In Situ Surface Temperature Measurement in a Conveyor Belt Furnace via Inline Infrared Thermography

Summary

This protocol describes how to install an infrared camera into a conveyor belt furnace, conduct a customer correction of a factory calibrated IR camera, and evaluate the spatial surface temperature distribution of an object of interest. The example objects are industrial silicon solar cells.

Abstract

Measuring the surface temperature of objects that are processed in conveyor belt furnaces is an important tool in process control and quality assurance. Currently, the surface temperature of objects processed in conveyor belt furnaces is typically measured via thermocouples. However, infrared (IR) thermography presents multiple advantages compared to thermocouple measurements, as it is a contactless, real-time, and spatially resolved method. Here, as a representative proof-of-concept example, an inline thermography system is successfully installed into an IR lamp powered solar firing furnace, which is used for the contact firing process of industrial Si solar cells. This protocol describes how to install an IR camera into a conveyor belt furnace, conduct a customer correction of a factory calibrated IR camera, and perform the evaluation of spatial surface temperature distribution on a target object.

Introduction

Process control and quality assurance of objects processed in conveyor belt furnaces¹ is important and accomplished by measuring the surface temperature of the object. Currently, the temperature is typically measured by a thermocouple¹. As thermocouple measurements require contact with the object, thermocouples inevitably damage the object. Therefore, it is common to choose representative samples of a batch for temperature measurements, which are not further processed since they become damaged. The measured temperatures of these damaged objects are then generalized to the remaining samples from the batch, which are further processed. Accordingly, production must be interrupted for thermocouple measurements. Furthermore, the contact is local, needs to be readjusted after each measurement, and influences the local temperature.

Infrared (IR) thermography² has a number of advantages over classic thermocouple measurements and represents a contactless, in-situ, real-time, time-saving, and spatially resolved temperature measurement method. Using this method, each sample of the batch, including those that are further processed, can be measured without interrupting production. In addition, the surface temperature distribution can be measured, which provides insight into temperature homogeneity during the process. The real-time feature allows correction of temperature settings on-the-fly. So far, the possible reasons for not using IR thermography in conveyor belt furnaces are 1) unknown optical parameters of hot objects (especially for nonmetals³) and 2) parasitic environmental radiation in the furnace (i.e., reflected radiation detected by the IR camera in addition to the emitted radiation from the object), which leads to false temperature output².

Here, as a representative proof-of-concept example of IR thermography in a conveyor belt furnace, we successfully installed an inline thermography system into an IR lamp powered solar firing furnace (Figure 1), which is used during the contact firing process of industrial Si solar cells (Figure 2A,B)^4,5. The firing process is a crucial step at the end of industrial solar cell production⁶. During this step, the contacts of the cell are formed^7,8, and surface passivation is activated⁹. To successfully achieve the latter, the time-temperature profile during the firing process (Figure 2C) must be accurately realized. Therefore, sufficient and efficient temperature control is required. This protocol describes how to install an IR camera into a conveyor belt furnace, conduct a customer correction of a factory calibrated IR camera, and evaluate the spatial surface temperature distribution of a target object.

Protocol

1. Installation of IR camera into a conveyor belt furnace

1. Decide which part of the furnace should be measured by the IR camera.
   NOTE: Here, the peak zone of the firing process is chosen (see the orange highlighted zone in the firing area of Figure 1A).
2. Define the temperature range of interest that the IR camera should detect (e.g., 700−900 °C, the typical peak temperature range of the firing process).
3. Determine, or at least estimate (through experiments or literature), the temperature, spectral, and angular dependent emissions of the object(s) of interest (e.g., silicon solar cell) to identify the wavelength range(s) of highest emission for the temperature range of interest (under a specific camera angle).
   NOTE: Here, the emission is estimated based on previous literature³ and a software called RadPro¹⁰, which calculates the spectral, angular, and temperature-dependent emissivity for materials of interest.
4. Deciding on the IR camera type
   NOTE: Here, a midwave infrared (MWIR) indium antimonide (InSb) camera (Table of Materials) is used.
   1. Choose a camera that can detect the temperature range of interest.
   2. Select a camera whose detection wavelength range matches the wavelength range of highest emission of the object of interest in the temperature range of interest.
   3. Avoid as much parasitic radiation detection by the camera as possible by avoiding objects that emit or reflect radiation into the camera field of view (e.g., IR lamps in a furnace). 
   4. Decide on the necessary spatial and temporal resolution of the camera (e.g., 640 px x 512 px and 125 Hz [full image] for the used camera here).
5. Realize a sufficient optical path from the IR camera to object (see Figure 1B).
   1. Avoid disturbing objects in the optical path (e.g., IR lamps causing direct or reflected light).
   2. Position the camera outside of the furnace chamber, if possible.
      NOTE: Most cameras have low operating temperatures (e.g., up to 50 °C). Make sure in advance that the camera position can be changed, if desired.
   3. Remove the furnace wall and isolation at the location where the optical path should be and replace the hole with an insulating IR window.
      1. Choose the appropriate material for the window that meets the following demands: 1) as transparent as possible for the detection wavelength (λ) range of the camera (e.g., quartz glass window for ~0.2 µm < λ < 3 µm, sapphire window for ~0.4 µm < λ < 4.2 µm) and 2) able to isolate the furnace chamber thermally.
         NOTE: The resulting temperatures of the window may influence the window transmission.
      2. Avoid damage of the IR window. Do not tighten the window to avoid breakage during heat expansion.
         NOTE: The window material should have a sufficient amount of space to expand when heated up.
6. Check the resulting field of view (FOV) of the IR camera by examining the thermography image via the IR camera software. Identify the targeted object and its temperature in the thermography image. Adjust the FOV, if necessary.

2. Global customer temperature correction of a fabrication calibrated IR camera

CAUTION: The fabrication of the IR camera is assumed to include a radiometric calibration.

1. Spot local optical artifacts, such as reflection and background radiation.
2. Conduct classic thermocouple measurements of the object while simultaneously recording the wafer including thermocouple with the IR camera.
   1. Check the validity of the used thermocouples. Search for known characteristic temperature points in the temperature profile of the processed object that can be clearly visibly detected (e.g., disruption in a smooth line). If the thermocouple measures these temperature points correctly, the thermocouple is most likely correctly calibrated.
   2. Example using silicon solar cells
      1. Place the thermocouple on the rear aluminum side of the wafer. Take a temperature profile for a standard firing process¹¹.
      2. Validate the thermocouples by determining whether there is a disruption in the temperature profile from step 2.2.2.1 around the Al-Si eutectic temperature of 577 °C in the form of a flatter curve (as is the case in Figure 2D).
         NOTE: If the disruption occurs at the temperature around 577 °C, it is a sign that the temperature measurement by the thermocouple is accurate. Use only validated thermocouples for the following steps.
   3. Conduct thermocouple measurements in the temperature range of interest at the same object spot (multiple times for statistical reasons), then at spatially various random spots (for statistical reasons) to obtain time-temperature profiles.
3. Determine the local uncorrected thermography object temperature underneath the thermocouples from the thermocouple measurements from step 2.2.3 while placing the thermocouple on the upper side of the object.
   1. Check for a possible local temperature drop around the contacting thermocouple (due to heat dissipation and shading). Assume the temperature in the vicinity of the thermocouple as the object temperature directly under the thermocouple, if a local temperature drop is not present.
   2. Perform the following steps if a local temperature drop is present.
      1. Determine the spatial temperature gradient of the present temperature drop in the part that is not covered by the thermocouple.
         NOTE: It is recommended to determine the gradient at multiple spots around the temperature drop (radially) and determine an average gradient.
      2. Estimate the contribution of possible optical artifacts induced by the thermocouple (example protocol for a case in which homogenous temperature along the cell depth direction is assumed, such as in Si solar cells).
         1. Place the thermocouple on the surface opposite to the measured surface and repeat the thermocouple and thermography measurement in this configuration (as shown in Figure 3A). Turn the object, including the thermocouple, around so that the thermocouple is not in the optical path between the camera and object.
            NOTE: If the gradient of the local temperature drop is the same for the thermocouple being inside and outside of the optical path (i.e., attached to the measured or opposite surface), it is a sign that the thermocouple most likely does not induce optical artifacts.
         2. Extrapolate the gradient of the temperature drop in the case of the thermocouple contacting the measured surface (i.e., inside optical path) to the area covered by the thermocouple to obtain the temperature of the object underneath the thermocouple.
         3. Repeat 2.3.2.2.2 for each measurement from step 2.2.3.
4. Alternative to 2.3: Determine the local uncorrected thermography object temperature underneath the thermocouples from the thermocouple measurements from step 2.2.3 while placing the thermocouple on the lower side of the object. To determine the local uncorrected thermography solar cell temperature under the thermocouple, extract the local temperature at the position of the thermocouple. 
   NOTE: Keeping the thermocouple on the rear side prevents the thermocouple from blocking the sight on the object by the camera. Therefore, on the one hand, the temperature correction is significantly simpler. On the other hand, thermocouples are usually not positioned on the lower side of the object during the firing process, thus might lead to operational complications, which is why this alternative needs to be carried out extra carefully. 
5. Correct the uncorrected thermography image with respect to the thermocouple measured temperatures with the data generated from steps 2.3 or 2.4.
   1. Plot the measured temperatures via thermocouples against the determined temperatures via uncorrected IR thermography. Conduct a curve fitting.
   2. Apply the obtained curve fit as a general uniform global correction formula for the uncorrected thermography image.
6. Repeat the temperature correction for each new object type or configuration, especially when the optical parameters differ.

3. Evaluation of spatial surface temperature distribution via IR thermography

NOTE: The firing conditions are assumed to be identical for this section.

1. Creation of a two-dimensional peak temperature distribution map (see Figure 4A)
   1. Write a script with an appropriate programing language to track the surface object temperature for each object surface spot along the entire camera FOV, i.e. acting as a "virtual thermocouple" placed at all object spots simultaneously.
      NOTE: Here, the script is written in MATLAB.
   2. Extract the maximum value, i.e. the peak temperature, for each object spot and plot these temperatures in a corresponding 2D distribution map.
2. Average temperature distribution in and perpendicular to the object throughput direction (see Figure 4B)
   1. In throughput direction: average the 2D temperature distribution in the dimension which is opposite to the throughput direction. What remains, is the average 1D temperature distribution in throughput direction.
   2. Perpendicular to the throughput direction: average the 2D temperature distribution in the dimension which is in throughput direction. What remains, is the average 1D temperature distribution perpendicular to the throughput direction.
      NOTE: It is recommended to leave out the last centimeter (at least) of the edge for the averaging since optical artifacts at the object edge might falsify the resulting temperature average.

Results

As shown in Figure 3B−D, the example object (here, a silicon solar cell; strictly speaking, a passivated emitter and rear cell [PERC]¹²; Figure 2A,B) can be clearly detected by the IR camera in different configurations⁴. The different configurations are monofacially metallized (Figure 3B), bifacially metallized¹³ (Figure 3C) and nonmetalized PERC samples (Figure 3D). The difference between the monofacial and bifacial configuration is that the former has a full area aluminum layer, whereas the latter has an H-pattern grid (similar to the silver front side) on the rear side. Here, the IR camera was positioned in a way that the camera FOV captured the peak temperature of the firing process. The peak phase is the most crucial phase during the firing process, since the contacts are actually formed during this phase¹⁴. Here, the temperature range of interest resembled the typical peak temperature range of the firing process (i.e., ca. 700–900 °C¹).

For the latter temperature range, the spectral emissivity is quite high and homogenous in the short, middle, and long wavelength infrared spectra³. A double sapphire layer was used as a transmissive window, allowing for good transmission in the short and middle IR wavelength spectra. In order to minimize detection of light from the IR lamps of the furnace (peak wavelength in short wavelength infrared range), an IR camera type with InSb as detector material was chosen, with a detection range of 3.7−4.1 µm (including filters). Only one-third of the wafer in the throughput direction can be detected at the same time. However, it was sufficient for this work, since the wafer passes the existing field of view entirely. Naturally, temperature corrected thermography images are shown here. Strictly speaking, the image is temperature-corrected with respect to the solar cells.

As can be seen in Figure 3A, the contacting thermocouple on the opposite side of the optical path caused a temperature drop around itself (with a temperature drop of 10 K), most likely due to heat dissipation and shading. The latter drop is important to estimate the cell temperature during firing without thermocouples, compared to the temperature measured by the thermocouple. Here, the cell was positioned onto a frame when contacted by a thermocouple (Figure 3E). The heat dissipation by the frame caused a temperature drop of around 10 K. Together with the additional heat drop by the thermocouple, the latter measured a 20 K lower temperature than what the cells displayed during standard processing (without the thermocouple equipment). It is important to estimate the latter offset for the used thermocouple system, which is performed with the help of thermography, as shown. The IR camera allows observation of the local heat dissipation of the cells by the conveyor belt if placed directly on the belt (Figure 3F). This is the reason why cells are usually placed on belt elevations to minimize contact between them and the belt.

Figure 4 shows the surface temperature distribution. Since silicon solar cells are typically around 160 µm thick and processed in the furnace for 30 s, it is likely that the temperature distribution along the cell depth is homogenous. Therefore, the results most likely suggest a temperature distribution rather than only a surface temperature distribution. Opposite to the throughput direction, an average temperature gradient of 1 K/cm was obtained. In the throughput direction, the incoming wafer quarter was substantially colder than the trailing wafer rest. The colder incoming portion experienced a gradient of 7 K/cm, while the hotter trailing part experienced a gradient of 0.5 K/cm.

In both directions, the cell edges (the remaining 2 cm) were ignored for determination of the gradients, since the detected temperature at the edges mixed with the colder outside boundary of the cells, resulting in false temperatures. Figure 4C shows a representative 2D temperature distribution of a monofacial solar cell, which was not metallized at the front side. The abovementioned trends in the same and opposite transport directions were observed here, as well. All in all, these results reveal that the solar cells in this work experienced a certain degree of spatial temperature inhomogeneity.

Figure 1: Most important equipment used in the protocol. (A) Lateral scheme of the conveyor belt furnace. This figure panel has been modified from Ourinson et al.⁴. (B) Zoomed-in last firing zone, visualizing the setup of the thermography system. 1) Furnace wall and isolation, 2) IR camera, 3) IR lamps, 4) insulating window, 5) object transport direction, 6) camera FOV, 7) transportation belt, 8) object, and 9) thermography software.  This figure panel has been modified from Ourinson et al.⁴. (C) The firing furnace used during this protocol. (D) Image illustrating the used IR camera and transmissive IR window positioned in the firing furnace. The numbers correspond to the numbers from panels A and B. Please click here to view a larger version of this figure.

Figure 2: Measured objects and their temperatures. (A) Schematic cross-section of a monofacial PERC solar cell. (B) Front (left) and rear (right) side view of an industrial PERC cell. (C) Thermocouple-measured industrial time-temperature profile of a PERC solar cell during the firing process, including segmentation into phases and section, which is covered by the camera field of view. This figure has been modified from Ourinson et al.⁵. (D) Demonstration of disruption around the eutectic temperature (T_EUT) of aluminium and silicon in a firing profile measured by a thermocouple, when the thermocouple is placed on the aluminium rear side of the solar cell. This figure has been modified from Ourinson et al.⁵. Please click here to view a larger version of this figure.

Figure 3: Representative temperature-corrected thermography images of PERC solar cells for identical firing conditions. (A) Visible local temperature drop caused by contact of a thermocouple from the rear side. (B) Thermography image of the upper one-third of a monofacially metallized PERC cell, including (1) visible busbars (2) positioned on the visible conveyor belt. T_AV shows the average temperature on the wafer. (C) Thermography image of a bifacially metallized PERC cell. (D) Thermography image of a nonmetallized PERC wafer. (E) Thermography image of a wafer placed on a thermocouple frame and contacted by a thermocouple. T_TC shows the wafer temperature measured by the thermocouple. (F) Thermography image of a wafer placed directly on the conveyor belt. (G) Color map of the temperature range measured by the IR camera. This figure has been modified from Ourinson et al.⁵. Please click here to view a larger version of this figure.

Figure 4: Temperature distribution of a PERC solar cell for identical firing conditions. (A) 2D peak temperature distribution of a monofacial PERC solar cell from the front side. (B) Average peak temperature distribution in (right picture) and perpendicular (left picture) to the cell transport direction.”Please click here to view a larger version of this figure.

Discussion

Commonly, thermography temperature is corrected via measuring and adapting the optical parameters of the object, transmissive window and path, and environmental temperature of the object and transmissive window². As an alternative method, a temperature correction technique based on thermocouple measurements is described in this protocol. For the latter method, knowledge of the parameters mentioned above is not required. For the application shown here, this method is sufficient. However, it cannot be guaranteed that the thermocouple method is sufficient for all thermography applications in a conveyor belt furnace.

In the protocol, a uniform global temperature correction of the thermography image is proposed; although, it is more precise to correct the spatially resolved temperature. However, it has been found that the uniform temperature correction is more appropriate in cases of moving objects. Furthermore, it is intended to correct the temperature of the object rather than the surrounding objects (e.g., the belt and walls).

As mentioned in step 2.3.2.2, the example provided here is assumed to have a homogeneous temperature distribution along the object depth. In cases of objects with inhomogeneous temperature distribution along their depths, the temperature on one surface does not resemble the temperature on the opposite surface. Thus, the steps described in section 2.3.2.2 do not apply for these cases. A solution for inhomogeneous temperature distribution along the object depth must be further studied.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Datalogger incl. Thermal barrier	Datapaq Ltd.
IR thermography camera "Image IR 8300"	InfraTec GmbH
IR thermography software "IRBIS Professional 3.1"	InfraTec GmbH
Solar cells	Fraunhofer ISE
Solar firing furnace "RFS 250 Plus"	Rehm Thermal Systems GmbH
Sheath thermocouples type K	TMH GmbH
Thermocouple quartzframe	Heraeus Noblelight GmbH