Best Instrument

Here are a few interesting mini-studies conducted as part of my 'baseline familiiarization' with HPLC hydraulics.
Diagnostic method for binary pumps

Diagnostic method for binary pumps

Pump Validation Method 1
Figure: High level verification of pump seal failure
Looking at the graph, we can see several charachteristics of the solvent flow.
  • Absolute flow accurracy for any given programmed flow rate.
  • Variance of flow at a given flow rate. (sometimes called 'pulsation')
  • 'Rise time', or time to reach the increased programmed flow.
  • 'Decay time', or time to reach the decreased programmed flow.
  • The differences in the above between pump A, and pump B

This method was developed in the lab of Daniel Sykes, Professor of Analytical Chemistry, at The Pennsylvania State University. The background is that John Best, of Best Instrument 'interned' in the summer of 2006 for Dr. Sykes, with the intention of both learning HPLC, testing products, and developing HPLC methods for use in analytical chemistry lab courses. It was noted that one of the two pumps on the HPLC system was not performing well. The 'method' was simply to digitize pump flow rate relative to a programmed range of flow rates, first for 'pump A', then 'pump B'. This method turned pump A on at a low flowrate for a period of time, then increased the flow rate in steps. Then, pump A qould be turned off, and pump B would be taken through the same flowrates.

Pump A, (the malfunctioning pump) runs from T=0 to about T=20 minutes. Pump B runs from 20 minutes until just over 40 minutes. The method file is apparent from the graph, first, step one of the pumps up (or down) through a range of flowrates, then do the same for the other pump. The graph is enlightening. Note the flow rates for pump 1 for which serious anomalies occur. In the end, these anomalies were attributed to a bad piston seal. After a new seal was obtained through Restek the method was re-run. Unfortunatley, the 'post correction' data has been misplaced, but in essence, the flow profiles for pump A and pump B were nearly identical.
In light of flow variance, we should discuss using precisely mixed solvents (isocratic), vs. mixing solvents 'on the fly', and that may become part of a different atricle.

HPLC pump pulsation

HPLC pump pulsation

Raw Pressures
Figure: Raw pressure data at 5 flow rates, normalized.

Pulse Dampers are useful to solve this problem, which is intrinsic to the reciprocating piston pump so often used in HPLC and UHPLC. In this case, a single piston pump is under test.

A ueful way to express pulsation is as a percent of overall pressure, (High Pressure - Low pressure) / Low Pressure.

A few things can be seen from the graph......

As mentioned it is possible to solve these pressure fluctuations by using a pulse dampener, but this is not the only solution, and posibly not the best solution for all situations. The pulse damper stores energy in the form of pressure which compresses a gas in the pulse damper. This energy can then be released to the HPLC column gradually while the pump piston is in 'refill' mode. That is, when the pump is drawing new solvent into the pump cylinder. The timing of this refill varies with flow rate, a very slow refill occurrs at low flow rates, allowing more time for pressure decay. This is evident in the graphs.

Another view of the pressure is to plot pulsation vs. flow rate.

Pulsation Vs. Flow
Figure: Pressure vs. flow rate, single piston pump.

A partial solution is very prevalent in the HPLC industry, that is, the use of two cylinder pumps, which are almost always delivering flow. Pump cams are ground to allow a longer delivery time than the refill time. That said, there is an issue of fluid compressibility which change the precise times check valves open and close, and the amount of cam 'cycle time' (time for a complete revolution of the cam) wasted on compressing the solvent before a check valve will open and the piston is actually delivering solvent.

On request, I will provide an excel spreadsheet which models pressure in both cylinders and the mixing 'T' for various cam profiles. These calculations were made to model a pump design with an adjustable pre-compression cylinder so better 'crossover' performance would be possible without a pulse damper.

Pump flow accurracy vs. driven hydraulic load

Pump flow accurracy vs. driven hydraulic load

Flow at various loads
Figure: Various hydraulic loads and resulting flow accurracy.

This graph of pump flow accurracy vs. load was captured by using 4 different lengths of narrow diameter tubing prior to the column, so that an additional pressure would be developed. Regardless of the particular solvent in use, a fraction of the pumps stroke must be lost while 'pre-compressing' the solvent before the output check valve can open. That said, the effect should be predictably worse for solvents with a high compressibility factor.

This effect shows the inherent challenge in calibrating the flow rate accurracy of HPLC pumps. If compressibility factors of the solvents are known, corrections can be made either by the pump's internal firmware, or by external software which adjusts the pumps programmed flow rate as required.

Separating Caffeine, Theobromine, and Theophyline

Separating Caffeine, Theobromine, and Theophyline

Separation of Caffeine, Theobromine, and Theophyline
Figure: Chromatogram, after method developed.

This HPLC separation method was developed for Dr. Dan Sykes of PSU. The separaion shown was acheived with an isocratic mixture of 10% acetonitrile in water as the solvent.

After developing the method, what is observed.

One might say that the separation is 'too good', in that the histogram is perhaps 'cleaner than necessary', and may therefore provide opportunity for additional speed optimization, even under HPLC conditions. 'HPLC conditions', as opposed to 'UHPLC conditions' is used to imply relatively high flow rates (mL/minute range vs. fractional mL/min) large sample volumes, and generally lower pressures.

Dispersion resulting from tubing slippage

Dispersion resulting from tubing slippage

Dispersion vs. Pressure
Figure: Resolution vs Pressure.

This study was performed at Supelco by John Best and Wendy Roe in 2006 on an Agilent 1100, and standard propritary column test chemistries. The flow profile of the method created ever increasing flow rates, resulting in corresponding increases in column head pressure.

The blue trace results from a 'no-slip', or 'slip-free' type of fitting which (when installed properly) hold the tubing securely at the bottom of the HPLC port. A high performance version of a cone-shaped ferrule makes the pressure seal. This approach was state-of-the-art at the time of the study, and while swept dead volume was minimized, there is an element of 'unswept', or 'indeterminantly swept' dead volume (about 400-600 nL) resulting in differences between the diameter of the tubing and the diameter of the bottom of the HPLC port.

Several observations can be made:
  • Inflection point 1 shows an initial slip of the PEEK tubing.
  • Inflection point 2 shows a second 'slip', resulting in additional sample dilution (or disperion)
  • Inflection point 3 shows failure, that is, the tubing slipped from the fitting completely causing a leak and loss of pressure.

It had been believed the slight loss of resolution of the UHPLC fittings (blue trace) at 5000 PSI is due to internal column dispersion. But it is possible an extra-column dispersion factor may be making itself visible.