A Statistical Framework for Laser-Induced Subsurface Modification and Cracking in 4H-SiC

Technical Summary: A Statistical Framework for Laser-Induced Subsurface Modification and Cracking in 4H-SiC

Problem Statement
The scaling of 4H-SiC technology for power electronics and quantum applications is bottlenecked by substrate manufacturing. Conventional mechanical sawing causes significant kerf loss, tool wear, and subsurface damage due to the material's extreme hardness and brittleness. Laser slicing (LS) offers a non-contact alternative by inducing controlled subsurface fractures. However, optimizing LS is challenging because the target is a continuous, coplanar field of overlapping basal cracks formed beneath the surface. Unlike surface ablation, where outcomes are directly observable, subsurface modification is buried, defect-sensitive, and statistically variable. Nominally identical laser parameters can yield different outcomes (modification only, crack onset, or stable cracking) due to local variations in defects, doping, and stress. Current process definitions often rely on deterministic thresholds, which fail to capture the probabilistic nature of brittle fracture and laser-induced damage in transparent semiconductors.

Methodology
The authors developed an event-based statistical framework to characterize laser-induced subsurface modification and cracking in commercial n-type 4H-SiC wafers (4° off-axis).

• Experimental Setup: Nanosecond laser pulses (1064 nm) were focused through high-NA optics to varying depths ($D$) below the surface. The study systematically varied three parameters: nominal fluence ($F_{nom}$), pulse-to-pulse pitch ($p$), and focal depth.
• Event Classification: Instead of measuring continuous crack lengths, the authors classified image-derived outcomes into three binary events:
  1. Modification (mod): Presence of a detectable subsurface modification.
  2. Crack Onset (onset): Appearance of isolated crack morphology.
  3. Stable Crack (stable): Formation of a persistent, connected crack extending along the writing direction.
• Statistical Analysis: The probability of each event occurring at a given fluence was modeled using a two-parameter Weibull distribution. Key metrics included the midpoint fluence ($F_{50}$), the shape parameter ($\beta$) indicating transition steepness, and the bootstrap confidence interval for uncertainty quantification.
• Morphological Validation: Cross-sectional Scanning Electron Microscopy (SEM) was employed to correlate statistical event classes with physical morphologies, specifically examining the transition from void-like modifications to wedge-like crack-forming inclusions.

Key Results

1. Pulse Pitch as the Governing Parameter: The pulse pitch ($p$) was identified as the primary factor controlling the transition from modification to stable cracking.
   • High Overlap (Small Pitch, e.g., 1–2 µm): Modification and stable cracking occur within nearly the same fluence range ($\Delta F_{50} \approx 0$ to $4$ J/cm²). The events are tightly coupled; once modification is detected, stable cracking is highly probable.
   • Low Overlap (Large Pitch, e.g., 4–5 µm): A distinct "modification-dominated" regime emerges. There is a substantial fluence gap ($\Delta F_{50}$ up to 48 J/cm²) where detectable subsurface modification occurs, but stable cracking remains improbable.
2. Transition Characteristics: The study revealed that the position of the threshold ($F_{50}$) and the shape of the transition ($\beta$) are independent. For instance, at large pitches, stable cracking can be delayed significantly relative to modification, yet once the threshold is crossed, the transition can be abrupt (high $\beta$). Conversely, at small pitches, the gap is small, but the transition can be broader.
3. Depth Dependence: While increasing focal depth ($D$) generally raised the fluence required for modification, it did not eliminate the stochastic response or the pitch-dependent trends.
4. Morphological Correlation: SEM analysis confirmed that the statistical separation corresponds to physical differences. At large pitches, low-fluence modifications appear as weak, void-like, or elliptical features. Stable cracking only occurs when the morphology evolves into a dense, wedge-like inclusion capable of generating sufficient stress concentration. At small pitches, the wedge-like morphology forms rapidly due to incubation effects and accumulated stress, explaining the tight coupling of events.

Significance and Claims
The paper claims to provide a process-relevant statistical description of laser-induced subsurface modification and cracking, moving beyond deterministic thresholds. By treating modification and cracking as probabilistic responses, the framework captures both systematic parameter trends and the excursions observed under nominally identical conditions.

• Process Optimization: The authors argue that process windows should be defined by the reproducibility of the desired crack response (probability of stable cracking) rather than a single energy threshold.
• Parameter Selection: The study demonstrates that selecting a larger pulse pitch allows for a process window where subsurface modification can be achieved without immediate, uncontrolled cracking, potentially offering more control over the fracture initiation. Conversely, smaller pitches facilitate rapid conversion to stable cracking, which may be desirable for efficient wafer separation if the stochasticity is managed.
• Methodological Contribution: The work establishes a combined statistical and morphological analysis approach that can be applied to other brittle materials where buried laser modification produces probabilistic fracture responses. The framework successfully isolates the dominant pitch-dependent trend while preserving local variability, avoiding the oversimplification of complex physical processes into single deterministic values.