📄 chemistry

Quantitative Evaluation of Silicon Wafer Backside Morphology and Profilometric Roughness by Optical Imaging During TMAH/IPA-Induced Surface Smoothing

This study demonstrates that TMAH-based etching effectively smoothens silicon wafer back surfaces through selective dissolution, with IPA addition accelerating the process kinetics, while validating that optical imaging combined with image analysis serves as a rapid, cost-effective alternative to profilometry for quantitatively monitoring surface roughness evolution.

Original authors: Fatma Betül Akgül Taner, Alican vatansever

Published 2026-09-24

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Fatma Betül Akgül Taner, Alican vatansever

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Silicon wafers are the thin, flat foundations upon which the modern world's electronics are built. While we often think of these chips as perfect, smooth squares, the reality of their manufacture is far more complex. A silicon wafer has two sides: the front, where intricate circuits are drawn, and the back, which is often left rougher and less scrutinized. Yet, this backside matters. Just as a rough road scatters car headlights, a rough silicon surface scatters light and disrupts the delicate electrical signals inside a chip. In advanced manufacturing, engineers need to know exactly how smooth that back surface is, and they need to know quickly. If the surface is too bumpy, it can cause light to bounce off in the wrong direction or create tiny electrical leaks that ruin the device. For years, measuring this roughness has required slow, expensive machines that touch the surface with a physical needle, tracing every bump and valley to create a map of the terrain.

A team of researchers at TUBITAK, a national research institute in Turkey, set out to find a faster, cheaper way to see these changes. They focused on a chemical process used to smooth silicon: washing the back of the wafer in a solution of tetramethylammonium hydroxide, a liquid that eats away at the silicon surface in a controlled way. They wanted to see if they could simply take a picture of the surface and use a computer to tell them how rough it was, without needing the heavy, slow machinery. To test this, they treated silicon wafers in this chemical bath at a warm temperature of 65 degrees Celsius. They ran the process for different lengths of time, from one minute up to twenty minutes, and in some cases, they added a common liquid called isopropyl alcohol to the mix to see if it changed how the surface smoothed out.

The researchers did not just rely on pictures; they also used a standard, high-precision machine to measure the actual height of the bumps and valleys on the surface, giving them a baseline of truth. They then compared these hard numbers to what their computer analysis of the optical images showed. The results were clear and consistent. As the wafers sat in the chemical bath longer, the surface became smoother. The average height of the bumps dropped significantly, and the surface began to reflect more light. In the pure chemical bath, the roughness decreased by nearly forty percent after twenty minutes, and the amount of light bouncing off the surface nearly doubled, rising from nineteen percent to thirty-seven percent. This confirmed that the chemical was successfully polishing the silicon, turning a jagged landscape into a flatter, more reflective plain.

When the researchers added the alcohol to the solution, the story changed slightly, but the outcome remained similar. The surface smoothed out just as well, reaching a final roughness that was almost identical to the pure chemical bath. However, the speed at which it happened was different. With the alcohol, the surface stopped changing much after about ten minutes, settling into a stable state. Without the alcohol, the surface continued to get smoother for the full twenty minutes. This suggests that the alcohol does not change the final destination of the surface's smoothness, but rather acts as a traffic controller, guiding the process to finish its work more quickly and then stopping it from changing further. It is as if the alcohol helps the chemical work more efficiently at the start, allowing the surface to reach its calm state sooner, whereas the pure chemical takes its time to achieve the same result.

The most significant discovery, however, was the link between the simple pictures and the complex measurements. The researchers used a computer program to analyze the images, looking at the size of the visible features and the variation in brightness across the surface. They found a powerful connection: as the features in the pictures grew larger and the brightness became more uniform, the actual physical roughness measured by the heavy machinery dropped. Specifically, the size of the visible patterns in the image was strongly linked to the smoothness of the surface. This means that a simple optical camera, combined with standard image analysis software, can act as a reliable stand-in for the expensive, slow machines. The researchers demonstrated that by looking at a picture, they could predict the surface roughness with high accuracy, offering a way for factories to check their work rapidly and cheaply.

The study concludes that this method of using light and images is not just a rough guess, but a precise tool. It shows that the way a silicon surface looks to a camera directly tells us how it feels to a touch-sensitive probe. By proving that these visual changes track perfectly with the physical changes, the researchers have provided a new, efficient way to monitor the quality of silicon wafers. This approach allows manufacturers to ensure their chips are built on the smoothest possible foundation without slowing down production with lengthy, mechanical checks. The work confirms that sometimes, the best way to measure the invisible texture of a surface is simply to look at it, provided you know exactly what to look for.

Technical Summary: Quantitative Evaluation of Silicon Wafer Backside Morphology and Profilometric Roughness by Optical Imaging During TMAH/IPA-Induced Surface Smoothing

Problem Statement
Silicon wafer backside morphology and surface roughness are critical parameters affecting wafer handling, thin-film deposition, and optical surface properties in semiconductor manufacturing. As surface roughness increases, light scattering and diffraction effects become pronounced, negatively impacting reflectance behavior and measurement accuracy. While dry etching (e.g., DRIE) offers high precision, it often introduces scalloping structures that limit smoothness. Wet chemical etching using Tetramethylammonium hydroxide (TMAH) is a preferred alternative due to its lack of metallic contaminants and controllable anisotropy. However, previous investigations have typically addressed backside roughness, optical behavior, etching kinetics, and surface morphology independently. There is a lack of comprehensive studies that simultaneously correlate profilometric roughness, image-derived morphological descriptors, and optical reflectance within a single experimental framework, particularly regarding the specific influence of Isopropyl Alcohol (IPA) additives on the kinetics and final state of surface smoothing.

Methodology
The study utilized N-type (100) silicon wafers subjected to wet etching at 65 °C under two conditions: pure TMAH (25 wt.%) and TMAH with 30 vol.% IPA. Treatment durations ranged from 1 to 20 minutes. The researchers employed a multi-modal characterization approach:

Profilometry: Surface roughness was quantified using a KLA-Tencor P-17 profilometer, measuring arithmetic average roughness ( $R_a$ ) and root-mean-square roughness ( $R_q$ ) across multiple line scans.
Optical Imaging and Image Analysis: Optical micrographs (500× magnification) were acquired and processed using ImageJ software. Morphological features were segmented using Otsu thresholding to calculate projected area ( $A$ ) and equivalent circular diameter ( $D_{eq}$ ). Additionally, grayscale intensity histograms were generated to determine the standard deviation ( $\sigma_{hist}$ ), serving as a metric for surface heterogeneity.
Optical Reflectance: Reflectance measurements were conducted at 365 nm using a Nanometrics system to assess the optical response to morphological changes.
Statistical Analysis: Pearson correlation analysis was performed to evaluate the relationships between the image-derived morphological parameters and the profilometric roughness values.

Key Results

Surface Smoothing: In both systems, surface roughness decreased progressively with increasing treatment time. In the pure TMAH system, $R_a$ decreased by approximately 39% (from ~1231 Å to ~776 Å) over 20 minutes, while reflectance increased from 19% to 37%.
Effect of IPA: The TMAH/IPA system also reduced roughness (by ~35% to ~823 Å) but exhibited different kinetics. Roughness values in the IPA-containing solution stabilized (reached a plateau) after approximately 10 minutes, whereas the pure TMAH system continued to smooth throughout the 20-minute period. Consequently, the final reflectance for the IPA system plateaued at ~31%, lower than the 37% achieved by pure TMAH.
Morphological Evolution: Image analysis revealed that as treatment time increased, the equivalent diameter ( $D_{eq}$ ) of surface features increased, while the histogram standard deviation ( $\sigma_{hist}$ ) decreased, indicating a transition from a heterogeneous surface with small asperities to a more uniform morphology with larger characteristic features.
Correlations: Strong linear correlations were established between image-derived metrics and profilometric roughness. Specifically, $D_{eq}$ showed a strong negative correlation with $R_q$ ( $R^2 > 0.90$ in both systems), and $\sigma_{hist}$ showed a strong positive correlation with roughness. This confirms that optical image analysis can effectively track surface evolution quantitatively.

Significance and Claims
The paper claims that the addition of IPA primarily influences the kinetics of surface evolution rather than the ultimate roughness level achievable under the tested conditions. The IPA accelerates the initial surface reorganization, leading to an earlier stabilization of the morphological state, whereas pure TMAH drives a continuous smoothing process over a longer duration.

The study demonstrates that TMAH-based etching promotes a surface smoothing mechanism driven by selective dissolution, where regions of higher curvature dissolve more rapidly. The addition of IPA modifies the progression of this process, likely by reducing surface tension and improving wetting, resulting in a more controlled and rapid morphological evolution.

Crucially, the authors assert that optical imaging combined with image-processing-based analysis (specifically using metrics like $D_{eq}$ and $\sigma_{hist}$ ) serves as a rapid and cost-effective complementary approach for wafer-scale surface assessment. While not intended to replace conventional profilometry, this methodology offers a viable tool for monitoring process trends and evaluating silicon wafer backside conditions in semiconductor manufacturing. The strong correlations validate that image-derived morphological parameters can reliably reflect changes in profilometric roughness and optical reflectance.

More like this