Anal. Chem. 1000, 60,1380-1387
1300
A two-beam arrangement was used to eliminate instabilities of the lamp. A chopper in combination with a lock-in amplifier arrangement enables measurements in normal daylight.
LITERATURE CITED Cocache. R. J . Phys. E 1988, 19, 401-412. Oyabu. T. J. Appl. Phys. 1082, 53, 2785-2787. Josowicz. M.:Janata. J. J. Anal. Chem. 1986, 58, 514-517. Honeybourne. C. L.; Houghton, J. D.; Ewen, R. J.; Hill, C. A. S. J . Chem. Soc., Faraday Trans. 11986, 82, 1127-1133. (5) Miasik. J. J.; Hooper, A,; Tofield, B. c. J , Chem. SOC., Faraday Trans. 1 1 9 8 8 , 82, 1117-1126. (1) (2) (31 (4)
(6) Veith, H. Kolloid-Z. 1957, 152, 36-41. (7) Sliwka. W. Angew. Chem., Int. Ed. Engl. 1975, 74, 539-550. (8) Brandner, G.; Dickert. F. L.; Lehmann, E. Ber. Bunsen-Ges. Phys. Chem. 1987, 91, 740-745. (9) Brandner, G.; Dickert, F. L.; Fackier, H. Z. Phys. Chem. (Munich) 1986, 148, 65-73. (10) Gutmann, V. Nectrochim. Acta 1978, 21, 661-670. (11) UVlvis Spectrophotometer-Lambda 5 Manual; Perkin-Elmer GmbH: Uberlingen, FRG.
RECEIVED
for review November 24, 1987. Accepted March
1, 1988.
Spectroscopic Investigation of the Mechanisms of the Alkali Bead Detector for Gas Chromatography Peter van de Weijer* and Bauke H. Zwerver Philips Research Laboratories, P.O. Box 80.000, 5600 J A Eindhoven, T h e Netherlands
Roderick J. Lynch Philips Scientific, York Street, Cambridge CB1 2PX, Great Britain
The mechanism of the alkali bead detector, as used to detect organonitrogen compounds in the column effluent of a gas chromatograph, has been investigated by emisslon and laser-induced fluorescence measurements. The emission spectra led to the conciudon that the addition of an aikail salt to the glass of the bead of the detector is not essential: the beads can be made of ordlnary soda glass. Laser-induced fluorescence measurements revealed that the alkali atoms are not lost from the bead by evaporation, but by an exchange with hydrogen. The appiicatlon of this combination of spectroscopic technlques during a long run test of the bead proved that the ionlratlon mechanism of the detector for both background and signal current is gas phase ionization of sodium atoms.
In order to detect organonitrogen and -phosphorus compounds in the column effluent of a gas chromatograph, two types of specific detectors may be used. The first one is the alkali flame ionization detector (AFID), represented schematically in Figure 1A. The detection is based on the observation of an increase in current through a hydrogen flame, which has been seeded with an alkali salt such as rubidium chloride, when the organonitrogen or -phosphorus compound is introduced. The hydrogen flame is used for both volatilization of the alkali salt and decomposition of the organic compound. The increase in current due to the introduction of the sample is caused by gas-phase ionization of alkali atoms by CN and PO or POz radicals which are believed to be one of the products of the pyrolysis (thermal decomposition) of the organonitrogen and -phosphorus compound, respectively. A disadvantage of this detector is its poor long-term stability owing to the increase of the distance between hydrogen flame and salt supply when it is evaporated. Of course, the position of the salt supply can be readjusted with respect to the burner jet, but it is hard to perform this operation in a reproducible way. The second nitrogen/phosphorus-selective detector was introduced by Kolb and Bischoff in 1974 ( I ) as an alternative
to the AFID (see Figure 1B). In this design the alkali salt is incorporated in a glass bead, mounted on an electrically heated platinum wire. The hydrogen flow is only 2-6 mL/min, a very low value in comparison with the 20-60 mL/min used with the AFIDs. Under these circumstances the hydrogen cannot form a flame but produces a plasma on the surface of the heated bead. The background current in this detector (i.e. the current in the absence of an organonitrogen or -phosphorus compound) is in the picoampere region. The signal current, due to the presence of a sample, can be a few orders of magnitude larger than the background current. Since the introduction of this alkali bead detector, several attempts have been made to explain its ionization mechanism. A main point of discussion is whether the ionization, responsible for the background and the signal current, occurs in the gas phase or a t the surface of the bead. Gas Phase Ionization. By analogy with the AFID Kolb and Bischoff originally proposed that the ionization process is gas phase ionization of rubidium. These atoms are assumed to leave the bead by evaporation. For the background current this gas phase ionization process may involve one of the following reactions (1,2): a recharging process with hydrogen atoms Rb
-
+ H30+
Rb+
--
+ H + H20
(1)
a three-body ionization process with two radicals Rb
+H +H
+ H + OH Rb + 0 + 0
Rb
-
Rb+ + H2 + e
+ H20+ e Rb+ + O2 + e Rb+
(2a) (2b)
or an ionization by the kinetic energy of nitrogen, which is used as the carrier gas Rb
+ N2
Rb+
+ N2 + e
(3)
In reactions 1and 2 the rubidium ground state and excited states may be involved from an energetic point of view (the ionization energy of rubidium is 4.2 eV, whereas the energy release by recombination of the two radicals in reactions 2a-c is 5.2, 4.5, and 5.2 eV, respectively). Reaction 3 probably
0 1988 American Chemical Society 0003-2700/88/0360-1380$01.50/0
ANALYTICAL CHEMISTRY, VOL. 60, NO. 14. JULY 15, 1988
1
column effluent
column effluent 1. schematic diagram of (A) me alkali %me ionization detecla (AFID) and (5)the modification of the AFID introduced by Kolb and BischOff (I), the alkalf bead detector. involves excited rubidium atoms since the fraction of nitrogen atoms which can accomplish this reaction, i.e. the high-energy tail of the nitrogen energy distribution, is substantially larger for excited rubidium atoms than for ground-state rubidium atoms. For the signal current the gas phase ionization process may be represented by
-
Rb* + CN/PO/P02 Rb+ + CN-/PO-/P02- (4) The electron affinity of CN radicals (3) is large enough to ionize rubidium from its first excited states but not large enough to ionize these atoms from the ground state. The electron affinity of PO and POz is not known. The reduction of the sensitivity during the life of the detector can be explained by the decrease in rubidium gas phase concentration when the bead gets depleted of rubidium salt. Surface Ionization. In the gas phase ionization proposal of Kolb and Bischoff, rubidium atoms are assumed to leave the bead by evaporation. This assumption is not consistent with the observation of a good sensitivity of beads containing nonvolatile salts (4, 5). Therefore, as an alternative to gas phase ionization, negative surface ionization of the radicals produced in the plasma was proposed by Patterson (6)for a rubidium-containing ceramic bead, by Olah et al. (7) for a rubidium-containing glass bead (similar to the bead as introduced by Kolb and Bischoff), and by Fujii and Arimoto (5) for a LaB, bead. In the theory of negative surface ionization the degree of ionization a t thermal equilibrium is described as (5)
n - / b =g_/ggW-d/kT
(5)
where noand n- are the densities of neutral and negative species (OH radicals for the background current, as suggested by Olah et al. (7) and CN/PO/POZ radicals for the signal current) desorbed from the negative surface a t temperature
-
1381
T,6 is the work function of the surface, E. is the electron affmity of the species, and goand g. are the statistical weights of the neutral and negative species, respectively. The rubidium atoms in the bead are responsible for a reduction of the work function of the glass. When the glass becomes depleted of rubidium atoms during the life of the bead, the work function will increase, resulting in a decreasing sensitivity of the detector. An additional argument for negative surface ionization has been given by Olah et al. (7). These investigators measured the current-voltage characteristic of the alkali bead detector. The most distinct feature of this current-voltage characteristic is its substantial asymmetry. For both the background and the signal current, these authors found that at negative bead polarity (normal operation mode) the current has a limiting value, whereas a t positive bead polarity the current keeps increasing with increasing voltage. This is in contrast with the current-voltage characteristic of flame ionization detectors, which is symmetric and shows limiting currents (8,9). When Olah et al. observed these features, they interpreted them in the following words: "The symmetry of the current-voltage characteristic of the flame ionization detector proves that the ionization process takes place in the homogeneous gas phase and therefore cannot be influenced by the magnitude of the electric field. The electrodes are passive, that is, they absorb only (collect) electric charges (electrons or ions). The limiting currents (all produced ions collected) are the same, independent of polarity. ... In the present case the asymmetry proves that at least one electrode must be active, that is, an emitting one. Such an ionization process can be influenced by the voltage applied". Following the arguments of Olah et al., the conclusion from their current-voltage characteristic, in our opinion, must be surface ionization for positive bead polarity (no limiting current) and gas phase ionization for negative bead polarity (limiting current). However, the conclusion of Olah et al. was totally different. They did not mention the mechanism for positive bead polarity and attributed the mechanism a t negative bead polarity to surface ionization! Though a surface ionization process may show a limiting curve too, as Olah et al. state correctly, we do not consider their proof for negative surface ionization at negative bead polarity (normal operation mode) very convincing. In order to investigate the ionization mechanism of both background and signal current. we performed spectroscopic measurements on the plasma around the bead. We used emission measurements to record the relative abundance of excited species and laser-induced fluorescenceto measure the relative abundance of ground-state species. EXPERIMENTAL SECTION Bead. In the experiments commercially available beads (diameter el mm) were used. These beads are prepared from a soda glass, enriched with rubidium chloride. In this work the alkali content of the bead was analyzed. In order to do so, the platinum wire is cut from the bead assembly and its weight is determined. After the bead is dissolved in aqueous HF solution (40%), the platinum wire is weighed again resulting in the weight ofthe bead. The solution is diluted, and the alkali content is determined by ion chromatography. Detector. The alkali bead detector in a gas chromatograph is heated externally to a temperature that must be higher than the column temperature in order to prevent condensation of the sample in the detector. For our spectroscopic investigation, however, we removed the detector from our gas chromatograph (PU 4550) and did not heat the detector externally. Moreover, we made three holes in the cylindrical detector body in a Tconfiguration: the two opposing holes allow a laser beam to pass through the detector, while the third one is occupied by an optical fiber to collect emission and fluorescence light. An effect of these alterations is that the performance of the detector might not be
1382
ANALYTICAL CHEMISTRY, VOL. 60, NO. 14, JULY 15, 1988
Table I. Standard Settings of the Detector nitrogen flow air flow hydrogen flow background current heating power" bead voltage
40 mL/min 185 mL/min 6 mL/min 80 pA 5-6 W -30 V
aElectrical heating power of the bead and the platinum wire, depending on the age of the bead.
I
Figure 2. Schematic diagram of the experimental arrangement for the laser-induced fluorescence measurements: BS, beam splitter; EM, energy meter; F, fiber; L, lens; LRM, low-resolution monochromator; M, mirror: ND, neutral density filter@);P, pinhole; PM, photomultiplier.
exactly the same as in normal use in a gas chromatograph. Our standing settings of the detector under investigation are given in Table I. Sample. We used nitrotoluene as a test sample. Instead of injecting a small amount of the sample in the column of a gas chromatograph, we used a constant sample flow in order to create a stationary state in the plasma of the detector. This was necessary since the spectroscopic measurements require a period of time which is long in comparison to the residence time of the sample in the plasma of the detector when detecting a chromatographic peak. So, when studying the mechanism of the background current, we used a flow of carrier gas (nitrogen) directly to the detector, without using a column. To investigate the ionization mechanism of the signal current, the nitrogen flow was led through a bottle over (liquid) nitrotoluene. This bottle was kept at constant temperature in a water bath. The temperature of the water was 18 OC. Assuming a saturated vapor pressure, this corresponds to a nitrotoluene flow of 1 nmol/s at the given nitrogen flow. Emission Measurements. The emission measurements were performed with an EG&G OMA I11 optical multichannel analyzer. In this system we used a 0.25-m Jarrell-Ash monochromator with a 150 grooves/mm grating. The detector was a photodiode array consisting of 1024 elements. In the spectral region 350-950 nm, the spectra were calibrated with mercury and rubidium discharge lamps. Laser-Induced Fluorescence. A schematic diagram of the experimental arrangement is given in Figure 2. A dye laser pumped by a pulsed nitrogen laser (Molectron DL I1 14 and UV 14, respectively) was used as a light source. Searching for fluorescence of the species Rb, Na, CN, and NOz, we used stilbene-l, Rhodamine 6G, 4,4"'-bis(butylacetyloxy)quarterphenyl (BiBuQ), and 7-(diethylamino)-4-methylcoumarin(7D4MC/ Cl/C47/C460). With these dyes 10-ns light pulses were made at 420.2 and 421.6 nm (for Rb), at 589.0 and 589.6 nm (for Na), around 388 nm (for CN), and around 545 nm (for NOz). The energy of the pulses was 25-100 pJ, depending on the dye and its age. The reason why we searched for fluorescence of these particular species will become clear in the section Results and Discussion. We will describe the experimental arrangement for the sodium fluorescence measurements. The arrangement differed only in detail when searching for fluorescence of the other species. Sodium atoms are excited by a dye laser pulse at 589.6 nm. In order to tune the laser to the correct wavelength, part of the laser beam is split off into a sodium vapor cell from which fluorescence can be visually observed. Another part of the beam
Figure 3. Simplified energy level diagram of sodium. Note that, for the sake of clarity, the energy difference between levels 1 and 2 is greatly exaggerated with respect to that between levels 0 and 1. Key: R, radiative transition; C, colllsionally induced transition.
is split off for monitoring the pulse energy with a Laser Precision Corp., Model Rj-7200 energy meter. Laser irradiation at 589.6 nm excites sodium atoms from the 32S!/2ground state to the 3?1/2 first excited state (see Figure 3). Sodium atoms in this state may decay radiatively to the ground state (R,) or collisionally to the 32P3jzsecond excited state (C12).Once this second level has been populated by the collisional process, it may decay in a similar way as the first excited state. The relative fluorescence intensities R1 and R2 depend on the ratio of the radiative and collisional decay rates, which is determined by the plasma conditions (IO). However, the sum of these two fluorescence intensities, at a certain laser energy, depends only on the sodium ground-state density. Therefore, the fluorescence of both excited states is detected with a low-resolution monochromator (0.25-m Jarrell-Ash with 1-mm slits) in combination with a photomultiplier. The output of the photomultiplier is fed into a PAR Model 160 boxcar. The time integrated fluorescence intensity is measured by use of a static gate. As the collected fluorescence wavelength interval contains the wavelength of the exciting h e r beam, precautions have to be taken to prevent stray light from the laser entering the detection system. Therefore two lenses and a pinhole are used to focus the laser beam in the plasma of the detector. In this way there is hardly any reflection of the laser beam from objects in the detector like the bead and the jet. With the pulse energy of our laser the sodium transition is saturated. This was checked by measuring the fluorescence intensity as a function of the laser energy which was varied with a set of neutral density fiiters. Due to this saturation the fractional excitation of sodium is independent of the sodium density. As a result the fluorescence signal is proportional to the sodium ground-state density.
RESULTS AND DISCUSSION Emission Measurements. An example of an emission spectrum from the interior of the detector is given in Figure 4. It shows the continuum emission of the bead and the platinum wire and a number of atomic emission lines. A prerequisite for the correctness of the assumption of gas phase ionization of excited rubidium atoms (1) is the presence of these atoms in the gas phase. Indeed we found rubidium emission from the lowest excited states at 780 and 794 nm. However, a much more intense sodium emission is present in the spectrum at 589.0 and 589.6 nm (not resolved by the system settings which we used). The presence of sodium in the gas phase is not surprising since the bead is made of soda glass (which apparently contains some potassium too, as can be seen from the emission spectrum). After correcting for the relative sensitivity of the photodiodes and the transition probabilities of the sodium and rubidium lines, we find that the density of excited sodium atoms is a factor of 5 larger than the density of excited rubidium atoms. As the first excited states of sodium are higher in energy than those of rubidium, the ground-state sodium to rubidium density ratio is likely to be even larger. Assuming that sodium and rubidium leave the bead at the same rate, it means that the sodium to ru-
ANALYTICAL CHEMISTRY, VOL. 60, NO. 14, JULY 15, 1988
I
Na
400
500
I
I
1
600
700
800
-h
1383
I
900
(nm)
Flgure 4. Emission spectrum of the interior of the alkali bead detector showing the continuum emission of the hot bead and platinum wire and a number of atomic emission lines. This emission spectrum does not change when nitrotoluene Is added to the gas flow.
bidium number density in the bead is larger than 5 as well.
This is confirmed by the result of the analysis of alkali content of the bead; the sodium to rubidium number density in the bead is 8. So far, we did not account for the presence of sodium when considering the ionization mechanism of the detector. However, sodium may be involved in the same way in these processes as rubidium. The proposed gas phase reactions can be the same (except reaction 2b, which can only involve excited sodium atoms since the ionization energy of ground-state sodium atoms is 5.1 eV). If the mechanism is surface ionization, sodium may be involved too; the sodium depletion during the life of the bead causes an increase of the work function of the bead, resulting in the observed lower sensitivity. In order to test the role of sodium in the ionization mechanism of the detector, we made some beads without rubidium chloride. As a result we found that the difference in performance of beads with and without rubidium chloride is not significant! This does not prove that rubidium is not involved in the ionization processes in the detector. Its possible contribution, however, is smaller than the variation in performance between different beads (e.g. variation in size). This observation proves that the ionization in the detector is mainly due to sodium. So, the presence of rubidium chloride in the bead is not essential; the nitrogen selective alkali bead detector can be made of ordinary soda glass! However, the addition of a substantial amount of rubidium chloride (or any other alkali salt) may lengthen the useful life of the bead. Laser-Induced Fluorescence. In order to measure the ground-state densities of rubidium and sodium, we performed laser-induced fluorescence (LIF) measurements. After excitation of sodium atoms at 589.6 nm, we found fluorescence a t 589.0 and 589.6 nm. Due to the much lower rubidium density we could not find any fluorescence from rubidium atoms in the plasma, after tuning the laser at 420.2 or 421.6 nm in a rubidium vapor cell. Figure 5 shows an example of an experiment where laserinduced fluorescence was used to monitor the sodium ground-state density. This diagram represents four different conditions of the detector. In the first one (column a) the hydrogen flow is 6 mL/min. In order to establish 80-pA background current at negative bead polarity ( i J , 5 W of electrical heating power (P)is necessary. For these standard settings of the detector we measured the background current at positive bead polarity (i+),the sodium fluorescence signal
a
b
C
d
Flgure 5. Diagram showing various quantities of the detector for four
different situations. The quantities are hydrogen flow, background current at 30 V negative bead polarity (i-), background current at 30 V positive bead polarity (i+), sodium fluorescence intensity ( F ) nortemperature malized to 100, electrical heating power of the bead (P), of the collector (T,,,,), and the number of 700-nm counts in the continuum emission of the bead (I,o,,) as a measure for the temperature b the hydrogen supply is switched off, at b of the bead. At a c the electrical heating power of the bead is increased in order to restore the original temperature of the bead and the collector and at c d the electrical heating power is increased to restore the original background current at negative bead polarity.
-
-
-
(F),the temperature of the collector (Tcon), using a thermocouple, and the number of counts at 700 nm in the emission spectrum as a measure of the temperature of the bead. The quantities F, TcOu, and 1700 are independent of the applied bead voltage. When the hydrogen supply is turned off (column b of Figure 5 ) , the plasma disappears and both the negative and the positive background current drop to zero. Due to the lack of hydrogen combustion the temperature of the collector and the bead decreases. The electrical bead heating current is kept constant. Due to the lower temperature the resistance of the platinum wire decreases. Therefore, the electrical heating power decreases slightly when the hydrogen supply is turned off. In the gas phase ionization theory of Kolb and Bischoff the alkali atoms are assumed to leave the bead by evaporation. When the hydrogen supply is turned off, the temperature of the bead decreases. Following Kolb and Bischoff the disappearance of the fluorescence signal can be explained by a lower evaporation rate of sodium. As a result the sodium density in the gas phase has decreased to such an extent that we are not able to detect any fluorescence. After increasing the electrical heating power of the bead to return the bead (and the collector) to the original temperature (column c of Figure 5 ) , we expected then to obtain the original fluorescence signal.
1384
ANALYTICAL CHEMISTRY, VOL. 60, NO. 14, JULY 15, 1988
However, we could not observe fluorescence at all (which means that the sodium density must be 2 orders of magnitude lower a t least). Even when we overcompensated the lack of hydrogen combustion by increasing the electrical heating power until the bead was white hot (instead of the normal reddish color), no fluorescence could be observed (column d). From this observation we concluded that sodium is not lost from the bead by evaporation but by a n interaction with the hydrogen plasma. The mechanism of this interaction could be an exchange of hydrogen and sodium atoms. From the experiments represented in Figure 5, we can draw some conclusions about the ionization mechanism in the detector. From column b to c the action is to increase the electrical heating power of the bead to such an extent that the temperature of the bead and the collector is the same as in the situation with the hydrogen flow on. After doing so, we observe that the current a t negative bead polarity is still zero, whereas the current at positive bead polarity has regained its original value. The ionization mechanism for this positive background cannot be gas phase ionization of sodium, as no sodium is present in the gas phase. Neither can it be surface ionization of OH radicals, as proposed by Olah et al. (7),since no hydrogen is present. The conclusion, in our opinion, must be that the ionization mechanism for the positive background is electron emission of the metal collector (which in this mode of operation is not a collecting anode but the emitting cathode). At first sight it seems rather odd that the cold “collector” cathode emits electrons, whereas the hot “bead“ cathode does not. However, we must realize that the surface of the collector is more than 2 orders of magnitude larger than the surface of the bead. Moreover, the shape of the collector is that of a cylinder. A cathode with such a shape is called a hollow cathode. A discharge with such a cathode has the property that the current density can be orders of magnitude larger than in a conventional discharge for the same cathode fall. The physical explanation for this phenomenon is that many of the products of ionization and excitation which in an ordinary discharge are lost to the wall will now fall upon the cathode and, by secondary processes, contribute to the current. An alternative mechanism for the positive background might be emission of positive sodium ions from the hot bead, which are optically inactive. However, during the life of the detector, the positive background current is constant. This confirms the idea that alkali atoms are not involved in its mechanism; the magnitude of the positive background current is determined by the temperature of the “collector”. The negative background current can also be restored to its original value by increasing the electrical heating power (column d of Figure 5 ) . For these conditions, where the bead is white hot, the mechanism is probably electron emission from the bead. However, at normal operational temperatures the mechanism is not electron emission. Variation of the Hydrogen Flow. Figure 6 shows some results of electrical and optical measurements as a function of the hydrogen flow. While the hydrogen flow is varied, the background current is kept constant by adjusting the electrical heating power of the bead. In the regime of hydrogen flows presented in Figure 6, the sensitivity of the alkali bead detector decreases with increasing hydrogen flow, which has been observed before ( I ) . For a lower hydrogen flow the temperature of the bead needs to be higher in order to maintain a constant background current (Figure 6B). However, the sodium density (ground state as well as first excited state), is lower in the situation where the bead is hotter (Figure 6, parts C and D). This observation confirms the conclusion drawn above that the sodium loss is not controlled by evaporation but by a sodium/hydrogen exchange. Therefore, to lengthen the useful life of the bead, a low hydrogen flow is recommended. Re-
background 0-3--0-3 ~
AI
,
4
,
I
6
6
4
8
8
-
H, flow (ml/min)
Flgure 6. Effect of a variation of the hydrogen flow through the detector (at constant background current) on (A) the signal current, (B) the number of 700-nm counts in the continuum emission of the bead, (C) the ground-state sodium density, and (D) the excited-state sodium density. The quantities represented in B-D are normalized to 100. Their magnitude is independent of the presence of nitrotoluene.
cently, it has been shown that extremely long lifetimes can be obtained in this way (11). The higher sensitivity of the detector at reduced hydrogen flow is obtained at a lower gas phase sodium density. This seems to contradict the alkali gas phase ionization mechanism, proposed by Kolb and Bischoff (1). However, by changing the conditions, we might have changed the efficiency of the CN radical production. If this efficiency is much higher a t lower hydrogen flows, the mechanism could be gas phase ionization. Therefore, it seemed useful to measure the CN radical density in the gas phase. However, neither emission
ANALYTICAL CHEMISTRY, VOL. 60, NO. 14, JULY 15, 1988
100 -
tI 50
Flgure 7. (A) Background current and (B) signal current as measured during 100 h of the life of the bead plotted versus the sodium density: 0, ground-state sodium density; X, excited-state sodium density. All densities and currents are normalized to 100 for a fresh bead.
nor laser-induced fluorescence of CN radicals could be found. Either the density of those radicals is too low to be measured or the pyrolysis of nitrotoluene results in another reactive species. An obvious alternative could be the formation of NOz radicals. The electron affinity of these species is large enough to ionize (excited) sodium atoms (3). Unfortunately, our attempts to observe NOz particles by emission or laser-induced fluorescence were unsuccessful too. Since we are not able to detect the reactive species responsible for the signal current through the plasma, we must be certain not to influence its density when relating the current and the sodium density. This might be possible by keeping all variables (like hydrogen flow and heating current) constant and observing the decreasing current and sodium density during the life of the bead. Long Run Test. During 100 h of the life of the bead, we have measured the decay of the background and the signal current in combination with the relative sodium excited state density (emission) and the sodium ground state density (laser-induced fluorescence). As observed before, the sodium densities are not affected by the presence of nitrotoluene. Figure 7 shows the result of this long run test. The decay of the ground-state density and decay of the excited-state density during the life of the bead are proportional. This indicates that the plasma conditions during the life of the bead are constant; the fractional excitation of sodium is constant. At constant plasma conditions we may make two reasonable assumptions. First, the efficiency of the pyrolysis of nitrotoluene is constant, and second, the fractional ionization of
1385
sodium is constant. These two assumptions, combined with the observation from Figure 7 that the decay of the background and signal current is proportional to the sodium densities, lead to the conclusion that both background and signal current through the detector are proportional to the sodium ion density. At constant plasma conditions this proves that the ionization mechanism is gas phase ionization of sodium atoms. Simple Model Description of the C u r r e n t d u r i n g a Long Run Test. While measuring the current through the detector during a long run test, we observed that the decay of the current cannot be described by a single exponential function; the relative decay for a new bead is faster than for an old bead, which is a well-known phenomenon (12). If, before starting the long run test, the bead is kept hot without the presence of hydrogen during a substantial period of time, the initial current and its relative decay rate are increased considerably. On the basis of the observations of the previous sections, we will try now to describe the decaying current during the life of the bead mathematically. The current through the detector (background as well as signal) is proportional to the sodium density in the plasma (Figure 7). The sodium density in the plasma is assumed to be proportional to the flow of sodium from the bead into the plasma. The flow of sodium from the bead into the plasma is assumed to be proportional to the density of sodium at the surface of the bead. Therefore, the current through the detector is proportional to the sodium density at the surface of the bead. If the bead is approximated as a sphere and only radial sodium diffusion is assumed, the equation describing the sodium density N in the bead is given by
where r is the radial coordinate in the bead and D is the diffusion coefficient of sodium. The sodium density in the bead is initially uniform with density No and there is a surface condition ( r = a )
-DaN/ar =
(YN
(7)
where a is the rate constant of the flow of sodium into the plasma (mathematically it does not matter whether this is an evaporation mechanism or an exchange with hydrogen atoms). The solution of eq 6 and 7 is (13)
where the (3, values are the roots of p, cot pn + L - 1 = 0
(9)
and
L = aa/D The density of sodium a t the surface of the bead can be obtained from eq 8 by putting r = a. Equation 8 shows that the decay of the current through the detector is described as the sum of exponential functions with time constants az/D(3,2. The values of (3, have the property that they increase when n increases. The decay of the (background) current during the first 100 h of operation of the bead is shown in the first part of Figure 8. Before the start of this experiment the bead has been kept at the operational temperature without hydrogen during 100 h. This procedure results in a very large background current and fast relative decay in the early life of the bead. During the life of the bead the relative decay rate of the current decreases until it is constant. Then the decay is determined
1386
ANALYTICAL CHEMISTRY, VOL. 60, NO. 14, JULY 15, 1988
coefficient, more experimental data would be necessary. The initial fast decay of the current during the first 10 h is only observed if the bead is kept hot during a long period of time without the presence of hydrogen, before starting the long run test. This indicates that during the production process some sodium is lost from the surface of the bead. In other words, without precaution the sodium density in the bead is not uniform, which is an assumption in our model description. So, the long-term stability of a bead without this heat treatment is typically reflected by the 10-100-h portion of the decay curves given in Figure 8. This long-term stability can even be improved considerably by operating the detector at lower hydrogen flows (11).
0
50
100 t (hours)
150
I 200
Figure 8. Decay of the background current during the life of the bead. The circles represented experimental data points, the full and dashed curves are calculated (for explanation see text). Before the long run
test was started, the bead was kept at the operational temperature without the presence of hydrogen during 100 h. As a result the initial decay of current is very fast during the first 10 h of the long run test. This fast decay is not observed at “normal” use of the bead. by only. Assuming that this situation is reached at t = 100 h, the measurement of the slope of the decay at 100 h will provide an estimate of the value of DPI2. Then, by trial and error we choose the relative values of D and PI, resulting in the numerical values for all other P,, L , and a. In this way we fitted the calculation to the observed decay of the current. During this procedure we had to adjust the value of DPI2, which means that the higher values still contribute significantly after 100 h. The full line in Figure 8 (from 0 to 100 h) is the result of the fitting procedure. With the values of D , P,, CY, and L found in this way, the sodium profile in the bead can be calculated. Though the density at the surface of the bead has dropped to 1% of the original value, the bead contains a substantial reservoir of sodium. According to the calculation of the sodium profile, the total sodium content has dropped only by 40% during the long run test. We continued the experiment by turning off the hydrogen supply. The resulting drop in temperature was compensated by increasing the electrical heating power of the bead. Then we waited for 100 h and switched the detector on again at the original settings. If evaporation was the loss mechanism of sodium from the bead, the expected current would be obtainable by extrapolating the calculated curve to t = 200 h, that is, a current of about 6 pA (dashed line in Figure 8). However, if sodium is lost by an interaction with the hydrogen plasma, as indicated by the laser-induced fluorescence experiments, no sodium has been lost during the last 100 h. Moreover, since the bead is kept hot, sodium can diffuse in the bead, restoring a flat profile again. According to our calculations this results in a sodium density at the surface of the bead which is 60% of the original density. The expected current would then be 600 PA. The observed current is 500 PA, which clearly demonstrates that no sodium is lost when the bead is hot without the presence of a hydrogen plasma. The relative decay of the 500-pA current should be the same as the decay of the original current, which is represented by the solid line from 100 to 200 h. The observed decay, however, is much faster. This is probably due to a gradual lowering of the sodium diffusion coefficient while the sodium content in the glass decreases (14). We feel that the main deficiency of our simple model description is the treatment of this sodium diffusion coefficient, which instead of being constant is a function of the sodium concentration. This means that it is a function of both time and radial coordinate in the bead. However, for a correct description of the sodium diffusion
CONCLUSIONS In this work we investigated an alkali bead detector that is made of glass enriched with rubidium salt. Until now the rubidium was considered to be responsible for the current in the detector. By emission measurements we found that sodium is present in the plasma of the detector in substantially larger amounts than rubidium. This observation, in combination with the tests of beads containing no rubidium, proved that the detector mainly works on sodium. Therefore, the nitrogen-selective alkali bead detector can be made of ordinary soda glass. Addition of alkali salt in substantial amounts could lengthen the useful life of the bead. Sodium is not lost from the bead by evaporation but by an interaction with the hydrogen plasma (sodium/ hydrogen exchange?). This has been proven by two observations. First, no sodium fluorescence signal is present without hydrogen, independent of the temperature of the bead. Second, the sodium content of the bead does not decrease when a bead is kept hot without hydrogen during a substantial period of time. As a consequence of this conclusion, the observation of a good sensitivity of sources containing nonvolatile alkali salts ( 4 , 5 ) does not necessarily imply a surface ionization mechanism. Further, we may lengthen the life of the bead by operating it at reduced hydrogen flow. This can be done without loss of sensitivity. At negative bead polarity (normal operation mode) the ionization mechanism of both background and signal current is gas phase ionization of sodium atoms. This conclusion is based on the observation of a proportionality between the current through the detector and the sodium density in the plasma, as measured during a long run test. However, in our simple model description, we argued that the current through the detector is also proportional to the sodium density at the surface of the bead. It is tempting to conclude from this statement that it indicates surface ionization. Following eq 5 this would imply where N, is the sodium density at the surface of the bead. This relationship is very unlikely and does not show a behavior observed for binary mixtures like alloys (15). At positive bead polarity the mechanism of the background current is electron emission from the “collector”. The mechanism of the signal current a t positive bead polarity is not revealed in this work. This is due to the fact that relatively small variations in the large positive background current obscured the positive signal current. We agree with Olah et ai. (7) that “at least one electrode must be active”. However, in our conclusion, this is true at positive bead polarity and not a t the (normal) negative bead polarity. Finally, we emphasize that the conclusions on the alkali bead detector are based on observations with nitrotoluene as a test sample. Though we feel that the conclusions are likely to apply to organophosphorus compounds as well, it is not strictly proven in this work.
Anal. Chem. 1988, 60,1387-1390
Note Added in Proof. When this paper was in press, we repeated the experiments using trimethyl phosphate as a test sample. Again we found that the performance of beads with and without rubidium chloride does not differ significantly. Further, a long run test showed the same proportionality between signal current and sodium density as for nitrotoluene. This proves that the conclusions drawn in this paper apply to organophosphorus compounds as well. Registry No. Na, 7440-23-5.
LITERATURE CITED
.,
Kolb. B.; Bischoff, J. J. Chromatogr. Sci. 1974, 72, 625-629. Alkemade, C. Th. J.; Hollander, Tj.; Snelleman, W.; Zeegers, P. J. Th. Metal Vapours In Flames; Pergamon: Oxford, 1962. Christodoulides, A. A.; McCorkle, D. L.; Christophorou, L. G. “Electron Affinities of Atoms, Molecules, and Radicals”; Electron -Molecule Interactlons and melr Appllmtions ; Christophorou, L. G., Ed.; Academic: London, 1984; Vol. 2, Chapter 6.
1307
(4) Braznlkov, V. V.; Gurev, M.; Sakcdynsky, K. I. Chromafogr. Rev. 1980, 12, 1-41. (5) Fujil, T.; Arimoto, H. h 8 l . Chem. 1985, 5 7 , 490-493. (6) Patterson, P. L. J. Chromafogr. 1978, 767, 381-397. (7) Olah, K.; Szoke, A.; Vajta, 2s. J. Chromafogr. Sci. 1979, 1 7 , 497-502. (8) Dewar, R. A. J. Chromafogr. 1981, 6 , 312-323. (9) McWllllam, I . 0. J . Chromtugr. 1061, 6 , 110-117. (10) Huennekens, J.; Gallagher, A. Phys. Rev. A 1983, 2 8 , 238-247. (11) Mans, F. A.; van Deift, R. J.; Frei, R. W.; Geerdink, R. B.; Brinkman, U. A. Th. Anal. Chem. 1986, 58, 1634-1638. (12) Patterson, P. L.; Howe, R. L. J. Chromatogr. Sci. 1978, 76, 275-280. (13) Crank, J. The Mathematics of Dlffusion, 2nd ed.; Clarendon: Oxford, 1975; Chapter 6. (14) Mollnelll. J.; Tomorawa, M.; Takata, M. J. Am. Ceram. Suc. 1985, 68, 165-168. (15) Riviere, J. C. “Work Function: Measurements and Results”; Solid State Surface Science; Green, M., Ed.; Marcel Dekker: New York, 1969; Chapter 4.
RECEIVED for review November 19, 1987. Accepted March 1, 1988.
Portable Double-Beam, Fiber-Optic-Based Photometric Comparator Paul R. Kraus, Adrian P. Wade,’ and S. R. Crouch* Department of Chemistry, Michigan State University, East Lansing, Michigan 48824
J. F. Holland Department of Biochemistry, Michigan State University, East Lansing, Michigan 48824
Brinton M. Miller Neogen Corporation, 620 Lesher Place, Lansing, Michigan 48912
An inexpensive and portable doubie-beam photometric comparator has been developed. I t has an advantage over existing singie-beam systems in that direct comparlsons about a crtticai decislon point can be made and the effects of stray radiatbn are mlnimlzed. A light-emttting diode has been used as the ilght source because of Its flxed wavelength, narrow emisskn proflie, low current requirements, and low cost. The comparator has been designed for photometric detenninations where It io deslrable to know if the analyte concentratlon is wlthin a certain range. The sample introduction system has been deslgned to accommodate mlcrotlter wells such as those often used in routlne Immunoassay procedures as the cuvettes. The unit has been designed to be operated by nontechnlcally oriented persons. I t can be battery powered for field use or operated from alternating current power.
The development and use of chemicals in agriculture, animal husbandry, and human health and nutrition have increased the awareness of the importance of the chemical environment and the effects of synthetic compounds in the flora and fauna of the earth. While modern research and service-oriented laboratories of industrialized nations are utilizing ‘Present address: Department of Chemistry, University of B r i t i s h C o l u m b i a 2036, M a i n M a l l , Vancouver, B C , Canada.
increasingly sophisticated instrumentation, there is a growing need for compact, rugged, inexpensive, and portable instruments that can perform determinations in regions isolated from the conveniences of modern laboratories. This is a consequence of the demand for instant analyses in areas remote from established facilities. In many of these situations a complete quantitative determination is not required. What is required is an indication of whether the analyte concentration in a sample is within or outside of an acceptable range. The determinations can answer such questions as: Are the active ingredients of a chemical treatment potent enough? Is the concentration of a trace contaminant below established levels? Is it safe to eat? Is it safe to discard? In such instances the analyte concentration in a sample need only to be compared to a standard. For example, allowable contaminant levels are defined for each toxin in the Code of Federal Regulations (1) and the Federal Register (2). Compliance to these rulings often reduces the determination of these species to a binary decision based upon the defined level. In this work we report the development of a double-beam, fiber-optic-based photometric instrument that directly compares two solutions. The unit compares the transmittance of a sample to that of a reference and indicates whether the sample contains a higher, lower, or approximately equivalent analyte concentration relative to the reference. Using the reference solution as the decision point calibration standard reduces the determination to a single measurement. The unit is lightweight and powered by either an automobile battery
0003-2700/88/0360-1387$01.50/0 0 1988 American Chemical Society