Micro LED Mass Transfer: An Analysis of 28 Patents

Moving something means sticking it down and letting it go. What switches that adhesion on and off is what splits mass transfer into five branches. Yet not one of the companies that had only the moving technology kept its own patents. They went to Apple right after collateral was released, through an assignment for the benefit of creditors, and to a bonding equipment maker.

In the previous piece on nano LEDs, one thing was deliberately set aside. Once a device grows larger, floating it in ink and laying it down with an electric field stops working, and each one has to be picked up and moved. At that moment count stops being a matter of area and becomes a matter of time. This piece is that place.

But while gathering patents, something unexpected came first. Not one of the companies that first built this technology holds those patents today.

1. Overview

1.1 Background and purpose

That mass transfer is the gate to micro LED displays is close to common knowledge in the industry. A single 4K panel carries about 8.3 million pixels, and with one chip per color that means 24.9 million to be moved. eLux wrote the situation into the background section of its own specification.

“In a display consisting of millions of devices, such a process can take several hours to complete, which is inefficient.” (US10418527B2, background)

This piece tracks who tried to pass that gate and how, using granted patents and USPTO assignment records. In the previous piece, assignment records confirmed that two nano LED lineages went to Samsung on a single day, 2018-09-07. The same material is used here for a different question. Where are the companies that built the technology now?

1.2 Scope

This is not an exhaustive survey. It is a landmark sample representing each company’s lineage. There are many more patents outside it.

1.3 Data sources and search

Everything was confirmed directly on Google Patents originals. News and market research reports were not used as grounds.

1.4 Classification

This piece splits the companies two ways. The criterion is whether they hold device patents.

• Transfer-only companies: they have the moving technology (transfer head, stamp, laser, fluidic assembly) and the inspection around it, but no patents on the structure or epitaxy of the light-emitting device itself
• Companies that hold devices as well: they hold LED structure, epitaxy and electrode patents alongside the moving technology

No ratios were used, only presence or absence. A small change to the regular expression classifying titles moved the same company’s ratio between 4 percent and 19 percent. A ratio is a number the classifier produces, and building a conclusion on it lets the classifier decide the conclusion.

2. Transfer means sticking and releasing

Moving something is, in the end, sticking it down and letting it go. It has to be released from where it was, stuck to a hand, and released from that hand as it is placed. Which means the adhesion has to be switchable on and off. Always stuck and you cannot put it down; never stuck and you cannot pick it up.

That is exactly where the methods of mass transfer divide. What switches the adhesion.

2.1 Each method runs into a different wall

These are not five ways of solving one problem. Each of the five ran into a different wall. And that wall generally shows up as speed.

An elastomer stamp controls adhesion by speed, so its speed cannot be raised freely. The principle is written into the Semprius specification.

Peel fast and the device comes along; peel slowly and it stays. The knob that switches adhesion is speed itself, so raising speed for throughput means failing to release when release is required. That is why Semprius claimed a shear offset.

Instead of pulling straight up, pushing sideways creates a difference in adhesion, which is a way out of the knob being speed.

The electrostatic head uses voltage as the knob rather than speed. Apply current to grip, cut it to release, independent of speed. It pays elsewhere. The head itself has to be built by semiconductor processing. That claim 1 of LuxVue’s US9620478B2 is not a transfer method but a method of making a transfer head is the evidence. Mesa structures on a substrate covered with a passivation layer is a process no different from making a device.

And voltage carries a price. Point Engineering’s specification names LuxVue directly and writes the problem down.

Point Engineering’s answer is vacuum suction. A porous film made by anodizing metal is given vertical pores running through it, with a vacuum chamber above drawing air through. No voltage is used, so there is neither charging nor damage.

The laser uses its knob exactly once. Uniqarta’s claim 1 makes the release layer multi-layered, with one layer a permanent adhesive and another a material sensitive to heat or ultraviolet light. Shine light and that layer decomposes, dropping the device. It is not switching on and off; switch off once and there is no going back.

Fluidic assembly abandons the idea of picking up at all. Float devices in liquid and let them flow, and they seat themselves in wells that match their shape. The parallelism is overwhelming, but there is a wall of speed here too. The eLux specification writes the dilemma down.

Too fast and the devices pass the wells; too slow and filling the whole surface takes a long time. Even in a method that does not pick up, speed is the gate in the end.

2.2 Which branch each of the ten companies stands on

All five without device patents are in someone else’s hands, and all five with them are in their own name.

2.3 The timeline of the five that changed hands

These are the dates left in the assignment records. All were confirmed on the originals.

Two of them show assignments to a bank and a venture lender. That means the patents were pledged as collateral. The Semprius chain also carries a 2011-08-30 assignment to the US Department of Energy, a mark that government money was involved.

An “assignment for the benefit of creditors” is an alternative to bankruptcy proceedings. The company hands its assets to a trustee to distribute among creditors and wind down.

These are the dates in the USPTO assignment records copied over as they stand. A diamond is the point at which patents were pledged as collateral, a square the point at which the owner changed. All five rows end in a square.

2.4 The numbers each of them put forward

Three companies wrote a processing scale into their specifications. Before setting them side by side, their different natures have to be seen first.

These three were not measured with the same ruler. One is an experimental condition, one a theoretical ceiling, one an extrapolation. Placed side by side they read as “electrostatic is 400 times faster than the stamp,” and there is no basis for that anywhere.

3. Qualitative analysis: 28 patents

They are arranged in the order of how adhesion is switched. Where the company that built each method now stands is seen in the same place.

3.1 Elastomer stamp: peel rate is the knob

Semprius’s patents passed through a creditor wind-down to the entity below, and that entity is now doing something else.

3.2 Electrostatic head: voltage is the knob, but the head must be built like a semiconductor

3.3 Laser: the knob is used only once

3.4 Vacuum suction: the side that uses no voltage

3.5 Fluidic assembly: giving up the idea of picking up

3.6 The side that holds the moving technology and the device together

3.7 What comes after transfer: inspection and placement

Once things are moved, defects have to be found and repaired. These are the companies that went for that place.

4. Comparative analysis

4.1 Not one of the builders is left

There are five companies that had only the moving technology. They are not the same five as the five branches in Section 2. The companies that built vacuum suction and fluidic assembly are on the side that remains, seen later, and in their place inspection (Tesoro) and lighting applications (Rohinni) enter here.

What the five have in common is neither nationality nor method. Elastomer stamp (Semprius), electrostatic (LuxVue), laser (Uniqarta), inspection (Tesoro), lighting applications (Rohinni) are all different methods. Founding dates are scattered from 2007 to 2017.

What they share is that they had only the moving technology. And all five had their patents pass into someone else’s name.

4.2 Collateral is recorded before the news

The assignment records for Semprius and LuxVue carry the names of banks and lenders.

Semprius pledged collateral to Silicon Valley Bank and Horizon Technology Finance on 2014-10-30, was released from it on 2017-04-07, and four days later, on 2017-04-11, passed over through an assignment for the benefit of creditors. A wind-down procedure.

LuxVue was pledged twice, to Comerica Bank on 2013-03-21 and TriplePoint Capital on 2013-05-20, and released on 2014-05-02. Then on 2016-04-25 the name became Apple’s.

These dates are not press releases but patent office records. A company need not announce its funding situation; a pledge of collateral is registered as an assignment and stays. It is the same path as the previous piece, where Glo AB was pledged to Fortress Credit, released, and then went to Samsung.

For Uniqarta, Tesoro and Rohinni, however, no collateral was identified. Whether they changed hands without collateral, or whether their small number of US grants kept it out of the sample, cannot be settled with this material.

4.3 The side that remains had something besides moving

All five that kept their own name hold device patents. PlayNitride and Mikro Mesa are Taiwanese LED companies, VueReal holds vertical device and epitaxial structure patents, eLux has nanostructure LED patents, and Point Engineering has micro LED structure patents alongside anodizing as its main business.

Here it becomes tempting to go one step further. What the survivors share might not be devices but “another leg to stand on besides moving.” Point Engineering’s leg is anodizing, not devices.

But this sample cannot say that much. Ten companies cannot separate “presence of device patents” from “presence of another business,” and in this sample the two move almost together. For the same reason causation is not claimed either. Whether they survived because they had device patents, or went as far as devices because they were the kind of company that would survive, is not something patents answer. A company with ample funding might have developed devices and transfer together, or device revenue might have carried the transfer development. Those are different stories, and patents record only the outcome.

What this piece states is co-occurrence. That the five without device patents did not remain, and that the five that remained all had device patents. That far.

4.4 There are layers inside “remained”

Keeping your own name and having a business that runs are different things. The priority distribution of the most recent 100 filings splits three ways.

eLux kept its name but has no new filings. What this piece calls “remained” means only that the patents are in its own name. The company’s finances and whether it is in volume production cannot be known from patents.

4.5 The direction the successor took

X Display Company inherited the Semprius transfer patents, but its recent grants are pixel driving and display memory architecture. Instead of carrying the transfer technology forward for sale, it moved toward display modules.

It is tempting to read this as another instance of “transfer alone does not stand,” but the business strategy of a successor entity cannot be judged from patent titles alone. Only the facts are recorded here.

5. Counter-evidence and limits

The conclusions of this piece hold only within the following five limits.

① The sample is ten companies. They were chosen after going through 348 US grants and 118 assignees, but this is not an exhaustive survey. Chinese and Japanese transfer specialists could not be included in this sample.

② Survivorship bias was not fully excluded. A search was made for “a company still independent while doing only transfer without devices” and none was found. All five (Semprius, LuxVue, Uniqarta, Tesoro, Rohinni) changed hands. But not finding one does not mean none exists. The range searched is as stated in ① above.

③ The scope was limited. Equipment majors such as Applied Materials, Kulicke and Soffa and Orbotech stand perfectly well without making devices. What this piece covers is companies founded on micro LED transfer technology, and equipment makers that already stood on another business are out of scope. Remove this limit and the conclusion does not hold.

④ The three numbers were not compared. 256, 100,000 and 56 million per hour are an experimental condition, a calculated ceiling and an extrapolation, so they cannot be placed on the same axis. Which method is faster cannot be answered from this material.

⑤ The reason for expiry was not confirmed. That two LuxVue compliant heads expired for non-payment of fees is a fact, but why Apple let them go is not in the specification.

And one thing is added on throughput. Specifications of 122 granted mass transfer patents were searched for throughput figures and none were found. However, a patent is not a document for recording performance. Writing performance into a specification narrows the claim scope in examination and litigation, so not writing it is normal practice. So this fact cannot be used to say “the industry does not know that number,” only that “the number is not in the place called a patent.”

6. Conclusions and implications

That mass transfer is the gate to micro LED was said fifteen years ago and is said now. Follow the patents of the companies that set out to open that gate, and not one of the builders holds them.

Semprius bet on solar and passed over through a creditor wind-down; LuxVue was pledged to two banks and went to Apple right after being released. Uniqarta went to a bonding equipment maker, Tesoro back to Apple, and Rohinni through an automotive components company to another entity. The methods of all five were different and the outcome is the same.

The companies that now hold those names as their own share one thing. They had something besides the moving technology. Taiwan’s PlayNitride and Mikro Mesa have LED devices, Canada’s VueReal has epitaxial structures and inspection cartridges, and Korea’s Point Engineering has anodizing as its main business.

This is not read as causation. Whether they held on because they had devices, or went as far as devices because they were the kind of company that would hold on, patents do not answer. What can be said is only that not one company without device patents remained in this sample.

And “remained” has to be used carefully too. eLux kept its name but has had almost no new filings since 2016, and VueReal’s filings are declining. A patent being in your own name and a business running are different things. That difference has to be seen not in patents but in disclosures, and it is outside the scope of this piece.

There were five ways to switch adhesion on and off. By peel rate, by voltage, by light, by vacuum, and by never picking up at all. A technology that split into five branches converges into one line in the owner’s name. That the methods differed did not make the outcomes differ.

7. Appendix: patent specification table

Status is copied as displayed in the legal status field on each patent’s detail page. “Expired” means expired for fee-related reasons.