Ionic Winds to Cool Computer Chips

A method to cool electronic components using a composite heat sink is already noted.

Here we shall discuss in detail, an interesting idea to cool electronic components like computer chips shaped over the last two years by Prof. Timothy Fisher and Prof. Suresh V. Garimella along with their students at the Cooling Technologies Research Center of the Purdue University. The idea is demonstrated successfully using prototype experiments, which has been discussed in a recent paper published in the September 2007 issue of the Journal of Applied Physics [abstract].

In standard forced convection cooling applications, a cooling fluid is blown over the hot surface to take away the heat and achieve cooling. In the case of electronics, say, a computer motherboard, the cooling fluid can be air blown at room temperature over the back side of the hot electronics board to take away the heat so that the electronics operate well below their critical safe operational reliability temperature (usually about 90 degree C).

The forced convection cooling is characterized and measured using a convection heat transfer coefficient that is defined through the familiar Newton�s law of cooling of a hot (solid) body, when cold fluid runs over it, given by

h = q" / delta T ... Eq. (1)

where the numerator on the RHS is the heat flux measured in Watts per square meter. The denominator in the present situation is the local temperature difference between the hot wall surface and the local reference fluid temperature (in the Fig. 1 below, far away in the vertical Y direction from the hot horizontal surface over which the flow persists).

Obviously, for a fixed heat flux - which is a good assumption for this situation where there is a finite fixed electronics sitting on the hot surface, releasing a fixed heat loss during operation of the device - lesser the temperature difference, higher the heat transfer coefficient. This means, for a fixed reference temperature of the fluid (room temperature air, in this case), lesser temperature difference translates to lesser hot surface temperature, telling us the forced convection cooling achieved is good.

One easy way of doing this is by blowing high speed cooling fluid (air, in this case). Because, as explained in an earlier essay, over the years it has been experimentally and theoretically proven that the single phase forced convection heat transfer coefficient increases with increase in local fluid speeds (mass flow rate). In other words, the degree of cooling achieved using this forced convection strongly depends on the local (near the hot surface) flow rate of the fluid.

However, there is a snag. In such a forced convection cooling configuration, while flowing over the hot surface, a small layer of the fluid gets stuck locally everywhere along the hot surface and is reduced of their speed as the hot surface is stationary. This results in a transverse fluid velocity profile that look more parabolic (when the bulk flow is laminar and not turbulent) as shown in the bottom left diagram in Figure 1, instead of, say, an identical large velocity value everywhere along the vertical y direction. This lower velocity near the hot surface obviously reduces the cooling potential of the bulk flow because of relatively lesser convection heat transfer coefficient.

So, if one were to somehow artificially blow more air very near the hot surface at any local region along the bulk flow direction, then locally at that point one would anticipate a velocity profile that would resemble something similar to that shown in the bottom right of Figure 1. This increased local air flow would result in possible increased local heat transfer coefficient. This is precisely what the researchers have done and report in this paper. And they bring in this local "wind" using a small scale corona discharge - sort of a mini lightning in the air flow, which gets doused by ionizing the surrounding air before it sustains an arc.

A quick digression into corona. A corona discharge, as we have read in our high school science book, results when a high electrical potential difference (hundreds to thousands volts) is applied between a sharp object (pin, wire or electrode) and a relatively blunt object (the collecting electrode) kept within millimeters from the sharp object. Such a configuration is shown in the top diagram of Figure 1. Due to the large surface area around the sharp object, the electric field "spills out" and ionizes the nearby neutral molecules by knocking off a few electrons from them. [According to Wikipedia link for corona discharge, the initially the neutral air near the sharp object is ionized by an exogenous environmental event (for example, as the result of a photon interaction), to create a positive ion and a free electron]. These knocked off electrons (which have a much higher charge/mass ratio and so are accelerated to a higher velocity) in turn accelerate and ionize further regions of the air creating an electron avalanche. The blunt electrode attracts the ion species created in this series of avalanches thus completing the circuit.

Figure 2

It is also established that the spacing between the electrodes determine whether ions are generated by corona discharge or electron field emission. Recent work [2, 3] suggest a 10 micro-meter gap distance or less for the impacting electrons to tunnel directly from the surface of the cathode into the atmosphere (see Figure 1 again, where the electrodes are separated by a distance of 10 micro-meter). The resulting local ionic wind is a secondary wind superposed over the existing bulk flow of air and influences the velocity and thermal boundary layer near the hot surface leading to an enhanced cooling of the surface. This is because, use of only microscale ionic winds is insufficient to cool the entire electronics system of say, a computer. Such ionic winds cannot sustain the necessary pressure drop across the entire system of electronics. In the design proposed in this paper, air blown with a fan is still used to cool the electronics. The "ionic wind engines" are supposed to be placed selectively for achieving local (spatial) heat transfer enhancement.

In the experiments conducted, the heat input to the hot surface of the chip was 4.3 Watts. The temperature of the cooling fluid near the local zone of corona is measured using an infra red camera device. A sample image is shown in Figure 2. indicating that the maximum cooling occurs upstream of the collecting electrode near the corona wire. The observed maximum temperature decreases relative to baseline conditions (with baseline defined as the absence of corona) were 27 C for u_bulk=0.2 m/ s and 25 C for u_bulk=0.3 m/ s. Initial cooling occurs near the corona wire, but as the system approaches steady state, both upstream and downstream are influenced by the ionic wind, as can be seen in the bottom right image taken at 10 minutes after the ionic winds were initiated.

To quantify the enhancement in the heat transfer, a slightly modified heat transfer coefficient from that of Eq. (1) is used in the paper, wherein the radiation effects are first subtracted as in

h(x) = q" (heater) - q"(wall, radiation) / T(wall, x) - T(reference) .... Eq. (2).

Using the above definition for measuring the local (at a particular x in Figure 1) heat transfer coefficient, the experiments showed a maximum value for the ratio

h(x, with corona) / h(x, without corona) = 2.6 ... Eq. (3)

near the sharp electrode (just above the collecting electrode in Figure 2), when current = 5.2 micro-Amperes and potential difference = 3297 Volts. Computer simulations presented in this paper have shown that a microscale ionic wind can enhance the local convection heat transfer coefficient of a 1 m/s bulk flow by approximately 50%.

Now for a technical aside for the interested readers. One ideally would wonder (at least I did, when I first browsed this work) when an ionic wind is created using a corona, the resulting plasma should be hot by itself. How is it then cooling the hot surface it is flowing over? Ideally, the generated ionic wind should have enough local kinetic energy to accelerate the convection from the hot surface but the wind by itself should not release enough electrical Joule heating so that it is already hot enough to receive convection enthalpy from the hot surface. The magnitude of the corona current seems to play a role in this aspect. To quote from the paper (a portion only under the fair copyright use)

When the corona current is increased by an order of magnitude, the temperature drop only increases by a factor of 4, and when the corona current increases by two orders of magnitude the temperature drop at the wire only increases by a factor of 6.5. One might expect that increasing the corona current by orders of magnitude would increase the body force and produce a similar effect on heat transfer. However, two reasons exist to explain why the relative impact on heat transfer is moderate. First, Joule heating of the air by the corona current also increases with corona current and decreases the relative heat transfer between the flow and the flat plate, though this effect is likely small based on the ratio of heater power to corona power. Second, the relationship between flow and heat transfer is not linear. Based on a laminar flow, flat plate assumption, the heat transfer rate should be proportional to the fourth-root of ion concentration. Based on the foregoing observations, we cannot conclusively ascertain whether the observed effects are due to the magnitude of the corona alone, or the interaction of the corona and the bulk flow.

Interesting.

To sum up, elsewhere in aerospace engineering, using ionic winds in the presence of a bulk flow to modulate an external boundary layer has been in vogue. Early experiments were performed by Velkoff and Godfrey [4] using an array of corona wires aligned with the flow and extended above a flat plate, which was also their collecting electrode. An increase in the Nusselt number for low-velocity bulk flows was shown but for higher velocities, the ionic wind impact was less. The microscale ionic wind devices discussed here enhances the bulk forced convection cooling over a local (spatial) hot spot caused by local high density of electronics. They are an attractive technology for hot-spot thermal management as they can be fabricated on an electronic chip, on a spreader, or on the skin of a computer notebook.

Update May 2009: Recent improvement on this technology reported by Technology Review - A Laptop Cooled with Ionic Wind

References

[1]Go, D., Garimella, S., Fisher, T., & Mongia, R. (2007). Ionic winds for locally enhanced cooling Journal of Applied Physics, 102 (5) DOI: 10.1063/1.2776164.

[2] M. S. Peterson, W. Zhang, T. S. Fisher, and S. V. Garimella, Plasma Sources Sci. Technol. 14, 654 (2005).

[3] W. Zhang, T. S. Fisher, and S. V. Garimella, J. Appl. Phys. 96, 6066 (2004). [ISI]

[4] H. R. Velkoff and R. Godfrey, J. Heat Transfer 101,157 (1979). [Inspec]

Related posts:

1. How to quickly cool a bottle of drink using seven equations
2. Composite Heat Sinks for Cooling Electronics
3. Free and Paid Convection
4. Free Convection and the Rayleigh Number
5. Objectives of Thermodynamics and Heat Transfer