ELECTRICAL CHARACTERIZATION OF THIN-FILM

P1: NBL/SDA/SNY P2: SDA/MKV QC: SDA

May 23, 1997 15:30 Annual Reviews AR034-08

Annu. Rev. Mater. Sci. 1997. 27:223–48Copyright c© 1997 by Annual Reviews Inc. All rights reserved

ELECTRICAL CHARACTERIZATIONOF THIN-FILMELECTROLUMINESCENT DEVICES

J. F. Wager and P. D. KeirDepartment of Electrical and Computer Engineering, Center for Advanced MaterialsResearch, Oregon State University, Corvallis, Oregon 97331-3211;e-mail: [email protected]

KEY WORDS: electroluminescence, AFTFEL, ZnS, SrS, electrical characterization

ABSTRACT

Electrical characterization methods for the analysis of alternating current thin-film electroluminescent (ACTFEL) devices are reviewed. Particular emphasis isdevoted to electrical characterization techniques because ACTFEL devices areelectro-optic display devices whose performance is to a large extent determinedby their electrical properties. A systematic procedure for ACTFEL electrical as-sessment is described. The utility of transient charge, voltage, current, and phos-phor field analysis is explained. Steady-state electrical characterization methodsdiscussed in this review include charge-voltage (Q-V), capacitance-voltage (C-V), internal charge-phosphor field (Q-Fp), and maximum charge-maximum ap-plied voltage (Qmax-Vmax) analysis. These electrical characterization methods areillustrated by reviewing relevant results obtained from the analysis of evaporatedZnS:Mn and atomic layer epitaxy (ALE) SrS:Ce ACTFEL devices.

INTRODUCTION

An alternating-current thin-film electroluminescent (ACTFEL) device is a solidstate display device in which some type of an alternating-current (ac) voltagewaveform is applied to the electrical contacts of a thin-film stack in orderto induce light emission (1). A typical ACTFEL stack consists of a phosphorlayer doped with a luminescent impurity (e.g ZnS:Mn, SrS:Ce, etc) sandwichedbetween two insulators that are contacted via one opaque and one transparentelectrode. The ACTFEL stack described employs a glass substrate, and light

2230084-6600/97/0801-0223$08.00



224 WAGER & KEIR

is viewed from the substrate side. Alternatively, in an inverted structure, theelectrode positions are reversed, and light is viewed from the opposite side ofthe ACTFEL device.

From a device physics perspective, a simple view of the operation of an ACT-FEL device involves six primary physical processes, as illustrated in Figure 1.Upon the application of a sufficiently large voltage to the ACTFEL device, thephosphor field is large enough that electrons trapped in interface states aretunnel-emitted into the phosphor conduction band (Process 1). Subsequently,these injected electrons gain energy from the field and are transported acrossthe phosphor (Process 2). As these hot electrons transit the phosphor layer, afraction of them excite luminescent impurities (Process 3) from their groundstate to their excited states; obviously, the hot electron must concomitantly loseenergy during this impact excitation process. Subsequently, the luminescentimpurity relaxes from its excited state to its ground state (Process 4), possiblygiving off a photon during this energy relaxation process (the energy relaxationprocess can also occur nonradiatively, in which case the potential energy ofthe luminescent impurity in its excited state is dissipated via the emission ofphonons to the lattice). The delay time between Processes 3 and 4 can be long(ms) or short (a fraction of a microsecond) depending on whether spin andparity selection rules for the radiative decay process are allowed or forbidden.

Figure 1 Energy band diagram illustrating the six primary physical processes of importance fordescribing the device physics operation of an ACTFEL device.



EL CHARACTERIZATION 225

Eventually, the transported electrons reach the phosphor/insulator anode inter-face and are trapped in interface states (Process 5). Finally, photons generatedby radiative recombination outcouple from the ACTFEL stack and are observedby the viewer (Process 6).

It is our contention that an ACTFEL device is primarily an electro-opticdevice because although an ACTFEL device is employed for visual display ap-plications, how it functions is primarily determined by its electrical properties.Thus electrical characterization of ACTFEL devices is the focus of this review.Moreover, the aim of this chapter is to review electrical characterization tech-niques that we have found particularly useful for unraveling the device physicsof ACTFEL devices. Therefore, this review is not meant to be a balanced andexhaustive overview of all the methods that have been employed for electricalcharacterization of ACTFEL devices.

First, we present a survey of ACTFEL electrical characterization techniques.Prior to discussing specific characterization techniques, the general circuit andwaveform we employ for two-terminal ACTFEL assessment are described.Then the electrical characteristics of evaporated ZnS:Mn and atomic layer epi-taxy (ALE) SrS:Ce ACTFEL devices are compared; these two kinds of ACTFELdevices are discussed because they exhibit distinctly different kinds of ACTFELdevice behavior. Additionally, ZnS and SrS are currently the technologicallymost important ACTFEL phosphors.

ELECTRICAL CHARACTERIZATION TECHNIQUES

Circuit and Waveform EmployedThe circuit we normally employ for two-terminal electrical characterizationof ACTFEL devices is shown in Figure 2. An arbitrary waveform generator(AWG) is used to provide the desired ac signal, which is then amplified viaa high-voltage amplifier. This voltage waveform drives a series combinationof a series element, the ACTFEL device, and a sense device. Typically, weemploy a 100� to 1 k� resistor as a series element. The purpose of thisseries element is to act as a current-limiter, which helps to protect the ACTFELdevice from catastrophic breakdown; the larger the series resistor, the morethe ACTFEL device is protected. However, although a series resistor helpsto protect the ACTFEL device from catastrophic breakdown, the resistor alsoincreases the time constant of the circuit, which obscures observation of thedynamic response of the ACTFEL device. Normally, we use a capacitor as asense element. The capacitance of this sense capacitor is much larger than thatof the ACTFEL device under test, so most of the applied voltage occurs acrossthe device. Essentially, our test circuit arrangement is that of a Sawyer-Towerarrangement (1–4) in which the device under test is monitored via measurementof the voltage across the sense capacitor.



226 WAGER & KEIR

Figure 2 Circuit used for two-terminal electrical characterization of an ACTFEL device.

As indicated in Figure 2, the transient voltage may be monitored, usually viaa digitizing oscilloscope, at three different points in the circuit. A wealth ofdevice physics information may be obtained from a measurement of these threevoltage transients (discussed below).

The standard applied voltage waveform we typically use to accomplish elec-trical characterization is illustrated in Figure 3 and consists of a 1 kHz sequenceof bipolar pulses with rise and fall times of 5µs and a pulse width of 30µs.Note that an A-J labeling scheme is used to designate certain important points inthe applied voltage waveform. The significance of the labeled points except Band G is obvious and requires no further explication; B and G refer to points inthe applied voltage waveform in which conduction in the phosphor is initiated.We use this labeling scheme throughout this review.

Transient Charge, Voltage, Current,and Phosphor Field AnalysisSeveral kinds of transient characteristics are often used for assessing the devicephysics operation of an ACTFEL device. The instantaneous external charge,




Figure 3 Standard applied voltage waveform used for ACTFEL electrical characterization. TheA–J labels are used to indicate certain important points in the applied voltage waveform.

qext(t), is usually determined by measuring the voltage across the sense capacitorshown in Figure 2 and multiplying by the sense capacitance to give

qext(t) = Csv3(t). 1.

The instantaneous voltage drop across the EL device,vEL(t) is simply given by

vEL(t) = v2(t)− v3(t), 2.

wherev2(t) andv3(t) are indicated in Figure 2. Althoughqext(t) andvEL(t) curvesare sometimes plotted and used directly for ACTFEL device assessment, it ismore common to use these curves as raw data input for the accomplishment ofQ-V, C-V, Q-Fp, or Qmax-Vmax measurements. We have recently begun using aprocedure in which both sets of curves are offset-adjusted prior to being furtherprocessed to accomplish other kinds of measurements (4a, 4b).

The idea underlying offset adjustment of theqext(t) curve (S Shih et al, paperin preparation) is that this curve is often found to be asymmetrically displacedfrom the origin of the voltage axis. This asymmetric displacement from thevoltage axis origin is referred to asqext(t) offset. Aqext(t) offset yields Q-V curvesthat are displaced vertically above or below the voltage axis and Q-Fp curves thatare displaced laterally above or below the phosphor field axis. Our procedurefor qext(t) offset adjustment is to simply displace theqext(t) curve so that it issymmetrically positioned from the voltage axis origin. The amount of voltageshift that is required to symmetrically displace theqext(t) curve is recorded andis regarded as a measure of the extent of theqext(t) offset. Simulation indicates



228 WAGER & KEIR

(4a) thatqext(t) offset may be associated with an asymmetry in the interfacestate depths at the two phosphor/insulator interfaces or with an asymmetry inthe location of space charge in the phosphor.

The second kind of offset procedure that we employ (4b) is to monitorvEL(t)and to reset the baseline ofvEL(t) to ensure that the baseline occurs at zero volts.We find thatvEL(t) curves typically require an offset adjustment of∼1–3% ofVmax due to small and random offset inaccuracies in the oscilloscope.vEL(t)offset is manifest in a Q-V curve as a displacement of the leakage charge fromthe charge axis.

qext(t) is the transient charge measured externally with respect to the phosphorof an ACTFEL device. From a device physics perspective, it would be moreuseful to know the internal transient charge present at the insulator/phosphorinterface, which we denoteq(t). q(t) is found (5, 6) from

q(t) = (Ci + Cp)

CiCsv3(t)− Cp[v2(t)− v3(t)], 3.

where Ci and Cp are the insulator and phosphor capacitances, respectively. Inorder to obtain a good estimate ofq(t), Ci and Cp must be accurately known.q(t) curves are primarily used as the input data for Q-Fp and Qmax-Vmax assess-ments. Note that the difference between aq(t) curve and aqext(t) curve involvesrescaling and subtraction of displacement charge.

An example of a transient current [i(t)] curve is shown in Figure 4.i(t) maybe found from the voltage drop across the series resistor shown in Figure 2,

i (t) = v1(t)− v2(t)

Rseries. 4.

Alternatively,i(t) may be found fromqext(t) by differentiation,

i (t) = dqext(t)

dt. 5.

Note thati(t) is the total current flowing in the external circuit; thus,i(t) iscomprised of a superposition of a displacement and a conduction current con-tribution, idisplace(t) and icond(t), respectively. Often, the conduction currentcontribution is the one of prime interest; this has led many researchers (2, 4, 7)to employ a bridge circuit in which the displacement current contribution is ze-roed below threshold so that only conduction current is measured directly. Wehave not used this approach extensively in our work because we find it extremelydifficult to accurately and reliably zero the displacement current. Therefore,the reader is referred to the literature (2, 4, 7) for further discussion of this tech-nique. If we need to evaluateicond(t), we prefer to do so via differentiation ofthe internal charge transient curve

icond(t) = dq(t)

dt. 6.




Figure 4 (a) vEL(t) andfp(t) and (b) qext(t) andi (t) curves for an evaporated ZnS:Mn ACTFELdevice operated at a Vmax∼ 40 V over threshold.

i(t) curves are used primarily as raw data for accomplishing C-V characteri-zation, for investigating whether trailing edge current is present (8, 9), and forexploring the detailed nature of the charge injection dynamics (10).

Consider the dynamic nature ofi(t) shown in Figure 4. From A to B, onlydisplacement current flows. At B, the flow of conduction current is initiated andtypically reaches its maximum value near point C. The magnitude of current



230 WAGER & KEIR

flow is reduced from C to D, during which time the applied voltage is heldconstant at its maximum value, Vmax. The phosphor field decreases or relaxesduring the CD portion of the waveform; thus the charge transferred during theCD portion of the waveform is referred to as relaxation charge. Finally, DE isthe falling edge portion of the waveform, during which time the applied voltageis reduced to zero. Most of the current flowing during this DE portion of thewaveform is displacement current.

An example of a transient phosphor field [fp(t)] curve is shown in Figure 4.fpis assessed using the circuit shown in Figure 2 and via the use of the followingequation (5, 6),

fp(t) = 1

dp

{Csv3(t)

Ci− [v2(t)− v3(t)]

}, 7.

wheredp is the phosphor thickness. Note that an accurate assessment offp(t)requires that the phosphor thickness and the insulator and phosphor capaci-tances be accurately determined.fp(t) curves are most often used as input datafor accomplishing Q-Fp analysis and for investigating the detailed dynamicsassociated with field-clamping, trailing edge emission, or field relaxation (seebelow).

Consider the nature of the dynamics of thefp(t) curve shown in Figure 4. Thiscurve possesses a small duration during the rising edge portion of the voltagepulse in which the phosphor field is nearly constant with time; this constantfield duration occurs at the peak of thefp(t) curve shown in Figure 4 but is toosmall to be noticeable. We define the phosphor field at this time as the steadystate field Fss. If Fss is independent of the maximum applied voltage, Vmax,above threshold, the device exhibits field-clamping. The idea of field-clamping(1–3, 5) is that for a large enough phosphor field, the emission rate of electronsfrom interface states and the concomitant rate of reduction of the phosphorfield is equal to the slew rate of the phosphor field from the increasing appliedvoltage. Although some literature suggests otherwise, it is our experience thattrue field-clamping is observed in very few ACTFEL devices. Rather, whatwe denote as pseudo-field-clamping is often found in many kinds of ACTFELdevices; by this we mean that Fssvaries slightly with Vmax. One sees in Figure 4that after the steady state field portion of thefp(t) waveform, the phosphor fieldrelaxes as a function of time. We refer to this portion of the curve as relaxation;in particular, we refer to the charge that is transferred across the phosphorduring this portion of the waveform as relaxation charge (11). Phosphor fieldrelaxation occurs during the CD or HI portions of the waveform, during whichtime the applied voltage is constant at its maximum value. Since the externalapplied voltage is constant, the phosphor field must relax as more charge is




transferred across the phosphor. Finally, during the DE or IJ portions of thewaveform, the polarity of the phosphor field reverses sign because the appliedvoltage is removed, and the field set up by the charge that was transferred bythe previous voltage pulse is of opposite polarity to that of the field set up bythe applied voltage of the previous pulse.

Charge-Voltage AnalysisCharge-voltage (Q-V) analysis is the classical technique (1–4) used to electri-cally characterize ACTFEL devices. Q-V analysis is accomplished by plottingqext(t) versusvEL(t), as given in Equations 1 and 2. Note that the charge Q plot-ted in a Q-V curve is really the external charge that is detected in the circuit andis measured as a voltage across the sense capacitor; this is in contradistinctionto the internal charge, which is the actual charge inside the ACTFEL device atone of the phosphor/insulator interfaces. As explicitly shown in Equation 3,internal and external charges are not identical. Also note that the voltage V ina Q-V curve is the voltage drop across the ACTFEL device. This is slightlydifferent than the applied voltage as measured at the output of the high-voltageamplifier because some of the applied voltage falls across both the series andthe sense elements. In order to avoid confusion, it would be more appropri-ate to refer to the Q-V technique as the external charge/EL voltage technique;however, we do not do so here because this approach is universally referred toas the Q-V technique.

A Q-V curve is obtained using a single voltage waveform in which Vmaxremains constant. Thus the voltage V in a Q-V measurement is actually theinstantaneous voltage drop across the ACTFEL device, and Q is an instanta-neous external charge monitored outside the ACTFEL device. Usually a Q-Vcurve is evaluated after the ACTFEL device has established a steady state withthe applied voltage waveform; typically, steady state is established after theACTFEL device has been subjected to∼3–4 periods of the applied voltagewaveform.

A somewhat idealized Q-V curve illustrating the primary features often ob-served in a ACTFEL device is shown in Figure 5. The A-J labeling schemeintroduced in Figure 2 is also used in Figure 5. Superscripts+ and− corre-spond to the polarity of the applied voltage pulse; our convention is that thepolarity of the pulse is defined with respect to the top electrode (i.e. the elec-trode farthest away from the substrate). The superscript e is used to designatethe relevant charge as an external charge. The absence of an e superscript in-dicates the charge is internal. An analysis of Figure 5 indicates the chargesshown are external when below turn-on and are internal when above turn-on.The idea (5) is that below turn-on the charge measured in the external circuit



232 WAGER & KEIR

Figure 5 An idealized Q-V curve in which certain important points and various charges andvoltages are labeled.

is not identical to the internal charge at the phosphor/insulator interface. How-ever, subsequent to the onset of strong conduction in the phosphor, the externalcharge transferred across the sense capacitor is identical to the charge that flowsacross the phosphor (more precisely, these two charges are identical only whenthe conduction current density is large enough so that the phosphor capacitancemay be assumed to be shunted out of the circuit).

Now consider the nature of the charges shown in Figure 5. Qcond is theconduction charge transported across the phosphor during the voltage pulse;this is the charge responsible for impact excitation of the luminescent impuritieswhich, hence, gives rise to light emission. Qe

pol is the polarization charge storedat the phosphor/insulator interface just prior to the onset of the subsequent pulseof opposite polarity. Qeleak is the leakage charge arising from the emission ofelectrons from shallow interface states during the zero voltage portion of thewaveform. Note that the magnitude of the leakage charge depends upon boththe magnitude of the polarization field at the electron-emitting interface and thedepth of the interface states. Finally, Qrelax is the relaxation charge that flows




across the phosphor during the portion of the waveform at which the appliedvoltage is constant at its maximum value; the phosphor field decreases duringthis portion of the waveform; hence the designation of relaxation charge.

In addition to the charges shown in Figure 5, two turn-on voltages, Vto, areindicated. Note that these voltages are referred to as turn-on voltages insteadof threshold voltages. The distinction between turn-on and threshold voltagesis as follows: The magnitude of a turn-on voltage depends upon the value ofVmax employed (the turn-on voltage depends upon Qpol and, hence, on Vmax),whereas a threshold voltage is a constant value that does not depend upon themagnitude of Vmax. It can be shown that the threshold voltage is equal to theturn-on voltage evaluated in the limit in which the polarization charge goes tozero.

Consider one period of a Q-V waveform. First, note that a Q-V loop precessesin a counter-clockwise manner. The point labeled A in Figure 5 corresponds tothe onset of a positive voltage pulse applied to the upper electrode (see Figure 3).A nonzero value of Q is observed at A because of polarization charge residingat the upper phosphor/insulator interface, which is left behind by the previouspulse of opposite polarity. The AB portion of the Q-V curve arises from therising edge of the external voltage when the magnitude is less than that requiredfor turn-on of the ACTFEL device. BC also occurs during the rising portion ofthe external voltage pulse but the voltage magnitude is greater than the turn-onvoltage for the BC portion of the Q-V curve. CD corresponds to the portionof the waveform in which the external voltage is held constant at its maximumamplitude. Section DE of the Q-V curve is obtained during the falling edge ofthe voltage pulse. EF corresponds to the segment of the waveform in whichno external bias is applied to the ACTFEL device. The remainder of the Q-Vcurve from F to A is similar to the A to F portion of the curve just describedexcept that the external applied voltage pulse is of opposite polarity.

Finally, the area enclosed within a Q-V curve is equal to the input electricalpower density delivered per pulse to the ACTFEL device (1).

Capacitance-Voltage AnalysisA capacitance-voltage (C-V) curve is generated by plotting the dynamic capac-itance as a function of the dynamic voltage drop across the ACTFEL device[i.e. c(v2−v3)] (12, 13). There are two typical ways of evaluating the dynamiccapacitance. One approach involves first measuring the current, using Equation4 or 5. The dynamic capacitance is then given by

c(vEL) = i (t)dvEL(t)

dt

. 8.



234 WAGER & KEIR

Alternatively, a C-V curve may be assessed as simply the derivative of a Q-Vcurve because the dynamic capacitance may also be defined as

c(vEL) = dqext(t)

dvEL(t). 9.

Note that this alternative way of viewing a C-V curve would seem to suggest thata C-V curve is of little utility because it is merely the slope of the Q-V curve, andthus, it would seem that all the relevant information is contained within a Q-Vcurve. In practice, we have found C-V analysis to be an important extensionof and a complement to the Q-V method because subtle details in a Q-V curveare amplified and become more readily evident upon differentiation of a Q-Vcurve to obtain the C-V curve. In this respect, C-V analysis is analogous tomodulation spectroscopy (14) in which the first, second, and sometimes eventhird derivative of the raw data are obtained in order to remove background andenhance subtle features in the measured data.

The primary problem associated with accomplishing a C-V measurementis that the raw data must be differentiated to obtain the dynamic capacitance.Because differentiation is a high-pass filtering operation, it is crucially importantto obtain raw data in a manner that maximizes the signal-to-noise ratio so thatthe lower frequency dynamic capacitance signal can be extracted from thehigher frequency noise. We have found the two approaches for estimating thedynamic capacitance (as specified by Equations 8 and 9) to be equivalent ifthe signal-to-noise ratio is identical for the two sets of raw data.

Three idealized C-V curves are shown in Figure 6 in order to illustrate thetype of information available from C-V analysis. These three curves are meantto indicate what such curves would look like for an ideal ACTFEL device sub-jected to three different Vmax; in practice, we typically use Vmax= 20, 40, and60 V above threshold for standard C-V characterization. First, note that thesub-turn-on capacitance, denoted Ccv

t , corresponds to the total capacitance ofthe ACTFEL stack (i.e. a series combination of the capacitances associated withthe phosphor and the two insulating layers). Usually we find very good agree-ment between Ccv

t and the total physical capacitance Cphyst (which we define as

the series combination of the phosphor and insulator capacitances, as calculatedfrom the expected thickness and dielectric constant of each individual thin-filmlayer). When there is a significant difference between Cphys

t and Ccvt , we find

that this is almost invariably due to an inaccuracy in estimating the thickness ofthe phosphor or insulator. This and other aspects of C-V analysis are very usefulfor ACTFEL process monitoring. In an ideal ACTFEL device, the above turn-on capacitance, Ccv

i , is equal to a series combination of the capacitances of thetwo insulators and is in good agreement with the physical insulator capacitanceCphys

i . In practice, there are few ACTFEL devices where Ccvi = Cphys

i .




Figure 6 Three idealized C-V curves obtained for three different Vmax. The arrow indicates thatthese curves shift to lower voltages as Vmax increases.

There are two kinds of deviation from this ideal situation (15). In the firstcase, we find Ccv

i < Cphysi . Such a situation arises (16) if the magnitude of the

conduction current is inadequate to completely shunt the phosphor capacitance.We rationalize such a C-V curve by saying that there is an inadequate density ofinterface states to reach a condition of field-clamping or pseudo-field-clamping.More precisely, such a C-V curve is a consequence of having an inadequatelylarge conduction current; the magnitude of the conduction current depends onthe magnitude of the phosphor field, the depth and density of the interfacestates, and the electron multiplication properties of the phosphor (i.e. electronmultiplication implies band-to-band or trap-to-band impact ionization, or trap-to-band field ionization). The second kind of insulator capacitance deviation iswhere Ccv

i > Cphysi . This situation arises when dynamic space charge generation

occurs within the phosphor (17, 18). Additionally, Ccvi is found to be greater

than Cphysi only if space charge is both created and trapped or localized within

the bulk region of the phosphor (19). Thus when electron multiplication byband-to-band impact ionization occurs, Ccv

i = Cphysi , since the created holes

are transported to the cathode phosphor/insulator interface in∼5 ps, which isessentially instantaneous with respect to the microsecond scale at which C-Vmeasurements are accomplished.



236 WAGER & KEIR

We have observed two types of C-V deviations in which Ccvi > Cphys

i (18).In the first case, the dynamic capacitance increases to a value of capacitancein excess of Cphys

i and then decreases to a lower value of capacitance but stilllarger than Cphys

i ; we refer to this as capacitance overshoot. The second kindof C-V deviation occurs when the dynamic capacitance increases to a valuegreater than that of Cphys

i and saturates. Simulation indicates that both C-Vdeviations in which Ccv

i > Cphysi are a consequence of dynamic space charge

generation; differences between whether overshoot or saturation are observedappear to be related to the dynamics of space charge generation (20), althougha detailed understanding of these differences is not available.

Each C-V curve possesses three turn-on voltages, Vto1, Vto2, and Vto3, cor-responding to the onset, mid-point, and saturation of the C-V transition, re-spectively. Vto1 is the voltage across the ACTFEL device corresponding to theonset of emission of electrons from the most shallow filled interface states orbulk traps. Vto3 corresponds to the initiation of field-clamping or pseudo-field-clamping. Vto2 corresponds most closely to the turn-on voltage found in a Q-Vmeasurement; thus, Vto2 corresponds to an average turn-on voltage. Note thatthese three voltages are referred to as turn-on voltages instead of thresholdvoltages (13, 21).

The slope of the C-V transition between Vto1 and Vto2 is envisaged as ameasure of the preclamping interface state density (13). In an ideal ACT-FEL device, the interface state density is negligible from the conduction bandminimum to∼1 eV below the phosphor conduction band minimum where thedensity increases abruptly. This type of interface state distribution yields a steepC-V transition. In contrast, suppose that the interface state density increasesgradually from the conduction band minimum to∼1 eV below the phosphorconduction band where the interface state density is large enough that field-clamping or pseudo-field-clamping occurs; in such a situation, a C-V transitionwith a smaller slope is observed. Quantitatively, the preclamping interface statedensity may be assessed (13) as

Qss= C2i

2qA

Ct

Cp

[1C

1V

]−1

, 10.

where A is the device area, and the term in the square bracket is the slope ofthe C-V transition.

There are two potential pitfalls inherent in this interpretation of the C-Vslope as a measure of the preclamping interface state density. First, the C-Vmeasurement is a dynamic measurement in which the voltage plotted on thex-axis is being slewed at a high rate (∼20–40 V/µs); therefore, only interfacestates with relatively short time constants are monitored in a C-V measurement.




True interface state assessment involves measurement of interface states withall possible time constants in an equilibrium-like manner (22), so a dynamicC-V measurement precludes accomplishing a true interface state measurement.Second, in some ACTFEL devices not all of the transferred charge derivesfrom interface states; some charge also arises from bulk states in the phosphor.For these reasons, our view is that the slope of the C-V transition should beconsidered a measure of the abruptness of turn-on. We find the C-V slope tobe a very useful relative measure of this abruptness if we compare ACTFELdevices measured under similar experimental conditions.

The turn-on voltages of the three C-V curves shown in Figure 6 shift rigidlyto smaller voltages with increasing Vmax. The rigid shift to smaller voltages ischaracteristic of an ACTFEL device in which the electrons derive exclusivelyfrom interface states. In contrast, we find that if a substantial fraction of thetransferred charge is from the phosphor bulk, Vto1 is fairly constant (i.e. inde-pendent of Vmax) at a small voltage, the initial portion of the C-V transition nearVto1 is rather washed out, and the C-V curves shift in a non-rigid manner inwhich the C-V slope usually decreases as a function of increasing Vmax. Thusthe shape of a C-V curve and its dependence upon Vmax are good indicators ofwhether the transferred charge is interface or bulk derived. In certain ACTFELdevices, we have observed the C-V characteristics to shift to larger voltageswith increasing Vmax. This unusual behavior is linked to an abnormal reductionin the polarization charge with increasing Vmaxand is invariably associated withspace charge generation in the phosphor.

Internal Charge-Phosphor Field AnalysisThere are two limitations of a Q-V curve in terms of the device physics informa-tion it contains. First, it is not easy to directly compare the relative magnitudesof various kinds of charge (e.g. polarization and conduction charge) within aQ-V curve because some of these charges are internal and others are external.Second, the x-axis of a Q-V curve involves the voltage applied across the entireACTFEL stack. From a device physics stand point, it is more useful to plotthe phosphor field on the x-axis. The internal charge-phosphor field (Q-Fp)technique was developed to address these two Q-V limitations (23).

A Q-Fp curve is generated from raw data obtained via the circuit shown inFigure 2 by plotting the instantaneous internal chargeq(t) versus the internalphosphor fieldfp(t), as obtained from equations Equations 3 and 7. These equa-tions arise from electrostatic equations used to describe the electrical behaviorof an ideal ACTFEL device that possesses no space charge generation (5, 17).If an ACTFEL device possesses space charge, these equations are still valid,but it should be recognized thatfp(t) is the average field across the phosphor



238 WAGER & KEIR

Figure 7 An idealized Q-Fp curve for an ACTFEL device.

and thatq(t) is equal to the actual internal charge flowing from one interface toanother plus a contribution that depends upon both the amount and location ofthe space charge generated in the phosphor (17).

An idealized Q-Fp curve for an ACTFEL device is shown in Figure 7. Notethat the same A-J labeling scheme used in Figure 5 to describe a Q-V curve isalso employed for a Q-Fp curve. Additionally, note that a Q-Fp loop is traversedin a clockwise manner, in contrast to a Q-V curve, which is traversed in acounter-clockwise direction. As indicated in Figures 5 and 7, the same kindsof charges (i.e. Qcond, Qpol, Qleak, Qrelax, and Qmax) are obtainable from a Q-Vor Q-Fp curve; however, since a Q-Fp curve involves only internal charge, themagnitudes of these various kinds of charge may be directly compared in aQ-Fp curve. Other device physics parameters available from a Q-Fp curve, butnot from a Q-V curve, are the steady state fields Fss, which are the constant ornearly constant fields occurring during the rising portion of the applied voltagewaveform above turn-on. Also, the polarization charges and fields are notlabeled in Figure 7 but correspond to the charges and fields at points A and Fin the Q-Fp curve.

One aspect of Q-Fp analysis we find most useful is the ability to evaluate theextent of field-clamping of a ACTFEL device. This may be readily accomplished




by plotting Q-Fp curves for three voltages above threshold (e.g. Vmax= 20, 40,and 60 V above threshold) and determining whether the steady state fields areidentical and independent of Vmax; if they are, the ACTFEL device is said toexhibit strong field-clamping.

Perhaps the most subtle aspect of Q-Fp analysis is the fact that the shape andaccuracy of a Q-Fp curve depend directly upon how well established are theinsulator and phosphor capacitances and the phosphor thickness, as is directlyevident from Equations 3 and 7. Any uncertainty in an estimate of the phosphorthickness is simply reflected as an uncertainty in the scaling of the x-axis ofthe Q-Fp curve. However, inaccuracies in the estimates of Ci and Cp lead todistortions in the shape of a Q-Fp curve (24). We have found that plottingQ-Fp curves is an excellent way to monitor the accuracy of estimates of Ci andCp. Typically, we prefer to employ Cphys

i and Cphysp for Q-Fp analysis. However,

sometimes it is advantageous when attempting to analyze ACTFEL devices thatpossess dynamic space charge generation to treat the Ci and Cp parameters in theQ-Fp equations as freely adjustable parameters and to adjust these parametersuntil the most ideal Q-Fp curve is obtained (ideal is defined as a Q-Fp curve thathas the most vertical and horizontal characteristics over the respective BC-GHand DE-IJ portions of the Q-Fp curve) (18, 24–26). We refer to these adjustedparameters as Cqfp

i and Cqfpp ; we believe that the deviation of these parameters

compared with Cphysi and Cphys

p is a measure of the amount and location of thedynamic space charge generated (18).

Maximum Charge–Maximum Applied Voltage AnalysisOur maximum charge-maximum applied voltage (Qmax-Vmax) measurement issimilar to that of a conventional transferred charge measurement (in the litera-ture, this measurement is sometimes referred to as a1Q-V measurement) (4).A 1Q-V measurement is accomplished by plotting Qcond versus Vmax. In thefollowing, we describe our procedure for accomplishing Qmax-Vmaxand externalmaximum charge-maximum applied voltage, Qe

max-Vmax, measurements.We define Qmax as indicated in Figure 7. Qmax corresponds to the net charge

located at the phosphor/insulator interface with respect to the flat band or neutralcharge level (i.e. the interface charge when the interface is electrically neutral).In order to accomplish a Qmax-Vmax or Qe

max-Vmax measurement, sets of offset-adjustedqext(t), q(t), andvEL(t) curves as a function of Vmax are obtained forboth positive and negative voltage pulses. To obtain a Qmax-Vmax curve, asorting routine is invoked to find Qmaxfrom eachq(t) curve and Vmaxfrom eachvEL(t) curve. Finally, Qmax is plotted as a function of Vmax to obtain the desiredQmax-Vmaxcurve. A Qe

max-Vmax curve is obtained in an identical fashion exceptthat a sorting routine is invoked to find Qe

max from eachqext(t) curve.



240 WAGER & KEIR

The intercept and slope are the primary device physics parameters of interestin a Qmax-Vmax curve. The intercept is equal to the threshold voltage. Thisthreshold usually correlates well with the threshold assessed from a luminance-voltage (L-V) curve. Thus the Qmax-Vmax method is an electrical means fordeterming the threshold of an electro-optic ACTFEL device.

The slope of a Qmax-Vmaxcurve is a measure of the charge transfer efficiencyof an ACTFEL device. The slope of a Qmax-Vmax curve may be greater than,less than, or equal to Cphys

i (15, 27). If the Qmax-Vmaxslope is greater than Cphysi ,

then some type of electron multiplication occurs in which more charge arrivesat the anode interface than is emitted from the cathode interface (27). Suchelectron multiplication processes consist of either bulk phosphor trap ionization(involving the creation of bulk dynamic space charge) or band-to-band impactionization (which does not involve bulk dynamic space charge generation).Both processes result in electron multiplication, but only trap ionization occursconcomitantly with C-V and Q-Fp overshoot or other manifestations of dynamicspace charge generation. Thus if an ACTFEL device has a Qmax-Vmax slopegreater than Cphys

i and exhibits no C-V or Q-Fp overshoot or other evidence fordynamic space charge generation, band-to-band impact ionization likely occursin the phosphor of this device. To date, evaporated ZnS:Mn ACTFEL devicesare the only devices we have tested that exhibit band-to-band impact ionizationbut no dynamic space charge generation.

If the Qmax-Vmax slope is less than that of Cphysi , this indicates that the ACT-

FEL device has insufficient charge transfer for field-clamping or pseudo-field-clamping. We have observed this kind of behavior in thiogallate:Ce ACTFELdevices, which also do not exhibit any evidence for dynamic space chargegeneration (16). If the Qmax-Vmax slope is equal to Cphys

i , this would implythat charge is efficiently transferred from the cathode interface to the anodeinterface but that no electron multiplication occurs during this transfer (again,we assume an absence of space charge generation). To date we have neverobserved an ACTFEL device with these characteristics. Finally, note thatsome ACTFEL devices do not exhibit a single, well-defined Qmax-Vmax slope(4b). For such devices, it is convenient to evaluate the slope of the Qmax-Vmaxcurve as a function of Vmax and to compare the magnitude of this slope toCphys

i .

ELECTRICAL CHARACTERIZATION EXAMPLES

Evaporated ZnS:MnThe electrical characteristics of an evaporated ZnS:Mn ACTFEL device witha phosphor thickness of∼600 nm are illustrated in Figures 8–11. Figure 8 is




Figure 8 C-V curves for an evaporated ZnS:Mn ACTFEL device for positive applied voltagepulses. Vmax= 20, 40, and 60 V above threshold. The arrow indicates that these curves shift tolower voltages as Vmax increases.

Figure 9 Q-V curves for an evaporated ZnS:Mn ACTFEL device. Vmax= 20, 40, and 60 V abovethreshold.



242 WAGER & KEIR

Figure 10 Q-Fp curves for an evaporated ZnS:Mn ACTFEL device. Vmax= 20, 40, and 60 Vabove threshold.

Figure 11 Qmax-Vmax curve for an evaporated ZnS:Mn ACTFEL device. The frequency of theapplied voltage waveform is 100 Hz. The dashed line is an extrapolated line that originates at thethreshold and whose slope corresponds to the physical capacitance.




a set of C-V curves obtained using positive applied voltage pulses. Similarnegative applied voltage pulse C-V curves are not shown; this similarity isone piece of evidence that the electrical characteristics of this ACTFEL deviceare very symmetrical with respect to the applied voltage polarity. The C-Vcurves indicated in Figure 8 are nearly ideal because Ccv

t and Ccvi are found to be

very similar to Cphyst and Cphys

i , and the C-V curves shift almost rigidly to lowervoltages with increasing Vmax.

Figures 9 and 10 present sets of Q-V and Q-Fp curves, respectively, forVmax= 20, 40, and 60 V above threshold. These curves are distinguished bystrong field-clamping, as evident from the fact that Fssis virtually independentof Vmax in the Q-Fp curves and that the above turn-on Q-V curves retraceeach other above threshold. Additionally, these sets of Q-V and Q-Fp curvesdemonstrate that Qcond, Qrelax, Qleak, and Qpol all increase with increasing Vmax,whereas Vto decreases with increasing Vmax.

Figure 11 is a plot of the Qmax-Vmax curve. This curve is taken at a fre-quency of 100 Hz in order to minimize effects associated with metastablehole trapping (4b). A least square fit to the linear portion of this curve justabove threshold results in an estimated threshold voltage of 122.8 V. This Qmax-Vmax intercept is in good agreement with the L-V threshold. The slope of thisQmax-Vmax is∼38 nF/cm2 for Vmax= 120–150 V and∼31 nF/cm2 for Vmax>

150 V. Thus just above threshold the Qmax-Vmax slope is significantly largerthan the physical insulator capacitance value of 29 nF/cm2, whereas at Vmax>150 V, the Qmax-Vmax slope is only slightly larger than that of the physi-cal insulator capacitance. The fact that the Qmax-Vmax slope is greater thanthe physical insulator capacitance, and yet there is no C-V or Q-Fp evidencefor dynamic space charge generation, indicates that electron multiplication viaband-to-band impact ionization occurs within the phosphor of this device (19).Moreover, the fact that the slope of the Qmax-Vmax curve changes as a functionof Vmax is evidence for the important role played by metastable hole trapping(4b).

Atomic Layer Epitaxy SrS:CeThe electrical characteristics of an atomic layer epitaxy (ALE) SrS:Ce ACTFELdevice with a phosphor thickness of 500 nm are collected in Figures 12–15.Figure 12 displays a set of C-V curves obtained using both positive and negativeapplied voltage pulses. There are various aspects of these curves that make themnon-ideal. First, note that the C-V curves depend strongly on the polarity of theapplied voltage pulse; this indicates that there is a large degree of asymmetryin the electrical properties of the upper and lower interfaces of this device.Next, notice that the above turn-on capacitance is significantly greater than thephysical insulator capacitance for both sets of C-V curves; this indicates that



244 WAGER & KEIR

Figure 12 C-V curves for an ALE SrS:Ce ACTFEL device for (a) positive voltage pulses and(b) negative voltage pulses. Vmax= 20, 40, and 60 V above threshold.

there is a large amount of dynamic space charge generation in the phosphorof this ACTFEL device; we attribute this dynamic space charge generation toCe-to-band and trap-to-band impact ionization. Moreover, note that there is alarge amount of overshoot in the positive voltage pulse curves shown in Figure12a, whereas the negative voltage pulse curves shown in Figure 12b saturate atcapacitances in excess of the physical insulator capacitance and do not display




Figure 13 Q-V curves for an ALE SrS:Ce ACTFEL device. Vmax= 20, 40, and 60 V abovethreshold.

Figure 14 Q-Fp curves for an ALE SrS:Ce ACTFEL device. Vmax= 20, 40, and 60 V abovethreshold.



246 WAGER & KEIR

Figure 15 Qmax-Vmax curve for an ALE SrS:Ce ACTFEL device. The frequency of the appliedvoltage waveform is 100 Hz. The dashed line is an extrapolated line that originates at the thresholdand whose slope corresponds to the physical capacitance.

typical overshoot characteristics in which the capacitance first increases andthen decreases. Finally, notice that these C-V curves first shift to lower voltageswith increasing Vmax but then shift to higher voltages. Also, these C-V shiftsare non-rigid. These observations indicate that a substantial fraction of thetransferred charge arises from the bulk phosphor via dynamic space chargegeneration and not exclusively from interface state injection.

Figures 13 and 14 present sets of Q-V and Q-Fp curves, respectively, forVmax= 20, 40, and 60 V above threshold. These figures offer more evidencethat this ALE SrS:Ce ACTFEL device behaves in a non-ideal manner. Theabove turn-on curvature of the Q-V curve and the Q-Fp overshoot are othermanifestations of dynamic space charge generation; Figures 13 and 14 confirmthat dynamic space charge generation is asymmetrical with respect to the voltagepulse polarity, in agreement with Figure 12. The Q-Fp curves show that field-clamping is not operative in this device. Moreover, the most striking aspect ofthese Q-Fp curves is the way in which the phosphor field decreases stronglyabove turn-on due to the formation of dynamic space charge. Another unusualaspect of Figure 14 is the fact that the magnitude of the charge decreases inthe DE and IJ regimes of the Q-Fp curve even though the applied voltage is notyet reduced to zero during this portion of the waveform. We refer to this kind




of phenomenon as charge collapse and attribute it to the presence of a largeinternal field that is set up by the transfer of charge and/or the storage of thischarge in very shallow states. When charge collapse occurs, the polarizationcharge begins to leak to the opposite interface even before the applied voltageis completely reduced to zero because the internal field is so large and/or thetrapped charge states are so shallow.

Figure 15 is a plot of the Qmax-Vmaxcurve. The Qmax-Vmax threshold voltageis found to be 77.8 V, which is in relatively good agreement with the thresholdfound from L-V measurements. Note that in contrast to the Qmax-Vmax curveshown in Figure 11, the slope of the Qmax-Vmaxcurve is constant above threshold.Additionally, the Qmax-Vmax slope shown in Figure 15 is∼73 nF/cm2, whichis significantly greater than the physical insulator capacitance of 39.8 nF/cm2.Differences in the slopes of the Qmax-Vmax curves shown in Figures 11 and 15are attributed to the different physical mechanisms responsible for Qmax-Vmaxslopes in excess of the physical insulator capacitance (i.e. band-to-band impactionization and metastable hole capture for the evaporated ZnS:Mn ACTFELdevice of Figure 11 and Ce-to-band and trap-to-band impact ionization for theALE SrS:Ce ACTFEL device of Figure 15).

ACKNOWLEDGMENTS

We wish to thank Sey-Shing Sun and Dick Tuenge of Planar America forproviding the samples used here. This work was supported by the US ArmyResearch Office under Contract No. DAAH04-94-G-0324, and by the DefenseAdvanced Research Projects Agency under the Phosphor Technology Centerof Excellence, Grant No. MDA 972-93-1-0030.

Visit the Annual Reviews home pageathttp://www.annurev.org.

Literature Cited

1. Ono YA. 1995.Electroluminescent Dis-plays.Singapore: World Scientific

2. Chen YS, Krupka DC. 1972.J. Appl. Phys.43:4089–96

3. Smith DH. 1981.J. Lumin.23:209–354. Ono YA, Kawakami H, Fuyama M, Oni-

sawa K. 1987.Jpn. J. Appl. Phys.26:1482–92

4a. Shih S, Keir PD, Hitt J, Wager JF. 1996.Appl. Phys. Lett.69:1921–23

4b. Myers R, Wager JF. 1997.J. Appl. Phys.81:506–10

5. Bringuier E. 1989.J. Appl. Phys.66:1314–25

6. Abu-Dayah A, Kobayashi S, Wager JF.1993.Appl. Phys. Lett.62:744–46

7. Neyts KA, De Visschere P. 1992.Solid-State Elect.35:933–36

8. Douglas AA, Wager JF, Morton DC, KohJ, Hogh CP. 1993.J. Appl. Phys.73:296–99

9. Morton DC, Koh J, Hogh CP, KhormaeiRI. 1993.Appl. Phys. Lett.63:863–65

10. Singh VP, Krishna S, Morton DC. 1991.J.Appl. Phys.70:1811–19

11. Douglas AA, Wager JF. 1992.SID 91 Dig.pp. 356–59

12. McArthur RC, Davidson JD, Wager JF,



248 WAGER & KEIR

Khormaei RI, King CN. 1990.Appl. Phys.Lett.56:1889–91

13. Davidson JD, Wager JF, Khormaei RI,King CN, Williams D. 1992.IEEE Trans.Electron DevicesED-39:1122–28

14. Cardona M. 1969.Modulation Spec-troscopy.New York: Academic

15. Pham LV. 1995.Electrical characteriza-tion, maximum charge-maximum voltage(Qmax-Vmax) analysis, aging and temper-ature studies of thiogallate thin-film elec-troluminescent devices. MS thesis. OregonState Univ., Corvallis. 96 pp.

16. Pham LV, Wager JF, Sun SS, Dickey E,Tuenge RT, King CN. 1994.SPIE Proc.2174:190–99

17. Keir PD, Ang WM, Wager JF. 1995.J.Appl. Phys.78:4668–80

18. Shih S, Keir PD, Wager JF, Viljanen J.1995.J. Appl. Phys.78:5775–81

19. Ang WM, Pennathur S, Pham L, Wager JF,Goodnick SM, Douglas AA. 1995.J. Appl.Phys.77:2719–24

20. Keir PD. 1995.Modeling phosphor spacecharge in alternating-current thin-filmelectroluminescent devices.MS thesis.Oregon State Univ., Corvallis, OR. 82 pp.

21. Wager JF, Douglas AA, Morton DC. 1992.In Electroluminescence, ed. VP Singh, JCMcClure, pp. 92–101. El Paso, TX: CincoPuntos

22. Sze SM. 1981.Physics of SemiconductorDevices.New York: Wiley

23. Abu-Dayah A, Kobayashi S, Wager JF.1993.J. Appl. Phys.74:5575–81

24. Abu-Dayah A. 1993.Internal charge-phosphor field analysis, electrical charac-terization, and aging studies of AC thin-film electroluminescent devices.MS thesis.Oregon State Univ., Corvallis. 72 pp.

25. Abu-Dayah A, Wager JF. 1994.J. Appl.Phys.75:3593–98

26. Thuemler RL, Wager JF. 1995.SID 95 Dig.pp. 473–75

27. Bringuier E. 1990.J. Appl. Phys.67:7040–44

Documents

ELECTRICAL CHARACTERIZATION OF THIN-FILM