Electrical heating synchronized with IR imaging to determine thin film defects

Kimmo Leppänen; Juha Saarela; Risto Myllylä; Tapio Fabritius

doi:10.1364/OE.21.032358

1. Introduction

Conductive thin films are used in many different electrical applications such as solar cells, flat-panel displays, electro chromic mirrors and windows, touch-panel controls and many others [1]. There is a wide variety of conductive films and a careful consideration is needed to fulfill application demands such as flexibility, rugged behavior and good conductivity [2–5]. Uniformity is an important feature of conductive thin films and it should be measurable during manufacturing or in the final applications. Laboratory scale characterization methods are typically either electrical or optical. Electrical conductivity standardized measurements are done by measuring a small area by using a four point probe [6]. Recently microscopic four-point probes have also been developed, which increase the lateral resolution and accuracy of conductance measurement [6]. Examples of traditionally used laboratory scale characterization tools are optical microscope [7–11], atomic force microscopy (AFM) [12], optical profilometer [12], high resolution transmission electron microscopy (HR TEM) [12], and Scanning Electron Microscope (SEM) [6,9,10,12,13]. The lateral resolution with these techniques varies from nanometer scale to micrometer scale. With these methods, it is possible to visualize the defect on the surface of the film in a relatively small area as these techniques are designed to samples far less than 1 mm² [6,8–13]. These techniques are limited to highlighting the changes only after deformation has occurred on the surface of the films. However it has also been suggested that before the crack becomes visible there are micro defects such as micro-cracks and micro detachment on the conductive films, which can be detected in the decreased conductance of these films [10].

Lock-in thermography is a technique whereby samples are heated and infrared camera is used to image the temperature profiles [14–17]. The technique uses modulated power to heat the object, and IR-imaging is synchronized with heat modulation. Rakotoniana et al. created a highly sensitive lock-in thermography system. The temperature resolution was down to noise level of the IR-camera, in this case 10 µK. Spatial resolution with special microscopic objective was 10 µm [14]. Besides electrical power, heat can be produced to the sample using other methods such as laser light [15]. Unfortunately laser cannot heat only the conductive film. Kunz et al. and Straube et al. were studying shunting problems due to sub-micron pinholes in solar cells. In these cases, the studied sample sizes were on the range of 10 mm [16, 17]. Rakotoniana et al. studied weak heat sources in electronic devices and solar cells. The sample sizes were 10 × 10 mm and 10 × 10 cm and the temperature spots were clearly visible [14]. When heating the sample in periods longer than a few seconds, the technique is not any more called lock-in thermography. When integrated circuits were heated with DC current for 10 seconds, almost the whole device heated up and the contrast of the IR-image was lost [14]. IR-camera has been used for several imaging applications [14–17], but a combination of IR-camera and electrical heating has not been demonstrated successfully outside of lock-in thermography.

Existing measuring techniques have problems when applied to production control of thin films. In the case of optical methods, it is not easy to detect defects on larger sample areas (cm² or bigger scale) and conductance measurements do not give the spatial information about samples with large size. Our aim was to build a robust measurement setup to localize and classify defects in conductive films in large areas and in high spatial resolution. Data processing speed is also important for implementing this method later in roll-to-roll production. To provide a suitable solution, we have built a system where DC-electrical heating is synchronized with IR-imaging.

2. Materials and methods

2.1 Indium tin oxide (ITO) samples and environmental conditions

ITO was chosen as sample material because it is a good example of conductive thin films. The benefits (high conductance, good transparency in visible wavelengths, relatively wide band gap and high work function) overcome the weaknesses (brittleness, cost). In addition it is at present one of the most widely used conductive thins film even in flexible applications [2–5].

The samples were cut from commercial PET-ITO roll, manufactured by DuPont, article code of DuPont Teijin film is Melinex ST 504. ITO specifications are nominal resistance 50 Ω / square, thickness 125 nm. The thickness of PET was 125 µm. Sample dimensions are summarized in Table 1. During IR measurements, the room temperature varied between 22.1 °C – 23.3 °C.

Table 1. Selected dimensions for samples and bending cylinders

View Table | View all tables in this article

The samples were subjected to bending by using cylinders with diameters given in Table 1. Five parallel samples were bent with every cylinder. Each sample was treated with a device, where the sample was bent over cylinder. Sample was pressed against the cylinder pulling with a one kilogram weight. As PET substrate was in contact with the cylinder, ITO remained free from accidental attrition, scratches or depression, and it was cracked in the sample handling device just by bending force. Additionally two different reference sample types were produced. Five references were pulled with 1 kg weight without any bending. Bending or pulling time was 3 minutes. Two additional references were not bent or pulled. Sample bending setup and procedure was the same as described by Leppänen et al. [18].

2.2 Synchronized thermography, measurement setup and devices

We define the arrangement where electrical heating of the samples is synchronized with IR-imaging as synchronized thermography (ST). First, the sample is placed between two power probes, which were 100 mm apart from each other. Triggering connects 20 V voltage over the sample. When the sample has been warmed by current for five seconds, an IR-camera takes the IR-image. The first tested parameter was synchronizing time between heating and IR-imaging. The tested times were 5-15 seconds and breakages were visible with all of them. Finally, five second heating time was chosen not to overheat the sample. Figure 1 shows the measurement setup.

Fig. 1 Measurement setup of the synchronized thermography (ST). Triggering starts heating the sample with electricity and after five seconds IR camera was triggered to take a picture the sample. In addition to ST, current and voltage were measured, marked in the figure with symbols A and V.

Download Full Size | PDF

IR- images were measured by using Flir b60 Infrared camera in temperature ranging from 21 °C to 46 °C, when giving 20V voltage for 5 seconds to the samples by ITT Instruments AX 323 power source. The IR camera was installed on a stable stand where the distance from the objective to the middle of the sample was 25 cm. In order to calibrate the IR-camera, a Lutron thermometer TM-903 was calibrated with ice water (0 °C) and boiling water (100 °C). The temperature reading of the IR camera was calibrated by Lutron thermometer on temperatures 20°C ( ± 1°C), 22 °C ( ± 1°C), 30 °C ( ± 1°C) and 40 °C ( ± 2°C). In Flir b60 specifications, accuracy is ± 2 °C. Measurements were done by using 256 temperature levels and image size 240 × 240 data pixels. Temperature range was set from 21°C to 46°C, which gave a temperature resolution better than 0.1 °C. In order to get temperature rise, room temperature was also measured at the same time as the temperature profile of sample. Room temperatures were measured with IR-camera from an unheated reference sample sheet. Simultaneously with measuring IR images, current and voltage were measured by using Agilent 34401A 6 ½ digit multimeter, and the conductance of the sample was calculated from the measured values. In this way conductance value and IR image were taken at the same moment. The original conductance of all samples was also measured before ST measurements by using the resistance measurement mode (4W) of the same multimeter.

Brucker Contour GT profilometer was used to see if the cracks appear on the sample. Parameters were as follows: measurement type: Phase-Shifting Interferometry (PSI), green light, objective 20X and measurement area 158 µm × 119 µm.

2.3 IR-image pre-handling

The original IR-Fig. included the sample, not powered reference sample and some surrounding areas. In order to achieve temperature rise of the region of interest (ROI), the following steps are needed:

1) Extracting sample area and reference temperature
2) To achieve temperature rise, sample specific room temperature (T_rt) is subtracted [Eq. (1)]: $T_{p} (x, y) = T_{extracted} (x, y) - T_{rt}, where x \in {1, 2, \dots 157} and y \in {1, 2, \dots 55}$
3) To exclude reflection from probes and cooling in the edges of the sample, ROI was selected, named (T_n(x,y)).

The details of the original image and three intermediate images together with pixel sizes and sizes in square millimeters are described in Table 2. The same steps are shown in Fig. 2.

Table 2. Data processing pre-handling steps

View Table | View all tables in this article

Fig. 2 From original IR-image to ROI. Blue is the coolest and orange is the warmest area. All samples were extracted to the size of T_n(x,y) to exclude temperature diffusion (in the case of the cracked samples, it was extracted on the side where the temperature was highest and the part where diffusion occurred was cut out).

Download Full Size | PDF

2.4 Defining major and minor breakages

The conductivity of ITO decreases dramatically, when defects appear as cracks on the surface [18]. In this article, defects without cracks on the surface are categorized as minor breakage. Respectively when cracks appear on the surface defects are called major breakage. Figures 3(a) and 3(b) show images of optical profilometer for not cracked and cracked samples. A clear difference can also be seen in the samples when they were measured with ST, and median temperature (Med(ΔT)) was calculated. First 49 medians were calculated [Eq. (2)] from every row of the sample map, and then median value from these medians was calculated [Eq. (3)]. This is later used to classify the breakages.

Fig. 3 Figures a and b show images measured by optical profilometer a) Not cracked sample (reference sample) b) Cracked sample (bent with 15 mm cylinder, arrows point to cracks). The dimensions of a) and b) were 158 µm (y-axis) × 119 µm (x-axis). c) Med(ΔT) of cracked and not cracked (Not cr.) samples, which were measured by ST. Samples are all the samples mentioned in section 2.1.

Download Full Size | PDF

Md, row (y) = Md [T_{n} (x, y) | x \in {1, 2, \dots 125}], where y \in {1, 2, \dots 49}

Med (Δ T) = Md (Md, Row (y) | y \in {1, 2, \dots 49})

In the Fig. 3(c) there is a clear gap on Med(ΔT) value between cracked or not cracked samples.

After definition of the cracked sample, it is possible to make the division in automatic data processing for every sample. Figure 4 shows data processing phases to the major and minor breakages and undamaged samples.

Fig. 4 Flow chart ending on major and minor breakages.

Download Full Size | PDF

The first step (IR-Fig. pre-handling) was described in section 2.3. The second step is to compare samples Med(ΔT) value to threshold value Med(ΔT)_th and separate samples with major breakages from the rest of the samples.

In Fig. 3(c), the PET-ITO samples have Med(ΔT) values. Cracked samples have significantly lower values. Based on this figure the threshold value Med(ΔT)_th is set to + 2 °C.

2.5 Data processing functions to qualify minor breakages and undamaged samples

Figure 5 shows how the data was processed to distinguish minor breakages and undamaged samples and characterize minor breakages.

Fig. 5 Data processing for minor breakage characterization

Download Full Size | PDF

The first step in the process is to subtract Md,row(y) from T_n(x,y) median value. By this Eq. (4) temperature diffusion was reduced on the direction of sample width.

T_{n - m} (x, y) = T_{n} (x, y) - Md, row (y); where x \in {1, 2, \dots 125} and y \in {1, 2, \dots 49}

The next step is to derive on the direction of sample width. To be able to derive the data map of the sample (T_n-m(x,y)), it was first duplicated. Both of these data matrices were modified by adding to them one row with zeros, for the other to the beginning and for the other to the end. The data matrices were at this moment T_n-m+(x,y) and T_n-m-(x,y), where $x \in {1, 2, \dots 125} a n d y \in {1, 2, \dots 50}$ . When calculating preliminary result D_p, the second data file was subtracted from the first one [Eq. (5)]:

D p (x, y) = T n - m + (x, y) - T n - m - (x, y), w h e r e x \in {1, 2, \dots 125} a n d y \in {1, 2, \dots 50}

After deriving, the sample data was also smoothed by taking median value of sliding 3 × 3 pixels with symmetric edges. Finally, the first and the last row were deleted as they did not represent real derivate values. After these additional functions, the data map had the following final form: $D_{d & s} (x, y), w h e r e x \in {1, 2, \dots 125} a n d y \in {1, 2, \dots 48}$ . At this moment it was possible to classify if the values of the pixels exceeded over the threshold of derivate value D_th and all pixels of the samples, where D_d&s(x,y) ≥ D_th, were marked as broken areas. Now the sample data was ready to corner pixels evaluation (explained in the next section) and then ΔT_minor and breakage area were calculated. ΔT_minor is the average temperature rise in the broken area of the sample [Eq. (6)]. Breakage area (BA_minor) is the sum of all broken parts of the sample [mm²] [Eq. (7)].

Δ T_{minor} = Ave (T_{n} (x, y) | x \in {1, 2, ...125} a n d y \in {1, 2, ...48} a n d D d & s (x, y) > D t h)

B A_{m i n o r} = 0.64 m m \times 0.64 m m \times n,

where n is the number of pixels, when

D_{d & s} (x, y) > D_{th},

where

x \in {1, 2, \dots 125}

and

y \in {1, 2, \dots 48}

and pixel size is 0.64 mm × 0.64 mm.

For PET-ITO samples the threshold value was set to D_th = 0.2. In the measurement setup temperature diffusion was slightly bigger on the corners of the samples than in the middle of the sample. That phenomenon was detectable on some individual corner pixels even on this already row median value subtracted data map. Therefore corners were evaluated separately. On every corner three rows and 6 columns were evaluated (18 pixels). A selection criterion was related to number of broken pixels. If at least 50% of the corner pixels (nine or more pixels) exceeded over D_th = 0.2 °C, the breakage was qualified as real breakage and the results were calculated by using all pixels. Otherwise breakage detection in the sample was continued but corner pixels were excluded from further evaluations. Now the sample map (D_d&s(x,y)) was evaluated again by using the same threshold criteria of derivative value D_th = 0.2°C. If none of ΔT_d&s(x,y) values were bigger than D_th = 0.2°C, the sample was classified as not broken. In this case no further evaluation was needed. When there was breakage in the sample, ΔT and breakage area were calculated.

2.6 Data processing functions to qualify major breakages

Figure 6 shows how the data was processed to separate major breakages.

Fig. 6 Data processing for major breakage characterization

Download Full Size | PDF

First procedure in major breakage evaluations was to subtract the median temperature value of the whole sample [Med(ΔT), Eq. (3)] from every data pixel [Eq. (8)]. That was done to set up not cracked part of the samples to the zero level.

T_{n - m (ws)} (x, y) = T_{n} (x, y) - Med (Δ T); where x \in {1, 2, \dots 125} and y \in {1, 2, \dots 49}

The second procedure was to calculate the median temperatures of the columns of the sample [Eq. (9)].

Md, col (x) = Md [T_{n - m (ws)} (x, y) | y \in {1, 2, \dots 49}], where x \in {1, 2, \dots 125}

Then these Md,col(x) values were compared to the threshold temperature value of column median pixel (CM_th, Fig. 6). When Md,col(x) > CM_th, where $x \in {1, 2, \dots 125}$ , the column of the sample was defined as broken part of the sample. To be able to calculate ΔT for major breakage, average values of the columns were first calculated [Eq. (10)]. For every broken column where breakage criteria fulfilled (Md,col(x) > CM_th), the average of these averaged column values were calculated and result was called ΔT_major [°C] [Eq. (11)].

Ave, col (x) = Ave (T_{n - m (ws)} (x, y) | y \in {1, 2, \dots 49}, where x \in {1, 2, \dots 125}

Δ T_{major} = Ave (Ave, col (x) | x \in {1, 2, \dots 125} and Md,col(x) > {CM}_{th}

The same criteria (Md,col(x) > CM_th, where $x \in {1, 2, \dots 125}$ ) was used, when breakage area [mm²] (BA) was calculated [Eq. (12)] as pixel size was 0,64 mm × 0,64 mm and 49 pixels were on every row.

{BA}_{major} = 0.64 mm \times 0.64 mm \times 49 \times n,

where n is the number of broken columns of the sample

For PET-ITO samples, the breakage threshold value was set CM_th = 0.2°C.

2.7 Data processing functions to mark the breakages in the sample map

Figure 7 shows how the data was processed to mark the breakages in the sample map.

Fig. 7 Data processing to mark the breakages in the sample map.

Download Full Size | PDF

All data pixels fulfilling breakage criteria were marked. Binary map of broken areas of the sample was created. Pixels of broken area had values one and pixels of unbroken areas had values zero. The area surrounding the area of breakage was processed by using MatLab function: imdilate{breakage-map(x,y),strel('square',3)}; where $x \in {1, 2, \dots 125}$ and $y \in$ {1,2,…48}, minor breakage or $y \in {1, 2, \dots 49}$ , major breakage. Then the area inside of these processed areas was marked to the value zero by using function MatLab function: bwperim (breakage-map(x,y)). This map was merged into the sample map (T_n(x,y)) and result is IR picture together with breakage borders.

3. Results

As a result of the above described method development, the first result was a method to define successively the breakage type and localize it on the sample map. Secondly, as an example, a set of PET-ITO samples were processed by the developed method. Outcome was ΔT, breakage area and breakage map for each sample. Additionally the correlation between average temperature and conductance of the samples is presented. The third result shows data processing time.

3.1 Automatic evaluating and breakage area visualization method

The summary of the automatic data processing method is shown in Fig. 8, including the automatic evaluating and breakage area localization method with all the steps mentioned. The specific selection values of the sample material (ITO) are inserted into this flow chart.

Fig. 8 Flow chart of automatic evaluating and breakage area localization method

Download Full Size | PDF

Different intermediate maps of one representative sample are shown in Figs. 9(a)–9(f) to illustrate the important steps in data processing.

Fig. 9 Different temperature and data figures of representative sample, which was bent with 25 mm and evaluated: a) Temperature increase of the sample; b) Sample map after horizontal medians were subtracted from every pixel; c) Derivative of the previous figure on the direction of sample width, d) Previous figure is smoothened using 3 × 3 median; e) Every data-pixel with absolute value over 0.2 °C after smoothing is shown as broken area; f) Temperature increase of the sample is shown together with broken areas. Red edges are surrounding the broken areas.

Download Full Size | PDF

Median value subtraction [Fig. 9(b)] decreases the effect of temperature diffusion as it is strongest on the sides of the sample. In Fig. 9(c) the areas where breakages do not occur have zero value on the large areas of the data map. Breakages are seen on the places where the pixel-values change rapidly from below zero to positive values. In these areas the current bypasses the broken areas and therefore the temperature of unbroken areas is increased. Median smoothing [Fig. 9(d)] enables better distinguishing of broken and unbroken areas as it reduces the noise of the measurement.

3.2 Results of evaluated ITO samples

Figure 10 shows typical temperature maps of samples with located breakages. Breakage areas are surrounded with red lines.

Fig. 10 Different types of breakages; a) Reference; b-d) samples with minor breakages; e, f) major breakage, cracked samples. Red lines border the breakage areas.

Download Full Size | PDF

Figure 10(a) shows the typical temperature profile of reference sample. Figure 10(b) (bent with 53 mm cylinder) shows the broken sample breakage where a large breakage is located on the edge of the sample. This breakage was seen in quite a large area [see red borders on Fig. 10(b)]. Figures 10(c) (bent with 30 mm cylinder) and 10(d) (bent with 25 mm cylinder) show the situation when the minor breakages (see red borders) are visible in the middle of the sample. These figures are typical pixel maps of those breakage types. Figures 10(e) (bent with 20 mm cylinder) and 10(f) (bent with 15 mm cylinder) show the typical pixel maps of major breakages. In these samples the area of cracks is in the middle of the pixel maps where bending occurred. The bigger cylinder diameter (20 mm) broke up more surface area. Because conductivity is decreased over the whole width of the sample, the temperature of that area increased more than in the undamaged area of the sample.

Figure 11(a) shows that the different breakages are located in different areas on ΔT and the breakage area coordinates. Figure 11(b) shows ΔT as a function of conductance for samples with major breakages.

Fig. 11 a) The broken area of different samples (mm²) is shown as function of ΔT (°C). Samples bent with 15 mm diameter: □; Samples bent with 20 mm diameter: ○; Samples bent with 25 mm diameter:▲; Samples bent with 30 mm diameter: + ; Samples bent with 53 mm diameter: × . Samples belonging to three distinguished groups on ΔT and broken area matrix are circled. The variations of ΔT for each bending cylinder are caused by differences in the structure of the ITO-layer, randomness in the breaking event and the measurement procedure used. b) ΔT (°C) of samples with major breakage is presented as a function of conductance (mS). ΔT is the average temperature increase of the broken segments. Samples bent with 15 mm diameter: □; Samples bent with 20 mm diameter: ○.

Download Full Size | PDF

When comparing the two most broken sample types [Fig. 11(a), marked with □ and ○], their ΔT and the broken areas are found significantly different. The area of breakage is different because the most broken samples were treated with smaller cylinder (15 mm vs. 20 mm). Thus the area of breakage is related to the diameter of the cylinder used. Temperature difference (ΔT) values are also significantly different. Leppänen et al. [18] have presented that there is correlation between a number of cracks and resistance, i.e. when the number of cracks increases, the value of current drops. When the current decreases, the temperature rise of the samples is reduced.

When comparing samples with minor breakages [Fig. 11(a), marked with ▲, + and × ] to the breakages where the cracks have already reached the surface of the sample (marked with ○ and □), the breakage area is significantly lower. When a crack reaches the surface of the sample, it continues through the whole width of the sample and increases significantly the breakage area. In the case of minor breakages, more evaluation is needed to distinguish the breakages.

As ΔT change was a meaningful factor on major breakages, Fig. 11(b) shows ΔT changes of broken area as a function of conductance for these major breakages. As there is a linear trend line between conductance and ΔT for all of the samples with major breakages (R² = 0,965, where R² is the square of the Pearson product moment correlation coefficient through data points in known y's and known x's, in this case the data pairs of ΔT and conductance), ΔT can be used as a criteria to define the degree of major breakage.

As it was not possible to separate samples with minor breakages from each other on Fig. 11(a), Fig. 12(a) shows how the broken area of these samples correlates with conductance. Reference conductance of the sample was measured also before IR shooting; and these values are used in this correlation. The squares represent the broken samples. Linear trend line has been drawn with these samples (R² = 0,860). Five samples which were bent with 25 mm cylinders represent the highest breakage areas and three samples, which were bent with 30 mm represent the lowest breakage areas above the zero level. Also two samples, which were bent 53 mm (marked with × ) cylinders fit well to the trend line between conductance and breakage area. The rest of the samples were not detected as not broken ones (marked with ▲, × , + and ○).

Fig. 12 a) Breakage area as a function of conductance. Samples bent with 25 mm and 30 mm cylinder and breakage appeared: □. Trend line is drawn with these samples. Not broken samples bent with 30 mm cylinder: ▲, Samples which were bent with 53 mm cylinder × , Reference samples prepared with pulling mode + , References without any sample preparation: ○ (conductance of these was measured at the time of IR measurement); b) Mean temperature increase as a function of conductance.

Download Full Size | PDF

A more thorough investigation of the samples which were bent with 30 mm cylinder [Fig. 12(a)] and which had less than 10 mm² breakage area clearly detected the breakage and it was clear that the breakages occurred in the specific point of the sample map [Fig. 10(c)]. However, with the same treatment two samples [Fig. 12(a), marked with ▲] did not show any breakage in the IR-image, and the conductance of these samples was slightly higher i.e. the bending did not affect all of the samples in the same dominancy. The same phenomenon is seen in the case of samples bent with 53 mm cylinder (marked with × ), but the reason for the breakage in these was probably something different.

Overall, there is a correlation between the minor breakage area and the conductance of the samples. Unbroken samples and references have higher conductance and none of the references were detected as broken ones. Furthermore, the samples which were bent with 25 mm, can be separated from the broken samples which were bent with 30 mm diameter. Additionally, all the samples, which were bent with 25 mm cylinder were classified as broken ones, but only three samples which were bent with 30 cylinder were classified as broken ones.

Figure 12(b) shows that the average temperature (Mean ΔT) of the samples increases at the same time as conductance increases. The conductance drop should be linear when the mean temperature of the data drops from the most conductive sample. This is proven by Fig. 12(b) (R²>0.99). This result confirms that the mean temperature increase of the samples correlates as expected (P = U²G), where P = electrical power (W), U = voltage (V) and G = conductance (S).

3.3 Data processing time

As the speed of processing the measurement is important in production control, the time of the used automatic evaluation process was tested. The computer had 3.4 GHz Intel Core i7 processor, 8 GByte RAM and 64 byte processing. The CPU-times (central processing unit-times) from pre-handled image to the final results (ΔT, breakage area, defect map) for the samples were 48.9 ± 17.5 (22 samples) millisecond for minor breakage evaluations (including also the samples with not detected breakages) and 46.8 ± 16.4 (10 samples) millisecond for major breakage evaluation. These values support the possibility of using this kind of evaluation in on-line applications.

4. Discussion

The automatic evaluation method presented here had the resolution suitable for separating different breakage types and localizing breakages in the temperature map of the sample. In the case of minor breakage [Figs. 10(b)–10(d)] the current is bypassing the broken areas. Therefore the temperatures of the places close to the broken areas are increased. For major breakages [Figs. 10(e) and 10(f)] the temperature of the middle segment of the sample is increased because the whole area is cracked. Now the current is bypassing the broken segment on the side of the sample. This is seen as increased temperature on that specific spot.

When discovering details of temperature rise in the broken area (ΔT) and the amount of breakage areas (BA), there are clear evidences of their usage to breakage level determination. When looking at the details of Fig. 11, ΔT is explicitly different for samples bent with 15 mm cylinder or with 20 mm cylinder. Furthermore Fig. 12(a) shows a linear correlation between the conductance of the samples and the temperature rise of the broken area. Therefore ΔT is a useful parameter to define the breakage level for major breakages. Respectively there is no trend of temperature rise in the broken area for not cracked samples [Fig. 11(a)], but there is a linear trend between the area of breakage and conductance [Fig. 12(a)]. Therefore the area of breakage is a useful parameter for defining the breakage level for minor breakages. When looking at the details of the samples bent with 53 mm cylinder [Fig. 12(a)], two of them had the breakages (marked with × ) and they were on the corners of the samples [Fig. 10(b)]. It seems possible that these samples developed a small breakage at the time they were cut to shape as the breakage is on the corner of these samples. The contradiction about the source of this kind of breakage does not reduce the value of the instrumental detection capabilities. It is rather shows that these kind of phenomena can be detected with ST and automated data processing. Furthermore, the explicit conductivity limit was close to 7.4 mS [Fig. 12(a)] for this set of PET-ITO samples, where this ST system started to find breakages i.e. where it was able to detect nano- or micro-defects in the samples.

These results have shown that it is possible to measure defects by using ST and automatic evaluation and finally calculate the degree of the breaking. This information is useful when defining quality control parameters and quality limits to the end products. When planning to produce quality control measurements for instance in roll-to-roll fabrication, the method needs to be suitable for its environment, and it has to be straightforward, automatic and robust. This methodology was built to meet these requirements. The instrumentation used is simple and it includes automatic evaluation. The size of the infrared-map was 240 × 240 pixels and it covered around 150 × 150 mm². As there are available instruments with pixel amounts of 1280 × 1024, it is possible to cover around 800 mm widths with one camera. This fits well to the roll-to-roll manufacturing environment, and thus it is a promising method to produce flexible electronics in mass production scale. As the average of CPU-time of the used automatic method was below 50 milliseconds, it is fast enough to be used in process control. In roll-to-roll fabrication there are several rolls and many layers are following each other. Therefore, it is possible to accidentally defect the quality of the film. The importance of these defects varies depending on the actual application.

5. Conclusions

A method to measure local defects and breakage levels in conductive thin films with simple measurement setup was implemented and evaluated. ITO was used as an example of films. ST together with automatic data processing distinguished the samples regarding minor and major breakages and breakages were located in a map of the sample. It was possible to show linear correlation between the degree of the breakage level (ΔT for major breakages and breakage area for minor breakages) and conductivity. ST method opens new possibilities for on-line production as the measuring speed, measured area and spatial resolution are appropriate. One such example is flexible solar cells production control.

Acknowledgments

We thank VTT Technical Research Centre of Finland for providing the PET-ITO sample material. This research was partially funded by Tekes Autosys project (1082/31/2011).

References and links

1. R. G. Gordon, “Criteria for choosing transparent conductors,” MRS Bull. 25(08), 52–57 (2000). [CrossRef]

2. J. Chae, S. Appasamy, and K. Jain, “Patterning of indium tin oxide by projection photoablation and lift-off process for fabrication of flat-panel displays,” Appl. Phys. Lett. 90(26), 261102 (2007). [CrossRef]

3. L. Yang, T. Zhang, H. Zhou, S. C. Price, B. J. Wiley, and W. You, “Solution-processed flexible polymer solar cells with silver nanowire electrodes,” ACS Appl. Mater. Interfaces 3(10), 4075–4084 (2011). [CrossRef] [PubMed]

4. C. J. M. Emmott, A. Urbina, and J. Nelson, “Environmental and economic assessment of ITO-free electrodes for organic solar cells,” Sol. Energy Mater. Sol. Cells 97, 14–21 (2012). [CrossRef]

5. N. Espinosa, R. García-Valverde, A. Urbina, and F. C. Krebs, “A life cycle analysis of polymer solar cell modules prepared using roll-to-roll methods under ambient conditions,” Sol. Energy Mater. Sol. Cells 95(5), 1293–1302 (2011). [CrossRef]

6. J. C. Li, Y. Wang, and D. C. Ba, “Characterization of semiconductor surface conductivity by using microscopic four-point technique,” Phys. Procedia 32, 347–355 (2012). [CrossRef]

7. K. Alzoubi, M. M. Hamasha, S. Lu, and B. Sammakia, “Bending fatigue study of sputtered ITO on flexible substrate,” J. Disp. Technol. 7(11), 593–600 (2011). [CrossRef]

8. Z.- Yu, J.- Zhao, F. Xia, Z.- Lin, D.- Zhang, J. Leng, and W. Xue, “Enhanced electrical stability of flexible indium tin oxide films prepared on stripe SiO2 buffer layer-coated polymer substrates by magnetron sputtering,” Appl. Surf. Sci. 257(11), 4807–4810 (2011). [CrossRef]

9. K. A. Sierros, D. S. Hecht, D. A. Banerjee, N. J. Morris, L. Hu, G. C. Irvin, R. S. Lee, and D. R. Cairns, “Durable transparent carbon nanotube films for flexible device components,” Thin Solid Films 518(23), 6977–6983 (2010). [CrossRef]

10. J. R. Lee, D. Y. Lee, D. G. Kim, G. H. Lee, Y. D. Kim, and P. K. Song, “Characteristics of ITO films deposited on a PET substrate under various deposition conditions,” Met. Mater. Int. 14(6), 745–751 (2008). [CrossRef]

11. M. M. Hamasha, K. Alzoubi, S. Lu, and S. B. Desu, “Durability study on sputtered indium tin oxide thin film on poly ethylene terephthalate substrate,” Thin Solid Films 519(18), 6033–6038 (2011). [CrossRef]

12. J.-W. Park, G. Kim, S.-H. Lee, E.-H. Kim, and G.-H. Lee, “The effect of film microstructures on cracking of transparent conductive oxide (TCO) coatings on polymer substrates,” Surf. Coat. Technol. 205(3), 915–921 (2010). [CrossRef]

13. O. Ourida and B. M. Said, “Influence of the blend concentration of P3HT: PCBM in the performances of BHJ solar cells,” SATRESET 1, 90–92 (2011).

14. J. P. Rakotoniana, O. Breitenstein, and M. Langenkamp, “Localization of weak heat sources in electronic devices using highly sensitive lock-in thermography,” Mater. Sci. Eng. B Adv. Funct. Solid State Mater. B 91–92, 481–485 (2002).

15. A. Wolf, P. Pohl, and R. Brendel, “Thermophysical analysis of thin films by lock-in thermography,” J. Appl. Phys. 96(11), 6306–6312 (2004). [CrossRef]

16. O. Kunz, J. Wong, J. Janssens, J. Bauer, O. Breitenstein, and A. G. Aberle, “Shunting problems due to sub-micron pinholes in evaporated solid-phase crystallised poly-Si thin-film solar cells on glass,” Prog. Photovolt. Res. Appl. 17(1), 35–46 (2009). [CrossRef]

17. H. Straube, J.-M. Wagner, J. Schneider, and O. Breitenstein, “Quantitative evaluation of loss mechanisms in thin film solar cells using lock-in thermography,” J. Appl. Phys. 110(8), 084513 (2011). [CrossRef]

18. K. Leppänen, B. Augustine, J. Saarela, R. Myllylä, and T. Fabritius, “Breaking mechanism of indium tin oxide and its effect on organic photovoltaic cells,” Sol. Energy Mater. Sol. Cells 117, 512–518 (2013). [CrossRef]

Dimension	Value
Sample length (total)	305 mm ± 0.5 mm
Sample width	35 mm ± 0.5 mm
Sample bending cylinder diameters	15,20, 25,30 and 53 mm

Step	Name of the IR-image	Data pixels	Size (mm²)	Description of the image
0	Original IR-image	240 × 240	153 × 153	The sample, not powered reference sample and some surrounding area
1	T_extracted(x,y)	157 × 55	100 × 35	IR-map of extracted sample between the probes
2	T_p(x,y)	157 × 55	100 × 35	IR-map of temperature rise of the sample
3	T_n(x,y)	125 × 49	80 × 31	IR-map of the ROI

Dimension	Value
Sample length (total)	305 mm ± 0.5 mm
Sample width	35 mm ± 0.5 mm
Sample bending cylinder diameters	15,20, 25,30 and 53 mm

Step	Name of the IR-image	Data pixels	Size (mm²)	Description of the image
0	Original IR-image	240 × 240	153 × 153	The sample, not powered reference sample and some surrounding area
1	T_extracted(x,y)	157 × 55	100 × 35	IR-map of extracted sample between the probes
2	T_p(x,y)	157 × 55	100 × 35	IR-map of temperature rise of the sample
3	T_n(x,y)	125 × 49	80 × 31	IR-map of the ROI

Electrical heating synchronized with IR imaging to determine thin film defects

Abstract

1. Introduction

2. Materials and methods

2.1 Indium tin oxide (ITO) samples and environmental conditions

2.2 Synchronized thermography, measurement setup and devices

2.3 IR-image pre-handling

2.4 Defining major and minor breakages

2.5 Data processing functions to qualify minor breakages and undamaged samples

2.6 Data processing functions to qualify major breakages

2.7 Data processing functions to mark the breakages in the sample map

3. Results

3.1 Automatic evaluating and breakage area visualization method

3.2 Results of evaluated ITO samples

3.3 Data processing time

4. Discussion

5. Conclusions

Acknowledgments

References and links

Cited By

Figures (12)

Tables (2)

Equations (12)

Optics Express