Introduction to Modern Liquid Chromatography, Third Edition part 89 pdf

A 1◦C increase in column temperature will normally decrease retention by 1–2% Section 2.3.2.2, so a method that is operated without column-temperature control will be subject to changes

17.4.3.5 Temperature Problems

Changes in column temperature affect values of t R and k A 1◦C increase in column temperature will normally decrease retention by 1–2% (Section 2.3.2.2), so a method that is operated without column-temperature control will be subject to changes in retention as the temperature of the laboratory changes during the day Temperature changes also can influence selectivity (Section 6.3.2), so shifts in relative retention may also be observed Many laboratories have stable daytime temperatures, but for energy conservation do not provide the same quality of temperature control at night Also, even though the laboratory temperature is relatively constant (as measured

at a wall-mounted thermostat), the local temperature can fluctuate significantly, especially if a heating duct directs air at or near the HPLC system For this reason problems related to temperature tend to be exhibited as cyclic changes in retention throughout the day Temperature-related retention problems can be corrected by using a column oven operated in a range where it has stable temperature control (Section 3.7) Inadequate column temperature control also can cause peak shape problems, as described in Section 17.4.5.3

17.4.3.6 Retention-Problem Symptoms

This section discusses retention-time problems in terms of symptoms; see the related items in Table 17.6

Abrupt changes in retention are usually easy to isolate If these occur when

the column is changed, the column itself is the most likely cause Re-installation

of the previous column should confirm this Column-to-column variation is much less common with today’s high-purity, type-B silica columns, but was commonplace with the lower-purity, type-A columns that may still be in use for some legacy methods Legacy methods may require adjustment of the mobile phase with each new column in order to meet system suitability; an alternative is to order several columns from the same batch of packing material Redevelopment of the method for a more robust separation is another solution, but it may not be economically feasible Substitution of an equivalent column (Sections 5.4.2, 6.3.6.1) that is more reproducible is another option Also, don’t overlook the possibility that the wrong column was inadvertently installed

If the change in retention occurred when a new mobile phase was formulated, the simplest solution is to make another batch of the mobile phase Be sure that the correct mobile-phase pH is used (Section 7.2.1), and that the pH is adjusted prior to the addition of organic solvent

Abrupt changes in retention are fairly common when a gradient method is transferred from one HPLC system to another This usually is due to differences in the system dwell-volume between different equipment (Section 9.2.2.4) Sometimes these differences can be compensated by a change in mobile-phase conditions, the injection timing, or modification of the system plumbing (Section 9.3.8.2; also Section 5.2.1 of [18])

If retention changes abruptly when none of the above conditions exist, and there is no obvious change in the system operating conditions, it is likely that there

is an equipment problem (e.g., check-valve failure), a leak (Table 17.3), a bubble (Table 17.4), or a column-temperature problem (Section 17.4.3.5)

Drifting retention times are a symptom of some instability in the system When

a method is set up, it is not uncommon for retention times to drift for the first few injections; this may be even more pronounced when a new column is installed The most likely cause of retention-time drift for RPC is incomplete equilibration of the mobile phase and column Incomplete equilibration can be especially pronounced for ion-pair separations, where 20 to 50 column volumes may be required for equilibration (Section 7.4) For most isocratic methods, however, retention times should stabilize after the first two or three injections For gradient elution, an increase

of the equilibration time between runs may be required to stabilize retention times,

especially if the first few peaks in the run are eluted close to t0(Section 9.3.7)

A less-common cause of retention-time drift is the presence of slowly equili-brating active sites on the column that become saturated after several injections When this is the case, several ‘‘priming’’ injections to deactivate the column (Sections 3.10.2.2, 13.3.1.4) may solve the problem Make several large-mass injections of the sample in a row (it usually is not necessary to make a complete run for each injection, just inject several times with perhaps a 30-second delay between injections), then allow the normal method cycle to run Sometimes priming injections are required just once for a column, whereas other samples may require priming injections each time the method is started

If retention time drifts in a continuous fashion over an entire sample batch,

it suggests that something is continuously changing in the method; for example, the mobile phase may be unstable The use of a volatile buffer (e.g., ammonium carbonate) coupled with helium sparging can result in evaporation of the buffer with

a change in mobile-phase pH Similarly loss of the organic component of the mobile phase can occur, but this is uncommon during the course of a day Re-formulation

of the mobile phase on a daily basis may be necessary for some methods If helium sparging is used (Section 3.3.2), note that it takes only one volume of helium to degas an equal volume of mobile phase (e.g., 1-L of He for 1-L of mobile phase), so

a few minutes of vigorous sparging is all that is needed If continuous sparging is necessary for pump or detector stability, turn down the helium supply to a trickle rather than allow vigorous sparging to continue If the presence of a small amount

of dissolved air is not a problem, in-line vacuum degassing (Section 3.3.3) usually

is more convenient and is adequately effective in most cases—without causing mobile-phase evaporation

Variable retention times for some or all peaks between chromatograms are

symptoms that some variable is not adequately controlled In one example where retention-time variation was observed only in the middle of gradient runs, the cause was related to a mobile-phase proportioning problem (see Section 5.5.4.1 of [18]) An intermittent check-valve failure will cause intermittent flow-rate, and thus retention changes Temperature fluctuations in the laboratory can change retention

on a run-to-run basis Usually the causes and fixes for variable retention times are similar to those for drifting retention

When retention times have decreased, several possible causes exist If

retention-time loss correlates with larger injected sample-mass and right-triangle

peak shapes (e.g., Fig 17.15a), mass overload of the column is likely Reduction of

the injected sample weight should correct this problem See the discussion of tailing and distorted peaks in Section 17.4.5.3 for more information on mass overload

When all peaks in the chromatogram show reduced retention, the problem

is associated with the column, mobile-phase, temperature, or flow rate Consult Table 17.5 and the appropriate discussion in Sections 17.4.3.1 through 17.4.3.5 for more information

When only some peaks in the run have shorter-than-normal retention times,

an unexpected change in the system chemistry is suggested; for example, a change in ionization of acidic or basic solutes Check the mobile-phase pH (prior to addition

of organic) Usually a change in the %B will affect all peaks in the run (though not necessarily in an identical way); if this is suspected, make a new batch of mobile phase Note also that the accuracy of on-line mixing of the mobile phase can vary among different HPLC systems An aging column can also affect the retention of just some peaks in the chromatogram; installation of a new column will serve to identify the column as the problem source

Inadequate retention of polar samples is sometimes a problem during RPC method development If the sample is ionic, it may be possible to change the mobile-phase pH so that the sample is converted to its non-ionized form, which will

be less polar and better retained (Section 7.3) An alternative is to use ion pairing to improve the retention of ionic samples (Section 7.4) If the sample is neutral, use of

a more polar mobile phase (less strong solvent) should increase retention However,

if the %-organic is ≤5%, column de-wetting may occur for alkyl-silica columns (Section 5.4.4.2), with resultant loss of retention Use of a column containing embedded polar groups or ‘‘AQ’’ type columns may be useful If other attempts

to retain polar compounds by RPC are not successful, a change to normal phase (Chapter 8) and especially hydrophilic interaction chromatography (HILIC, Section 8.6) may provide the desired results See the additional discussion regarding poor retention of polar solutes in Section 6.6.1

Retention times that are too long usually have similar causes as those that

are too short Refer to Table 17.5, Sections 17.4.3.1 through 17.4.3.5, and the discussion of smaller than expected retention

17.4.4 Peak Area

With today’s data systems, quantification by peak area is much more common than

by peak height (Section 11.2.3), so we will assume peak-area measurements for the current discussion; however, the same troubleshooting process can be used for either peak-height or area problems If a change in retention accompanies a peak-area problem, first correct the retention problem before addressing the peak-size problem Peak-area response for most methods will be very consistent over time For

example, repetitive injections of the same, well-retained sample (e.g., k >2) with UV

detection and a signal-to-noise ratio of S /N >100, peak area should vary <1%

between runs (Section 3.10.1.3) However, smaller peaks, shorter retention times, and/or the use of some other detectors may generate less reproducible results The following discussion of peak-area related problems is organized by (1) peaks that are larger than expected (Section 17.4.4.1), including peaks in blanks and carryover, (2) smaller than expected peaks (Section 17.4.4.2), and (3) peak areas that are variable from run to run (Section 17.4.4.3) A summary of symptoms and solutions is listed

in Table 17.7 In this section, we will assume that the method had been working properly for previous sample batches

17.4.4.1 Peak Area Too Large

For peak areas that are too large, the first step is to determine if the problem

is reproducible, and if it is related to just one sample or solute, or all samples Answers to these questions usually will require re-injecting one or more samples and/or examining several chromatograms from a batch of samples If the area is not reproducible between several injections of the same sample, see Section 17.4.4.3 (variable areas) If the sizes of all peaks vary in the same proportion, check to be sure that the correct injection volume is selected Another possible cause is faulty sample preparation—check to be sure that the dilution or concentration steps were done properly If the areas for different peaks in the chromatogram have changed

by different proportions, the detector settings may be at fault Check the detector wavelength (UV detector, Section 4.4), interface adjustments (evaporative detectors, Sections 4.12–4.14), time constant, and so forth

Peaks that appear in a blank injection generally come from one of two sources: late elution or carryover A peak that is not fully eluted in one run can appear in the next (or later) run; if the sample contains other components, the extra peak will be much broader than the neighboring peaks This is illustrated in Figure 17.5, where in

a a broad peak X (arrow) appears at approximately 2 minutes in the chromatogram.

In Figure 17.5b, the run of Figure 17.5a is extended, showing peak X both in the

previous run (at≈2 min) and at its normal place in the chromatogram (≈7 min) If

peak X must be quantified in the run, the run can be extended as in Figure 17.5b to

include the peak in the correct run If the peak is not of interest, several options are

available The run can be extended as in Figure 17.5b, the run time can be adjusted

so that the peak appears in the following chromatogram in a region where no other peaks are present, a step-gradient can be used to flush the peak from the column,

or sample cleanup can be modified to remove the peak from the sample prior to injection Carryover results when a small portion of the sample is trapped in or adsorbed on the surfaces of the autosampler and shows up when a blank is injected Check for carryover as described in Section 17.2.5.10

(a)

(b)

X

10

Time (min)

Figure17.5 Example of late elution (a) Broad peak (X) appears out of place in chro-matogram; (b) entire chrochro-matogram; extended run time allows peak to elute in proper position

in chromatogram (≈7 min)

17.4.4.2 Peak Area Too Small

Peak areas that are smaller than expected can have the same root cause as peak areas that are too large, and the process discussed above (Section 17.4.4.1) can be followed to isolate and identify problems due to small peaks Of course, carryover and late-elution problems are less applicable for peaks that are too small Other less common causes of small peaks are a detector time constant that is too large (Section 4.2.3.1), a data sampling rate that is too slow (Section 11.2.1.1), peaks that are off scale (underintegrated), or peaks that are improperly integrated (Section 11.2.1.4)

17.4.4.3 Peak Area Too Variable

If the precision of a method is worse than it has been historically, this will appear

as peak areas (or heights) that are more variable than expected If there also is a retention-time problem, it is best to correct it first (Section 17.4.3) There are many possible causes of variability in peak areas, some of which are also discussed in Section 11.2.4 Nearly any step in sample preparation and analysis can contribute

to peak-area variation Some of the more likely sources are discussed below The first step is to determine if the results from a single sample are consistent

If replicate injections of the same sample give consistent peak areas, all the processes from sample injection onward are working properly The source of the problem then has to be something prior to placing the sample in its vial Possible problems

of this kind include sampling, equipment, and sample preparation errors Sampling

is the process of selecting a representative (in this case, equivalent) sample (Section 16.3)— if the master sample is not homogeneous, subsamples may not be equivalent

If volumetric or gravimetric laboratory equipment is not accurate or operating

properly, error can be introduced, a common source of such error is a pneumatic pipette that is worn beyond acceptable tolerances The typical sample-preparation process (Chapter 16) has multiple steps in each of which small errors are possible that can affect analyte recovery (e.g., filtration, evaporation, dilution) In a stepwise manner modify the sample preparation process or circumvent specific steps to isolate the source of the problem

If replicate injections of the same sample give inconsistent peak areas, the problem is likely due to the processes that take place from sample injection onward The most likely sources are the autosampler, pump, detector, or data-processing

steps First check the autosampler by rerunning the reproducibility test of Section

3.10.1 to see how it compares to past tests (Section 17.2.4); make any necessary

repairs Pump malfunction can lead to a change in mobile-phase flow rate, another

possible source of peak-area variation (check this by running a flow-rate test,

Section 3.10.1.3) Detection problems, such as detector overload or poor wavelength

selection might affect one peak and not another If detector overload is suspected (very large peaks, e.g., >1 AU for a UV detector), dilute the sample or inject a

smaller volume to see if smaller peaks give more consistent areas For LC-MS detectors with an electrospray interface (Section 4.14.1.1), a poorly performing spray tip can result in different amounts of sample getting into the MS at different

times in the chromatogram The integration and data workup process might have

problems, such as if a peak had a start or stop time improperly set, or the data sampling rate was too slow (Section 11.2.1) Another occasional case of variable peak area can occur if a frozen sample is not properly thawed and/or mixed prior

to injection A gradient of analyte concentrations may then occur from the top to the bottom of a vial In this case replicate injections from such a sample may show

a descending or ascending (depending on the nature of analyte and matrix) series of peak areas

17.4.5 Other Problems Associated with the Chromatogram

In addition to the symptoms discussed in the preceding sections, chromatograms often exhibit obvious defects in appearance which can be used to isolate the cause

of the problem This section covers three of these:

• baseline drift

• baseline noise

• peak shape

17.4.5.1 Baseline Drift Problems

Baseline drift is defined as a continuous rise and/or fall of the chromatographic baseline extending over a period of tens of minutes to hours (Section 4.2.3.1) Drift can occur in a rising, falling, or cycling pattern, as well as exhibit other characteristics Some of the symptoms and causes of drift are summarized in Table 17.8 It should

be noted that some drift is expected;, for example, one UV detector specifies drift

of≤2 × 10−4 AU/hr at 250 nm at constant room temperature and with air in the cell and≤3 × 10−4 AU/hr with a room temperature fluctuation of≤2◦C [19]

Periodic drift is characterized by a cyclic pattern, with the baseline rising and

then falling (or vice versa) over one or more runs This is most common with gradient elution within a single run, as a result of a mismatch of the detector response to the mobile phase A- and B-solvents This is illustrated in the baselines of Figure 17.6

[20] Baseline (Fig.17.6a) is for a gradient run from 5–80% water/MeOH at 215 nm,

with drift of≈0.9 AU (because MeOH has much stronger absorbance than water

at this wavelength; see data of Table I.2, Appendix I) Such drift is normal and

0.0

0.1

(b,c)

1.0

(a)

time (min)

(b) 215 nm

(c) 254 nm (a) 215 nm

Figure17.6 Baselines obtained using water-methanol or phosphate-methanol gradients,

5–80% B in 10 minutes (a) Gradient at 215 nm and 1.0 AU full-scale; solvent A: water; sol-vent B: methanol; (b) same as (a), except solsol-vent A: 10 mM potassium phosphate (pH-2.8) and 0.1 AU full-scale; (c) same as (a), except 254 nm and 0.1 AU full-scale Adapted from [20].

is a problem only if it precludes accurate integration of the chromatogram If the drift is unacceptable, there are three general approaches for addressing the problem One option is to add a UV-absorbing reagent to the A-solvent In the example of

Figure 17.6b, the use of 10-mM phosphate buffer (pH-2.8) instead of water reduced the drift of Figure 17.6a by nearly 30-fold Because drift will be less severe at longer

wavelengths, another option is to increase the detection wavelength, provided that the sample response is acceptable at the new wavelength (UV detection is assumed; other detectors may offer other options) The effect of a wavelength change is seen

by comparing Figure 17.6a (215 nm) with Figure 17.6c (254 nm) Alternatively,

a less-absorbing organic solvent might be chosen In this case ACN could be used instead of MeOH (not shown); ACN has negligible drift at 215 nm and may be used successfully for gradients at 200 nm or above Of course, a change in mobile-phase

A or B can change the chromatographic selectivity, so further adjustments in the method may be necessary (only applicable for method development)

Negative baseline drift can be a greater problem because data systems typically

stop integrating when the detector reads less than −0.1 AU (−10% drift) Thus,

if the gradient-elution baseline of Figure 17.7a [20] is encountered, it is likely that

the baseline will drop off scale in a negative direction, with loss of the data (it was possible to collect this baseline only by turning off the auto-zero function and manually setting the baseline start at +1 AU) As in Figure 17.6a, c, the drift of Figure 17.7 is much less at 254 nm (Fig 17.6c) than 215 nm (Fig 17.6a) The negative drift of Figure 17.7a could be converted into a (more acceptable) positive drift by adding a UV-absorbing buffer to the B-solvent (Fig 17.8a [20]) Another

possible fix with some data systems is to adjust the scale of the data channel to a range of 0.0 to−1.0 AU.

In some cases, however, the use of mobile-phase additives as in Figure 17.8a cannot correct severe, negative drift In the example of Figure 17.9a, the baseline

for this ammonium bicarbonate-methanol gradient exhibits a negative dip in the

−0.5

−1.0 0.0

time (min)

(a) 215 nm

(b) 254 nm

Figure17.7 Baselines obtained using ammonium acetate-methanol gradients Solvent A: 25-mM ammonium acetate (pH-4); solvent B: 80% methanol in water; gradient: 5–100%

B in 40 minutes (a)215-nm detection; (b)254 nm Adapted from [20].

40 0

0.2

0.1

0.0

−0.1

time (min)

(b) 254 nm

(a) 215 nm

30

Figure17.8 Baselines obtained using equimolar ammonium acetate-methanol gradients as in Figure 17.8, but with buffer added to B-solvent Solvent A: 25-mM ammonium acetate (pH-4)

in 5% methanol; solvent B: 25-mM ammonium acetate in 80% methanol; gradient 0–100% B

in 40 minutes (a) 215-nm detection; (b)254 nm Adapted from [20].

−0.2

0.0

−0.1 0.1

time (min)

(a) 215 nm

(b) 254 nm

Figure17.9 Baselines obtained using ammonium bicarbonate-methanol gradients Solvent A: 50-mM ammonium bicarbonate (pH-9); solvent B: methanol; gradient: 5–60% B in 10

minutes (a) 215-nm detection; (b)254 nm Adapted from [20].

middle at 215 nm Adjustment of the absorbance of either the A- or B-solvent cannot solve this problem Although this mobile phase is unacceptable for detection

at 215 nm (Fig 17.9a), detection at 254 nm (Fig 17.9b) poses no problem An

alternative detector might also be used; for example, bicarbonate mobile phases are commonly used with LC-MS, without creating baseline problems Fluorescence detection is another option used to obtain flat baselines for gradient elution of fluorescent analytes

A change in temperature of the column (and mobile phase) is another major

cause of periodic baseline drift A change in mobile-phase temperature changes the refractive index of the mobile phase and the transmission of light through the UV-detector cell If the column is operated without adequate temperature control (Section 3.7.1), the baseline is likely to drift as the laboratory temperature changes Temperature-related baseline drift can be confirmed by related changes

in retention times with temperature See Section 17.4.3.5 for further discussion of temperature-related problems

Other types of isocratic baseline drift are not cyclic, and these may arise from

different causes Slow system equilibration after a change of conditions (mobile

phase, column, column temperature, flow rate, etc.) will result in initial baseline drift that usually subsides within 30 to 60 minutes Baseline drift associated with equilibration may be accompanied by retention-time drift Similarly, when a detector

is first turned on, the detector response may drift for a few minutes or even hours as the lamp, electrodes, or other detector elements warm up and stabilize

17.4.5.2 Baseline Noise Problems

Disturbances in the baseline are referred to as baseline noise The characteristics

of baseline noise can help identify its source Baseline disturbances can be periodic

or random, and the duration of the disturbances can be shorter (short–term noise)

or longer (long-term noise) than the width of a chromatographic peak Moreover, baseline noise is superimposed upon any baseline drift In addition to the discussion below, consult Table 17.9 as well as Sections 3.3.1 (degassing), 3.8.3 (data rates), 4.2.3 (noise), 11.2.1.1 (data sampling contributions), 11.2.4.2 (chromatographic sources), and 11.2.4.3 (detection sources)

High-frequency short-term noise shows up as the ‘‘buzz’’ on the baseline (e.g.,

Fig 4.5) resulting from electronic noise on the electrical circuits This has a period

of 60 Hz (North America) or 50 Hz (most of the rest of the world), depending

on the frequency of the alternating-current electrical supply High-frequency noise usually can be significantly reduced as discussed in Section 4.2.3.1 by the use of

a cleaner electrical supply (e.g., use an uninterruptable power supply, UPS) and/or selection of a larger detector time-constant Figure 4.5 shows the reduction of noise

by approximately 300-fold by the use of a simple noise filter

Random and low-frequency short-term noise can result from several different

sources Insufficient degassing can lead to the introduction of air bubbles into

the HPLC system Bubbles trapped in the pump head(s) can also cause baseline disturbances as the pressure fluctuates from one piston stroke to the next, giving a regular pattern to the baseline noise Bubbles in the pump should be accompanied by pressure fluctuations as described in Section 17.4.2.3 Bubbles that make it through the pump, or that are formed after the pump by mixing inadequately degassed mobile phase in high-pressure-mixing systems, often will be kept in solution due to the system pressure However, when the dissolved air leaves the column, the pressure

is greatly reduced and the bubbles may reform As the bubbles pass through the detector, random, sharp spikes may appear, especially with optical detectors (e.g., UV-visible, Section 4.4; fluorescence, Section 4.5; refractive index, Section 4.11) Detectors that evaporate the mobile phase (e.g., Sections 4.12–4.14) are, of course, not susceptible to mobile-phase bubble problems If the bubble is trapped in the flow

cell, a large shift in baseline may result Adding a back-pressure restrictor after the detector (Section 4.2.1) may solve bubble problems in optical detectors

Electrical spikes are similar to bubbles But to distinguish their presence from

bubbles, turn off the pump flow and monitor the baseline If the spiking continues, the problem is electronic; if the spiking stops and the baseline remains steady, the problem is due to a bubble The use of better degassing procedures (Section 3.3.1)

is the first line of defense against bubbles A back-pressure restrictor (Section 4.2.1) will keep bubbles in solution until after they leave the detector

The selection of a data collection rate that is too fast can result in excessive

short-term baseline noise As described in Section 3.8.3, the data rate should be set to collect≈20 points across the peak Higher data rates will increase the baseline noise while having little benefit on the amount of signal collected, so the signal-to-noise ratio (Section 4.2.3) will worsen Lower data rates may reduce baseline noise, but this risks reducing the signal as well, so the signal-to-noise ratio may suffer

Long-term noise shows up as baseline disturbances that are comparable in size

(or wider) to normal peaks One common source of long-term noise is the presence

of late-eluted materials in the sample (see the discussion of Fig 17.5 in Section

17.4.4.1) As retention time increases for solutes or background interferences in the sample, the band width increases and the peak height decreases Late-eluting peaks from prior separations can accumulate over time, resulting in a drifting and erratic baseline A strong-solvent flush of the column (e.g., 25 mL of methanol

or acetonitrile) often will remove strongly retained material from the column For this reason a strong-solvent flush is recommended following each batch of samples (isocratic separation assumed) For some methods a column flush may be needed more often Gradient methods usually are less susceptible to late-eluted interferences because they have a strong-solvent column-wash built into every run Heroic efforts

to remove strongly retained materials (e.g., flushing with acid, base, chaotropes,

or methylene chloride) can be effective but can also damage the column A better approach is to use improved sample pretreatment (Chapter 16) to reduce the sample burden of late-eluted materials Remember, the column is a consumable item Once

500 or so samples are analyzed, the cost per sample for the column becomes a trivial portion of the overall analysis cost, so column replacement often is a better choice than extensive column cleaning or sample pre-treatment

Sometimes long-term noise shows up as regular baseline fluctuations, as in

Figure 17.10 (note that the y-axis is 1 mAU full scale) Usually cyclic baseline

disturbances are caused by pump problems and will be accompanied by pressure

Time (min)

Figure17.10 Cyclic baseline noise that was attributed to interference from an electronic air filter in the laboratory Adapted from [21]