Cooling Rate After Reflow: What It Sets in the Joint

The cooling segment is the part of a reflow profile that most shops never specify. The ramp, the soak and the peak are tuned and recorded, and then the board is allowed to cool at whatever the last few zones and the ambient air produce.

That is a missed control, because cooling sets two things that matter: the grain structure of the solidified joint and the mechanical stress the assembly carries away from the oven. Both are visible in the field as fatigue life and as cracks that appear without a thermal cause.

Why the Cooling Segment Is Part of the Recipe

Solidification happens over a narrow temperature range, and the rate at which the joint passes through it decides how the grains form. Fast cooling produces many small grains, slow cooling produces fewer large ones, and the boundaries between those grains are where fatigue cracks begin.

The rate also decides how long the joint stays hot enough for the intermetallic layer to keep growing. Every extra second above about 150 °C adds thickness to that layer, and the layer is brittle, so the joint that spends longer cooling is the joint that is more prone to fracture at the interface. That layer is also why a joint reworked twice behaves differently from one reworked once.

The Rates in Use

Most lead-free profiles cool at two to six degrees per second between the peak and the solidus, and a rate below about one degree per second is slow enough to allow coarse grain growth and a thicker intermetallic layer.

Faster than about six degrees per second begins to look like thermal shock, particularly for large ceramic capacitors and glass-bodied devices, which have their own rate limits that are usually more restrictive than the solder requires. Those limits are worth collecting into one table, because they are scattered across component datasheets. The profile has to satisfy the most sensitive component on the board rather than the majority.

Grain Structure and Joint Strength

A fine grain structure distributes strain across many boundaries rather than concentrating it on a few, and joints with finer grains generally survive thermal cycling better. This is one of the reasons a controlled cooling profile improves the life of a joint without changing the alloy, and the finished structure is the product of the alloy, the pad finish and the cooling rate together.

The effect is easy to demonstrate and hard to see in production, because a section at high magnification is needed to compare grains. It becomes visible in the laboratory as a difference in cycles to failure, which is the measurement that justifies the control in the first place, as discussed in the notes on joint fatigue.

Intermetallic Growth and Time at Temperature

The intermetallic layer grows with the square root of time at temperature, so the largest part of its growth happens during the time the joint spends near the peak rather than during the ramp. Cooling rate contributes by extending or shortening the tail of that exposure.

A joint cooled slowly from the peak spends longer in the range where growth is fast, and the resulting layer is thicker and more prone to brittle fracture. The relationship between that layer and joint ageing is described in the notes on intermetallic growth.

Thermal Shock and What It Damages

Cooling too quickly produces a temperature difference between the surface of a component and its interior, and the differential expansion that follows can crack a ceramic body, lift a bond wire or delaminate a package. The damage is mechanical and it appears immediately, which is what distinguishes it from a slow cooling problem that never cracks a body.

The components at risk are the ones with a large coefficient of thermal expansion bonded to a material with a small one. Multilayer ceramic capacitors and glass diodes are the usual victims, and their suppliers publish a maximum cooling rate that should be taken as the limit for the profile.

Warpage and the Cooling Gradient

A board cools from the surfaces inwards, and the copper distribution decides how unevenly that happens. A board with a large copper plane on one side and sparse copper on the other cools asymmetrically and comes out of the oven bowed.

The effect is worst where the copper is unbalanced by design, and it is made worse by a fast cooling rate because the gradient across the board is larger. A board that is already close to its flatness limit can be pushed over it by the cooling segment alone. The relationship between cooling, copper balance and the resulting shape is covered in the notes on warpage and reflow.

How the Oven Produces Cooling

Cooling is delivered by the last zones of the oven, which run at low temperature with fans, and sometimes by a water-cooled section. The conveyor speed sets the dwell in each zone, so the cooling rate is changed by changing the speed or the zone temperature rather than by a dedicated control in most ovens. An oven with a separate cooling section gives more freedom, and its set points belong in the same record as the heating zones.

That coupling is why cooling is often an accident of the ramp settings. Where cooling matters, the last two zones have their own set points and the profile is measured with a thermocouple attached to the product, not read from the display, which is the practice described in the notes on thermocouple placement.

Measuring and Recording the Rate

The rate is calculated from the recorded profile as the slope between the peak and the temperature at which the joint is solid, and it should be reported in degrees per second with the thermocouple position stated. A rate calculated over the whole descent from peak to ambient hides the part that matters. The slope is best reported between the peak and about 150 degrees, where the joint is still changing.

Two measurements, one on a large thermal mass and one on a small component, show whether the board is cooling evenly. A large difference between them means the cooling is being driven by mass rather than by the oven, and the rate on the small part is the one that governs the component limits.

Setting a Limit for a Product

The limit should be taken as a window rather than a maximum. Too slow allows grain growth and intermetallic thickening, while too fast risks mechanical damage, so the specification has both a floor and a ceiling.

A window of two to four degrees per second satisfies most assemblies, and it should be narrowed only where a component supplier or a reliability requirement demands it. Once the window is set, it is verified in the same profile run that confirms the peak and the time above liquidus, so it costs nothing extra to control. The window should be printed on the profile sheet next to the peak and the time above liquidus, so that all three are reviewed together.

FAQ

Does cooling rate really matter if the joints look good? It does. The joints look good either way, and the difference appears in grain size, intermetallic thickness and the fatigue life measured in cycles rather than in appearance.

What is a safe cooling rate? Two to six degrees per second suits most lead-free work. Below about one degree per second the joint coarsens, and above six the risk of thermal shock to ceramic parts rises.

How is the rate controlled? Through the last zones of the oven and the conveyor speed, then confirmed with a thermocouple on the product. The oven display alone does not report it.