126 EQUIPMENT The influence of temperature on chromatographic separation suggests that in most cases some means of column temperature control is necessary. Even though the laboratory as a whole may have adequate climate control, an individual location in the laboratory may experience temperature fluctuations of several degrees, as the heating, ventilating, and air-conditioning (HVAC) system cycles. The use of a column oven on every HPLC system is strongly recommend to maintain constant temperature. If the column oven does not preheat the mobile phase (generally required only for methods operating at > 40 ◦ C), a simple preheater can be fabricated from 0.5 m of 0.005-in i.d. stainless-steel tubing. This tubing should be connected to the column inlet and mounted inside the column oven, preferably in contact with a heated surface. 3.7.2 Oven Designs Three column-oven designs are popular: • block heater •airbath • Peltier heater Water-bath heaters are a simple alternative, but these are inconvenient and seldom used today. If, for some reason, a column oven is not used, temperature fluctuations should be minimized by wrapping the column in an insulating material (e.g., foam pipe-wrap). 3.7.2.1 Block Heater The block heater relies on direct contact of the column with a heat source. Commonly the heat is transferred from a grooved aluminum block into which the column is clamped, with heat provided by a cartridge-type heater. The column and heating block are contained in an insulated compartment. In another format, a flexible blanket of heating tape is wrapped around the column. The direct contact of the column with the heater provides efficient heating [12]. If additional solvent preheating is necessary, a preheater can be used as described above. 3.7.2.2 Air Bath An air-bath heater is constructed the same way as a gas chromatography oven; this design is effective at controlling the column temperature. Air is a less efficient conductor of heat than metal, so column-temperature equilibration may take longer than with a block heater 3.7.2.3 Peltier Heater Peltier heaters are popular for HPLC work. In addition to heating, they can control the column oven at ambient temperatures or below—without an auxiliary cooling mechanism. However, most Peltier-heated column ovens are not efficient heat conductors because the column seldom is in intimate contact with the heated surface. Most Peltier oven designs include a preheater, which is a piece of capillary tubing embedded in the aluminum heater block of the oven. Often this preheater 3.8 DATA SYSTEMS 127 provides most of the heating for the column [15]; if a preheater is not used with a Peltier oven, it is unlikely that the set temperature will match the true column temperature. A well-designed Peltier oven will include a grooved block that clamps the column to the heated surface (or some other mechanism to ensure column contact) and an embedded preheater enclosed in an insulated compartment. 3.8 DATA SYSTEMS In the second edition of this book, a total of three paragraphs were devoted to chromatographic data systems. Perhaps no other factor in HPLC practice has been impacted more in the interceding years than data handling. In 1979, dedicated data integrators were available from a few suppliers, but the personal computer (PC) and its ripple-down effect of software to support HPLC were not even on the horizon. At one point HPLC systems comprised separate modules, each controlled by manual settings. Gradually control settings changed from switches and rheostats to microprocessors, and today’s modules each contain many microprocessors. These developments allowed the operational control of all the HPLC hardware functions (flow rate, detector wavelength, autosampler control, etc.) by a dedicated system controller. At the same time the system controller directed the operation of the HPLC, and chromatographic data were collected, processed, and displayed by a separate data system (or data processor). Today, system control and data processing have merged to the point that the device (and its associated software) performs all these functions is simply called a data system; we will use this terminology unless there is a need for clarification. Today’s chromatographic data systems serve many functions In the HPLC operation. Most or all of the following capabilities are available in today’s data systems: • experimental aids (Section 3.8.1) • system control (Section 3.8.2) • data collection (Section 3.8.3) • data processing (Section 3.8.4) • report generation (Section 3.8.5) • regulatory functions (Section 3.8.6) In general, to access all its control capabilities, a data system must be used with HPLC equipment from the same manufacturer as the data system. However, several data-system manufacturers offer control capabilities for instrumentation from the major HPLC equipment suppliers. Data collection and processing are functions that are more universal—many laboratories standardize on one brand of data collection and processing system, and use a central system to collect and process data from a variety of brands and types of laboratory equipment. 3.8.1 Experimental Aids During method development, a systematic approach will reduce the amount of work involved and increase the likelihood of obtaining an acceptable separation. 128 EQUIPMENT Chapter 10 discusses the use of computer simulation as a tool to guide the method development process. Software for method development is available as standalone software that can be used with any HPLC system (e.g., DryLab ® ), or the software may be specifically designed for use with one brand and model of equipment. When a computer is connected to the HPLC system, other useful software may be available for tasks that may not be directly related to the separation. Wizards, help files, and other information can assist the user in isolating and solving system problems quickly. Electronic laboratory notebooks and databases can simplify recordkeeping and system maintenance records. Some systems have reminders that can be set to indicate when pump-seal changes or other maintenance should be undertaken. Several manufacturers supply audiovideo files that can be used to guide certain maintenance tasks. The Laboratory Information Management System (LIMS) is another category of software that allows users to track the flow of samples and sample results through the laboratory. Some of these functions will help with regulatory compliance (Section 3.8.6 and Chapter 12), some are used to help coordinate work activities, and some may be used to support the financial aspects of the laboratory business (e.g., initiate the billing process when sample analysis is complete). For example, consider the application of a LIMS system in a bioanalytical service laboratory that analyzes drugs in plasma. When the samples arrive at the lab, the samples are assigned a number, and information for the sample is added to a sample record in a database. For example, the date and time of receipt, sample condition, sample tracking number, and patient identification (ID) might be entered. The sample tube would have a bar-code label added (if it did not already have one) and would be transferred to a freezer (freezer location, date, and time logged) for holding until analysis. When it is time to analyze the sample, a sample-analysis table is created from the database by the analyst, and samples are pulled from the freezer and moved to the sample preparation lab (date and time recorded). After the sample preparation takes place, the bar code from the original sample container is correlated to the bar code on the sample vial for injection. Additional data might be added to the database during sample preparation, such as the lot numbers of various reagents. Any remaining raw sample would be returned to the freezer (date, time, and location recorded) A sample-analysis sequence table is created that correlates the autosampler-tray position with the sample ID. After analysis, the data are processed to generate a report of sample concentration for each sample; these data would be automatically entered into data tables in a report template. The analyst would review the data and transfer the report to the quality unit for further review before he sends it to the client (each approval or review would be recorded). When the report was ready to send, another report would be sent to the accounting department to bill the client for the analyzed samples. At any future point, customized reports could be created for special purposes, such as chain of custody, identification of problems (e.g., correlation of a batch of reagents with aberrant results), number of samples run per instrument per month, and so forth. The capabilities of LIMS systems are often customized for a particular laboratory or application, so the possibilities are practically limitless. In some cases the boundary between a LIMS and a data system gets blurred; LIMS software can duplicate, replace, or take advantage of separate data-system software. For more information, 3.8 DATA SYSTEMS 129 consult one of the LIMS vendors; a Google search of ‘‘LIMS vendors’’ identified one listing of > 200 companies that supply LIMS and related software or hardware. 3.8.2 System Control One of the most widespread uses of the computer in support of HPLC is to control the system. The system-control function provides a single point of control for all operational settings; for example, flow rates, mobile phase composition, column temperature, and detector wavelength. All the settings for a specific method can be stored for easy retrieval and setup when the method is next used, as well as for archival purposes, or for transfer to another HPLC system. Method modification and the development of new methods are simplified, because an existing method can be used as a template, so only the necessary items have to be changed. Most controller software contains features that make it easy to make a permanent electronic record of the specific instrument settings used for each sample run within a laboratory, information that may help when troubleshooting system problems arise and may simplify regulatory audits. 3.8.3 Data Collection HPLC detectors convert a sample peak into a stream of x- (time) and y- (intensity) data, and sometimes an additional z-variable (e.g., wavelength). These data are generated in an analog or digital format that is then sent to the data system for collection and recording. Most data systems can accept either analog or digital signal inputs and most detectors have both analog and digital outputs. For systems in which the detector and data system are from the same manufacturer, the connection may be as simple as a fiber-optic cable. When the data system and detector are not from the same manufacturer, generally an electrical connection of some kind is required. Besides the detector output, the data system may record other system settings, such as the pump flow rate, detector wavelength, system pressure, and other settings. In its most sophisticated form the data system can reconstruct the exact conditions used to analyze each sample, including the column and equipment serial numbers, mobile-phase batch numbers, and all of the chromatographic results. The sampling rate (data rate) for the data system must be sufficiently high so that the collected data accurately represent the peaks; if the data rate is too fast, excessive noise will be collected. A good compromise is to collect ≈20 data points across a peak. For narrow peaks, such as those generated by short, small-particle columns, the data system must be capable of a high enough sampling rate to gather 20 data points across a peak. For example, in Table 3.3, the 50 × 4.6-mm column packed with 1.8-μm particles generates a 30-μLpeakfork = 0.5. When operated at 1-mL/min, this means that the peak is ≈2-sec wide, so a data rate of ≈10 Hz is required. However, when operated in the high-throughput mode, the flow rate may be much higher. At 5 mL/min, the same peak would be just 0.4-sec wide, requiring adatarateof≈50 Hz for adequate sampling. Larger retention times and lower flow rates will generate broader peaks, so the sampling rate does not have to be as fast. Many data systems will adjust the sampling rate during a run so that approximately the same number of data points is collected across each peak. If in doubt, it is always better to collect data at too high a data rate, since data averaging (bunching) to simplify the data set (and reduce noise) can be done during data processing; data 130 EQUIPMENT processing will be unable to create additional data points when the original sampling rate is too slow. See Section 11.2.1 for additional information. 3.8.4 Data Processing Once data are stored by the data system, they can be processed at any future time. Most analysts use a graphical chromatogram for a visual inspection of the quality of the data. As laboratories move to paperless records, the chromatogram may be viewed only on a computer monitor. For qualitative and quantitative analysis (Chapter 11), the collected data are processed to create a simplified output table of retention time and peak area (or height). Additional processing may take advantage of the results of calibration runs to convert the time-and-area results into concentration data (Section 11.4.1). Further processing of the data may provide additional information to the operator, such as UV spectra, MS, or MS/MS data, and other information from specialty detectors, such as light-scattering detectors. 3.8.5 Report Generation Most data systems are PC-based, and as such are capable of running any PC-compatible software. This makes it easy to transfer data from the data processing portion of the data system into Excel ® and other report-generation software. Thus available tables, chromatograms, graphs, statistical results, and other processed data can be incorporated into a formal report. Some software packages may be able to transfer results in an automatic or semi-automatic fashion so that reports are generated automatically. Besides the convenience and time-saving nature of this process, it can reduce transcription and other operator-related errors. If the software programs have been validated, the amount of time spent checking reports for errors can be greatly reduced. 3.8.6 Regulatory Functions As more laboratories become subject to regulatory guidelines by government or other regulatory agencies, the integrity of the chromatographic results will become more important. Specific requirements for electronic records, such as 21 CFR Part 11 [18], require electronic audit trails for data when they are stored electronically—often without paper copies. Data systems that are ‘‘Part 11 compliant’’ have built-in functions that remove the burden of record keeping from the individual operator. For example, such systems require that any manual adjustment of baselines during data processing include a record of the reason for the change, the name of the operator, and a time stamp. Data systems that track all the system settings for each injection can simplify the process of proving the validity of a specific result, when a regulatory auditor reviews the data. It is expected that as additional regulatory requirements are placed on the HPLC laboratory, software manufacturers will continue to provide products that help users comply with the regulations. 3.10 MAINTENANCE 131 3.9 EXTRA-COLUMN EFFECTS The observed peak volume V p in a chromatogram is influenced by band broadening inside the column V p0 , as well as by other system-related factors, V p = (V p0 2 + 4V 2 s + V pl 2 + V det 2 + V ds 2 ) 0.5 (3.1) including the volume contributions of the injection process V s , the tubing and fittings used to plumb the system V pl , the detector cell V det , and the data-system time-constant V ds . The various contributions add as in Equation (3.1) to give the overall peak volume V p andtherelatedpeakwidthW. Peak broadening that results from factors other than the separation (V p0 ) arises from extra-column effects. For a given method setup, extra-column effects will be constant, but the column contribution will vary with k for the solute. In isocratic separation extra-column peak broadening will therefore be more pronounced for early peaks in the chromatogram with smaller values of V p0 . The use of column conditions that generate smaller peak volumes V p0 —short, narrow-diameter, small-particle columns (see Table 3.3)—will make extra-column effects more important. The large peak volumes generated by 150 × 4.6-mm columns packed with 5-μm particles tolerate relatively large sample injections, larger tubing diameters, and larger detector flow cells. However, HPLC systems that rely on sub-2-micron particles in small-volume columns must be specially designed to minimize extra-column effects, or the system will be unable to provide satisfactory results. For more information on the influence of the various contributions to peak broadening, see V p (Section 2.4), V s (Sections 2.6, 3.6.3), V pl (Sections 3.4.1.2, 3.4.2.2), V det (Section 4.2.4), and V ds (Sections 3.8.3, 11.2.3). 3.10 MAINTENANCE The quality of the analytical results obtained from an HPLC system and method depends heavily on the ability of the HPLC system to perform reliably and according to specifications. Three main areas need to be addressed for reliable system operation: • ensure the HPLC system works properly (Section 3.10.1) • prevent as many problems as possible Section 3.10.2) • make efficient and effective repairs (Section 3.10.3) These tasks are covered in the present section. Although no HPLC system is ever problem free, the user can take an active role to minimize problems through the use of some of the techniques described here. 3.10.1 System-Performance Tests Quantitative measurements of HPLC system performance will allow the user to compare performance over time. This can ensure that the instrument works when it is new, help anticipate future problems, and show that a repair was effective. 132 EQUIPMENT 3.10.1.1 Installation Qualification, Operational Qualification, and Performance Qualification One way to demonstrate that a new HPLC system functions properly is to follow the practice of the pharmaceutical industry and perform installation qualification (IQ), operational qualification (OQ), and performance qualification (PQ) tests prior to the release of the system for routine work. The IQ test ensures that the instrument is installed according to the manufacturer’s procedures. IQ often is done by the vendor, if installation is included with the purchase of the system. The documentation accompanying the system will outline the IQ test. OQ demonstrates that the instrument meets the manufacturer’s specifications, or some subset of them. The OQ test results also may be included in the documenta- tion. Alternatively, the performance tests outlined in Sections 3.10.1.2 and 3.10.1.3 may be used to compare the test results to the manufacturer’s specifications (gener- ally found in the back of one or more of the operator’s manuals). The OQ or PQ test often includes a column-performance test (Section 3.10.1.3) to confirm that all the system hardware is working well as a unit. The PQ test generally is user-designed and may range from extensive tests, such as the performance tests described below, to the analysis of a few mock samples to show that the expected results can be obtained for a specific HPLC method. Once these three tests (IQ, OQ, and PQ) have been performed, the system should be ready for routine use. 3.10.1.2 Gradient Performance Test The most important part of the test suite for the evaluation of system performance is a pair of experiments to determine the linearity and accuracy of gradient formation, as well as measure the system dwell-volume. These tests apply to system operation whenever on-line mixing is to be used, for both isocratic or gradient methods. The column is removed, and replaced with a piece of narrow-bore connecting tubing. For example, ≈1 m of 0.005-in (≈0.13-mm) i.d. tubing can be used to connect the injector and detector. This provides sufficient pressure to enable reliable operation of the pump check-valves and results in insignificant dead-volume (≈12 μL) or dispersion of the gradient for most systems. Next water is placed in the A-reservoir, and water that contains 0.1% acetone is placed in the B-reservoir. (An alternative is to use methanol in A and 0.1% acetone in methanol in B, but the same base solvent must be used in both reservoirs.) A UV detector is used, with the wavelength set to 265 nm. Gradient Linearity. The system is programmed first to run a full-range gradient (0–100% B); a 20-minute gradient time is recommended. The flow rate should be set such that the system generates sufficient pressure for reliable check-valve operation; generally, 1 to 3 mL/min will be satisfactory. The autosampler should be in the inject mode, so that mobile phase is pumped through the loop. Because the injector loop normally is in the flow stream during a run (Fig. 3.17b), the loop-volume contributes to the dwell-volume. (If the system is usually run with the loop out of the flow stream, this test should be run with the injection valve in the load mode [Fig. 3.17b] rather than the inject mode.) The test gradient is then carried out. A plot of the baseline should appear as an S-shaped curve, as illustrated by the solid 3.10 MAINTENANCE 133 Figure 3.26 Gradient profile for water to water-acetone gradient. curve of Figure 3.26. This blank gradient can be used as a rough check of gradient linearity, and also measures the system dwell-volume (see below). Gross deviations from gradient linearity can be checked by comparison to a straight line that fits the middle of the gradient profile (dashed line in Fig. 3.26). The actual gradient (solid curve) should be smooth and not deviate from the line except for the slight ‘‘gradient rounding’’ at the beginning and end of the gradient. If deviations are seen, extra attention should be paid to the gradient step-test (see below). For low-pressure-mixing systems, the gradient-proportioning valve test (see below) also will help isolate gradient-linearity problems [19]. If visible deviations from linearity are observed, these often will appear as a distinct shift or angle in the gradient plot (e.g., Fig. 17.23 and Section 5.5.3.3 of [20]). It is wise to carry out additional step-test measurements in the region where deviations are observed. Most HPLC systems can be programmed to generate gradient rates of 10%/min or higher. If the system is to be used with steep gradients—such as for high-throughput applications—or a very different flow rate, the gradient linear- ity can be checked by running the blank gradient test under these conditions. For example, instead of the 20-minute, 2-mL/min gradient of Figure 3.26, it may be more appropriate to test the system with a 5-minute gradient run at 0.5 mL/min if these are more representative operating conditions. Gradient rounding, which usually is not a significant problem for most gradient conditions, tends to be more serious for very steep gradients (or smaller gradient-volumes t G F,wheret G is the gradient time and F is the flow rate; also see Sections 8.1.6.2 and 9.2.2 of [20]). Usually gradient rounding can be reduced by decreasing the system dwell volume (see discussion of Fig. 17.28) while increasing the gradient volume t G F. 134 EQUIPMENT Dwell-Volume Determination. The gradient profile (as in Fig. 3.26) can be used to determine the system dwell-volume. Dwell-volume can be measured by means of one of two techniques. The first method is to extend the linearity test line (dashed line in Fig. 3.26) until it intersects the extended baseline. The time between this intersection and the start of the gradient is the dwell-time, as shown in Figure 3.26. Dwell-time t D can be converted to dwell-volume V D by multiplying by the flow rate F : V D = t D F. This method to determine dwell-volume is simple, but it is subject to any errors that can result from inaccuracy in drawing the linearity-test line through the gradient. It also may be inconvenient to make this measurement directly on a computer monitor from a data system output. A second method for measuring the dwell-volume is less error-prone and more convenient to perform on the computer monitor. This is shown graphically in Figure 3.26. Determine the detector response at the initial baseline (0% B) and at the end of the gradient (100% B). From these two values, locate the point on the plot that the response has reached 50% B and note the time t 1/2 it took to reach this point. The dwell time t D equals t 1/2 minus half the gradient time (e.g., 10 minutes for a 20-minute gradient). The dwell-volume equals t D F. Gradient Step-Test. This test uses the same system setup and A- and B-solvents. The gradient step-test determines the accuracy of solvent proportioning for selected solvent (or mobile-phase) mixtures. If the system also is used for isocratic methods, this test will check the accuracy of on-line mixing. The system controller is set to deliver a series of solvent mixtures in a stair-step design. A good choice for this test is to use a 10% step size, so that mixtures of 0, 10, 20, , 80, 90, 100% B are formed—each for an interval of 3 minutes. Problems are encountered most commonly near 50% B, so an additional step at 45% B and 55% B should be added for a total of 13 steps. The remaining conditions are the same as those described for the gradient linearity test (above). The results for the 40–60% B portion of this test for a well-behaved system are shown in Figure 3.27. Next calculate actual %B for each step (as in Fig. 3.27) by measuring its height from the baseline (0% B) and dividing by the distance between the 100% B step and the 0% B step. The %B for each step should compare favorably to the programmed Figure 3.27 Results of a gradient step-test. Measured response (%B) shown below each step. 3.10 MAINTENANCE 135 value for the step. For example, the 40% B setting in Figure 3.27 actually delivered 39.95% B. Typically manufacturers specify accuracy of ±0.5–1% B throughout the mixing range. For example, in Figure 3.27, the 55% B step actually delivered 54.04% B—(barely) within a ±1% criterion. For applications with gradient rates of ≥1%/min, accuracy of ±1% usually is sufficient. When shallower gradients are used, smaller deviations may be required. It may also be possible to improve proportioning accuracy by premixing solvents. For example, the proportioning accuracy of a 15–25% B gradient can be improved 10-fold from ±1% to ±0.1% by the replacement of 100% aqueous solvent in reservoir A with hand-mixed 15/85 organic/aqueous, and 100% organic in B with 25/75 organic/aqueous, plus revision of the program to generate a 0–100% B gradient instead of 15–25% B. The accuracy of on-line mixing for isocratic elution can be similarly improved with this technique. Gradient Proportioning Valve (GPV) Test. A third test with a similar system setup as above is useful for low-pressure-mixing systems but does not apply to high-pressure mixing. This test checks the accuracy of the proportioning-valve system and its associated control software. For example, consider a four-solvent system (A, B, C, and D). The A and B inlet lines are placed in a reservoir that contains water and the C and D lines are placed in a reservoir that contains 0.1% acetone in water. The baseline is generated by delivery of a 50/50 mixture of A and B. The various combinations of solvents are checked by generation of blends of 90% of A or B with 10% of C or D. For example, the test results shown in Figure 3.28a (for an acceptable test result) are for the sequence shown in the caption. The height above baseline of each 90/10 plateau is measured. The difference between the height of the highest and lowest plateaus is divided by the average plateau height to determine the %-range for the various proportioning valve combinations. A plateau range of ≤2% is usually acceptable, although ranges of ≤1% are common for well-behaved systems (Fig. 3.28b; see Fig. 17.25 for an example of a failed GPV test). 3.10.1.3 Additional System Checks In addition to the accuracy and linearity of gradient formation, other factors affect LC-system reliability. The tests listed below can be used on a periodic basis (e.g., annually or semiannually) to help ensure that the system is operating properly. Typical performance values for these tests are listed in Table 3.4. Flow-Rate Check. Although a small change in the flow rate F usually results in only minor changes in separation, large errors in F can be more serious. Con- sequently a check of flow-rate accuracy on a periodic basis is recommended. The second-by-second flow-rate accuracy during a run is difficult to measure without specialized equipment (and is not important), but a longer term volumetric check of flow rate can be made easily by carrying out a timed collection of mobile phase in a 10-mL volumetric flask at a flow rate of 1 mL/min under isocratic conditions. For high-pressure-mixing systems the flow rate can depend on solvent compressibility, so it is best to check the flow for representative solvents. For example, check the flow of 100% A with water and 100% B with acetonitrile or methanol. Typical manufacturer’s system specifications are ±1% for flow-rate accuracy. A measured flow-rate accuracy should fall within this range for routine operation. In Table 3.4 . a graphical chromatogram for a visual inspection of the quality of the data. As laboratories move to paperless records, the chromatogram may be viewed only on a computer monitor. For qualitative. result, when a regulatory auditor reviews the data. It is expected that as additional regulatory requirements are placed on the HPLC laboratory, software manufacturers will continue to provide products. through the laboratory. Some of these functions will help with regulatory compliance (Section 3.8.6 and Chapter 12), some are used to help coordinate work activities, and some may be used to support