Instrumentation - American Chemical Society


Instrumentation - American Chemical Societypubs.acs.org/doi/pdfplus/10.1021/ac00266a878Similara> T. This is the most...

0 downloads 118 Views 9MB Size

Instrumentation M. R. Sharpe, Pye Unicam Ltd. Cambridge CB1 2PX United Kingdom

Stray Light in UV-VIS Spectrophotometers Ultraviolet-visible (UV-VIS) spec­ trophotometers have been used as a tool for chemical analysis for many years. They find application in many fields, including industry, medicine, and university research. It is therefore important that such instruments be reliable in operation, convenient to use, and very accurate. Since spectro­ photometers were first produced com­ mercially their performance has stead­ ily improved. Most currently avail­ able instruments now meet the above requirements, so t h a t analysts can apply them to their work with confi­ dence. In recent years one particularly no­ ticeable change in these instruments has been the use of microprocessors to give much greater flexibility and con­ venience of operation, as well as allow­ ing automation of many analyses. Less noticeably, there have been significant improvements in optical components, which have resulted in longer instru­ ment life and greatly improved photo­ metric accuracy, particularly because the effects of "stray light" have been significantly reduced. Not many years ago the UV stray-light specification for a spectrophotometer was usually about 1%. Now, however, figures of less than 0.01% are often quoted, 0003-2700/84/0351 -339A$01.50/0 (c) 1984 American Chemical Society

which is such a low level of stray light that its effects can be ignored for most analyses. Not all analysts, however, have access to the latest instruments, and consequently there are in use a wide variety of spectrophotometers which vary greatly in age and perfor­ mance. All users of spectrophotome­ ters should be aware of the practical limitations of their instruments, even if they are the latest models. This article will show how stray light can affect analytical accuracy and will discuss the sources of stray light and how recent improvements in optical components have benefited in­ strument performance and life. Stray Light and Its Effects on Measurement Accuracy Figure l a shows the basic compo­ nents of a simple single-beam spectro­ photometer. T h e monochromator en­ trance slit is illuminated by a light source. T h e monochromator is adjust­ ed to the analytical wavelength, and the light emerging from the exit slit is passed through a sample. T h e light flux transmitted by the sample is mea­ sured by a detector, usually a photo­ cell, photodiode, or photomultiplier. If /„ = light flux incident on sample and / = light flux transmitted by the

/

/

>

X\x VX Sχ "\.

1

\

sample, then the transmittance Τ at the analytical wavelength is Τ =

h

(1)

or in terms of absorbance units, which are used because the absorbance A is proportional to the product of the sample path length and concentration, A = -log10T

(2)

If measurements are made over a range of wavelengths, then the absorb­ ance spectrum of the sample can be obtained. T h e monochromator acts as a filter

ANALYTICAL CHEMISTRY, VOL. 56, NO. 2, FEBRUARY 1984 · 339 A

bandwidth, centered at the analytical wavelength, and a proportion a of the stray light from outside the spectral bandwidth, then the total transmitted light flux is TI„ + alx. The measured transmittance TM of the sample at the analytical wave­ length is then TI„ + ctl. (3) TM —

l„ + h

Putting the fractional stray light S as S =

h + Is

(4)

gives

Figure 1. (a) A single-beam spectrophotometer; (b) spectral bandwidth

for the light source and passes light in a small band of wavelengths centered at the required wavelength. This small range of wavelengths is determined by the monochromator spectral band­ width, which is proportional to the monochromator slit widths. When the entrance and exit slits are of equal width, the bandwidth has approxi­ mately a triangular energy profile. The spectral bandwidth is customarily defined as the width of the triangular profile at half maximum energy, as shown in Figure lb. In practice a monochromator is not a perfect device, and it can transmit a small flux of light over the entire wavelength range of the light source.

This "unwanted" component of light flux outside the spectral bandwidth is known as stray light, and it can cause serious measurement errors for the unwary analyst. Because the light transmitted by most samples varies with wavelength, the proportion of stray light transmit­ ted by a sample will not be equal to the sample transmittance Τ at the an­ alytical wavelength. If the "wanted" light flux incident on the sample with­ in the monochromator spectral band­ width is /„ and the unwanted stray light flux is /.,, then the total light flux incident on the sample is /„ + /,. If the sample transmits a proportion Τ of the light within the spectral

Figure 2. The effect of stray light on absorbance measure­ ments for several stray-light levels

TM = T+ S(a T) (5) It is thus clear that stray light can give rise to a difference between the true sample transmittance Τ and the mea­ sured transmittance TM. To illustrate the practical effects of stray light some particular cases will be examined. a> T. This is the most common case in practice. In quantitative analy­ sis one usually measures the absorb­ ance of a sample at a peak of absorb­ ance, when the sample transmittance is comparatively small. In conse­ quence a — Τ in Equation 5 is posi­ tive, most of the stray light being transmitted. If one takes the rather ideal case when a = 1 and all the stray light is transmitted by the sample, then Fig­ ure 2 shows the measured absorbance as a function of the true absorbance, for several values of stray light. In the absence of stray light, the concentra­ tion of a sample is proportional to the measured absorbance (Beer-Lambert law). However, an instrument which, for example, has 0.1% stray light shows measurable deviations from lin­ earity at only 1.5 A, where A = ab­ sorbance units. At higher concentra­ tions there is an increasing loss of sen-

Figure 3. Absorbance error for several stray-light levels

340 A · ANALYTICAL CHEMISTRY, VOL. 56, NO. 2, FEBRUARY 1984

sitivity and, consequently, measure­ ment precision, until at 3 A increasing sample concentration produces no fur­ ther increase in absorbance. To show these effects more clearly Figure 3 shows the absorbance error as a func­ tion of the true absorbance for several values of stray light. a < T. This is a more unusual ex­ ample and corresponds to a sample that absorbs relatively little light at the measured wavelength, but absorbs most of the stray light. This results in a — Τ being negative, and conse­ quently the measured absorbance is higher than the true absorbance. A practical example is the spectrum of benzene vapor at about 250 nm, which is sometimes used as a spectral resolu­ tion test for instruments. Stray light can cause the absorbance minima be­ tween the benzene absorption peaks to be partially "filled in" by the positive absorbance error, apparently degrad­ ing the spectral resolution. α =ϋ Τ. If the sample has the same transmittance outside the spectral bandwidth as at the analytical wave­ length, then (« — Τ) ~ 0, and there is no absorbance error. Calibrated neu­ tral glass filters, used to measure ab­ sorbance accuracy of spectrophotome­ ters, are an example of such a materi­ al, which has been deliberately chosen to eliminate the effects of stray light when checking photometric accuracy of an instrument. In conclusion, it should be noted from these examples and from Equa­ tion δ that the effective amount of stray light is highly dependent on the absorption spectrum of the sample being measured. Instrument manufac­ turers usually specify the stray light where its effects are greatest in the UV, using a particular form of test sample to be discussed later. The ef­ fective stray light with other samples will usually be less than the specifica­ tion figure, but the analyst should be aware that this depends on the sample spectrum (1,2). Origins of Stray Light Figure 4 shows a typical diffraction grating monochromator. The entrance slit is illuminated by the light source, and light from this slit is focused to a parallel beam by the collimating mir­ ror, this beam being incident on the grating. The grating is rotated to dif­ fract light of the required wavelength onto the focusing mirror, which in turn focuses it onto the exit slit. The main source of stray light in most spectrophotometers is usually the dispersing element in the mono­ chromator, either a prism or a diffrac­ tion grating. Scattering of light and unwanted reflections from other opti­ cal elements can also contribute sig­ nificantly to the stray light, depending

. Collimating Mirror

From Light — » Source

• Focusing Mirror

Scattering at Walls from Other Orders

Figure 4. A diffraction grating monochromator, illustrating the sources of stray light

on the relative quality of the dispers­ ing element and how carefully the monochromator has been designed and internally baffled. The strong zero-order spectrum can be particular­ ly troublesome when reflected and scattered at walls and mirrors. These various sources of stray light are indi­ cated in Figure 4. Good monochroma­ tor design depends on ensuring that the mirrors and dispersing element are of high quality with little scatter and that scattered light is minimized by baffling so that it cannot reach the exit slit. In addition, light can be dif­ fracted twice or more by the grating on reflection from the mirrors. This can also be avoided by careful design and baffling. Higher order spectra are usually removed by suitable filters. It is also possible for stray light to arise outside the monochromator, such as from light leaks in the instru­ ment, allowing some light directly to the sample or detector from outside the instrument or directly from the light source. However, in a well-de­ signed, well-constructed instrument this latter source of stray light should be completely negligible and will not be further considered. Most modern instruments now use a diffraction grating as a dispersing ele­ ment in the monochromator, as prisms in general have a poorer stray light performance and require complicated precision cams to give a linear wave­ length scale. Replica gratings can now be produced more cheaply than prisms and require only a simple sine bar mechanism for the wavelength scale.

342 A · ANALYTICAL CHEMISTRY, VOL. 56, NO. 2, FEBRUARY 1984

A diffraction grating of the type used in a UV-VIS spectrophotometer consists of a glass or silica substrate on which there is a layer of resin for a ruled replica grating or a layer of pho­ toresist for a holographic (or interfer­ ence) grating. This surface is covered with fine parallel grooves produced by the relevant manufacturing process and is finished with a reflecting layer of aluminum, which in modern instru­ ments is also covered with a trans­ mitting protective film to prevent oxi­ dation or contamination of the alumi­ num. The profile of the grating grooves is usually a shallow triangle, with the wide faces of each groove tilt­ ed at an angle known as the blaze angle, which results in the grating having maximum diffraction efficien­ cy at a certain wavelength, usually in the UV. Figure 5 shows electron mi­ crographs of some gratings, and the slightly inclined wide groove faces can be clearly seen. A typical reflection grating in a UV-VIS spectrophotometer may have 1200 grooves/mm, which means the grooves are spaced at about 800-nm intervals. The grating may have a width of 20 mm or more, giving a total of at least 24 000 grooves. To obtain constructive interference across this number of grooves with little light scattering, the spacing and form of the grooves must be accurate to within a few nanometers to give a high-quality grating. Mechanical diamond ruling of gratings to near this accuracy is one of the marvels of modern technology. Holographic gratings, generated from a laser interference pattern, have ex-

(b)

(a)

Figure 5. (a) A conventional ruled grating; (b) a blazed holographic grating

tremely precise groove spacings and smoother groove surfaces, resulting in an order of magnitude or more lower stray light than can be achieved by mechanical ruling. In consequence, most modern instruments now use ho­ lographic gratings in preference to ruled gratings. For further informa­ tion on gratings the reader will find a wealth of literature on the subject— for example, Reference 3. Quantitative Aspects of Stray Light

If the entrance slit of a monochromator is illuminated with monochro­ matic light, for example by a laser,

then as the instrument wavelength setting λ/ is scanned through the monochromatic wavelength Xc the en­ ergy from the exit slit appears as shown in Figure 6. This shows the out­ put peaking at both the first and sec­ ond orders of diffraction, the first order being that normally used. For a perfect monochromator one would ex­ pect the curve to be the triangular bandwidth function shown previously as Figure lb. This is shown as a dashed line. The excess energy outside the bandwidth function is the small fraction of the monochromatic light scattered in the monochromator to

Relative Energy R(\,, K)

Wavelength Setting λ;

give stray light at other wavelength settings. It is convenient to consider the energy from the monochromator shown as Figure 6 as a function R(X/,XC), the ratio of the energy at wavelength setting λ/ to that when set to the first-order wavelength \c. The function R(X/,XC) can therefore be put as

R(Xj,Xc) = RAhX)

((6) where R„(\j,\c) = energy within tri­ angular bandwidth and Rs(\/,\c) = stray light energy Curves of the type shown as Figure 6, obtained for a series of monochro­ matic wavelengths, can be used to as­ sess the stray-light performance of a monochromator when it is used with a continuum light source, as is normally the case in a spectrophotometer. In practice, however, it is difficult to re­ late these results to practical measure­ ments of transmittance or absorbance on a sample, as so many other factors are involved in a complete spectropho­ tometer. In addition, as discussed ear­ lier, the stray-light errors depend very much on the absorption spectrum of the sample. Figure 7 illustrates how the various components of a spectrophotometer influence its performance. If 7(X) = source spectrum E(\) = grating efficiency M(X) = mirror reflectance P(X) = detector sensitivity then the output voltage from the de­ tector is proportional to the product of all these factors, so that at a wave­ length setting X/ the detector output D(X/) is given by D(\,)

Figure 6. Monochromator energy output when illuminated by a monochromatic light source 344 A · ANALYTICAL CHEMISTRY, VOL. 56, NO. 2, FEBRUARY 1984

+ RAh,K)

= /(λ/)£(λ/)Λί"(λ/)Ρ(λ/)

(7)

where η is the number of reflecting surfaces in the light beam. (continued on p. 348 A)

100 Γ Best Sample 00

^ ^ r · ^ " T u n g s t e n Lamp

10 Λ ^ 1

Μ

• • • Deuterium Arc

fi%

~//C •

' Λ * " * * »

.

200

400

600

60

800

100 100

200 (Days)

Figure 8. The effect of aging on the reflectance of aluminum at 200 nm. R = re­ flectance at 200 nm 200

400

600

800

400

600

800

100 80 60 40 20 0

200

Figure 7 shows these factors for typ­ ical components: tungsten lamp and deuterium arc sources, a grating blazed in the UV, an aluminum mir­ ror, and an S20 photocathode. The detector output D(X) can be obtained from a single-beam instru­ ment if one takes a series of readings at wavelengths over the instrument range, without altering the zero con­ trol. Alternatively, for a double-beam instrument the same measurements have to be made with it working in a single-beam mode. Using Equations 6 and 7 it is possi­ ble to calculate how the transmittance of a sample is affected by stray light (4). If

100

TMM

= measured sample transmittance Τ(λ) = true sample transmittance Β = instrument spectral bandwidth

10

then at wavelength λ/ 200

400

600

800

TM(\,)

= τ(λ/) + | f'jnxc)

-Τ(λ / ))Κ ϊ (λ /> λ ε ) ^ r M i X c 100

10

200

400

600

800

Wavelength (nm)

Figure 7. Performance of spectropho­ tometer components

(8)

This equation is equivalent to the simpler Equation 5 given earlier. Thus, to predict the stray-light error of a transmittance or absorbance mea­ surement from the monochromator stray-light function J?(X/,XC), a very laborious calculation is required based on difficult measurements. This em­ phasizes why the effect of stray light cannot be easily specified by instru­ ment manufacturers for practical measurements on particular samples. What else can we learn from Equa­ tion 8? As in Equation 5 it shows that the transmittance error depends on

348 A · ANALYTICAL CHEMISTRY, VOL. 56, NO. 2, FEBRUARY 1984

the sample transmittance Τ(λ/) at the analytical wavelength and on the transmittance T(XC) of the sample over the whole instrument range. In addition it shows how the instrument response D(X) can magnify the straylight error if the analytical wavelength Xj is chosen to be where D(\j) is small, for example near the end of the usable range of a particular light source, when the stray light from other wavelengths where D(XC) is large is greatly magnified in effect by the ratio of the values of D(X). This type of effect becomes very evident when, for example, the light output of the deuterium arc source falls off with age or mirrors become dirty or con­ taminated, both effects resulting in D(X) decreasing in the UV and, in consequence, the UV stray light in­ creasing in effect. It will be noticed from Equation 8 that the stray-light error is apparently inversely proportional to the spectral bandwidth B. However, because the stray-light function Rs(Xi,Xc) is also proportional to the bandwidth, as it depends on the area of slit into which light is scattered, the stray-light error in practice is largely independent of the bandwidth when a continuum light source is used. It should also be noted that the height of the mono­ chromator slits should be kept to a minimum compatible with other re­ quirements, such as allowing enough energy for adequate signal-to-noise ratio, so that the slit area into which stray light can be scattered is kept to a minimum (2,4). Recent Improvements in Stray-Light Performance In recent years there have been very significant reductions in stray-light

100

Reflectance

80

60 Al - Fresh UV Irradiated for 170 Days Then Washed

0%

40

20 _i 200

300

400

600

ι

L_L

800 1000

X(nm)

Figure 9. The effect of UV irradiation and subsequent washing on the reflectance of aluminum

levels quoted by the manufacturers of most commercial spectrophotometers. The two main reasons for this are the introduction of holographic diffrac­ tion gratings and the use of mirrors coated with a protective film. Coated optics. The mirrors in a spectrophotometer are normally coat­ ed with aluminum, the metal with the

100

highest reflectance over the required wavelength range. Freshly deposited aluminum slowly grows a layer of alu­ minum oxide on its surface, which re­ sults in a progressive loss of reflec­ tance at UV wavelengths, which is ac­ celerated by exposure to UV radiation. Figure 8 shows how at 200 nm the re­ flectance of aluminum in air decreases

Reflectance

80

60 R%

40 -

^ ^ AI + Si0 2 - Fresh •^" - UV Irradiated for 530 Days - Then Washed

20 200

300

400

-ι—ι ι ι 600 800 1000

\(nm)

Figure 10. The effect of UV irradiation and subsequent washing on the reflectance of silica-coated aluminum 350 A · ANALYTICAL CHEMISTRY, VOL. 56, NO. 2, FEBRUARY 1984

by about 5% per year and at a faster rate when exposed to a flux of UV ra­ diation similar to that from a deuteri­ um arc source in a spectrophotometer. In some instruments there are up to 12 mirror reflections before the light beam reaches the detector, so that even a 5% loss of reflectance at each mirror can halve the energy. The oxide layer growth can be pre­ vented by overcoating the aluminum with a thin transparent layer. Some manufacturers use a layer of magne­ sium fluoride for this purpose, but it is not very satisfactory as it is relatively soft and has poor chemical resistance, and thus cannot be easily cleaned. A better solution to the problem is to use a silica or synthetic quartz coat­ ing, which is hard and chemically re­ sistant. The correct coating thickness will also enhance the reflectance at UV wavelengths by constructive inter­ ference effects within the thin film. Aluminum mirrors coated with sili­ ca do not age like bare aluminum. If they become dirty, they can be washed with a mild detergent and distilled water to restore the original high re­ flectance. Figure 9 shows that washing an uncoated mirror after it has been ex­ posed to a deuterium arc for 170 days does not restore its reflectance. Con­ trast this with Figure 10 for a silicacoated mirror; after 530 days' expo­ sure to the same environment, wash­ ing fully restores the original reflec­ tance. The use of silica-coated alumi­ num mirrors thus ensures long mirror life with enhanced reflectance in the UV and minimum deterioration of stray-light performance. Holographic diffraction grat­ ings. In recent years a new process for making diffraction gratings has been developed: the holographic or interfer­ ence method. The grating is made by first coating a glass substrate with a layer of photoresist, which is then ex­ posed to interference fringes generat­ ed by the intersection of two collimated beams of laser light. When the pho­ toresist is developed it yields a surface pattern of parallel grooves. When coated with aluminum this becomes a diffraction grating. Compared with a ruled grating, the grooves of a holographic grating are much more uniformly spaced, smooth, and uniformly shaped, resulting in much lower stray light levels. The dia­ mond ruling process requires days to rule one grating, whereas a holograph­ ic grating requires only a few minutes' exposure time. Simple holographic gratings have a sinusoidal groove profile with no welldefined blaze wavelength and a maxi­ mum diffraction efficiency of 40%. With suitable interferometer geome­ try it is possible to produce blazed (continued on p. 356 A)

gratings with the required triangular groove profile and high blaze efficien­ cy of the order of 80%. Figure 5 shows electron micrographs of a convention­ al ruled grating and of a blazed holo­ graphic grating. It is evident that the holographic process produces grooves that are an order of magnitude smoother and more regular. Holographic gratings used in com­ mercial spectrophotometers are either original master gratings produced di­ rectly by an interferometer or replica gratings, which are reproduced from a master holographic grating by mold­ ing its grooves onto a resin surface on a glass or silica substrate. The replica­ tion process can produce gratings that are almost as good as master gratings. Both types of gratings are coated with an aluminum reflecting surface and usually also with a protective layer of silica or magnesium fluoride, as de­ scribed previously for mirrors. Referring back to Figure 6 for a typ­ ical UV-VIS grating 20 mm wide with 1200 grooves/mm, the theoretical min­ imum value of β(λ/,λ 0 ) for monochro­ matic light of 600 nm, with a spectral bandwidth of 3 nm, is of the order of 10 - 7 between diffracted orders (4), be­ cause of limiting diffraction effects. A ruled grating of this type usually has a corresponding minimum of between 1 0 - 5 and 10~6, whereas holographic gratings can have minima of less than 1 0 - 6 and sometimes as low as 3 X 10 - 7 . The holographic process is thus capable of producing gratings that al­ most reach the theoretical stray-light minimum. Standard Tests for Stray Light To compare the stray-light perfor­ mance of spectrophotometers, stan­ dard tests are recommended by the American Society for Testing and Ma­ terials (ASTM Designation E387-72). These tests have been almost univer­ sally adopted by spectrophotometer manufacturers. The tests are based on the principle of measuring the transmittance of a sample that has virtually zero trans­ mittance at the wavelength at which the stray light is to be measured and a high transmittance at wavelengths from which the stray light originates. The measured transmittance of the test sample is then a measure of the stray light, as only stray light is trans­ mitted by the sample. This can also be deduced from Equation 5. If the sam­ ple transmittance Τ at the set wave­ length is zero, then the measured transmittance Τ Μ - OLS. If a, the pro­ portion of stray light transmitted, is 1, then the measured transmittance TM = S, the total stray light. For most test samples a can differ appreciably from 1, and therefore the standard tests are really only compar­

ative. Thus, for many practical sam­ ples the effective stray light can differ significantly from that given by the test. Stray light is usually at a maximum where the instrument energy is at a minimum. For this reason most manu­ facturers quote stray light at 220 nm, where the deuterium arc lamp energy is small, and at 340 nm, close to a min­ imum of the tungsten lamp energy, as shown in Figure 7. Also, 340 nm is chosen as it is an important and wide­ ly used biochemical wavelength. At 220 nm the ASTM test measures the transmittance of a 10-g/L solution of Nal in a 10-mm path length cell. This solution has a transmittance of less than 1 0 - 1 0 at 220 nm, but trans­ mits most of the energy at wave­ lengths longer than 265 nm. At 340 nm a 50-g/L solution of N a N 0 2 in a 10-mm path length cell is used. Some doubts have been ex­ pressed about this test at 340 nm (5), because most spectrophotometers use a bandpass filter around 340 nm to re­ duce stray light. Consequently, the ASTM test can give an optimistic measurement because the test solu­ tion only measures a small proportion of the stray light in the filter passband. Unless the test is modified it is to be expected that manufacturers will continue to use the recommended test. A word of warning on the test solu­ tions given above: It is preferable to use fresh solutions, particularly when the stray-light levels being measured are very low (for example,