11762
J. Phys. Chem. 1993,97, 11762-1 1768
Electroluminescence of Tb3+-DopedAnodic TazOs Films Eric A. Meulenkamp' and John J. Kelly Condensed Matter-Debye Institute, Utrecht University, P.O.Box 80.000, 3508 TA Utrecht, The Netherlands Received: April 6, 1993; In Final Form: September 9, 1993.
The electroluminescence (EL) of Tb3+-dopd anodic Ta2Os films in the presence of H202 was studied. Since the EL of the dopant Tb3+ion shows the same potential (U) dependence as the background EL, light emission from Tb3+can be used to monitor electrochemical processes at the oxide/solution interface. In the rising part of the (EL,U) curve, both the hole injection rate and the radiative recombination efficiency increase. This gives rise to a quadratic dependence of the intensity on the current. The EL is not influenced by the rotation rate; ring-disk experiments show the hole injection current to be kinetically limited. At more negative potential, quenching of the emission occurs. Comparison of the photoluminescence (PL) and EL intensities and spectra, which depend on the surroundings of the Tb3+ ion, indicates that the quenching must be due to incorporation of hydrogen from the oxide/solution interface. The anodic dark current resulting from cathodic polarization can be used as a measure of the concentration of the H-induced recombination centers.
1. Introduction The n-type semiconductor electrodes can show electroluminescence (EL) in the presence of a strong oxidizing agent. These agentsare reduced by means of valence band electrons;this process can also be described as "hole injection". The holes recombine with the majority carriers, Le., the conduction band electrons. If recombinationoccurs radiatively, light emission is observed. EL of many materials, including compound semiconductors and oxides, has been reported.' The intensity of the EL depends on three parameters: the radiative quantum efficiency of the luminescent center, the concentration of injected holes, and the concentrationof other nonradiativerecombination centers. These quantities are related to electrochemicalprocesses and thus vary as a function of the electrode potential. EL can respond to the chemical nature of the interface in two ways. In the case of small-bandgap materials, such as n-type 111-V semiconductors, the competitionfor holes between surface recombination, which has been shown to be nonradiative,2and bulk recombination determines the EL intensity. For example, thelargevariations of the EL intensity in then-GaAs/Fe(CN)& system were ascribed to the presence of a hydride-covered or a radical-like surface, which exhibit a low or a high surface recombination rate, re~pectively.~In the case of large-bandgap materials, a second type of surface sensitivity is sometimes found. In particular at Ti02, radiative recombination via surface states has been described.& The emission can, in principle, be used to determine the energy and the concentration of the surface states. A few years ago, Haapakka et al. described characteristic EL of Tb3+-doped Tal05 films.' In previous work we showed that this EL also originates partly from the surface.8 The Ta205 film contained approximately0.1% Tb3+ and had a thickness ofabout 23 nm.8 The bandgap was 4.3 eV. The current-potential characteristics resemble those of an n-type semiconductor.8 On the basis of a comparison between photoluminescence (PL) and EL of Eu3+-and Tb3+-dopedlayers, we concluded that the Tb3+ ion acts as a recombination center for free electrons and holes.8 In this paper, we want toconsiderto what extent thecharacteristic EL of the Tb3+ ion can be used as a probe to monitor electrochemical processes such as hydrogen in-diffusion under cathodic bias and H202 reduction. The use of Tb3++-doped layers provides extra information because of the spectral features of this rare earth ion and because of possible comparison with ~
*Abstract published in Advance ACS Abstracts, October 15, 1993.
0022-3654/93/2097-11762$04.00/0
undoped layers. The emphasis is on the potential dependence of the EL intensity and the EL spectrum. 2. ExperimentalSection
Tantalum rod of 99.99%purity was purchased from Highways International (Baarn, The Netherlands) and cut into disks. The anodizationprocedureconsistedof a symmetricdouble-step pulse potential (8/0/-8/0V, 104 Hz) followed by a dc polarization at 10 V. This procedure was chosen because it offered a high and reproducible emission intensity.'~~A detailed description of the preparation of the layer as well as of the pretreatment and the anodization setup has been given elsewhere.* Measurements were performed in a three-electrodeconfiguration using rotating disk electrodes (4.0-mm diameter), a large Pt counter electrode, and a saturated calomel electrode (SCE) as reference. All potentials are given with respect to the SCE. Ring-disk electrodes consisted of a tantalum disk (3.8-mm diameter) and a platinum black ring (inner and outer diameter 4.6 and 5.8 mm, respectively). The theoretical collection efficiencywas 35%? Diffusion-limited H202 reduction and oxidation waves were observed at the ring. Either a home-built potentiostat or a Heka (Lambrecht, Germany) bipotentiostat was used. EL was measured with rotating electrodes (unless otherwise stated) on Perkin-Elmer MPF44B and SPEX DM3000F Fluorolog spectrofluorometers. PL experimentswere performed with stationary electrodesin the SPEX Fluorolog. In EL experiments data were sampled by computer using a MetraByte DASl6G1 analog-to-digital converter. All chemicals were of analytical grade. Doubly-distilled deionized water was used. All solutionswere purged with nitrogen. The experiments were carried out at room temperature in a solution containing 1.O M NaC104 as the supporting electrolyte and H202 as the oxidizing agent. The scan rate was 50 mV/s.
3. Results
3.1. Current-Potentia1 Characteristics. In Figure 1 the (iJ) characteristics in indifferent electrolyte are shown for three pH values. Doped layers, Le., containing Tb3+,and undoped layers gave identical results. The current, which is due to H2 evolution, increases rather slowly as a function of potential. This is often found for relatively thick, large-bandgap anodic oxide layers. lo This slow increase is not related to Ohmic drop: the series resistance, as determined from impedance measurements, was about 30 0. The current onset shifts in the negative direction as the pH is increased. This is commonly found for semiconducting CP 1993 American Chemical Society
Electroluminescence of Tb3+-DopedAnodic Ta2O5 Films
The Journal of Physical Chemistry, Vol. 97, No. 45, 1993 11763 60
2-
h
P
40
z” .g
E,
4
\
E
v
20
.e
.e
? I
-2.5
-2.0
-1.5
-1.0
’
0
-5 50
0
u (VI Figure 1. Current-potential curves measured in 1.0 M NaClO4 at the
b
indicated pH values. The rotation rate is 3000 rpm. materials because the flatband potential shifts with pH. The onset potential varied from electrode to electrode. For example, the entire (i,U) curve at pH = 3.1 in Figure 1 could be displaced by 0.2 V in either direction. However, the shape and the relative positions of the curves of Figure 1 remained unaltered. For pH values between 2.5 and 4.0 considerable hysteresis is observed. A current plateau due to H+ reduction, which is often diffusion-limited, is found during the forward scan from positive tonegative potential. Themagnitudeoftheplateaucurrent agrees well with that calculated using the Levich e q ~ a t i o n .At ~ pH = 5.0, the plateau current can also be observed when acetic acid is added to the solution. The current depends almost linearly on the concentration of acetic acid. The acid acts as a proton source because of the acid-base equilibrium prior to the electron-transfer step. During the reverse scan, H+ reduction no longer seems to occur, as the (i,U) curve isdimilar to that at pH = 9.0. H+ reduction at intermediate pH values has been observed before at a large-bandgap semiconductor, viz. ZnS.ll The (i,U) curve does not change when H202 up to a concentration of 2 X le2 M is added to an indifferent electrolyte solution. That H202 is being reduced during cathodic polarization is, however, clear from ring-disk measurements. The reduction has been shown to take place in two steps.12 First, an electron is transferred from the conduction band to H202, creating a hydroxyl ion and a hydroxyl radical. This radical then injects a hole into the valence band. The H202 consumptionat the Ta205 disk was calculated by measuring the decrease in the H202 oxidation current at a platinum black ring (shielding mode9). In Figure 2a the decrease in the ring current, corrected for the collection efficiencyNO, is shown as a function of the disk current during a forward scan for a solution of pH 3.1. The dashed line = i D ) shows the result expected if the entire disk current were due to H202 reduction. Clearly, H2 evolution and H202 reduction occur simultaneously. The ratio of the H202 reduction current to the total current is a constant 35% up to about 100-pA disk current. Figure 2b shows the Hz02 reduction current at the disk during a forward scan as a function of the rotation rate. At -1.9 V, which is in the rising part of the (i,v)curve in this case (cf. Figure l), H202 reduction is kinetically limited. At a more cathodic potential a significant influence of the rotation rate is found. However, the current is much smaller than the expected diffusionlimited current of 330 pA at loo0 rpm. It proved impossible to measure the H202 current at very negative disk potential because the large amount of H2 produced at the disk led to some H2 oxidation at the ring. At potentials positive with respect to -1.0 V,a small anodic current is found in the reverse scan with a maximum at about 0.0 V. This is not expected for an n-type semiconductor in the dark. At the same time, an anodic sub-bandgap photocurrent can be observed for wavelengthsup to 700 nm. The anodic dark current and the photocurrent slowly decay to zero at anodic potential, implying that they are related to the oxidation of some reduced species formed at negative bias. Both dark current and the
100
200
150
/-2 . 60
-
-2.3 v -1.9 v
Y
40
r-
z“ \
:.
20
M
?
0
01/2
(rpm112)
Figure 2. Rotating ring-disk experiments in 1.0 M NaClO4 (pH = 3.1) + 2 X lC3M HzOz. (a) Change in the ring current Ai- due to H202 oxidation, corrected for the collection efficiency NO,as a function of the disk current during a forward potential scan. The ring potential is 1.0 V. The rotation rate is 2000 rpm. The dashed line represents Ai*/No expected if the total disk current were due to H202 reduction. (b) Change in the corrected ring current, A i ~ / N oas, a function of the rotation rate w (in revolutionsper minute) at two potentials during a forward potential
scan.
1
0
0
20
40
60
80
100
cathodic pulse time (s) Figure 3. EL intensity at 545 nm during a cathodic pulse (lJ= -2.3 V) and anodic dark charge (V = 2.0 V) after a cathodic pulse as a function of the cathodic pulse time. The meaning of A Q is explained in the text. Experimental conditions as in Figure 2, w = lo00 rpm.
photocurrent increase as the charge passed during the preceding cathodic polarization increases. This is clearly seen in a potential step experiment. First, the potential is stepped to a negative value; subsequently, a positive potential is applied, and the resulting dark current is measured. In Figure 3 the anodic charge measured in the dark is shown as a function of the preceding cathodic pulse time. The EL intensity, which is also indicated, is discussed below. The charge has been corrected by taking the anodic charge after a 10-ms cathodic pulse to be only due to double-layer charging. Since the cathodic current of about 2.5 mA cm-2 is almost constant in the time interval shown, the x axis can also be read as a cathodic charge. 3.2. Characteristic TV+ E h i o n vs Background Emidon. Undoped TazO$films show a very weak broad EL emission band in the presence of H202. The emission ranges from about 330 to 850 nm. This background emission is due to recombination
11764 The Journal of Physical Chemistry, Vol. 97, No. 45, 1993
-2.5
-2.0
-1.5
-1.0
u (VI Figure 4. Comparison bctwecn the (EL,LI) curves of a Tb3+-doptd electrodeat A = 623 nm (Tb3+emission)and at h = 420 nm (background emission). Experimental conditions as in Figure 3.
Meulenkamp and Kelly
0.0
0.5 i2
1 .o
1.5
(mA2cm-4)
Figwe 5. EL intensity at 545 nm as a function of the square of the cathodic current during a forward scan. Experimental conditions as in Figure 3.
intensitiesstarts at about 60%of the maximumintensity. During of electrons with holes injected by OH radicahs This emission the reverse scan a similar relation was often observed, although can also be discerned at Tb3+-dopedelectrodes at wavelengths in a much smaller current range. The slope of the line increased shorter than 460 nm. At longer wavelengths Tb3+ emission as the scan rate was lowered: the EL intensity at a given potential dominates the spectrum. Tb3+acts as a recombination center. The characteristic emission due to transition from the 5D4to 7 F ~ increased while the current remained rather constant. The background emission showed the same EL a: i2 relation. levels within the 4fa shell is easily recognized. The main emission lines are at 490, 545,585, and 623 nm. The 5D4-7F5 transition 3.4. Compnrisoa of EL and PL. The strong quenching of the EL at negative potential is evident in Figure 4. A decrease was at 545 nm is the strongest.s also observed in photoluminescence (PL) experiments. The In Figure 4 the EL intensity-potential (EL,U) curve is shown. potential dependence of the PL and EL at 545 nm is compared The emission has been normalized with respect to the maximum in Figure 6a for a stationary electrode. PL was excited using intensity. The potential dependence is rather complex and shows supra-bandgap radiation of 260 nm. The electron-hole pairs marked hysteresis. Typically, a slow rise starting at about -1.5 created recombine on the Tb3+ ion and give rise to the V is observed during a forward scan (from positive to negative characteristic luminescence from the 5D4 excited state? potential). At around -2.2 V the maximum intensity is reached, The main features of the (EL,U) curve in Figure 6a resemble and further scanning in the negative direction gives rise to strong those of Figure 4. The differences reflect, in part, the spread in quenching. The degree of quenching depends on the reversal potential (-2.75 V in Figure 4), which determines the cathodic the (EL,U) characteristics from electrode to electrode. However, the different H202 concentration and rotation rate also have a charge. Figure 3 clearly illustrates that the EL intensity during a cathodic pulse decreases as the anodic charge in the following marked effect on the (EL,U) curve. Despite the fact that the anodic pulse increases. During the reverse scan the intensity shape of the (EL,U) curve differed somewhat from sample to often shows a maximum. In the following sections the potential sample, the effects of rotation rate, HzO2 concentration, and dependence of the emission is discussed in more detail. emission wavelength were similarfor all samples. The PL intensity shows a slight increase during the scan from +2.0 to about -2.2 The (EL,U) curve shows the same pH shift as the (i,U)curve. The highest intensity is obtained for pH values between 3 and 5. VIat which potential a sudden decrease is observed. The relative Outside this optimum range, the EL intensity rapidly decreases decrease is 15-2096 at the reversal potential. During the reverse to about 10%of the maximum value at pH = 1 and 10. The pH scan the intensity rises slowly to the original value. The (EL,U) curve shows an intensity drop at the same potential as the (PL,U) effect is probablyrelated to the pH-dependent ratio of H2 evolution curve. However, the relative decrease of about 75% at -3.0 V and the H202 reduction rate (see Figure 1). A similar effect has been described in the case of S20s2- induced EL of ZnS.11J3The is much larger than in the case of PL. surface charge of the oxide (the point of zero charge of Ta2O5 Tb3+emission can also be obtained by direct excitation of the is about 2.914), which influences the H202 adsorption rate, may 5D3 level using 378-nm r a d i a t i ~ n . ' In ~ Figure 6b the potential also play a role. dependenceof the Tb3+PL, excited at 260 and 378 nm, is shown. To establish the role of the Tb3+EL as a probe for processes The intensity has been normalized with respect to the value at occurring at the oxide/electrolyte interface, we investigated positive potential. The emission intensity with b,,, = 378 nm is whether the potential dependence of the characteristic 5D4 about 200 times weaker than with b,,, = 260 nm, owing to the emission mirrors that of the background emission. Clearly, the much smaller oscillator strength of the intraconfigurational (EL,I/) curves of Figure 4, which were measured with the same t r a n s i t i ~ n .The ~ ~ PL intensity is almost independent of potential Tb3+-dopedelectrode, are very similar. It can be concluded that during the forward scan up to about -2.0 VI where quenching the processes responsible for the potential dependence of the starts. The relative decrease of the PL excited at 378 nm is much background emission also affect the Tb3+emission in the same smaller than that excited at 260 nm. The differences between way. The Tb3+emission and the background emission show the the (PL,U) curves of Figure 6a,b reflect the spread in theelectrode same dependence on pH, H202 concentration, and rotation rate. characteristics. Since the Tb3+emission is about 500 times stronger than the In the potential range in which quenching occurs the EL background emission, Tb3+-doped layers were mainly used to emissionspectrum changes. In Figure 7 the potential dependence assess the relationship between EL intensity and electrochemical of the relative intensity of the various Tb3+emission lines with processes. respect to the 5454111 emission is shown. The intensity of this 3.3. Current-EL Relation. The EL intensity rises more steeply emission line is, to a good approximation, independent of the with potential than the current. Actually, in a certain potential surroundings of the luminescent ion (see Discussion). It can be range EL a z2. Figure 5 shows the EL intensity of Tb3+-doped used as an internal reference intensity. The scatter at relatively Ta205 at 545 nm as a function of the square of the cathodic positive potentialisdueto thelow intensity. Therelativeintensities current during a forward scan. The x axis corresponds to a at 490 and 585 nmdecrease by 10-2096 at potentialsmorenegative potential range from -1.5 to -2.2 V. The deviation at higher than -2.2 V, where quenching of the emission starts. The relative
Electroluminescence of Tb3+-DopedAnodic TazOs Films
a
The Journal of Physical Chemistry, Vol. 97, No.45, 1993 11765
I'
2.5
?
-
2.0
?
Q
1
x 1 -
1.5
+ i
.CI U
2
If! c
1.0
.d
0.5 0.0 -2.0
-3.0
-2.5
-1.0
-2.0
-1.5
u (VI b 1.05
- 1
b' 4.0
n
j
4 0.95
U
-+-
260
-*-
378
n
h . YI
v1
3 0.85
j 3.0
cd v h
Y
c
. U I
. I
Y
2.0
c
0.75
. I
-3.0
-1.0
-2.0
1.o
u (VI
F i p c 6 . Normalid (EL,U) and (PL,U) curvesof a stationary electrode in 1.0 M NaClO4 (pH = 3.1). The emission was measured at 545 nm. (a) Comparison of the quenching of PL (excitation at 260 nm) and EL in the presence of 2 X lo-' M H~02.The dashed line indicates the onset of quenching. Note the y-axis break and scales. (b) Comparisonof the quenching of PL excited at 260 and 378 nm.
-2.0
-1.5
-1.0
u (VI +
curvesin l.OMNaC104(pH= 3.1)+2X 1~MHzO~atthreerotation rates: (1) 200, (2) 900, and (3) 4000 rpm.
0.40
G;0.20 d 3 U
0.15
.
.*j .*-
623 0.10 -3.0
-2.5
Figure 8. Influence of the rotation rate on the (EL,U) curves at 545 nm. (a) (EL,U) curves in 1.0 M NaC104 (pH = 9.0) 2 X lW3M H202 at three rotation rates: (1) 250, (2) 1000, and (3) 4000 rpm. (b) (EL,U)
0.50
U
0.0
. *
-2.5
. -2.0
, .-:*.:. .I , e ' & . $ .. . >e:
-1.5
u (VI Figure 7. Comparison of the (EL,U) curves at various wavelengths.The intensity has been taken relative to the emission intensity at 545 nm. Experimental conditions as in Figure 3. For the sake of clarity only the forward scan is shown. Note the y-axis break and scales.
intensity at 623 nm is essentially independent of potential. The spectral change shown in Figure 7 cannot be explained by absorption by Ta205 or some other property of the oxide itself for two reasons. First, absorption would be strongest at shorter wavelengths;this would imply a continuousincrease of the amount of quenching of the relative intensity at negative potentials with decreasing wavelength, in contrast with the irregular wavelength dependence observed. Second, the 420420-nm intensity ratio of the background emission of an undopcd electrodeis independent of potential. Hence, the spectral properties of Tb3+ have to be taken into account. We were not able to measure reliably the PL
spectrum as a function of potential, owing to the presence of H2 bubbles and simultaneous weak EL due to the residual 0 2 concentration. 3.5. Effect of Rotation Rate and HZOZConcentration. The effect of the rotation rate and the H202 concentration on the (i,v) characteristics has already been described. Here we focus on the effect on the EL intensity. The EL intensity should be expected to show a linear dependence on the hole injection rate, provided that the concentration and rotation rate only affect the injection rate. However, the (EL,LI) curve depends on both in a highly nonlinear manner. The pH, the rotation rate, and the concentration are independent variables in the ranges from pH = 1.2 to 9.0, from 100 to 4000 rpm, and from 2 X 1k5to 2 X 1 k 2M HzOz. For example, the effect of the concentration on the (EL,U) curve is similar at pH = 1.2 or 9.0 and at 250 or 2000 rpm. Hence, the effects of these parameters can be considered separately. Two cases could be distinguished with respect to the influence of the rotation rate. Figure 8a shows the effect of the rotation rate (03 > w2 > w1) on the 545-nm emission when no current plateau due to H+ reduction was present. In this case the pH is 9.0. During the forward scan, a slight rotation rate dependence isobservedintherisingpartofthe(EL,U)curve. At morenegative potential, the three curves diverge and the largest intensity is obtained for the highest rotation rate. During the reverse scan
11766 The Journal of Physical Chemistry, Vol. 97, No. 45, 1993
-2.50
-2.25
-2.00
u
-1.75
-1.50
-1.25
(VI
Figure 9. Effect of the H202 concentration on the (EL,U) curve at 545 nm in 1.0 M NaC104 (pH = 3.1): (1) 2 X l P 5 , (2) 2 X l ( r , (3) 2 X 10-3, and (4) 2 X 1 P 2 M H202. The rotation rate is 2000 rpm.
the (EL,U) curve is again nearly independent of rotation rate at potentials more positive than -1.75 V. Figure 8b shows the effect of the rotation rate (w3 > w2 > w1) in the case of a diffusion-limited H+ reduction plateau current at pH = 3.1. Similar to the situation at a pH of 9.0,the (EL,U) curves coincide during the forward scan. A small divergence is again observed when the intensity is close to the maximum value. In therangefrom-1.8 to-2.3V, whichcorresponds to the potential range close to the current plateau, the highest EL intensity is found for the lowest rotation rate. During the reverse scan, the shape of the (EL,U) curve at high rotation rate is markedly different from that at lower rotation rates. At relatively positive potential, a huge intensity increase is observed upon a 4-fold increase of the rotation rate. The influenceof the H202 concentration is also complex. Figure 9 shows the (EL,U) curves at four concentrations (4 > 3 > 2 > I), each differing by a factor of 10. To show more clearly the results at low concentration, a part of curve 4 is not included. Its maximum intensity is about 6 at -2.25 V. Obviously, the maximum intensity in the (EL,U) curve is not proportionalto the H202 concentration. The onset of the EL and the peak potential shift in the negative direction as the concentration increases. Consequently, the degree of quenching at the reversal potential varies from 100%at 2 X 1e5M to only 40% at 2 X 1t2 M H202. At the lower concentrations, (almost) no increase of the intensity is found during the reverse scan. At high concentration, on the other hand, the (EL,U) curve shows a maximum intensity of the same order of magnitude as during the forward scan. 4. IHscussion
4.1. Curreat-Potential Characteristics. According to Schultze et a1.,16 who studied the reduction of Fe3+,Ta205 layers show strongly inhibitedcathodiccharge transfer. They described films thicker than 15 nm as insulators. Indeed,the Hz evolution current (Figure 1) increases very slowly with potential. Schultze et al. proposethat the cathodic current is limited by theelectron-transfer reaction or by the transport of electrons within the oxide, presumablyvia a hopping mechanism. The latter limitation would explain the identical (i,U) curves in the absence and presence of H202. However, diffusion-limited H+ reduction seems to indicate that electron transport within the solid cannot be rate limiting. It is also possible that the H+ and H202 reduction processes are coupled. For example, at intermediate pH values, the OH- ions produced by reduction of H202 can react with H+, thereby decreasing the diffusion flux of H+ toward the electrode surface and, hence, the H+ reduction current. Kankare et al.17have also reported identical (i,v) curves for this system. A similar effect was described in the case of S20a2- reduction at Zn0.I8 Interestingly, the H+reduction shows a large hysteresis (see Figure l), which has also been reported before.17
Meulenkamp and Kelly Figure 2a,b shows that the reduction rate of H202 increases as the potential is scanned in the negative direction. Actually, a significant dependence on rotation rate is observed at -2.3 V, while the current only corresponds to about 20% of the diffusionlimited value. This can be due to a mixed diffusion-kinetic limitati~n.~ Another possible explanation involves the concept of a spatially inhomogeneous electrode surface. In a series of papers,I9Scheller et al. have reported on the effect of the rotation rate on the current at a model electrode provided with a pattern of active and inactive sites. If the dimensions of the active sites and the distance between neighboring sites are large compared to the Nemst diffusion layer thickness, the current is proportional to the square root of the rotation rate; the current is, however, much smaller than that calculated from the geometric area. In the present system, inspection by optical microscopy revealed inhomogeneous light emission at a stationary electrode. This may be related to a spatially inhomogeneous reduction rate of H202. The anodic dark current resulting from cathodic polarization (see Figure 3) is of the same order of magnitude as that reported by Stimming,20 who also observed a sub-bandgap photocurrent. During cathodic polarization, partial reduction of the metal ions in the oxide layer, e.g., Ta(V) to Ta(IV), and incorporation of hydrogen into the lattice may occur. It has been proposed that this proceeds via an electron-proton reduction mechanism, yielding HTa205.21 In the case of valve metal oxides, such as WO,, V205, and Nb2O5, the reversible takeup of hydrogen has been studied in detail.22 This phenomenon is responsible for the electrochromism. The color changeof the above-mentionedoxides as a function of hydrogen content clearly reflects the drastic influence of hydrogen on the optical properties. In the present case, the sub-bandgap photocurrent indicates that hydrogen induces states within the bandgap. These are likely to promote recombination. Therefore, the charge determined from the anodic dark current or from the photocurrent with s u b bandgap illumination can be used as a measure of the number of recombination centersintroduced during the preceding cathodic polarization. 4.2. Current-EL Relation. A quadratic dependence of the EL intensity on the cathodic current has been described for a number of electrode materials, such as ZnS,13 and GaP.2' A common feature of the (i,U) curves is that a diffusion-limited reduction wave of the oxidizing agent is not observed. Instead, the current increases with potential more or less exponentially. A mechanism often proposed to account for the EL 0: P relation involves the Williams and Prener model.25 This model describes recombination of electronsand holes via deep donor and acceptor centers. If the capture of charge carriers by the deep centers is rate-limiting, the recombination rate is proportional to the concentration of electrons and holes. The hole concentration is taken to be linearly dependent on the total cathodic current. A further assumption is that the cathodic current is proportional to the electron concentration. Thus, the recombination rate and the EL intensity vary as the square of the current. There are several arguments why the use of this scheme is rather questionable. First, in some of the electrode materials described (ZnS, Sic) EL takes place in the bulk of the semiconductor. Obviously, as the current is proportional to the surface electron concentration and not to the bulk electron concentration, the EL intensity should depend linearly on the current. Similarly, any mechanism which involves the product of the electron and hole concentrationseems implausible. Second, it is difficult to prove that the trapping of charge carriers by deep centers is the rate-limiting step in the emission process. Interestingly,the quadratic dependence of EL on i is also known in solid-state devices. In the case of nonideal junction characteristics, the relation EL a inholds, with 1 In I2.26 For low minority carrier injection levels n is 2, while for a high injection level n is 1. The nonideality can be due to tunnel currents or to the presence of a large density of defects in the space charge
Electroluminescenceof Tb3+-DopedAnodic Taz05 Films
The Journal of Physical Chemistry, Vol. 97, No, 45, 1993 11767
region. In a p n junction LED, for example, i a exp(eU/ZkT) if the defects have energylevels close to the center of the bandgap. The EL intensity is simply proportional toexp(eU/kT),26 yielding the EL a izrelation. The Occurrence of this relation in the case of well-defined semiconductors may be explained in this way. However, the validity of applying these concepts to the present case is rather doubtful as no potential-free bulk region exists. It is important to realize that the electrodeis not in a stationary state during a potential scan. When the electrode potential is pulsed to a negative value at which, in a potential scan, the EL a P relation hdlds, a nearly linear increase of the emission intensity with time is observed which can last some tens of seconds. This increase cannot be caused by a time-dependent hole injection rate since the shielding current at the ring (H202 oxidation) in such an experiment is constant. Hence, the nonradiative recombination rate at competing centers must decrease. Fan et al." observed a similar time dependence, although on a much smallertime scale, for the EL of ZnS, which they explained by filling of electron traps. In a similar way, we explain the increase of the intensity with time (characteristic time constant T ) by removal of nonradiative recombinationcenters. When t / r goes to zero, implying that nonradiative recombination is dominant, the following expression for the ratio &ad of the radiative recombination rate and the total recombination rate can be derived: Since the rate of the linear increase is strongly dependent on the potential and is related to filling of electron traps, we take the constant T to be inversely proportional to the cathodic current i. Hence, the product it represents a cathodic charge Qath. In a previous section (see Figure 2a) we showed that the hole injection current ih is proportional to the cathodic current during a potentiodynamic scan:
ih a i As far as we are aware, this relation, which is essential in any explanation of the EL a i2relation, has never been verified in the l i t e r a t ~ r e . ~ Obviously, ~ . ~ ~ . ~ ~the EL intensity is proportional to both the efficiency of radiative recombination and the hole injection current:
(3) Here, Y denotes thescan rate. Sincethe current at a given potential does not depend on the scan rate, Q-th is proportional to i/u. The combination of eqs 1 and 2 gives the EL a i2relation. Equation 3 also explains the observed increase of the slope of the (EL,i2) curve with decreasing scan rate. 4.3. Comparison between EL and PL. We first discuss the difference in the degree of quenching of the Tb3+PL excited at two wavelengths; see Figure 6b. The quantum efficiency of emission upon excitation of the 5D3level at 378 nm is close to 1, because the energy gap between the 5D4 excited state and the next-lower level is too large for efficient multiphonon emi~si0n.l~ This energy gap is only determined by the energy levels of Tb3+ and is independent of the surroundings of the ion. No intensity change is expected as a function of potential if the concentration of Tb3+ ions bound to the oxide remains constant. The return of the intensity to the original value during the reverse scan indicates that this is the case. The small decreaseshown in Figure 6b is probably due to an enhancement of the nonradiative decay rate during cathodic polarization. H incorporation induces formation of OH groupsz1 which are known to decrease the emission quantum efficiency.15 From the foregoing, it is clear that the relatively strong quenching of the PL (Lxc= 260 nm) and the EL cannot be due to a decrease of the radiative quantum efficiency. Obviously, a change in the hole injection rate can also be excluded since that would not affect the PL intensity. Because the potential at which
quenching starts is identical for PL and EL (see Figure 6a), a common feature must be responsible for the quenching. The 260-nm excited PL and the EL originate from electron-hole recombination. Hence, thequenching must bedue toan increased nonradiative recombination rate. This is confirmed by the (PL,LI) curve of Eu3+-doped electrodes. This ion does not act as a recombination center. Accordingly, the characteristic PL shows the same small degree of quenching under excitation at 395 and 280 nm. The first wavelength corresponds to a transition within the 4f shell and the latter to a charge transfer. Increased nonradiativerecombination has often been suggested to account for EL quenching under large cathodic As discussed in a previous section, we can use the anodic charge after acathodicpulseasa measure for thedensityof recombination centers. The simultaneous decrease of the EL intensity and increase of the anodic charge in Figure 3 is in accordance with the generation of new recombination centers. A linear relationship was sometimes found between the EL intensity and the anodic charge. However, for a large series of experiments no simple relationship between these quantities was obvious. Hydrogen evolution takes place at the oxide/electrolyte interface. Hence, diffusion of hydrogen into the oxide layer also starts at the interface. Since in EL experiments the holes are injected from solution, considerable quenching of the intensity is expected. In the case of PL excitation at 260 nm, on the other hand, electron-hole pairs are created throughout the layer as the penetration depth of the light is comparable to or larger than the thickness of the oxide layer.* Since the diffusion length of holes is expected to be extremely small in this amorphous layer, PL, on average, originates from deeper inside the oxide layer than EL. Therefore, the degree of quenching is much smaller. The difference in the spatial origin of EL and PL has also been demonstrated on the basis of a difference in the amount of selfabsorption of the emission.2* Generation of new recombination centers near the oxide/ electrolyte interface can also explain the change in the EL spectrum under large cathodic bias; see Figure 7. In our previous paper, we argued that the difference between the PL and EL spectrum was due to a higher contribution to EL from surfacebound ions. The EL spectrumshowed relatively strongeremission at 490 nm (SD4-7F6) and 585 nm (5D4-7F4). These transitions are hypersensitive; Le., their intensity strongly depends on the deviation from inversion symmetry of the surroundings of the luminescent ion. This deviation is largest at the surface. Correspondingly, the decrease of the relative intensity of these two emission lines in Figure 7 points to a decrease of the contribution to EL of the surface-boundTb3+ions, in accordance with a quenching process located mainly at the surface. 4.4. Effect of Rotation Rate and HZOZConcentration. The effect of the rotation rate on the (EL,U) curves is complex. We first discuss the data of Figure 8a. In the rising part of the (EL,U) curve during a forward scan, the EL is only slightly dependent on the rotation rate. This can easily be accounted for, since the ring-disk measurements demonstrate that HzOz reduction is kinetically limited. When the potential is scannedin thenegative direction, a significant effect of the rotation rate is found, in accordancewith the improved HzO2 reduction kinetics; see Figure 2b. In the case of diffusion-limited H+ reduction (Figure 8b) similar characteristics are found. The decrease of the EL with increasing rotation rate in the potential range from -1.8 to -2.3 V is related to the H+ reduction current plateau. Since H2 evolution induces formation of new recombination centers, we ascribe this anomalous ( E L , U ~ /relation ~) to strongly enhanced quenching at high rotation rate. Not much attention has been paid in the literature to the influenceof the rotation rateon the ELof large-bandgapmaterials. However, in the cases described, e.g., in the systems Talos/ HZO2z9 and Ti02/S20&,6 a plot of the EL intensity vs the square root of the rotational rate often shows large y-axis intercepts and varying slopes, comparable to the present case at potentials
11768 The Journal of Physical Chemistry, Vol. 97, No. 45, 1993
negative with respect to the quenching potential. With regard to this complex dependence we can only speculate on some interaction between the hole injection rate and the generation and removal of H-induced recombination centers. Someelectrode heterogeneity, which was mentioned in the section on the (i,U) characteristics, may also play a role. The concentration dependenceof the EL is also highly nonlinear. The different shapes of the (EL,U) curves of Figure 9 point to a complex interplay between the H+and H202 reduction currents and the generation rate of new recombination centers. Interestingly, the onset of the (EL,U) curve shifts in negative direction as the H202 concentrationis increased, with a concurrent increase in H202 consumption as measured by using a ring-disk electrode. This can only be accounted for by a greatly increased nonradiative recombination rate. Thus, the centers responsible for recombination must be induced by H202. Actually, various authors have suggested surface states which result from the interaction of the oxidizing agent with the electrode surface.c6 4.5. EL as a Robe for Electrochemical Processes. In this section, we want to consider briefly the role of EL as a probe for processes at the electrode/electrolyte interface. The Occurrence of EL proves that H202 is being reduced, although this reduction does not show in the (i,U) curve. At potentials positive with respect to the EL maximum during a forward scan the (EL,U) relation is well-understood. The steep rise of the intensity with potential is related not only to an increased hole injection rate but also to an increased relative recombination rate at Tb3+.The role of these processes was partly elucidated by using rotating ring-disk electrodes (RRDE). The slight dependenceof the EL on the rotation rate points to a kinetically limited hole injection rate. This was confirmed by the RRDE experiments. At potentials negative with respect to the EL maximum and during the reverse scan, the (EL,U) relation becomes increasingly more complex. The combined results of PL and EL experiments show that the quenching of the emission must be correlated with the creation of new nonradiativerecombinationcenters at or near the oxide/electrolyte interface. The rotation rate dependence of theELand the RRDEresults points toa process whichcounteracts this quenching, viz., more efficient H202 reduction kinetics. A further complication is that the x axis in a potentiodynamic experiment represents not only a potential axis but also a time axis. As a consequence, we are not able to analyze without undue speculation the potential dependence of the EL with respect to effects of e.g. hole injection rate and competitive recombination rate. We would like to point out that the nonlinear effects, on the other hand, show that interpretation of the (EL,U) curve always requires studyof the intluenceofsuch quantities as rotation rate and concentration. Potential pulse experiments yield qualitatively similar results as the potentiodynamicexperiments. In this paper the emphasis is on the potential scan experiments because of the rather poor reproducibilityof pulse experimentsfor polarizationtimes longer than 10 s. It is interesting to compare the present results with those obtained using a symmetric double step pulse squence (3/0/-3/0 V, 62.5 Hz*). A rather linear relation between EL and HzO2 concentration and rotation rate is often found. Since the trapping kinetics of the Ta2O5 layer are too slow to adapt to the rapidly changing potential, a quasistationary state is obtained which in a way shows a linear dependence on the hole injection rate. 5. C ~ l u S i O u E
We have shown that the characteristic Tb3+emission of doped anodic Ta205 filmscan be used to investigate the relation between EL and electrochemical processes since it mirrors the background
Meulenkamp and Kelly EL of undoped Ta2O5. The advantage of using Tb3+ is 2-fold. First, its intensity is much higher than the background emission. Second, the spectral features of this rare earth ion yield extra information: a study of the EL and PL at various excitation and emission wavelengths showed that quenching of the emission at very negative potential must be due to H penetration into the oxide layer from the oxide/electrolyte interface. The potential dependence of the EL could be explained on the basis of two processes: hole injection rate .and generation and removal of new recombination centers. In the region close to the flatband potential the (EL,U) curve could be explained in detail with the help of rotating ring-disk experiments. At very negative potential, a complicated dependence on the concentration and rotation rate prevented such analysis.
Acbwledgmmt. We thank Prof. G. Blasse (Utrecht University) for helpful discussions. This work was supported by the Netherlands Foundation for Chemical Research (SON),with financial aid from the Netherlands Organization for Scientific Research (NWO).
References and Notes (1) Scee.g.: (a)Pettinger, B.;Sch~ppel,H.-R.;Ge~ha,H. Be?. BunsenGes. Phys. Chem. 1976,80,849.(b) Nakato, Y.; Tsumura, A,; Tsubomura,
H. Chem. Phys. Lett. 1982,85,387. (2) Smandek, B.; Gerischer, H. Electrochim. Acta 1985,30,1101. (3) Decker. F.; Pettinger, B.; Gerischer, H . J . Electrochem. Soc. 1983, 130, 1335. (4) Smandek, B.; Gerischer, H. Electrochim. Acta 1989,34, 1411. (5) Ponyak, S. K.; Sviridov, V. V.; Kulak, A. I.; Sa”, M. P. J. ElectroanaI. Chem. 1992,340,73. (6) Nogami, G.;Murakami, K.; Ohkubo, S.;Hamasaki, Y. 1.Elecrrochem. Soc. 1992, 139, 1777. (7) Haapakka, K.; Kulmala, S . Anal. Chim. Acta 1988, 208,69. (8) Meulenkamp, E. A.; Kelly, J. J.; Blasse, G. J. Phys. Chem. 1992,%, 1819. (9) Albery, W. J.; Hitchman, M . L. Ring-Disc Electrodes; Clarendon Press: Oxford, U.K., 1971. (10) Schmickler, W.; Schultze, J. W. Mod. Aspects Electrochem. 1986, 17,357. (11) Fan, F.-R. F.; Leempoel, P.; Bard, A. J. J . Electrochem. Soc. 1983, 130, 1866. (12) Memming, R. J. Electrochem. Soc. 1969,116,785. (13) Ouyang, J.; Fan, F.-R. F.; Bard, A. J. J. Electrochem. Soc. 1989, 136, 1033. (14) Butler, M.A.; Ginley, D. S . J . Electrochem. Soc. 1978,125,228. (15) Henderson, B.; Imbusch, G. F. Optical Spectroscopy of Inorganic Solids; Clarendon Press: Oxford, U. K., 1989. (16) Macagno, V.;Schultze, J. W. J. Electroanal. Chem. lW, 180,157. Schultze, J. W.; Macagno, V. A. Electrochim. Acta 1986,31,355. (17) Kankare, J.; Haapakka, K. In Extended Abstracts; 38th Meeting of the International Society of Electrochemistry, Maastricht, 1987;p 656. (18) Fichou, D.;Kasanyi, J. J. Electrochem. Soe. 1986,133, 1607. (19) (a) Scheller, F.; Landsberg, R.; Muller, S.J. Electroanal. Chem. 1969,20, 375. (b) Scheller, F.; Landsberg, R.; Wolf, H. Z . Phys. Chem. (Munich) 1970,213,345. (c) Scheller, F.; Landsberg, R. 1bld. 1970,244, 273. (20) White, T.; Stimming, U. Ae reported by: Stimming, U. Electrochim. Acta 1986,31,415. (21) Gusinskii, G. M.; Karpukhina, L. G.; Muzhdaba, V. M.;Naidenov, V. 0.;Tomilenko, G. F.; Khanin, S . D. Sou. Phys. SolldStcrte 1987,29,1867. (22) (a) Ord, J. L.; Bishop, S.D.; De Smet, D. J. J. Electrochem. Soc. 1991, 138,208. (b) Reichman, B.; Bard, A. J. Ibid. 1980,127,241. (23) Manivannan, A.; Itoh, K.;Hashimoto, K.;Sakata, T.; Fujishima, A. J. Electrochem. Soc. 1990, 137,3121. (24) Butler, M.A.; Ginley, D. S.Appl. Phys. tcrt. 1980,36,845. (25) Prener, J. S.;Williams, F. E. J . Electrochem. Soc. 1956,103,342. (26) (a) Kresclel, H.; Butler, J. K. Semlconducror h e r s and Heterojunction LEDs; Academic: New York, 1977;Chapter 2. (b) Sze, S. M. Semiconductor Devices Physics and Technology;Wilcy: New York, 1985; Chapter 3. (27) Yamase, T.; Gerischer, H. Bur. Bunsen-Ges.Phys. Chem. 1983,87, 349. (28) See e.&: Streckert, H. H.; Tong, J.-R.; Ellis, A. B. J. Am. Chem. SOC.1982,104,581. (29) Kankare, J. J.; Ryan, D. E.;Fiirst, B. J. Can. J. Chem. 1977,55, 1193.
