Why chips are stacked
Miniaturisation has gained outstanding importance in modern semiconductor technology.
In addition, there is the increased mobility of the end devices - especially with mobile electronics, the form factor plays an increasingly prominent role.
The goal of the semiconductor industry is therefore to integrate performance within the smallest volumes.
In this respect, the process leads from 2D to 3D integration. This involves stacking components in the third dimension. This is achieved by stacking different components on top of each other within a single enclosure. The classic printed circuit board with the semiconductor components located next to each other can thus be integrated into a single housing (chip). Various individual components (dies) are then placed on top of each other on this chip. The technological requirements for such 3D integration are very high and industrial manufacturing standards must be implemented in the solution. This is where the new process for the three-dimensional integration of semiconductor components offers an innovative solution.
The printing process in seven steps
Printed 3D interposers completely dispense with lithographic processes. They are based on 3D printing. It is an additive process in which - contrary to usual printing processes - it is not only about ink coverage, but in which a volume is printed on. The aim is to create a three-dimensional solid.
First, an insulating layer is printed, which is not applied to the surfaces to be contacted. Here, the printing medium is cured immediately after printing by means of suitable radiation (UV light). This first layer also ensures homogenised surface properties on the wafer surface, which can be better controlled, for example, with regard to their adhesion properties. In addition, this first layer is also intended to compensate for the zone of influence of a laser.
After printing this first insulating layer, these contact surfaces are contacted by means of the metallic ink and a first layer of conductive tracks is printed. A jet printing process is used and the printing medium is cured or sintered by means of a directly tracked laser.
Subsequently, another insulating layer is printed, this time leaving the contact areas open at other predetermined points.
Thanks to the layers lying on top of each other and insulated from each other except for the contact surfaces, it is possible for conductive paths to intersect. Once all layers have been printed, the last layer in turn has contact surfaces that are to be used for contacting a component lying on top.
To facilitate contacting, a vertical elevation of the contacts can be printed analogous to the usual bumps by printing metallic ink several times and sintering each time.
Existing methods
Status quo:
Already today, semiconductor components are assembled into stacks.
The challenge:
- Connection - how to connect the different stacked components electrically.
- Heterogeneity - the higher the number and the more varied the geometric and circuit design of the stacked components, the more difficult stacking becomes.
Procedures to date:
Wire bonding:
Wire bonding was initially used to make electrical contact between the components of a stack.
Disadvantages of this process:
- The requirements, such as radii of curvature and accessibility to the contact surfaces, set clear limits.
- The electrical contacting can generate disturbing electrical fields or electrical fields can disturb the contacting.
TSV (Through-Silicon-Via) technology:
In this process, small openings (vias) are etched into the thin wafer using a deep-etch process (e.g. BOSCH process). These are subsequently filled with metal. With this method, it is now possible to lead an electrical contact from the front side of a wafer to its rear side in an extremely short path.
Disadvantages of this process:
- The geometric placement of the contact points is complex.
- This method is often not suitable for heterogeneous stacks.
- Thermal management is challenging sometimes.
Si or glass interposer:
Currently, Si or glass interposers are used for rewiring stacked wafers. In this way, contact points can be geometrically relocated or supplemented using lithographically applied conductor tracks.
Disadvantages of this process:
- Lithographic processes are costly.
- Lithographic processes are inflexible.
- Time-to-market is high due to the effort involved.
- Heterogeneous stacks - stacks of several different components in unequal geometrical layout are difficult to wire.
Printed interposers
The principle
Wiring is printed onto the surfaces of the semiconductor components using metal-enriched nano inks. Subsequently, the components can be stacked very easily - similar to storage elements.
Advantages:
- As it is a 3D printing process, conductive traces can be easily crossed and contact points can be adjusted as desired.
- This process can be done within a single plant and involves low material costs only.
Advantages of printed interposers
Unfavourably positioned contact points of a semiconductor component are moved by printed structures so that they are in the ideal position before stacking.
Advantages:
- For the first time, it is possible to stack heterogeneous wafers almost as desired.
- The high flexibility of the printing process means that changes can be carried out quickly and easily - minimising development costs and time.
Technological requirements for printed interposers
It is only through the availability of two innovative technologies that it has become possible to print interposers:
- Aerosol continuous jet printer
- Metallic nano-particle inks
Today already, passive components are being printed successfully, which means that empirical values are available. A number of companies have specialised in such printing processes and inks.
A 3D printed interposer consists of dielectric and conducting materials.
Printed wiring layers and isolation layers are coplanar.
Printing methods
- The printing process for both the conductive and the insulating material can be based on the ink jet principle, which is already widely used and is also applied in 3D printing.
- In addition, there are a number of other printing processes and nozzles that appear suitable.
- A conflict of objectives arises due to the fact that, for economic reasons, a surface or a distance has to be printed in the shortest possible time and due to the fact that the demand for geometric accuracy can only be achieved by very measured application.
- It can therefore make sense to work with several print nozzles that are optimised differently or can be controlled flexibly and change their print parameters in the process.
- The combination of several printing processes, such as screen printing and aerosol jet printing, also appears promising.
Screen printing
Large areas that are to be provided with an electrical layer can be printed using the screen-printing process. The combination with aerosol printing is interesting - this would enable a quick and economical change of print jobs with high flexibility.
Inkjet printing
The advantage of inkjet printing is the high accuracy combined with the possibility to print specific areas thicker. However, there is a risk that isolated drops may not be printed - conductive structures in a printed interposer would be interrupted and the contacting would be faulty.
Aerosol printing
Another nozzle printing process is aerosol printing. This new technology is becoming increasingly important.
As a continuous jet printing process, the risk of discontinuity of material application is greatly reduced. At the same time, the reliability and quality of the conductive structures is substantially increased.
Comparison with silicon and glass interposers
Technical data
| Lihographically produced interposer | Printed interposer | |
|---|---|---|
| Size of structure | as small as required | 5–20 µm |
| Wiring layers | 2 maximum | any desired |
| Passive components | Yes | Yes |
| Contacting by | Bumps | Bumps |
| Thickness of the interposer | approx. 50 µ | 5-30 µ |
| Interposer on wafer level | Yes | Yes |
| Interposer on die level | No | Yes |
| Production costs of Interposer | high to very high | low (single-step process) |
| Material costs | high (Si wafer) | low |
| Adjustment | difficult | very easy |
| Costs | high | low |
| Heat dissipation | by bumps only | full surface on one side |
| Complex heterogeneous stacks | difficult to impossible | very easy |
| Challenges | Complexity, costs and effort | Optimisation of the printing process |
Manufacturing and cost
| Today: Lithographically produced Si interposer | In the future: Printed interposer | |
|---|---|---|
| Series production / cost-volume ratio |
|
|
| Production time / production speed |
|
|
| Time-to-market |
|
|
| Flexibility |
|
|
Integration into production
Optimal integration into the process chain
The printing of 3D interposers can easily be integrated into the existing processes at semiconductor manufacturers:
- On the one hand, it is basically possible to print the interposer on already singular components. This is a very practical approach for smaller quantities. This method would be less ideal for large quantities, as the handling and alignment of the individual components would take a considerable amount of time.
- A more practical approach would be to print the interposer on the homogeneous wafer or on the already separated - i.e. sawn - wafer, as long as the components are still present in the wafer compound. In this case, the advantage would be that the wafer would only have to be handled and aligned once.
- For optimal accuracy, it would be advantageous if the wafer was not yet sawn, but already laminated on the sawing foil.
The printing of an interposer can take place within a technical system in which all the necessary steps take place without the wafer or the component having to be removed again until the interposer is finished.