• No results found

Bonding Setup

3.2 Development and Description of Setup

3.2.1 Bonding Setup

To enable vacuum packaging, one of the key objectives of the new bonding setup is to move the mechanism to apply the force inside the evacuated volume which requires a considerably larger bonding chamber than previously available. The upper part of the former setup could be integrated into the new setup without any major changes as it was intended to be vacuum-tight in its original design phase already. All interfaces are fitted with O-ring seals (see Figure 3.2). The original load glass of thickness 4 mm and diameter 100 mm was replaced with one of thickness 13 mm so it can withstand the atmospheric pressure without bending or the risk of breaking once the setup is evacuated. Borofloat glass is used for the load glass because it has very good transmission at the laser wavelength of 940 nm (see Figure 3.18).

The lower part of the bonding setup was completely redesigned to accommodate a pneumatic cylinder to apply the required force for bonding. One important design rule when constructing a vacuum apparatus is to keep its dimensions as small as possible to facilitate the evacuation of the chamber and to minimise the surface area where water molecules can adhere to. A review of the options available to apply this force has shown that pneumatic cylinders are the ideal solution. They are the most compact devices which can still exert forces in the order of a few hundred Newton and they are inexpensive. Other alternatives like vertical lift stages or piezo actuators are expensive and either rather large or can only apply forces of a few Newton which would not suffice for wafer-level bonding. The only risk of using pneumatic cylinders, a device which is operated with compressed air, inside a vacuum is a potential for slight air leaks. A compact double action pneumatic cylinder from SMC (CDQMB25-30) with a stroke of 30 mm and a maximum force of 500 N was chosen, which should be sufficient for bonding up to 100 samples at a time. In case higher forces should be required the cylinder can always be exchanged for a more powerful one of similar outer dimensions

which would still fit into the chamber. A sketch of the bonding setup is shown in Figure 3.3.

Figure 3.3: Sketches of new bonding setup with enlarged bonding chamber and integrated pneumatic cylinder

The pneumatic cylinder (a) is centred in the lower part of the bonding chamber (b), which is basically a hollow aluminium cylinder with a solid base. To further enhance the benefit of localised laser heating the samples are placed onto a copper block (c) for bonding to restrict the lateral heat flow and to keep the temperature in the sensitive device area in the centre of the package as low as possible. Active water-cooling is integrated into the copper block to keep it at a constant temperature even during the extended time period of several tens of minutes of some bonding processes. In the previous setup the upper part of the bonding setup (d) was used to exert the bonding force and the copper block as part of the lower one was locked into position. With the new setup, the copper block, which is mounted on top of the pneumatic cylinder, is used to press the samples against the load glass (e) inside the upper part of the setup (lid).

The lid is sealed (pressed) flush against the bonding chamber and is held in a fixed position by a linear stage, and in case of bonding in vacuum additionally by the atmospheric pressure acting on top of the lid. Hence, the bonding force is independent of the level of vacuum inside the chamber as only the force acting onto the lid is increased once the chamber is pumped down; the force acting onto the samples remains unchanged.

(a)

(b) (d) (e)

(h) (f)

(g) (c)

To measure the applied force, a miniature compression load cell (CDF-0.5kN) supplied by Omni Instruments with a range of 0-0.5 kN (f) is placed between the copper block and the pneumatic cylinder. The load cell is connected to a load cell indicator (Tracker 243 by Data track) which displays the load with an accuracy of ±0.01 kg. The load cell is sandwiched between two aluminium plates; one containing a recess to centre it and the other one is to provide a flat, hard surface to press against to minimise measuring errors. Prior experiments have shown that for high quality bonds with full contact of the interfaces along the bond line it is essential that the load glass and the copper block are absolutely parallel. Since several components (pneumatic cylinder, load cell with aluminium plates and copper block) are stacked on top of each other there is a risk of a slight tilt between the load glass and the copper heat sink if the connection between all components is rigid. Hence, the components are not bolted together tightly, but held in place and guided by rods to permit a slight play (see Figure 3.4). In addition, a pliable cushion material (g) is placed between the copper block and the aluminium plate above the load cell which compensates for any tilt in the assembly ensuring that the two are aligned in parallel.

Figure 3.4: Photograph and schematic sketch of copper block and load cell on pneumatic cylinder

All the feed lines for cooling water, compressed air, etc. are put through feedthroughs (h) in the bottom of the bonding chamber. Instead of sealing those in directly vacuum-tight coupling interfaces were designed which allow easy replacement of the feed lines in case they break without the need for elaborate repairs. For that reason four holes of diameter 25 mm are drilled into the base of the bonding chamber; one of them is fitted with a DN-ISO FN16 flange to which the vacuum pump is connected and the remaining ones are sealed with customised plugs machined out of brass. For the feed lines for the cooling water and compressed air one of these plugs is equipped with four

through-Load cell

Guiding rods Aluminium

plates

Pneumatic cylinder Copper block

Copper block Cushion

holes fitted with push-in fittings (SMC) on either side (Figure 3.5). The connection pipes on the in- and outside of the bonding chamber can easily be connected and replaced if needed. According to the supplier (SMC) the fittings seal well down to pressures of at least 10 mbar which should suffice for this purpose.

Figure 3.5: Photograph of bonding setup with vacuum-sealed feed lines

A second plug is used for the electrical connections to the load cell and the wires are sealed using a vacuum compatible epoxy (Torr Seal, Varian). The remaining three holes are used for the electrical connections of thermocouples. Instead of sealing the thermocouples in directly interchangeable couplings with connectors on either side were made (Figure 3.6). The wires for the electrical connection of the thermocouple were fed through the centre hole of such a threaded push-in fitting, as already used for the pipe connections, and sealed in using the vacuum epoxy. These couplings have two major advantages rather than sealing in the thermocouples directly: if a cable break occurs within the coupling it can simply be unscrewed and replaced by a new one, and also in case a thermocouple breaks it can be exchanged by simply unplugging it. The fourth plug is a blank which is left in case any further feedthroughs, for example nitrogen for back-filling the bonding chamber, are needed in future.

Couplings for TCs

Water + air feed lines Load cell cable

Vacuum port

Figure 3.6: Photograph of vacuum-tight, interchangeable coupling for thermocouple

A rotary vacuum pump (DUO 5M) and a Pirani gauge (PT441 930-T) by Pfeiffer Vacuum Ltd. are used to evacuate the chamber and to monitor the level of vacuum in the system. A bleed valve is fitted to break the vacuum prior to re-opening the chamber after bonding. Initial vacuum tests of the setup have shown that a final pressure of around 3x10-2 mbar was achieved with the pneumatic cylinder moving up and down inside and the water-cooling switched on. Despite all the feedthroughs and potential, slight air and water leaks – fittings only specified for vacuum down to ~10 mbar – a final pressure inside the bonding chamber close to the limit of rotary pumps (typically around <5x10-3 mbar) was achieved. Thus a very flexible and modular bonding setup has been developed where various, interchangeable connections are fed into a vacuum chamber without breaking the vacuum. The entire bonding chamber is placed onto a compact lab jack (LJ750/M, Thorlabs) so its height can be adjusted into the focal position below the scan head of the laser system.