Anal. Chem. W Q l , 63, 1933-1942
1933
Increased Analytical Precision in the Hollow Cathode Discharge Emission Source by Improved Discharge Current Control Jih-Lie Tseng and J. C. Williams* Department of Chemistry, Memphis State University, Memphis, Tennessee 38152 Robert B. Bartlow and Steven T. Griffin Department of Electrical Engineering, Memphis State University, Memphis, Tennessee 38152 James C. Williams, Jr. Department of Anatomy and Cell Biology, Medical University of South Carolina, Charleston, South Carolina 29425
A significant improvement In the precision of the hollow cat8s an embsbn source k reporled. Precision of 1% or iem has been observed several times over periods of several hours. An average bng-term staMllty of 4.3% for U and 6.0% for Na in the emhion signal from microsamples ( G O nL) depodted in the hollow cathode dlscharge source b reported. The improved precision is attrlbuted prlmarily to the tntroductkn of dectronlccl that hold the discharge current more nearly conrtant and to the shielding of ail wiring to the source from the power supply. A currentcontrolled swttch that b capable of driving a hollow cathode discharge in etther dc or pulsed mode is described. This switch is capable of generathy current pukes as rhort as 2 ps through a mslstlve load and greatly improves the discharge stablitty and repeatability at turn-on in both the dc and puis6d modes. The characteristics of pulses produced by this switch are presented; however, analyticai performance is reported only for the dc mode. temporal current plots are presented for the new instrumentation and compared to plots taken with commonly used currentcontrolled power supplies. Instrumentation, operation, and sample preparation procedures are dercrlbed. Typical temporal profWe8 of the emlsrkn signal from microsampies deposited in Ai and stainless steel hollow cathodes are given.
INTRODUCTION The hollow cathode discharge (HCD) is a widely-used source of radiation for atomic spectroscopy. Even though the HCD is relatively free of matrix effects (1-3), has an extremely high analytical sensitivity (4-8),and provides a spectrum with very narrow emission lines (9),its use as an atomic emission source has been limited because of the poor precision that commonly has been obtained (10). It was demonstrated recently (11,12) that the simultaneous determination of Na, K, Ca, Mg, C1, and P in microsamples of common physiological fluids may be possible with the HCD with a precision of a few percent. This good precision, however, has been obtained only intermittently over the past few years (11,12). During this time, the precision that one would most likely get with the HCD remained at the 10-30% level. Such wide fluctuations in precision suggest that some parameters of the discharge were not well-controlled and that proper control might significantly improve precision. The emission signal from microsamples excited in the HCD reaches its maximum within 10-200 ms after the discharge is begun and dissipates completely within a few seconds (11, 12). Therefore, a stable discharge must be maintained from 0003-2700/9 110383-1933$02.50/0
its initiation in order to produce reliable analytical data from this transient signal. In the procedure used here for several years, a discharge is initiated at an auxiliary cathode and then switched to the hollow cathode by using a solid-state switch. This switch has no voltage or current control capability and simply provides a ground connection to the cathode. This switch is referred to as the "voltage switch" in subsequent text. A new switch was designed to control the discharge current. This improved the stability of the discharge. The new switch (Figure 1) will be referred to as the 'current-controlled switch" in subsequent text. This device allows an orderly initiation of the discharge and can maintain a relatively stable current for several seconds. Alternatively, the discharge can be terminated as quickly as 2 ps, thus giving it pulse capability. A dc and a pulsed discharge have very similar requirements with respect to control of the discharge current. In both cases, it is required that the discharge current rise smoothly to the designated limit and remain constant during the pulse period. In this.sense,the dc mode is just a very long pulse. The pulsed mode may offer distinct advantages over dc operation (13-18). However, evaluation of the pulsed mode has not be completed and only the dc analytical performance of the current-controlled switch is reported here.
EXPERIMENTAL SECTION The water-cooled HCD source used here is essentially the same as that recently reported (12). Analytical data were taken by using hollow cathodes of Cu, Al, or stainless steel (SS) (1.5 mm diameter X 5 mm deep) in argon at 11 Torr. The source has an auxiliary cathode that allows the discharge to be started and stabilized before beginning the discharge at the hollow cathode. The source is mounted on an optical rail and the light divided among a series of narrow-band interference filters (11) or monochromators (12), which are mounted along the rail. In addition to these two instruments, a third instmment usea an EG&G PARC Model 1421 photodiode array (PDA) attached to an ARC Model AM-511 spectrograph to observe spectra from the HCD source. The reciprocal linear dispersion of 0.62 nm/mm yields a resolution of 0.015 nm/photodiode. The adjustable slit was set to 10 pm for the data reported here. Sample solutions were prepared from doubly deionized water (Millipore deionized HzO, 18MQcm resistivity) with chloride salts of Na, K, and Li. Hollow cathode electrodea were drilled and then sputtered clean in the discharge chamber. The sample deposition procedure has been described recently (12). Briefly, nanoliter samples were deposited in the hollow cathode by using a micropipet mounted on a micromanipulator and a low-power stereomicroscope (Bausch & Lomb, Stereo Zoom 4). A "clean bench" (NUAIRE, Model NU-201-S24) was used as a work space for sample preparation to reduce environmental contamination. After deposition, the cathode was heated in an aluminum block for 2 min at 95 "C and then inserted into the source housing while still hot. The vacuum was applied immediately in order to rapidly 0 1991 American Chemical Society
1934
ANALYTICAL
CHEMISTRY, VOL. 63,NO.
18, SEPTEMBER 15, 1991
VBB
Flguro 1. Schematlc diagram for the current-controlled switch wlth dual current limits. See Table I for a list of the components and values.
remove volatile deposits from the warm cathode. Subnanogram quantities of analytes are quickly depleted in the discharge. This produces a temporal emission signal that lasts from only a few tenths of a second for some elements to several seconds for others. This transient signal was recorded for each optical channel by using a multichannel gated integrator constructed with Evans 4130A integrator boards and 4163A dual amplifier boards. The gated integrator, the HCD high-voltage power supply, and the timing of the discharge were controlled by Metrabyk's PIO-12 and Dash-16 boards in a Zenith PC using Pascal programs written here. The discharge procedure used in this report consisted of four steps. In the first step, a discharge was initiated at the auxiliary cathode and continued for 500 ms. In the second step, the discharge at the auxiliary cathode was terminated and, simultaneously, a low-current discharge of -5 mA was begun at the hollow cathode and continued for 200 ms. The discharge current was increased to 80 mA at the beginning of the third step and held constant for the desired excitation period. In the final step, the hollow cathode was cleaned by repeating the first three steps until the emission signal decreased no further. The initial discharge serves two purposes. First, the discharge does not begin immediately when power is applied; however, it will begin during the 500 ms allotted. This delayed discharge or lag has been discussed (19) and was reviewed recently (12). Secondly, the initial discharge period serves to ionize the gas in the source chamber, which allows a discharge to begin at the hollow cathode at a predictable time. In this manner, breakdown occurs well in advance of the analytical signal. Thus,effects of a variable time-lag in breakdown are avoided and a temporally consistent signal is generated from the small amount of analyte. Temporal emission profiles for each sample are stored on disk for subsequent data reduction. As before (11, 12), a KEPCO BHK1000-0.2M high-voltage power supply was used to drive the hollow cathode source. All connections from the KEPCO power supply to the hollow cathode source were made with coaxial cable (Belden, M8214 1C11,RG8, 50 Q,E34972). Both earlier instruments (11,12)were fitted with the new cable. A common grounding point was chosen on the metal rack holding each instrument. The current was monitored by observing the voltage drop acrosa a 50-Qresistor, as shown in Figure 2. One of the six analytical channele available in the gated integrator was used to make a temporal record of the current during every discharge period. Data from all channels were presented graphidy on the CRT and optionallyby a dot matrix printer. The data were stored on disk for subsequent analysis. The integration times used to record the temporal analytical signals ranged from 1to 6 ms. Thus,only relatively low frequency changes in the current could be observed in this manner. A Tektzonix 2430A digital oscilloscopeequipped with a camera was used to make oscillographsof the current during the discharge, giving information about much higher frequency changes in the
BALLAST
r-+r--!
PCWER SUPPLY
j
CATHODE
500k
(A) BALLAST
ANCDE j
i
CATHJDE
rfi (B) F w e 2. Block diagrams of the two instrumental setups used to test the current-controlled switch. The switch is the diamond-shaped symbol. The hollow cathode in (A) Is a 1OX scale-up of the cathode used to take the analytical data reported here. The dlscharge voltage c a n be monitored in (A). The configuration In (e) glves the hlghest operation speed. This small cathode was 1.5 X 5 mm and was used
to take analytlcal data reported here.
discharge current. Note that the time-base dependent, single-shot bandwidth of modem digital scopes may not reveal high-frequency signals when set for very slow scan rates. Some models have the ability of storing the high and low value for each time interval, thus revealing the presence of high-frequency signal. However, several data points must be taken during a transient period in
ANALYTICAL CHEMISTRY, VOL. 83, NO. 18, SEPTEMBER 15, 1991
order to obtain detailed information of ita magnitude and frequency. Thus,because the oscilloscope memory is usually only 1024 pointa, the highest frequency components can only be observed with the fastest mans (20). Design of the Current-Controlled Switch. Using either a pulsed or a dc diachage for quantitative analysis requiresa switch with reliable transient behavior. Driving a discharge in a currenbcontdled mode ia desirable because the discharge has higher stability when the current density is controlled. In addition, dual current limits, a low standby current and a higher main discharge current, have been used to enhance system stability (21). These requirements are best met by incorporating current regulation into the switch. This provides a more consistent transient response as compared to approaches that depend on the current-limiting circuitry in dc power supplies to control the current. The power supply currenblimiting circuitry is designed for overload protection and can neither control high-speed current transients nor generate high-speed current pulses. Prior efforts at regulatedcurrent switch design include one by Malmstadt (22),which was designed for atomic fluorescence spectroscopy (23).His design exhibited excellent long-term stability, but transient behavior was not examined. It is likely that common-mode effects in the current-sensing portion of the circuit could be considerable. Other schemes have been used to stabiliza the discharge current, but their usefulness in pulsed discharges is limited. The most common uses a voltage-regulated power supply Vm, a ballast resistor R-, and a nonregulated current switch in series (19). The discharge current is calculated as ID
VPS- VD -
4”t
where VD is the voltage drop across the discharge. This configuration has two principal limitations. First, the current limit, and therefore the discharge stability, depends on the discharge voltage. This voltage is dependent in a highly nonlinear manner on other discharge parameters such as pressure, fill gas, and electrode material and geometry. Hence, small parametric variations can result in large changes in the discharge voltage. Though these problems may be treated by using an extremely high-voltage power supply and a large ballast resistor, the values are so extreme in high-current applications as to be impractical. The second limitation occura when the switch transistor is driven into saturation. This slows ita response to the point of rendering the approach unusable for generating short pulses. A similar pulsing technique uses a power MOSFET to shunt a portion of the supply current around the discharge (24). As with the previous technique, the current limit depends on the discharge voltage, leading to the same problems with parametric variations. Another common confiation, referred to here as the voltage switch ( I I , 1 2 ) , is implemented with a current-regulated power supply and a switch that has no current control. The transient response of this configuration depends on the current-regulating circuitry of the dc power supply being fast and well-behaved. The interaction of the current regulation circuitry and the non oneb o n e correspondence between the discharge voltage and current produces a classic relaxation oscillator. The oscillation period depends on the aforementioned discharge parameters plus the ballast resistor. The analytical consequences of this are significant (see Figure 8). As a result, the approach detailed below has been implemented to generate current-regulated analytical discharge excitation pulses with a short risetime. The current-controlled switch depicted in Figure 1 features a current-limiting feedback circuit with dual set pointa and optoisolation of the digital control signals. The main-switch transistor Q2 is on when the output transistor of the ON/OFF o p toisolator Q1is in cutoff. Conversely, when Q1 is in saturation, Q2is forced into cutoff and the switch is off. A diode D1 has been included to ensure that Q2 goes completely into cutoff by not allowing a forward bias on the emitter-basejunction. A 5 V power supply VBBprovides a forward bias to the emitter-base junction of Q2through a pull-up resistor Rpl. The maximum base current available to Q2is,,Z VBB/R,~and is realized only when the feedback transistor &s is in cutoff. Negative feedback is accomplished by redirecting a portion of the current from the base of
-
IS35
Table I. List of Component Values for the Current-Controlled Switch Shown in Figure 1 component 4 1
Qz
value 100 a (1/2 W,2% tolerance) SK9411 (B 7, SVc, = 1500 V,Pf~
100 W,
4 MHz
8133444 (B = 150,fT = 100 MHz) a series combination of a 20 and a 5004 15-turn potentiometer a series combination of a 2 4 and a 104 15-turn R2 potentiometer 1.5 ka RPZ 1N4007 D1 optoisolators HllL2
Q Q6
R1
Qz through Qsto the switch common, which is at the negative side of V D The maximum base current to Q2 as set by feedback is calculated as
where VBEnis the voltage across the base-emitter junction of Q, and V, is the voltage drop across D,. Thus,the discharge current can be calculated as ID= I B d 2 , where p2 is the current gain of
Qz.
The low, or standby, discharge current setting is selected when the mode selection transistor Q is in cutoff. This occurs when the output transistor of the HI/LOW optoisolator Q, is in saturation. In this mode, the discharge current is sampled by the primary sampling resistor R1. When the voltage across the base-emitter junction of % reachea approximately 0.67 V, &s turns on, reducing the base current of Q2to ID/&. The standby discharge current ID,, is calculated by eq 1.
The high discharge current setting is selected when Q is in saturation, giving a discharge current of ZD = ID,, + Z~ditl.In this mode, additional discharge current I- is required to turn Q3 on because a portion of the current through R1has been redirected through the secondary current sampling resistor R2 and Q. This additional load current is calculated in eq 2, where VC- is the potential difference between the collector and emitter of Qa.
The output transistor of each optoisolator Q1and Q, is driven by a Schmitt trigger. The output of a Schmitt trigger is a binary function of the input signal, greatly simplifying system design. In contrast,the output current of standard phototransistor-output optoisolators is a more linear function of the input current, thus making the system design depend on the electrical characteristics of the circuit driving the optoisolator. Furthermore, the output transition time for the Schmitt-trigger optoisolator used here (HllL2) is roughly 0.5 ps (though faster unita are available), whereas the output transition time for a phototransistor-output optoisolatoris limited to roughly 3 pa. The components selected to allow the switch to have low and difference current settings of roughly 1-30 mA and 35-205 mA, respectively, are given in Table I. Two different HCD sources were used to test the switch. The first consisted of a stainless steel (SS) hollow cathode (11mm diameter X 25 mm deep, type-304) in argon at 1.5 Torr. This scaled-up version was used here only for testing of the currentcontrolled switch (25). The second was the type used to take the analytical data for this paper; it consisted of hollow cathodes of Cu, Al, and SS (1.5 mm diameter X 5 mm deep) in argon at 11 Torr. The product of the hollow diameter and the fill gas pressure for each setup is 1.65 Torr cm. The constant PD product gives
1936
ANALYTICAL CHEMISTRY, VOL. 63, NO. 18, SEPTEMBER 15, 1991
approximately the same excitation conditions in the hollow cathode (19). The setups used to test this switch are depicted in Figure 2. Both configurations have a 50-Qresistor in series with the discharge between the switch common and ground for measuring the discharge current with an oscilloscope. The two resistors in series (500 kQ and 50 R) across the discharge in Figure 2A act as a voltage divider to view the discharge voltage on an oscilloscope. The Configuration depicted in Figure 2B is faster than that in Figure 2A due to decreased stored charge in the switch. The stored charge is equal to the product of the stray capacitance in the switch and the voltage from the collector of Q2to ground. This voltage is much higher for the configuration in Figure 2A than in Figure 2B. The stray capacitance is unchanged. All instrumentation cables were either RG-58/U or RG-l74/U (504 impedance). The clock in the first setup was a multivibrator integrated circuit (TLC555)configured with separate controls for pulse width and interpulse delay. The second setup was controlled by a gate-delay generator (ORTEC Model 416A) gated by a microcomputer. Purpose of Ballast. In the most common power supply configuration used to drive a discharge, the ballast resistor limits the discharge current. In the configuration used here, the ballast reduces the amount of heat that the main switch transistor must dissipate. The discharge current, as calculated by using eqs 1 and 2, is a function of &. fI2 is, in turn, a function of the temperature. Reducing the power dissipated by the main switch transistor enhances long-term current stability by reducing transistor heating. The heat dissipated by a transistor is the product of the collector-emitter voltage and the collector current. For example, consider a discharge (ID = 200 mA, V D = 300 V) driven by an 800-V power supply. With no ballast,the mainswitch transistor dissipates (800 V - 300 V)(O.2 A) = 100 W of power. The addition of a 2-kil ballast, dissipating (0.2 A)2(2000Q) = 80 W,requires the main switch transistor to dissipate only 20 W of power. However, a 2.5-kR ballast would be too large: It would force the main switch transistor into saturation and defeat the current regulation. Cost. The total cost for this switch was about $35. Over half of this figure covered the cost of the 5-V power supply, the main switch transistor, and the heat sink for the main switch transistor. This price includes neither a printed circuit board nor a case. A detailed parts list is available upon request.
RESULTS AND DISCUSSION Validation of the Current-Controlled Switch. The range over which the current could be set was found experimentally to be 8-180 mA. The difficulties encountered a t limit currents greater than roughly 180 mA were due to the current gain, 8, of the main switch transistor (Table I) being closer to 6 than the value used for selecting IB2 and RP1of 7. This limitation was not a problem for this project. Below about 8 mA, the low current limit behavior of some of the switches was oscillatory. One, however, was stable to less than 5 mA. The difference is attributed to variations in the physical layout of components and wiring. Long-term drift in the discharge current limit was noted when the switch was used to generate pulses longer than about 1s. This was due primarily to thermally induced changes in the current gain of the main switch transistor. For low duty cycles and relatively short pulses (
