Seed Crystal, Dash Necking and Crystal Orientation: How a Czochralski Ingot Starts

Where growth begins

A Czochralski pull starts with high-purity polysilicon melted in a fused-quartz crucible, held just above silicon's melting point of about 1,414 °C under an inert gas. Nothing in that melt tells the silicon how to arrange its atoms. Left to cool on its own it would freeze into many grains with random orientations. The seed crystal is what turns a pot of molten silicon into a single crystal.

The seed is a small rod of single-crystal silicon cut so that a known crystal direction runs along its axis. It is lowered until it touches the melt surface, allowed to equilibrate, and then slowly raised while the seed and crucible rotate. Silicon freezes onto the seed and copies its lattice, so the orientation of a two-metre ingot is decided by a piece of silicon a few centimetres long.

Why dislocations appear at the seed

Dipping a cold seed into silicon near its melting point is a thermal shock. Teaching material from the University of Kiel notes that this shock will almost certainly create dislocations, and the seed may already carry some of its own. A dislocation is a line defect in the crystal lattice. If it is allowed to run along the growth direction it propagates into the body of the ingot, and a crystal that loses its dislocation-free structure is of little use for high-efficiency cells or chips.

W. C. Dash of General Electric studied where these dislocations come from and published the remedy in the Journal of Applied Physics in 1959. His paper, Growth of Silicon Crystals Free from Dislocations, is still the reference that the process is named after.

Dash necking: grow thin, then widen

The remedy is to pull fast enough, for a short distance, that the crystal narrows to a thin neck. In a thin neck the dislocations are not aligned with the growth direction, so they run out to the surface and terminate instead of continuing down the crystal. Once a length of dislocation-free neck has been grown, the pull rate and heater power are adjusted so the crystal widens.

The neck is not usable material, and it has to carry the full weight of the finished ingot, which is one reason neck length and diameter are tightly controlled recipe parameters on large modern pullers. Exact neck dimensions vary by machine, crystal size and supplier, and published values should not be treated as universal.

• Seed dip and thermal equilibration
• Neck: thin, fast section that clears dislocations
• Crown and shoulder: controlled widening to target diameter
• Body: constant-diameter section that becomes wafers
• Tail: tapered end that lets the crystal leave the melt without shock

Crown, shoulder, body and tail

After the neck, the crystal is widened through the crown and shoulder until it reaches the target diameter. The body is then grown at constant diameter, which is the section later squared and sliced. At the end, the diameter is reduced into a tail cone so the crystal can separate from the melt without a sudden temperature change that could generate dislocations moving back up the body.

The seed end, shoulder and tail are cut away as crop ends. They are solid silicon and are often cleaned and remelted, unlike kerf powder lost in slicing.

Crystal orientation: <100> vs <111>

Crystal orientation is written with Miller indices. <100> and <111> name directions in the cubic silicon lattice, and (100) or (111) name the wafer surface that results when the ingot is sliced perpendicular to its axis. PV-Manufacturing.org gives <100> as the industry-standard orientation for Czochralski silicon.

For solar the reason is texturing. Alkaline etchants attack silicon planes at different rates, so a (100) surface develops a random field of small pyramids bounded by slower-etching (111) faces. Those pyramids reduce reflection and trap light. A (111) surface does not texture this way in alkaline solution, so it is not used for mainstream solar wafers. <111> material appears in some power-device and specialty uses, and Dash's 1959 work grew dislocation-free crystals along both axes.

Orientation is set by the seed and cannot be changed after growth. A wafer specification should therefore state orientation and its tolerance explicitly, and an ingot supplier should be able to show how seeds are qualified and tracked.

Why the crucible still matters here

Necking protects the start of the crystal, but dislocation-free growth can still be lost later in the run. Particles shed from the crucible wall, including cristobalite flakes or material released from bubbles in a poor inner layer, can reach the growth interface and end dislocation-free structure partway through the body. That is why crucible inner-surface quality and devitrification control sit alongside pull control in any discussion of ingot yield.

What buyers should ask an ingot supplier

Buyers rarely need neck dimensions, but they do need evidence that growth is controlled.

• Stated orientation and tolerance, and how it is verified
• How dislocation-free growth is confirmed and how a structure loss is handled
• How seeds are sourced, qualified and traced to ingots
• What crucible grade and inner-surface treatment is used
• How crop ends are cut, recorded and recycled