156 DETECTION N S Figure 4.7 Measurement of chromatographic signal (S) and noise (N). from the middle of the baseline noise to the top of the peak (Fig. 4.7). The contribution of S/N to precision can be estimated as CV = 50 S/N (4.1) where CV is the coefficient of variation (equivalent to %-relative standard deviation, %-RSD). The lower limit of detection (LLOD) often is described as S/N = 3, which would give CV ≈16%, whereas the lower limit of quantification (LLOQ) is S/N = 10, for CV ≈5% (see also Section 11.2.4). These values of CV are the contribution of S/N to the overall imprecision of the method, so the overall method precision is expected to be worse than the S/N contribution. As long as the imprecision attributable to S/N (or any other single contributor to error) is less than half of the desired method imprecision, S/N will have a minor (<15%) influence on the overall method precision (see the discussion of Eq. 11.2). For example, if the overall method requires imprecision of no more than 2%, a contribution of S/N of <1% should be satisfactory. This suggests that a S/N value of ratio 50:1 or more is required for an overall method imprecision of <2%. The signal-to-noise ratio can be improved by increasing the signal, reducing the noise, or both, as summarized in Table 4.1. An increase in signal for a given peak or sample may be available from a change in detector setting; for example, the use of a UV wavelength that corresponds to maximum sample absorptivity. A more sensitive detector also may be available, of either the same or different type. Derivatization (Section 16.12) or other modification of the analyte may make it more responsive to the chosen detector. A more common means of increasing the signal is to inject a larger weight of sample (either a larger sample volume or a sample concentrate; Section 3.6.3). However, column or detector overload will eventually limit the possible increase in signal in this way. Any reduction in peak width should translate into a proportional increase in peak height (area is assumed to be constant); smaller k-values (increase in %B; see examples of Fig. 2.10b), narrow-diameter columns, or more efficient small-particle columns can each be used for this purpose. 4.2 DETECTOR CHARACTERISTICS 157 Table 4.1 Improvement of Signal-to-Noise Ratio Increase Signal Decrease Noise Better wavelength (or other detector adjustment) Increase detector time constant More sensitive detector More data-system signal averaging Analyte derivatization Better temperature control Inject larger weight of sample Higher reagent/solvent purity Reduce peak width (volumetric) Better sample cleanup Smaller k Constant, pulse-free flow Smaller column volume Column switching Larger plate number Any reduction of baseline noise also can improve S/N, for example, signal smoothing by an increase in the detector time-constant or data-system sam- pling rate (Section 4.2.3.1). Excessive smoothing, however, can reduce the signal intensity. Better temperature control of the column, detector, and general instru- ment environment also can reduce noise, especially for detectors sensitive to refractive index changes. Purer solvents (e.g., HPLC grade) and better sam- ple cleanup can reduce the introduction of noise-generating contaminants. For gradient applications, changes in the system are sometimes attempted in order to reduce the dwell-volume (Section 9.2.2.4) and the gradient delay time. The mixer-volume comprises a major fraction of the dwell-volume in many systems, but reduction of the mixer-volume can increase baseline noise. Some HPLC sys- tems have optional mixers that can be added to smooth the baseline and reduce noise—these devices can be especially advantageous for isocratic methods run at maximum detector sensitivity. Column switching (Section 16.9) can be used to transfer a desired fraction from a cleanup column to the analytical column, thereby diverting unwanted contaminants to waste, so as to reduce baseline noise. 4.2.4 Detection Limits Although the signal-to-noise ratio is a measure of the inherent quality of the detector signal, the minimum detectable mass or concentration often is the limiting factor in the usefulness of a detector for a particular application. The term sensitivity often is used interchangeably with detection limit when describing an HPLC detec- tor. However, in proper use, sensitivity is the slope of a calibration plot, that is, the change in signal per unit change in concentration (or mass) of analyte, whereas detection limit refers to the minimum concentration (or mass) that can be measured. HPLC detectors respond either to the concentration of the sample in the detector cell (e.g., UV detection) or the mass of sample in the detector (e.g., LC-MS). Detection limits, discussed more thoroughly in Section 11.2.5, are defined as follows: The limit of detection LOD is the smallest signal that can be discerned 158 DETECTION from the noise—with confidence that a peak really is present. Often a S/N of 3 is equated to the LOD.Thelower limit of quantification LLOQ (sometimes called limit of quantification or limit of quantitation, LOQ) is the smallest signal that can be measured with the required precision for the method. The LLOQ often is defined as S/N ≥ 10, but a value of S/N ≥ 50 may be chosen for high-precision methods. There is a never-ending need for lower and lower detection limits for trace analysis, and assays for which on-column injections of <1 ng are becoming more and more common. The LOD and LLOQ are directly related to the concentration (or mass) of sample in the detector cell. Thus a longer path-length cell for UV detection is favored in terms of signal intensity. However, the detector cell should be designed with a minimum volume that is compatible with other requirements of the detector. Excess cell volume will result in additional extra-column peak broadening (Section 3.9). This is especially true for small-volume columns, columns packed with small particles, and peaks with k<2. For example, with a 50 × 4.6-mm column packed with 3-μm particles and k<3, significant peak broadening was observed for an 8-μL UV-detector cell when compared with a 1-μL cell [14]. To minimize the broadening of early-eluted peaks, the detector cell volume V det should be less than approximately one-tenth of the final volume of the peak of interest V p (V p = WF,whereW is the baseline width of the peak [min], and F is flow rate [mL/min]) [15]: V det < 0.1V p (4.2) (For other peak-broadening contributions to V p , see Eq. 3.1 in Section 3.9.) Some examples of the column contribution to peak volume V p0 for early-eluted peaks (k = 2) for some popular column configurations are shown in Table 4.2. (In a well-behaved system, according to Eq. 2.27 and 3.1, the observed peak volume V p should not be much larger than V p0 .) Table 4.3 lists the detector cell volumes for several UV-detector configurations. For UV detectors (Section 4.4), signal intensity is proportional to path length, so longer path flow cells will have lower detection limits. However, for detector cell diameters <1 mm, signal loss due to light scattering in the cell can be a problem, so special cell designs (e.g., total internal reflectance) are necessary for smaller cell diameters (see the discussion of Section 4.4). The data of Tables 4.2 and 4.3 show that column lengths L ≥ 100 mm with a diameter d c = 4.6 mm, packed with 5- or 3-μm d p particles, will work well with the standard 10 × 1.0-mm UV cell (see (Eq. 4.2), but any combination of smaller column dimensions or smaller particles requires smaller cell volumes to avoid unnecessary extra-column peak broadening. (Note that Eq. 3.1 is an approximation, so peak-broadening calculations based on Eq. 3.1, and therefore conclusions based on Tables 4.2 and 4.3, also are approximations.) 4.2.5 Linearity For quantitative analysis by HPLC (Section 11.4), the detector response must be related to the amount of analyte present. If analyte response y is plotted against analyte concentration x, the simplest, most convenient, and most reliable relationship is y = mx, where the slope m is a constant defined as the sensitivity. Such a relationship between analyte response and analyte amount is termed linear. 4.2 DETECTOR CHARACTERISTICS 159 Table 4.2 Typical Peak Volumes V p0 L (mm) d c (mm) d p (μm) V p0 (μL) a 250 4.6 5 212 150 4.6 5 164 3 127 2.1 3 26 100 4.6 3 104 2.1 3 22 1.0 3 5 24 50 4.6 3 73 2.1 3 15 2.1 2 12 1.0 2 3 a Assumes k = 2; reduced plate height h = 2.5(seeEqs.2.27and3.1). Table 4.3 UV-Detector Cell Volumes Path Length (mm) Inner Diameter (mm) Volume (μL) 10 1.0 8 0.5 2 0.25 0.5 514 0.5 1 110.8 0.5 0.2 For best use over a wide range of sample concentrations, a wide linear dynamic range (the concentration range over which the detector output is proportional to analyte concentration, e.g., 10 5 for UV detection) is desired, so that both major and trace components can be determined in a single analysis over a wide concentration range. For example, with a stability-indicating method, peaks ≥ 0.1% of the response of theactiveingredient(= 100%) must be reported, which would require a linear range of at least 100/0.1 = 10 3 . Some detectors (e.g., evaporative light scattering) have a narrow linear range of 1 to 2 orders of magnitude. Although less convenient and reliable, a nonlinear calibration curve (e.g., quadratic) can be used—as long as the detector response changes in a predictable manner with sample concentration (or mass). 160 DETECTION Table 4.4 HPLC Detectors Sample-Specific (Sections 4.4–4.10) Bulk Property (Sections 4.11–4.13) Hyphenated (Sections 4.14–4.15) Reaction (Section 4.16) UV-visible Refractive index Mass spectrometric Reaction Fluorescence Light scattering Infrared Electrochemical Corona discharge Nuclear magnetic resonance Radioactivity Conductivity Chemiluminescent nitrogen Chiral 4.3 INTRODUCTION TO INDIVIDUAL DETECTORS The remainder of this chapter (Sections 4.4–4.16) provides a discussion of most HPLC detectors in use today. In Table 4.4, detectors are grouped by technique (sample specific, bulk property, etc.) in approximate order of popularity within each group. Sample-specific detectors will be treated first, and reaction detectors last—with only limited discussion of less-used detectors. Within each section, principles of detector operation are discussed first, followed by one or more example applications. Where appropriate, a comparison with other detectors is included. A detailed discussion of every detector is beyond the scope of this book. In addition to the references cited in each section, a more general discussion of HPLC detectors can be found in [16, 17]. 4.4 UV-VISIBLE DETECTORS The most widely used detectors in modern HPLC are photometers based on ultravio- let (UV) and visible light absorption. These detectors have a high sensitivity for many solutes, but samples must absorb in the UV (or visible) region (e.g., 190–600 nm). Sample concentration in the flow cell is related to the fraction of light transmitted through the cell by Beer’s law: log  I o I  = εbc (4.3) where I o is the incident light intensity, I is the intensity of the transmitted light, ε is the molar absorptivity (or molar extinction coefficient) of the sample, b is the cell path-lengthincm,andc is the sample concentration in moles/L. Light-absorption 4.4 UV-VISIBLE DETECTORS 161 HPLC detectors usually are designed to provide an output in absorbance that is linearly proportional to sample concentration in the flow cell, A = log  I o I  = εbc (4.4) where A is the absorbance. Properly designed UV detectors are relatively insensitive to flow and temper- ature changes. UV photometers that are linear to > 2 absorbance units full scale (AUFS) with <1 × 10 −5 AU noise are commercially available. With this perfor- mance, solutes with relatively low absorptivities can be monitored by UV, and it is possible to detect a few nanograms of a solute having only moderate UV absorbance. The wide linear range of UV detectors (≈10 5 ) makes it possible to measure both trace and major components in the same chromatogram. UV detectors commonly use flow cells of the Z-path design of Figure 4.3a, with a 1-mm diameter and 10-mm path length (for a volume of ≈8 μL). This cell volume is adequate for ≥100 ×≥4.6-mm columns packed with ≥3-μm particles (Section 4.2.4), but smaller volume and/or smaller particle columns may experience unacceptable extra-column peak broadening in an 8-μL cell. Shorter path-length cells will reduce the cell volume, but the signal is proportional to the path length (Eq. 4.4)—so sensitivity must be balanced against extra-column peak broadening in choosing the flow cell dimensions. If the refractive index (RI) within the cell changes (e.g., during gradient elution), the amount of energy reaching the photodetector can change; when a light ray hits the side of the flow cell, the ratio of reflected to absorbed light depends on the refractive-index ratio of the mobile phase and cell wall (and the angle of the light ray hitting the cell wall). The latter refractive-index effect plus imperfections in optical alignment make it difficult to successfully use cell diameters smaller than ≈1 mm. One innovation that can minimize this problem is a flow cell design as in Figure 4.3b, where the internal surface of the flow cell is coated with a reflective coating—light that strikes the sides of the flow cell is reflected so as to still reach the photodetector [18, 19]. The use of this light-pipe technique allows the cell diameter to be reduced for smaller cell volumes (e.g., 0.25 mm × 10 mm ≈0.5 μL), and thus less peak spreading for use with very small-volume, small-particle columns. Alternatively, a longer, narrower diameter flow cell can be used (increasing b in Eqs. 4.3 and 4.4) for more absorbance in a smaller volume cell (e.g., 0.25 mm i.d. × 50 mm long, with a volume of ≈2.5 μL). It is not necessary to operate a UV detector at the absorption maximum of the analyte. A hypothetical example of wavelength selection is shown in Figure 4.8. The spectra for two analytes, X and Y, are shown in Figure 4.8a, with UV maxima at ≈250 nm and ≈270 nm, respectively. At 280 nm, Y has much stronger absorbance, so it has a much larger peak (Fig. 4.8b, same mass on column). At 260 nm, the absorbances of X and Y are approximately equal, so the peaks are of approximately equal size (Fig. 4.8c). At 210 nm, both compounds have even stronger absorbance and generate much larger peaks (Fig. 4.8d). Notice also the appearance of a new peak Z, which was not observed at higher wavelengths. This general increase in sensitivity at lower wavelengths is one reason for the widespread use of ≤ 220 nm for detection (near-universal detection). The corresponding loss of detector selectivity at lower wavelengths can be a disadvantage for other separations, where it might 162 DETECTION 012345 Time (min) 280 nm 260 nm 210 nm XY X Y X Y Z 210 220 230 240 250 260 270 280 Wavelength (nm) Absorbance (a) (b) (c) (d ) X Y Figure 4.8 Wavelength selectivity for UV detection. (a) Absorbance spectra for two hypo- thetical compounds X and Y. Chromatograms at (b) 280 nm, (c) 260 nm, and (d) 210 nm. be undesirable to ‘‘see’’ certain sample constituents (e.g., arising from the sample matrix). UV-visible spectrophotometric detectors can respond throughout a wide wave- length range (e.g., 190–600 nm), which enables the detection of a broad spectrum of compound types. Almost all aromatic compounds absorb strongly below 260 nm; 4.4 UV-VISIBLE DETECTORS 163 lamp filter beam splitter sample & reference flow cells photocells Figure 4.9 Schematic of a fixed-wavelength UV detector. Dashed lines show optical path. compounds with one or more double bonds (e.g., carbonyls, olefins) can be detected at wavelengths of <215 nm, while the preponderance of aliphatic compounds possess significant absorbance at ≤205 nm. Reversed-phase mobile phases of ace- tonitrile plus water or phosphate buffer can be used routinely for detection at 200 nm, whereas methanol-containing mobile phases cannot be used below ≈210 to 220 nm, depending on methanol concentration (see Appendix I, Table I.2). The proper selection of the mobile phase makes it possible to operate UV detectors in a near-universal detection mode in the 200- to 215-nm region, where most organic compounds exhibit some UV absorbance. Because of the relatively small absorbance differential between water (or phosphate buffer) and acetonitrile at > 200 nm or methanol at > 220 nm, UV detectors are also quite useful for gradient elution. Mobile phases with large differences in UV absorbance, such as tetrahydrofuran and water at <240 nm, may not be amenable for use with gradients and UV detection. UV detectors come in three common configurations. Fixed-wavelength detec- tors (Section 4.4.1) rely on distinct wavelengths of light generated from the lamp, whereas variable-wavelength (Section 4.4.2) and diode-array (Section 4.4.3) detec- tors select one or more wavelengths generated from a broad-spectrum lamp. 4.4.1 Fixed-Wavelength Detectors Figure 4.9 is a generic schematic of a fixed-wavelength UV detector. These detectors were the mainstay of UV detection prior to the introduction of the variable- and diode-array detectors, but they are not widely used today. Their current appeal is low price and simple construction, and they tend to be more popular in the educational environment or other budget-limited settings. UV radiation at 254 nm from a low-pressure mercury lamp passes through a band-pass filter and beam splitter, and shines on the entrance of the flow cell. Light transmitted through the flow cell strikes the photodetector (usually a photodiode) and is converted to an electronic signal. Most UV detectors operate in a differential absorbance mode, where light also passes through a reference cell, and the difference between the light passing through the sample and reference cells is converted to absorbance, according to Equation (4.4). Although some detectors enable the reference cell to be filled with mobile phase, an air reference is most commonly used, which allows for correction of variations in light intensity from the source lamp, but not for changes in the mobile-phase absorbance. 164 DETECTION lamp diffraction grating slit flow cell photocell 200 nm 360 nm Figure 4.10 Schematic of a variable-wavelength UV detector; reference flow cell not shown. Dashed lines show optical path. The 254-nm line from the low-pressure mercury lamp is the most popular wavelength for use with the fixed-wavelength UV detector. For historical reasons this wavelength is still popular for applications that use variable- and diode-array detectors, although there is no real reason to use this particular wavelength. Through the use of other lamps (e.g., zinc), phosphors, and other lines in the mercury lamp output, detection at 214, 220, 280, 313, 334, and 365 nm can be accomplished with the fixed-wavelength detector. 4.4.2 Variable-Wavelength Detectors UV spectrophotometers (variable-wavelength and diode-array detectors) offer a wide selection of UV and visible wavelengths. Such devices have the versatility and convenience of operation at the absorbance maximum of a solute or at a wavelength that provides maximum selectivity, as well as the ability to change wavelengths during a chromatographic run. The most widely used detector in HPLC today is the variable-wavelength UV detector shown schematically in Figure 4.10. A broad-spectrum UV lamp (typically deuterium) is directed through a slit and onto a diffraction grating. The grating spreads the light out into its component wavelengths, and the grating is then rotated to direct a single wavelength (or narrow range of wavelengths) of light through the slit and detector cell and onto a photodetector. These detectors usually use a sample and reference cell configuration (Section 4.4) for differential detection. For detection in the visible region, a tungsten lamp is used instead of deuterium. The use of a variable-wavelength detector allows one to program a change in the detection wavelength during a chromatogram. Thus one peak can be detected at 280 nm and another at 220 nm. Although it is possible for many detector models to change the wavelength quickly, so as to generate a UV spectrum for a peak, the results are complicated by the change in analyte concentration during the spectral scan and may be of limited value. 4.4 UV-VISIBLE DETECTORS 165 lam p slit flow cell photodiode array diffraction grating 200 nm 360 nm Figure 4.11 Schematic of a diode-array UV detector. Dashed lines show optical path. 4.4.3 Diode-Array Detectors A schematic of the diode-array detector (DAD, also called photodiode-array, PDA) is shown in Figure 4.11; it has a similar optical path to the variable-wavelength detector, except that the white light from the lamp passes through the flow cell prior to striking the diffraction grating. This allows the grating to spread the spectrum across an array of photodiodes, hence the name photodiode array (PDA). The number of photodiodes varies with the specific brand and model of detector, but detectors with 512 or 1024 diodes are common. The signals from the individual photodiodes are processed to generate a spectrum of the analyte. Because the spectra are generated at the same time (vs. single-wavelength monitoring with the variable-wavelength detector), the DAD can contribute to peak identification. The DAD can be operated to collect data at one or more wavelengths across a chromatogram, or to collect full spectra on one or more analytes in a run. Of course, the data-file size is much larger for full-spectra runs, but data compression techniques and inexpensive data storage make this less of a concern than it was in the past. If two closely eluted peaks have sufficiently different spectra, it may be possible to distinguish the two peaks spectrally. The utility of the DAD to distinguish between two peaks can be understood in conjunction with Figure 4.12, where a partial chromatogram for a closely eluted peak pair X and Y is shown (Fig. 4.12a). If the solutes have spectra as shown in Figure 4.12b and are monitored at a wavelength where both have significant absorbance, such as 260 nm, the resulting chromatogram will look like a single peak (X + Y in Fig. 4.12a); the corresponding peaks are shown for the solutes injected individually. Even though the peaks appear to overlap completely at 260 nm, if other wavelengths are monitored, it may be possible to distinguish between the peaks. For example, if 240 nm is used, only X will respond, whereas only Y will respond at 280 nm. The added selectivity of the detector can be used to compensate, at least in part, for inadequate chromatographic separation. Thus the DAD could simultaneously collect data at 240 and 280 nm during the chromatographic run, and individual chromatograms plotted at 240 or 280 nm would allow quantification of X and Y, even under the partially overlapped conditions of Figure 4.12. Another common application of the DAD is for peak-purity determination. The software accompanying the DAD accomplishes this by calculating an absorbance . magnetic resonance Radioactivity Conductivity Chemiluminescent nitrogen Chiral 4.3 INTRODUCTION TO INDIVIDUAL DETECTORS The remainder of this chapter (Sections 4.4–4.16) provides a discussion of most HPLC detectors in use today. In Table 4.4, detectors are grouped. monitoring with the variable-wavelength detector), the DAD can contribute to peak identification. The DAD can be operated to collect data at one or more wavelengths across a chromatogram, or to. Chromatograms at (b) 280 nm, (c) 260 nm, and (d) 210 nm. be undesirable to ‘‘see’’ certain sample constituents (e.g., arising from the sample matrix). UV-visible spectrophotometric detectors can