Electroluminescence of Dy3+ and Sm3+ Ions in Polycrystalline

(MSM) structure, is similar to a varistor and is able to operate under either ac or dc mode. The excitation of the trivalent rare earth ions is a cons...
0 downloads 0 Views 707KB Size
J. Phys. Chem. 1995, 99, 5674-5679

5674

Electroluminescence of Dy3+ and Sm3+ Ions in Polycrystalline Semiconducting Zinc Oxide S. Bachir, J. Kossanyi,.C. Sandouly, P. Valat, and J. C. Ronfard-Haret* Laboratoire des Matiriaux Moliculaires, CNRS, 2-8, rue H. Dunant, F 94320, Thiais, France Received: December 23, 1994@

The electroluminescence of Dy3+ and Sm3+ ions inserted in polycrystalline semiconducting zinc oxide was studied using a rare earth-doped sintered zinc oxide disc sandwiched between two metallic solders. This device, which corresponds to a metal-semiconductor-metal (MSM) structure, is similar to a varistor and is able to operate under either ac or dc mode. The excitation of the trivalent rare earth ions is a consequence of a hot electron impact process. The rare earth ions are located near the grain boundaries. Their luminescence acts as a probe and evidences hot electrons near the grain boundaries where, in accordance with the model of varistors, they are produced by thermionic emission.

Introduction The use of ZnS or ZnSe in electroluminescent devices is the subject of a very large number of studies,' and surprisingly, the ability of ZnO to be used in such devices received only little For both pure and rare earth-doped ZnO devices, the results are disappointing and difficult to interpret. Most of the articles report the same broad, more or less structured spectra, centered around 550 nm and attributed incorrectly when doped by rare earth5 to transitions between donor levels of the rare earth and acceptor levels of ZnO. Only our recent studies on the electroluminescenceof sintered semiconducting polycrystalline rare earth-doped ZnO electrodes report the emission spectra characteristic of transitions between the 4f levels of trivalent rare earth ions (RE3+).8-'2 These spectra were recorded in an electrochemical cell using electrodes operating under anodic bias at polarization potentials of ca. 10 V vs saturated calomel electrode (SCE), in contact with an aqueous electrolyte. A direct impact excitation mechanism was proposed to explain the dependence of the intensity of the emitted light upon both the applied potential and the RE3+ concentration. Unfortunately, the presence of the liquidsolid interface causes serious limitations to the potential use of this system in electroluminescent displays. In order to overcome this difficulty, we have developed a new device of metal-semiconductor-metal (MSM) structure, identical to that of a varistor in which a sintered polycrystalline ZnO pellet doped with RE3+ ions is sandwiched between two metal contacts. The preliminary results conceming the electroluminescence of Ho3+ ions have been published recently. l3 They show the potentiality of polycrystalline semiconducting ZnO as a host matrix in electroluminescent systems. To our knowledge, the present study is the first report on the observation of the luminescence of Dy3+ and Sm3+ ions in ZnO using such a structure. Experimental Section ZnO (Koch-Light, 99.99% pure) and rare earth oxide (RE2O3) (Rh8ne-Poulenc, 99.5% pure) powders are mixed together in an agate mortar. Pellets are made by pressing the mixture of the two oxides in a Specac press (4 t/cm2) in the presence of a small amount of ethanol. After sintering for 6 h at temperatures comprised between 1100 and 1200 "C under air, pellets of ca. 1 cm diameter and 1.0 3~ 0.1 mm thickness are obtained. The @

Abstract published in Advance ACS Absrracts, March 15, 1995.

0022-365419512099-5674$09.00/0

TABLE 1: Characteristics of the Samples sample 1

2

3

composition 0.5 at. % Dy3+ 1.0 at. % Dy3+ 1.0 at. % Sm3+ sintering temp, "C 1100 1150 1200 mean grain size, pm not measured 6.0 6.5 n

167

154

sintered discs are then electroded by covering both surfaces with In-Ga alloy and mounted in a Bakelite holder where the electric contacts are achieved by means of copper wires. The voltage and current are measured with a classical voltmeter and microammeter. For luminescence measurements, the pellet is located either in a Perkin-Elmer MPF-44 spectrofluorimeter or in a SLM Aminco 8000 C single-emission monochromator for variable-temperature study, in such a way that the image of the edge of the pellet is focused on the entrance slit of the emission monochromator. For each experiment the voltage, the current, and the relative light intensity are measured simultaneously in order to allow accurate comparisons. The scanning electron microscopy measurements were performed at the Service de Microscopie Electronique of the Universitk Pierre et Marie Curie of Paris with a JEOL CXII apparatus equipped with an ASID 4D scanning system. The electrical power dissipated in the pellets is not negligible; it reaches 600 mW during the experiment at room temperature at the highest applied voltages. This can lead to an increase of the temperature of the pellets, which will depend upon the applied voltage. Thus, the i-Vand B-Vcurves presented in this study are not free of a temperature shift effect. In order to reduce this effect, the pellets were cooled by blowing gaseous nitrogen in the room temperature measurements. Unfortunately, such a process is not possible in the measurements performed at low temperature (208, 223,243, and 265 K). The temperature controlled by the cryostat is that of the holder, and a difference can occur between this temperature and the actual temperature of the pellet. The temperature of the pellet is higher than that of the holder, and the difference will increase with the electrical power dissipated into the pellet. The results described below correspond to three RE3+-doped ZnO pellets whose characteristics are presented in Table 1. Results and Discussion Sample Characterization. The Sm3+- or Dy3+-doped sintered polycrystalline ZnO have been studied by scanning 0 1995 American Chemical Society

J. Phys. Chem., Vol. 99, No. 15, 1995 5675

Electroluminescence of Dy3+ and Sm3+ in ZnO

the RE3+ions. By analogy with the RE3+-dopedZnO electrodes and the RE3+-doped ZnS electroluminescent systems,17 the bands centered around 480, 580, 670, and 750 nm observed with the Dy3+-doped samples (samples 1 and 2 ) are attributed to the 4F9/z 6Hi5/z94F9/z 6Hi3/z,4F9/z 6Hii/z1and 4F9/z 6H9/2transitions of the Dy3+ ion, and in the spectrum recorded with the Sm3+-doped sample (sample 3), the bands centered around 575,615, and 665 nm are attributed to the 4G~/2 %5/2, 4G5/2 %7/2, and 4G5/2 6H9/2 transitions of the Sm3+ ion. The device being symmetrical, it can operate under both ac and dc modes, and the spectra recorded within both conditions are the same, although the system appears to be less luminescent under ac mode. Under constant applied voltage there is no variation with time of the emitted light intensity, and no degradation seemed to occur after long operating times (ca. 100 h) . Under our operating conditions we did not observe other patterns attributable to ZnO itself, just as for the RE3+-doped ZnO electrodes under anodic bias in contact with an aqueous e l e c t r ~ l y t e . ~This J ~ point will be discussed below since the characteristic luminescence pattern of ZnO was already observed with ZnO varistors,18 using a device containing no rare earth, similar to ours. Another important difference with the RE3+doped ZnO electrodes in contact with an aqueous electrolyte under positive bias lies in the voltage used to induce the electroluminescence of the RE3+ions. A polarization potential of ca. 10 V vs SCE was used for electrodes in contact with an aqueous electrolyte whereas applied voltages up to 300 V are used in the present study. For a direct electron impact excitation mechanism, by the use of simple models of the electron distribution and of the field configuration, two laws describing the dependence of the emitted light intensity B versus the applied voltage V were deduced: l9 the Destriau law

-

-

-

J

I

I

I

I

I

I

450

500

550

600

650

700

750

Wavelength (nm)

Figure 1. Electroluminescence spectra of sample 1 (upper) and sample 3 (lower) at room temperature (0.5 at. % Dy3+-dopedZnO pellet and 1.O at. % Sm3+-dopedZnO pellet, respectively) submitted to an applied voltage equal to 180 V dc.

electron micrography (SEM). The micrographs are similar to that already reported in other cases.11J2,14,15 The samples show a compact granular structure with polycrystalline-like irregular grains. RE3+ oxide aggregates appear as white spots on the micrograph. Holes at the grain boundaries, which have been attributed to the presence of the rare earth, result from the grain growth inhibiting effect of the dopant. The size of the ZnO grains is not uniform. It ranges from 3 to 15 p m (Table l), and a mean value is obtained by counting the number of grains on a selected distance. In order to reduce the error in this value, several different countings were performed; provided that the distance used to count the number of grains is large compared to the grain size, the result remains constant. An average number n of grain boundaries in series through the sample thickness, equal to the thickness of the pellets divided by the average grain diameter, is also obtained. However, all these values are only statistical, and in polycrystalline ceramics, n should be better described by a distribution around a mean value.16 Furthermore, depending upon both the shape of the grains and their size distribution, a proportionality constant has to be applied so that these n values are only approximate. The results are presented in Table 1. ElectroluminescenceMeasurements. An emission of light arising from the pellets is observed when they are submitted to an applied voltage. The spectra recorded with the pellets doped with Dy3+ and Sm3+ are presented in Figure 1. They are identical to those recorded using RE3+-doped ZnO electrodes in contact with an aqueous electrolyte under positive bias, in which a direct electron impact excitation mechanism was evidenced.lOJ1 They show only the characteristic patterns of

-

-

-

-

and the Alfrey-Taylor law

B = Bo e~p[-(VdV)'/~] where Bo and VO are constants. Our results are presented in Figure 2. For samples 1 and 2 the intensity of the emitted light is measured at 550 nm (4F9/2 %13/2 transition of the Dy3+ ion17),and for sample 3 the emitted light intensity is measured at 665 nm (4G5/2 6H9/2transition of the Sm3+ ion17). For the three samples, the B-V plots seem to better agree with the Alfrey-Taylor relationship. But, both the Destriau and AlfreyTaylor laws refer to homogeneous systems while polycrystalline ZnO electrical properties are dominated by grain boundaries.20 The emitted light intensity and the applied voltage are not the primary quantities. The light output depends on the current i as well as on the efficiency of turning current into light. As soon as a minimum voltage threshold is passed, the field is a more fundamental quantity than the v01tage.l~ Therefore, it is difficult to draw firm conclusions from the B-V laws. The electrical properties of polycrystalline semiconductors were the subject of a very large number of s t u d i e ~ . ~ ~ - ~ ~ Specially, previous studies on ZnO varistors have shown that double Schottky barriers are present at the grain-grain contact and that the current, for low applied voltages, is controlled by the grain boundaries where these double Schottky barriers occur. ZnO grains are good conductors so that the entire voltage is sustained across narrow regions at the grain boundaries, and the total applied voltage Vis divided into n individual potential drops u21.23

-

-

Bachir et al.

5616 J. Phys. Chem., Vol. 99, No. 15, 1995

loooo 1000

5

1

0.1 0.05

O.

mae

0.07

0.09

0.11

0.13

0.15

v-112

loooo

1 c

Figure 3. Energy band diagram near a grain boundary under an applied voltage U. The impact excitation of a TR3+ion by electrons injected in the positively biased grain is shown. E,, E,, and Et are the conduction

J

I

band, valence band, and the quasi-Fenxi level, respectively.

loo0

b

2.0 103

e

2

v

0.1

0

0.005

0.01

0.015

0.02

0.025

0.03

1IV Figure 2. Alfrey-Taylor (upper) and Destriau (lower) plots, at room temperature, of the emitted light intensity, observed at 580 nm for sample 1 (0.5 at. % Dy3+)and sample 2 (1.0 at. % Dy3+) and observed at 665 nm for sample 3 (1.0 at. % Sm3+):(A)sample 1, dc mode; (4) sample 1, ac mode; (0)sample 2, dc mode; ( 0 )sample 3, dc mode. For the ac mode the voltage is given in r m s values.

V=nU

(3)

each corresponding to one grain boundary barrier. Using the n values reported in Table 1 and, for applied voltages lower than 300 V used in this study, U remains always lower than 2 V. The structure of the current-voltage characteristics of a grain boundary submitted to a low applied voltage is described by the theory of Pike and Seager24 on the basis of a double depletion layerkhermionic emission model (Figure 3). The relationship between the current density J through a grain boundary and the potential U applied to this grain boundary is

A* is the Richardson constant (A* = 30 A cm-2 K-* for ZnO), T the temperature, 5 the energy difference between the Fermi level and the conduction band outside the depletion layers, (Pg the banier height of the grain boundary potential barrier, q the charge of the electron, and k the Boltzmann constant. Equation 4 predicts an increase of the conductivity with the temperature. For varistors an increase of the conductivity with the applied voltage is also observed. The dc mode currentvoltage characteristics, in the 0-300 V range, at room temperature for the three devices, and at different temperatures for sample 3 are shown in Figure 4. Sample 2 presents a very weak non-ohmic behavior, whereas samples 1 and 3 show a larger increase of their conductivity with the applied voltage, and in

.-

1.5 103

t

0

O

A

1.0 103

5.0 102

0 0

50

100

150

200

250

300

350

v (VI Figure 4. Current-voltage curves under dc mode for sample 1 (0.5 at. % Dy3+) at room temperature (A),sample 2 (1.0 at. % Dy3+)at room temperature (0),and sample 3 (1.0 at. % Sm3+) at room and 243 K (m). temperature (O), 265 K (0). addition, a variable-temperature experiment with sample 3 shows an increase of the conductivity of this sample with the temperature. Figure 2 and eq 3 show that the luminescence of the RE3+ ions occurs when the drop U of the applied voltage ranges from 0.5 to 2.0 V by grain or V by grain boundary barrier. These values correspond to energies close to that of the emitting levels of the RE3+ ions. They are sustained across narrow regions at the grain boundaries where the polarization is large. In electroluminescent RE3+-doped ZnS systems the field lies around 2 x lo6 V cm-', and this is a condition for producing the hot electrons necessary to induce the impact excitation of the RE3+ ions.lg In the ZnO systems under study, the pellets are submitted to an average field lower than 3 x lo3 V cm-', but high local fields can be reached near the grain boundaries. Electron trapping and detrapping occur at the grain boundaries, and those injected in the positively biased grains can pick up a large amount of kinetic energy in the high electric field and impact-excite the rare earth ions. In ZnO, the emission of longitudinal optical phonons leads to a strong cooling of the hot electrons, and only a field in the order of lo6 V cm-' is able to produce the hot electrons suited for the impact excitation of the rare earth ions.20

J. Phys. Chem., Vol. 99, No. 15, 1995 5677

Electroluminescence of Dy3+ and Sm3+ in ZnO 100

1400 1200

80

1000

5

60

800

40

600

v

m

p

"t. \

5

5

400 20

200

a

0

0

0.5

1

1.5 i (mA)

2

2.5

3

Figure 5. Intensity E (in arbitrary units) of the emitted light versus the electrical current flowing through sample 2 (1.0 at. % DyS+) observed at 550 nm (0)and sample 3 (1.0 at. % Sm3+)observed at 665 nm (0)at room temperature.

Structure studies on RE3+-doped polycrystalline ZnO have shown that the RE3+ ions act as ZnO grain growth inhibitors. They do not substitute homogeneously for Zn2+ ions in the semiconducting lattice, but rather remain near the surface of the grains and in agregates at the grain ~ o m e r s . ~ l -The l ~ SEM and energy dispersive spectrometry results already presented12 do not allow to establish whether the RE3+ion diffuse into the ZnO grains close to their surface or form intergranular films between the ZnO grains. For rare earth-doped zinc oxide varistors, no intergranular glassy films could be detected.15For BizOs/ZnO varistors, Bi atoms were detected in low concentration inside ZnO grains, near the surface at distances up to 40 nm.26 The size of the Bi3+ ions is larger than that of the RE3+ ions, and their charge is identical. Thus, it seems likely that the RE3+ ions, as the Bi3+ ions, diffuse into the ZnO grains near the surface. The observed luminescence would be a consequence of the presence of both the RE3+ ions and the hot electrons near the grain boundaries, the hot electrons being responsible for the impact excitation of the RE3+ions. Electroluminescence was already observed in ZnO varistors containing no rare earth.18 Both ZnO band-gap and subbandgap luminescences were characterized. The band-gap luminescence, arising from electron-hole pairs recombinations, was taken as experimental evidence of holes production during the breakdown of the varistors. In the same way, the luminescence of RE3+ions makes an experimental verification of the presence of hot electrons. In the present case, the RE3+luminescence occurs for lower applied voltages ( U 2 V) which correspond to the prebreakdown region of the varistors. The observation of the RE3+luminescence at voltages lower than those necessary to produce the breakdown of varistors excludes any excitation of the RE3+ ions by an energy transfer mechanism following an electrodhole pair recombination. No simple relation can be deduced from the dependence of the emitted light upon the current intensity (Figure 5). However, as already reported for ZnO electrode^^^" above a certain threshold, the intensity of the emitted light becomes proportional to the current flowing through the pellet. The intensity of the emitted light, in electron impact electroluminescent systems, is proportional to the number of hot electrons of energy compatible with the impact excitation of the luminescent centers.27 Thus, the thresholds observed in Figure 6 would be a consequence of the energy distribution of the hot electrons relative to the energy of the emitting levels of the RE3+ ions (4F9/2for the Dy3+ ion and 4G5,2for the Sm3+ion). For weak currents, corresponding to low applied voltages, the mean energy of the hot electrons is low, and only a small number of hot electrons is able to

0 ~ " ~ ' " " v" ' " ~ ' ' ~ ~ 0 0.005 0.01 0.015 liV

0.02

(VI)

Figure 6. Plot of ln(E/i) vs 1/V at room temperature for sample 2 (1 .O at. % Dy3+) observed at 550 nm (0)and sample 3 (1.0 at. % Sm3+) observed at 665 nm (0).

impact-excite the RE3+ ions to their emitting levels. For high currents, corresponding to high applied voltages, the mean energy of the hot electrons is high as compared to the energy necessary to populate the emitting levels of the RE3+ ions; the intensity of the emitted light tends to be proportional to the current. The electroluminescence emission intensity due to a specific transition is proportional to the number N* of luminescent centers in their excited state. This number may be ~ r i t t e n ~ ~ ~ * ~ as N*

OC

j- Nn$(E) v(E)c(E)dE

(5)

where N and ne are the number of luminescent centers and hot electrons, respectively, f ( E ) is the energy distribution function of the hot electrons, v(E) is the velocity of the hot electrons, and a(@ is the impact cross section of the considered luminescence center. In this expression, the product n 8 . Q represents the number of hot electrons which are able to induce an impact excitation of the luminescent centers into their emitting level I of energy Et. In semiconductors, the hot electrons energy distribution is given by the Boltzmann's statistic^,^^ and a rough approximation, neglecting the variations of v(E) and a(E)in the integration of (6), leads to express the B/i ratio with a Destriau-type relationship

B/i = exp( -EI/Ek)

(6)

in which Ek is the average kinetic energy picked up by the hot electrons near the boundary in the positively biased grain. The maximum value EOof this energy is given by

Eo=qU+

@B

(7)

As already stated by eq 3, the total applied voltage is divided into n individual potential drops, each of them corresponding to one grain boundary. The voltage range corresponding to the observation of the luminescence of the rare earth implies qU >> kT; then the current is controlled by the first exponential of eq 4, which describes the lowering of the barrier. Typical @B values of cu. 1 eV corresponding to a current density of the order of 1 p A cm-2 at 400 K are reported for ZnO varistors.20 In the present case the i-V curves shown in Figure 3 are not highly nonlinear, and for an applied voltage equal to 0.6 V by grain boundary the current is close to 1 mA; this could indicate that the barriers are lower than those usually reported for ZnO varistors.20 An estimate of @B using eq 4, at room temperature for sample 3, leads to a decrease from 0.55 to 0.50 eV when

5678 J. Phys. Chem., Vol. 99,No. 15, 1995 the voltage applied to the sample increases from 100 to 210 V. In this voltage range, Arrhenius plots of the current lead to activation energies close to 0.35 eV. Thus, neglecting @B toward qU in eq 7 is possible only for the highest applied voltages. An approximation, discussed later, leads to consider E k = EO;then a plot of ln(B/i) vs 1/V should give a straight line with a slope equal to the energy of the emitting level of the RE3+ ion multiplied by the number of baniers in series through the thickness of the pellets. Figure 6 shows the results corresponding to samples 2 and 3 at room temperature. The range of variation of both B/i and 1/V is larger for sample 3 than for sample 2, and in the 140300 V range, the variation of ln(B/i) vs 1/V is linear for both samples. The slope of the ln(B/i) vs 1/V plot is equal to -760 V for sample 2, and for sample 3 it decreases from -210 to -593 V when the voltage applied to the pellet increases from 50 to 210 V. The luminescence of the Dy3+ ion, observed at 580 nm, corresponds to an emission from the 4F9/2level, the energy Er of which is 2.0 x IO4 cm-’ (2.55 eV).30 The luminescence of the Sm3+ ion originates from the 4GW2level (Er = 2.20 eV).30 The theoretical values of the slopes of the ln(B/i) vs 1/V plots are calculated by multiplying the energy (E1 or Er) of the emitting levels of the RE3+ions by the number n of grain boundary barriers in series through the thickness of the pellets. Using the n values of Table 1, one finds -425 and -340 V for samples 2 and 3, respectively. The calculated values are in the same range as the experimental ones. In the proposed model, the density of states at the grain boundary and in the semiconductor near the bottom of the conduction band are neglected. The density of states at the grain boundary alters the current-voltage characteristics, and its effect is included in the measure of the current. The density of states near the conduction band of the semiconductor alters the distribution function of the hot electrons and, as a consequence, should alter the ln(B/i) vs 1/V plots. However, its effect is difficult to take into account and is usually neglected when Ek is small compared with E1 (or E I . ) . ~ ~ Several explanations can be proposed to account for nonlinearity and the slope of the plots of Figure 6 . Since sample 3 presents the largest variations, we shall limit the discussion to this sample. The low-voltage part (50-100 V) of the ln(B/i) vs l/Vplot is quite linear with a slope equal to -210 V, a value higher than the calculated one (-340 V). This indicates that the actual mean value Ek of the kinetic energy of the hot electrons is underestimated within this voltage range (Ek> qV/ n). In the positively biased grain, the hot electrons can pick up the maximum average kinetic energy EOgiven by eq 7. Thus, the underestimation of the Ek can arise from the neglect of @B and/or from the use of an erroneous n value. The lower the polarization, the greater the extent of @B in EO. Therefore, the low-voltage tail of the ln(B/i) vs 1/V plot is a consequence of the barrier. When the polarization increases, @B decreases while qU increases. The extent of @B in EOdecreases, leading to an decrease of the slope of the ln(B/i) vs W p l o t . The underestimation of EO can also be a consequence of an erroneous n value, and overestimating n leads to underestimate U. As already outlined, the size of the ZnO grains is not uniform but is described by a distribution around the mean value n which must be corrected depending upon the shape of the grains. The correction leads always to increase the actual size of the grains with respect to the measured value and, therefore, to lower the number of grain boundaries. The value of n is overestimated by a factor which can reach 2.25.16 Then, the low value (-210 V) of the slope of the ln(B/i) vs l/Vplot in the low-voltage tail is a consequence of the lack of correction to n and/or a

Bachir et al. consequence of the distribution due to the presence of larger grains. A reduction in the breakdown voltage of varistors was already proposed to result from this grain size d i s t r i b u t i ~ nIn .~~ the present case this explanation cannot be excluded. However, any correction to n, proposed to account either for the irregular grains shape or for the distribution, will decrease its value and imply a lower extent of the barrier height @B toward qU in the value of EO. In the high-voltage range the slope of the ln(B/i) vs 1/V plot equals -593 V. Clearly this value exceeds the calculated one (-340 V) and, unlike the low-voltage range, corresponds to an overestimation of the mean kinetic energy Ek of the hot electrons which impact-excite the Sm3+ ions. The actual value of this energy is lower than that resulting from the ln(B/i) vs 1/V plot. Owing to the reasons developed above, this overestimation of the energy does not result from an approximation or from a mistake on @B or n. It means only that the mean energy of the hot electrons which impact-excite the Sm3+ ions is lower than the energy difference qU between the valence (or conduction) bands of two adjacent grains, outside the depletion layers (Ek < qU,(PBis neglected and qU = EO). This can have several origins: (i) In the approximation EO = Ek. Due to a cooling of the hot electrons which loose their excess energy by optical phonon scattering, it results Ek < EO. This cooling of the hot electrons always occurs. In ZnO, when the field increases from lo5 to lo6 V cm-l, the rate of energy loss decreases from 0.50 to 0.15 eVJ100 A.2o (ii) In an overestimation of EO. Due to an insulating character of the barriers, it results EO < qU. The lack of observation of intergranular glassy films in FE3+-Zn0 varistors does not favor this hypothesis. (iii) In a limitation of the kinetic eneryg of the hot electrons which impact-excite the Sm3+ ions. This limitation of the kinetic energy can result from the localization of the Sm3+ions near the grain boundaries. The mean kinetic energy of the hot electrons increases with the length of their trajectories within the positively biased grain. For a donor density equal to 1OI8 and an applied potential U equal to 1.5 V by grain boundary, the width of the depletion layer in the positively biased grain is equal to 55 nm. If the rare earth ions are located close to the boundaries at distances lower than this 55 nm value, the mean kinetic energy of the hot electrons which excite the Sm3+ ions is lower than Ek. These three limitations of the kinetic energy of the hot electrons can occur at any voltage, but it is difficult to estimate their extent in the low-voltage part of the ln(B/i) vs 1/V plots. Finally, note that in the 140-300 V range, both samples present the same ratio between the experimental and calculated values of the slopes (593/340 s 760/425) of the ln(B/i) vs 1/V plots. Equation 4 indicates that the current flowing through the pellets depends upon the temperature. It is known, in semiconductors, that the hot electrons distribution, given by Boltzmann’s statistics, depends also upon the temperature. For sample 3, Figure 4 shows that in accordance with eq 4 the current is temperature dependent. Activation energies close to 0.35 eV were measured for the current in the 100-210 V range. Consequently, the emitted light intensity must also depend upon the temperature. The plots of ln(B/i) vs 1/V at different temperatures presented in Figure 7 have the same shape as the one obtained at room temperature. In the 150-300 V range, the three variations at 208,223, and 243 K are linear and parallel with the same slope (-595 V) as the one obtained at room temperature with a different instrumentation and deduced from Figure 6. The slopes of the low-voltage tails are equal to -305

J. Phys. Chem., Vol. 99, No. 15, 1995 5679

Electroluminescence of Dy3+ and Sm3+ in ZnO

5.5

4

4*5

32 *. 53

k

the presence of the hot electrons near the grain boundaries where the rare earth ions are located. Owing to the numerous simplifications, the correlation between the microscopic and the phenomenological measurements is only approximative.

e

t

1

References and Notes (1) Electroluminescence; Proceedings of the Fourth Intemational Workshop, Tottori Japan, October 11 -14, 1988; Shionoya, S., Kobayashi, H., Eds.; Springer-Verlag: Berlin, 1989. (2) Thomas, B. W.; Walsh, D. Electron. Lett. 1973, 9, 362-3. (3) Takata, S . ; Minami, T.; Nanto, H. J. Lumin. 1988,40&41,794-5. (4) Takata, S . ; Minami. T.; Nanto, H. Jpn. J. Appl. Phys. 1981, 20, 1759-60. (5) Bhushan, S.; Kaza, B. R.; Pandey, A. N. Pramana 1978, 11, 6772. (6) Bhushan, S.; Pandey, A. N.; Kaza, B. R. J . Lumin. 1979,20, 2938. (7) Tripathi, L. N.; Chaubey, B. R.; Mishra, C. P. Acta Phys. Pol. A 1981, 59, 15-25. (8) Ronfard-Hmt, J. C.; Kouyate, D.; Kossanyi, J. Solid State Commun. 1991, 79, 85-88. (9) Kouyate, D.; Ronfard-Haret, J. C.; Kossanyi, J. J . Electroanal Chem. 1991, 319, 145-60. (10) Kouyate, D.; Ronfard-Haret, J. C.; Kossanyi, J. J . Lumin. 1991, 50, 205-10. (11) Kouyate, D.; Ronfard-Haret, J. C.; Kossanyi, J. J . Mater. Chem. 1992, 2, 727-32. (12) Ronfard-Haret, J. C.; Azuma, K.; Bachir, S.; Kouyate, D.; Kossanyi, J. J. Mater. Chem. 1994, 4, 139-44. (13) Bachir, S.; Kossanyi, J.; Ronfard-Haret, J. C. SolidState Commun. 1994, 88, 795-801. (14) Mukae, K.; Tsuda, K.; Nagasawa, I. Jpn. J. Appl. Phys. 1977, 16, 1361-8. (15) Williams, P.; Krivanek, 0. L.; Thomas, G.; Yodogawa, M. J.App1. Phys. 1980, 51, 3930-4. (16) Mendelson, M. I. J . Am. Ceram. SOC. 1969, 52, 443-6. (17) Chase, E. W.; Hepplewhite, R. T.; Krupka, D. C.; Kahng, D. J . Appl. Phys. 1969, 40, 2512-9. (18) Pike, G. E.; Kurtz, S . R.; Gourley, P. L.; Philipp, H. R.; Levinson, L. M. J . Appl. Phys. 1985, 57, 5512-8. (19) Allen, J. W. J . Lumin. 1981, 23, 127-39. (20) Greuter, F.; Blatter, G. Semicond. Sci. Technol. 1990, 5, 111-37. (21) Levinson, L. M.; Philipp, H. R. Ceram. bull. 1986, 65, 639-46. (22) Eda, K. IEEE Electrical Insulation Mag. 1989, 5, 28-41. (23) Gupta, T. K. J. Am. Ceram. Sot. 1991, 73, 1817-40. (24) Pike, G . E.; Seager, C. H. J. Appl. Phys. 1979, 50, 3414-22. (25) Emtage, P. R. J . Appl. Phys. 1979, 50, 6833-7. (26) Olsson, E.; Falk, L. K. L.; Dunlop, G. L.; Oesterlund, R. J . Mater. Sci. 1985, 20, 4091-8. (27) Krier, A,; Bryant, F. J. Phys. Status Solidi A 1984, 83, 315-22. (28) Allen, J. W.; Ayling, S . G. J. Phys. C: Solid State Phys. 1986, 19, L369-73. (29) Baraff, G. A. Phys. Rev. 1962, 128, 2507-17. (30) Krupka, D. C. J . Appl. Phys. 1972, 43, 476-81.

5

0.002

0.005

1/v

0.008 (V.')

0.011

Figure 7 . Plot of ln(B/i) vs 1IVfor sample 3 (1.0 at. % Sm3+)observed 243 (W), and 265 K (0). at 665 nm at 208 (A), 223 (e),

and -230 V for the experiments performed at 243 and 265 K, respectively. For the experiments performed at 208 and 223 K, the light emitted at voltages lower than 150 V is too weak to be measured. For the 265 K experiment, the disagreement with other experiments, which affects mainly the highest voltage points, can be attributed to a temperature shift effect related to a higher electrical power dissipation in the pellet at this temperature. In the low-voltage range the precision of the measurements is too low to allow an analysis of any possible variation of the slopes of the plots. Nevertheless, in the 150-300 V range, the good agreement between the results obtained at 208, 223, and 243 K and at room temperature is striking. It indicates that the temperature has no effect on the variation of ln(B/i) vs 1/V. Consequently, this variation can be attributed to a specific property of the sample. However, this last result does not enable one to corroborate any of the hypotheses proposed above, explaining the discrepancy between the experimentalresults and the theoretical model.

Conclusion We have shown that polycrystalline ZnO is able to be used as a host matrix to observe the electroluminescence of trivalent rare earth ions. The luminescence of rare earth ions is the result of an electron impact excitation process, and the hot electrons responsible for the impact excitation of the rare earth are produced by thermionic emission at the ZnO grain boundaries. The luminescence of the rare earth, acting as a probe, evidences

JP943398S