How to Choose an Orthogonal HPLC Column for Impurity Profiling

Leave a Comment / By / October 9, 2026

Choose an orthogonal HPLC column for impurity profiling by its selectivity, not its label. Pick a phase whose interactions differ from the primary column, change pH or organic modifier where the impurities allow, and test both methods on the same stressed and spiked samples. A useful orthogonal method finds peaks the primary method hides.1,2

What makes a second HPLC column orthogonal for impurity profiling?

In method development, an orthogonal method is an additional method with very different selectivity from the primary one. Its job is to check that the primary method stays specific as new impurities appear, and it can keep doing that through development.2 ICH Q2(R2) gives it a formal role: specificity can be shown by the absence of interference or by comparing results with an orthogonal procedure, and where one procedure does not discriminate well enough, a combination of two or more can be used.1

Orthogonality is therefore a property of the separation, not the box. Two C18 columns from different suppliers carry different names, yet the hydrophobic-subtraction model shows that many alkylsilica columns are equivalent for most separations.3 What matters is whether the second method moves the peaks that can hide behind each other on the first.

The payoff is concrete. In three development case studies, Zheng and Rasmussen found impurities that the primary method missed: two impurities coeluting in one batch, a 0.40% peak that was really two coeluting impurities, and an impurity at 0.10% (w/w) hidden under the drug-substance peak.2

Diagnose this automatically

Chromatography Troubleshooting Decision Engine

Any HPLC symptom, one starting point — the engine narrows hundreds of failure modes to the few that fit your evidence.

Start a diagnosis
No account needed

Which stationary phase is most likely to give orthogonal selectivity?

Start from the interactions the primary column does not offer. Alkylsilica (C8, C18) columns are described well by five interactions in the hydrophobic-subtraction model; phenyl, cyano, pentafluorophenyl (PFP) and embedded-polar-group (EPG) phases add interactions the model does not capture, and the model’s error in predicting their retention grows accordingly.3 Dolan and Snyder note that the model is less reliable for phenyl and cyano columns, so compare those phases by experiment.4

Table 1 lists the candidates by the extra interactions they bring, with the model’s average error for retention as a measure of how far each departs from alkylsilica behavior.

Table 1. Candidate second phases for an alkylsilica primary method. Interactions and average prediction errors as reported by Snyder, Dolan and Stoll; a larger error means more selectivity the alkylsilica model does not describe.3
Phase Interactions beyond alkylsilica Average error in predicted k Use as a second column
Alkylsilica (C8, C18) None ±1% Another alkylsilica column is often equivalent; choose one only from a large Fs
Phenyl π–π 7% Candidate; adds π–π interaction
Cyano π–π, dipole 10% Candidate; compare by experiment
PFP π–π, dipole, fluoro-hydrophobicity ±15% First choice orthogonal to alkylsilica
EPG Polar-group interactions ±20% Candidate; pairs well with PFP in two-dimensional work

On that basis, Snyder, Dolan and Stoll suggest a PFP column as the first choice of a phase orthogonal to an alkylsilica column.3 Dolan’s practical rule is similar: a cyano, phenyl or EPG column often changes the separation from a starting C8 or C18 column.5 For the mechanism behind each phase, see how stationary-phase chemistry changes selectivity.

Figure 1 maps these candidates against the alkylsilica primary column, with the levers that act on the mobile phase.

Map of second-phase choices for an alkylsilica primary column. Top band: five columns for alkylsilica (C8, C18, the primary), phenyl, cyano, PFP and EPG, each with a silica particle and a schematic ligand (alkyl chain, phenyl ring, nitrile, pentafluorophenyl ring, alkyl chain with an embedded polar group), the interactions it adds beyond alkylsilica, a bar for its average error in predicted k (±1%, 7%, 10%, ±15%, ±20%) and how to use it; PFP is marked first choice orthogonal to alkylsilica. Bottom band: the five model parameters H, S*, A, B and C with the Fs guidelines (35 or more ignoring C for nonionized samples, 100 or more with ionized solutes), methanol and acetonitrile bottles for the organic-modifier lever, and a pH ruler with a bracket marking about 1 to 1.5 units from the pKa.
Figure 1. Which second phase changes selectivity? The primary alkylsilica (C8, C18) column is described by five hydrophobic-subtraction parameters: hydrophobicity (H), steric resistance (S*), hydrogen-bond acidity (A), hydrogen-bond basicity (B) and cation exchange (C). Phenyl, cyano, PFP and EPG phases add interactions the model does not capture; each bar is its average error in predicted k, a measure of how far it departs from alkylsilica behavior, and PFP is the suggested first choice.3 For two alkylsilica columns, the orthogonal Fs guideline is 35 or more (C term ignored) for predominantly nonionized analytes and 100 or more with ionized solutes; these are guidelines, not limits.3,4 Changing from methanol to acetonitrile can change selectivity, and combined with a column change it raises, but does not guarantee, the chance of an orthogonal separation.5 pH moves ionizable impurities most within about 1–1.5 units of their pKa; neutral compounds are essentially unaffected (see the pH section). Ligands are schematic.

How do you use the hydrophobic-subtraction Fs value to compare columns?

For alkylsilica columns, the hydrophobic-subtraction model characterizes each column by five parameters: hydrophobicity (H), steric resistance (S*), hydrogen-bond acidity (A), hydrogen-bond basicity (B) and cation exchange with ionized silanols (C).4 The column-comparison function Fs is a weighted distance between two columns in that five-parameter space:6

Fs = {[12.5(H2 − H1)]2 + [100(S*2 − S*1)]2 + [30(A2 − A1)]2 + [143(B2 − B1)]2 + [83(C2 − C1)]2}1/2

Subscripts 1 and 2 denote the two columns. The parameters for several hundred columns are published in the USP and PQRI column databases, which are built on this model.4 Table 2 gives the authors’ guideline values. They apply to alkylsilica columns and are guidelines, not limits.

Table 2. Guideline Fs values for alkylsilica columns, as stated by the model’s authors.3,4
Fs Reading Scope
≤ 3 Equivalent for most separations Alkylsilica columns
3 to ≈10 Likely equivalent Especially for few solutes with Rs well above 2
≥ 35, with the C term ignored Orthogonal Samples of predominantly nonionized analytes
≥ 100 Orthogonal Samples containing ionized solutes

The general recommendation is the alkylsilica column with the largest possible Fs that still meets the method’s other requirements.3 Earlier guidance used a different single cut-off (about 65), so quote the value and its scope from the source you use.5 A large Fs predicts different selectivity; only the impurity samples show whether the difference falls where you need it.

Should you change pH or organic modifier as well as the column?

Often, yes. Changing from methanol to acetonitrile can change selectivity strongly, and combining a solvent change with a column change further increases the chance of an orthogonal separation, although it does not guarantee one.5 Mobile-phase pH is an additional lever for ionizable impurities.5

pH acts on the analyte, not the column. It is most effective within about 1–1.5 pH units of an analyte’s pKa, while neutral compounds are essentially unaffected; in one example, the separation factor between two anilines fell from 1.64 at pH 2.0 to 1.05 at pH 5.5.7 Choose a buffer with a pKa within 1 unit of the target pH, stay inside the column’s documented pH range, and for a robust final method keep the pH more than 1.5 units from the analytes’ pKa.7 The case studies used all three levers: a C8 method with formic acid was checked by a PFP method with trifluoroacetic acid, and a C18 method at low pH by a C18 method with ammonium acetate.2

How do you judge orthogonality from an impurity sample?

Judge it on samples that contain the impurities you care about. Zheng and Rasmussen screened six broad gradients on each of six columns, 36 conditions per sample, and selected forced-degradation samples degraded 5–15%, because samples degraded further may contain secondary degradation products.2

Compare the two chromatograms of the same identified sample. Peaks that change order, a peak that splits into two and a new peak that emerges from under the main component are the evidence.2 For a critical pair, express the change as the separation factor; a pair whose order reverses passes through α = 1 between the two conditions, as covered in separation factor in HPLC.

Figure 2 shows the pattern these case studies report, drawn as schematic chromatograms of one sample on the two methods.

Two schematic chromatograms of the same stressed sample, stacked. Top, primary method (C8, formic acid): Imp 1, a single peak that is Imp 2 and Imp 3 coeluting (dashed outlines show the two components), the API peak with Imp 4 hidden under its tail, and Imp 5; a note on the API peak says DAD purity can pass here but is not proof. Bottom, orthogonal method (PFP, trifluoroacetic acid): Imp 2, Imp 1, Imp 3, the API, Imp 4 and Imp 5 as separate peaks, with Imp 2, Imp 3 and Imp 4 highlighted. Dotted lines link each compound between the traces: Imp 1 and Imp 2 cross (order change), Imp 2 and Imp 3 fan out (a split) and Imp 4 moves out from under the API (a new peak). Below, three numbered tiles name these kinds of evidence, each with a small before-and-after trace sketch.
Figure 2. One stressed sample, two methods. One stressed sample shows the pattern Zheng and Rasmussen report: impurities hidden by coelution on the primary method are revealed by the orthogonal method.2 Imp 2 and Imp 3 coelute and Imp 4 hides under the API on the primary method; the orthogonal method resolves them and swaps Imp 1 and Imp 2. Peaks are assumed to be matched by identity (for example with spiked samples where impurities are available); conditions as in one of the reported case studies; the peak pattern is schematic. Each compound keeps its peak area on both traces, and a merged peak’s area equals the sum of its components. DAD purity software tests for one spectroscopic signature, so the API peak can pass if Imp 4’s spectrum is similar to the API’s or its absorbance is much lower.8 Schematic traces, not data; peak positions are not retention predictions.

Why is a passing peak-purity result not enough?

Diode-array peak-purity software answers a narrow question: is the peak made of compounds with a single spectroscopic signature? Impurities often have spectra highly similar to the parent compound’s, and an impurity present at much lower absorbance can go undetected; in a published example a lot passed spectral purity although an impurity of about 0.13% of the area was present.8 Stoll and co-workers recommend combining diode-array detection with mass spectrometry and screening complementary column selectivities.8 A passing purity result supports specificity; it does not prove it.

How does an orthogonal method support specificity in validation?

For separation methods, ICH Q2(R2) states that specificity can be demonstrated by the resolution of the two components that elute closest to each other, and that discrimination can be established by stressing or spiking the product.1 Where impurities cannot be obtained, results on samples containing typical impurities or degradants can be compared with an orthogonal procedure, and the approach should be justified.1 The guideline describes that second procedure as well-characterized; see the ICH Q2(R2) guideline, section 3.1.

Table 3 turns the steps into a decision table. Each step ends in a record that the next step uses.

Table 3. Choosing and confirming an orthogonal method. Steps are sequential.1,2,3,8
Step What to do Record Move on when
1. Screen Run a stressed sample (5–15% degraded) on candidate phases unlike the primary, with a change of pH or organic modifier Conditions for every run A condition separates the sample differently from the primary method; if none does, screen another phase or condition
2. Compare Compare identified peaks on both methods: order changes, splits, new peaks Peak identities and critical pairs on each method The second method resolves what the primary may hide
3. Confirm specificity Check resolution of the closest-eluting pair and stressed or spiked samples; use DAD and MS purity as supporting evidence Resolution, spiking and purity results Interference is excluded or explained
4. Document Record the comparison with the orthogonal procedure and the justification Both procedures, samples and results Keep the orthogonal method to evaluate the primary method on new batches

Figure 3 follows Table 3 step by step, including the return to screening when no condition separates the sample differently.

Four-step workflow matching Table 3, in two bands, with columns for what to do, what to record and when to move on. Step 1, Screen (column and mobile-phase bottle icons): run a 5–15% degraded stressed sample on candidate phases with a pH or organic-modifier change; a dashed loop returns to screening if no condition separates the sample differently. Step 2, Compare (a single peak above the same two compounds resolved, each keeping its area): compare identified peaks for order changes, splits and new peaks. Step 3, Confirm specificity (DAD and mass spectrometer icons): resolution of the closest-eluting pair, stressed or spiked samples, DAD and MS purity as support. Step 4, Document (record page): record the comparison and justification and keep the orthogonal method for new batches.
Figure 3. Choosing and confirming an orthogonal method. The figure follows Table 3 row for row; steps are sequential. Screen a stressed sample (5–15% degraded) on candidate phases unlike the primary, with a change of pH or organic modifier, and record the conditions for every run; move on when a condition separates the sample differently from the primary method, and if none does, screen another phase or condition. Compare identified peaks on both methods for order changes, splits and new peaks. Confirm specificity from the resolution of the closest-eluting pair and stressed or spiked samples, with DAD and MS purity as supporting evidence, until interference is excluded or explained. Document the comparison with the orthogonal procedure and the justification, and keep the orthogonal method to evaluate the primary method on new batches.1,2,8 Schematic icons and traces, not data; each compound keeps its area.

For where column choice sits in the full development sequence, see how to develop an HPLC method. For validated or compendial methods, the applicable procedure and regulatory framework take precedence over the general rules of thumb given here.1

Frequently asked questions

Is a second C18 column from another supplier orthogonal?

Usually not by default. Many alkylsilica columns are equivalent for most separations; an alkylsilica column qualifies only when its Fs against the primary column is large for the kind of sample you have. A phase with different interactions, such as PFP, is the more likely candidate. Look up both columns in the USP or PQRI database before you spend time screening.3,4

Is a column with a different USP L-code automatically orthogonal?

Not automatically. Orthogonality means very different selectivity from the primary method, and the USP and PQRI column databases describe that selectivity with the hydrophobic-subtraction parameters rather than with the packing class. Check the parameters where they exist, and compare the impurity samples on both methods. The stressed and batch samples that matter to the method are the real test.2,4

Which Fs value means two columns are orthogonal?

For alkylsilica columns, the model’s authors give guidelines of 35 or more (ignoring the C term) for mostly nonionized samples and 100 or more when ionized solutes are present. These are guidelines for that column family, not limits, and older guidance used a different value. Decide first whether your impurities are ionized at the method pH, because that decides which value applies.3,5

Can I change only the mobile phase instead of the column?

Sometimes. pH can move ionizable impurities strongly when it is within about 1–1.5 units of their pKa, but it leaves neutral compounds essentially unchanged. A change of organic modifier combined with a column change gives a better chance of an orthogonal separation. For the final method, keep the pH more than 1.5 units from the pKa so small pH errors do not move the peaks.5,7

Can DAD peak purity replace an orthogonal method?

No. Purity software tests for a single spectroscopic signature, and impurities with similar spectra or low absorbance can pass. In one published example, a lot passed spectral purity although an impurity of about 0.13% of the area was present. Use it with MS and an orthogonal separation, not instead of them, and record it as supporting evidence.8

Which samples should I use to compare the two methods?

Use forced-degradation samples degraded about 5–15%, plus batches with known impurities, so both methods see the peaks that matter without secondary degradants. Samples degraded further may contain secondary degradation products. Spiked samples help where impurities are available. Run the same sample on both methods so that peaks can be matched by identity.1,2

The takeaway

An orthogonal column earns its place by finding what the primary method misses. Choose a phase with different interactions, usually PFP or another non-alkyl phase when the primary is C18, consider pH and organic modifier too, and prove the difference on stressed and spiked samples rather than on a label or a purity score.1,2,3

References

  1. International Council for Harmonisation. ICH Q2(R2) Validation of Analytical Procedures. Step 4, 1 November 2023. https://database.ich.org/sites/default/files/ICH_Q2%28R2%29_Guideline_2023_1130.pdf — specificity by absence of interference or orthogonal procedure comparison (3.1.1); resolution of the closest-eluting pair, stressing or spiking, comparison with an orthogonal procedure when impurities are unavailable (3.1.2.2).
  2. Zheng F, Rasmussen HT. Use of orthogonal methods during pharmaceutical development: case studies. LCGC Supplements. 2009;27(4). https://www.chromatographyonline.com/view/use-orthogonal-methods-during-pharmaceutical-development-case-studies — purpose of an orthogonal method; 36-condition screen; 5–15% degraded samples; three cases of impurities missed by the primary method.
  3. Dolan JW, Stoll DR, Snyder LR. How to take maximum advantage of column-type selectivity. LCGC Europe. 2017;30(9). https://www.chromatographyonline.com/view/how-take-maximum-advantage-column-type-selectivity-0 — Fs guideline values for alkylsilica columns; largest Fs recommendation; PFP as first orthogonal choice; extra interactions and prediction errors by phase.
  4. Dolan JW, Snyder LR. The hydrophobic-subtraction model for reversed-phase liquid chromatography: a reprise. LCGC North America. 2016;34(9). https://www.chromatographyonline.com/view/hydrophobic-subtraction-model-reversed-phase-liquid-chromatography-reprise — the five column parameters; USP and PQRI databases; Fs guidelines; lower reliability for phenyl and cyano columns.
  5. Dolan JW. The perfect method, part V: changing column selectivity. LCGC North America. 2007;25(10). https://www.chromatographyonline.com/view/perfect-method-v-changing-column-selectivity — cyano, phenyl or EPG versus C8/C18; earlier Fs value of about 65; combining solvent and column changes; pH as an additional lever.
  6. Stoll DR. Selectivity in reversed-phase liquid chromatography: 20 years of the hydrophobic subtraction model. LCGC Europe. 2020;33(4). https://www.chromatographyonline.com/view/selectivity-reversed-phase-liquid-chromatography-20-years-hydrophobic-subtraction-model-0 — Fs as a weighted distance and its weights (12.5, 100, 30, 143, 83).
  7. Dolan JW. Back to basics: the role of pH in retention and selectivity. LCGC North America. 2017;35(1). https://www.chromatographyonline.com/view/back-basics-role-ph-retention-and-selectivity — pH most effective within about 1–1.5 units of pKa; neutral compounds unaffected; buffer pKa within 1 unit; aniline example (α 1.64 to 1.05).
  8. Stoll DR, Rutan SC, Venkatramani CJ. Peak purity in liquid chromatography, part I: basic concepts, commercial software, and limitations. LCGC North America. 2018;36(2). https://www.chromatographyonline.com/view/peak-purity-liquid-chromatography-part-i-basic-concepts-commercial-software-and-limitations — what purity software tests; similar spectra and low-level impurities; false-negative example; DAD with MS and complementary columns.

Further reading

  • Snyder LR, Kirkland JJ, Dolan JW. Introduction to Modern Liquid Chromatography. 3rd ed. Wiley.
  • Neue UD. HPLC Columns: Theory, Technology, and Practice. Wiley-VCH.

Reviewed against cited sources. Specificity statements follow ICH Q2(R2); column-selectivity statements follow the cited LCGC articles by the hydrophobic-subtraction model’s authors; case-study figures are as reported by Zheng and Rasmussen. For validated or compendial methods, the applicable procedure and regulatory framework take precedence over the general rules of thumb given here. Evidence review: October 2026.

Leave a Comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Scroll to Top