Monday, May 27, 2013

Turning Lasers into Robotic Optical Hands for Manipulating Biological Cells

Newton’s scientific accomplishments are truly astonishing. One of his remarkable theories stated that light has momentum. If light has momentum, then it should be possible to move objects by shining a light on them. I am sure that it sounded like a crazy idea when Newton first proposed it.

Over the years, people have done numerous experiments to confirm this theory. This idea is so captivating that it even influenced the great George Lucas. Star Wars movies featured famous Lightsabers that utilized the special properties of the light to create a powerful Jedi weapon. But we have not seen such fantastic spectacles of light and matter interaction in our everyday macroscale world. Light has very small momentum. So moving a heavy couch by shining a laser on it remains in the realm of science fiction. Unfortunately, if you make the laser too powerful, it will simply evaporate the couch and set your house on fire.

A different picture emerges at the microscale. It is certainly possible to move tiny objects by shining a laser on them. But this mode of interaction does not offer much control. Ashkin in 1986 figured out a better way. He created optical traps that were able to hold tiny particles in place. The basic idea was to bend and focus a laser beam tightly using an objective lens. Once the object enters the laser beam, the laser starts interacting with it and pushing it towards the focal point where it gets trapped.

We can imagine the laser as a collection of rays. These rays are reflected and refracted by objects that intercept them. As the rays are bent, their momentum changes and they exert force on the object. This phenomenon can be visualized as interactions between a stationary ball and a moving ball. The direction of the motion of the moving ball changes as it strikes the stationary ball. Hence it exerts a force on the stationary ball. Ashkin found out that once the effect of all the rays in the laser beam was accounted for, the direction of the resultant force was such that the object was pushed towards the focal point of the beam. So as the object entered the laser beam, it was simply pushed towards the focal point and once it reached that point it remained there. In essence, the focused laser beam created a particle trap. A trapped particle can be moved by moving the laser beam. Thus, the laser has been turned into a tweezing tool for grabbing small particles and moving them. Moving optical traps are often referred as optical tweezers.

Numerous groups have used optical traps to manipulate biological cells and study them. In fact, many important discoveries in biology have been made using optical traps. Biologists are primarily interested in fundamental scientific discoveries. So they are happy to create and move optical traps using tele-operation. Just like tele-operated robots, tele-operated optical traps have many inherent limitations. They are slow, require significant expertise, and limit what kind of manipulation is possible.

I was introduced to optical tweezers in 2004 during my sabbatical at the National Institute of Standards and Technology. Thank you Arvind Balijepalli and Tom LeBrun! I am interested in robotics. So once I learned about optical traps, I became interested in turning them into robotic hands for automatically manipulating biological cells. In many situations, directly trapping biological cells can cause complications. Cells might have an irregular shape and they might be susceptible to damage due to direct exposure to the laser.

We decided to take a different path. Rather than building robotic optical tweezers, we wanted to build robotic optical hands. The idea was to use the laser to trap and move microspheres made out of silica or polystyrenes. These microspheres can serve as “fingers” for gripping or pushing cells. So in our idea, the laser would act as an optical hand (i.e., hand of a ghost) and microspheres would act as “fingers”. This idea enabled free floating “fingers” with no physical hand attached to them. This would be truly an alien hand with no biological counterpart on Earth! We could have as many “fingers” as we wanted by splitting the laser beam to create multiple traps. We could also have multiple hands if we wanted! It was a crazy idea. But it removed many constraints associated with conventional microscale robotic grippers and offered several new possibilities. Soon we were hooked to make this idea a reality.

There were numerous challenges. Microspheres and cells float in the liquid medium and exhibit Brownian motion. We had to detect these objects in the scene, plan the next trap location, and make sure that microspheres and cells moved the way we wanted them to move. However, we only had a few milliseconds to do image analysis, planning, and control. Images are noisy and the environment has significant uncertainty. Moreover, the motions of all the hands and fingers need to be exquisitely coordinated. So this was a really tough robotics problem. +ashis banerjee , +Sagar Chowdhury , +Petr Svec , and +Atul Thakur worked incredibly hard to solve the challenging planning, perception, and control problems to realize this vision. They built upon the basic software capability provided by Andrew Pomerance. Wolfgang Losert and Chenlu Wang provided valuable help in conducting the experiments.
Thank you National Science  Foundation for supporting this work! 
Our adventures in this area began by concurrently trapping multiple microspheres and moving them into an ensemble. We then used that ensemble to hold a cell and move it. We also developed the capability to move the cell into its desired location by pushing on it using a microsphere. If a cell is very sensitive to the laser, then we can use an intermediate microsphere as a tool, so that the microsphere “finger” being trapped by the laser does not touch the cell and ensures physical separation between the cell and the laser. Please see below the video of our robotic optical hand.

Hopefully, our colleagues in biology and medicine would be able to think about new scientific theories  that can be enabled by the above described robotic optical hands and their variants. Possibilities range from understanding the behavior of cancer cells to understanding how cells communicate.

Robotics solutions that enable precise automated manipulation of individual cells are expected to revolutionize medicine and biology. Our explorations in this area have taught us that robotics at the microscale requires out-of-the-box thinking. We are now busy coming up with even crazier ideas to marry robotics and biology. So stay tuned for updates.


  1. This is a great example of robotics in conjunction with biology which can be used to solve advance medical problems. I am seriously looking forward for an actual practical implementation of this work. Good work by Optical Tweezers team..!!

  2. This is an amazing piece of work, very lucidly described. The video/demo is very striking. If it is possible to manipulate individual cells in this way, major applications could result. The video has peaked my curiosity, and I'm wondering about the following questions:
    1) I presume that it is possible to insert and remove the silica microspheres from the fluid at will. Is this true? So in any real
    life biological application, the optical tweezers "leave no trace" except move the desired cells to the desired locations.
    2) Are the cells sufficiently large that the effects of Brownian motion can be neglected? Are there situations where small cells could be significantly affected by Brownian motion, therby complicating the control/planning/obstacle avoidance?
    3) Is this work and its potential applications two dimensional? If so, is the laser/optical tweezer originating from outside the two dimensional plane in which the cells live?
    4) Is the analogous problem in three dimensions much harder? Does one have to consider the interaction of the optical tweezer with the fluid medium, and whether the optical tweezer may hit other unrelated cells in the fluid medium?
    5) The video suggests that the control and movement is happening very quickly in real time. Is it reasonable to expect that the techniques developed here will be applicable to situations like the control of multiple cells insider a human artery/vein (where the fluid medium may be moving with respect to the walls)?

    I suspect that an immense amount of work has gone into this video - the robotics control/planning, the experimental work, the imaging technology, the collaboration between the experimentation, the robotic calculations, and the laser manipulation knowledge. It is very impressive.

  3. I can help you clarifying some of your queries:
    1. You can control the concentration of silica microspheres in the solution. Optical tweezers here is used as a hand and microspheres as fingers to move the cell to the desired locations.
    2. Cells are of size 4-6 microns or more. But we have done experiments with even smaller objects of around 2 micron. Our planning is very fast and hence can handle very dynamic obstacles.
    3. It can potentially be used for manipulating cells in 3D. Optical tweezers can trap objects in 3D. Right now we are imaging only from top and do not have any depth perception. Hence, we have to restrict our planning on a 2D plane.
    4. You need to have depth perception to move objects in 3D. In our experiments all the cells are lying more or less on the same plane. We did not observe any effect on other cells present in scene which are not manipulated.
    5. Yes, the manipulation is real time. Currently OT is only used for in vitro experiments due to its design constraints.

  4. I'd like to add a couple of things to Sagar's replies.

    - The group is looking at coupling optical tweezers with microfluidics to achieve high throughput and precise motion simultaneously. In such a hybrid set-up, the interaction between laser and fluid flow will be significant. The experiments shown in the video were conducted in fluid at rest.

    - Apart from the sensing instrumentation issue, real-time planning and control becomes computationally challenging for 3D manipulation due to the growth in state-action space (3 degrees of freedom to 6) and the need to account for additional physical phenomena such as the shadowing effect of laser beams and the downward pull of gravity.

  5. This looks really cool, and especially useful for medical research. Here's my question though: when can we see these cells play soccer, or dance or something? Better yet, what if people could play Pong against a living cell?

  6. The idea of 'optical hands' is fascinating and the group's efforts in making it a reality are very impressive! I will be much interested in using this technology to study biological swarming; with an ability to isolate individuals and precisely increase or decrease population sizes, we can systematically characterize crucial local properties (e.g, neighbor interactions) and global properties (e.g., phase transitions) in microbial swarms.