A tryptic digest presents the mass spectrometer with thousands of ions sharing a nominal mass, and a reversed-phase gradient delivers batches of those that also share a retention time. Ion mobility adds a dimension that responds to neither: the time an ion needs to cross a gas-filled region under an electric field, set by how much neutral gas it has to push through. That time is converted into a collision cross section, and the cross section — not the arrival time — is what transfers between instruments. What the conversion assumes decides whether you are holding a structural measurement or a number.
Mobility is measured; cross section is derived
The primary observable is the reduced mobility, K0. A community reporting guide is explicit that K0 values depend intrinsically on ion structure, the nature of the bath gas, temperature, and E/N, and that ion mobility does not measure molecular surfaces directly, but collision cross section (CCS) values are derived from mobility values using a physical model [1]. The same guide states that methods relying on calibration are empirical, and may therefore give method-dependent results, when the gas nature, temperature or E/N cannot be matched to those of the primary method [1]. A CCS is a model output evaluated at stated conditions. A CCS reported without gas, temperature and field is not reusable by anyone else.
The drift gas is not a formality. Bush and co-workers measured cross sections directly in both helium and nitrogen for 113 peptide ions in a low-pressure, ambient-temperature drift cell with radio-frequency ion confinement; the set spans masses of 231–2969 Da, ΩHe of 89–616 Å2 and ΩN2 of 151–801 Å2 [2]. Nor can one gas be rescaled to the other with a single factor. Across a database of denatured peptide, denatured protein, native-like protein and native-like protein complex ions, helium and nitrogen cross sections can be well correlated for larger biomolecular ions, but non-correlated differences for smaller ions can be more significant [3] — and short peptides sit at the small end of that range.
Three separators, three calibration burdens
In a uniform-field drift tube the drift time yields the mobility directly, which is why the reporting guide treats determination of K0 by drift-tube ion mobility as carrying the least uncertainty [1]. Everything downstream of that is a comparison against it.
A travelling-wave separator works on a different principle. Giles and colleagues described superimposing a voltage pulse on the confining RF of one electrode in a stacked-ring ion guide and then moving that pulse to the adjacent electrode and onward along the guide, producing a travelling voltage wave; operated near 0.2 mbar the device functions as a mobility separator [4]. The cost is that the separation parameter is not linear in mobility. Shvartsburg and Smith developed the theory and identified the governing parameter c, the ratio of ion drift velocity at the steepest wave slope to the wave speed. At low c, they report, the ion transit velocity is proportional to the squares of mobility (K) and electric field intensity (E), as opposed to linear scaling in drift tube (DT) IMS, while at the highest c the transit velocity asymptotically approaches the wave speed [5]. Resolving power becomes mobility-dependent, scaling as K1/2 in the low-c limit and less at higher c, and because the transit time depends nonlinearly on mobility, the true resolving power of TW IMS differs from that indicated by the spectrum; near-optimum resolution is achievable over an approximately 300–400% range of mobilities [5]. Every travelling-wave cross section is therefore a calibrated cross section, and its accuracy is inherited from the calibrants.
Trapped ion mobility inverts the geometry: the ion is held by an electric field against a moving bath gas rather than driven through a stationary one. In the original coupling to mass spectrometry, transmission and mobility separation were characterised as a function of source conditions, bath gas velocity, analysis scan speed, RF confinement and downstream ion optics, with the mobility resolution set by the scan speed and the bath gas velocity, and the separation could be tuned from high speed (~25 ms) for rapid analysis to slower scans for higher mobility resolution (R > 80) [6]. Applied to model peptides, the analyser reached resolving power above 250, where many new features, otherwise hidden by lower resolution IMS analyzers, are revealed — up to roughly 3 and 8 times the resolving power of the drift-tube and travelling-wave instruments of the day respectively [7]. That study also draws the distinction that matters for isomer work: unlike differential ion mobility, trapped ion mobility both resolves congested conformational features and returns information about their relative size through the ion-neutral cross section [7].
Cyclic travelling-wave devices buy resolution with path length. A 98 cm closed-loop separator reached a mobility resolving power of approximately 750 over 100 passes, demonstrated on a reverse-sequence peptide pair, while also supporting mobility selection, activation, storage and IMSn in custom combinations [8].
Calibrate with something that resembles the analyte
Calibrated platforms need standards whose cross sections were established on a drift tube, and the choice of standard is not neutral. In the database assembled for exactly that purpose, errors in travelling-wave cross sections determined for native-like protein complexes calibrated using other native-like protein complexes were significantly less than those calibrated using denatured proteins [3]. Match the calibrant class — charge-state range, mass range and compactness — to the analyte class, and record in the method which class you used, because that choice is a term in the reported uncertainty.
For peptides, polyalanine is the workhorse calibrant. Lietz and co-workers tested the assumption that breaks first in practice: high-sensitivity acquisitions run elevated transmission voltages that can activate ions, whereas polyalanine calibration had previously been validated only at reduced voltages chosen to avoid gas-phase activation. They found the additional error introduced under high transmission voltages was not significant, and that the chromatography-coupled acquisition contributed negligible error [9]. The same work carries a warning about library transfer: terminal lysine position, set by whether the digest used LysC or LysN, influenced which of the two structural sub-families triply charged ions preferentially populated, the LysN peptides showing the stronger preference for the compact sub-family [9]. A cross section library is conditioned on sample chemistry, not on sequence alone.
Reading a CCS distribution
A mobility peak is a population, not a molecule. Resolved peaks mean resolved conformer families; a single peak means only that whatever families are present were not separated at the resolving power used. This is why raising resolving power keeps exposing structure that lower-resolution analysers merged, and why even above R = 250 the analysis remains, in the authors' words, principally limited by the plurality of conformational states [7]. Before treating a monomodal arrival-time distribution as evidence of a single conformer, establish what your analyser would have done to a known two-state case at the same settings.
What positions those peaks for peptides is now characterised at scale. Trapped ion mobility with parallel accumulation–serial fragmentation, run over whole-proteome digests of five organisms, produced more than two million cross section values covering roughly 570,000 unique combinations of sequence, charge state and modification, and identified hydrophobicity, proline content and histidine position as main determinants [10]. The directions are specific: peptides with histidines close to the N-terminus are more likely to adopt an extended conformation and peptides with histidines closer to the C-terminal lysine or arginine are more compact; peptides with more prolines had somewhat smaller cross sections on a global scale; and the triply charged species split into two sub-populations [10].
A complementary intrinsic-size-parameter model, built from about 134,000 peptides in charge states 1+ to 4+ acquired by two-dimensional chromatography coupled to trapped ion mobility on digests generated with seven proteases, supplies the mechanism. Its authors find that ion mobility is strongly affected by the peptide's ability to solvate the positively charged sites: internal positioning of polar residues and proline decreases cross sections because it improves charge solvation, and, critically, this ability decreases with increasing peptide charge because of electrostatic repulsion; higher helical propensity and peptide hydrophobicity push peptides towards extended structures with higher than predicted cross sections [11].
Charge state is a variable, not a nuisance
Because charge solvation competes with coulombic repulsion within the same ion, a sequence observed at 2+ and at 3+ presents two different structural problems, and a cross section comparison across charge states compares two different balances. Comparison within a single charge state is the only comparison the measurement licenses without additional argument.
For proteins, the relationship between charge and compactness has been examined directly: the evidence shows that there is often a strong correlation between the degree of folding and charge state, with the most compact conformations observed for the lowest charge states, yet higher charge states do not necessarily indicate unfolding, and a folded, compact state can survive in the gas phase at least on a millisecond timescale [12]. Charge state is a hypothesis about structure, not a readout of it.
It is also a separation axis you can drive deliberately. In epimer work, d/l pairs that co-eluted in one charge state were resolved in another, and pairs that merged as protonated species separated once metalated [13]. If a mixture fails to resolve, changing the charge state distribution or the cationisation mode is a cheaper first move than changing analysers.
The gas phase is not solution
The standing caveat is not that gas-phase structures are artefacts. It is that they are a different ensemble, and the relationship to the solution ensemble has to be argued case by case. Cryogenic ion mobility-mass spectrometry followed the undecapeptide substance P through the final stages of electrospray and found that the anhydrous conformers originate from evaporation of cluster ions — [SP+2H]2+(H2O)n with n from 0 to about 50, and [SP+3H]3+(H2O)n with n from 0 to about 30 — and that major structural changes do not occur during the evaporative process [14]. For the triply charged ion, the same study reported that a compact dehydrated conformer population can be kinetically trapped on the time scale of several milliseconds, even when an extended gas phase conformation is energetically favorable [14].
Kinetic trapping cuts both ways. It is the reason solution-derived structure can survive desolvation at all, and equally the reason an observed population can be dominated by a state that is not the gas-phase minimum. For native-like protein and protein complex ions — a class that does not include the short peptides discussed above, and on which the figure was measured — the average effective density was 0.6 g cm-3, significantly lower than that for the solvent-excluded regions of proteins, which was taken to suggest that such ions can retain significant memory of their solution-phase structures rather than collapsing to globular structures [3]. What a CCS distribution licenses is a statement about the relative compactness of desolvated ions under the conditions you ran. It does not, on its own, license a statement about solution conformer populations.
Collision-induced unfolding as a stability probe
Collisional activation applied ahead of the mobility cell and ramped in steps, with the cross section recorded at each step, turns the instrument into a stability assay. The output is an unfolding trajectory: a series of cross section plateaus and the activation energies at which one gives way to the next. The method enables rapid differentiation of subtly different protein isoforms on the basis of their unfolding patterns and stabilities [15].
Two boundaries govern how far the result travels. The technique was developed on proteins and multiprotein assemblies [15]; a short peptide with few resolvable conformational states offers little to unfold, and a flat trajectory is uninformative rather than evidence of rigidity. And the ramp interrogates the resistance of a desolvated, multiply charged ion to collisional heating on the instrument's own timescale, which is a gas-phase quantity by construction. Use it to rank variants run back-to-back under identical activation settings, charge state and calibration, and treat any absolute stability reading as unsupported.
Isomers chromatography cannot deliver separately
The clearest use of the mobility dimension is where the other two dimensions are degenerate by construction. Inserting gas-phase mobility separation between reversed-phase chromatography and time-of-flight analysis, applied to a combinatorial library designed to contain 4000 peptides randomised over ten amino acids, made it possible to resolve many peptide isomers that have identical retention times and masses [16].
Sequence inversions are a major such class in bottom-up work. Their separation by ion mobility had been limited to isomeric pairs until baseline separation of all seven 8-mer tryptic peptide isomers was achieved using differential ion mobility, with peak-capacity analysis implying that larger libraries should be resolvable for heavier peptides in higher charge states [17]. Note what that technique costs: differential ion mobility separates on the field dependence of mobility and does not return a cross section [7], so it buys resolution and surrenders the structural coordinate. Multipass cyclic travelling-wave separation keeps the coordinate, and a reverse-sequence peptide pair was the demonstration case for resolving power near 750 [8].
Stereochemical isomers are the harder case. Peptides containing a d-amino acid leave both parent and fragment masses unchanged, and their separation by ion mobility was impeded by small collision cross section differences (commonly ~1%) [13]. Trapped ion mobility at resolving power up to about 340 nonetheless achieved broad baseline separation of such peptides with up to about 30 residues, with linear quantification down to 0.25%, establishing utility for real samples across a large inter-isomeric dynamic range [13]. The ceiling has since moved considerably: baseline resolution of the largest known d/l peptides, a 72-residue peptide from Homarus americanus, was demonstrated across trapped, cyclic and field-asymmetric ion mobility with a dynamic range up to 100, and the d and l spectra proved strikingly similar except for a uniform shift of the separation parameter, indicating epimer-specific structural elements conserved across multiple charge states and conformers [18].
The operational lesson is that resolving power is the currency. A 1% cross section difference requires an analyser that can deliver better than 1% separation, and a co-eluting pair that failed to split at low resolving power has not been shown to be a single species.
Cross section as an orthogonal identification coordinate
Because a calibrated cross section is reproducible and orthogonal to both mass and retention time, it can be scored alongside them. The 113-peptide reference set was assembled so that rapidly and accurately calibrated values could be used as an additional parameter to increase the specificity of assignments in proteomics experiments [2]. Prediction closes the loop: a deep recurrent network trained on the million-scale trapped ion mobility dataset predicted cross sections from sequence alone with a median relative error of 1.4% on independent validation peptides, and its authors report that ion mobility can substantially reduce the number of potential peptides that need to be considered and that isobaric peptide sequences can be distinguishable by their different cross sections [10]. The intrinsic-size-parameter approach reached a reported prediction accuracy of about 0.981 across its full peptide population [11].
Read those numbers against one another before designing a filter. A 1.4% median prediction error [10] is the same order as the roughly 1% cross section difference typical of d/l peptide pairs [13]. Predicted CCS is therefore a scoring feature that removes candidates whose experimental value sits far from expectation; it is not a discriminator between near-isomeric alternatives. Separating those requires measured cross sections acquired on one instrument under one calibration at sufficient resolving power — reported with the charge state, drift gas, temperature and field at which they were obtained [1].