Pressing the LOQ down

[Mattias]

Hi,

How would you do to create an LC-UV method with the best possible LOQ using the following assumptions:

You don't need to separate the main peak from anything but the solvent front. You know the optimal wavelength. You don't want to have any extra sample prep (as SPE).

The obvious is to inject as much as possible, and to use a weaker solvent in the sample than in the mobile phase. And to use a non-Uv absorbing mobile phase. But would you go for a short or long column, isocratic or gradient elution?

[DR]

Isocratic/gradient - doesn't matter, as long as you have decent peak shape. Column length - however long you need to acheive separation and some minimum value of K will do.

Overloading the column w/ sample can actually be counterproductive, depending on what comes along with the analyte of interest.

Other keys to low LOQ - having low system noise, and possibly increasing sensitivity. Using a fresh lamp and a flow cell with a longer path length can help.

[Mattias]

If we assume that the noise is more or less constant, I guess the QL will only be dependant on the peak height (or width)? High flow and small particles and/or steep gradients should produce the narrowest peaks?

We have a 50 mm flowcell in house, but I doubt that it will increase my S/N that much.

We use RIA now for cleaning validation, and I just want to see how far I can go with HPLC instead.

[Mark Tracy]

Well, I pushed LOQ for one method by using a 4.3mm guard column as a concentrator cartridge, then eluting that onto a 2.1 mm analytical column. That let me inject 1 mL onto the column. You need an auxiliary pump to load and wash the cartridge, and an auxiliary valve to switch into line with the analytical column. It is described in the LC/GC applications notebook Feb 2007 on page 21; sorry there was not room enough for a diagram. Your local Dionex sales rep would be happy to show you the setup.

[Bruce Hamilton]

As you say, a large injection volume, weaker solvent ( but not at the expense of the baseline or peak shape - I'd inject in mobile phase if necessary and, as DR notes, overloading can counterproductive ), really-really good quality solvents that don't absorb, isocratic with premixed and degassed solvent, detector settings at most stable and sensitive, and a column that best matches the noise characterics of your HPLC system, probably 2mm ID and 3um at 50mm length, targetting a K value of 1.

Good luck,

Bruce Hamilton

[tom jupille]

First of all, it may be easier to separate out injection volume issues. The LOD or LOQ in mass terms can be approximated by:

Wm ≈ (1000 * MW * (k’+1) * (S/N) * N’ * L * dc^2) / (ε *N^0.5 * Lc)

Where:
Wm is mass-on-column (in milligrams)
MW is analyte molecular weight
S/N = target signal to noise ratio (≈ 10 for LOQ; ≈ 3 for LOD)
N' = baseline noise
L = column length (in cm)
dc = column internal diameter (in cm)
ε = molar extinction coefficient
N = plate number
Lc = flow cell path length

(from the Snyder, Kirkland, & Glajch Practical HPLC Method Development book)

The most effective single thing you can do is to reduce baseline noise. Next most effective is to increase the flow cell path length (other things being equal, your 50 mm cell should drop LOQ by a factor of 5 compared to a standard 10 mm cell).
Make the k' as low as you can get away with.
Use the narrowest possible column (but be careful about extra-column effects, which can limit your injection volume!).

If you are expressing LOD in concentration terms, then the remaining variable is the injection volume (as you have pointed out, the weaker the diluent, the more you can inject).

[Bruce Hamilton]

Maths was never fun, and I know the equation components can't be considered in isolation, but if you just halved/doubled each parameter in the equation, wouldn't reducing the Column ID have the most profound effect?.

Please keep having fun,

Bruce Hamilton

[DR]

Yes, but that's the term closest to the landmine.

[Mark Tracy]

Halving the column i.d. produces a fourfold reduction in LOQ only if you can maintain the original injection volume, and the extra-column effects are sufficiently small.

[HW Mueller]

Neat, this chain allows me to erase the questionmarks in my old, 1988 version of Snyder, Glajch, Kirkland, Practical HPLC Method Development. It has the dc^2 term in the denominator of the equation for Wm, which is contrary to all the arguments for smaller diameter columns. Furthermore, a few pages byond the equation there is a sample calculation (equ. is 4.1 on page 87, the sample calc. is in Table 4.3 on page 90) which uses dc, not dc^2, in the denominator (double mistake, Tom´s Wm with dc^2 in the numerator must be correct).
Hopefully this helps other people who have access to the 1988 version.

[HW Mueller]

Whew.... found another discrepancy:
Tom, you use L (column length) in the equation above, Snyder, et al, in 1988, use L^0.5. In that case it appears that the latter could be a better estimate?

[DR]

Halving the column i.d. produces a fourfold reduction in LOQ only if you can maintain the original injection volume, and the extra-column effects are sufficiently small.

Landmines with hair triggers... analyte capacity, extracolumn effects - both of which can suffer if injection volume, diluent composition etc. are not right.

[tom jupille]

Tom, you use L (column length) in the equation above, Snyder, et al, in 1988, use L^0.5. In that case it appears that the latter could be a better estimate?

Intuitively, the "L * dc^2" factor is proportional to the column volume (larger volume = more dilution). In the equation, the "L^0.5" dependence is captured by the plate number in the denominator.

Of course, using a shorter column will reduce the plate number (unless you use smaller particles, which requires higher pressure, which . . . etc. -- that's what makes chromatography interesting! ).

[tom jupille]

Tom, you use L (column length) in the equation above, Snyder, et al, in 1988, use L^0.5. In that case it appears that the latter could be a better estimate?

Intuitively, the "L * dc^2" factor is proportional to the column volume (larger volume = more dilution). In the equation, the "L^0.5" dependence is captured by the plate number in the denominator.

Of course, using a shorter column will reduce the plate number (unless you use smaller particles, which requires higher pressure, which . . . etc. -- that's what makes chromatography interesting! ).

[HW Mueller]

In other words, the equation has more qualitative than quantitative use?

(On dc: I tend to agree with DR on that, probably, because I tried a 2mm dc column too early in this development. It had horrendous backpressure, etc., etc. Anyway it was a very short and intense experience which prompted me to go in the direction of concentrating samples more, even though this required more extensive cleanup or mullti-step HPLC).