PEAK PURITY Determination by HPLC Diode Array Chromatography Software (UV/VIS): Limitations and Uses

"Peak Purity" software determination by HPLC UV/VIS detection is one of the most abused and easily misunderstood features found in advanced liquid chromatography systems (e.g. HPLC, UHPLC and CE).

For HPLC, one or more inline detectors can be used which provide additional data about a fully resolved peak’s physical or chemical properties. The data obtained can be compared to that of a pure standard, or known impurity. For compounds which absorb light in the region of most UV/VIS detectors (~ 200 to 900 nm), a single wavelength detector (e.g. UV/VIS) provides a very limited second dimension of data (retention time is the first dimension), but a scanning, multi-wavelength UV/VIS detector can add a second and third dimension of data to the retention time. Scanning detectors, commonly known as Diode-Array Detectors (aka: DAD or PDA) are commonly used in HPLC and CE analysis (they are required for routine method development). A scanning DAD can provide detailed sample UV/VIS spectra across a range of wavelengths for each peak, at any retention time recorded, allowing for a 3D plot of the spectra to be recorded much like a “fingerprint”. "Pure" compounds which do absorb light across a pre-defined wavelength range should show identical spectral profiles (“slices”) across the upslope, apex and down slope of the resolved peak. "Impure" peaks may show dissimilar spectra across the width of the peak revealing the presence of a co-eluting peak or impurity. Impure peaks may also NOT show any dissimilar spectra at all (because some compounds may not be detected). When a properly developed HPLC analysis method is used to evaluate the purity of a sample, the single dimension of “retention time” is evaluated with additional dimensions of analysis such as the UV/VIS peak spectra. "Peak Purity" relies on the detection of a sample's spectral profile to detect the presence of an "impurity" (that may have co-eluted with the sample). This additional dimension of analysis (full Spectra) is required to improve the confidence level that a peak may in fact be correctly identified (qualitatively) and does not contain any co-eluting compounds. IOW: "Peak Purity" does not actually test for purity.

Diode-Array 'Software' based Peak purity determination by HPLC is a qualitative assessment of the impurity profile of the sample. It is designed to reveal impurities, NOT prove peak purity. BTW: We really should rename it “Peak Spectral Impurity Assessment" because that is in fact what we are measuring. The algorithm used for Peak Purity determination is designed to confirm the presence of one or more impurities by comparing spectral data slices (multiple slices taken at the apex and both the upslope and down slope sections of the peak).  A mismatch would indicate the peak has not been fully resolved (one or more co-eluting peaks are present). In other words, it is impure by UV/VIS analysis. Note: It does not indicate that the compound is impure, but rather 'the peak' being measured is. As you can see, the concept makes sense, but the how it is used in many laboratories is flawed leading to invalid reports and data.

• “Peak Purity” does not in fact indicate the actual purity of the compound, but instead indicates when a peak may be found to contain impurities. It is an estimated measure of PEAK Impurity.

In simple terms, IF the spectral slices obtained from one peak are not identical, than the peak may contain one or more impurities. Co-elution is the most likely reason for this.

Points to consider when using "Peak Purity" software:

• The absence of any spectral differences across the sample peak are not an indication of actual purity;
• Compounds similar to your sample may have similar absorbance profiles (fooling the system);
• The relative concentration of actual impurities may not be high enough to detect;
• The compounds / impurities may not absorb light at the wavelengths scanned;
• The HPLC method used, the software settings and the parameters that you chose in the ‘Peak Purity’ software menu have a huge effect on the results obtained. Different people often get different results for the same sample. Inputting poor quality settings or using a poor quality method often leads to misleading purity results. This is an advanced software feature requiring many years of training to use. Again, it does NOT test for purity.
  
• The peak of interest must be retained on the column (K prime > 2) and resolved apart from any observed peaks. Don't use peak purity to analyze peak(s) which elute at or near the column void volume (Low K prime values may demonstrate that good chromatography fundamentals were ignored. Poor quality methods fail validation). Poor quality HPLC method and poorly selected DAD "Purity" settings result in invalid results (audits, recalls etc may result from reliance on a subjective "software" feature).

We prefer to think of HPLC 'Peak Purity Assessment' as a null test. If the recorded peak spectral data slices are different, than you probably have co-elution and/or impurities present (so try and develop a better method to resolve the peaks apart). If no differences in the spectra are seen (they are similar), then the peak may be pure or may contain compounds with similar spectra as are commonly seen with related reaction synthesis products or compounds. So only when you detect differences in the acquired spectra can you be confident that there IS a qualitative difference or impurity present. You will not know what percentage of impurity level is (since you do not know what it is).

When configuring the Peak Purity parameters for your sample, you must start with a very high quality HPLC method (A "validated method" is not necessarily a high quality method. "Validation" does not in fact insure that the method follows good chromatography fundamantals). The correct detector sample rate, threshold, slope, signal wavelength and bandwidths need to have been properly selected and used (Reference Wavelength always OFF). The peaks shown in your chromatogram should have excellent symmetry with good on-column retention (K-prime, as applicable to mode), baseline separation (> 2.0 for non-SEC modes) and very low baseline noise levels. The two Peak Purity spectral reference points should be manually selected and placed at times before and after the peak of interest in clear baseline areas where no other peaks or spectra are seen (never use the instrument default settings for reference points!). Select at least 7 spectra from the sample peak for comparison (more detail can be provided with more spectra, but be careful not to select spectra near the baseline or the noise limits). If your method and chromatogram are not of the highest quality, then please do not use the automated "peak purity" analysis feature, instead spend time improving your method.

SUMMARY: 

The HPLC UV/VIS Peak Purity Analysis (“Peak Spectral Purity”) feature is very complex and has many software settings which must be set up correctly to obtain any scientifically useful data regarding possible peak impurity levels. 

• Do NOT use the system default settings / values for 'Peak Purity' ! They are just place holders for actual values (which you must calculate and fill in the correct values for your method).
  

 

* Due to a general lack of formal training, I often see this software feature being used incorrectly by most chromatographers. This is worth repeating... the HPLC method used to obtain the original data must be of the highest quality and the training of the operator must also be at the highest level. To use this advanced software feature successfully, an advanced understanding of the fundamentals of chromatography are required as are a detailed understanding of all of the peak purity software features (how to set the correct threshold, obtain reference baselines, Set sampling rate, noise levels, signal extraction, normalization settings…). Routine HPLC training classes do not cover these types of tasks. Years of specialized training and practical experience are required to use these tools. Never use the “automated” versions or the manufacturer’s default values to find “Peak Purity”. The only correct way to use these features is to manually tune the method and settings to your specific sample. Failure to customize the method and settings used may result in invalid data and incorrect "purity" determinations. 

Due to very complex software setup needed for "Peak Purity" determination by UV/VIS spectra, the requirement for a high quality HPLC method and a high quality data-set,it is our opinion that few should ever use it. In general, the recommendation for most chromatographers is to not use this feature unless first having demonstrated the required skills and advanced understanding of the fundamentals of chromatography. Most of the methods that we professionally review where "Peak Purity" data have been used as part of the method have been found to be based on invalid methods, resulting in any "purity statements" issued as unscientific and invalid. Please proceed cautiously and request professional review of any methods which employ it BEFORE committing to relying on it.

©Copyright, March 1, 1996 by William Letter of Chiralizer Services (Plainsboro, NJ) from a portion of material presented in an HPLC Diode Array Method Development Class.

Determining the Data Acquisition Rate (Sampling Rate) For Your HPLC Detector

Another common question I am asked is how to set-up the HPLC detector’s sampling rate. This article is specific to commonly used UV/VIS, not mass selective detectors (Mass Spectrometer detectors are set-up in a similar manner, but you also want to take into account the numbers of MRM transitions for each peak and dwell time to account for the scanning delay. Typical values for MS are >10 points with 15-20 being best). 

Most HPLC (UHPLC) instrument manufacturer’s provide default sampling rate values within their software packages. Please do not use them as the values shown were just put there to fill in the data field and may not apply to your application or method. Many chromatographer's use these values without first understanding if they are appropriate for their own methods. This is a common mistake. Just as the manufacturer does not know what wavelength, flow rate or mobile phase you will use, they also do not know what sample(s), method and/or conditions are appropriate for your specific application. As such, they provide numerous default values in these data entry fields to satisfy the software's requirement. Just as you select an appropriate wavelength and bandwidth, you should always calculate and enter the correct detector data acquisition rate value yourself which is appropriate for your specific application, detector type and method. 

The Peak shape's role during integration: For each chromatographic analysis you must determine the optimum sampling rate for the chosen detector. An accurate value is critical for proper instrument set-up, quantification and integration of your sample(s) peaks. In the most basic sense, the area under a perfectly Gaussian peak requires at least ten points to describe it with some detail. Ten points will provide basic data about the shape of an ideal peak to the computer. Since peaks are rarely perfectly symmetrical, a larger number of points will provide more accurate integration of the peak’s actual shape and total area. This will improve run-to-run reproducibility and quantification. We suggest you include twenty to thirty data points to allow for a more detailed fit to the peak. Too few points across a peak and you lose detail and sacrifice reproducibility. Too many points and you start to introduce noise into the system. 


With these facts in mind we can next think about calculating the detector’s data acquisition rate. You must select a data rate (sampling rate) that is sure to provide the recommended 20 to 30 data points across the peak width (we use the commonly calculated peak width at half height as the time measurement). Select a detector sampling rate that will provide you with this degree of detail and resolution. This is best accomplished by initially looking at an actual chromatogram of your sample. Look at the chromatogram and use the narrowest sample or standard peak past the void time, with good retention as an example to determine the best acquisition rate. The narrowest peak will be the worst-case scenario and will insure that you have enough points across all of the remaining peaks in the sample. It's width is often measured in units of time (seconds/minutes). This data can often be read directly off of a generated data acquisition report.

Examples:

(a) If your narrowest peak has a peak width of 1.00 minute (60 seconds), then divide 30 points into 60 seconds for a result of 2 seconds per data point. The preferred sampling rate would be 2 seconds, 0.03 minutes or 0.5 Hz (depending on the units used by your detector).

(b) If your narrowest peak has a peak width of 0.20 minutes (12 seconds), then divide 30 points into 12 seconds  for a result of 0.4 seconds per data point. This equals a sampling rate of 2.5 samples per second or 2.5 Hz.

Summary:  

     To Determine the Data Acquisition Rate For Your Detector You Need To:

• Calculate the best data rate for each method and not use a generalized value (though similar methods will often use the same rate).
• Use your existing sample integration data results to identify the narrowest chromatographic peak in your analysis (at the baseline or half-height).
• Record the width value of this peak (usually in units of time).
• Divide this number by thirty (30) to determine the preferred sampling rate.
• Use this value, or a value close to it, for your detector’s sampling rate.

UV / VIS, VWD, DAD, PDA HPLC DETECTOR SIGNAL BANDWIDTH (bw) SELECTION

Modern chromatography UV/VIS detectors offer the operator a choice of one to several hundred different signal wavelength choices (as is the case for Diode Array Detectors). Besides being able to specify a single wavelength, you can often choose a signal BANDWIDTH (bw) to associate with each wavelength [e.g. for a 280 nm signal with 10 nm bandwidth. This is often written as: 280 (10) or [280:10]. In many detectors, Signal Bandwidth is a variable, not fixed and represents the total number of nanometers across the specified signal value chosen. For example: If you select a signal wavelength of 280 nm and choose a bandwidth value of 10 nm, then you are actually gathering all signal data between 275 nm and 285 nm (5 nm to the left of the apex and 5 nm to the right for a total of 10 nm). Using a narrow bandwidth has the advantage of increasing the signal selectivity of the detector as you are only collecting data within a tight window. If you were to increase the bandwidth to 60 nm in the same example you would now be collecting data between 250 nm and 310 nm. The additional data collected over this wider range may reduce the total noise (by averaging it over a wide range), improve the S/N ratio (which may increase sensitivity), but it also reduces the selectivity. Large bandwidths also increase the chance you may include peak signal data from other co-eluting components into your signal data. You must select a bandwidth range for each signal wavelength which is located 'safely' away from any other potentially interfering peak. As with many things in life, balance is important. In this case, bandwidth choice is the balance between selectivity and sensitivity.

• When developing new methods we recommend that you choose an initial bandwidth value of 10 nm for each signal. This provides a nice balance between selectivity and sensitivity. It is also a common bandwidth value used on many older UV/VIS detectors which have a fixed signal bandwidth (such as many single or variable wavelength detectors).

• If you have determined the exact signal maximum for your sample and you would like to gain additional sensitivity for your sample (and thus decrease selectivity), re-run the analysis using several different, but increasing signal bandwidth values (e.g. 10, 20, 30, 50 and 100 nm). Choose bw values that are safely within the range of the detector, within the limits of the mobile phase's absorption region and also away from any potential co-eluting peaks. *To confirm which value is best, be sure and calculate the actual measured signal to noise ratio of the peak of interest after each analysis. This is a critical step! Do not be fooled by increases in the peak height or area alone as these changes are not always synonymous with better signal to noise ratios. Only by measuring the actual baseline noise level for each run and comparing it with the actual peak signal obtained will you be able to determine if increasing the bandwidth has provided you with better noise reduction and signal strength.

• To increase spectral signal selectivity choose a bw value that is very narrow. A value such as 2 or 4 nm would allow the detector to collect only signal data that is at or near the apex of your selected wavelength. This can be very useful when trying to discriminate your signal from nearby signal peaks, especially at low wavelengths such as 210 nm.

• When reporting your method conditions always include the wavelength AND bandwidth used for each signal. In order to accurately reproduce your method, this information is needed. *The flow cell dimensions, wavelength and bandwidth should always be included in your method.