Electron-beam Lift-off : Collection Efficiency And Paths To Improvement
Liftoff coatings require specific rules of deposition (normal or 90o) incidence angle be maintained to prevent metal deposition onto resist via sidewalls.
Liftoff carriers orient wafers to maintain the perpendicularity rule from a point source at the electron beam impact point in the crucible to a point on the center of each wafer in a carrier.
Considering perfect 90o incidence to an essentially “flat wafer center" it is then obvious that incidence from the same point source at the e-gun to the outer edge of the wafer creates an imperfection to perfect normal incidence will occur. The lack of perpendicularity represents a process limit for “ideal liftoff conditions".
These constraints lead to larger diameter wafers requiring greater distance between the E-beam source and the wafers surface to maintain near normal incidence at the wafers edge. In general, +/- 5o from 90o incidence is an accepted rule or good liftoff processing geometry.
Figure 1 maps the radiating zones of gold vapor flux by effective deposition rate at locations above an E-gun. Most simply, this is the flux topography above an E-gun. It is important to note, the map is for a specific set of evaporation conditions as described at the top of the chart. The flux lines would be different if a material other than gold were evaporated even under the exact same E-Gun conditions.
This image can be mathematically described as cosine radiation from a point source. The product of the equation will be deposition rate which is highest close and directly above the source. This material flux, under similar conditions, is high reproducible and this is what offers engineers the benefit of a predictable production environment. However the flux pattern’s shape is ruled by cosine curve mathematics and is highly dependent on control of the conditions listed below:
· Bulk Characteristics of the material being evaporated: Density, Tc, Te, etc melting point,
· Maximum power capacity of the material being evaporated,
· Size of crucible used (thermal mass),
· Deposition Power,
· Beam Shape,
· Use of sweep,
Cosine curves are expressed in exponential form as cosine to the “n" power of an angle theta describing a position above the (point) source. As the exponent “n" changes (becoming >1) the cosine curve exhibits a characteristic elongation changing its shape from a unit circle (n=1) to what may be better described as an inverted and long teardrop as n exceeds a value of ~3.
Temescal has collected hundreds of vapor flux maps using data from E-Gun deposition tests. As a result of this work, we have developed software providing both a prediction coating performance from E-Guns and the automatic design of uniformity masks used to create batch coatings of uniform thickness. Figure 2 shows how a uniformity mask functions in an e-gun system. Figure 3 shows such a prediction for an unmasked film deposited from an e-gun onto a small liftoff carrier with two tiers of wafers.
Across a single-axis-of-rotation carrier, the thickness distribution of an unmasked liftoff film reflects the form of radiation from a point source. That is, a rich inner flux zone deposits thicker films on wafers carried in the inner-tier positions, while a weaker outer flux zone deposits thinner films onto wafers carried in outer-tier positions.
In order for a mask to improve uniformity in this environment, it must selectively block material moving to the substrate. This allows a mask to deliver a uniform thickness film to all the wafers on a rotating carrier. System providers regularly re-design equipment, often to accommodate larger wafers by increasing source-to-substrate distance or to provide larger batch sizes. However as the next figure demonstrates, mask designs for liftoff must also change with any change occurring to the wafer carrier. When a carrier’s size is increased (to hold larger, or more wafers) its’ uniformity mask must also grow to balance its’ ability to block the rich inner flux zone to a now larger yet weaker outer flux region of the carrier.
Figure 4 illustrates how the declining trend in the outer tier flux rate would continue as a carrier is enlarged (a continuation of the violet dotted line). This obligates the mask also be enlarged to increase material capture from the central zone to balance the entire batch coating to a common thickness.
Figure 4. Masked vs. Unmasked Deposition
In Figure 5, a carrier’s spherical dome radius (the source-to substrate distance) has been superimposed onto the flux map. This indicates the surface where wafers would be carried to make liftoff films. Note the span between different flux or deposition rate zones and their decline as the diameter of a carrier is increased. A reference mask position is also shown indicating the horizontal plane where a mask would be located. The mask’s projected shadow onto the carrier (and wafers) would then trim flux to achieve good uniformity over a specific carrier/wafer batch.
Figure 5 shows flux, the wafer carrier and mask position as viewed through a chamber door looking horizontally into the system. To clearly show mask shape and its shadow effect onto the dome and wafers better, it is more appropriate to change the view to one looking down from the chamber’s top, straight through the wafer and mask onto the centrally mounted E-Gun source.
Figure 6 uses such a top-down projection showing a sequence of growing carrier diameters and their correspondingly growing masks for this specific gold deposition. The carrier is represented by a gray circle in the diagram. The mask starts as a gray outline leaf shape in 6-i) and grows in shape and size as the carrier’s diameter increases. Numbers displayed on a helical arc radiating from the center offer the deposition rate, in Å/s, that would result at the outer edge of the carrier as its’ diameter increases. Note: The process evaporation power is held constant in the model, only the carrier and mask change.
Figure 6-i through 6-iv illustrate the impact of increasing a carrier’s surface area. Carrier size growth is expressed in terms of the half-angle created by the carrier’s outer edge and a vertical center-line dropped directly to the E-Gun source. The first diagram shows a dome with a 25° half angle, and in the succeeding diagrams the half angle increases in increments of 10°.
Figure 7 presents three graphs of modeling data in more detail for a substrate carrier as it grows from a small 15ohalf-angle to a larger 60ohalf-angle size.
· dome surface area
· the declining rate of deposition at the outer edge of the carrier with each step of growth
· a calculation of the ideal collection of material beneath the growing dome surface, normalized for liftoff incidence and the use of a mask to make the full batch a uniform material thickness.
Again, this is for ideal cosine gold deposition from a flux cloud of cosine shape n = 3.3. This represents a typical production deposition condition in a fab.
The graphs in Figure 7 show that a single-axis-of-rotation substrate carrier with a half-angle of approximately 40oyields the best possible collection efficiency. Dome growth beyond ~40oresults in more material being lost to the mask than gained on the carrier or wafers. Beyond that point, material collection on the dome declines as a percentage of total material evaporated. The decline is due to the increasing need for uniformity mask growth to compensate for the low deposition rate at a carriers’ outer edge.
Ideal* collection is never realized as a small percentage of evaporated materials are inevitably lost in conditioning and other stages of actual production runs. However, generalized evaporation losses, from an ideal*, are experienced by equally by all sources regardless of system geometry or application. Therefore they can be considered to represent a normalized loss coefficient across all processes being modeled for comparative efficiency purposes.
Continued modeling now reveals that maximum collection efficiency in a masked, single axis rotation liftoff process is available from a system with a half-angle of ~40°. “Conceptual construction" of such a tool indicates it would carry 60 wafers in a 53 inch wide carrier, as shown in Figure 8.
While large, it’s conceivable a segmented carrier could be made for loading as in other production tools. “Conceptual modeling" indicates this batch would collect Au material in identical flux conditions at an effective 2.1 Angstrom/second rate (due to larger radius of outer wafer tier). An ideal* cosine model of the cloud indicates 26.7% of the gold evaporated will be collected on the 60 wafers. To offer a comparison this figure would represent 38% Au collection improvement over Temescal’s FC-4400’s collection efficiency which carries 30 wafers/batch operating under the same flux conditions.
Modeling the “ideal geometry" gives rise to another question which is, ‘Is there any way for liftoff coating collection to be improved beyond this?’
In 2002, Temescal’s work in flux mapping software compelled a focus on mask related losses. Experiments were made attempting to reduce the need for uniformity masks in systems. An alternative concept for liftoff wafer motion was developed called HULA an acronym for High Uniformity Liftoff Assembly. The system maintains the wafer’s orthogonal relationship needed for liftoff while also providing planetary motion which moves the wafer between high and low (inner and outer radius) flux zones during deposition. Time averaging wafer residence between the zones offers uniform films without the need for masks. HULA fixtures demonstrated that wafers can be coated with high uniformity at effectively increased deposition rates without the need for masks. Figure 9 shows such a HULA fixture.
Figure 9. HULA Substrate Carrier
Conventional single-axis-of-rotation carriers fix wafers in inner and outer tier positions. Outer radius wafers get low deposition rates by virtue of location and single-axis-rotation. Inner radius wafers get rich flux and are coated faster. They therefore require selective masking reducing collection to match that of the outer tier wafer. In Figure 10, a mask is represented by the green triangle.
Once installed, masks allow all wafers to be coated at the same rate over a fixed number of carrier revolutions. Variations in different material/turret-pocket flux may be addressed by rate changes or through supplemental masks used to trim a select material/turret-pocket. Regardless of the mask employed, the maximum deposition rate in these systems is established by the deposition rates at the outer-tier position of the carrier for each material. The yellow arrow In Figure 11 indicates that this mask, optimized for gold uniformity resulted in a nominal 1000 Angstrom films.
By contrast, a HULA averages a wafer’s time spent in a range of flux zones by rotation and eliminates the need for the mask. This provides a normalized and effectively higher average deposition rate to all wafers on the carrier. It improves collection efficiency and uniformities of all different materials in a turret gun.
Unmasked batch uniformities of