HPLC Peak Splitting. Common Reasons For It

True "Split" HPLC peaks, not resulting from co-elution of another peak, can be caused by a number of chromatography problems. Here are a few examples and their solutions:

1. Sample overload. Sample overloading is one of the most common reasons for observing peak "splitting". Reduce the sample concentration by factors of ten to see if the peak shape improves. 
2.  A poor quality HPLC method. Poor quality methods which do not use mobile phase solutions which are at an appropriate pH (*If the pH of the mobile phase is close to the pKa of the sample, then split peaks may result); which does not dissolve the sample in (should be fully soluble) or are unstable, show sample or mobile phase precipitation can cause this effect. Always check solubility before starting.
3. A partially plugged or fouled column. A dirty or fouled column (from not washing down properly with a solution which is STRONGER than the mobile phase). Analysis methods should be followed by separate wash methods to remove all bound material and any late eluters,
4. Wrong injection solution. Peak splitting may be the result of dissolving and injecting your sample in a solution that is stronger than your mobile phase. Dissolve and inject samples in the mobile phase or in a solution which is a slightly weaker solution (not stronger).
5. A poorly packed column, void at column inlet, a dirty frit or poor mechanical connection (i.e. improperly swaged fitting). These types of structural or mechanical defects can each result in peak "splitting" (all of these are less common today than in the past using modern HPLC columns). When present, a dirty inlet frit can be replaced with a new one, or the column can sometimes be backflushed to remove any accumulated material. Connections should always be double checked.
6. Detector data rate set too low. Too few peaks collected over time may result in integration errors and inaccurate peak symmetry problems. Read more about how to determine the best data collection rate at this link.

Modern HPLC Method Development Tips (PART II):

This is the second of two articles (Part I) which will provide suggestions on how to improve the HPLC (UHPLC) method development process. - PART II

INITIAL METHOD DEVELOPMENT:

1. Before you start, learn what you can about your sample, its hazards, solubility and properties. Conduct a quick literature and/or keyword search on the web using a popular search engine (e.g. GOOGLE). You can often find many journal articles, white papers, application notes and chemical data on the web in just a few minutes.
2. Determine which liquids your compound(s) are soluble in. Use a pipette and several glass vials or tubes with different solutions (pH is important too). This will narrow down the types of mobile phase and chromatography modes that you can use.
3. Column choice is the most important part of method development. The stationary phase that you choose has the single greatest effect on selectivity! Select the right column and mode of chromatography. Most RP methods should start with at least a modern ultra high purity, metal free type B silica column with a C8 or C18 support (*But you must select the column type that best suits your application). For small molecules (<1,000 Daltons) select a standard sized analytical column with a 2.5 to 5.0 micron particle size and pore size between 60 and 120 Å (e.g. 4.6 x 150 mm; 3.0 x 100 mm). *For larger peptide or protein molecules you will need a particle with a large pore size (~ 300 Å). For optimal results, columns with very small internal volumes should be paired with HPLC systems with similar ultra-low internal delay volumes. Your HPLC system should be optimized to balance the needs of good mixing, low delay volume and proper sampling rate for your method. If the method is likely to be transferred to different types of HPLC systems, then you may want to initially stay away from the < 2.1 u particle supports for your method development. At this time, these tiny particles can not be reliably packed into columns as well as the larger sized particles and generally have much poorer %RSD values than larger particles (i.e. 2.5 to 5 u). Method development is initially easier & generally more rugged using conventional particle sizes [Keep things simple when starting out. You can always change later on]. Once you have selected a column and decided on a flow rate, make sure you calculate and measure the actual column void volume to find the 'Tzero value' (column dwell volume). You will need this value to find out if your compound has been retained on the column (K prime) and to also determine most of the needed performance calculations and parameters.
4. Once you have selected an HPLC column (if possible, please start with a brand new column) you will need to test it and establish a baseline to show that it meets both the manufacturer’s specifications and, far more importantly, that it meets your method’s requirements. In most cases, do not use the manufacturer’s QC test solution for this. Those stds are designed to allow the column to easily pass manufacturing QC, not the more critical requirements that you may need to prove. We prefer to use a real sample mixture (~2 compounds) that are similar to the type proposed for the method. They must be well retained on the column (K prime of > 2), easy to prepare, stable and reliable. This std and the specific method you develop for it, will be used to prove the column’s performance (i.e. R, S, K prime, Tzero, Plates). It will be used again, when the column’s performance is called into question.  
5. Notes on Mobile phase and additives: Keep it simple. Avoid the use of any additives such as ion-pairing reagents when first starting out. They are overused in general and can cause problems later on. If required, they can always be added later on. Use only HPLC grade solvents, fresh RO HPLC grade Water and the highest purity acids, bases and/or additives, if required. Use only filtered (0.22u) products. Always dissolve samples in the mobile phase (or a weaker solution) for injection. For many RP methods a low pH mobile phase (~ pH 2.5) provides a good starting place. Samples with a pKa far enough away from this value are likely to be retained, stable and unaffected by small changes in pH. *pH is usually one of the last parameters which is optimized.
6. Use a Gradient Method to find the approximate elution conditions of the mixture AND make sure you are detecting all of the compounds injected onto the column [For dedicated RP isocratic methods, start with a high organic mobile phase % (i.e. 95%) and record the results. Continue to reduce the organic content in steps of 10% and observe where the best compromise exists between retention and elution]. For RP gradient methods, start with a very high aqueous % (e.g. 98%) and run your gradient, slowly, to a very high organic % (e.g. 95% or 98%, not 100%!) to make sure you retain, hold, and then elute everything. Your mobile phase conditions must be strong enough to elute everything off the column during the gradient portion of the run (long enough hold time). Please Do NOT include the gradient reversal portion back to initial conditions (wash and re-equilibration) of the gradient as part of the same analysis method. The method should end after the gradient has reached and been held at its maximum level for a period of time (we refer to this as the "hold time"). It should NOT switch back to the initial conditions to re-equilibrate the column. Don't make this novice mistake. Column flushing and re-equilibration should be a separate method and/or step, separate from your analysis method. If you include your column flushing (washing) and re-equilibration steps as part of your analysis method, you are also forcing the baseline's slope to change radically. This may interfere with integration plus include extra peaks and baseline changes that you are going to have to integrate, identify and explain to others (e.g. auditors). Additionally, including these wash steps as part of the same method wastes time as you must wait for these steps to complete before you can start to process the data obtained in the analysis method. Summary: Develop methods like the professionals do and create separate Analysis and Flush/Re-Equilibration Methods (or steps).
7. If your compounds can be seen with a UV/VIS detector, then make sure you are using a Diode Array Detector (aka, DAD or PDA) set to scan a wide range of wavelengths (e.g. 210 - 410 nm). Method development should not be performed using a single or multi-wavelength detector. This invites errors, limits the utility of the method and does not result in any cost savings. You must have a full scanning detector so you can detect all possible peaks at the same time. If you use a single or dual wavelength detector you may not know if an impurity has been introduced or if you have a co-eluter (because you will have no way to detect it). Generic starting settings for Wavelength Bandwidth should be ~ 8 nm, software based Reference Wavelength OFF and the sampling rate must be set to collect at least 20+ peaks/sample peak of the narrowest sample peak observed and integrated (at ½ height). Run with scanning turned 'on' for all analysis methods and review the spectral data for each run. This provides important qualitative data about the compounds which may be used for purity determination and also to demonstrate how the method is selective for the compound(s) of interest. *If your compounds do not absorb well (weak chromophores), then you may still want to have a UV/VIS detector inline (first) with a secondary detector second. The UV/VIS detector will still be very useful for troubleshooting and detecting other compounds. Select an appropriate type of detector and compatible mobile phase as required for the secondary detector (e.g. RID, EC, FLD, ELSD, CAD, MS...). Note: If available, LC-DAD-MS is one of the most useful instrument setups for LC method development.
8. This needs to be repeated (from Part I)....  Before starting ANY HPLC analysis, the HPLC pump must be running and operating with no problems, achieving a stable baseline, steady flow rate with as little pulsation as possible (~ 1% ripple or less). Accuracy depends on this. Do not begin any HPLC analysis unless the HPLC pumping system is working perfectly.