CHARACTERIZATION OF HIGH POWER UV-LED

Transcription

1 HELSINKI UNIVERSITY OF TECHNOLOGY Department of Signal Processing and Acoustics Metrology Research Institute CHARACTERIZATION OF HIGH POWER UV-LED Special Assignment in Metrology S (5ov) Henri Nykänen 61301J Instructor: Lic. Sc. (Tech.) Pasi Manninen Department of Signal Processing and Acoustics Espoo 2008

2 Contents 1 Introduction 1 2 Theory Bandgap in semiconductors Properties of pn-junctions Radiative recombination in pn-junction Inuence of temperature on LEDs Typical LED structures Homojunction LED Heterojunction LED Quantum well LED Measurements and results Current-voltage curve Spectrum Current dependence Temperature dependence Duty cycle dependence Conclusions 20 ii

3 List of Figures 2.1 Illustration of valence- and conductionband in semiconductor and the band gap in between Illustration of bent valence- and conduction band with an impurity state inside the forbidden energy gap working as a recombination center Illustration of pn-junction and its energy band structure. Here E c and E v are conduction and valence band energies respectively and E f is the Fermi energy. For details see refs [3, 2, 5] Illustration of typical Current-Voltage curve of a pn-junction diode. [1] Illustration of radiative recombination Illustration of the theoretical spectrum of LED pn-homojunction structure. A) Doping by diusion or ion implantation. B) Epitaxial growth pn-heterojunction structure and the corresponding band structure A) Practical structure of modern GaN-based LED. B) Simpli- ed illustration of MQW LEDs energyband structure Current of the LED as function of applied voltage in room temperature. The C-V curve of the forward voltage is shown in the inset Normalized spectrum of the UV-LED with 350, 500 and 700mA current in logaritmic scale Normalized spectrum of the UV-LED with 350, 500 and 700mA current iii

4 3.4 Graph representing the diode voltage as a function of temperature Graph representing the diode spectrums peak intensity as a function of temperature Wavelength of LED spectrum peak as a function of temperature with 50mA current Illustration of the signal used to determine the eect of pulse ratio in LED FWHM of the LED spectrum as a function of pulse-ratio Peak wavelength of the LED emission spectrum as a function of pulse-ratio iv

5 Chapter 1 Introduction Study of light emitting diodes (LEDs) is very important these days since more and more applications of these energy ecient devices are used around the globe. There are immense number of possible uses for light sources which can be used as energy saving alternatives for traditional lamps or integrated easily with modern microfabrication technology. This assignment is about dening basic electrical and optical properties of a certain high power ultraviolet (UV) LED. It can be used as a comparison for other comparable devices or as an brief introduction to operation of LED's. The basic physical principles of LED operation are explained along with the most commonly used LED structrures. The most important characterization measurements are introduced and then applied to the UV-LED in question. 1

6 Chapter 2 Theory This chapter will briey introduce the basic theory of LEDs. The derivation of formulas and the advanced quantum mechanics are neglected and they can be found in the references. The idea is to show the most important parts of the theory which explain the results of the measurements obtained for the test LED. 2.1 Bandgap in semiconductors In crystalline semiconductor materials, possible energy states of electrons are divided by the forbidden energy gap. In absolute zero temperature, all the electrons are located below the forbidden energy gap in valence band. When temperature rises or some other mechanism gives the electrons enough excess energy they can transit over the gap to the conduction band which is empty in absolute zero. The band structure of a semiconductor is a direct result from crystalline structure. It can be derived starting with the Schrödinger equation and using the boundary conditions of a periodic material. The derivation will result to an equation for the allowed energy states of electrons which cannot be true for certain areas in wave vector domain. These areas are the forbidden energy gaps.[3] The wider the band gap the larger the energy is required to lift electrons to the conductor band. This means that materials with a wider band gap are better insulators than the ones with a smaller band gap. One could think that in a band that is full the electrons do not have any space to move and a band that is empty has nothing to carry the charge. Metals do not have 2

7 CHAPTER 2. THEORY 3 Figure 2.1: Illustration of valence- and conductionband in semiconductor and the band gap in between. a forbidden energy gap and are therefore good conductors. The dierence between insulator and semiconductor is not very accurate. Certain literature sources claim that if band gap is wider than 3eV then the material is an insulator but this is only a rounded direction giving value. When electrons transit away from valence band they leave empty spaces. These spaces are called holes and they can be treated as charge carriers with positive charge. Thus a current through a semiconductor can be considered to consist of two parts: the electron movement in the conduction band and the hole movement in the valence band. When there are electrons in the conduction band and holes in the valance band recombination of these is possible. Hence the electron drops back to the valence band to ll the hole. The excess energy of the electron is released as a photon or a phonon. Phonons are lattece vibrations and can be interpret as heat. Recombination involving a photon is discussed in the next section. In real semiconductor materials the bands are bent to parabolic shape. The valence band has maximum and conduction band has minimum at some values of wave vektor k. Because it is energetically favourable the electrons gather to conduction band minimum and the holes gather to valence band maximum. If these maximum and minimum are at the same k value the material is said to have a direct band gap. Otherwise the material has indirect band gap. In direct band gap materials the recombination is easy because

8 CHAPTER 2. THEORY 4 there is no need to change the wave vector value of the electron while transition happens. With indirect band gap materials the recombination is harder and must involve either an excess phonon which changes the k-value of the electron or some kind of platform inside the forbidden energy gap that the electron can use as a middle station during transition. In pure semiconductor material the band gap is free of defects. But when there are impurities in the material or the lattice structure is broken for some reason such as dislocation of lattice atoms or surface interface there are allowed states also inside the energy gap. These states can hold charge carriers and can act as recombination centers. In some cases this is desirable for example when using indirect band gap semiconductors in light emitting devices or adjusting the colour of emitted light. Usually the defects in band gap result to non-radiative recombination which deteriorate the performance of the device. Figure 2.2: Illustration of bent valence- and conduction band with an impurity state inside the forbidden energy gap working as a recombination center. 2.2 Properties of pn-junctions A pn-junction is an interface of two semiconductor materials with dierent types of majority charge carriers. For example there could be a junction of two silicon bulks with dierent doping materials which introduce holes or

9 CHAPTER 2. THEORY 5 electrons into the silicon. Dopant materials removing excess electrons from bulk materials are called acceptors and dopants giving up electrons are called donors. The pn-junction has the property of letting current ow only in one direction. When these two dierently doped semiconductors are joined together a depletion region is formed. This means that the free electrons introduced by the donors diuse to the other side of the junction where there are holes and recombine with them. This process leaves the junction area depleted of free charge carriers. Only the ionized dopants are left in each side of the junction. These dopants form the space charge region which then produces a potential called diusion voltage [5] V D = kt ln N AN D, (2.1) q e where q e is the charge of electron, N A and N D are the acceptor and donor concentratinos respectively and n i is the intrinsic carrier concentration of the semiconductor. n 2 i Figure 2.3: Illustration of pn-junction and its energy band structure. Here E c and E v are conduction and valence band energies respectively and E f is the Fermi energy. For details see refs [3, 2, 5] The I-V characteristics of the pn-junction can be derived from the Shockley equation [5] and the equation obtained for pn-junction diode current is

10 CHAPTER 2. THEORY 6 I = ea ( Dp τ p n 2 i N D + Dn n 2 i τ n N A ) ( ) (e ev 1) Dp Dn kt = ea N A + N D e e(v V D ) kt, τ p τ n (2.2) where D n,p and τ n,p are the electron and hole diusion constants and the electron and hole minority carrier lifetimes respectively. From this equation it can be seen that the current increases rapidly after the diode voltage reaches V V D. The treshold voltage is V T H V D. Figure 2.4: Illustration of typical Current-Voltage curve of a pn-junction diode. [1] 2.3 Radiative recombination in pn-junction When a pn-juction is exposed to forward current the injected minority carriers are recombining with majority carriers of the junction materials. This recombination can result to luminous behaviour of the diode which is then obviously called the light-emitting diode. The radiative recombination is the base of practically every semiconductor light emitting device. An electron transiting from conductino band to valence band gives up the amount of energy that corresponds to the transition distance in energy scale. The emitted photon has an energy of

11 CHAPTER 2. THEORY 7 E = hf, (2.3) where h is the Planck Constant, E is the excess energy released from the transition and f is the frequency of the photon. Thus the wavelength of the emitted light corresponds to the electron transition distance. Depending on the radiative recombination mechanism a certain colour of light is then obtained. Figure 2.5: Illustration of radiative recombination. The most important factor in determining the wavelength of emitted photons is the length of the forbidden energy gap. The larger the gap is the more excess energy the electrons in conduntion band have and hence the smaller is the wavelength of the emitted photon when electron transits back to valence band. This is called the band-to-band recombination. An important part of the light emitting devices which emit photons with small wavelength are made from Gallium Nitride (GaN) based materials because of its wide band gap. Other radiative recombination mechanisms include for example donoracceptor transitions, band-impurity transitions and exciton transitions. 2.4 Inuence of temperature on LEDs Usually LEDs are driven with constant current I. The temperature dependence of diode voltage can be determined by solving the voltage from 2.2

12 CHAPTER 2. THEORY 8 [5]: V (T ) = kt e ln I I S + E g(t ). (2.4) e The rst term represents the shift of the Fermi level due to temperature change and the second term represents the changes of band gap energy. The band gap energy gets smaller as the temperature rises and this results to decreasing voltage and increasing wavelength. For details see reference [5]. The emission intensity depends on temperature as [5] I = I 0K e T/T 1, (2.5) where T 1 is the characteristic temperature that describes the temperature dependence of the LED. The higher T 1 is the smaller the temperature dependence which is desirable. When T increases the intensity decreases. The emission spectrum of a LED consists of the photons emitted by recombining electron-hole pairs. The maximum intensity is obtained at wavelength corresponding to the band gap energy added with the thermal energy. However the bands are parabolic and not all recombinatinos happen exactly at the smallest gap. Thus the spectrum is spread to smaller wavelenghts. Obviously the longest wavelength is at the minimum gap energy but this can also change if there are some other recombination centres inside the band gap which are always present in real life devices. The emission intensity is a function of energy and temperature [5] I(E, T ) E E g (T )e E k b Tc, (2.6) where k b is the Boltzmann constant, and E g is the forbidden energygap. The theoretical spectrum is illustrated in gure 2.6. The square root part is the dependence resulting from density of states (DOS) and the exponential part is the Boltzmann distribution of carrier states. [5]. 2.5 Typical LED structures Homojunction LED The homojunction LED is the simplest of LED structures. It is basically a pn-junction with both sides p and n made out of the same material but other

13 CHAPTER 2. THEORY 9 Figure 2.6: Illustration of the theoretical spectrum of LED. side doped with donor and other with acceptor material. This can be done by doping n-type bulk material at certain area with enough acceptors to make the area p-type (gure 2.7 A) and vise versa. Also an epitaxial growth is possible when introducing donors and acceptors to the source gasses in turns (gure 2.7 B). These materials form a luminous junction when exposed to forward current Heterojunction LED The problem with the homojunction LED is that when the junction is located deep beneath the materials the emitted photons are reabsorbed. If the junction is placed near the surface the non-radiative surface recombination takes place thus greatly reducing the device eciency. Also diusion of charge carriers out of the junction decreases eciency as the recombination volume gets larger. These problems can be solved by using the double hetero junction LED. The idea is that there are two junctions of two dierent materials with dierent band gap so that the material with a smaller band gap is placed between materials with a larger band gap. The material in between

14 CHAPTER 2. THEORY 10 Figure 2.7: pn-homojunction structure. A) Doping by diusion or ion implantation. B) Epitaxial growth. is kept thin. This creates an area with potential barriers in both sides which prevent the charge carriers from diusing away from the junction. Also the emitted light is not reabsorbed because of the larger band gap of the surrounding material and the photons are not strong enough to lift electrons to the conduction band. Figure 2.8: pn-heterojunction structure and the corresponding band structure Quantum well LED To get further connemet of charge carriers and better overlapping of electron and hole wave-functions a quantum well (QW) active area is used. To the

15 CHAPTER 2. THEORY 11 Figure 2.9: A) Practical structure of modern GaN-based LED. B) Simplied illustration of MQW LEDs energyband structure. active area one or more quantum well structures are grown with specic dimensions depending on application. Particle energies in quantum well are [3] E n = n 2 π 2 h 2 2m(2a) 2 V 0, n = 1, 2, 3... (2.7) where h is the Plank constant, a is the width of the well and V 0 is the well potential. Thus the energies are accurately dened in quantum well with certain dimensions and the recombination of charge carriers produces a narrow spectrum of light. The most used structure in practical applications is the multiple quantum well (MQW) where a stack of QWs is applied in active region. [5]

16 Chapter 3 Measurements and results 3.1 Current-voltage curve To measure the current-voltage curve a DC source and voltage meter are connected to the LED. First the current is set to a few hundred ma's with negative polarity and then increased with constant step size towards the forward current maximum. At every step the voltage over the diode is measured. From this data the current-voltage curve can easily be drawn. The measured I-V curve is shown in gure 3.1. From this picture it is easy to see that the current starts to ow eciently at about 3.2V. This is the voltage from which the LED starts to emit light. Energy of 3.2eV for a photon responds to 390nm wavelength which is approximately the peak wavelength of the UV-LED. The junction begins to conduct the charge carriers at 3.2V thus starting to emit light. At large enough negative bias the junction starts to leak current in reverse direction. This happens at about 2V reverse voltage. 3.2 Spectrum Current dependence The spectrum measurements of a LED are done with a monochromator based spectroradiometer Bentham DM150. The LED is connected to DC source and the current is set to be convenient for the used equipment. An optical bre is placed towards the LED to lead the emitted radiation to the monochromator. 12

17 CHAPTER 3. MEASUREMENTS AND RESULTS Current (ma) Current (ma) ,8 2,9 3,0 3,1 3,2 3,3 3,4 3,5 Voltage (V) Voltage (V) Figure 3.1: Current of the LED as function of applied voltage in room temperature. The C-V curve of the forward voltage is shown in the inset. Comparing the theoretical spectrum in gure 2.6 to the measured spectrum in gure 3.2 it is clear that the shapes are similar in measured wavelength area. However the measured LED has unidealities which defect the curve from the ideal case. At larger wavelengths in gure 3.2 inset the descenting curve is not linear but has a slightly bumpy behaviour which diers from the ideal case. This is due to recombination centers inside the forbidden energy gap which result to emission with larger wavelength. These unidealities originate from crystal defects. The peak wavelength shifts slightly towards larger wavelengths when the current is increased. Also the larger the current the wider the spectrum. This is consistent with the LED warming up with higher currents. The minimum wavelength stays approximately constant at increasing current which is obvious since the maximum recombination energy is dened by the constant energy gap width.

18 CHAPTER 3. MEASUREMENTS AND RESULTS ,5E-4 3E-4 Intensity (a.u.) ,1 0,01 1E-3 Intensity (a.u.) Wavelength (nm) 1E Wavelength (nm) 2,5E-4 2E-4 1,5E-4 1E-4 350mA 500mA 699mA Figure 3.2: Normalized spectrum of the UV-LED with 350, 500 and 700mA current in logaritmic scale. Intensity (a.u.) 1,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 Intensity (350mA) Intensity (500mA) Intensity (699mA) 0, Wavelength (nm) Figure 3.3: Normalized spectrum of the UV-LED with 350, 500 and 700mA current.

19 CHAPTER 3. MEASUREMENTS AND RESULTS Temperature dependence The eect of temperature change to light intensity and the peak wavelength of the emission spectrum can be measured with the same spetroradiometer setup as in previous section. Only dierence is that the LED is now placed on temperature-controlled plate. The temperature can be set for example to 40C and increased with steps of ten degrees towards the maximum safe value given by manufacturer. From the drawn spectrum the peak wavelength and intensity can easily be read. It was shown in theory chapter that the voltage of the LED should decrease with increasing temperature (equation 2.4). As shown in gure 3.4 this is the case. The descent angle of this curve is constant in this interval. The tanget calculated from the curve is approximately 2 mv which is exactly the value C given for 350mA forward current in the manufacturer datasheet. 3,20 3,12 Voltage (V) 3,04 2, Temperature (C) Figure 3.4: Graph representing the diode voltage as a function of temperature. According to equation 2.5 the emission intensity of LED should decrease with increasing temperature. This can be seen in gure 3.5. The value of wavelength in the spectrum with which the maximum intensity is gained is also slightly temperature dependent. Actually the peak seems to be shifting to higher wavelength as the temperature increases. This is to be expected as more recombinations with lower energy can occur with increasing lattice vibrations.

20 CHAPTER 3. MEASUREMENTS AND RESULTS Intensity 3200 Intensity (Abs) Temperature (C) Figure 3.5: Graph representing the diode spectrums peak intensity as a function of temperature Duty cycle dependence It is of interest to see how switching current aects the peak wavelength and bandwidth of the LED. For this the LED is again connected to the spectroradiometer but now the DC source is replaced with signal generator that can create rectangular waves. Also the ratio between voltage-up and voltage-down time can be changed in between 2-85%. A convenient frequency of about 1kHz is then applied to the LED with dierent pulse ratios and the required information is stored. The full width at half maximum (FWHM) value of the spectrum was measured at 350mA with 1kHz frequency. From gure 3.8 it can be seen that the FWHM value increases as the pulse ratio increases. The junction exposed to longer pulses of forward bias thus emits a slightly broader spectrum of light than a junction exposed to shorter pulses. The explanation is that the LED has more time to heat and less time to cool down with longer pulses thus breadening the spectrum. The emission spectrum peak wavelength changes as a function of pulse ratio. This can be seen in gure 3.9. As the pulse ratio decreases the peak shifts to lower wavelength. One possible explanation for this is that the piezoelectric elds inside the crystal have more time to take eect on the spectrum with higher pulse ratio and because these phenomenon are known to shift

21 CHAPTER 3. MEASUREMENTS AND RESULTS 17 Peak Wavelength 404 Peak Wavelength (nm) Temperature (C) Figure 3.6: Wavelength of LED spectrum peak as a function of temperature with 50mA current. Table 3.1: Comparison of peak wavelength of the LED spectrum with dierent pulse ratios and dierent frequencies. Pulse-Ratio Peak with 1000Hz Peak with 180Hz 2% nm nm 5% nm nm 24% nm nm the spectrum towards red this can explain this phenomenon atleast partly. This might best explain the increasing peak wavelength when pulseratio is less than 5%. The most probable explanation for increasing peak wavelength with increasing pulse ratio is simply that the LED is heated more because the average current increases with larger pulse-ratio. Heating decreases the value of E g thus lowering the emitted photon energy and increasing the wavelength. A few pulse ratio measurements were also done with 180Hz frequency for comparison. The used values for pulse ratio were 2, 5 and 24%. It can be seen from table 3.1 that the behaviour of the peak is similar with both frequencies. The peak has the lowest wavelength at 5% pulse ratio and increases in both directions.

22 CHAPTER 3. MEASUREMENTS AND RESULTS 18 Figure 3.7: Illustration of the signal used to determine the eect of pulse ratio in LED. 12,20 12,15 12,10 FWHM (nm) 12,05 12,00 11,95 11,90 11,85 11, Pulse Ratio (%) Figure 3.8: FWHM of the LED spectrum as a function of pulse-ratio.

23 CHAPTER 3. MEASUREMENTS AND RESULTS ,70 395,65 Peak Wavelength (nm) 395,60 395,55 395,50 395,45 395,40 395,35 0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 Pulse Ratio (%) Figure 3.9: Peak wavelength of the LED emission spectrum as a function of pulse-ratio.

24 Chapter 4 Conclusions The data gathered from the measurements of the UV-LED had the properties which could be expected to be obtained in light of theory and manufacturer specications. The behaviour of the dierent curves was similar to the simplied curves given by theory chapter although more detailed theory was neglected. Thus it can be assumed that the measurements were accurate and correct and the used measuring devices worked well. For measuring the current-voltage characteristics a simple setup of current source - voltage meter seemed to be precise enough. The spectrum measurements required better, more accurate equipment since a trial with more robust device left the spectrum curve rippled. Thus especially the intensitypeak wavelength was impossible to determine accurately without well calibrated monochromator-based measurements. There are much more interesting properties which could be determined from LED's. The dynamics of UV-LED could be studied more. Also many variations of the measurements already made could be done such as position of C-V curve as function of temperature and pulsed input bias with changing temperature. Unidealities could be determined such as parasitic capasitance and break down voltage. However this work was intented to provide a short introduction to LED theory and oers some insight on how to determine the characteristics of LED's which was a success. 20

25 Bibliography [1] (Author: unknown). Internet Source. [2] Bhattacharya, P. Semiconductor Optoelectronic Devices, second ed. Pearson Education Company, [3] Griffits, D. J. Introduction to Quantum Mechanics, international ed. Pearson Education International, [4] Prolight Opto Technology Corporation. ProLight PG1N-1LLX 1W UV Power LED. Data Sheet. [5] Schubert, E. F. Light-Emitting Diodes. Cambridge University Press,