Fiber Laser Amplifier
by: Ryan Marshall
Supervisor: Dr. Corwin
This program is funded by the National Science Foundation through grant number PHY-0851599.
Welcome. This page summarizes my experience doing research for the summer 2011 in Lumos, Dr. Corwin’s lab.
Below, I give an Introduction, I describe the Project Goals, my Research Strategy, and my Research Progress (Maximizing Ouput Power, Minimizing Pulse Duration, FROG, Autocorrelator, Broadening the Spectrum).
Introduction: Frequency references are important for telecommunications and anything else where it is necessary to know a time or frequency precisely. One way to make a frequency reference is to lock a frequency combed tooth to a molecular transition in a gas. A frequency comb is a pulsed laser that is made up of discrete, equally spaced frequencies. By locking a tooth to a molecular transition, the comb will be stabilized, and it will be possible to measure unknown frequencies very accurately. Our goal is to do direct comb spectroscopy of acetylene in hopes of improving the accuracy and portability of frequency references. To do direct comb spectroscopy, the frequency comb has to be amplified, and that was my project. I built a fiber laser preamplifier, which involved cutting and splicing fiber in attempt to maximize output power and minimize pulse duration.
Summary of what I did: My project was to build a fiber laser preamplifier that can be used for direct comb spectroscopy of acetylene in photonic crystal fibers. The combs used in the lab are homemade and portable, and it is necessary for them to be amplified in order to do spectroscopy. The comb I used was centered at 1560nm wavelength, and my goal was to get significant gain at 1532nm. I used two backward pumping laser diodes at 1480nm to pump erbium-doped fiber (EDF), my gain fiber. First, I optimized the length of gain fiber to get the maximum possible average output power. This required cutting or splicing on EDF in small increments and measuring the power and spectrum each time. Next, I minimized the pulse duration to maximize the peak power. I used two different kinds of fiber in the preamplifier: the erbium-doped gain fiber, and single-mode fiber (SMF). These two fibers have opposite signed dispersions, so I cut or spliced on SMF to cancel out the dispersion from the EDF. Finally I added highly non-linear fiber (HNLF) to broaden the spectrum to get significant gain at 1532nm. Since the comb is at 1560nm, the majority of the gain before the HNLF is at that same wavelength, so the HNLF is needed. In the future, the preamplifier can be used for direct comb spectroscopy of acetylene, or it could be modified to do spectroscopy of other gases.
Project Goals: My main goal is to build a fiber laser amplifier that can be used for direct comb spectroscopy of acetylene. Some secondary goals include learning about, setting up, and using a frequency-resolved optical gating (FROG) device, an optical spectrum analyzer, and an autocorrelator.
Research Strategy: My strategy is to use erbium-doped fiber (EDF) as a gain medium. I start with a long piece of EDF and cut back in small steps, making power measurements after each cut-back. When the power is maximized, I will have the optimal length of EDF. Once that is accomplished, I will need to measure the pulse duration using the FROG and the autocorrelator. The different types of fiber have different dispersions, so I should be able to make a transform-limited pulse by adding or cutting back single-mode fiber (SMF). Finally, I will need to broaden the spectrum to achieve significant gain at 1532 nm wavelength. The comb laser is 1560 nm, so the amplification will be centered near that wavelength. I will use highly non-linear fiber (HNLF) to broaden the spectrum.
Here is a cartoon of the fiber laser preamplifier from when I first started working on it:
Fiber laser comb: This is a frequency comb, which means that the frequency spectrum of this mode-locked laser had discrete, equally-spaced peaks that are separated by the repetition rate. The repetition rate of the comb I used was 98.6 MHz.
WDM: Wavelength division multiplexor. This is to fiber optics as a dichroic mirror is to free-space optics. That is, it passes light of certain wavelengths, but reflects light from other wavelengths. This component has three ports: a pass port, a reflective port, and a common port. The pass port passes certain wavelengths through. The pass port of the WDM that I used was connected to the fiber laser comb, and passed the light at 1560nm. The reflective port was connected to the laser diodes, and reflected light at 1480nm into the common port. The common port was connected to the rest of the amplifier, and carried the signals from the pass port and the reflective port.
LD: Laser diodes. I used two 500mW laser diodes at 1480nm to pump the gain fiber (Liekke, ER-110). The laser diodes are connected to PM (polarization maintain). PM fiber is a special kind of fiber that does not change the polarization of the light within the fiber. The two laser diodes are combined into one fiber before they are sent to the reflective port of the WDM.
SMF: Single-mode fiber. This is what I would call the generic fiber. All unlabeled black lines are single-mode fiber. All of the “x’s” indicated splices between fibers. The green and grey colored arrows represent fiber connectors.
Liekke, Er-110: Erbium doped fiber. This is used as the gain medium. The pump photons from the two pumping laser diodes excite the erbium ions into higher energy states, and then the comb laser stimulates the emission of coherent photons at 1560nm.
Power meter: Simply measures the average output power of the amplifier. The output of the amplifier is split so that 95% of the power goes to the power meter, and 5% of the power goes to the OSA. This way, the spectrum and the power can be measure simultaneously.
OSA: Optical spectrum analyzer. Measures the average spectrum output of the amplifier. Average meaning that it does not measure the spectrum of individual pulses, but it averages the spectrum over a certain scanning time.
The first major task was to get the maximum possible power out of the amplifier. We were shooting for an average power output of around 100mW. With forward pumping, however, we were only achieving around 20mW output power. Forward pumping means that the laser diodes are connected to the front of the gain fiber. This way, the pump would be absorbed first at the front of the gain fiber and there would be less absorption toward the end of the gain fiber. We decided to try backward pumping because we were not getting enough output power. This meant moving the WDM to the other side of the gain fiber:
With backward pumping, we were able to get an output power of 100mW. This was good, but we still wanted to maximize the output power. To do this, we made cut-backs on the erbium doped fiber. There is an optimum length of gain fiber that will maximize the output power. This length is the point where all of the pump is absorbed in the gain fiber, and there is no more extra gain fiber after that point. If the gain fiber is shorter than the optimum length, not all of the pump power will be absorbed in the gain fiber, and so some percentage of the pump is being wasted. If the gain fiber is longer than the optimum length, all of the pump power is being absorbed in the gain fiber, but there is extra loss in the fiber that does not absorb any pump.
After each cut-back, we took power and spectral measurements. The most important thing that we care about at this point is the power. A plot of the spectra after cut-backs is seen in the following figures (the lengths listed for each color are the lengths of the gain fiber, and the power listed is the total amplifier output power):
After finding the optimum length for the gain fiber, we got an output power of 115mW.
We want to minimize the duration of the pulse to that we have the highest peak power.
Fibers have dispersion. The two different fibers that we used, to this point (EDF and SMF) have different dispersions. In fact, they have opposite signed dispersions. This is good because it allows us to optimize the length of SMF used to exactly cancel out the dispersion from the EDF so that we have the shortest pulse possible. Therefore, we made cut-backs on the SMF and measured the pulse duration after each cut-back. We want to cut-back SMF and not EDF because the length of EDF is already optimized to achieve the maximum power possible for the amount of pump power that we use.
We used two different devices to measure pulse duration: FROG and autocorrelator.
FROG: (Frequency resolved optical gating). One way to measure pulse duration is with a FROG. FROG is nice because it will not only tell you the intensity vs. time of pulses, but it will also tell you the wavelength vs. time of pulses. Like an autocorrelator, FROG uses a gated pulse to measure itself. However, the FROG measures the spectrum vs. delay. The FROG then uses a complicated algorithm to retrieve the intensity vs. delay of the pulse.
Here is an example of what the FROG will tell you:
The upper left corner shows the Measured FROG Trace. This is what the FROG camera sees. The measured trace is put through the FROG algorithm, and returns the Retrieved FROG Trace. Because the FROG device measures the second harmonic, this trace will always be symmetric in delay.
The Control Panel shows the center wavelength of the input pulse, the degree of saturation of the FROG camera, and the noise subtraction level.
Temporal Intensity and Phase shows the pulse intensity and phase versus time. Again, due to the measuring of the second harmonic, the sign of the pulse in time is unknown. For example, in this screenshot, there is a wing at the beginning of the pulse. FROG cannot determine whether that wing actually comes before the pulse, or follows it.
Spectral Intensity and Phase shows the intensity and phase versus wavelength. This is simply a Fourier Transform of the Temporal Intensity.
The Results Panel shows some key characteristics of the pulse, including FROG error that determines how well the retrieved FROG trace matches the measured FROG trace.
The Alerts window lets you know of certain things that may be affecting how well the retrieved FROG trace matches the measured FROG trace. For example, in this screenshot, the program has determined that the measured FROG trace does not fill enough of the camera. Therefore, it suggests decreasing the grid size of what is measured, so that a greater percentage of the grid measures the actual pulse.
The Autocorrelation window
shows the Intensity versus time of the autocorrelation. I will talk about autocorrelation in more
This is how we set up the FROG to measure the pulse:
The isolator only allows light to travel in one direction. This is necessary in the set-up so that no reflections can go back into the comb and take it out of mode lock. I learned this the hard way. At first, I did not have an isolator. I would measure the spectrum with the OSA, and all would be well, but then I would change the polarization and the spectrum would turn to a single spike at 1560nm. This is because the comb went out of mode lock and became CW. The isolator, to our relief, fixed that problem.
PC is polarization controller. This piece allows you to change the polarization of the light
ND filter is a neutral density filter. This attenuates the beam before it goes to the FROG camera and the autocorrelator.
LP is a linear polarizer. The FROG will only take horizontally polarized light, so a linear polarizer is useful.
˝ WP is a half wave plate, and this allows you to rotate the polarization of linearly polarized light.
The thick black lines represent fiber, and the thin black lines represent free-space laser beam.
With this set-up, we can measure the
FROG and the spectrum from the OSA simultaneously.
Although the FROG is potentially a very useful in that it gives you the spectrum and intensity of a pulse, we were unable to get retrieve FROG traces that we could trust, and we decided that it would be a better use of our time to just use the autocorrelator.
Autocorrelator: Although the FROG also calculates the autocorrelation of the pulse, we were unable to get the FROG working well enough to trust it.
This is an example of the set-up of an autocorrelator:
An autocorrelator uses a pulse to measure itself. The pulse is split into two different paths, and one path goes through a variable delay. In the autocorrelator that we used, the variable delay was a set of rotating mirrors. The two beams are then focused only a nonlinear device, in our case, a second harmonic generation crystal. The second harmonic is picked up by a detector, and the autocorrelation signal can be read from an oscilloscope.
The second harmonic is only sent to the detector when all or parts of the beams are overlapping. When there is zero delay, the intensity of the second harmonic will be the greatest.
The autocorrelator pretty much measures the convolution of the intensity of the pulse and itself. This is not the actual electric field intensity of the pulse in time, but if the shape of the pulse is known (sech^2, Gaussian, etc.) the actual pulse intensity can be determined from the intensity autocorrelation.
Here are some autocorrelation results from cutting back the length of SMF to try to achieve the shortest pulse possible (the left plot is at full power 111mW amp output, the right plot is at 60mW amp output). The lengths listed are the lengths of SMF after the EDF. In all cases, there is 1.05m of SMF before the EDF. The times listed on the 111mW plot are the widths of the pulse at full width half maximum. The shortest pulse we generated was around 110fs.
What we really care about is the pulse width at full power, but taking measurements at lower powers can give us information about what we are dealing with.
What I have failed to mention up to now are the effects of nonlinear dispersion. At the 60mW power, the pulse width is decreasing at first, but starts to increase again with the green curve. Although there are still some nonlinear effects at 60mW, they have much less of an effect on the pulse relative to the linear dispersion than at 111mW. Because at a lower power the pulse duration is increasing and at higher power the pulse duration is still decreasing, we have presumably passed the optimal length if there were no nonlinear effects.
The very-noticeable wings on the pulses in the 111mW plot are non-linear effects of a third-order soliton.
A soliton is a pulse where the temporal (and spectral) shape of the pulse does not change as it propagates, or returns to its original shape after a soliton period. The higher the order of a soliton, the weirder it behaves in between each soliton period. The soliton order, N, can be calculated by
where γ is a nonlinear parameter, is the peak power, is the width of the pulse, and is the dispersion coefficient of the fiber. In our case, γ = 1.43 1/Wkm, = 2.7 kW, = 255 fs, and = -2.5 x 10^4 fs^2/m.
Therefore, plugging in the numbers gives us a third-order soliton.
The soliton will return to the same shape after one soliton period length. The soliton period, , can be calculated by
Plugging in the numbers gives us a soliton period of 4.05m. The means that, theoretically, if we added 4.05m of SMF fiber to the red curve in the 111mW plot, we would get the same winged result. If we added it to the pink curve, we would get the same non-winged result.
We need power at 1532.8nm for spectroscopy of Acetylene. These are the output spectra:
The right plot is a log scale, and after a 1533nm Fiber Bragg Grating (FBG). The FBG filters out about a 0.5nm bandwidth centered at a certain wavelength. As seen in the plots, we have a little gain at the desired 1532.8nm. However, we must check to make sure that the power we see is actually from the pulse. To do this, we hooked the amplifier up to an electronic spectrum analyzer to check to see if we could see the comb’s repetition rate. We saw a peak at 98.6MHz, of -17 dBm. Converting this to a photodetector voltage (using the impedance of the photodetector) we get 0.03 V. What we measured on the photodetector was 0.076 V. Therefore, that fraction (3/76) of the power that we see is part of the pulse. It is good that we see power at 1532.8nm in the pulse, but it is still not enough power for direct comb spectroscopy.
We can add highly nonlinear fiber to try to broaden the spectrum to shift some of the bigger gain from the higher wavelengths down to around 1533nm.
Final Presentation: Click here to download my presentation in powerpoint format
About Me: I am from Plymouth, Minnesota. I go to school at Macalester College in St. Paul, Minnesota. I am on the soccer team at Macalester. I also enjoy games, playing guitar and writing songs. I plan to go to graduate school for physics.