Molybdenum Components for Semiconductor Crystal Growth and Diffusion: Selection and Contamination Control

The interior of a crystal growth furnace or a diffusion furnace is often treated as a matter of thermal field design, yet it is the metal structural parts that actually separate heat, atmosphere and material. Molybdenum support columns, heat shields and alloy fixtures sit in the furnace for long periods under radiant heat and heating-cooling cycles, and their condition usually shows up in the product as crystal defects, within-wafer anomalies or particle contamination; by the time traceability begins, the clues are hard to reconstruct. Managing these parts as components with a defined service life and acceptance conditions costs less than investigating an entire furnace after a problem appears.

The Service Envelope the Furnace Environment Sets for Molybdenum Parts

Czochralski, zone melting and epitaxy equipment differ in atmosphere, but they share high temperature with low pressure or a protective gas. Molybdenum has a melting point of about 2620℃ and a density of about 10.2g/cm³, and retains good high-temperature strength in vacuum and inert atmospheres, which makes it suitable for load-bearing and reflecting functions in a hot zone. Its limits are equally clear: above about 500–600℃, trace oxygen and moisture oxidize it to MoO₃, which volatilizes and thins the surface layer by layer; in long contact with carbon-based insulation, molybdenum carbide may form at the interface and make the area locally brittle. In engineering layout, molybdenum parts are usually kept away from the free surface of the melt and out of the path of condensate falling from cold zones, and the furnace dew point is controlled.

A hydrogen atmosphere is often assumed to be completely safe. Hydrogen is itself reducing, but trace moisture in industrial hydrogen still reacts with molybdenum at high temperature, so dew point control matters more than the type of atmosphere. In vacuum furnaces, watch the venting stage when the furnace is broken open: at the moment a hot molybdenum part meets air, the oxidation rate is clearly higher than at room temperature. Understanding these conditions is a basic issue to settle before selection, not something to make up for after a part fails.

Another layer of boundary comes from assembly relationships. Molybdenum parts rarely work alone; they always contact graphite, quartz, insulation felt or metal flanges. Differences in thermal expansion between dissimilar materials at high temperature turn into contact stress, and it is not rare for a quartz part to be scratched by a molybdenum part or a graphite part to be cracked by a molybdenum screw. Allowing clearance for thermal expansion in the design solves the problem better than raising the material grade.

In actual operation, molybdenum parts do not carry a constant load. Each charging, melt joining and shutdown brings a temperature cycle, and structural parts expand and contract through it while bolt preload fluctuates. With too much preload, thin plates are pressed into dimples; with too little, vibration and thermal cycling gradually loosen the connection. Giving a torque range, specifying the tightening sequence and recording the number of assembly cycles in the assembly procedure reduces this kind of slow failure.

Supports, Shielding and Fasteners in the Crystal Growth Hot Zone

Molybdenum parts in the hot zone fall roughly into three categories: load-bearing supports and pulling connections, heat shields that control heat flow, and fasteners that hold graphite and insulation felt in place. The indicators that matter differ between the three, and using one acceptance basis for all of them makes it easy to miss the critical items.

Inspection and acceptance basis for the indicators above: electrode products are controlled to YS/T 1312-2019 Molybdenum Electrodes, molybdenum rods follow ASTM B387 / GB/T 17792, molybdenum plates and discs follow ASTM B386 / GB/T 3876, and a certificate of quality is supplied with each shipment so that batches can be traced.

Heat shields are mostly built as stacked thin plates with gaps left between layers to prevent heat from short-circuiting along the contact surface; molybdenum screws and nuts are used at the fixing points, tightened within the specified torque range so that the thin plates are not pressed into dimples. The number of layers and the layer spacing are usually given as a reference range from thermal field simulation, and the actual configuration still has to be adjusted to the furnace type and power; typical values are subject to actual operating conditions and the contract.

The effect of machining accuracy is often underestimated. If the edge of a heat shield has burrs or waves, the local temperature field shifts and crystal diameter control becomes difficult; if the coaxiality of a support column is out of tolerance, the runout during crucible rotation is transmitted to the melt surface. The flatness, hole positions and surface roughness of molybdenum plates and molybdenum alloy structural parts need to be aligned with the thermal field simulation results at the drawing stage, not corrected during assembly.

Carriers and Baffles in Diffusion and Annealing

Molybdenum parts in a diffusion furnace are mostly carriers and baffles: wafer boat supports, flow equalizing plates and furnace mouth heat shield plates. These parts sit close to the wafers and are highly sensitive to particles and metal contamination. If cutting oil remains on the surface or cleaning is incomplete, it can turn into a contamination source that is hard to remove after heating. A sound approach is to specify the cleaning procedure, the packaging method and the storage time limit after unpacking, and to run an empty-furnace wafer check before the first loading.

Diffusion and annealing temperatures are mostly between 800–1200℃, and some low-pressure processes also involve frequent evacuation and gas filling. Under these conditions the main change in molybdenum parts is slow deformation and deterioration of the surface state rather than sudden fracture. Whether a part can still be used should be judged by whether flatness, hole positions and surface roughness are still within the drawing tolerance, not simply by whether cracks are visible. The support points of the carrier also need attention: too few supports leave slip marks on the wafers under their own weight at high temperature.

Controlling Cleanliness and Metal Contamination

The source of contamination is often not the material itself but the machining and handling steps. Cutting fluid, mold release agents, sweat from gloves and metal chips on the inner wall of a transfer box can all leave iron, nickel, chromium and other elements on the surface of molybdenum parts. At high temperature these elements diffuse into the wafer far faster than molybdenum does, and once they enter the device the electrical performance drifts in ways that are hard to explain. The control philosophy should be to keep contamination out of the machining step rather than rely on cleaning before loading to make up for it.

The cleaning sequence is usually degreasing, acid pickling, deionized water rinsing and drying. Pickling for too long attacks grain boundaries and forms micropits that are hard to see, and these micropits become traps for contamination in later use; incomplete drying leaves water marks, and oxidation at the water marks begins early in the heating cycle. For packaging, use vacuum or clean bags and mark the cleaning date and batch number on the bag; parts that exceed the storage limit should be cleaned again before loading, which is more reliable than judging by experience.

The interval between cleaning and loading also needs managing. If a part is left overnight in an ordinary workshop after unpacking, dust and moisture can re-adsorb on the surface; if multi-layer transfer boxes are not cleaned regularly, debris collected at the bottom falls onto the parts below. Arranging unpacking, assembly and loading within a tight time window, and setting a cleaning cycle for the transfer boxes, is a simple way to control particles.

How Machining Accuracy Propagates into Crystal Quality

The hot zone is a coupled system. A flatness deviation in a heat shield changes local radiative heat exchange, which in turn changes the temperature difference between the crucible edge and center; a verticality deviation in a support column makes the crucible wobble periodically as it rotates, which shows up in the crystal as periodic diameter variation. If parts are tightened forcibly during assembly to eliminate gaps, assembly stress remains inside the structure and is released after heating, changing the positional relationships. Such problems are often invisible in a cold check.

The machining process itself also needs to allow stress relief. When a thick plate is cut directly from the rolled state, residual stress is unevenly distributed between surface and core, and the part tends to warp gradually after machining; the more reliable route is a vacuum anneal after rough machining, followed by finish machining and finishing. If the damaged layer left by wire cutting is not removed, it may oxidize preferentially at high temperature and form pitting. The surface treatment method should be stated on the drawing so that the supplier does not choose blasting or polishing by habit.

Pre-Loading Confirmation and Trial Run

Before introducing molybdenum parts into a semiconductor process, each item can be confirmed in a fixed order and the results recorded on the same traveler.

1. Whether the material grade and batch can be traced, and whether the material certificate matches the measured composition;
2. Whether dimensions and geometric tolerances match the drawing, and whether hole positions and mating surfaces have been rechecked;
3. Whether the surface has been degreased, pickled or blasted, and whether the roughness meets particle control requirements;
4. Whether there are burrs at edges and hole openings, and whether they would scratch quartz parts or the wafer boat after assembly;
5. Whether the cleaning water and drying method match, and whether the packaging is vacuum or clean-room bagged;
6. Whether a low-risk batch trial run is arranged for first use, with the thermal field and product performance recorded.

The value of this checklist is not the number of items but that the same basis is used every time. Once the basis is unified, differences between batches become comparable, and when an anomaly appears it is easier to judge whether the cause is the material, the machining or the assembly.

Batch Traceability and Change Management

Traceability of small parts is often neglected. A molybdenum screw is not worth much, but if it comes from a different billet, an anomaly in the whole furnace becomes hard to attribute. A practical approach is to keep four codes: billet number, machining batch, heat treatment batch and dispatch batch, indexable against each other in the ledger. The marking method can be laser marking, electro-etching or a high-temperature label, chosen with regard to service temperature and cleaning method so that the mark does not disappear in use.

Change management matters just as much. Changing the billet supplier, adjusting the heat treatment schedule or switching the type of cleaning agent are changes that cannot be seen on a drawing but may alter the surface state and internal structure. Run a small-batch validation before the change and compare the thermal field and particle data with historical records; that is far more economical than discovering the problem in volume production. Change records should be kept together with the batch ledger for later trace-back.

Conditions to Write into Drawings and Contracts

In purchasing communication, grade, purity, density, grain size, dimensional tolerance, surface roughness, edge treatment and packaging method should all be written into the drawing rather than left as verbal agreements. The usual purity range is 99.95%–99.99%, and density and grain size are selected according to the loads on the part; these parameters are typical values or reference ranges, subject to operating-condition calculation and the contract. The contract should also define the acceptance method, sampling ratio, shipping documents and the procedure for handling anomalies, so that the two sides are not working from different assumptions on arrival.

What semiconductor processes need from molybdenum parts is essentially stability: little deformation and little volatilization at high temperature, few particles shed from the surface, and little variation between batches. Putting these requirements into the design and purchasing stages costs less than replacing parts afterwards.