,z) \to (-x,-y,-z)$. If every atom maps onto an atom of the same kind, that point is a center of symmetry, and $i$ is present. As with $\sigma$, applying $i$ twice restores the original: $i^2 = E$. Benzene and $\text{XeF}_4$ have inversion centers; $\text{NH}_3$ and $\text{CH}_4$ do not.

Improper rotation, $S_n$. Rotate by $360°/n$ about an axis, then immediately reflect through the plane perpendicular to that axis — a single composite operation, not two separate steps that could be checked independently. If the result is indistinguishable from the original, an $S_n$ axis is present. Staggered ethane, $\text{CH}_3\text{CH}_3$, is the classic teaching example: rotating by $60°$ alone does not superpose the molecule on itself (it swaps eclipsed/staggered character between the two carbons' substituents), but rotating by $60°$ and then reflecting through the perpendicular plane does — so ethane has an $S_6$ axis coincident with its $C$–$C$ bond, even though $C_6$ itself is absent.

The proper/improper split is not bookkeeping — it has a real physical consequence. A right hand is related to a left hand by reflection, not by any rotation; no amount of turning a right glove in space makes it fit a left hand. Consequently, any molecule that possesses only $E$ and $C_n$ operations (no $\sigma$, $i$, or $S_n$ at all) is chiral: it is not superposable on its mirror image, and the two mirror forms are distinguishable enantiomers. A molecule possessing any improper operation is achiral. This is the deepest chemical payoff of assembling the complete operation list: chirality is decided entirely by whether the improper subset $\{\sigma, i, S_n\}$ is empty or not.

The formal statement of completeness rests on three facts about how these operations combine and generate each other. Writing $R$ for a general symmetry operation and $\hat n$ for a unit vector along an axis, define rotation by angle $\theta$ about $\hat n$ as $C(\theta,\hat n)$ and reflection through the plane normal to $\hat n$ as $\sigma(\hat n)$.

where: $S_n$ is improper rotation by $360^\circ/n$; $C_n$ is proper rotation by the same angle about the same axis; $\sigma_h$ is reflection in the plane perpendicular to that axis. The two operations commute because rotation about an axis and reflection in the plane perpendicular to that same axis act on independent coordinates (the rotation mixes $x,y$; the reflection only flips $z$).

Reading: this equation is the reduction of the five-member list to three generators. Setting $n=1$ ($360°$ rotation, i.e. $C_1=E$) gives $S_1 = \sigma_h \cdot E = \sigma$ — reflection is improper rotation by one full turn. Setting $n=2$ ($180°$ rotation) gives $S_2 = \sigma_h\cdot C_2 = i$ — inversion is improper rotation by a half turn. Every occurrence of $\sigma$ or $i$ in a molecule's inventory can therefore be re-expressed as a special case of $S_n$; the operator zoo has only two truly distinct proper/improper building blocks, $C_n$ and $S_n$ (with $E$ as the trivial $n\to\infty$ or $n=1$ rotation).

where: repeated application of $C_n$ $n$ times returns to start (a full $360°$ turn); repeated $S_n$ behaves differently depending on parity of $n$ because each application both rotates by $360°/n$ and flips the reflection.

Reading: for odd $n$, applying $\sigma_h$ an odd number of times ($n$ times) leaves a net single reflection, but combined with the full $360°$ rotation, it works out that $S_n^n$ collapses to $E$ only after accounting for the accompanying rotation — concretely this is verified operation by operation for $S_3$ below. For even $n$, $\sigma_h$ applied $n$ times ($n$ even) cancels to $\sigma_h^{\,n}=E$ separately from the rotation, but the rotation alone after $n$ applications is a full turn, so what survives depends on the exact power; the clean, checkable statement is that $S_n^2 = C_n^2$ (reflection-squared is identity, so two applications of $S_n$ leave pure rotation by $2\times 360°/n$).

where: this counts how many distinct operations a single $S_n$ axis generates before the sequence closes back on $E$.

Reading: an $S_n$ axis with $n$ even always implies the presence of a genuine $C_{n/2}$ axis coincident with it (because $S_n^2 = C_n^2 = C_{n/2}$), while an $S_n$ axis with $n$ odd implies the presence of both a genuine $C_n$ axis and a separate $\sigma_h$, because $S_n^n = \sigma_h$ for odd $n$ forces $\sigma_h$ to be a symmetry element on its own. This asymmetry between odd and even $S_n$ is exactly the kind of detail that trips people up when cataloguing operations, and it is worked line-by-line below.

Example 1 (foundational): water, $\text{H}_2\text{O}$. Place oxygen at the origin, with the molecular plane as the $yz$-plane and the $C_2$ axis along $z$, bisecting the H–O–H angle ($104.5°$ experimentally, but the exact angle is irrelevant to the symmetry census as long as it is not $180°$ or $0°$). Rotation search: rotating $180°$ about $z$ swaps the two hydrogens with each other while leaving oxygen fixed — since both hydrogens are identical atoms, this reproduces an indistinguishable configuration. So $C_2$ is present. No axis of order $3$ or higher exists (there is no way to divide $360°$ into three or more equal rotations that superpose two atoms related by only a two-fold relationship). Inventory so far: $E, C_2$. Reflection search: the plane containing all three atoms (the molecular plane, here $yz$) trivially reflects every atom onto itself — call this $\sigma_v(yz)$. A second plane, perpendicular to the first, also containing the $C_2$ axis, and bisecting the H–O–H angle from the other side (the $xz$-plane) swaps the two hydrogens with each other while leaving O fixed — again indistinguishable, since the hydrogens are identical. Call this $\sigma_v'(xz)$. No horizontal plane exists (there is nothing above and below to swap; $\sigma_h$ would require an atom set symmetric top-to-bottom relative to a plane perpendicular to $C_2$, which water's bent shape does not have). Inversion search: inverting through O sends each H at $(0,\pm y_0, z_0)$ to $(0,\mp y_0,-z_0)$ — the sign flip on $z$ moves each hydrogen to where no atom sits (there is no hydrogen at negative $z$ in a bent molecule with both H atoms on the same side as the H–O–H opening). So $i$ is absent. $S_n$ search: check $S_2$ first, since $S_2=i$ which was already ruled out. Check $S_4$ (coincident with the $C_2$ axis, testing if a $90°$ rotation-reflection works): rotating $90°$ about $z$ moves the hydrogens out of the $yz$-plane entirely, and no subsequent single reflection through a plane perpendicular to $z$ can bring both back onto the original H positions simultaneously (this reflection only flips $z$, not the required in-plane rotation mismatch). So $S_4$ is absent, and by similar reasoning no higher $S_n$ survives either. Final inventory for $\text{H}_2\text{O}$:

Four operations, no improper $S_n$ beyond the trivial reflections already counted, no inversion. This is the operation set defining point group $C_{2v}$. As a limiting-case check: if the H–O–H angle were artificially opened to exactly $180°$ (a hypothetical linear triatomic $\text{H–O–H}$), the molecule would gain a $\sigma_h$ and a full $C_{\infty v}$ or $D_{\infty h}$ character depending on whether the two H–O bond lengths remain equal — a genuinely different, higher-symmetry inventory, confirming that the bent geometry is what limits water to just these four operations.

Example 2 (applied): staggered ethane, $\text{CH}_3\text{CH}_3$, focus on the $S_6$ operation. Take the C–C bond along $z$. In the staggered conformation, the three hydrogens on the "front" carbon sit at azimuthal angles $30°, 150°, 270°$ while the three on the "back" carbon sit at $90°, 210°, 330°$ — offset by exactly $60°$ from the front set (this offset is precisely what "staggered" means). Test $C_6$ alone: rotating everything by $60°$ about $z$ takes the front hydrogens from $\{30°,150°,270°\}$ to $\{90°,210°,330°\}$ — which is exactly the back hydrogens' azimuthal set — but the front hydrogens are now sitting at the back carbon's height ($z$ unchanged by rotation, so actually the front H's stay at the front carbon's $z$-coordinate). This is a problem: the front carbon's hydrogens have rotated into the correct azimuthal positions for the back set, but they are still attached at the front carbon's height. Since the back carbon does not sit at the front carbon's height, a pure $60°$ rotation does not superpose the molecule onto itself. So bare $C_6$ is absent, consistent with there being no sixfold rotational relationship between atoms at different heights along $z$. Now test $S_6$: rotate by $60°$ about $z$ (moving front H's azimuthally onto the back H's angular positions, as above) and then reflect through the plane $\sigma_h$ perpendicular to $z$, bisecting the C–C bond at its midpoint. That reflection flips $z \to -z$, swapping "front becomes back" in height as well as leaving the already-matched azimuthal angle. Now both azimuthal angle and height match the original back-carbon hydrogen positions, and by the equivalent argument, the original back hydrogens land where the front hydrogens started. The net composite operation superposes the whole molecule onto itself exactly. So:

Because $6$ is even, Step 3 of the parity derivation above applies: $S_6^2 = C_6^2 = C_3$ must also be a genuine symmetry operation — and indeed staggered ethane has a $C_3$ axis along C–C (rotating $120°$ maps each carbon's three hydrogens onto each other, since $120°$ is a symmetry of any freely-standing methyl group's three-fold local arrangement). This matches the actual point group of staggered ethane, $D_{3d}$, which lists $S_6$ and $S_6^5$ explicitly alongside $C_3$, $C_3^2$, three $C_2'$ axes, three $\sigma_d$ planes, $i$, and $E$ — twelve operations in total.

Example 3 (exam/competition level): counting the complete operation inventory of $\text{PF}_5$ (trigonal bipyramidal, point group $D_{3h}$), asymmetric bond-length version. Real $\text{PF}_5$ has two chemically distinct fluorine environments — two axial F at $180°$ to each other along $z$, three equatorial F in the $xy$-plane at $120°$ to each other — with axial P–F bond length $158\ \text{pm}$ and equatorial P–F bond length $153\ \text{pm}$ (these differ because of Bent's-rule-type $s/p$ hybridization differences, not because of any symmetry argument, but the inequality of bond lengths is exactly what a naive student worries breaks the symmetry — so this example deliberately keeps the two lengths unequal to show that operation counting does not care about absolute bond length, only about which atoms map to which). Rotations: the $z$-axis through P, perpendicular to the equatorial plane, is a genuine $C_3$ axis: rotating $120°$ permutes the three equatorial F among themselves (all three have the same 153 pm bond length and the same chemical environment, so permuting them is truly indistinguishable) while leaving both axial F fixed (they lie exactly on the rotation axis). So $C_3, C_3^2$ present. Perpendicular to $z$, passing through P and one equatorial F, is a $C_2$ axis: $180°$ rotation about this line swaps the two axial F with each other (both are 158 pm, identical environment, so this swap is allowed) while swapping the other two equatorial F's with each other and leaving the equatorial F on the axis fixed. There are three such $C_2$ axes, one through each equatorial F. Inventory: $E, 2C_3, 3C_2$. Reflections: the equatorial plane itself ($xy$, containing P and all three equatorial F) is $\sigma_h$ — every atom either lies in the plane (unmoved) or is one of the two axial F reflected into each other (both 158 pm, identical, allowed). Three vertical planes, each containing $z$ and one equatorial F while bisecting the angle between the other two, are $\sigma_v$ planes: each contains one equatorial F and both axial F (unmoved) while reflecting the remaining two equatorial F's into each other. Inventory adds: $\sigma_h, 3\sigma_v$. Inversion: test whether a center at P works. Inverting an axial F at $(0,0,+158)$ sends it to $(0,0,-158)$ — which is exactly where the other axial F sits (same bond length, same element), so that swap is fine. But inverting an equatorial F at, say, $(153,0,0)$ sends it to $(-153,0,0)$ — is there an F atom there? The equatorial fluorines sit at $120°$ intervals starting from some reference angle; $(-153,0,0)$ corresponds to azimuthal angle $180°$, which is not one of $0°, 120°, 240°$ unless the equatorial set happens to include that angle — and for three equally spaced substituents starting at $0°$, the angles are $0°, 120°, 240°$, none of which is $180°$. So inversion fails for the equatorial set: $i$ is absent in $D_{3h}$ $\text{PF}_5$, correctly reflecting the known point group (trigonal bipyramids lack inversion symmetry because three equatorial positions cannot be centrosymmetric — an odd number of equatorial substituents can never inversion-pair among themselves). Improper rotations: test $S_3$ along $z$ (rotate $120°$, then reflect through $\sigma_h$). The $120°$ rotation alone already permutes the equatorial F's correctly (shown above) while leaving axial F's on-axis; the subsequent $\sigma_h$ reflection swaps the two axial F's into each other (allowed, both 158 pm) and leaves the (already-permuted) equatorial F's in the equatorial plane unmoved by the reflection. Net result: full superposition. So $S_3, S_3^5$ present (and since $3$ is odd, the parity derivation says $S_3^3=\sigma_h$ must be an independent element too — and indeed $\sigma_h$ is already separately in the list, consistently). Complete tally for $\text{PF}_5$:

As a limiting-case check: if the two axial bond lengths were made unequal to each other (say $158$ pm on one side, $160$ pm on the other — a genuinely distorted, physically artificial molecule), the $C_2$ axes through the equatorial fluorines would immediately fail (a $180°$ rotation would try to swap two axial atoms of now-different bond length, which is not indistinguishable), and $\sigma_h$ would fail for the same reason, collapsing the point group down to $C_{3v}$ with inventory only $\{E, 2C_3, 3\sigma_v\}$ — six operations, less than half of the symmetric case. This confirms the census procedure is sensitive to exactly the geometric equalities it should be sensitive to, and insensitive to the specific numerical value of any unbroken equal bond length.

The systematic language of symmetry operations as applied to molecules descended from nineteenth-century crystallography and group theory, but its rigorous application to molecular structure and spectroscopy is inseparable from the work of Hans Bethe on crystal-field splitting (1929) and, more directly for chemists, from the mid-twentieth-century synthesis carried out by F. Albert Cotton, whose textbook Chemical Applications of Group Theory (1963) became the standard vehicle through which working inorganic chemists learned to enumerate $E$, $C_n$, $\sigma$, $i$, and $S_n$ systematically and connect the resulting point groups to vibrational spectroscopy and bonding models.