Mixing and Mixers in Liquid Chromatography—Why, When, and How Much? Part I, The Pump

What types of mixers are needed with different types of LC pumps? Are different mixers needed for different applications?

With some topics I have written about in the past year, I have tried to draw attention to aspects of our practical work with LC that I think are often underappreciated-for example, the value of using inline mobile phase filters, and best practices around reducing contaminants when using LC–MS. There are other topics where the issues I observed in talking with people about their LC work are more about misunderstandings, misconceptions, and myths. This month, I have chosen to tackle a topic of this kind, focused on mobile-phase mixing and mixers. I think there is some confusion among users of LC about what happens when we combine two (or more) different flow streams in LC systems, and when mixers are needed to blend the different fluids. Hopefully, the discussion will enable a more detailed understanding of why they are needed, and when and how one might consider using something other than the default mixer that is used at the time of installation.

Why do we need mixers in LC pumps in the first place?

Common Pump Designs

The vast majority of LC pumps in use today can be approximately characterized by one of two designs, shown schematically in Figure 1. These designs have been discussed in the "LC Troubleshooting" column in the past (1), and there are good educational resources for learning about them elsewhere (see www.chromacademy.com). Readers interested in much more detail will find an extensive discussion of modern pumps by Frank Steiner in "The HPLC Expert II," edited by Stavros Kromidas (2). For the sake of discussion here, we need to establish that we will discuss the designs at a particular level of detail to appreciate why and when a "mixer" is needed. The design in Figure 1a is typically referred to as a high-pressure mixing system, meaning that each of the mobile-phase constituents is first pushed with enough force to meet the inlet pressure of the column, and then these flows are brought together in a fluidic junction (referred to here as the convergence point) upstream from the point where the sample is injected. These systems are usually binary pumps, meaning that they have two independent high-pressure pumps that can deliver two different solvents to the convergence point at high pressure. The design shown in Figure 1b is typically referred to as a low-pressure mixing system, meaning that two or more fluids are brought together at the convergence point while at low pressure, and then the single stream of combined fluid is pushed by a single high-pressure pump to meet the inlet pressure of the column. The major advantage of the high pressure mixing design in Figure 1a is that the time it takes for a change in solvent composition made by the pump (as in gradient elution) to appear at the column is typically much shorter than it is for low-pressure mixing systems. We refer to this as the gradient delay time, and we'll discuss this more later on in the article. On the other hand, the major advantage of the low-pressure mixing system illustrated in Figure 1b is that these typically come with a lower purchase price, simply because there is one high-pressure pump involved instead of two. The fact that this type of pump can also produce ternary (three fluids) and even quaternary (four fluids) mobile phases can also be a great asset for some applications (for example, making pH gradients online using the pump).

Figure 1: Block diagrams for the two most commonly used designs of HPLC pumps in use today: a) binary pump with high-pressure mixing; and b) quaternary pump with low-pressure mixing.

Imagining the Convergence of Two Fluids in the Pump

Now, to the question at hand: Why do we need a mixer? First, it is helpful to define the meaning of mixer in the context of this discussion. Here, I will use the term mixer to describe any device that is deliberately used in a LC system for the purpose of mixing two or more fluids. Of course, in real LC systems there are many elements of the system that promote mixing of the mobile phase (and sample), whether or not this is the intent of that element. For example, some mixing can occur in a straight, open connecting capillary, and a fair amount of mixing happens in all chromatography columns (although we try to prevent this by design!). Now, the ways that different fluid streams converge in a LC system can be very different, depending on the context in which this happens. This month we will consider two scenarios: 1) convergence in a high-pressure mixing pump (Figure 1a); and 2) convergence in a low -pressure mixing pump (Figure 1b). Next month, we will consider the convergence of sample and mobile phase when a sample is injected into a column (Figure 2, scenario 3). Figure 2 illustrates, albeit in an idealized way, the major differences between these three scenarios in terms of how two fluids will converge. In the first scenario, the two fluids are brought very close together in a single small-diameter tube such that molecules of fluid A only have to travel a short distance (hundreds of micrometers at most) by diffusion or convection, or both, to actually mix with molecules of fluid B. In this case, there is probably quite a lot of disordered flow that happens in the tube as it makes turns and goes through connections to valves and other fittings. This disordered flow is probably sufficient to mix two fluid streams if they really are consistent over time, as shown in scenario 1 of Figure 2. This ideal is never exactly achieved in pumps that use reciprocating pistons; we'll address this in a bit more detail below.

Figure 2: Idealized representation of the different ways two fluids converge under different circumstances in LC systems: 1) convergence in a binary high-pressure mixing pump; 2) convergence at the outlet of a solvent proportioning valve in a low-pressure mixing pump; 3) convergence when a sample is injected into a mobile-phase stream that will carry the sample to the LC column.

Moving on to scenario 2 in Figure 2, we see that the way the fluids converge here is entirely different from the scenario in Figure 1. In the low-pressure mixing design, the composition of the mobile phase is determined by the duration of opening and closing of solenoid valves that are mounted on the solvent proportioning valve. For example, in the case where the desired mobile-phase composition is 25:75 A:B (as in Figure 2), solenoid A may open for 150 ms and then close, followed by solenoid B opening for 450 ms and then closing. These two events constitute one cycle of the valve that allows the pump to draw in a packet of the 25:75 A:B mixture, and this process repeats over time. The important difference between scenarios 1 and 2 is that in scenario 2 the fluids are brought together in a serial fashion, like slices of bread assembled into a loaf, whereas in scenario 1 the two streams are brought together in a parallel fashion. In scenario 2, this means that for the two fluids to actually mix to produce a homogeneous mixture, the molecules of fluid A have to be moved relatively long distance to mix with the molecules of fluid B, as shown by the curved arrow in Figure 2. This distance depends on a number of factors, including flow rate, the desired mixture composition, and the diameter of the tube connecting the solvent proportioning valve to the high pressure pump. Quite a bit of mixing will occur as the mixture is drawn into and pumped out of the high-pressure pump head, but generally speaking pumps with this low-pressure mixing design require mixers with larger volumes, mainly because this is what is required to effectively make the mixture of fluids homogeneous. There is one additional and important detail relevant to scenario 1 that is relevant here. As stated above, most pumps in use today that use the high-pressure mixing design use dual reciprocating pistons in each high-pressure pump. Although there are many variations on this design (2), the basic idea is that one piston draws fluid in from the solvent bottle while the other piston pushes fluid out at high pressure toward the column. These pistons accumulate and displace a finite volume, and so at the end of a piston stroke the travel of both pistons is reversed. At the end of the stroke, a check valve is used to prevent flow of mobile phase backward from the column into the pump. While this all works very well in principle, the important point here is that short-term deviations in the flow from each high-pressure pump are very hard to avoid. A simple view of the consequence of this is shown in Figure 3. When the flow of fluid A is temporarily lower at the end of a stroke, the composition of the combined fluid will be enriched in fluid B, and vice versa. In cases where the mobile-phase components are transparent to the detector (for example, water in the case of UV detection, or methanol in fluorescence detection), we are blind to these deviations in mobile phase composition. However, if the detector is even slightly sensitive to these changes (for example, when using formic acid in water with UV detection at 214 nm, or acetonitrile and water with refractive index detection), these deviations will be detected as a wavy pattern in the baseline. Here again, as was the case with low-pressure mixing systems, a mixer is needed in this case to move fluid molecules from one enriched region of the stream to another as shown by the arrow in Figure 3. And so while very little mixing is needed for high-pressure mixing designs in the ideal case (Figure 2, scenario 1), non-ideal behavior of this type of pump requires that a mixer be added to smooth out the short-term variations in mobile phase composition.

Figure 3: Illustration of the effect of inconsistent flow from individual high-pressure pumps in a binary high pressure mixing system on variation of the mobile-phase composition over time, and the effect that this can have on detector baselines.

Impact of Inadequate Mixing on Chromatographic Performance

The two major practical consequences of inadequate mixing are noisy detector baselines and poor retention time precision. The extent to which each of these is actually a problem in practice depends on a large number of experimental details, and a thorough discussion of all of them is far beyond the scope of this piece. However, a comparison of detector signals under typical operating conditions with and without mixing will illustrate the point that the extent of deliberate mixing in the system can have a big effect on the observed signal. Figure 4 shows absorbance signals from a UV detector when pumping 25:75 acetonitrile:50 mM ammonium formate in water from a high-pressure mixing pump, both with and without a mixer installed. The effect of the mixer is to smooth out short-term variations in mobile phase compositions, which obviously makes the detector baseline appear much less noisy. It is easy to imagine how these kinds of variations without mixing could lead to poor precision in retention time as well.

Figure 4: Comparison of UV absorbance signals (220 nm) obtained without a column installed, while pumping 25:75 acetonitrile:50 mM ammonium formate at 0.5 mL/min. The pump is a high-pressure binary mixing system (Agilent 1290, Infinity II). The outlet of the convergence point was connected directly to the detector with 100 cm × 120 µm i.d. PEEK tubing, with or without the mixer installed.

When Should We Consider Using a Different Mixer?

As with a lot of other method development decisions in LC, the analyst must decide whether the mixer in use yields a level of performance for the method that is sufficient for the application at hand, and explore alternatives when it is not. Some applications are considerably more demanding than others in this regard. One that is particularly challenging, yet very important and in widespread use, involves the use of trifluoroacetic acid (TFA) in the eluent, reversed-phase LC columns, and gradient elution. Such conditions are very commonly used for separations of complex peptide mixtures, as well as mixtures of other hydrophilic amine-containing compounds. This is a particularly challenging case because not only does TFA absorb UV light at the same wavelengths that are used to detect peptides (most commonly 214 nm), but TFA itself is slightly retained by reversed-phase columns in water-rich mobile phases. This means that short-term variations in the mobile phase like that depicted in Figure 3 can lead to waves of TFA moving through the column as pulses of acetonitrile-rich mobile phase cause local decreases in retention of TFA and thus transient increases in elution of TFA from the column. Figure 5 shows a comparison the UV absorbance baselines observed under these conditions, with two different mixers in use. This example clearly shows the benefit of using a larger mixer in this case, as the baseline becomes much smoother.

Figure 5: Comparison of UV absorbance signals (214 nm) obtained with different mixers in use. The same pump was used as in Figure 5, but now with a column (30 mm x 2.1 mm i.d. Agilent SB-C18) in use, and gradient elution. Solvent A is 0.1% trifluoroacetic acid in water, solvent B is 0.1% trifluoroacetic acid in acetonitrile, and the gradient runs from 2 to 40% B in 4 min.

How Much Mixing is Enough?

One thing I have not touched on specifically in this article is mixer design. This is, in part, because there are too many different designs in use to cover in one article! However, although some designs undoubtedly are superior to others, it is wise in most cases to stick with the design supported by the manufacturer of your pump where possible. In cases where the pump manufacturer has limited offerings for different mixers, then, in principle, it is possible to mix and match pumps and mixers because almost all mixers in use today are passive devices-that is, they do not require power and do not have any moving parts. Readers interested in more details associated with specific mixer designs will find quite a lot of discussion on this point in Kromidas' book (3).

Although I just made the case that there should not be anything standing in the way of adding a large mobile-phase mixer to an LC system, it is very important that the user understand the consequence of adding a substantial mixing volume to the system. Any element that adds volume to the flow path between the fluid convergence point in a pump and the LC column will increase the gradient delay time. Again, this is the difference between the time that a change in mobile-phase composition is made at the convergence point and the time that this change arrives at the column inlet. Adding a mixer with a volume of 35 µL to a system that is running at 1 mL/min. will not be very consequential because it will only change the gradient delay time by 0.035 min (~2 s). But, adding a larger volume (say, 200 µL) to a system running at 100 µL/min will have a much bigger effect, as this will increase the gradient delay time by 2 min! The effects of a large change like this will be numerous when gradient elution is used, and so should be made with care. A short list of changes that should be expected includes: significant (absolute) increases in retention times, significant changes in selectivity (relative retention), and increases in analysis time because of the slower arrival of the gradient start at the column, and longer time required to flush strong solvent from the system at the end of the gradient elution program.

Summary

The performance of LC pumps has been improved dramatically in terms of mobile-phase composition precision and accuracy compared to what was possible in the earlier days of modern LC. This has led to a significant reduction in mixer volume and concomitant decreases in gradient delay time, which has been very helpful for increasing throughput of LC methods, reducing solvent waste, and enabling higher performing two-dimensional LC methods. However, even with the best available pumps, mixers are still needed, and the size of the mixer that is optimal varies somewhat by application. A more detailed understanding of why mixers are needed and when a change in mixers is warranted should lead to better method performance and improved method transferability in the future.

References

(1) J.W. Dolan, LCGC North Am. 34, 400–407 (2016).

(2) F. Steiner, in The HPLC Expert II: Find and Optimize the Benefits of Your HPLC/UHPLC, S. Kromidas , Ed. (Wiley-VCH, Weinheim, Germany, 2017), pp. 101–170.

(3) S. Kromidas, Ed., The HPLC Expert II: Find and Optimize the Benefits of Your HPLC/UHPLC (Wiley-VCH, Weinheim, Germany, 2017).

ABOUT THE COLUMN EDITOR

Dwight R. Stoll

Dwight R. Stoll is the editor of "LC Troubleshooting." Stoll is a professor and co-chair of chemistry at Gustavus Adolphus College in St. Peter, Minnesota. His primary research focus is the development of 2D-LC for both targeted and untargeted analyses. He has authored or coauthored more than 50 peer-reviewed publications and three book chapters in separation science and more than 100 conference presentations. He is also a member of LCGC's editorial advisory board. Direct correspondence to: LCGCedit@ubm.com