Extraction of built-in potential of organic diodes from current-voltage characteristics
EExtraction of built-in potential of organic diodes fromcurrent-voltage characteristics
Prashanth Kumar M, Saranya R, Soumya Dutta a Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai 600036, India
Abstract
Physics based analytic equations for charge carrier profile and current density are de-rived by solving the carrier transport and the continuity equations for metal-intrinsicsemiconductor-metal diodes. Using the analytic models a physics based method is devel-oped to extract the built-in potential V bi from current density-voltage ( J - V ) character-istics. The proposed method is thoroughly validated using numerical simulation results.After verifying the applicability of the proposed theory on experimentally fabricated or-ganic diodes, V bi is extracted using the present method showing a good agreement withthe reported value. Keywords:
Organic diode, Organic solar cell, Device simulation, Built-in potential,Injection limited current, Space charge limited current
1. Introduction
In organic diodes and solar cells, organic semiconductors are typically sandwichedbetween two dissimilar metal electrodes. The electric field and the charge profile underequilibrium are governed by the metal contacts of these devices. Moreover, the workfunction difference of the electrodes determines V bi [1], which is a crucial metric forboth organic diode and solar cells as far as device performance is concerned. However,the charges that get injected from the contacts cause an excess potential drop near themetal-semiconductor junctions [2]. This excess drop in potential leads to a reduction of V bi [3] to a lower value, say V (cid:48) bi , which is typically measured instead of V bi for a giventemperature T .In this work we propose a physics-based analytical model for estimating V (cid:48) bi from J - V characteristics which takes the effect of injected charge into account. Our model is furtherextended to establish a method for extracting V bi by using T dependency of V (cid:48) bi . Finally,the method is employed to estimate V bi of poly(3-hexylthiophene) (P3HT):phenyl-C61-butyric acid methyl ester (PCBM) based organic diodes, which are fabricated in ourlaboratory. ∗ Corresponding author
Email address: [email protected] (Soumya Dutta)
Preprint submitted to . May 19, 2017 a r X i v : . [ c ond - m a t . o t h e r] M a y . Simulation and analytical results E g C a t h o d e f f f f V b i LUM OHOM O
A n o d e d
Figure 1: A schematic of equilibrium energy-band diagram of an organic diode, with organic semiconduc-tor of thickness d and bandgap E g , where LUMO and HOMO being the lowest unoccupied and highestoccupied molecular orbitals, respectively. We use the Metal-Insulator-Metal methodology for numerical simulation [4, 5, 6]. Thenumerical simulations are done using the commercially available the Sentaurus tech-nology computer-aided design (TCAD) tool [7]. In numerical simulations, we considerSchottky contacts at anode and cathode with barriers φ ( φ ), φ ( φ ) for electrons(holes) respectively [Fig. 1].Therefore the carrier concentrations at the contacts are determined by thermionicemission process and are given as n = N C exp (cid:16) − φ qV t (cid:17) , n d = N C exp (cid:16) − φ qV t (cid:17) ,p = N V exp (cid:16) − φ qV t (cid:17) , p d = N V exp (cid:16) − φ qV t (cid:17) , (1)where V t is the thermal voltage, n ( p ), n d ( p d ) are the electron (hole) concentration atanode and cathode respectively, N C ( N V ) is the effective density of states for electrons(holes). At equilibrium (applied voltage V = 0 V ) the dissimilar metal work-functions arealigned leading to band bending [Fig. 1] which sets up a built-in electric field insidethe device. The strength of the built-in electric field depends on V bi , d and the injectedcharge. Depending on the magnitude of the injected charge and its effect on the electricfield, the diodes can be classified into two categories: (1) Low space charge (LSC) case.(2) High space charge (HSC) case. The injected charge can be modified by changing N C ( N V ), barrier for electrons (holes) and the temperature. In this particular study we keep N C ( N V ) unchanged and vary the barriers for electrons (holes) and the temperature toexplain LSC and HSC cases. The parameters associated with the simulation for LSCand HSC cases are given in Fig. 2. 2 . 0 0 0 . 0 3 0 . 0 6 0 . 0 9- 8 0- 6 0- 4 0- 2 002 04 0- 5 . 5- 5 . 0- 4 . 5- 4 . 0- 3 . 5- 3 . 0 ( b ) 0 . 6 5 V 0 . 9 5 V 0 V 0 . 3 5 V Electric field (kV cm-1)
D i s t a n c e f r o m a n o d e ( m m )
0 V 0 . 6 5 V 0 . 9 5 V
H O M O
Energy (eV)
L U M O ( a )
Figure 2: (a) The band diagram for LSC case for different V (TCAD results), (b) Electric field profileof LSC (lines) and HSC (symbols) cases for different V . The parameters used for the simulation of LSCcase are E g = 1 . φ = 1 . V bi = 0 . µ n = µ p = 1 × − cm /Vs, N C = N V = 1 × cm − , ε = 3.3 ε , d = 100 nm and T = 300 K. HSC case is resembled by changing φ to 1.15 eV andkeeping all other parameters same as that of LSC. In LSC case, the field due to the injected charge is very less compared to the electricfield generated due to work-function difference. Hence the electric field is expected tobe uniform by maintaining linear band bending [Fig. 2(b)] inside the device [Fig. 2(a)].The injected carriers (from metals) undergo diffusion and drift concurrently in oppositedirection to each other, as a consequence of concentration gradient and electric fieldrespectively. Thus in order to model the carrier profiles and the current density, both driftand diffusion have to be considered simultaneously. The transport equation, describingelectron current density, can be expressed as J n = qnµ n E + qµ n V t ∂n∂x , (2)where q is the electron charge, E ( x ) is the electric field, n ( x ) is the electron carrierconcentration and µ n is the mobility of electron. As discussed above, E ( x ) is uniform3nd it is represented as E ( x ) = − ( V bi − V ) d . (3)In order to arrive at analytic solution under steady state conditions, we consider threeassumptions: (1) Semiconductor is intrinsic, (2) Carrier mobilities ( µ n and hole mobility µ p ) are constant with respect to V and T , (3) There is no carrier generation andrecombination. The last assumption modifies the continuity equation for electrons as ∂J n ( x ) ∂x = 0 . (4)Using Eqs. (2), (3) and (4), a second order differential equation is developed for n ( x ) as ∂ n ( x ) ∂x + E ( x ) V t ∂n ( x ) ∂x = 0 . (5)By employing the thermionic emission boundary condition for electrons [Eq. 1], ananalytic solution for n ( x ) is obtained as n ( x ) = n d − n exp (cid:16) V bi − VV t (cid:17) + ( n − n d ) exp (cid:16) V bi − VV t xd (cid:17) − exp (cid:16) V bi − VV t (cid:17) . (6)Similar approach can be used to obtain an analytic solution for holes. D i s t a n c e f r o m a n o d e ( m m ) Electron density (1013 cm-3)
Figure 3: Electron profile for different V as annotated, using TCAD simulation (symbols) and Eq. 6(solid lines). The extracted electron profile inside the device using TCAD simulation (symbols) andEq. 6 (solid lines) for different V are compared in Fig. 3 which ensures that Eq. 6 isin good agreement with the TCAD results. Further, the analytic solution for current4ensity can be expressed using Eqs. (2), (3), (6) and their hole counterparts as J = q ( µ n n + µ p p d ) ( V bi − V ) (cid:104) exp (cid:16) VV t (cid:17) − (cid:105) d (cid:20) − exp (cid:18) − V bi − VV t (cid:19)(cid:21) . (7)The variation of J with respect to V , based on TCAD simulation (symbols) and Eq. 7 - 0 . 4 0 . 0 0 . 4 0 . 8 1 . 21 0 - 1 1 - 8 - 5 - 2 Current density (mA cm-2)
V o l t a g e ( V ) ( c )
T C A D A n a l y t i c a l
Current density (mA cm-2) V b i Figure 4: J - V characteristics for LSC case where the symbols are TCAD and solid lines are model (Eq.7). (solid line) are displayed in Fig. 4, showing excellent consistency. A similar equation hasbeen reported by different groups in the literature [4, 8]. Where S Jung et al. arrived ata similar analytical equation for less disordered organic materials with Gaussian densityof states. However, the present method is completely rest upon charge based model withcoherent device physics considering the effective density of states.According to Fig. 3, charge increases exponentially from anode to cathode for V < V bi (0.7 V in particular). The exponential nature of charge along with linear variation of E ( x ) with V results in exponential variation of J with respect to V . On the otherhand, for V > V bi , the charge carrier profile changes significantly [Fig. 3] by virtue offield reversal [Fig. 2]. Charge carrier concentration increases from anode to cathodelike a logistic function [Fig. 3], maintaining its spatially uniform nature inside the deviceexcept near the anode-semiconductor junction. The uniform nature of both charge carrierconcentration and electric field profiles leads to a linear variation of current. In LSC case,the current is typically injection limited current (ILC) since dominant part of the currentis controlled by the injected charge carriers. In HSC case, the electric field due to the injected charge becomes comparable to theelectric field associated with band bending, expressed by Eq. 3. Hence the net electricfield within the device becomes non-uniform. HSC case can be realized by reducingthe barrier height for electrons (holes) or by increasing N C ( N V ) or by increasing thethickness of the semiconductor. However, in this study HSC is realized by reducing5 - 1 2 - 9 - 6 - 3 ( d ) T C A D A n a l y t i c a l
Current density (mA cm-2)
V o l t a g e ( V ) J L i n e a r f i t
Current density1/2 [ (mA cm-2)1/2 ] V ' b i Figure 5: J - V characteristics for HSC case, where the symbols are TCAD and solid lines are model (Eq.9). the barrier height at anode-semiconductor junction in particular. Under equilibrium, auniform electric field is observed within the device except near the anode-semiconductorjunction where injected charge is high [Fig. 2(b)]. However, for V < V bi the magnitudeof uniform electric field is slightly less than that of LSC case. Thus, J - V characteristicsmaintain the same exponential nature as that of LSC case, exhibiting a reduction inbuilt-in potential [Fig. 5]. Hence it is essential to modify the electric field in case ofHSC by reducing V bi to V bi − φ (i.e., V (cid:48) bi ) where φ accounts for the reduction in E ( x )due to the injected charge. However, in case of V > V (cid:48) bi , the non-linearity in electricfield profile near the anode-semiconductor junction becomes predominant upon applyingvoltage and spreads throughout the device differing drastically from that of LSC case.As a consequence, the electric field and the carrier concentration become interdependent,leading to non-linear J - V characteristics. The current density varies with square of V (for V > V (cid:48) bi ) as evidenced by a linear nature of √ J -V characteristics [Fig. 5]. As thecurrent is controlled by the space charge, it is space charge limited current (SCLC).For V < V (cid:48) bi , in the uniform electric field region, the electric field strength can bemodeled as E ( x ) = − ( V (cid:48) bi − V ) d . (8)Using Eq. 8, Eq. 7 can be modified as J = q ( µ n n + µ p p d ) ( V (cid:48) bi − V ) (cid:20) exp (cid:18) VV t (cid:19) − (cid:21) d (cid:20) − exp (cid:18) − V (cid:48) bi − VV t (cid:19)(cid:21) . (9) J - V characteristic using Eq. (9) shows a good agreement with TCAD results under V < V (cid:48) bi for φ = 0.0544 V [Fig. 5(d)], where φ is obtained by fitting TCAD results withEq. (9). 6 .2. Extraction of Built-in potential In LSC case, J changes its nature from exponential to linear for V > V bi , whereas inHSC case, J changes its nature from ILC to SCLC for V > V (cid:48) bi . Most of the practicalorganic diodes belong to HSC case. To understand more about current transition fromexponential to linear or ILC to SCLC, we adopt a function G , proposed by Mantri et al.[9], where G is defined as G = ∂ ln( J ) ∂ ln( V ) . (10)The variation of G with respect to V shows three distinct regions signifying three different V (cid:1) G V o l t a g e ( V )
T C A D ( e
V / V t - 1 ) ( V (cid:2) - V ) m R e g i o n 1 R e g i o n 2 R e g i o n 3
Figure 6: G - V characteristics for LSC. V (cid:1) (V) T e m p e r a t u r e ( K ) V b i Figure 7: V α variation with temperature for different V bi with φ = 0 .
45 eV (TCAD results). nature of current [Fig. 6]. In region-1, the variation of G can be fitted using a simpleexponential function of V as [exp( V /V t ) − follows a power law as ( V γ − V ) m with an exponent m, where V γ and m are fittingparameters. Transition between these two regions (1 and 3) occurs through region-2,showing a combined effect of exponential and power law. Moreover, in region-2, G - V characteristics exhibit a peak at a voltage, termed as V α . TCAD simulation results forthe variation of V α with respect to temperature as a function of different V bi is depictedin Fig. 7. It emphasizes that modeling V α variation with V bi will help in determining V bi from J - V characteristics. Using Eq. (9) and Eq. (10), a unified expression for G isdeveloped as G = VV t (cid:20) − exp (cid:18) − V (cid:48) bi − VV t (cid:19)(cid:21) + VV − V (cid:48) bi . (11)Eq. (11) consists of two terms which correspond to two different nature of current. First G V o l t a g e ( V )
Figure 8: G - V characteristics for LSC and HSC cases, where the symbols are TCAD and solid lines aremodel (Eq. 11). term shows exponential nature, whereas the second term represents a power law with V γ = V (cid:48) bi and m = 1. Eq. (11) shows an excellent match with the TCAD results for LSCcase, where V (cid:48) bi = V bi [Fig. 8]. In case of HSC, there is indeed a good agreement betweenEq. (11) and TCAD results throughout regions 1 and 2. However, G - V characteristicsdeviate from Eq. (11) in region-3 due to the presence of SCLC, which cannot be capturedby the present model. Thus Eq. 11 is in good agreement with the TCAD results forboth LSC and HSC cases throughout region-1 and region-2. Hence Eq. (11) can be usedfor extracting V α for both LSC and HSC cases. Using Eq. (11) and equating its firstderivative to zero at V = V α , we obtain V (cid:48) bi V t − ( V (cid:48) bi − V α ) (cid:20) V α V t exp (cid:18) − V (cid:48) bi − V α V t (cid:19)(cid:21) = 0 . (12)Eq. (12) can be solved numerically to get V (cid:48) bi . For higher values of V bi , the second terminside the square brackets of Eq. (12) can be neglected. Therefore, a compact analytical8quation is realized for V α as V α = V (cid:48) bi − ( V (cid:48) bi V t ) / . (13)Subsequently V (cid:48) bi is calculated as V (cid:48) bi = (cid:18) √ V t + √ V t + 4 V α (cid:19) . (14)Using Eq. (14) and the value of V α (extracted from G - V characteristics), V (cid:48) bi can be cal- V b i V' bi (V)
T e m p e r a t u r e ( K )
Figure 9: V (cid:48) bi variation with respect to T for different V bi with φ = 0 .
45 eV, where the symbols areTCAD and solid lines are model (Eq. 14). (cid:1)
Vbi (V) (cid:1) ( e V ) Figure 10: Variation of extracted V bi and φ with respect to φ , for different φ , where the symbols areTCAD and solid lines are model (Eq. 16). V (cid:48) bi increases with the decrease in T and saturatesto V bi . The variation of V (cid:48) bi with respect to T arises due to φ , which in turn depends ondominant injected charge near metal-semiconductor junction [ p = N V exp( − φ / ( kT ))] and thereby on T . Upon decreasing temperature, the amount of injected charge de-creases, resulting φ tends to be zero and hence V (cid:48) bi approaches to V bi . φ can be calculatedfrom the relation φ = V bi − V (cid:48) bi for different T and V bi . From the variation of φ withrespect to T , a semi-empirical model is developed for φ as φ = kT ln qεN V exp (cid:16) − φ kT (cid:17) rC g V t + 1 , (15)where C g = ε/d , ε is the dielectric constant and r is a fitting parameter being independentof T . Using Eq. (15), V (cid:48) bi can be written as V (cid:48) bi = V bi − kT ln qεN V exp (cid:16) − φ kT (cid:17) rC g V t + 1 . (16) V bi and φ can be obtained by solving Eq. (16) self-consistently with T dependentvariation of V (cid:48) bi . The extracted parameters are in good agreement with TCAD resultsand it is validated for different combinations of φ and φ , which shows the robustnessof our model [Fig. 10].
3. Experimental results
In order to validate our model, V bi is extracted from experimental results of the organicsolar cell, fabricated in our laboratory. Organic solar cells consisting of P3HT:PCBMas active material with aluminum (Al) as cathode and indium tin oxide (ITO)/poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) as anode were fabricatedinside a nitrogen glove box and characterized in a vacuum probe station. In order tovalidate the versatility of our model, we used three different thicknesses, 173 nm (DeviceA), 154 nm (Device B) and 106 nm (Device C) of P3HT:PCBM, resulting from the spinspeed of 850 rpm, 1000 rpm and 1500 rpm respectively. J - V characteristics of device Aas a function of temperature is plotted in semilogarithmic scale and linear scale, as shownin Fig. 11. It is observed that the forward current increases with increase in temperatureas expected. 10 . 0 0 . 3 0 . 61 0 - 3 - 1
51 01 5 2 7 0 K 2 9 0 K 3 1 0 K 3 3 0 K 3 5 0 K
Current density (mA cm-2)Current density (mA cm-2)
V o l t a g e ( V )
Figure 11: Experimental J - V characteristics for device A. ln( J ) V t ( e V ) - 1 Figure 12: ln( J ) variation with 1 /V t for different applied voltages, where symbols are experimental andsolid lines are linear fit to the experimental data for device A. - 0 . 4- 0 . 3- 0 . 2- 0 . 1 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1 . 0 D e v i c e A D e v i c e B D e v i c e C
V o l t a g e ( V ) S Figure 13: Variation of S with applied voltage for different P3HT:PCBM spin speeds, where symbolsare experimental and solid lines are linear fit to the experimental data. .1. Model validation for experimental results To check whether the experimental results follow the proposed model, we have toanalyze the ln ( J ) variation with 1 /V t for different V . For 0 < V < V (cid:48) bi , by considering µ n n > µ p p d and using Eq. 1 and Eq. 9 ln ( J ) can be written asln ( J ) = ln (cid:20) qµ n N C ( V (cid:48) bi − V ) d (cid:21) + SV t , (17)where S = V − φ η . (18)According to the proposed model ln ( J ) varies linearly with 1 /V t with a slope S beingdependent on V . According to Eq. (18), S varies linearly with V having a slope (1 /η )equal to one and the intercept gives the value of one of the barrier potential ( φ ). Table 1: The extracted parameters for P3HT:PCBM diode with different thickness.
Device η φ / V bi φ / (eV) (V) (eV)A 0.9903 0.8087 0.660 0.2888B 1.006 0.7826 0.661 0.2871C 1.004 0.7779 0.654 0.2795The experimental variation of ln( J ) for Device A (symbols) is shown in Fig. 12 andit confirm the linear variation of ln( J ) with 1 /V t for different applied voltages. Hencethe experimental results are fitted with linear variation to find S and the intercept andthis study is extended for Devices B and C. The variation of S with V is shown in Fig.13, one can notice from the figure that for 0 . < V < . S varies linearly with V .Moreover the linear fit in that particular regime gives a slope which is nearly equal toone (Table 1), which is in consistent with the proposed model. Hence this confirm theapplicability of the proposed model on these experimental results. In addition we extractone of the barrier potential φ (or φ ) for different devices which is nearly equal to 0 . . 4 0 . 6 0 . 8 1 . 024681 01 21 4 2 7 0 2 9 0 3 1 0 3 3 0 3 5 0 G V o l t a g e ( V )
Figure 14: Experimental G - V characteristics for device A. V (cid:1) (V) T e m p e r a t u r e ( K )
D e v i c e A D e v i c e B D e v i c e C
Figure 15: Experimental variation of V α with temperature for devices with different P3HT:PCBMthickness. Using Eq. (10), we obtain G - V plot for Device A, as shown in Fig. 14. V α is extractedfrom the peak position of G - V plot for different temperatures and shown in Fig. 15 fordifferent devices. Subsequently, V (cid:48) bi , calculated using Eq. (12) for devices A,B and C,are shown in Fig. 16 with symbols. As explained earlier, by solving Eq. (16) in coherentmanner, V bi and φ are extracted and presented in the Table 1.13 . 6 20 . 6 40 . 6 6 2 8 0 3 0 0 3 2 0 3 4 0 3 6 00 . 6 20 . 6 40 . 6 20 . 6 40 . 6 6 V'bi (V)
D e v i c e A M o d e l ( E q . 1 6 ) ( a )( c )
V'bi (V)
T e m p e r a t u r e ( K )
D e v i c e C M o d e l ( E q . 1 6 ) ( b )
V'bi (V)
D e v i c e B M o d e l ( E q . 1 6 )
Figure 16: Experimental variation of V (cid:48) bi with temperature for devices with different P3HT:PCBMthickness, where symbols are experimental and the solid lines are model. It is important to note that the extracted values of V bi are almost same for differentthickness. Hence the V bi obtained using the present model is independent of thickness, asexpected. Moreover the extracted values of V bi are in consistent with reported values forP3HT:PCBM device [10], which validates our model and ensures the method of extractingbuilt-in potential from J - V characteristics of organic diode or solar cell.
4. Conclusion
In summary, we developed analytic models for injected charge profile, J - V charac-teristics and V (cid:48) bi . V bi is estimated using temperature dependent variation of V (cid:48) bi . Theextracted values of V bi are in good agreement with TCAD results. The extracted V bi Acknowledgements
The authors would like to acknowledge Department of Science & Technology (DST,Govt. of India), Nissan and IIT Madras for financial support and Centre for NEMS andNanophotonics (CNNP) for providing fabrication facility.
References [1] G. G. Malliaras, J. R. Salem, P. J. Brock, and C. Scott, “Electrical characteristics and efficiency ofsingle-layer organic light-emitting diodes,”
Phys. Rev. B , vol. 58, pp. R13411–R13414, Nov 1998.[2] J. Simmons, “Theory of metallic contacts on high resistivity solidsi. shallow traps,”
J. Phys. Chem.Sol , vol. 32, no. 8, pp. 1987 – 1999, 1971.[3] M. Kemerink, J. M. Kramer, H. H. P. Gommans, and R. A. J. Janssen, “Temperature-dependentbuilt-in potential in organic semiconductor devices,”
Applied Physics Letters , vol. 88, no. 19,p. 192108, 2006.[4] P. de Bruyn, A. H. P. van Rest, G. A. H. Wetzelaer, D. M. de Leeuw, and P. W. M. Blom, “Diffusion-limited current in organic metal-insulator-metal diodes,”
Phys. Rev. Lett. , vol. 111, p. 186801, Oct2013.[5] G. A. H. Wetzelaer and P. W. M. Blom, “Ohmic current in organic metal-insulator-metal diodesrevisited,”
Phys. Rev. B , vol. 89, p. 241201, Jun 2014.[6] L. J. A. Koster, E. C. P. Smits, V. D. Mihailetchi, and P. W. M. Blom, “Device model for theoperation of polymer/fullerene bulk heterojunction solar cells,”
Phys. Rev. B , vol. 72, p. 085205,Aug 2005.[7] . Sentaurus TCAD Release H-2013.03 Ed.Synopsys Mountain View CA USA[8] S. Jung, C.-H. Kim, Y. Bonnassieux, and G. Horowitz, “Injection barrier at metal/organic semicon-ductor junctions with a gaussian density-of-states,”
Journal of Physics D: Applied Physics , vol. 48,no. 39, p. 395103, 2015.[9] P. Mantri, S. Rizvi, and B. Mazhari, “Estimation of built-in voltage from steady-state currentvoltagecharacteristics of organic diodes,”
Organic Electronics , vol. 14, no. 8, pp. 2034 – 2038, 2013.[10] M. Mingebach, C. Deibel, and V. Dyakonov
Phys. Rev. B , vol. 84, p. 153201, Oct 2011., vol. 84, p. 153201, Oct 2011.