HPLC Back to Basics

A Laboratory Companion for Liquid Chromatographers #1

• Retention Factor
• Column Dead-time
• Retention Tme Diagnostics
• Selectivity
• Efficiency
• Tailing
• Resolution
• Fundamental Resolution Equation
Retention Factor
The retention factor is a measure of the distribution of the half a unit, which means that an estimate is quite adequate.
sample between the mobile phase and the stationary phase. The The retention factor is estimated simply as follows. Note that
calculation is a simple one, as shown in Figure 1. tR is the retention the numerator of the equation (tR – t0) tells us to subtract t0 from
time and t0 is the The calculation is a simple one, as shown in the retention time. Or simply throw away everything before t0
Figure 1. tR is the retention time and t0 is the column dead-time. and start measuring from t0. The denominator (t0) says to use t0
Retention is measured from the time the sample is injected to as our unit of measure, instead of time or distance. So we just
the highest point on the peak. Measurement of the column mark off the baseline in units of t0, starting at t0, as shown at the
dead-time is most easily measured as the first disturbance in the bottom of Figure 1. When we do this, you can see that for the
chromatogram (the “solvent front”) – we’ll consider t0 in a later three peaks, k ≈ 1, ≈2 and ≈3, respectively.
article. As both tR and t0 are in the same units (min, sec, furlongs As we’ll cover in a later article, we get the “best”
or fortnights), the units cancel out and k is a dimensionless chromatography if 2 < k < 10, and usually 1 < k < 20 is
quantity. acceptable for isocratic separations. If the retention range is wider
The calculation of k is a simple one, but it still requires a than this, it is likely that a gradient will be required. And if k < 1,
calculator, and this activation-energy barrier is too much for we tend to have more problems – less stable separations and a
many of us (where is that calculator…?), so we don’t bother. But higher chance of chromatographic interferences at the beginning
for many purposes, we don’t need to know k to much better than of the chromatogram.

tR
t R - t0
k=
t0

t0

minutes

k= 0 1 2 3 4

Figure 1: Retention factor can be calculated using the standard equation or estimated by using t0 as a ruler.

HPLC Back to Basics: A Laboratory Companion for Liquid Chromatographers #1

Column Dead-Time
In the previous article we looked at the retention factor, k, and column length (in cm) by 0.1. Thus, a 150 x 4.6 mm i.d. column
how to calculate or estimate it. In order to perform either of these would be 15 cm long, so VM ≈ 0.1 x 15 ≈ 1.5 mL. This should be
processes, we need to know the column dead-time, t0 . If we are within about 10% of the true column volume. To convert column
using a UV detector and a “real” sample, usually there is an obvious volume to dead-time, just divide by the flow rate, F. So for the
disturbance in the baseline, as illustrated in Figure 1. If the sample current example, at 2 mL/min, t0 ≈ (1.5 mL) / (2 mL/min) ≈
is very clean, t0 might appear as a little zig-zag of the baseline 0.75 min.
[Figure 1(a)], but in most cases there is a significant “solvent” So far, so good, but what if you use a column that is not 4.6
or “garbage” peak at the beginning of the chromatogram, as in mm i.d.? You can use the equation at the bottom of Figure 1. The
Figure 1(b). As we’re only interested in a good estimate of t0 at exponent means that a calculator is needed, which will probably
this point, just pick a consistent place to measure it. The arrows mean you won’t bother with this calculation. But wait! The
in Figure 1(a) and 1(b) show where I’d pick to estimate t0 – just second most common column after 4.6 mm i.d. is 2.1 mm i.d.,
where the peak starts to rise above the baseline – this will be and the estimate is simple. The column volume will be directly
easy to measure consistently. related to the change in cross-sectional area of the column,
An alternate, but less convenient, way to measure the column which is proportional to the square of the ratio of the column
dead-time is to inject something that you know is unretained. diameters. So if we consider the two most popular columns, the
This is what the column manufacturer does. Uracil is the most factor is (4.6 / 2.1)2 = 4.8 ≈ 5. I like 5s and 10s, because I can do
popular t0-marker, because it has good UV response and is the math in my head. Consider the most popular LC-MS column:
unretained for mobile phases of ≥60% methanol/water that are 50 x 2.1 mm i.d. If this were a 4.6 mm i.d. column, it would
typical of most column test conditions. Thiourea is another good have a volume of 50 mm = 5 cm x 0.1 ≈ 0.5 mL according to
marker, especially if weaker mobile phases are desired for column the estimate discussed earlier. We know that the 2.1 mm column
testing. is smaller diameter than the 4.6 mm one, so we divide by the
But what happens if you don’t see a disturbance at the conversion factor, 5. this gives us VM ≈ 0.1 mL = 100 µL. All in
beginning of the run? This can happen, for example, if you are our head — pretty simple, huh?
using a MS-detector. As mentioned last week, estimates are We’ve looked at several ways to estimate t0 by using the
good enough for our current needs, so we can estimate t0 in one chromatogram or the column size. These are good enough for
of two ways. If you are using a 4.6 mm i.d. column, which is the general purposes of method development and troubleshooting.
most common column diameter, the estimate is very simple: In the next article we’ll see how we can use t0 to help diagnose
the column volume, VM, can be estimated by multiplying the problems.

(a) (b)

minutes minutes

• inject a non-retained solute (e.g., uracil)

• calculation:
VM ≈ 0.1 L(cm) (for 4.6 mm i.d. columns)
t0 = VM / F
VM ≈ 0.5 L dc2 (all units in cm)

Figure 1: Various ways to estimate the column dead-time.

Separation Science - www.sepscience.com
tR-t0 Diagnostics
If we look at change in the chromatogram and concentrate on in pH adjustment or measuring the methanol. Yes, the mobile
what happens to the retention time, tR, and the t0, there are phase can deteriorate or evaporate over time, causing changes,
only four possible combinations: both t0 and tR change together, but these are much less common. And they are noticed when a
only tR changes, only t0 changes (highly unlikely), and neither new batch of mobile phase is made, causing the step-change in
variable changes (no change = no problem). Only the first two retention.
are of interest, so these are shown in the table in Figure 1. Column packing changes tend to be slow in development
First, let’s consider what could be the matter if both retention over hundreds or thousands of samples, and one-way in nature.
and the column dead-time change together. It has to be either Retention times tend to increase or decrease gradually as the
a flow-rate-related change or a change of the column size. Last column ages. Column changes usually are accompanied with an
time I checked, the column doesn’t change size spontaneously increase in system pressure.
– you don’t put a 150 x 4.6 mm i.d. column on the HPLC system Temperature changes often show up as a diurnal change,
for an overnight run and come back in the morning to find that particularly if the column is not operated in a column oven. As
it has shrunk to 123 x 4.3 mm. So column-size changes are the the temperature increases, retention decreases – approximately
result of operator error and should be obvious. We are left with 2% / 1 ºC temperature increase. I remember working in one
changes in flow rate. Increased flow rate again usually is operator lab without air conditioning that would heat up 5-10 ºC in
error, but it is (remotely) possible that a controller malfunction the summer as the sun blazed in the south-facing windows
could occur. The most likely causes of a lower flow than normal and heated up the brick facing on the outside of the building.
are a leak, a faulty check valve, a bubble in a pump head, or a Retention times decreased when this occurred, but they
bad pump seal. increased again at night when the lab cooled off. Even labs
If only the retention time changes, we can see from Figure 1 with better climate control may have different day and night
that we probably have a problem with the mobile phase, the thermostat settings, and this can cause temperature cycles
column packing, or the temperature. As these are the most likely that correlate with retention changes. Use a column oven and
causes of problems, you might think that this diagnostic table keep the HPLC system away from drafts and you’ll minimize
isn’t of much help. However, each of these failure modes has its temperature-related problems.
own characteristics that can help us to isolate the source of the So we’ve seen that a simple examination of the
problem. chromatograms for changes in retention time and the column
Mobile phase problems generally appear in a step-wise dead-time can help us to diagnose possible problem sources
fashion. For example, you make up a new batch of mobile phase with the aid of an understanding of the chromatographic
and the retention times shift because you made a small error behaviours summarized in Figure 1.

t0 tR
flow rate X X

mobile phase X

column packing X

column size X X

temperature X

Figure 1: Use of the column dead-time, t0, and retention time, tR, as diagnostic tools to help
isolate problems.

HPLC Back to Basics: A Laboratory Companion for Liquid Chromatographers #1

Selectivity
Selectivity is the ability of an HPLC method to separate two can separate them – if they interact in the same manner, a single
analytes from each other. Selectivity usually is abbreviated with peak results. Some of the more important variables that affect
the Greek letter α, and is calculated as: selectivity are the solvent strength and type, the temperature
of the column, the buffer and other additives, and the type
α = k2 / k1 of column packing. Let’s look briefly at the way each of these
variables influences reversed-phase separation under isocratic
Where k2 and k2 are the retention factors, k, of the first and (constant solvent-strength) conditions.
second peaks of a peak pair. The calculation of α is shown in By solvent strength, usually we are referring to the ratio of the
Figure 1. For the two peaks with k-values of 1.95 and 2.15, α = aqueous and organic components of the mobile phase. Stronger
1.10. Although this chromatogram looks like a good separation, solvents are those that elute compounds more quickly so that
with a little baseline space between the two peaks, α is a poor retention is smaller. Thus, the less acetonitrile (ACN) or methanol
way to determine the quality of the chromatogram. The reason (MeOH) in the mobile phase, the larger will be the retention
for this is that it does not take the peak width into account. It times and longer the run. This will also cause peaks to broaden,
is easy to imagine that if the two peaks of Figure 1 were twice and because area is constant, to be shorter. Weaker mobile
as wide, the valley between the two peaks would not reach phases also tend to improve the separation, but this is not true in
baseline. This would be a much poorer separation, yet the every case.
α-value would be unchanged. This means that we need a way to The solvent type also affects the peak spacing. For mobile
measure the width of the peaks, as will be covered in next week’s phases of equal strength, that is ones that give the same average
discussion. retention times, peak spacing usually will differ with different
type solvents. The three most common organic solvents used
What Influences Selectivity? for reversed-phase HPLC are acetonitrile, methanol, and less
We might wonder of what use α is, if it doesn’t give us a measure commonly tetrahydrofuran (THF). Although changing from one
of chromatographic quality. In general, if α ≥ 1.1, we should be of these solvents to another is almost guaranteed to change the
able to get baseline separation for a good quality column, but α-value, at least for some of the peaks in a chromatogram, there
there are better ways to measure the separation, as we’ll see in is no way of knowing in advance if a specific change will improve
later discussions. For the moment, let’s look at some of the things or worsen a separation.
that can be used to change α – that is, how can we move peaks Column temperature works in a similar manner to mobile-
around relative to each other? phase strength in that higher column temperatures reduce
Selectivity is changed when we change the chemistry of the retention times and lower temperatures decrease them.
chromatographic system. Changes in the chemistry of the system Selectivity often changes with a change in column temperature,
influence how a sample solute interacts with the stationary but it is not possible in advance to predict whether the separation
phase and mobile phase. If two compounds interact with the of one pair of peaks will improve or get worse. The changes in
column or mobile phase in a sufficiently different manner, we separation with a change in temperature often are different than

Separation Science - www.sepscience.com

those when solvent type or solvent strength is changed. chromatogram. Even a change from one brand to another within
The mobile-phase pH is an important variable when ionic or a type of packing can change chromatographic selectivity).
ionizable compounds are present. Ionized analytes tend to have
smaller retention times than non-ionized ones or ones for which So What?
ionization is suppressed. Buffer strength, or molarity, does not We have seen above that there are many different ways to
have a major effect in most reversed-phase separations, but if too change the relative peak spacing in a chromatogram. Often
little buffer is present, peak tailing can be worse. Other additives, the challenge is how to pick which variable to choose when
such as ion-pairing reagents, also can affect the separation of you desire to make such a change. In contrast, with so many
two peaks under the proper circumstances. ways to change selectivity, we should also be aware that this
Finally, the column packing type can influence the peak means that there are many ways to ruin a satisfactory separation
spacing in a chromatogram. There are numerous stationary by unintentionally changing a variable if we aren’t careful to
phases available, including C18, C8, C4, cyano, phenyl, amino, control the chromatographic conditions. The susceptibility of a
embedded polar phases, and fluoro phases. Each of these has chromatogram to changes in conditions gives us myriad topics
different chemical characteristics, so a change in column type for future discussions.
is likely to change peak spacing for at least some peaks in a

HPLC Back to Basics: A Laboratory Companion for Liquid Chromatographers #1

Efficiency
In this article we’ll consider the last of the initial set of intersect the baseline at ±2 standard deviations (σ). A well-
measurements, the column plate-number, N. This is also called behaved chromatographic peak should be Gaussian in shape,
the column efficiency, and is calculated as so we can conclude that the peak width at baseline is 4σ. If you
substitute 4σ for wb in the first equation, you’ll see that N is
N = 16 (tR / wb)2 merely the square of the retention time divided by

where wb is the column width at baseline between tangents N = (tR / σ)2

drawn to the sides of the peak, as in Figure 1.
An alternate way to calculate N is to use the peak width at half Now with that and a couple of dollars, you can buy a cup of
the peak height, w0.5: coffee at the local coffee shop – aren’t you glad you asked?

N = 5.54 (tR / w0.5)2 So What?

What is so great about N? Why not just use the peak width to
The half-height method is easier to determine directly from the describe a peak? The beauty of the column plate-number from
computer monitor, because you don’t have to draw tangents. It a practical standpoint is that it is approximately the same for all
also enables determination of N if the peak is not fully separated peaks in a chromatogram. And N is possible to calculate from first
from a neighbouring peak, as long as the valley between the principles, so we can get a good idea of what N should be for a
peaks is lower than the half-height. Half-height measurements given column, mobile phase, temperature, and solute without
commonly are the method of choice for automatic determination doing the actual experiment. And while this is not accurate
of N by data systems. enough to make exact comparisons, it allows us to evaluate
You might wonder what the plate number represents. In a whether or not the plate number obtained from a given column
convoluted way, it is related to distillation technology, where is reasonable for the conditions. Plate numbers that are lower
a distillation column can have different stages, or plates, but than expected, say by 30% or more, may be the result of column
the best spinning-band distillation columns may have only a aging, unwanted chemical effects, poor system plumbing, or
hundred or so plates, whereas even a poor HPLC column will other problems. In future ‘HPLC Solutions’ discussions, we’ll look
generate thousands of plates. So relating N to distillation is a at how N can be combined with k and α to evaluate resolution
stretch. And it is not the number of plates in the buffet line at the during method development.
local restaurant – in fact, I think that number commonly is (n-1), I’ll end with a cautionary note – calculation of N, as in
because the person in front of me just took the last plate! equations 1 and 2 is only possible for isocratic separations. This
On a more serious note, the plate number can be related to the technique is not appropriate for gradients – it is best just to use
statistical broadening of a peak. Recall from your basic statistics peak width to describe gradient peaks.
class that tangents drawn to the sides of a Gaussian distribution

tR

N = 16 (tR / wb)2
N = 5.54 (tR / w0.5)2 w0.5

N = (tR / σ)2

4σ
wb

Separation Science - www.sepscience.com

Tailing
So far in this Back to Basics series, all the peaks we’ve looked factor instead of the tailing factor to measure peak tailing. The
at are symmetric. The ideal peak in an HPLC chromatogram is asymmetry factor is based on the front and back half-widths of
Gaussian in shape, with an equal amount of distortion on the the peak, but these are measured at 10% of the peak height. The
front and back edge of the peak. However, in the real world, asymmetry factor is determined by dividing the back half-width
this is rarely the case – most peaks tail. Because excessive peak by the front half-width (Figure 1, right side). As with the tailing
tailing is an indication that something is wrong, it is a good factor, a Gaussian peak will have equal half-widths, so the value
idea to include a measure of peak tailing as part of the system of the asymmetry factor for a Gaussian peak will be 1.0.
suitability measurements. One might wonder why there are two different ways to
Peak tailing is most commonly measured in one of the two measure peak tailing, and I really don’t have a clue. In fact, there
ways shown in Figure 1. In the pharmaceutical industry, the are other less popular methods, too, such as the tau-sigma
tailing factor, TF, is used. This may also be referred to as the technique. You can see from the data of Table 1 that there isn’t
USP tailing factor or the EP tailing factor, for the United States much difference between the values of As and TF at values less
Pharmacopoeia or European Pharmacopoeia, two of the than approximately 2, but the numbers diverge as they get
regulatory bodies generating guidelines for pharmaceutical larger. Which one is better to use? I don’t think it makes much
HPLC methods. The tailing factor is determined by drawing a difference, so it is best to employ the technique used most
perpendicular line from the peak centre to the baseline of the commonly in your industry or dictated by your company’s
peak. Then the peak width and the front half-width are measured policies.
for the peak at 5% of the height of the peak. The tailing factor is The important practice is to calculate the peak tailing
simply the entire peak width divided by twice the front half- consistently, and on a regular basis, such as in your system
width (Figure 1, left side). For a perfectly Gaussian peak, the suitability test. As we’ll see next week, peak tailing (As or TF) of
front half-width will be exactly half the entire peak width, so the less than ≈2 usually is acceptable. An increase in tailing can be
tailing factor will be 1.0. an indication of column failure, poor mobile phase preparation,
The non-pharmaceutical world tends to use the asymmetry or some other chemical change in the system.

As TF
(at 10%) (at 5%)
1.0 1.0
1.3 1.2
1.6 1.4
1.9 1.6
2.2 1.8
2.5 2.0

As = BC/CA10%

Figure 1. Calculation of peak tailing. Left, USP or EP tailing factor, TF; right, asymmetry factor, As.
HPLC Back to Basics: A Laboratory Companion for Liquid Chromatographers #1
Resolution
In previous articles of this Back-to-Basics series, we have looked however, we are most concerned about the resolution when
at the retention factor, k, the selectivity, α, and the column plate peaks are marginally separated. This means that the peaks may
number, N. Nice as these measurements are, they don’t tell us overlap a bit at the bottom, and measurement of the peak width
much about the quality of the separation. For example, you can at baseline is not possible. Just as we saw that it is possible
have k in the ideal 2 < k < 10 region, or α = 1.2, or N = 15,000, to measure N at half the peak height, we can take the same
but none of these values on their own make the separation any approach for calculation of resolution:
good. As a measure of separation quality, we need to determine
the resolution, Rs, which is most commonly calculated as: Rs = (tR2 – tR1) / ((1.7 * 0.5 (w0.5,1 + w0.5,2))

Rs = (tR2 – tR1) / ((0.5 * (w1 + w2)) where w0.5,1 and w0.5,2 are the peak widths measured at half
the peak height. Note that the factor of 1.7 is added to the
We need to measure the retention times for the two peaks of denominator to adjust for the difference in width at the half-
interest (tR1 and tR2) and the widths of the two peaks at baseline height. The half-height technique is the way many data systems
(w1 and w2) between tangents drawn to the sides of the peaks. measure resolution, because it is simpler to measure than the
For the steroid separation of Figure 1, we can calculate baseline width.
So what represents a good value of resolution? If the peaks
Rs = (3.15 – 2.95) / (0.12) = 1.67 ≈ 1.7 have only minor tailing, a value of Rs = 2.0 should be adequate,
but as tailing increases, the resolution requirement increases, as
If the peaks are perfectly symmetric, which is unlikely, the well. Rs ≥ 2.0 may not be possible, and values of Rs ≥ 1.7 are
valley between the peaks should just touch the baseline when Rs fine in many cases. But don’t put all your trust in a number for
= 1.5. So with well-shaped peaks, as in Figure 1, Rs = 1.7 looks resolution – confirm visually that the valley touches the baseline
pretty good, with a little baseline between the peaks. However, between the two peaks. If Rs < 1.7, or you otherwise see peak
nearly every peak shows some degree of tailing, so to allow for overlap, you might want to consider quantifying the peaks by
a small amount of tailing and still retain a bit of flat baseline peak height, rather than area. The peak height measurement
between the peaks, Rs ≥ 2.0 generally is desired. can tolerate much more peak overlap than area can before errors
The use of equation 1 is convenient and gives good results, occur.
but it is useful only if the peaks are baseline-resolved. Often,

Separation Science - www.sepscience.com

Fundamental Resolution Equation, Part 1
In the previous article we looked at the way we measure First, let’s look at the influence of the plate number on
resolution, Rs, from a chromatogram using the following resolution
equation:
Rs = f(N0.5) (3)
Rs = (tR2 – tR1) / ((0.5 * (w1 + w2)) (1)
We can see that resolution is a function of the square-root
Where t1 and t2 are the retention times of a pair of peaks and of the plate number. You’ll recall that N can be increased most
w1 and w2 are their corresponding baseline peak widths. This easily by increasing the column length, reducing the particle
calculation, or a similar equation using widths at half the peak size, or both. Flow rate has a minor effect on N for samples with
height, is great for reporting resolution values, such as with molecular weights <1000 Da in real applications, so we won’t
system suitability measurements, but there is another resolution consider it here. The square-root function means that if we have
equation that is more useful in guiding us in the method a separation such as that shown on the left of Figure 1, with
development process. resolution of 0.8 and we want baseline resolution of 1.6, we
What is often called “the fundamental resolution equation,” need to increase N by a factor of 4. What will it take to make this
expresses resolution as change? Perhaps we should hook four columns together in series
– an unlikely event, because this will increase both run time
Rs = ¼ (N)0.5 (α-1) (k / [1+k]) (2) and pressure by fourfold, to say nothing of our column budget!
Maybe it would make more sense to reduce the particle size?
Where N is the column plate number or efficiency, α is Ahh, switch from 5 µm diameter particles to 1.25 µm particles.
the selectivity, and k is the retention factor. The utility of this Anyone know where to get such particles? … me neither. So
equation cannot be overemphasized, and is the basis of method the bottom line is that doubling resolution by changing N is not a
development software, such as DryLab (Molnar Institut, Berlin). very practical approach.
We’ll look at equation 2 in more detail here.

4xN

HPLC Back to Basics: A Laboratory Companion for Liquid Chromatographers #1

Fundamental Resolution Equation, N
In the previous article we began our study of the fundamental This generally is sufficient for LC-MS applications and allows
resolution equation shorter run times than a 100 or 150 mm long column.
What about the use of sub-2-µm particles as a means to
Rs = ¼ (N)0.5 (α-1) (k / [1+k]) (1) increase N and thus resolution? The plate number is inversely
related to particle size, so a change from 5 µm to 2 µm particles
Where N is the column plate number or efficiency, α is the would give a 5/2 or 2.5-fold increase in N. But this would
selectivity, and k is the retention factor. We saw that, because N increase resolution only by (2.5)0.5 ≈ 1.6-fold. This doesn’t seem
contributes to resolution as a square-root function, using N by too bad until we consider the relationship between particle size
itself is not a very powerful approach to make large increases in and pressure. Pressure increases inversely with the square of the
resolution, such as by twofold or greater. particle size change. That is, the 2.5-fold reduction in particle size
If we are going to use equation 1 as a guide for method would generate a 6.25-fold increase in pressure. A 150 x
development, we first need to pick a starting place. This means 4.6 mm column operating at 2000 psi (140 bar) with 5 µm
that we have to choose a column. If we plot the dependence particles would generate 12,500 psi (860 bar) if it were filled
of resolution on N, we can obtain the relationship shown in with 2 µm particles. This would require either the purchase of
Figure 1. Here I have marked an N = 10,000 plate column as a a higher-pressure HPLC system or reduction of the flow rate by
recommended starting point. This is a 100 x 4.6 mm, 3 µm or >2-fold (with a corresponding increase in run time) to keep
150 x 4.6 mm, 5 µm column. Either of these columns can be the pressure within the operating limits of a conventional HPLC
run at 2 mL/min, so the run times will not be excessive. And system. Of course, a shorter (e.g., 100 mm) 2 µm column might
they have sufficient peak capacity to separate approximately 12 be an acceptable compromise for a smaller gain in resolution
sample components without a huge challenge. If we choose a with a smaller penalty in pressure.
250 x 4.6 mm, 5 µm column, we gain only ≈30% in resolution The bottom line is that the plate number is not a very
for a 65% increase in retention time and pressure – not a very powerful way to leverage resolution because of the square-root
good tradeoff. If we want to maintain the original pressure, we dependence of Rs on N. It is best to pick a column that is likely to
would have to make a corresponding reduction in flow rate, separate the complexity of sample you are using and then take
further increasing the run time – even less desirable. In contrast, advantage of k and α for changes in selectivity to increase Rs. For
if we are doing LC-MS, the detector has separation power that most of us, this means a 150 x 4.6 mm, 5 µm or 100 x 4.6 mm,
complements that of the chromatographic column, so we usually 3 µm column operated at 1-2 mL/min.
choose a 50 x 2.1 mm, 3 µm column that generates N ≈ 5000.

Figure 1: The relationship between the column plate number and resolution.

Separation Science - www.sepscience.com

Fundamental Resolution Equation, Alpha
In the last two articles we’ve looked at the fundamental stationary phase, or sample. A change in mobile phase pH is the
resolution equation most powerful way to change α, but it only works for ionizable
compounds. A more universal way to change α is to change the
Rs = ¼ (N)0.5 (α-1) (k / [1+k]) (1) mobile phase solvent – acetonitrile (ACN) to methanol (MeOH)
or tetrahydrofuran (THF) are the most common choices.
Where N is the column plate number or efficiency, α is the Another option is to change the column chemistry, but the
selectivity, and k is the retention factor. In using equation 1 as column labels of C18, CN, or AQ aren’t necessarily useful in
a method development guide, we saw that the first step is to this regard. To a lesser extent, a change in the solvent strength
choose a column with a plate number that is likely to separate (%-organic solvent), gradient time, or temperature can be used
the complexity of sample we have. Often a N ≈ 10,000 is to change α.
sufficient for our needs. This can be obtained with a 150 x Although a change in α is the most powerful way to change
4.6 mm, 5 µm or 100 x 4.6 mm, 3 µm column operated at resolution, because of the relationship shown in Figure 1, it is
1-2 mL/min for reasonable run times. Now let’s move to the not very predictable without prior knowledge. That is, I’m willing
right in equation 1 and look at the separation factor, α. to bet that the chromatogram will look different if I change the
In a similar way to what we did for N, if we plot Rs vs α, we mobile phase organic solvent from ACN to MeOH, but I can’t tell
see the relationship shown in Figure 1. This is a very unexciting you if the separation will get better or worse. And although an
graph, but what it tells us is that as we increase a, we get a overall change in selectivity for the chromatogram is likely with
direct increase in Rs. This catches my attention – if I double alpha, such a change, for a given peak pair, the separation may increase,
I double resolution. That’s a much better return than doubling decrease, or not change at all. Because of the unpredictability of
the plate number only to receive an optimistic 40% increase in changing α without prior knowledge of the system chemistry,
resolution. improving the separation by changing α is best left until other
In an earlier article we saw that α is changed by making variables have been explored – specifically a change in the
chemical changes in the system – in the mobile phase, retention factor, k, as will be explored in the next article.

HPLC Back to Basics: A Laboratory Companion for Liquid Chromatographers #1

Fundamental Resolution Equation, k
In the last three articles we’ve looked at the fundamental see curves such as the blue one of Figure 1 – for example with
resolution equation reaction completion. We like to work up on the flat portion of the
curve, because we can tolerate small changes in the independent
Rs = ¼ (N)0.5 (α-1) (k / [1+k]) (1) variable (k in this case) with insignificant changes in the
dependent variable (Rs). If 2 < k < 10, we are in the flat portion
Where N is the column plate number or efficiency, α is the of the curve, so we are likely to get more robust separations than
selectivity, and k is the retention factor. We saw that because if k < 2. Also, the contribution of k to Rs is in the 70-90% range
of the poor resolution leverage gained by the square-root of the maximum, which means that we have obtained most of
dependence of N, it is best to pick a column with a reasonable the “leverage” of k as a variable to improve resolution when
plate number in the beginning, but to wait for further 2 < k < 10.
investigations of N until later. In the previous article we looked For values of k < 2, for example, k = 1, we are taking
at α, and found that, although it is the most powerful variable less advantage of k as a variable than at larger values. More
to use to change resolution, it is also not very predictable, so it is importantly, we are in a steep part of the k vs Rs curve, meaning
best left until later to investigate thoroughly. In this article we’ll that small changes in k will result in larger changes in Rs, for a
look at the relationship between resolution and k. less robust separation. Finally, with small values of k, the peaks
Let’s look at the k-term of equation 1. If k is small, for example, will come out quite early in the chromatogram and are more
1, we get a value of 1/(1+1) or 0.5. If k is quite large, for likely to encounter interference from the unretained junk that
example, 100, we get 100/101 ≈ 1. So, as k is increased, Rs is elutes at the front of the chromatogram. Looking at the other end
increased, but not in a linear fashion. As we did in the previous of the 1 < k < 20 range, there is no particular advantage of large
discussions for N and α, we can plot Rs against k, as shown in k-values in terms of resolution. And the separation suffers from
Figure 1. Here you can see that there is a steep rise in resolution longer run times and broader peaks, which in turn mean shorter
at small values of k which levels off as k is increased. peaks and poorer detection limits.
I often recommend that you will get the “best” chromatography Because of the relationship of k and Rs shown in Figure 1,
if you can obtain k-values in the range of 2 < k < 10, or if the it usually is best to start method development by choosing a
sample polarity range is larger so that all the sample peaks column with N ≈ 10000, then to adjust k to obtain 2 < k < 10 if
won’t fit into this range, 1 < k < 20 also is acceptable. These possible, or 1 < k < 20 if necessary. Once k is optimized, then a
recommendations are based on Figure 1. In chemistry, we often can be addressed if more selectivity is needed.

Figure 1: The relationship between the retention factor, k, and resolution.

Separation Science - www.sepscience.com

 produced by

in collaboration with
John Dolan has written more than 130 papers on
HPLC and related topics. He is the coauthor of
Troubleshooting HPLC Systems, High-Performance
Gradient Elution, and Introduction to Modern Liquid
Chromatography, three authoritative books in the
field. Dr Dolan has taught HPLC techniques to more than 10,000 of
students over the last 30 years.

more on-line learning from John Dolan at

Eclipse Business Media Ltd

2 Clarendon Road, Ashford, Middlesex TW15 2QE, United Kingdom