Measurement methods for the spatially resolved charac- charac-terization of diode breakdown

Diode breakdown comes along with several physical interactions within the pn-junction, which can be used for the spatially resolved characterization of diode breakdown behav-ior in solar cells. More specifically, two ramifications are useful:

A) Avalanche breakdown, localized within microplasma channels, emits bright light (section 6.1.1). Internal field emission itself is a non-radiative transition, which, however, is often followed at slightly higher reverse biases by a multiplication process which causes visible light emission.

B) In breakdown sites, a reverse current is transported across the pn-junction. This causes the dissipation of heat due to Joule heating.

6.2.1 Measurements based on an electroluminescence setup

Light in the visible spectral range can be imaged with the help of silicon charge coupled device cameras usually used in electroluminescence (EL) measurement setups. On 156 x 156 mm² solar cells for example, a good resolution in the order of 100 µm is ob-tained which allows e.g. for a good comparison between breakdown sites and sites of increased recombination activity. In the EL setup used at the Fraunhofer ISE, the solar cell is connected to a voltage generator being able to supply up to 20 V via two or three rows of pins brought onto the busbars.

In order to gain more insight into the microscopic reasons for the local breakdown, the spatially resolved determination of the local breakdown voltage is of great interest. Mak-ing use of the very short measurement times of the EL setup (in the order of 10 s or less

per image), we achieved local breakdown maps with EL images taken at consecutively increasing reverse voltages. The measurements were fed into an image processing algo-rithm which recognizes the position of breakdown light emitting sites. For this, an inten-sity threshold needs to be defined in order to distinguish between breakdown spots and background noise. Breakdown was detected when the intensity was larger than

<I> + 3σ, <I> being the average background noise level of a dark image and σ denoting its standard deviation. Finally, the algorithm assigned to each camera pixel the lowest reverse voltage at which BD light was detected, which corresponds to the definition of the local breakdown voltage used in this thesis (formulated in section 6.3.2).

In this approach, the local breakdown voltage is systematically overestimated; this error is the larger the softer the local breakdown characteristic because of the shallow slope.

Moreover, due to refraction and reflection at textured surfaces, the spots appear much larger to the EL camera (some hundred µm) than they are in reality (well below 5 µm in diameter, see Figure 6.17), depending on the surface treatment of the wafer. Both error sources should nevertheless allow for comparison of the local breakdown voltage since the same analytic procedure is applied to each measurement.

As a result, breakdown voltage maps of the solar cells were generated. Alternatively, the fraction of the breakdown light emitting area versus the reverse voltage was estimated.

This was done by calculating the ratio between the number of camera pixels which show BD light emission according to the above definition and the total pixel number covering the solar cell area.

For high resolution as well as for spectral measurements, a dedicated PL- / EL-measurement tool comprising a confocal microscope was used, equipped with a Si cam-era for microscopic images. The microscope is coupled via glass fibers and a 600 G/mm grating to a U-VIS spectrometer (Si line camera) for the spectral wavelength range be-tween 400 – 1000 nm or alternatively via a 150 G/mm grating to an IR spectrometer (InGaAs line camera) for the spectral range between 900 – 1500 nm. Charge carriers can be excited by a green laser (532 nm) coupled into the microscope optics, generating pho-toluminescence. Alternatively, electroluminescence and breakdown light emission are generated by biassing the sample with the help of a voltage generator, allowing for volt-ages up to 20 V. The sample is held on a high-precision nanometer-resolution scanning table. An overall resolution in the order of 1 µm can be achieved.

With the help of the PL- / EL-measurement tool, the spectra of singular BD light emitting spots can be analyzed. In order to account for the varying sensitivities of the detectors and for the influences of the optical path on the spectrum, the setup is calibrated using a calibration lamp with a known spectrum. For details on the microscope setup, refer to [120].

6.2.2 Measurements based on thermography

The heat generated in breakdown sites due to the reverse current flow can be measured via dark lock-in thermography (DLIT) [171]. It is directly proportional to the power dissi-pation and thus to the current in the breakdown sites, yielding the relevant information concerning the solar cell operated in a module.

The DLIT images can be calibrated to the local current density [171] according to

A T T I

J_local = _local ^total (6-4),

when the total current flowing through the entire solar cell I_total is known; A stands for the entire solar cell area and T is the average temperature signal.

As a result of the direct proportionality, DLIT measurements can be combined in order to yield information about the breakdown type [172]. Most importantly, with the help of the so-called TC-DLIT method, the local temperature coefficient (TC) can be estimated (see sections 6.1.1 and 6.1.2). This is done by combining two measurements which are per-formed at the same reverse voltage but at two different temperatures. The difference in the measured reverse current directly indicates the TC.

In the DLIT experiments, the sample is mounted to a copper chuck which is thermostat-controlled. The bus-bars are contacted via four pins which are connected to a voltage generator allowing for biases up to 20 V and 20 A. The generated heat is then detected by a stirling-cooled silicon CCD camera sensitive in a wavelength range between 3 and 5 µm.

6.2.3 Comparison of both measurement methods

Both measurement methods, EL as well as DLIT, present different opportunities and drawbacks which have to be kept in mind during the investigation of the reverse bias dependence of mc-Si solar cells:

As breakdown usually happens in small sites with diameters in the order of a few µm, EL imaging-based investigations provide the possibility to correlate breakdown sites with other solar cell-related issues such as recombination activity or dislocation luminescence with a relatively high resolution in the order of 100 µm. In addition, the acquisition time is in the order of only a few seconds.

DLIT is based upon the thermal waves generated by a periodical voltage variation (30 Hz in most cases, yielding a resolution in the order of 1 mm). To obtain high resolution DLIT measurements also in the order of 100 µm, the frequency of the periodical voltage varia-tion has to be increased. This leads to very low signal-to-noise ratios, prolonging the measurement time to several minutes per DLIT image.

At this point, it needs to be said that in both, camera-based EL measurements and DLIT images, each pixel can contain the signal of dozens of breakdown sites, risking to obliter-ate any properties of singular breakdown channels.

While the light generated in breakdown sites is an indirect consequence of many charge carriers in a large electric field, for which the correlation with the local reverse current is a priori not clear (see section 6.3.2), the heat is directly proportional to the local reverse current. For measurements involving the current density, thus DLIT-related methods should be preferred. However, due to the broken symmetry at the solar cell edge, special care has to be taken when interpreting the DLIT signal in these solar cell regions [171].

During EL investigations, the reverse bias applied to the solar cell is kept constant for a couple of seconds in order to suppress noise. Therefore, depending on the measurement chuck design, the sites of the breakdown are subject to different heating conditions dur-ing the DLIT and EL measurements. Whether or not the increased heatdur-ing in EL meas-urements has an influence on the breakdown behavior under different reverse biases is difficult to tell. Taking DLIT measurements as reference, supposing that they are less influenced by the reverse current induced heating, no significant impact has been no-ticed.