27-29 September 2011, Paris, France

Characterization of solar cells by thermal transient testing

B. Plesz, Gy. Horvth, A. Vass-Vrnai
Budapest University of Technology and Economics, Department of Electron Devices H-1117 Magyar tudsok krt. 2, Budapest, Hungary E-mail: <plesz|horvath|vassv>@eet.bme.hu
Abstract- The current paper deals with the application of thermal transient testing as a characterization tool for solar modules. Based on the measurement of a representative sample a concentrator module - we prove the applicability of this measurement technique. Metrics such as junction-to-base plate thermal resistance are derived and can serve as a basis of a multi domain solar module model. The used technique also enables us to verify the quality of attachment layers in a solar module allowing fair quality control and reliability analysis of these devices.

I.

INTRODUCTION

Solar cells are one of the most prospering fields nowadays with yielding an average growth of over 30% per year over the last decade. Thanks to this superior growth R&D in this field has increased, and produced a wide variety of different solar cell concepts, e. g. crystalline silicon cells, amorphous silicon cells, semiconductor alloy multi-junction cells for concentrator applications, non-silicon thin film solar cells or organic solar cells. All of these cell types share a common feature, namely that they convert light into electric energy, but due to changes in the ambient conditions (such as temperature and irradiation) their behavior can be extremely varying. As an example, the power curves of crystalline silicon and organic solar cells are quite different: while the power generated by a silicon solar cell decreases with temperature [1,2], the temperature dependence of organic cells shows a positive coefficient [3]. Current standard measurements [4,5] do not incorporate the variation of temperature, although the power generated by photovoltaic devices shows a strong dependence on these parameters. Thus there is a strong need for measurement techniques that can be applied for combined electro thermal investigations. Characterization methods where the variation of the temperature is taken into account, but only steady states are investigated, can be summarized as static electrothermal methods. By extending todays characterization techniques with such testing methods more accurate models for PV structures can be derived. As for the interlink between electric and thermal behavior of solar cells it is characteristic that the electric response to changing irradiation conditions is by orders of magnitudes faster than the thermal response. Thus for example in case of rapidly changing weather conditions the characteristics measured with static electrothermal methods can result in

false predictions if a feedback on the cell temperature is not available. Since the cells are normally incorporated dust- and waterproof into the module, temperature sensors would have to be integrated during the fabrication process, if feedback on the cell temperature is needed. This would lead to additional costs and changes in manufacturing processes. It can be seen that prediction models based on static electrothermal characterization need three input parameters, namely irradiance, ambient temperature and cell temperature where the cell temperature is a feedback signal and thus does not allow explicit black box modeling. By using transient electro-thermal measurements and appropriate models, a more sophisticated description of device behavior can be given. Using thermal-transient measurement the heatflow path between the heat generation source (the cell) and the ambient can be measured. Contrary to static electrothermal characterization methods this way not only the thermal resistance, but the heat capacitances of the solar module can be measured additionally. If in addition the heat generation in the solar cells is determined from the absorbed irradiation and the ambient temperature is known, a precise prediction can be given at every single moment of operation. With models based on thermal transient measurements no cell temperature feedback is needed, only the irradiance level and the ambient temperature have to be introduced as input parameters to the model. In addition to precise complex solar module models thermal transient testing can also provide valuable data for the thermal design and reliability testing of solar cell modules e.g. in concentrator and space applications. Most of the space vehicles orbiting around the Earth (e.g.: satellites) have solar cells as primal source of energy. For this reason the functioning and thus the mission of these space vehicles depends on reliability of the solar cells. The most frequent reason why these solar cell modules fail or their efficiency significantly decreases is the delamination of the solar cell from the bearer. II. EXPERIMENTAL

For the accurate thermal modeling of an encapsulated solar module an approach similar to those used for the characterization of power electronics can be applied. Thermal transient testing is an appropriate tool for

EDA Publishing/THERMINIC 2011

ISBN:

27-29 September 2011, Paris, France mapping the heat conduction path between a semiconductor value to convert the measured voltage data to temperature junction and its ambient. This technique is widely used to data. derive package parameters such RthJA and RthJC in case of This number indicates that the tested device was a triple power devices. In these cases the temperature sensor can be junction cell. a dedicated temperature sensitive diode or a transistor. Calibration curve Since a solar cell is basically a series of PN junctions with 2.6 a temperature sensitive forward voltage, the thermal Recorded points 2.55 transient methodology is suitable for their testing in Regression curve practically the same way [6]. 2.5 However there are some differences which should be 2.45 considered in case of the measurement of solar cells. 2.4 The packaging of solar modules is not primarily designed 2.35 to enhance the cooling path from the junction, but for transmittance at the upper side and mechanical stability at y = -0.0062x + 2.7132 2.3 the backside. For this reason the heat paths in both directions 2.25 are comparable. Concentrator cells are an exception as due 20 30 40 50 60 70 to the high power density of the irradiation additional Temperature (oC) measures have to be taken to increase the thermal conductivity between the cell and the ambient. Fig. 2. The calibration curve. In addition the current of the solar cell varies with illumination, so it is essential to perform the temperature In the actual test arrangement we used the modification of sensitivity calibration under dark conditions. For practical the so-called transient dual interface methodology [7] to considerations we performed the measurements in a dark identify the thermal resistance from the junction to the environment, too. bottom of the base-plate. In order to follow the standard the The silicon and other structural materials such as the glass measurements were carried out at two different boundary and aluminum are easy to incorporate into models due to conditions. First good thermal contact was established their well known material properties and geometries. The between the MCPCB and a temperature stabilized cold plate thermal resistance and capacitance of attachment materials using heat conducting grease. As a second step the grease between these layers are on the other hand difficult to was removed so the thermal resistance between the MCPCB calculate as their parameters are usually not well defined. and the cold plate increased. With the measured two For the actual tests we used a GaAs concentrator solar different boundary conditions we can identify the divergence cell with a 10x10 mm surface area, mounted on an aluminum point of the curves, and read the characteristic thermal based MCPCB. The structure shown in Fig. 1. below is resistance values. similar to those used in case of power LED-s, therefore we may assume that the heat-conduction path is mainly one dimensional in this particular case.
Voltage (V)

Fig. 1. Structure diagram.

The temperature sensitive parameter (TSP) value has to be calibrated in case of each device under test to measure the temperature dependence of the forward voltage of the diode. For device calibration we prefer to use a temperature controlled environment to set up different ambient temperature values. A constant sensor current is driven through the diode, while we measure the forward voltage values at each settled temperature point. With trial measurements we selected 10 mA constant sensor current for biasing the diode, to obtain an optimal signal-to-noise ratio. The temperature of the reference plate was controlled from 25 to 65 C in 10 C steps, and the forward voltages were measured. The calibration curve can be viewed on Fig. 2. Based on the slope of the curve, we used -6.198 mV/K

Fig. 3. Smoothed response.

The temperature transient curves corresponding to these two boundary conditions are shown in Fig. 3. You can note that until one time point, approximately 0.1 second both functions run together, indicating that the main trajectories of the heat flow spread in the same structure. If a TIM material is used, the junction temperature elevation is 4.9 C above the cold-plate temperature. If no TIM material

EDA Publishing/THERMINIC 2011

ISBN:

27-29 September 2011, Paris, France The flat section between point 1 and 2 represents the is used, the temperature elevation is 2.7 C higher, i.e. attachment layer as it shows a significant increase in its 7.6 C. thermal resistance at a very low capacitance elevation. The Fig. 4. shows the cumulative structure functions of the next step should represent the copper layer. In order to prove curves belonging to the two different boundary conditions. this we determined the actual volume of the copper and We can identify the point of separation at about 3.14 K/W compared it to the measurement result. The measured value which is the junction-to-MCPCB thermal resistance of the was 1.47 mm3 while the calculated volume based on the real structure. The total thermal capacitance is about 1.4 Ws/K. geometry (0.035 mm thick copper) was 1.4 mm3. This simple calculation proved that the region between points 1 and 2 is the die attach layer. As we can accurately determine the resistance of the DA layer in a module, we assume that variations and changes of its resistance can be measured too. This results in a method for testing and characterization of the quality of the die attach and other interface materials in the module. In order to prove our assumption we have created some devices with artificially altered die attach resistance to simulate possible die attach failures. The samples used to simulate delamination of solar cells were processed from commercially available single junction solar cells. The cells were diced into pieces of 30 x 20 mm sample size, and were fixed on a copper plate with an ordinary TIM material containing silver balls. The TIM was applied in a thickness of 500 um. To alter the thermal resistance of the die attach spacers cut from 500 m thick borosilicate glass wafers were used, see Fig. 6. For the investigations three samples were processed. As a reference one cell was attached directly on the copper plate without any TIM. The remaining two samples were processed as described above, with different ratios between the TIM and the glass area. The two samples had a glass to TIM ratio of 0 (grease only) and 1:3 (25 % of the solar cell area) percent respectively. This test represents the delamination of the solar cells in different scales.

Fig. 4. Comparison of the structure functions corresponding to the different boundary conditions.

We tried to find the correspondence between the layers of the solar receiver structure shown in Fig. 1. and the different parts of the structure functions. The real dimensions of the chip are (based on our measurements) 5.67x6.88x0.18 mm, giving a total volume of 7.02 mm3. In comparison the volume of the chip calculated based on the thermal capacitance value identified from the early, steep part of the structure function and the specific heat of GaAs was 7.16mm3. (see point 1 in Fig. 5.) The approximate Rth of the TIM layer can be measured with the help of the cumulative structure function, too as it can be seen in Fig. 5.

Fig. 6. Structure of the sample simulating DA problems

Fig. 5. Rth of the TIM layer.

The resulting structure functions can be viewed in Fig. 7. The silicon volume is identical in all three cases, and equals to 150 mm3. As the heat-spreading is ideally one dimensional in this case this value can be precisely read from the structure functions. The reference sample and the sample having 100% TIM coverage behave in a similarly, but the reference shows slightly better performance. The sample with 25% glass coverage and 75% TIM coverage shows a definite increase in its die attach resistance. The change in the die attach resistance can be easily observed by comparing thermal resistance values corresponding to a characteristic thermal capacitance in the

EDA Publishing/THERMINIC 2011

ISBN:

27-29 September 2011, Paris, France APPLIED PHYSICS, 90 (10): 5343-5350 NOV 15 2001, ISSN: system. In our example the 10 Joule/K value is characteristic 0021-8979 to the copper base plate. The curve corresponding to the 25% [4] IEC 60904-1:2006, Photovoltaic devices, Part 1: Measuremenet of glass coverage reaches this value approximately 0.1 K/W photovoltaic current-voltage characterstics later than the sample with total TIM coverage. [5] IEC 60904-8:1998, Photovoltaic devices, Part 8: Measurement of spectral response of a photovoltaic (PV) devices The overall thermal resistance of this system is very low, [6] Siegal, B.; , "Solar Photovoltaic Cell thermal measurement issues," and the thermal conductivity difference between the glass Semiconductor Thermal Measurement and Management and the TIM is approximately 1:4. The fact that despite all Symposium, 2010. SEMI-THERM 2010. 26th Annual IEEE , vol., these the difference can be well measured shows the good no., pp.132-135, 21-25 Feb. 2010 [7] D. Schweitzer, The junction-to-case thermal resistance: A resolution of the presented method.
T3Ster Master: cumulative structure function(s) 25%glass_6A_5mA - Ch. 0 100%TIM_6A_5mA - Ch. 0 Reference_6A_5mA - Ch. 0

boundary condition dependent thermal metric, 26th Annual IEEE SEMI-THERM Symposium, March 2010, San Jose, CA,USA, pp. 151

1000 100

Cth [Ws/K]

10 1 0.1

Increased DA resistance 150 mm Silicon

3

0.01 0.001

0.1

0.2

0.3 Rth [K/W]

0.4

0.5

0.6

Fig. 6. Structure functions corresponding to the solar cell models

III.

CONCLUSIONS

We demonstrated that thermal transient testing is applicable to the characterization of solar modules. As it was shown thermal transient measurements performed on solar cells can provide cell-to-base plate thermal resistance metrics. This can be used as a base for multi-domain solar module models providing precise predictions for a broad spectrum of operating conditions. In addition the inner structural layers can be characterized as shown in the case of the die attach in the discussed test arrangement. Due to this thermal transient testing can be applied to e.g. quality management of fully assembled modules or module testing during operation. Such a test can be essential in cost- and reliability sensitive applications such as space vehicles or devices operating under harsh environmental conditions.
REFERENCES [1] [2] Green, M. A. (2003), General temperature dependence of solar cell performance and implications for device modelling. Progress in Photovoltaics: Research and Applications, 11: 333340. Temperature Dependence of Protocrystalline Silicon/Microcrystalline Silicon Double-Junction Solar Cells, Kobsak Sriprapha, Seung Yeop Myong, Akira Yamada and Makoto Konagai, Jpn. J. Appl. Phys. 47 (2008) 1496 Katz, EA; Faiman, D; Tuladhar, SM; Kroon, JM; Wienk, MM; Fromherz, T; Padinger, F; Brabec, CJ; Sariciftci, NS: Temperature dependence for the photovoltaic device parameters of polymerfullerene solar cells under operating conditions, JOURNAL OF

[3]

EDA Publishing/THERMINIC 2011

ISBN: